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"
37 #include "inode-map.h"
39 #include "rcu-string.h"
41 #include "dev-replace.h"
46 #include "compression.h"
47 #include "space-info.h"
48 #include "delalloc-space.h"
49 #include "block-group.h"
52 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
53 * structures are incorrect, as the timespec structure from userspace
54 * is 4 bytes too small. We define these alternatives here to teach
55 * the kernel about the 32-bit struct packing.
57 struct btrfs_ioctl_timespec_32 {
60 } __attribute__ ((__packed__));
62 struct btrfs_ioctl_received_subvol_args_32 {
63 char uuid[BTRFS_UUID_SIZE]; /* in */
64 __u64 stransid; /* in */
65 __u64 rtransid; /* out */
66 struct btrfs_ioctl_timespec_32 stime; /* in */
67 struct btrfs_ioctl_timespec_32 rtime; /* out */
69 __u64 reserved[16]; /* in */
70 } __attribute__ ((__packed__));
72 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
73 struct btrfs_ioctl_received_subvol_args_32)
76 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
77 struct btrfs_ioctl_send_args_32 {
78 __s64 send_fd; /* in */
79 __u64 clone_sources_count; /* in */
80 compat_uptr_t clone_sources; /* in */
81 __u64 parent_root; /* in */
83 __u64 reserved[4]; /* in */
84 } __attribute__ ((__packed__));
86 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
87 struct btrfs_ioctl_send_args_32)
90 /* Mask out flags that are inappropriate for the given type of inode. */
91 static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
94 if (S_ISDIR(inode->i_mode))
96 else if (S_ISREG(inode->i_mode))
97 return flags & ~FS_DIRSYNC_FL;
99 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
103 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
106 static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags)
108 unsigned int iflags = 0;
110 if (flags & BTRFS_INODE_SYNC)
111 iflags |= FS_SYNC_FL;
112 if (flags & BTRFS_INODE_IMMUTABLE)
113 iflags |= FS_IMMUTABLE_FL;
114 if (flags & BTRFS_INODE_APPEND)
115 iflags |= FS_APPEND_FL;
116 if (flags & BTRFS_INODE_NODUMP)
117 iflags |= FS_NODUMP_FL;
118 if (flags & BTRFS_INODE_NOATIME)
119 iflags |= FS_NOATIME_FL;
120 if (flags & BTRFS_INODE_DIRSYNC)
121 iflags |= FS_DIRSYNC_FL;
122 if (flags & BTRFS_INODE_NODATACOW)
123 iflags |= FS_NOCOW_FL;
125 if (flags & BTRFS_INODE_NOCOMPRESS)
126 iflags |= FS_NOCOMP_FL;
127 else if (flags & BTRFS_INODE_COMPRESS)
128 iflags |= FS_COMPR_FL;
134 * Update inode->i_flags based on the btrfs internal flags.
136 void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
138 struct btrfs_inode *binode = BTRFS_I(inode);
139 unsigned int new_fl = 0;
141 if (binode->flags & BTRFS_INODE_SYNC)
143 if (binode->flags & BTRFS_INODE_IMMUTABLE)
144 new_fl |= S_IMMUTABLE;
145 if (binode->flags & BTRFS_INODE_APPEND)
147 if (binode->flags & BTRFS_INODE_NOATIME)
149 if (binode->flags & BTRFS_INODE_DIRSYNC)
152 set_mask_bits(&inode->i_flags,
153 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
157 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
159 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
160 unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags);
162 if (copy_to_user(arg, &flags, sizeof(flags)))
168 * Check if @flags are a supported and valid set of FS_*_FL flags and that
169 * the old and new flags are not conflicting
171 static int check_fsflags(unsigned int old_flags, unsigned int flags)
173 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
174 FS_NOATIME_FL | FS_NODUMP_FL | \
175 FS_SYNC_FL | FS_DIRSYNC_FL | \
176 FS_NOCOMP_FL | FS_COMPR_FL |
180 /* COMPR and NOCOMP on new/old are valid */
181 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
184 if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
187 /* NOCOW and compression options are mutually exclusive */
188 if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
190 if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
196 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
198 struct inode *inode = file_inode(file);
199 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
200 struct btrfs_inode *binode = BTRFS_I(inode);
201 struct btrfs_root *root = binode->root;
202 struct btrfs_trans_handle *trans;
203 unsigned int fsflags, old_fsflags;
205 const char *comp = NULL;
208 if (!inode_owner_or_capable(inode))
211 if (btrfs_root_readonly(root))
214 if (copy_from_user(&fsflags, arg, sizeof(fsflags)))
217 ret = mnt_want_write_file(file);
222 fsflags = btrfs_mask_fsflags_for_type(inode, fsflags);
223 old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags);
225 ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
229 ret = check_fsflags(old_fsflags, fsflags);
233 binode_flags = binode->flags;
234 if (fsflags & FS_SYNC_FL)
235 binode_flags |= BTRFS_INODE_SYNC;
237 binode_flags &= ~BTRFS_INODE_SYNC;
238 if (fsflags & FS_IMMUTABLE_FL)
239 binode_flags |= BTRFS_INODE_IMMUTABLE;
241 binode_flags &= ~BTRFS_INODE_IMMUTABLE;
242 if (fsflags & FS_APPEND_FL)
243 binode_flags |= BTRFS_INODE_APPEND;
245 binode_flags &= ~BTRFS_INODE_APPEND;
246 if (fsflags & FS_NODUMP_FL)
247 binode_flags |= BTRFS_INODE_NODUMP;
249 binode_flags &= ~BTRFS_INODE_NODUMP;
250 if (fsflags & FS_NOATIME_FL)
251 binode_flags |= BTRFS_INODE_NOATIME;
253 binode_flags &= ~BTRFS_INODE_NOATIME;
254 if (fsflags & FS_DIRSYNC_FL)
255 binode_flags |= BTRFS_INODE_DIRSYNC;
257 binode_flags &= ~BTRFS_INODE_DIRSYNC;
258 if (fsflags & FS_NOCOW_FL) {
259 if (S_ISREG(inode->i_mode)) {
261 * It's safe to turn csums off here, no extents exist.
262 * Otherwise we want the flag to reflect the real COW
263 * status of the file and will not set it.
265 if (inode->i_size == 0)
266 binode_flags |= BTRFS_INODE_NODATACOW |
267 BTRFS_INODE_NODATASUM;
269 binode_flags |= BTRFS_INODE_NODATACOW;
273 * Revert back under same assumptions as above
275 if (S_ISREG(inode->i_mode)) {
276 if (inode->i_size == 0)
277 binode_flags &= ~(BTRFS_INODE_NODATACOW |
278 BTRFS_INODE_NODATASUM);
280 binode_flags &= ~BTRFS_INODE_NODATACOW;
285 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
286 * flag may be changed automatically if compression code won't make
289 if (fsflags & FS_NOCOMP_FL) {
290 binode_flags &= ~BTRFS_INODE_COMPRESS;
291 binode_flags |= BTRFS_INODE_NOCOMPRESS;
292 } else if (fsflags & FS_COMPR_FL) {
294 if (IS_SWAPFILE(inode)) {
299 binode_flags |= BTRFS_INODE_COMPRESS;
300 binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
302 comp = btrfs_compress_type2str(fs_info->compress_type);
303 if (!comp || comp[0] == 0)
304 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
306 binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
313 trans = btrfs_start_transaction(root, 3);
315 ret = PTR_ERR(trans);
320 ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
323 btrfs_abort_transaction(trans, ret);
327 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
329 if (ret && ret != -ENODATA) {
330 btrfs_abort_transaction(trans, ret);
335 binode->flags = binode_flags;
336 btrfs_sync_inode_flags_to_i_flags(inode);
337 inode_inc_iversion(inode);
338 inode->i_ctime = current_time(inode);
339 ret = btrfs_update_inode(trans, root, inode);
342 btrfs_end_transaction(trans);
345 mnt_drop_write_file(file);
350 * Translate btrfs internal inode flags to xflags as expected by the
351 * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are
354 static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags)
356 unsigned int xflags = 0;
358 if (flags & BTRFS_INODE_APPEND)
359 xflags |= FS_XFLAG_APPEND;
360 if (flags & BTRFS_INODE_IMMUTABLE)
361 xflags |= FS_XFLAG_IMMUTABLE;
362 if (flags & BTRFS_INODE_NOATIME)
363 xflags |= FS_XFLAG_NOATIME;
364 if (flags & BTRFS_INODE_NODUMP)
365 xflags |= FS_XFLAG_NODUMP;
366 if (flags & BTRFS_INODE_SYNC)
367 xflags |= FS_XFLAG_SYNC;
372 /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */
373 static int check_xflags(unsigned int flags)
375 if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME |
376 FS_XFLAG_NODUMP | FS_XFLAG_SYNC))
382 * Set the xflags from the internal inode flags. The remaining items of fsxattr
385 static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg)
387 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
390 simple_fill_fsxattr(&fa, btrfs_inode_flags_to_xflags(binode->flags));
391 if (copy_to_user(arg, &fa, sizeof(fa)))
397 static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg)
399 struct inode *inode = file_inode(file);
400 struct btrfs_inode *binode = BTRFS_I(inode);
401 struct btrfs_root *root = binode->root;
402 struct btrfs_trans_handle *trans;
403 struct fsxattr fa, old_fa;
405 unsigned old_i_flags;
408 if (!inode_owner_or_capable(inode))
411 if (btrfs_root_readonly(root))
414 if (copy_from_user(&fa, arg, sizeof(fa)))
417 ret = check_xflags(fa.fsx_xflags);
421 if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0)
424 ret = mnt_want_write_file(file);
430 old_flags = binode->flags;
431 old_i_flags = inode->i_flags;
433 simple_fill_fsxattr(&old_fa,
434 btrfs_inode_flags_to_xflags(binode->flags));
435 ret = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
439 if (fa.fsx_xflags & FS_XFLAG_SYNC)
440 binode->flags |= BTRFS_INODE_SYNC;
442 binode->flags &= ~BTRFS_INODE_SYNC;
443 if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE)
444 binode->flags |= BTRFS_INODE_IMMUTABLE;
446 binode->flags &= ~BTRFS_INODE_IMMUTABLE;
447 if (fa.fsx_xflags & FS_XFLAG_APPEND)
448 binode->flags |= BTRFS_INODE_APPEND;
450 binode->flags &= ~BTRFS_INODE_APPEND;
451 if (fa.fsx_xflags & FS_XFLAG_NODUMP)
452 binode->flags |= BTRFS_INODE_NODUMP;
454 binode->flags &= ~BTRFS_INODE_NODUMP;
455 if (fa.fsx_xflags & FS_XFLAG_NOATIME)
456 binode->flags |= BTRFS_INODE_NOATIME;
458 binode->flags &= ~BTRFS_INODE_NOATIME;
460 /* 1 item for the inode */
461 trans = btrfs_start_transaction(root, 1);
463 ret = PTR_ERR(trans);
467 btrfs_sync_inode_flags_to_i_flags(inode);
468 inode_inc_iversion(inode);
469 inode->i_ctime = current_time(inode);
470 ret = btrfs_update_inode(trans, root, inode);
472 btrfs_end_transaction(trans);
476 binode->flags = old_flags;
477 inode->i_flags = old_i_flags;
481 mnt_drop_write_file(file);
486 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
488 struct inode *inode = file_inode(file);
490 return put_user(inode->i_generation, arg);
493 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
496 struct btrfs_device *device;
497 struct request_queue *q;
498 struct fstrim_range range;
499 u64 minlen = ULLONG_MAX;
503 if (!capable(CAP_SYS_ADMIN))
507 * If the fs is mounted with nologreplay, which requires it to be
508 * mounted in RO mode as well, we can not allow discard on free space
509 * inside block groups, because log trees refer to extents that are not
510 * pinned in a block group's free space cache (pinning the extents is
511 * precisely the first phase of replaying a log tree).
513 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
517 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
521 q = bdev_get_queue(device->bdev);
522 if (blk_queue_discard(q)) {
524 minlen = min_t(u64, q->limits.discard_granularity,
532 if (copy_from_user(&range, arg, sizeof(range)))
536 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
537 * block group is in the logical address space, which can be any
538 * sectorsize aligned bytenr in the range [0, U64_MAX].
540 if (range.len < fs_info->sb->s_blocksize)
543 range.minlen = max(range.minlen, minlen);
544 ret = btrfs_trim_fs(fs_info, &range);
548 if (copy_to_user(arg, &range, sizeof(range)))
554 int __pure btrfs_is_empty_uuid(u8 *uuid)
558 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
565 static noinline int create_subvol(struct inode *dir,
566 struct dentry *dentry,
567 const char *name, int namelen,
568 struct btrfs_qgroup_inherit *inherit)
570 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
571 struct btrfs_trans_handle *trans;
572 struct btrfs_key key;
573 struct btrfs_root_item *root_item;
574 struct btrfs_inode_item *inode_item;
575 struct extent_buffer *leaf;
576 struct btrfs_root *root = BTRFS_I(dir)->root;
577 struct btrfs_root *new_root;
578 struct btrfs_block_rsv block_rsv;
579 struct timespec64 cur_time = current_time(dir);
585 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
588 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
592 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
596 ret = get_anon_bdev(&anon_dev);
601 * Don't create subvolume whose level is not zero. Or qgroup will be
602 * screwed up since it assumes subvolume qgroup's level to be 0.
604 if (btrfs_qgroup_level(objectid)) {
609 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
611 * The same as the snapshot creation, please see the comment
612 * of create_snapshot().
614 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false);
618 trans = btrfs_start_transaction(root, 0);
620 ret = PTR_ERR(trans);
621 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
624 trans->block_rsv = &block_rsv;
625 trans->bytes_reserved = block_rsv.size;
627 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
631 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
637 btrfs_mark_buffer_dirty(leaf);
639 inode_item = &root_item->inode;
640 btrfs_set_stack_inode_generation(inode_item, 1);
641 btrfs_set_stack_inode_size(inode_item, 3);
642 btrfs_set_stack_inode_nlink(inode_item, 1);
643 btrfs_set_stack_inode_nbytes(inode_item,
645 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
647 btrfs_set_root_flags(root_item, 0);
648 btrfs_set_root_limit(root_item, 0);
649 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
651 btrfs_set_root_bytenr(root_item, leaf->start);
652 btrfs_set_root_generation(root_item, trans->transid);
653 btrfs_set_root_level(root_item, 0);
654 btrfs_set_root_refs(root_item, 1);
655 btrfs_set_root_used(root_item, leaf->len);
656 btrfs_set_root_last_snapshot(root_item, 0);
658 btrfs_set_root_generation_v2(root_item,
659 btrfs_root_generation(root_item));
660 generate_random_guid(root_item->uuid);
661 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
662 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
663 root_item->ctime = root_item->otime;
664 btrfs_set_root_ctransid(root_item, trans->transid);
665 btrfs_set_root_otransid(root_item, trans->transid);
667 btrfs_tree_unlock(leaf);
668 free_extent_buffer(leaf);
671 btrfs_set_root_dirid(root_item, new_dirid);
673 key.objectid = objectid;
675 key.type = BTRFS_ROOT_ITEM_KEY;
676 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
681 key.offset = (u64)-1;
682 new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
683 if (IS_ERR(new_root)) {
684 free_anon_bdev(anon_dev);
685 ret = PTR_ERR(new_root);
686 btrfs_abort_transaction(trans, ret);
689 /* Freeing will be done in btrfs_put_root() of new_root */
692 btrfs_record_root_in_trans(trans, new_root);
694 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
695 btrfs_put_root(new_root);
697 /* We potentially lose an unused inode item here */
698 btrfs_abort_transaction(trans, ret);
702 mutex_lock(&new_root->objectid_mutex);
703 new_root->highest_objectid = new_dirid;
704 mutex_unlock(&new_root->objectid_mutex);
707 * insert the directory item
709 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
711 btrfs_abort_transaction(trans, ret);
715 ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
716 BTRFS_FT_DIR, index);
718 btrfs_abort_transaction(trans, ret);
722 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
723 ret = btrfs_update_inode(trans, root, dir);
725 btrfs_abort_transaction(trans, ret);
729 ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
730 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
732 btrfs_abort_transaction(trans, ret);
736 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
737 BTRFS_UUID_KEY_SUBVOL, objectid);
739 btrfs_abort_transaction(trans, ret);
743 trans->block_rsv = NULL;
744 trans->bytes_reserved = 0;
745 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
747 err = btrfs_commit_transaction(trans);
752 inode = btrfs_lookup_dentry(dir, dentry);
754 return PTR_ERR(inode);
755 d_instantiate(dentry, inode);
761 free_anon_bdev(anon_dev);
766 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
767 struct dentry *dentry, bool readonly,
768 struct btrfs_qgroup_inherit *inherit)
770 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
772 struct btrfs_pending_snapshot *pending_snapshot;
773 struct btrfs_trans_handle *trans;
776 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
779 if (atomic_read(&root->nr_swapfiles)) {
781 "cannot snapshot subvolume with active swapfile");
785 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
786 if (!pending_snapshot)
789 ret = get_anon_bdev(&pending_snapshot->anon_dev);
792 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
794 pending_snapshot->path = btrfs_alloc_path();
795 if (!pending_snapshot->root_item || !pending_snapshot->path) {
800 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
801 BTRFS_BLOCK_RSV_TEMP);
803 * 1 - parent dir inode
806 * 2 - root ref/backref
807 * 1 - root of snapshot
810 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
811 &pending_snapshot->block_rsv, 8,
816 pending_snapshot->dentry = dentry;
817 pending_snapshot->root = root;
818 pending_snapshot->readonly = readonly;
819 pending_snapshot->dir = dir;
820 pending_snapshot->inherit = inherit;
822 trans = btrfs_start_transaction(root, 0);
824 ret = PTR_ERR(trans);
828 spin_lock(&fs_info->trans_lock);
829 list_add(&pending_snapshot->list,
830 &trans->transaction->pending_snapshots);
831 spin_unlock(&fs_info->trans_lock);
833 ret = btrfs_commit_transaction(trans);
837 ret = pending_snapshot->error;
841 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
845 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
847 ret = PTR_ERR(inode);
851 d_instantiate(dentry, inode);
853 pending_snapshot->anon_dev = 0;
855 /* Prevent double freeing of anon_dev */
856 if (ret && pending_snapshot->snap)
857 pending_snapshot->snap->anon_dev = 0;
858 btrfs_put_root(pending_snapshot->snap);
859 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
861 if (pending_snapshot->anon_dev)
862 free_anon_bdev(pending_snapshot->anon_dev);
863 kfree(pending_snapshot->root_item);
864 btrfs_free_path(pending_snapshot->path);
865 kfree(pending_snapshot);
870 /* copy of may_delete in fs/namei.c()
871 * Check whether we can remove a link victim from directory dir, check
872 * whether the type of victim is right.
873 * 1. We can't do it if dir is read-only (done in permission())
874 * 2. We should have write and exec permissions on dir
875 * 3. We can't remove anything from append-only dir
876 * 4. We can't do anything with immutable dir (done in permission())
877 * 5. If the sticky bit on dir is set we should either
878 * a. be owner of dir, or
879 * b. be owner of victim, or
880 * c. have CAP_FOWNER capability
881 * 6. If the victim is append-only or immutable we can't do anything with
882 * links pointing to it.
883 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
884 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
885 * 9. We can't remove a root or mountpoint.
886 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
887 * nfs_async_unlink().
890 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
894 if (d_really_is_negative(victim))
897 BUG_ON(d_inode(victim->d_parent) != dir);
898 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
900 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
905 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
906 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
909 if (!d_is_dir(victim))
913 } else if (d_is_dir(victim))
917 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
922 /* copy of may_create in fs/namei.c() */
923 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
925 if (d_really_is_positive(child))
929 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
933 * Create a new subvolume below @parent. This is largely modeled after
934 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
935 * inside this filesystem so it's quite a bit simpler.
937 static noinline int btrfs_mksubvol(const struct path *parent,
938 const char *name, int namelen,
939 struct btrfs_root *snap_src,
941 struct btrfs_qgroup_inherit *inherit)
943 struct inode *dir = d_inode(parent->dentry);
944 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
945 struct dentry *dentry;
948 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
952 dentry = lookup_one_len(name, parent->dentry, namelen);
953 error = PTR_ERR(dentry);
957 error = btrfs_may_create(dir, dentry);
962 * even if this name doesn't exist, we may get hash collisions.
963 * check for them now when we can safely fail
965 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
971 down_read(&fs_info->subvol_sem);
973 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
977 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
979 error = create_subvol(dir, dentry, name, namelen, inherit);
982 fsnotify_mkdir(dir, dentry);
984 up_read(&fs_info->subvol_sem);
992 static noinline int btrfs_mksnapshot(const struct path *parent,
993 const char *name, int namelen,
994 struct btrfs_root *root,
996 struct btrfs_qgroup_inherit *inherit)
999 bool snapshot_force_cow = false;
1002 * Force new buffered writes to reserve space even when NOCOW is
1003 * possible. This is to avoid later writeback (running dealloc) to
1004 * fallback to COW mode and unexpectedly fail with ENOSPC.
1006 btrfs_drew_read_lock(&root->snapshot_lock);
1008 ret = btrfs_start_delalloc_snapshot(root);
1013 * All previous writes have started writeback in NOCOW mode, so now
1014 * we force future writes to fallback to COW mode during snapshot
1017 atomic_inc(&root->snapshot_force_cow);
1018 snapshot_force_cow = true;
1020 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
1022 ret = btrfs_mksubvol(parent, name, namelen,
1023 root, readonly, inherit);
1025 if (snapshot_force_cow)
1026 atomic_dec(&root->snapshot_force_cow);
1027 btrfs_drew_read_unlock(&root->snapshot_lock);
1032 * When we're defragging a range, we don't want to kick it off again
1033 * if it is really just waiting for delalloc to send it down.
1034 * If we find a nice big extent or delalloc range for the bytes in the
1035 * file you want to defrag, we return 0 to let you know to skip this
1038 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
1040 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1041 struct extent_map *em = NULL;
1042 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1045 read_lock(&em_tree->lock);
1046 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
1047 read_unlock(&em_tree->lock);
1050 end = extent_map_end(em);
1051 free_extent_map(em);
1052 if (end - offset > thresh)
1055 /* if we already have a nice delalloc here, just stop */
1057 end = count_range_bits(io_tree, &offset, offset + thresh,
1058 thresh, EXTENT_DELALLOC, 1);
1065 * helper function to walk through a file and find extents
1066 * newer than a specific transid, and smaller than thresh.
1068 * This is used by the defragging code to find new and small
1071 static int find_new_extents(struct btrfs_root *root,
1072 struct inode *inode, u64 newer_than,
1073 u64 *off, u32 thresh)
1075 struct btrfs_path *path;
1076 struct btrfs_key min_key;
1077 struct extent_buffer *leaf;
1078 struct btrfs_file_extent_item *extent;
1081 u64 ino = btrfs_ino(BTRFS_I(inode));
1083 path = btrfs_alloc_path();
1087 min_key.objectid = ino;
1088 min_key.type = BTRFS_EXTENT_DATA_KEY;
1089 min_key.offset = *off;
1092 ret = btrfs_search_forward(root, &min_key, path, newer_than);
1096 if (min_key.objectid != ino)
1098 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1101 leaf = path->nodes[0];
1102 extent = btrfs_item_ptr(leaf, path->slots[0],
1103 struct btrfs_file_extent_item);
1105 type = btrfs_file_extent_type(leaf, extent);
1106 if (type == BTRFS_FILE_EXTENT_REG &&
1107 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1108 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1109 *off = min_key.offset;
1110 btrfs_free_path(path);
1115 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1116 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1120 if (min_key.offset == (u64)-1)
1124 btrfs_release_path(path);
1127 btrfs_free_path(path);
1131 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1133 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1134 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1135 struct extent_map *em;
1136 u64 len = PAGE_SIZE;
1139 * hopefully we have this extent in the tree already, try without
1140 * the full extent lock
1142 read_lock(&em_tree->lock);
1143 em = lookup_extent_mapping(em_tree, start, len);
1144 read_unlock(&em_tree->lock);
1147 struct extent_state *cached = NULL;
1148 u64 end = start + len - 1;
1150 /* get the big lock and read metadata off disk */
1151 lock_extent_bits(io_tree, start, end, &cached);
1152 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len);
1153 unlock_extent_cached(io_tree, start, end, &cached);
1162 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1164 struct extent_map *next;
1167 /* this is the last extent */
1168 if (em->start + em->len >= i_size_read(inode))
1171 next = defrag_lookup_extent(inode, em->start + em->len);
1172 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1174 else if ((em->block_start + em->block_len == next->block_start) &&
1175 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1178 free_extent_map(next);
1182 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1183 u64 *last_len, u64 *skip, u64 *defrag_end,
1186 struct extent_map *em;
1188 bool next_mergeable = true;
1189 bool prev_mergeable = true;
1192 * make sure that once we start defragging an extent, we keep on
1195 if (start < *defrag_end)
1200 em = defrag_lookup_extent(inode, start);
1204 /* this will cover holes, and inline extents */
1205 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1211 prev_mergeable = false;
1213 next_mergeable = defrag_check_next_extent(inode, em);
1215 * we hit a real extent, if it is big or the next extent is not a
1216 * real extent, don't bother defragging it
1218 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1219 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1223 * last_len ends up being a counter of how many bytes we've defragged.
1224 * every time we choose not to defrag an extent, we reset *last_len
1225 * so that the next tiny extent will force a defrag.
1227 * The end result of this is that tiny extents before a single big
1228 * extent will force at least part of that big extent to be defragged.
1231 *defrag_end = extent_map_end(em);
1234 *skip = extent_map_end(em);
1238 free_extent_map(em);
1243 * it doesn't do much good to defrag one or two pages
1244 * at a time. This pulls in a nice chunk of pages
1245 * to COW and defrag.
1247 * It also makes sure the delalloc code has enough
1248 * dirty data to avoid making new small extents as part
1251 * It's a good idea to start RA on this range
1252 * before calling this.
1254 static int cluster_pages_for_defrag(struct inode *inode,
1255 struct page **pages,
1256 unsigned long start_index,
1257 unsigned long num_pages)
1259 unsigned long file_end;
1260 u64 isize = i_size_read(inode);
1267 struct btrfs_ordered_extent *ordered;
1268 struct extent_state *cached_state = NULL;
1269 struct extent_io_tree *tree;
1270 struct extent_changeset *data_reserved = NULL;
1271 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1273 file_end = (isize - 1) >> PAGE_SHIFT;
1274 if (!isize || start_index > file_end)
1277 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1279 ret = btrfs_delalloc_reserve_space(BTRFS_I(inode), &data_reserved,
1280 start_index << PAGE_SHIFT,
1281 page_cnt << PAGE_SHIFT);
1285 tree = &BTRFS_I(inode)->io_tree;
1287 /* step one, lock all the pages */
1288 for (i = 0; i < page_cnt; i++) {
1291 page = find_or_create_page(inode->i_mapping,
1292 start_index + i, mask);
1296 page_start = page_offset(page);
1297 page_end = page_start + PAGE_SIZE - 1;
1299 lock_extent_bits(tree, page_start, page_end,
1301 ordered = btrfs_lookup_ordered_extent(BTRFS_I(inode),
1303 unlock_extent_cached(tree, page_start, page_end,
1309 btrfs_start_ordered_extent(inode, ordered, 1);
1310 btrfs_put_ordered_extent(ordered);
1313 * we unlocked the page above, so we need check if
1314 * it was released or not.
1316 if (page->mapping != inode->i_mapping) {
1323 if (!PageUptodate(page)) {
1324 btrfs_readpage(NULL, page);
1326 if (!PageUptodate(page)) {
1334 if (page->mapping != inode->i_mapping) {
1346 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1350 * so now we have a nice long stream of locked
1351 * and up to date pages, lets wait on them
1353 for (i = 0; i < i_done; i++)
1354 wait_on_page_writeback(pages[i]);
1356 page_start = page_offset(pages[0]);
1357 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1359 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1360 page_start, page_end - 1, &cached_state);
1361 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1362 page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
1363 EXTENT_DEFRAG, 0, 0, &cached_state);
1365 if (i_done != page_cnt) {
1366 spin_lock(&BTRFS_I(inode)->lock);
1367 btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
1368 spin_unlock(&BTRFS_I(inode)->lock);
1369 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1370 start_index << PAGE_SHIFT,
1371 (page_cnt - i_done) << PAGE_SHIFT, true);
1375 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1378 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1379 page_start, page_end - 1, &cached_state);
1381 for (i = 0; i < i_done; i++) {
1382 clear_page_dirty_for_io(pages[i]);
1383 ClearPageChecked(pages[i]);
1384 set_page_extent_mapped(pages[i]);
1385 set_page_dirty(pages[i]);
1386 unlock_page(pages[i]);
1389 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1390 extent_changeset_free(data_reserved);
1393 for (i = 0; i < i_done; i++) {
1394 unlock_page(pages[i]);
1397 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1398 start_index << PAGE_SHIFT,
1399 page_cnt << PAGE_SHIFT, true);
1400 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1401 extent_changeset_free(data_reserved);
1406 int btrfs_defrag_file(struct inode *inode, struct file *file,
1407 struct btrfs_ioctl_defrag_range_args *range,
1408 u64 newer_than, unsigned long max_to_defrag)
1410 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1411 struct btrfs_root *root = BTRFS_I(inode)->root;
1412 struct file_ra_state *ra = NULL;
1413 unsigned long last_index;
1414 u64 isize = i_size_read(inode);
1418 u64 newer_off = range->start;
1420 unsigned long ra_index = 0;
1422 int defrag_count = 0;
1423 int compress_type = BTRFS_COMPRESS_ZLIB;
1424 u32 extent_thresh = range->extent_thresh;
1425 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1426 unsigned long cluster = max_cluster;
1427 u64 new_align = ~((u64)SZ_128K - 1);
1428 struct page **pages = NULL;
1429 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1434 if (range->start >= isize)
1438 if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES)
1440 if (range->compress_type)
1441 compress_type = range->compress_type;
1444 if (extent_thresh == 0)
1445 extent_thresh = SZ_256K;
1448 * If we were not given a file, allocate a readahead context. As
1449 * readahead is just an optimization, defrag will work without it so
1450 * we don't error out.
1453 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1455 file_ra_state_init(ra, inode->i_mapping);
1460 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1466 /* find the last page to defrag */
1467 if (range->start + range->len > range->start) {
1468 last_index = min_t(u64, isize - 1,
1469 range->start + range->len - 1) >> PAGE_SHIFT;
1471 last_index = (isize - 1) >> PAGE_SHIFT;
1475 ret = find_new_extents(root, inode, newer_than,
1476 &newer_off, SZ_64K);
1478 range->start = newer_off;
1480 * we always align our defrag to help keep
1481 * the extents in the file evenly spaced
1483 i = (newer_off & new_align) >> PAGE_SHIFT;
1487 i = range->start >> PAGE_SHIFT;
1490 max_to_defrag = last_index - i + 1;
1493 * make writeback starts from i, so the defrag range can be
1494 * written sequentially.
1496 if (i < inode->i_mapping->writeback_index)
1497 inode->i_mapping->writeback_index = i;
1499 while (i <= last_index && defrag_count < max_to_defrag &&
1500 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1502 * make sure we stop running if someone unmounts
1505 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1508 if (btrfs_defrag_cancelled(fs_info)) {
1509 btrfs_debug(fs_info, "defrag_file cancelled");
1514 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1515 extent_thresh, &last_len, &skip,
1516 &defrag_end, do_compress)){
1519 * the should_defrag function tells us how much to skip
1520 * bump our counter by the suggested amount
1522 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1523 i = max(i + 1, next);
1528 cluster = (PAGE_ALIGN(defrag_end) >>
1530 cluster = min(cluster, max_cluster);
1532 cluster = max_cluster;
1535 if (i + cluster > ra_index) {
1536 ra_index = max(i, ra_index);
1538 page_cache_sync_readahead(inode->i_mapping, ra,
1539 file, ra_index, cluster);
1540 ra_index += cluster;
1544 if (IS_SWAPFILE(inode)) {
1548 BTRFS_I(inode)->defrag_compress = compress_type;
1549 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1552 inode_unlock(inode);
1556 defrag_count += ret;
1557 balance_dirty_pages_ratelimited(inode->i_mapping);
1558 inode_unlock(inode);
1561 if (newer_off == (u64)-1)
1567 newer_off = max(newer_off + 1,
1568 (u64)i << PAGE_SHIFT);
1570 ret = find_new_extents(root, inode, newer_than,
1571 &newer_off, SZ_64K);
1573 range->start = newer_off;
1574 i = (newer_off & new_align) >> PAGE_SHIFT;
1581 last_len += ret << PAGE_SHIFT;
1589 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1590 filemap_flush(inode->i_mapping);
1591 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1592 &BTRFS_I(inode)->runtime_flags))
1593 filemap_flush(inode->i_mapping);
1596 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1597 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1598 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1599 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1607 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1608 inode_unlock(inode);
1616 static noinline int btrfs_ioctl_resize(struct file *file,
1619 struct inode *inode = file_inode(file);
1620 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1624 struct btrfs_root *root = BTRFS_I(inode)->root;
1625 struct btrfs_ioctl_vol_args *vol_args;
1626 struct btrfs_trans_handle *trans;
1627 struct btrfs_device *device = NULL;
1630 char *devstr = NULL;
1634 if (!capable(CAP_SYS_ADMIN))
1637 ret = mnt_want_write_file(file);
1641 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
1642 mnt_drop_write_file(file);
1643 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1646 vol_args = memdup_user(arg, sizeof(*vol_args));
1647 if (IS_ERR(vol_args)) {
1648 ret = PTR_ERR(vol_args);
1652 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1654 sizestr = vol_args->name;
1655 devstr = strchr(sizestr, ':');
1657 sizestr = devstr + 1;
1659 devstr = vol_args->name;
1660 ret = kstrtoull(devstr, 10, &devid);
1667 btrfs_info(fs_info, "resizing devid %llu", devid);
1670 device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true);
1672 btrfs_info(fs_info, "resizer unable to find device %llu",
1678 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1680 "resizer unable to apply on readonly device %llu",
1686 if (!strcmp(sizestr, "max"))
1687 new_size = device->bdev->bd_inode->i_size;
1689 if (sizestr[0] == '-') {
1692 } else if (sizestr[0] == '+') {
1696 new_size = memparse(sizestr, &retptr);
1697 if (*retptr != '\0' || new_size == 0) {
1703 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1708 old_size = btrfs_device_get_total_bytes(device);
1711 if (new_size > old_size) {
1715 new_size = old_size - new_size;
1716 } else if (mod > 0) {
1717 if (new_size > ULLONG_MAX - old_size) {
1721 new_size = old_size + new_size;
1724 if (new_size < SZ_256M) {
1728 if (new_size > device->bdev->bd_inode->i_size) {
1733 new_size = round_down(new_size, fs_info->sectorsize);
1735 if (new_size > old_size) {
1736 trans = btrfs_start_transaction(root, 0);
1737 if (IS_ERR(trans)) {
1738 ret = PTR_ERR(trans);
1741 ret = btrfs_grow_device(trans, device, new_size);
1742 btrfs_commit_transaction(trans);
1743 } else if (new_size < old_size) {
1744 ret = btrfs_shrink_device(device, new_size);
1745 } /* equal, nothing need to do */
1747 if (ret == 0 && new_size != old_size)
1748 btrfs_info_in_rcu(fs_info,
1749 "resize device %s (devid %llu) from %llu to %llu",
1750 rcu_str_deref(device->name), device->devid,
1751 old_size, new_size);
1755 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
1756 mnt_drop_write_file(file);
1760 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1761 const char *name, unsigned long fd, int subvol,
1763 struct btrfs_qgroup_inherit *inherit)
1768 if (!S_ISDIR(file_inode(file)->i_mode))
1771 ret = mnt_want_write_file(file);
1775 namelen = strlen(name);
1776 if (strchr(name, '/')) {
1778 goto out_drop_write;
1781 if (name[0] == '.' &&
1782 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1784 goto out_drop_write;
1788 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1789 NULL, readonly, inherit);
1791 struct fd src = fdget(fd);
1792 struct inode *src_inode;
1795 goto out_drop_write;
1798 src_inode = file_inode(src.file);
1799 if (src_inode->i_sb != file_inode(file)->i_sb) {
1800 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1801 "Snapshot src from another FS");
1803 } else if (!inode_owner_or_capable(src_inode)) {
1805 * Subvolume creation is not restricted, but snapshots
1806 * are limited to own subvolumes only
1810 ret = btrfs_mksnapshot(&file->f_path, name, namelen,
1811 BTRFS_I(src_inode)->root,
1817 mnt_drop_write_file(file);
1822 static noinline int btrfs_ioctl_snap_create(struct file *file,
1823 void __user *arg, int subvol)
1825 struct btrfs_ioctl_vol_args *vol_args;
1828 if (!S_ISDIR(file_inode(file)->i_mode))
1831 vol_args = memdup_user(arg, sizeof(*vol_args));
1832 if (IS_ERR(vol_args))
1833 return PTR_ERR(vol_args);
1834 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1836 ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1837 subvol, false, NULL);
1843 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1844 void __user *arg, int subvol)
1846 struct btrfs_ioctl_vol_args_v2 *vol_args;
1848 bool readonly = false;
1849 struct btrfs_qgroup_inherit *inherit = NULL;
1851 if (!S_ISDIR(file_inode(file)->i_mode))
1854 vol_args = memdup_user(arg, sizeof(*vol_args));
1855 if (IS_ERR(vol_args))
1856 return PTR_ERR(vol_args);
1857 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1859 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1864 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1866 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1867 if (vol_args->size > PAGE_SIZE) {
1871 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1872 if (IS_ERR(inherit)) {
1873 ret = PTR_ERR(inherit);
1878 ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1879 subvol, readonly, inherit);
1889 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1892 struct inode *inode = file_inode(file);
1893 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1894 struct btrfs_root *root = BTRFS_I(inode)->root;
1898 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1901 down_read(&fs_info->subvol_sem);
1902 if (btrfs_root_readonly(root))
1903 flags |= BTRFS_SUBVOL_RDONLY;
1904 up_read(&fs_info->subvol_sem);
1906 if (copy_to_user(arg, &flags, sizeof(flags)))
1912 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1915 struct inode *inode = file_inode(file);
1916 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1917 struct btrfs_root *root = BTRFS_I(inode)->root;
1918 struct btrfs_trans_handle *trans;
1923 if (!inode_owner_or_capable(inode))
1926 ret = mnt_want_write_file(file);
1930 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1932 goto out_drop_write;
1935 if (copy_from_user(&flags, arg, sizeof(flags))) {
1937 goto out_drop_write;
1940 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1942 goto out_drop_write;
1945 down_write(&fs_info->subvol_sem);
1948 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1951 root_flags = btrfs_root_flags(&root->root_item);
1952 if (flags & BTRFS_SUBVOL_RDONLY) {
1953 btrfs_set_root_flags(&root->root_item,
1954 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1957 * Block RO -> RW transition if this subvolume is involved in
1960 spin_lock(&root->root_item_lock);
1961 if (root->send_in_progress == 0) {
1962 btrfs_set_root_flags(&root->root_item,
1963 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1964 spin_unlock(&root->root_item_lock);
1966 spin_unlock(&root->root_item_lock);
1968 "Attempt to set subvolume %llu read-write during send",
1969 root->root_key.objectid);
1975 trans = btrfs_start_transaction(root, 1);
1976 if (IS_ERR(trans)) {
1977 ret = PTR_ERR(trans);
1981 ret = btrfs_update_root(trans, fs_info->tree_root,
1982 &root->root_key, &root->root_item);
1984 btrfs_end_transaction(trans);
1988 ret = btrfs_commit_transaction(trans);
1992 btrfs_set_root_flags(&root->root_item, root_flags);
1994 up_write(&fs_info->subvol_sem);
1996 mnt_drop_write_file(file);
2001 static noinline int key_in_sk(struct btrfs_key *key,
2002 struct btrfs_ioctl_search_key *sk)
2004 struct btrfs_key test;
2007 test.objectid = sk->min_objectid;
2008 test.type = sk->min_type;
2009 test.offset = sk->min_offset;
2011 ret = btrfs_comp_cpu_keys(key, &test);
2015 test.objectid = sk->max_objectid;
2016 test.type = sk->max_type;
2017 test.offset = sk->max_offset;
2019 ret = btrfs_comp_cpu_keys(key, &test);
2025 static noinline int copy_to_sk(struct btrfs_path *path,
2026 struct btrfs_key *key,
2027 struct btrfs_ioctl_search_key *sk,
2030 unsigned long *sk_offset,
2034 struct extent_buffer *leaf;
2035 struct btrfs_ioctl_search_header sh;
2036 struct btrfs_key test;
2037 unsigned long item_off;
2038 unsigned long item_len;
2044 leaf = path->nodes[0];
2045 slot = path->slots[0];
2046 nritems = btrfs_header_nritems(leaf);
2048 if (btrfs_header_generation(leaf) > sk->max_transid) {
2052 found_transid = btrfs_header_generation(leaf);
2054 for (i = slot; i < nritems; i++) {
2055 item_off = btrfs_item_ptr_offset(leaf, i);
2056 item_len = btrfs_item_size_nr(leaf, i);
2058 btrfs_item_key_to_cpu(leaf, key, i);
2059 if (!key_in_sk(key, sk))
2062 if (sizeof(sh) + item_len > *buf_size) {
2069 * return one empty item back for v1, which does not
2073 *buf_size = sizeof(sh) + item_len;
2078 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2083 sh.objectid = key->objectid;
2084 sh.offset = key->offset;
2085 sh.type = key->type;
2087 sh.transid = found_transid;
2089 /* copy search result header */
2090 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2095 *sk_offset += sizeof(sh);
2098 char __user *up = ubuf + *sk_offset;
2100 if (read_extent_buffer_to_user(leaf, up,
2101 item_off, item_len)) {
2106 *sk_offset += item_len;
2110 if (ret) /* -EOVERFLOW from above */
2113 if (*num_found >= sk->nr_items) {
2120 test.objectid = sk->max_objectid;
2121 test.type = sk->max_type;
2122 test.offset = sk->max_offset;
2123 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2125 else if (key->offset < (u64)-1)
2127 else if (key->type < (u8)-1) {
2130 } else if (key->objectid < (u64)-1) {
2138 * 0: all items from this leaf copied, continue with next
2139 * 1: * more items can be copied, but unused buffer is too small
2140 * * all items were found
2141 * Either way, it will stops the loop which iterates to the next
2143 * -EOVERFLOW: item was to large for buffer
2144 * -EFAULT: could not copy extent buffer back to userspace
2149 static noinline int search_ioctl(struct inode *inode,
2150 struct btrfs_ioctl_search_key *sk,
2154 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2155 struct btrfs_root *root;
2156 struct btrfs_key key;
2157 struct btrfs_path *path;
2160 unsigned long sk_offset = 0;
2162 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2163 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2167 path = btrfs_alloc_path();
2171 if (sk->tree_id == 0) {
2172 /* search the root of the inode that was passed */
2173 root = btrfs_grab_root(BTRFS_I(inode)->root);
2175 root = btrfs_get_fs_root(info, sk->tree_id, true);
2177 btrfs_free_path(path);
2178 return PTR_ERR(root);
2182 key.objectid = sk->min_objectid;
2183 key.type = sk->min_type;
2184 key.offset = sk->min_offset;
2187 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2193 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2194 &sk_offset, &num_found);
2195 btrfs_release_path(path);
2203 sk->nr_items = num_found;
2204 btrfs_put_root(root);
2205 btrfs_free_path(path);
2209 static noinline int btrfs_ioctl_tree_search(struct file *file,
2212 struct btrfs_ioctl_search_args __user *uargs;
2213 struct btrfs_ioctl_search_key sk;
2214 struct inode *inode;
2218 if (!capable(CAP_SYS_ADMIN))
2221 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2223 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2226 buf_size = sizeof(uargs->buf);
2228 inode = file_inode(file);
2229 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2232 * In the origin implementation an overflow is handled by returning a
2233 * search header with a len of zero, so reset ret.
2235 if (ret == -EOVERFLOW)
2238 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2243 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2246 struct btrfs_ioctl_search_args_v2 __user *uarg;
2247 struct btrfs_ioctl_search_args_v2 args;
2248 struct inode *inode;
2251 const size_t buf_limit = SZ_16M;
2253 if (!capable(CAP_SYS_ADMIN))
2256 /* copy search header and buffer size */
2257 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2258 if (copy_from_user(&args, uarg, sizeof(args)))
2261 buf_size = args.buf_size;
2263 /* limit result size to 16MB */
2264 if (buf_size > buf_limit)
2265 buf_size = buf_limit;
2267 inode = file_inode(file);
2268 ret = search_ioctl(inode, &args.key, &buf_size,
2269 (char __user *)(&uarg->buf[0]));
2270 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2272 else if (ret == -EOVERFLOW &&
2273 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2280 * Search INODE_REFs to identify path name of 'dirid' directory
2281 * in a 'tree_id' tree. and sets path name to 'name'.
2283 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2284 u64 tree_id, u64 dirid, char *name)
2286 struct btrfs_root *root;
2287 struct btrfs_key key;
2293 struct btrfs_inode_ref *iref;
2294 struct extent_buffer *l;
2295 struct btrfs_path *path;
2297 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2302 path = btrfs_alloc_path();
2306 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2308 root = btrfs_get_fs_root(info, tree_id, true);
2310 ret = PTR_ERR(root);
2315 key.objectid = dirid;
2316 key.type = BTRFS_INODE_REF_KEY;
2317 key.offset = (u64)-1;
2320 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2324 ret = btrfs_previous_item(root, path, dirid,
2325 BTRFS_INODE_REF_KEY);
2335 slot = path->slots[0];
2336 btrfs_item_key_to_cpu(l, &key, slot);
2338 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2339 len = btrfs_inode_ref_name_len(l, iref);
2341 total_len += len + 1;
2343 ret = -ENAMETOOLONG;
2348 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2350 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2353 btrfs_release_path(path);
2354 key.objectid = key.offset;
2355 key.offset = (u64)-1;
2356 dirid = key.objectid;
2358 memmove(name, ptr, total_len);
2359 name[total_len] = '\0';
2362 btrfs_put_root(root);
2363 btrfs_free_path(path);
2367 static int btrfs_search_path_in_tree_user(struct inode *inode,
2368 struct btrfs_ioctl_ino_lookup_user_args *args)
2370 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2371 struct super_block *sb = inode->i_sb;
2372 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2373 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2374 u64 dirid = args->dirid;
2375 unsigned long item_off;
2376 unsigned long item_len;
2377 struct btrfs_inode_ref *iref;
2378 struct btrfs_root_ref *rref;
2379 struct btrfs_root *root = NULL;
2380 struct btrfs_path *path;
2381 struct btrfs_key key, key2;
2382 struct extent_buffer *leaf;
2383 struct inode *temp_inode;
2390 path = btrfs_alloc_path();
2395 * If the bottom subvolume does not exist directly under upper_limit,
2396 * construct the path in from the bottom up.
2398 if (dirid != upper_limit.objectid) {
2399 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2401 root = btrfs_get_fs_root(fs_info, treeid, true);
2403 ret = PTR_ERR(root);
2407 key.objectid = dirid;
2408 key.type = BTRFS_INODE_REF_KEY;
2409 key.offset = (u64)-1;
2411 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2414 } else if (ret > 0) {
2415 ret = btrfs_previous_item(root, path, dirid,
2416 BTRFS_INODE_REF_KEY);
2419 } else if (ret > 0) {
2425 leaf = path->nodes[0];
2426 slot = path->slots[0];
2427 btrfs_item_key_to_cpu(leaf, &key, slot);
2429 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2430 len = btrfs_inode_ref_name_len(leaf, iref);
2432 total_len += len + 1;
2433 if (ptr < args->path) {
2434 ret = -ENAMETOOLONG;
2439 read_extent_buffer(leaf, ptr,
2440 (unsigned long)(iref + 1), len);
2442 /* Check the read+exec permission of this directory */
2443 ret = btrfs_previous_item(root, path, dirid,
2444 BTRFS_INODE_ITEM_KEY);
2447 } else if (ret > 0) {
2452 leaf = path->nodes[0];
2453 slot = path->slots[0];
2454 btrfs_item_key_to_cpu(leaf, &key2, slot);
2455 if (key2.objectid != dirid) {
2460 temp_inode = btrfs_iget(sb, key2.objectid, root);
2461 if (IS_ERR(temp_inode)) {
2462 ret = PTR_ERR(temp_inode);
2465 ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC);
2472 if (key.offset == upper_limit.objectid)
2474 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2479 btrfs_release_path(path);
2480 key.objectid = key.offset;
2481 key.offset = (u64)-1;
2482 dirid = key.objectid;
2485 memmove(args->path, ptr, total_len);
2486 args->path[total_len] = '\0';
2487 btrfs_put_root(root);
2489 btrfs_release_path(path);
2492 /* Get the bottom subvolume's name from ROOT_REF */
2493 key.objectid = treeid;
2494 key.type = BTRFS_ROOT_REF_KEY;
2495 key.offset = args->treeid;
2496 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2499 } else if (ret > 0) {
2504 leaf = path->nodes[0];
2505 slot = path->slots[0];
2506 btrfs_item_key_to_cpu(leaf, &key, slot);
2508 item_off = btrfs_item_ptr_offset(leaf, slot);
2509 item_len = btrfs_item_size_nr(leaf, slot);
2510 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2511 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2512 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2517 /* Copy subvolume's name */
2518 item_off += sizeof(struct btrfs_root_ref);
2519 item_len -= sizeof(struct btrfs_root_ref);
2520 read_extent_buffer(leaf, args->name, item_off, item_len);
2521 args->name[item_len] = 0;
2524 btrfs_put_root(root);
2526 btrfs_free_path(path);
2530 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2533 struct btrfs_ioctl_ino_lookup_args *args;
2534 struct inode *inode;
2537 args = memdup_user(argp, sizeof(*args));
2539 return PTR_ERR(args);
2541 inode = file_inode(file);
2544 * Unprivileged query to obtain the containing subvolume root id. The
2545 * path is reset so it's consistent with btrfs_search_path_in_tree.
2547 if (args->treeid == 0)
2548 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2550 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2555 if (!capable(CAP_SYS_ADMIN)) {
2560 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2561 args->treeid, args->objectid,
2565 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2573 * Version of ino_lookup ioctl (unprivileged)
2575 * The main differences from ino_lookup ioctl are:
2577 * 1. Read + Exec permission will be checked using inode_permission() during
2578 * path construction. -EACCES will be returned in case of failure.
2579 * 2. Path construction will be stopped at the inode number which corresponds
2580 * to the fd with which this ioctl is called. If constructed path does not
2581 * exist under fd's inode, -EACCES will be returned.
2582 * 3. The name of bottom subvolume is also searched and filled.
2584 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2586 struct btrfs_ioctl_ino_lookup_user_args *args;
2587 struct inode *inode;
2590 args = memdup_user(argp, sizeof(*args));
2592 return PTR_ERR(args);
2594 inode = file_inode(file);
2596 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2597 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2599 * The subvolume does not exist under fd with which this is
2606 ret = btrfs_search_path_in_tree_user(inode, args);
2608 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2615 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2616 static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2618 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2619 struct btrfs_fs_info *fs_info;
2620 struct btrfs_root *root;
2621 struct btrfs_path *path;
2622 struct btrfs_key key;
2623 struct btrfs_root_item *root_item;
2624 struct btrfs_root_ref *rref;
2625 struct extent_buffer *leaf;
2626 unsigned long item_off;
2627 unsigned long item_len;
2628 struct inode *inode;
2632 path = btrfs_alloc_path();
2636 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2638 btrfs_free_path(path);
2642 inode = file_inode(file);
2643 fs_info = BTRFS_I(inode)->root->fs_info;
2645 /* Get root_item of inode's subvolume */
2646 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2647 root = btrfs_get_fs_root(fs_info, key.objectid, true);
2649 ret = PTR_ERR(root);
2652 root_item = &root->root_item;
2654 subvol_info->treeid = key.objectid;
2656 subvol_info->generation = btrfs_root_generation(root_item);
2657 subvol_info->flags = btrfs_root_flags(root_item);
2659 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2660 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2662 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2665 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2666 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2667 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2669 subvol_info->otransid = btrfs_root_otransid(root_item);
2670 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2671 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2673 subvol_info->stransid = btrfs_root_stransid(root_item);
2674 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2675 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2677 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2678 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2679 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2681 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2682 /* Search root tree for ROOT_BACKREF of this subvolume */
2683 key.type = BTRFS_ROOT_BACKREF_KEY;
2685 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2688 } else if (path->slots[0] >=
2689 btrfs_header_nritems(path->nodes[0])) {
2690 ret = btrfs_next_leaf(fs_info->tree_root, path);
2693 } else if (ret > 0) {
2699 leaf = path->nodes[0];
2700 slot = path->slots[0];
2701 btrfs_item_key_to_cpu(leaf, &key, slot);
2702 if (key.objectid == subvol_info->treeid &&
2703 key.type == BTRFS_ROOT_BACKREF_KEY) {
2704 subvol_info->parent_id = key.offset;
2706 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2707 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2709 item_off = btrfs_item_ptr_offset(leaf, slot)
2710 + sizeof(struct btrfs_root_ref);
2711 item_len = btrfs_item_size_nr(leaf, slot)
2712 - sizeof(struct btrfs_root_ref);
2713 read_extent_buffer(leaf, subvol_info->name,
2714 item_off, item_len);
2721 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2725 btrfs_put_root(root);
2727 btrfs_free_path(path);
2733 * Return ROOT_REF information of the subvolume containing this inode
2734 * except the subvolume name.
2736 static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2738 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2739 struct btrfs_root_ref *rref;
2740 struct btrfs_root *root;
2741 struct btrfs_path *path;
2742 struct btrfs_key key;
2743 struct extent_buffer *leaf;
2744 struct inode *inode;
2750 path = btrfs_alloc_path();
2754 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2755 if (IS_ERR(rootrefs)) {
2756 btrfs_free_path(path);
2757 return PTR_ERR(rootrefs);
2760 inode = file_inode(file);
2761 root = BTRFS_I(inode)->root->fs_info->tree_root;
2762 objectid = BTRFS_I(inode)->root->root_key.objectid;
2764 key.objectid = objectid;
2765 key.type = BTRFS_ROOT_REF_KEY;
2766 key.offset = rootrefs->min_treeid;
2769 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2772 } else if (path->slots[0] >=
2773 btrfs_header_nritems(path->nodes[0])) {
2774 ret = btrfs_next_leaf(root, path);
2777 } else if (ret > 0) {
2783 leaf = path->nodes[0];
2784 slot = path->slots[0];
2786 btrfs_item_key_to_cpu(leaf, &key, slot);
2787 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2792 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2797 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2798 rootrefs->rootref[found].treeid = key.offset;
2799 rootrefs->rootref[found].dirid =
2800 btrfs_root_ref_dirid(leaf, rref);
2803 ret = btrfs_next_item(root, path);
2806 } else if (ret > 0) {
2813 if (!ret || ret == -EOVERFLOW) {
2814 rootrefs->num_items = found;
2815 /* update min_treeid for next search */
2817 rootrefs->min_treeid =
2818 rootrefs->rootref[found - 1].treeid + 1;
2819 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2824 btrfs_free_path(path);
2829 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2833 struct dentry *parent = file->f_path.dentry;
2834 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2835 struct dentry *dentry;
2836 struct inode *dir = d_inode(parent);
2837 struct inode *inode;
2838 struct btrfs_root *root = BTRFS_I(dir)->root;
2839 struct btrfs_root *dest = NULL;
2840 struct btrfs_ioctl_vol_args *vol_args = NULL;
2841 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2842 char *subvol_name, *subvol_name_ptr = NULL;
2845 bool destroy_parent = false;
2848 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2849 if (IS_ERR(vol_args2))
2850 return PTR_ERR(vol_args2);
2852 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2858 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2859 * name, same as v1 currently does.
2861 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2862 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2863 subvol_name = vol_args2->name;
2865 err = mnt_want_write_file(file);
2869 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2874 err = mnt_want_write_file(file);
2878 dentry = btrfs_get_dentry(fs_info->sb,
2879 BTRFS_FIRST_FREE_OBJECTID,
2880 vol_args2->subvolid, 0, 0);
2881 if (IS_ERR(dentry)) {
2882 err = PTR_ERR(dentry);
2883 goto out_drop_write;
2887 * Change the default parent since the subvolume being
2888 * deleted can be outside of the current mount point.
2890 parent = btrfs_get_parent(dentry);
2893 * At this point dentry->d_name can point to '/' if the
2894 * subvolume we want to destroy is outsite of the
2895 * current mount point, so we need to release the
2896 * current dentry and execute the lookup to return a new
2897 * one with ->d_name pointing to the
2898 * <mount point>/subvol_name.
2901 if (IS_ERR(parent)) {
2902 err = PTR_ERR(parent);
2903 goto out_drop_write;
2905 dir = d_inode(parent);
2908 * If v2 was used with SPEC_BY_ID, a new parent was
2909 * allocated since the subvolume can be outside of the
2910 * current mount point. Later on we need to release this
2911 * new parent dentry.
2913 destroy_parent = true;
2915 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2916 fs_info, vol_args2->subvolid);
2917 if (IS_ERR(subvol_name_ptr)) {
2918 err = PTR_ERR(subvol_name_ptr);
2921 /* subvol_name_ptr is already NULL termined */
2922 subvol_name = (char *)kbasename(subvol_name_ptr);
2925 vol_args = memdup_user(arg, sizeof(*vol_args));
2926 if (IS_ERR(vol_args))
2927 return PTR_ERR(vol_args);
2929 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
2930 subvol_name = vol_args->name;
2932 err = mnt_want_write_file(file);
2937 subvol_namelen = strlen(subvol_name);
2939 if (strchr(subvol_name, '/') ||
2940 strncmp(subvol_name, "..", subvol_namelen) == 0) {
2942 goto free_subvol_name;
2945 if (!S_ISDIR(dir->i_mode)) {
2947 goto free_subvol_name;
2950 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2952 goto free_subvol_name;
2953 dentry = lookup_one_len(subvol_name, parent, subvol_namelen);
2954 if (IS_ERR(dentry)) {
2955 err = PTR_ERR(dentry);
2956 goto out_unlock_dir;
2959 if (d_really_is_negative(dentry)) {
2964 inode = d_inode(dentry);
2965 dest = BTRFS_I(inode)->root;
2966 if (!capable(CAP_SYS_ADMIN)) {
2968 * Regular user. Only allow this with a special mount
2969 * option, when the user has write+exec access to the
2970 * subvol root, and when rmdir(2) would have been
2973 * Note that this is _not_ check that the subvol is
2974 * empty or doesn't contain data that we wouldn't
2975 * otherwise be able to delete.
2977 * Users who want to delete empty subvols should try
2981 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2985 * Do not allow deletion if the parent dir is the same
2986 * as the dir to be deleted. That means the ioctl
2987 * must be called on the dentry referencing the root
2988 * of the subvol, not a random directory contained
2995 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
3000 /* check if subvolume may be deleted by a user */
3001 err = btrfs_may_delete(dir, dentry, 1);
3005 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3011 err = btrfs_delete_subvolume(dir, dentry);
3012 inode_unlock(inode);
3014 fsnotify_rmdir(dir, dentry);
3023 kfree(subvol_name_ptr);
3028 mnt_drop_write_file(file);
3035 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
3037 struct inode *inode = file_inode(file);
3038 struct btrfs_root *root = BTRFS_I(inode)->root;
3039 struct btrfs_ioctl_defrag_range_args *range;
3042 ret = mnt_want_write_file(file);
3046 if (btrfs_root_readonly(root)) {
3051 switch (inode->i_mode & S_IFMT) {
3053 if (!capable(CAP_SYS_ADMIN)) {
3057 ret = btrfs_defrag_root(root);
3061 * Note that this does not check the file descriptor for write
3062 * access. This prevents defragmenting executables that are
3063 * running and allows defrag on files open in read-only mode.
3065 if (!capable(CAP_SYS_ADMIN) &&
3066 inode_permission(inode, MAY_WRITE)) {
3071 range = kzalloc(sizeof(*range), GFP_KERNEL);
3078 if (copy_from_user(range, argp,
3084 /* compression requires us to start the IO */
3085 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3086 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
3087 range->extent_thresh = (u32)-1;
3090 /* the rest are all set to zero by kzalloc */
3091 range->len = (u64)-1;
3093 ret = btrfs_defrag_file(file_inode(file), file,
3094 range, BTRFS_OLDEST_GENERATION, 0);
3103 mnt_drop_write_file(file);
3107 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
3109 struct btrfs_ioctl_vol_args *vol_args;
3112 if (!capable(CAP_SYS_ADMIN))
3115 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
3116 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3118 vol_args = memdup_user(arg, sizeof(*vol_args));
3119 if (IS_ERR(vol_args)) {
3120 ret = PTR_ERR(vol_args);
3124 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3125 ret = btrfs_init_new_device(fs_info, vol_args->name);
3128 btrfs_info(fs_info, "disk added %s", vol_args->name);
3132 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3136 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
3138 struct inode *inode = file_inode(file);
3139 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3140 struct btrfs_ioctl_vol_args_v2 *vol_args;
3143 if (!capable(CAP_SYS_ADMIN))
3146 ret = mnt_want_write_file(file);
3150 vol_args = memdup_user(arg, sizeof(*vol_args));
3151 if (IS_ERR(vol_args)) {
3152 ret = PTR_ERR(vol_args);
3156 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
3161 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3162 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3166 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
3167 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
3169 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3170 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3172 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3175 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3176 btrfs_info(fs_info, "device deleted: id %llu",
3179 btrfs_info(fs_info, "device deleted: %s",
3185 mnt_drop_write_file(file);
3189 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3191 struct inode *inode = file_inode(file);
3192 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3193 struct btrfs_ioctl_vol_args *vol_args;
3196 if (!capable(CAP_SYS_ADMIN))
3199 ret = mnt_want_write_file(file);
3203 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3204 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3205 goto out_drop_write;
3208 vol_args = memdup_user(arg, sizeof(*vol_args));
3209 if (IS_ERR(vol_args)) {
3210 ret = PTR_ERR(vol_args);
3214 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3215 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3218 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3221 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3223 mnt_drop_write_file(file);
3228 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3231 struct btrfs_ioctl_fs_info_args *fi_args;
3232 struct btrfs_device *device;
3233 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3237 fi_args = memdup_user(arg, sizeof(*fi_args));
3238 if (IS_ERR(fi_args))
3239 return PTR_ERR(fi_args);
3241 flags_in = fi_args->flags;
3242 memset(fi_args, 0, sizeof(*fi_args));
3245 fi_args->num_devices = fs_devices->num_devices;
3247 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3248 if (device->devid > fi_args->max_id)
3249 fi_args->max_id = device->devid;
3253 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
3254 fi_args->nodesize = fs_info->nodesize;
3255 fi_args->sectorsize = fs_info->sectorsize;
3256 fi_args->clone_alignment = fs_info->sectorsize;
3258 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
3259 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
3260 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
3261 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
3264 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
3265 fi_args->generation = fs_info->generation;
3266 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
3269 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
3270 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
3271 sizeof(fi_args->metadata_uuid));
3272 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
3275 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3282 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3285 struct btrfs_ioctl_dev_info_args *di_args;
3286 struct btrfs_device *dev;
3288 char *s_uuid = NULL;
3290 di_args = memdup_user(arg, sizeof(*di_args));
3291 if (IS_ERR(di_args))
3292 return PTR_ERR(di_args);
3294 if (!btrfs_is_empty_uuid(di_args->uuid))
3295 s_uuid = di_args->uuid;
3298 dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid,
3306 di_args->devid = dev->devid;
3307 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3308 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3309 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3311 strncpy(di_args->path, rcu_str_deref(dev->name),
3312 sizeof(di_args->path) - 1);
3313 di_args->path[sizeof(di_args->path) - 1] = 0;
3315 di_args->path[0] = '\0';
3320 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3327 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3329 struct inode *inode = file_inode(file);
3330 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3331 struct btrfs_root *root = BTRFS_I(inode)->root;
3332 struct btrfs_root *new_root;
3333 struct btrfs_dir_item *di;
3334 struct btrfs_trans_handle *trans;
3335 struct btrfs_path *path = NULL;
3336 struct btrfs_disk_key disk_key;
3341 if (!capable(CAP_SYS_ADMIN))
3344 ret = mnt_want_write_file(file);
3348 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3354 objectid = BTRFS_FS_TREE_OBJECTID;
3356 new_root = btrfs_get_fs_root(fs_info, objectid, true);
3357 if (IS_ERR(new_root)) {
3358 ret = PTR_ERR(new_root);
3361 if (!is_fstree(new_root->root_key.objectid)) {
3366 path = btrfs_alloc_path();
3371 path->leave_spinning = 1;
3373 trans = btrfs_start_transaction(root, 1);
3374 if (IS_ERR(trans)) {
3375 ret = PTR_ERR(trans);
3379 dir_id = btrfs_super_root_dir(fs_info->super_copy);
3380 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3381 dir_id, "default", 7, 1);
3382 if (IS_ERR_OR_NULL(di)) {
3383 btrfs_release_path(path);
3384 btrfs_end_transaction(trans);
3386 "Umm, you don't have the default diritem, this isn't going to work");
3391 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3392 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3393 btrfs_mark_buffer_dirty(path->nodes[0]);
3394 btrfs_release_path(path);
3396 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3397 btrfs_end_transaction(trans);
3399 btrfs_put_root(new_root);
3400 btrfs_free_path(path);
3402 mnt_drop_write_file(file);
3406 static void get_block_group_info(struct list_head *groups_list,
3407 struct btrfs_ioctl_space_info *space)
3409 struct btrfs_block_group *block_group;
3411 space->total_bytes = 0;
3412 space->used_bytes = 0;
3414 list_for_each_entry(block_group, groups_list, list) {
3415 space->flags = block_group->flags;
3416 space->total_bytes += block_group->length;
3417 space->used_bytes += block_group->used;
3421 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3424 struct btrfs_ioctl_space_args space_args;
3425 struct btrfs_ioctl_space_info space;
3426 struct btrfs_ioctl_space_info *dest;
3427 struct btrfs_ioctl_space_info *dest_orig;
3428 struct btrfs_ioctl_space_info __user *user_dest;
3429 struct btrfs_space_info *info;
3430 static const u64 types[] = {
3431 BTRFS_BLOCK_GROUP_DATA,
3432 BTRFS_BLOCK_GROUP_SYSTEM,
3433 BTRFS_BLOCK_GROUP_METADATA,
3434 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3442 if (copy_from_user(&space_args,
3443 (struct btrfs_ioctl_space_args __user *)arg,
3444 sizeof(space_args)))
3447 for (i = 0; i < num_types; i++) {
3448 struct btrfs_space_info *tmp;
3452 list_for_each_entry_rcu(tmp, &fs_info->space_info,
3454 if (tmp->flags == types[i]) {
3464 down_read(&info->groups_sem);
3465 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3466 if (!list_empty(&info->block_groups[c]))
3469 up_read(&info->groups_sem);
3473 * Global block reserve, exported as a space_info
3477 /* space_slots == 0 means they are asking for a count */
3478 if (space_args.space_slots == 0) {
3479 space_args.total_spaces = slot_count;
3483 slot_count = min_t(u64, space_args.space_slots, slot_count);
3485 alloc_size = sizeof(*dest) * slot_count;
3487 /* we generally have at most 6 or so space infos, one for each raid
3488 * level. So, a whole page should be more than enough for everyone
3490 if (alloc_size > PAGE_SIZE)
3493 space_args.total_spaces = 0;
3494 dest = kmalloc(alloc_size, GFP_KERNEL);
3499 /* now we have a buffer to copy into */
3500 for (i = 0; i < num_types; i++) {
3501 struct btrfs_space_info *tmp;
3508 list_for_each_entry_rcu(tmp, &fs_info->space_info,
3510 if (tmp->flags == types[i]) {
3519 down_read(&info->groups_sem);
3520 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3521 if (!list_empty(&info->block_groups[c])) {
3522 get_block_group_info(&info->block_groups[c],
3524 memcpy(dest, &space, sizeof(space));
3526 space_args.total_spaces++;
3532 up_read(&info->groups_sem);
3536 * Add global block reserve
3539 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3541 spin_lock(&block_rsv->lock);
3542 space.total_bytes = block_rsv->size;
3543 space.used_bytes = block_rsv->size - block_rsv->reserved;
3544 spin_unlock(&block_rsv->lock);
3545 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3546 memcpy(dest, &space, sizeof(space));
3547 space_args.total_spaces++;
3550 user_dest = (struct btrfs_ioctl_space_info __user *)
3551 (arg + sizeof(struct btrfs_ioctl_space_args));
3553 if (copy_to_user(user_dest, dest_orig, alloc_size))
3558 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3564 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3567 struct btrfs_trans_handle *trans;
3571 trans = btrfs_attach_transaction_barrier(root);
3572 if (IS_ERR(trans)) {
3573 if (PTR_ERR(trans) != -ENOENT)
3574 return PTR_ERR(trans);
3576 /* No running transaction, don't bother */
3577 transid = root->fs_info->last_trans_committed;
3580 transid = trans->transid;
3581 ret = btrfs_commit_transaction_async(trans, 0);
3583 btrfs_end_transaction(trans);
3588 if (copy_to_user(argp, &transid, sizeof(transid)))
3593 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3599 if (copy_from_user(&transid, argp, sizeof(transid)))
3602 transid = 0; /* current trans */
3604 return btrfs_wait_for_commit(fs_info, transid);
3607 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3609 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3610 struct btrfs_ioctl_scrub_args *sa;
3613 if (!capable(CAP_SYS_ADMIN))
3616 sa = memdup_user(arg, sizeof(*sa));
3620 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3621 ret = mnt_want_write_file(file);
3626 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3627 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3631 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3632 * error. This is important as it allows user space to know how much
3633 * progress scrub has done. For example, if scrub is canceled we get
3634 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3635 * space. Later user space can inspect the progress from the structure
3636 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3637 * previously (btrfs-progs does this).
3638 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3639 * then return -EFAULT to signal the structure was not copied or it may
3640 * be corrupt and unreliable due to a partial copy.
3642 if (copy_to_user(arg, sa, sizeof(*sa)))
3645 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3646 mnt_drop_write_file(file);
3652 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3654 if (!capable(CAP_SYS_ADMIN))
3657 return btrfs_scrub_cancel(fs_info);
3660 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3663 struct btrfs_ioctl_scrub_args *sa;
3666 if (!capable(CAP_SYS_ADMIN))
3669 sa = memdup_user(arg, sizeof(*sa));
3673 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3675 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3682 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3685 struct btrfs_ioctl_get_dev_stats *sa;
3688 sa = memdup_user(arg, sizeof(*sa));
3692 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3697 ret = btrfs_get_dev_stats(fs_info, sa);
3699 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3706 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3709 struct btrfs_ioctl_dev_replace_args *p;
3712 if (!capable(CAP_SYS_ADMIN))
3715 p = memdup_user(arg, sizeof(*p));
3720 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3721 if (sb_rdonly(fs_info->sb)) {
3725 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3726 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3728 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3729 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3732 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3733 btrfs_dev_replace_status(fs_info, p);
3736 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3737 p->result = btrfs_dev_replace_cancel(fs_info);
3745 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3752 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3758 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3759 struct inode_fs_paths *ipath = NULL;
3760 struct btrfs_path *path;
3762 if (!capable(CAP_DAC_READ_SEARCH))
3765 path = btrfs_alloc_path();
3771 ipa = memdup_user(arg, sizeof(*ipa));
3778 size = min_t(u32, ipa->size, 4096);
3779 ipath = init_ipath(size, root, path);
3780 if (IS_ERR(ipath)) {
3781 ret = PTR_ERR(ipath);
3786 ret = paths_from_inode(ipa->inum, ipath);
3790 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3791 rel_ptr = ipath->fspath->val[i] -
3792 (u64)(unsigned long)ipath->fspath->val;
3793 ipath->fspath->val[i] = rel_ptr;
3796 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3797 ipath->fspath, size);
3804 btrfs_free_path(path);
3811 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3813 struct btrfs_data_container *inodes = ctx;
3814 const size_t c = 3 * sizeof(u64);
3816 if (inodes->bytes_left >= c) {
3817 inodes->bytes_left -= c;
3818 inodes->val[inodes->elem_cnt] = inum;
3819 inodes->val[inodes->elem_cnt + 1] = offset;
3820 inodes->val[inodes->elem_cnt + 2] = root;
3821 inodes->elem_cnt += 3;
3823 inodes->bytes_missing += c - inodes->bytes_left;
3824 inodes->bytes_left = 0;
3825 inodes->elem_missed += 3;
3831 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3832 void __user *arg, int version)
3836 struct btrfs_ioctl_logical_ino_args *loi;
3837 struct btrfs_data_container *inodes = NULL;
3838 struct btrfs_path *path = NULL;
3841 if (!capable(CAP_SYS_ADMIN))
3844 loi = memdup_user(arg, sizeof(*loi));
3846 return PTR_ERR(loi);
3849 ignore_offset = false;
3850 size = min_t(u32, loi->size, SZ_64K);
3852 /* All reserved bits must be 0 for now */
3853 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3857 /* Only accept flags we have defined so far */
3858 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3862 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3863 size = min_t(u32, loi->size, SZ_16M);
3866 path = btrfs_alloc_path();
3872 inodes = init_data_container(size);
3873 if (IS_ERR(inodes)) {
3874 ret = PTR_ERR(inodes);
3879 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3880 build_ino_list, inodes, ignore_offset);
3886 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3892 btrfs_free_path(path);
3900 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3901 struct btrfs_ioctl_balance_args *bargs)
3903 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3905 bargs->flags = bctl->flags;
3907 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3908 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3909 if (atomic_read(&fs_info->balance_pause_req))
3910 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3911 if (atomic_read(&fs_info->balance_cancel_req))
3912 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3914 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3915 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3916 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3918 spin_lock(&fs_info->balance_lock);
3919 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3920 spin_unlock(&fs_info->balance_lock);
3923 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3925 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3926 struct btrfs_fs_info *fs_info = root->fs_info;
3927 struct btrfs_ioctl_balance_args *bargs;
3928 struct btrfs_balance_control *bctl;
3929 bool need_unlock; /* for mut. excl. ops lock */
3932 if (!capable(CAP_SYS_ADMIN))
3935 ret = mnt_want_write_file(file);
3940 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3941 mutex_lock(&fs_info->balance_mutex);
3947 * mut. excl. ops lock is locked. Three possibilities:
3948 * (1) some other op is running
3949 * (2) balance is running
3950 * (3) balance is paused -- special case (think resume)
3952 mutex_lock(&fs_info->balance_mutex);
3953 if (fs_info->balance_ctl) {
3954 /* this is either (2) or (3) */
3955 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3956 mutex_unlock(&fs_info->balance_mutex);
3958 * Lock released to allow other waiters to continue,
3959 * we'll reexamine the status again.
3961 mutex_lock(&fs_info->balance_mutex);
3963 if (fs_info->balance_ctl &&
3964 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3966 need_unlock = false;
3970 mutex_unlock(&fs_info->balance_mutex);
3974 mutex_unlock(&fs_info->balance_mutex);
3980 mutex_unlock(&fs_info->balance_mutex);
3981 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3986 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
3989 bargs = memdup_user(arg, sizeof(*bargs));
3990 if (IS_ERR(bargs)) {
3991 ret = PTR_ERR(bargs);
3995 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3996 if (!fs_info->balance_ctl) {
4001 bctl = fs_info->balance_ctl;
4002 spin_lock(&fs_info->balance_lock);
4003 bctl->flags |= BTRFS_BALANCE_RESUME;
4004 spin_unlock(&fs_info->balance_lock);
4012 if (fs_info->balance_ctl) {
4017 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4024 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4025 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4026 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4028 bctl->flags = bargs->flags;
4030 /* balance everything - no filters */
4031 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4034 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4041 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to
4042 * btrfs_balance. bctl is freed in reset_balance_state, or, if
4043 * restriper was paused all the way until unmount, in free_fs_info.
4044 * The flag should be cleared after reset_balance_state.
4046 need_unlock = false;
4048 ret = btrfs_balance(fs_info, bctl, bargs);
4051 if ((ret == 0 || ret == -ECANCELED) && arg) {
4052 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4061 mutex_unlock(&fs_info->balance_mutex);
4063 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4065 mnt_drop_write_file(file);
4069 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4071 if (!capable(CAP_SYS_ADMIN))
4075 case BTRFS_BALANCE_CTL_PAUSE:
4076 return btrfs_pause_balance(fs_info);
4077 case BTRFS_BALANCE_CTL_CANCEL:
4078 return btrfs_cancel_balance(fs_info);
4084 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4087 struct btrfs_ioctl_balance_args *bargs;
4090 if (!capable(CAP_SYS_ADMIN))
4093 mutex_lock(&fs_info->balance_mutex);
4094 if (!fs_info->balance_ctl) {
4099 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4105 btrfs_update_ioctl_balance_args(fs_info, bargs);
4107 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4112 mutex_unlock(&fs_info->balance_mutex);
4116 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4118 struct inode *inode = file_inode(file);
4119 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4120 struct btrfs_ioctl_quota_ctl_args *sa;
4123 if (!capable(CAP_SYS_ADMIN))
4126 ret = mnt_want_write_file(file);
4130 sa = memdup_user(arg, sizeof(*sa));
4136 down_write(&fs_info->subvol_sem);
4139 case BTRFS_QUOTA_CTL_ENABLE:
4140 ret = btrfs_quota_enable(fs_info);
4142 case BTRFS_QUOTA_CTL_DISABLE:
4143 ret = btrfs_quota_disable(fs_info);
4151 up_write(&fs_info->subvol_sem);
4153 mnt_drop_write_file(file);
4157 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4159 struct inode *inode = file_inode(file);
4160 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4161 struct btrfs_root *root = BTRFS_I(inode)->root;
4162 struct btrfs_ioctl_qgroup_assign_args *sa;
4163 struct btrfs_trans_handle *trans;
4167 if (!capable(CAP_SYS_ADMIN))
4170 ret = mnt_want_write_file(file);
4174 sa = memdup_user(arg, sizeof(*sa));
4180 trans = btrfs_join_transaction(root);
4181 if (IS_ERR(trans)) {
4182 ret = PTR_ERR(trans);
4187 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
4189 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
4192 /* update qgroup status and info */
4193 err = btrfs_run_qgroups(trans);
4195 btrfs_handle_fs_error(fs_info, err,
4196 "failed to update qgroup status and info");
4197 err = btrfs_end_transaction(trans);
4204 mnt_drop_write_file(file);
4208 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4210 struct inode *inode = file_inode(file);
4211 struct btrfs_root *root = BTRFS_I(inode)->root;
4212 struct btrfs_ioctl_qgroup_create_args *sa;
4213 struct btrfs_trans_handle *trans;
4217 if (!capable(CAP_SYS_ADMIN))
4220 ret = mnt_want_write_file(file);
4224 sa = memdup_user(arg, sizeof(*sa));
4230 if (!sa->qgroupid) {
4235 trans = btrfs_join_transaction(root);
4236 if (IS_ERR(trans)) {
4237 ret = PTR_ERR(trans);
4242 ret = btrfs_create_qgroup(trans, sa->qgroupid);
4244 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
4247 err = btrfs_end_transaction(trans);
4254 mnt_drop_write_file(file);
4258 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4260 struct inode *inode = file_inode(file);
4261 struct btrfs_root *root = BTRFS_I(inode)->root;
4262 struct btrfs_ioctl_qgroup_limit_args *sa;
4263 struct btrfs_trans_handle *trans;
4268 if (!capable(CAP_SYS_ADMIN))
4271 ret = mnt_want_write_file(file);
4275 sa = memdup_user(arg, sizeof(*sa));
4281 trans = btrfs_join_transaction(root);
4282 if (IS_ERR(trans)) {
4283 ret = PTR_ERR(trans);
4287 qgroupid = sa->qgroupid;
4289 /* take the current subvol as qgroup */
4290 qgroupid = root->root_key.objectid;
4293 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
4295 err = btrfs_end_transaction(trans);
4302 mnt_drop_write_file(file);
4306 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4308 struct inode *inode = file_inode(file);
4309 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4310 struct btrfs_ioctl_quota_rescan_args *qsa;
4313 if (!capable(CAP_SYS_ADMIN))
4316 ret = mnt_want_write_file(file);
4320 qsa = memdup_user(arg, sizeof(*qsa));
4331 ret = btrfs_qgroup_rescan(fs_info);
4336 mnt_drop_write_file(file);
4340 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4343 struct btrfs_ioctl_quota_rescan_args *qsa;
4346 if (!capable(CAP_SYS_ADMIN))
4349 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
4353 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4355 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
4358 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4365 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4368 if (!capable(CAP_SYS_ADMIN))
4371 return btrfs_qgroup_wait_for_completion(fs_info, true);
4374 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4375 struct btrfs_ioctl_received_subvol_args *sa)
4377 struct inode *inode = file_inode(file);
4378 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4379 struct btrfs_root *root = BTRFS_I(inode)->root;
4380 struct btrfs_root_item *root_item = &root->root_item;
4381 struct btrfs_trans_handle *trans;
4382 struct timespec64 ct = current_time(inode);
4384 int received_uuid_changed;
4386 if (!inode_owner_or_capable(inode))
4389 ret = mnt_want_write_file(file);
4393 down_write(&fs_info->subvol_sem);
4395 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4400 if (btrfs_root_readonly(root)) {
4407 * 2 - uuid items (received uuid + subvol uuid)
4409 trans = btrfs_start_transaction(root, 3);
4410 if (IS_ERR(trans)) {
4411 ret = PTR_ERR(trans);
4416 sa->rtransid = trans->transid;
4417 sa->rtime.sec = ct.tv_sec;
4418 sa->rtime.nsec = ct.tv_nsec;
4420 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4422 if (received_uuid_changed &&
4423 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4424 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4425 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4426 root->root_key.objectid);
4427 if (ret && ret != -ENOENT) {
4428 btrfs_abort_transaction(trans, ret);
4429 btrfs_end_transaction(trans);
4433 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4434 btrfs_set_root_stransid(root_item, sa->stransid);
4435 btrfs_set_root_rtransid(root_item, sa->rtransid);
4436 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4437 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4438 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4439 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4441 ret = btrfs_update_root(trans, fs_info->tree_root,
4442 &root->root_key, &root->root_item);
4444 btrfs_end_transaction(trans);
4447 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4448 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4449 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4450 root->root_key.objectid);
4451 if (ret < 0 && ret != -EEXIST) {
4452 btrfs_abort_transaction(trans, ret);
4453 btrfs_end_transaction(trans);
4457 ret = btrfs_commit_transaction(trans);
4459 up_write(&fs_info->subvol_sem);
4460 mnt_drop_write_file(file);
4465 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4468 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4469 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4472 args32 = memdup_user(arg, sizeof(*args32));
4474 return PTR_ERR(args32);
4476 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4482 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4483 args64->stransid = args32->stransid;
4484 args64->rtransid = args32->rtransid;
4485 args64->stime.sec = args32->stime.sec;
4486 args64->stime.nsec = args32->stime.nsec;
4487 args64->rtime.sec = args32->rtime.sec;
4488 args64->rtime.nsec = args32->rtime.nsec;
4489 args64->flags = args32->flags;
4491 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4495 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4496 args32->stransid = args64->stransid;
4497 args32->rtransid = args64->rtransid;
4498 args32->stime.sec = args64->stime.sec;
4499 args32->stime.nsec = args64->stime.nsec;
4500 args32->rtime.sec = args64->rtime.sec;
4501 args32->rtime.nsec = args64->rtime.nsec;
4502 args32->flags = args64->flags;
4504 ret = copy_to_user(arg, args32, sizeof(*args32));
4515 static long btrfs_ioctl_set_received_subvol(struct file *file,
4518 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4521 sa = memdup_user(arg, sizeof(*sa));
4525 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4530 ret = copy_to_user(arg, sa, sizeof(*sa));
4539 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4544 char label[BTRFS_LABEL_SIZE];
4546 spin_lock(&fs_info->super_lock);
4547 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4548 spin_unlock(&fs_info->super_lock);
4550 len = strnlen(label, BTRFS_LABEL_SIZE);
4552 if (len == BTRFS_LABEL_SIZE) {
4554 "label is too long, return the first %zu bytes",
4558 ret = copy_to_user(arg, label, len);
4560 return ret ? -EFAULT : 0;
4563 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4565 struct inode *inode = file_inode(file);
4566 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4567 struct btrfs_root *root = BTRFS_I(inode)->root;
4568 struct btrfs_super_block *super_block = fs_info->super_copy;
4569 struct btrfs_trans_handle *trans;
4570 char label[BTRFS_LABEL_SIZE];
4573 if (!capable(CAP_SYS_ADMIN))
4576 if (copy_from_user(label, arg, sizeof(label)))
4579 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4581 "unable to set label with more than %d bytes",
4582 BTRFS_LABEL_SIZE - 1);
4586 ret = mnt_want_write_file(file);
4590 trans = btrfs_start_transaction(root, 0);
4591 if (IS_ERR(trans)) {
4592 ret = PTR_ERR(trans);
4596 spin_lock(&fs_info->super_lock);
4597 strcpy(super_block->label, label);
4598 spin_unlock(&fs_info->super_lock);
4599 ret = btrfs_commit_transaction(trans);
4602 mnt_drop_write_file(file);
4606 #define INIT_FEATURE_FLAGS(suffix) \
4607 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4608 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4609 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4611 int btrfs_ioctl_get_supported_features(void __user *arg)
4613 static const struct btrfs_ioctl_feature_flags features[3] = {
4614 INIT_FEATURE_FLAGS(SUPP),
4615 INIT_FEATURE_FLAGS(SAFE_SET),
4616 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4619 if (copy_to_user(arg, &features, sizeof(features)))
4625 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4628 struct btrfs_super_block *super_block = fs_info->super_copy;
4629 struct btrfs_ioctl_feature_flags features;
4631 features.compat_flags = btrfs_super_compat_flags(super_block);
4632 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4633 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4635 if (copy_to_user(arg, &features, sizeof(features)))
4641 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4642 enum btrfs_feature_set set,
4643 u64 change_mask, u64 flags, u64 supported_flags,
4644 u64 safe_set, u64 safe_clear)
4646 const char *type = btrfs_feature_set_name(set);
4648 u64 disallowed, unsupported;
4649 u64 set_mask = flags & change_mask;
4650 u64 clear_mask = ~flags & change_mask;
4652 unsupported = set_mask & ~supported_flags;
4654 names = btrfs_printable_features(set, unsupported);
4657 "this kernel does not support the %s feature bit%s",
4658 names, strchr(names, ',') ? "s" : "");
4662 "this kernel does not support %s bits 0x%llx",
4667 disallowed = set_mask & ~safe_set;
4669 names = btrfs_printable_features(set, disallowed);
4672 "can't set the %s feature bit%s while mounted",
4673 names, strchr(names, ',') ? "s" : "");
4677 "can't set %s bits 0x%llx while mounted",
4682 disallowed = clear_mask & ~safe_clear;
4684 names = btrfs_printable_features(set, disallowed);
4687 "can't clear the %s feature bit%s while mounted",
4688 names, strchr(names, ',') ? "s" : "");
4692 "can't clear %s bits 0x%llx while mounted",
4700 #define check_feature(fs_info, change_mask, flags, mask_base) \
4701 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
4702 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4703 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4704 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4706 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4708 struct inode *inode = file_inode(file);
4709 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4710 struct btrfs_root *root = BTRFS_I(inode)->root;
4711 struct btrfs_super_block *super_block = fs_info->super_copy;
4712 struct btrfs_ioctl_feature_flags flags[2];
4713 struct btrfs_trans_handle *trans;
4717 if (!capable(CAP_SYS_ADMIN))
4720 if (copy_from_user(flags, arg, sizeof(flags)))
4724 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4725 !flags[0].incompat_flags)
4728 ret = check_feature(fs_info, flags[0].compat_flags,
4729 flags[1].compat_flags, COMPAT);
4733 ret = check_feature(fs_info, flags[0].compat_ro_flags,
4734 flags[1].compat_ro_flags, COMPAT_RO);
4738 ret = check_feature(fs_info, flags[0].incompat_flags,
4739 flags[1].incompat_flags, INCOMPAT);
4743 ret = mnt_want_write_file(file);
4747 trans = btrfs_start_transaction(root, 0);
4748 if (IS_ERR(trans)) {
4749 ret = PTR_ERR(trans);
4750 goto out_drop_write;
4753 spin_lock(&fs_info->super_lock);
4754 newflags = btrfs_super_compat_flags(super_block);
4755 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4756 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4757 btrfs_set_super_compat_flags(super_block, newflags);
4759 newflags = btrfs_super_compat_ro_flags(super_block);
4760 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4761 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4762 btrfs_set_super_compat_ro_flags(super_block, newflags);
4764 newflags = btrfs_super_incompat_flags(super_block);
4765 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4766 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4767 btrfs_set_super_incompat_flags(super_block, newflags);
4768 spin_unlock(&fs_info->super_lock);
4770 ret = btrfs_commit_transaction(trans);
4772 mnt_drop_write_file(file);
4777 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
4779 struct btrfs_ioctl_send_args *arg;
4783 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4784 struct btrfs_ioctl_send_args_32 args32;
4786 ret = copy_from_user(&args32, argp, sizeof(args32));
4789 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4792 arg->send_fd = args32.send_fd;
4793 arg->clone_sources_count = args32.clone_sources_count;
4794 arg->clone_sources = compat_ptr(args32.clone_sources);
4795 arg->parent_root = args32.parent_root;
4796 arg->flags = args32.flags;
4797 memcpy(arg->reserved, args32.reserved,
4798 sizeof(args32.reserved));
4803 arg = memdup_user(argp, sizeof(*arg));
4805 return PTR_ERR(arg);
4807 ret = btrfs_ioctl_send(file, arg);
4812 long btrfs_ioctl(struct file *file, unsigned int
4813 cmd, unsigned long arg)
4815 struct inode *inode = file_inode(file);
4816 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4817 struct btrfs_root *root = BTRFS_I(inode)->root;
4818 void __user *argp = (void __user *)arg;
4821 case FS_IOC_GETFLAGS:
4822 return btrfs_ioctl_getflags(file, argp);
4823 case FS_IOC_SETFLAGS:
4824 return btrfs_ioctl_setflags(file, argp);
4825 case FS_IOC_GETVERSION:
4826 return btrfs_ioctl_getversion(file, argp);
4827 case FS_IOC_GETFSLABEL:
4828 return btrfs_ioctl_get_fslabel(fs_info, argp);
4829 case FS_IOC_SETFSLABEL:
4830 return btrfs_ioctl_set_fslabel(file, argp);
4832 return btrfs_ioctl_fitrim(fs_info, argp);
4833 case BTRFS_IOC_SNAP_CREATE:
4834 return btrfs_ioctl_snap_create(file, argp, 0);
4835 case BTRFS_IOC_SNAP_CREATE_V2:
4836 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4837 case BTRFS_IOC_SUBVOL_CREATE:
4838 return btrfs_ioctl_snap_create(file, argp, 1);
4839 case BTRFS_IOC_SUBVOL_CREATE_V2:
4840 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4841 case BTRFS_IOC_SNAP_DESTROY:
4842 return btrfs_ioctl_snap_destroy(file, argp, false);
4843 case BTRFS_IOC_SNAP_DESTROY_V2:
4844 return btrfs_ioctl_snap_destroy(file, argp, true);
4845 case BTRFS_IOC_SUBVOL_GETFLAGS:
4846 return btrfs_ioctl_subvol_getflags(file, argp);
4847 case BTRFS_IOC_SUBVOL_SETFLAGS:
4848 return btrfs_ioctl_subvol_setflags(file, argp);
4849 case BTRFS_IOC_DEFAULT_SUBVOL:
4850 return btrfs_ioctl_default_subvol(file, argp);
4851 case BTRFS_IOC_DEFRAG:
4852 return btrfs_ioctl_defrag(file, NULL);
4853 case BTRFS_IOC_DEFRAG_RANGE:
4854 return btrfs_ioctl_defrag(file, argp);
4855 case BTRFS_IOC_RESIZE:
4856 return btrfs_ioctl_resize(file, argp);
4857 case BTRFS_IOC_ADD_DEV:
4858 return btrfs_ioctl_add_dev(fs_info, argp);
4859 case BTRFS_IOC_RM_DEV:
4860 return btrfs_ioctl_rm_dev(file, argp);
4861 case BTRFS_IOC_RM_DEV_V2:
4862 return btrfs_ioctl_rm_dev_v2(file, argp);
4863 case BTRFS_IOC_FS_INFO:
4864 return btrfs_ioctl_fs_info(fs_info, argp);
4865 case BTRFS_IOC_DEV_INFO:
4866 return btrfs_ioctl_dev_info(fs_info, argp);
4867 case BTRFS_IOC_BALANCE:
4868 return btrfs_ioctl_balance(file, NULL);
4869 case BTRFS_IOC_TREE_SEARCH:
4870 return btrfs_ioctl_tree_search(file, argp);
4871 case BTRFS_IOC_TREE_SEARCH_V2:
4872 return btrfs_ioctl_tree_search_v2(file, argp);
4873 case BTRFS_IOC_INO_LOOKUP:
4874 return btrfs_ioctl_ino_lookup(file, argp);
4875 case BTRFS_IOC_INO_PATHS:
4876 return btrfs_ioctl_ino_to_path(root, argp);
4877 case BTRFS_IOC_LOGICAL_INO:
4878 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4879 case BTRFS_IOC_LOGICAL_INO_V2:
4880 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4881 case BTRFS_IOC_SPACE_INFO:
4882 return btrfs_ioctl_space_info(fs_info, argp);
4883 case BTRFS_IOC_SYNC: {
4886 ret = btrfs_start_delalloc_roots(fs_info, -1);
4889 ret = btrfs_sync_fs(inode->i_sb, 1);
4891 * The transaction thread may want to do more work,
4892 * namely it pokes the cleaner kthread that will start
4893 * processing uncleaned subvols.
4895 wake_up_process(fs_info->transaction_kthread);
4898 case BTRFS_IOC_START_SYNC:
4899 return btrfs_ioctl_start_sync(root, argp);
4900 case BTRFS_IOC_WAIT_SYNC:
4901 return btrfs_ioctl_wait_sync(fs_info, argp);
4902 case BTRFS_IOC_SCRUB:
4903 return btrfs_ioctl_scrub(file, argp);
4904 case BTRFS_IOC_SCRUB_CANCEL:
4905 return btrfs_ioctl_scrub_cancel(fs_info);
4906 case BTRFS_IOC_SCRUB_PROGRESS:
4907 return btrfs_ioctl_scrub_progress(fs_info, argp);
4908 case BTRFS_IOC_BALANCE_V2:
4909 return btrfs_ioctl_balance(file, argp);
4910 case BTRFS_IOC_BALANCE_CTL:
4911 return btrfs_ioctl_balance_ctl(fs_info, arg);
4912 case BTRFS_IOC_BALANCE_PROGRESS:
4913 return btrfs_ioctl_balance_progress(fs_info, argp);
4914 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4915 return btrfs_ioctl_set_received_subvol(file, argp);
4917 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4918 return btrfs_ioctl_set_received_subvol_32(file, argp);
4920 case BTRFS_IOC_SEND:
4921 return _btrfs_ioctl_send(file, argp, false);
4922 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4923 case BTRFS_IOC_SEND_32:
4924 return _btrfs_ioctl_send(file, argp, true);
4926 case BTRFS_IOC_GET_DEV_STATS:
4927 return btrfs_ioctl_get_dev_stats(fs_info, argp);
4928 case BTRFS_IOC_QUOTA_CTL:
4929 return btrfs_ioctl_quota_ctl(file, argp);
4930 case BTRFS_IOC_QGROUP_ASSIGN:
4931 return btrfs_ioctl_qgroup_assign(file, argp);
4932 case BTRFS_IOC_QGROUP_CREATE:
4933 return btrfs_ioctl_qgroup_create(file, argp);
4934 case BTRFS_IOC_QGROUP_LIMIT:
4935 return btrfs_ioctl_qgroup_limit(file, argp);
4936 case BTRFS_IOC_QUOTA_RESCAN:
4937 return btrfs_ioctl_quota_rescan(file, argp);
4938 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4939 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
4940 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4941 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
4942 case BTRFS_IOC_DEV_REPLACE:
4943 return btrfs_ioctl_dev_replace(fs_info, argp);
4944 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4945 return btrfs_ioctl_get_supported_features(argp);
4946 case BTRFS_IOC_GET_FEATURES:
4947 return btrfs_ioctl_get_features(fs_info, argp);
4948 case BTRFS_IOC_SET_FEATURES:
4949 return btrfs_ioctl_set_features(file, argp);
4950 case FS_IOC_FSGETXATTR:
4951 return btrfs_ioctl_fsgetxattr(file, argp);
4952 case FS_IOC_FSSETXATTR:
4953 return btrfs_ioctl_fssetxattr(file, argp);
4954 case BTRFS_IOC_GET_SUBVOL_INFO:
4955 return btrfs_ioctl_get_subvol_info(file, argp);
4956 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
4957 return btrfs_ioctl_get_subvol_rootref(file, argp);
4958 case BTRFS_IOC_INO_LOOKUP_USER:
4959 return btrfs_ioctl_ino_lookup_user(file, argp);
4965 #ifdef CONFIG_COMPAT
4966 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4969 * These all access 32-bit values anyway so no further
4970 * handling is necessary.
4973 case FS_IOC32_GETFLAGS:
4974 cmd = FS_IOC_GETFLAGS;
4976 case FS_IOC32_SETFLAGS:
4977 cmd = FS_IOC_SETFLAGS;
4979 case FS_IOC32_GETVERSION:
4980 cmd = FS_IOC_GETVERSION;
4984 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));