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>
29 #include <linux/fileattr.h>
33 #include "transaction.h"
34 #include "btrfs_inode.h"
35 #include "print-tree.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,
158 * Check if @flags are a supported and valid set of FS_*_FL flags and that
159 * the old and new flags are not conflicting
161 static int check_fsflags(unsigned int old_flags, unsigned int flags)
163 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
164 FS_NOATIME_FL | FS_NODUMP_FL | \
165 FS_SYNC_FL | FS_DIRSYNC_FL | \
166 FS_NOCOMP_FL | FS_COMPR_FL |
170 /* COMPR and NOCOMP on new/old are valid */
171 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
174 if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
177 /* NOCOW and compression options are mutually exclusive */
178 if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
180 if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
186 static int check_fsflags_compatible(struct btrfs_fs_info *fs_info,
189 if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL))
196 * Set flags/xflags from the internal inode flags. The remaining items of
197 * fsxattr are zeroed.
199 int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
201 struct btrfs_inode *binode = BTRFS_I(d_inode(dentry));
203 fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode->flags));
207 int btrfs_fileattr_set(struct user_namespace *mnt_userns,
208 struct dentry *dentry, struct fileattr *fa)
210 struct inode *inode = d_inode(dentry);
211 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
212 struct btrfs_inode *binode = BTRFS_I(inode);
213 struct btrfs_root *root = binode->root;
214 struct btrfs_trans_handle *trans;
215 unsigned int fsflags, old_fsflags;
217 const char *comp = NULL;
220 if (btrfs_root_readonly(root))
223 if (fileattr_has_fsx(fa))
226 fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags);
227 old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags);
228 ret = check_fsflags(old_fsflags, fsflags);
232 ret = check_fsflags_compatible(fs_info, fsflags);
236 binode_flags = binode->flags;
237 if (fsflags & FS_SYNC_FL)
238 binode_flags |= BTRFS_INODE_SYNC;
240 binode_flags &= ~BTRFS_INODE_SYNC;
241 if (fsflags & FS_IMMUTABLE_FL)
242 binode_flags |= BTRFS_INODE_IMMUTABLE;
244 binode_flags &= ~BTRFS_INODE_IMMUTABLE;
245 if (fsflags & FS_APPEND_FL)
246 binode_flags |= BTRFS_INODE_APPEND;
248 binode_flags &= ~BTRFS_INODE_APPEND;
249 if (fsflags & FS_NODUMP_FL)
250 binode_flags |= BTRFS_INODE_NODUMP;
252 binode_flags &= ~BTRFS_INODE_NODUMP;
253 if (fsflags & FS_NOATIME_FL)
254 binode_flags |= BTRFS_INODE_NOATIME;
256 binode_flags &= ~BTRFS_INODE_NOATIME;
258 /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */
259 if (!fa->flags_valid) {
260 /* 1 item for the inode */
261 trans = btrfs_start_transaction(root, 1);
263 return PTR_ERR(trans);
267 if (fsflags & FS_DIRSYNC_FL)
268 binode_flags |= BTRFS_INODE_DIRSYNC;
270 binode_flags &= ~BTRFS_INODE_DIRSYNC;
271 if (fsflags & FS_NOCOW_FL) {
272 if (S_ISREG(inode->i_mode)) {
274 * It's safe to turn csums off here, no extents exist.
275 * Otherwise we want the flag to reflect the real COW
276 * status of the file and will not set it.
278 if (inode->i_size == 0)
279 binode_flags |= BTRFS_INODE_NODATACOW |
280 BTRFS_INODE_NODATASUM;
282 binode_flags |= BTRFS_INODE_NODATACOW;
286 * Revert back under same assumptions as above
288 if (S_ISREG(inode->i_mode)) {
289 if (inode->i_size == 0)
290 binode_flags &= ~(BTRFS_INODE_NODATACOW |
291 BTRFS_INODE_NODATASUM);
293 binode_flags &= ~BTRFS_INODE_NODATACOW;
298 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
299 * flag may be changed automatically if compression code won't make
302 if (fsflags & FS_NOCOMP_FL) {
303 binode_flags &= ~BTRFS_INODE_COMPRESS;
304 binode_flags |= BTRFS_INODE_NOCOMPRESS;
305 } else if (fsflags & FS_COMPR_FL) {
307 if (IS_SWAPFILE(inode))
310 binode_flags |= BTRFS_INODE_COMPRESS;
311 binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
313 comp = btrfs_compress_type2str(fs_info->compress_type);
314 if (!comp || comp[0] == 0)
315 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
317 binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
324 trans = btrfs_start_transaction(root, 3);
326 return PTR_ERR(trans);
329 ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
332 btrfs_abort_transaction(trans, ret);
336 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
338 if (ret && ret != -ENODATA) {
339 btrfs_abort_transaction(trans, ret);
345 binode->flags = binode_flags;
346 btrfs_sync_inode_flags_to_i_flags(inode);
347 inode_inc_iversion(inode);
348 inode->i_ctime = current_time(inode);
349 ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
352 btrfs_end_transaction(trans);
357 * Start exclusive operation @type, return true on success
359 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
360 enum btrfs_exclusive_operation type)
364 spin_lock(&fs_info->super_lock);
365 if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) {
366 fs_info->exclusive_operation = type;
369 spin_unlock(&fs_info->super_lock);
375 * Conditionally allow to enter the exclusive operation in case it's compatible
376 * with the running one. This must be paired with btrfs_exclop_start_unlock and
377 * btrfs_exclop_finish.
380 * - the same type is already running
381 * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller
382 * must check the condition first that would allow none -> @type
384 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
385 enum btrfs_exclusive_operation type)
387 spin_lock(&fs_info->super_lock);
388 if (fs_info->exclusive_operation == type)
391 spin_unlock(&fs_info->super_lock);
395 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info)
397 spin_unlock(&fs_info->super_lock);
400 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info)
402 spin_lock(&fs_info->super_lock);
403 WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE);
404 spin_unlock(&fs_info->super_lock);
405 sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation");
408 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
410 struct inode *inode = file_inode(file);
412 return put_user(inode->i_generation, arg);
415 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
418 struct btrfs_device *device;
419 struct request_queue *q;
420 struct fstrim_range range;
421 u64 minlen = ULLONG_MAX;
425 if (!capable(CAP_SYS_ADMIN))
429 * btrfs_trim_block_group() depends on space cache, which is not
430 * available in zoned filesystem. So, disallow fitrim on a zoned
431 * filesystem for now.
433 if (btrfs_is_zoned(fs_info))
437 * If the fs is mounted with nologreplay, which requires it to be
438 * mounted in RO mode as well, we can not allow discard on free space
439 * inside block groups, because log trees refer to extents that are not
440 * pinned in a block group's free space cache (pinning the extents is
441 * precisely the first phase of replaying a log tree).
443 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
447 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
451 q = bdev_get_queue(device->bdev);
452 if (blk_queue_discard(q)) {
454 minlen = min_t(u64, q->limits.discard_granularity,
462 if (copy_from_user(&range, arg, sizeof(range)))
466 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
467 * block group is in the logical address space, which can be any
468 * sectorsize aligned bytenr in the range [0, U64_MAX].
470 if (range.len < fs_info->sb->s_blocksize)
473 range.minlen = max(range.minlen, minlen);
474 ret = btrfs_trim_fs(fs_info, &range);
478 if (copy_to_user(arg, &range, sizeof(range)))
484 int __pure btrfs_is_empty_uuid(u8 *uuid)
488 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
495 static noinline int create_subvol(struct inode *dir,
496 struct dentry *dentry,
497 const char *name, int namelen,
498 struct btrfs_qgroup_inherit *inherit)
500 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
501 struct btrfs_trans_handle *trans;
502 struct btrfs_key key;
503 struct btrfs_root_item *root_item;
504 struct btrfs_inode_item *inode_item;
505 struct extent_buffer *leaf;
506 struct btrfs_root *root = BTRFS_I(dir)->root;
507 struct btrfs_root *new_root;
508 struct btrfs_block_rsv block_rsv;
509 struct timespec64 cur_time = current_time(dir);
517 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
521 ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
525 ret = get_anon_bdev(&anon_dev);
530 * Don't create subvolume whose level is not zero. Or qgroup will be
531 * screwed up since it assumes subvolume qgroup's level to be 0.
533 if (btrfs_qgroup_level(objectid)) {
538 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
540 * The same as the snapshot creation, please see the comment
541 * of create_snapshot().
543 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false);
547 trans = btrfs_start_transaction(root, 0);
549 ret = PTR_ERR(trans);
550 btrfs_subvolume_release_metadata(root, &block_rsv);
553 trans->block_rsv = &block_rsv;
554 trans->bytes_reserved = block_rsv.size;
556 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
560 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
561 BTRFS_NESTING_NORMAL);
567 btrfs_mark_buffer_dirty(leaf);
569 inode_item = &root_item->inode;
570 btrfs_set_stack_inode_generation(inode_item, 1);
571 btrfs_set_stack_inode_size(inode_item, 3);
572 btrfs_set_stack_inode_nlink(inode_item, 1);
573 btrfs_set_stack_inode_nbytes(inode_item,
575 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
577 btrfs_set_root_flags(root_item, 0);
578 btrfs_set_root_limit(root_item, 0);
579 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
581 btrfs_set_root_bytenr(root_item, leaf->start);
582 btrfs_set_root_generation(root_item, trans->transid);
583 btrfs_set_root_level(root_item, 0);
584 btrfs_set_root_refs(root_item, 1);
585 btrfs_set_root_used(root_item, leaf->len);
586 btrfs_set_root_last_snapshot(root_item, 0);
588 btrfs_set_root_generation_v2(root_item,
589 btrfs_root_generation(root_item));
590 generate_random_guid(root_item->uuid);
591 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
592 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
593 root_item->ctime = root_item->otime;
594 btrfs_set_root_ctransid(root_item, trans->transid);
595 btrfs_set_root_otransid(root_item, trans->transid);
597 btrfs_tree_unlock(leaf);
599 btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
601 key.objectid = objectid;
603 key.type = BTRFS_ROOT_ITEM_KEY;
604 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
608 * Since we don't abort the transaction in this case, free the
609 * tree block so that we don't leak space and leave the
610 * filesystem in an inconsistent state (an extent item in the
611 * extent tree without backreferences). Also no need to have
612 * the tree block locked since it is not in any tree at this
613 * point, so no other task can find it and use it.
615 btrfs_free_tree_block(trans, root, leaf, 0, 1);
616 free_extent_buffer(leaf);
620 free_extent_buffer(leaf);
623 key.offset = (u64)-1;
624 new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
625 if (IS_ERR(new_root)) {
626 free_anon_bdev(anon_dev);
627 ret = PTR_ERR(new_root);
628 btrfs_abort_transaction(trans, ret);
631 /* Freeing will be done in btrfs_put_root() of new_root */
634 ret = btrfs_record_root_in_trans(trans, new_root);
636 btrfs_put_root(new_root);
637 btrfs_abort_transaction(trans, ret);
641 ret = btrfs_create_subvol_root(trans, new_root, root);
642 btrfs_put_root(new_root);
644 /* We potentially lose an unused inode item here */
645 btrfs_abort_transaction(trans, ret);
650 * insert the directory item
652 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
654 btrfs_abort_transaction(trans, ret);
658 ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
659 BTRFS_FT_DIR, index);
661 btrfs_abort_transaction(trans, ret);
665 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
666 ret = btrfs_update_inode(trans, root, BTRFS_I(dir));
668 btrfs_abort_transaction(trans, ret);
672 ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
673 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
675 btrfs_abort_transaction(trans, ret);
679 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
680 BTRFS_UUID_KEY_SUBVOL, objectid);
682 btrfs_abort_transaction(trans, ret);
686 trans->block_rsv = NULL;
687 trans->bytes_reserved = 0;
688 btrfs_subvolume_release_metadata(root, &block_rsv);
690 err = btrfs_commit_transaction(trans);
695 inode = btrfs_lookup_dentry(dir, dentry);
697 return PTR_ERR(inode);
698 d_instantiate(dentry, inode);
704 free_anon_bdev(anon_dev);
709 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
710 struct dentry *dentry, bool readonly,
711 struct btrfs_qgroup_inherit *inherit)
713 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
715 struct btrfs_pending_snapshot *pending_snapshot;
716 struct btrfs_trans_handle *trans;
719 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
722 if (atomic_read(&root->nr_swapfiles)) {
724 "cannot snapshot subvolume with active swapfile");
728 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
729 if (!pending_snapshot)
732 ret = get_anon_bdev(&pending_snapshot->anon_dev);
735 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
737 pending_snapshot->path = btrfs_alloc_path();
738 if (!pending_snapshot->root_item || !pending_snapshot->path) {
743 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
744 BTRFS_BLOCK_RSV_TEMP);
746 * 1 - parent dir inode
749 * 2 - root ref/backref
750 * 1 - root of snapshot
753 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
754 &pending_snapshot->block_rsv, 8,
759 pending_snapshot->dentry = dentry;
760 pending_snapshot->root = root;
761 pending_snapshot->readonly = readonly;
762 pending_snapshot->dir = dir;
763 pending_snapshot->inherit = inherit;
765 trans = btrfs_start_transaction(root, 0);
767 ret = PTR_ERR(trans);
771 spin_lock(&fs_info->trans_lock);
772 list_add(&pending_snapshot->list,
773 &trans->transaction->pending_snapshots);
774 spin_unlock(&fs_info->trans_lock);
776 ret = btrfs_commit_transaction(trans);
780 ret = pending_snapshot->error;
784 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
788 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
790 ret = PTR_ERR(inode);
794 d_instantiate(dentry, inode);
796 pending_snapshot->anon_dev = 0;
798 /* Prevent double freeing of anon_dev */
799 if (ret && pending_snapshot->snap)
800 pending_snapshot->snap->anon_dev = 0;
801 btrfs_put_root(pending_snapshot->snap);
802 btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
804 if (pending_snapshot->anon_dev)
805 free_anon_bdev(pending_snapshot->anon_dev);
806 kfree(pending_snapshot->root_item);
807 btrfs_free_path(pending_snapshot->path);
808 kfree(pending_snapshot);
813 /* copy of may_delete in fs/namei.c()
814 * Check whether we can remove a link victim from directory dir, check
815 * whether the type of victim is right.
816 * 1. We can't do it if dir is read-only (done in permission())
817 * 2. We should have write and exec permissions on dir
818 * 3. We can't remove anything from append-only dir
819 * 4. We can't do anything with immutable dir (done in permission())
820 * 5. If the sticky bit on dir is set we should either
821 * a. be owner of dir, or
822 * b. be owner of victim, or
823 * c. have CAP_FOWNER capability
824 * 6. If the victim is append-only or immutable we can't do anything with
825 * links pointing to it.
826 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
827 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
828 * 9. We can't remove a root or mountpoint.
829 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
830 * nfs_async_unlink().
833 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
837 if (d_really_is_negative(victim))
840 BUG_ON(d_inode(victim->d_parent) != dir);
841 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
843 error = inode_permission(&init_user_ns, dir, MAY_WRITE | MAY_EXEC);
848 if (check_sticky(&init_user_ns, dir, d_inode(victim)) ||
849 IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
850 IS_SWAPFILE(d_inode(victim)))
853 if (!d_is_dir(victim))
857 } else if (d_is_dir(victim))
861 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
866 /* copy of may_create in fs/namei.c() */
867 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
869 if (d_really_is_positive(child))
873 return inode_permission(&init_user_ns, dir, MAY_WRITE | MAY_EXEC);
877 * Create a new subvolume below @parent. This is largely modeled after
878 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
879 * inside this filesystem so it's quite a bit simpler.
881 static noinline int btrfs_mksubvol(const struct path *parent,
882 const char *name, int namelen,
883 struct btrfs_root *snap_src,
885 struct btrfs_qgroup_inherit *inherit)
887 struct inode *dir = d_inode(parent->dentry);
888 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
889 struct dentry *dentry;
892 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
896 dentry = lookup_one_len(name, parent->dentry, namelen);
897 error = PTR_ERR(dentry);
901 error = btrfs_may_create(dir, dentry);
906 * even if this name doesn't exist, we may get hash collisions.
907 * check for them now when we can safely fail
909 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
915 down_read(&fs_info->subvol_sem);
917 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
921 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
923 error = create_subvol(dir, dentry, name, namelen, inherit);
926 fsnotify_mkdir(dir, dentry);
928 up_read(&fs_info->subvol_sem);
932 btrfs_inode_unlock(dir, 0);
936 static noinline int btrfs_mksnapshot(const struct path *parent,
937 const char *name, int namelen,
938 struct btrfs_root *root,
940 struct btrfs_qgroup_inherit *inherit)
943 bool snapshot_force_cow = false;
946 * Force new buffered writes to reserve space even when NOCOW is
947 * possible. This is to avoid later writeback (running dealloc) to
948 * fallback to COW mode and unexpectedly fail with ENOSPC.
950 btrfs_drew_read_lock(&root->snapshot_lock);
952 ret = btrfs_start_delalloc_snapshot(root, false);
957 * All previous writes have started writeback in NOCOW mode, so now
958 * we force future writes to fallback to COW mode during snapshot
961 atomic_inc(&root->snapshot_force_cow);
962 snapshot_force_cow = true;
964 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
966 ret = btrfs_mksubvol(parent, name, namelen,
967 root, readonly, inherit);
969 if (snapshot_force_cow)
970 atomic_dec(&root->snapshot_force_cow);
971 btrfs_drew_read_unlock(&root->snapshot_lock);
976 * When we're defragging a range, we don't want to kick it off again
977 * if it is really just waiting for delalloc to send it down.
978 * If we find a nice big extent or delalloc range for the bytes in the
979 * file you want to defrag, we return 0 to let you know to skip this
982 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
984 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
985 struct extent_map *em = NULL;
986 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
989 read_lock(&em_tree->lock);
990 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
991 read_unlock(&em_tree->lock);
994 end = extent_map_end(em);
996 if (end - offset > thresh)
999 /* if we already have a nice delalloc here, just stop */
1001 end = count_range_bits(io_tree, &offset, offset + thresh,
1002 thresh, EXTENT_DELALLOC, 1);
1009 * helper function to walk through a file and find extents
1010 * newer than a specific transid, and smaller than thresh.
1012 * This is used by the defragging code to find new and small
1015 static int find_new_extents(struct btrfs_root *root,
1016 struct inode *inode, u64 newer_than,
1017 u64 *off, u32 thresh)
1019 struct btrfs_path *path;
1020 struct btrfs_key min_key;
1021 struct extent_buffer *leaf;
1022 struct btrfs_file_extent_item *extent;
1025 u64 ino = btrfs_ino(BTRFS_I(inode));
1027 path = btrfs_alloc_path();
1031 min_key.objectid = ino;
1032 min_key.type = BTRFS_EXTENT_DATA_KEY;
1033 min_key.offset = *off;
1036 ret = btrfs_search_forward(root, &min_key, path, newer_than);
1040 if (min_key.objectid != ino)
1042 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1045 leaf = path->nodes[0];
1046 extent = btrfs_item_ptr(leaf, path->slots[0],
1047 struct btrfs_file_extent_item);
1049 type = btrfs_file_extent_type(leaf, extent);
1050 if (type == BTRFS_FILE_EXTENT_REG &&
1051 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1052 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1053 *off = min_key.offset;
1054 btrfs_free_path(path);
1059 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1060 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1064 if (min_key.offset == (u64)-1)
1068 btrfs_release_path(path);
1071 btrfs_free_path(path);
1075 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1077 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1078 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1079 struct extent_map *em;
1080 u64 len = PAGE_SIZE;
1083 * hopefully we have this extent in the tree already, try without
1084 * the full extent lock
1086 read_lock(&em_tree->lock);
1087 em = lookup_extent_mapping(em_tree, start, len);
1088 read_unlock(&em_tree->lock);
1091 struct extent_state *cached = NULL;
1092 u64 end = start + len - 1;
1094 /* get the big lock and read metadata off disk */
1095 lock_extent_bits(io_tree, start, end, &cached);
1096 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len);
1097 unlock_extent_cached(io_tree, start, end, &cached);
1106 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1108 struct extent_map *next;
1111 /* this is the last extent */
1112 if (em->start + em->len >= i_size_read(inode))
1115 next = defrag_lookup_extent(inode, em->start + em->len);
1116 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1118 else if ((em->block_start + em->block_len == next->block_start) &&
1119 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1122 free_extent_map(next);
1126 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1127 u64 *last_len, u64 *skip, u64 *defrag_end,
1130 struct extent_map *em;
1132 bool next_mergeable = true;
1133 bool prev_mergeable = true;
1136 * make sure that once we start defragging an extent, we keep on
1139 if (start < *defrag_end)
1144 em = defrag_lookup_extent(inode, start);
1148 /* this will cover holes, and inline extents */
1149 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1155 prev_mergeable = false;
1157 next_mergeable = defrag_check_next_extent(inode, em);
1159 * we hit a real extent, if it is big or the next extent is not a
1160 * real extent, don't bother defragging it
1162 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1163 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1167 * last_len ends up being a counter of how many bytes we've defragged.
1168 * every time we choose not to defrag an extent, we reset *last_len
1169 * so that the next tiny extent will force a defrag.
1171 * The end result of this is that tiny extents before a single big
1172 * extent will force at least part of that big extent to be defragged.
1175 *defrag_end = extent_map_end(em);
1178 *skip = extent_map_end(em);
1182 free_extent_map(em);
1187 * it doesn't do much good to defrag one or two pages
1188 * at a time. This pulls in a nice chunk of pages
1189 * to COW and defrag.
1191 * It also makes sure the delalloc code has enough
1192 * dirty data to avoid making new small extents as part
1195 * It's a good idea to start RA on this range
1196 * before calling this.
1198 static int cluster_pages_for_defrag(struct inode *inode,
1199 struct page **pages,
1200 unsigned long start_index,
1201 unsigned long num_pages)
1203 unsigned long file_end;
1204 u64 isize = i_size_read(inode);
1208 u64 start = (u64)start_index << PAGE_SHIFT;
1213 struct btrfs_ordered_extent *ordered;
1214 struct extent_state *cached_state = NULL;
1215 struct extent_io_tree *tree;
1216 struct extent_changeset *data_reserved = NULL;
1217 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1219 file_end = (isize - 1) >> PAGE_SHIFT;
1220 if (!isize || start_index > file_end)
1223 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1225 ret = btrfs_delalloc_reserve_space(BTRFS_I(inode), &data_reserved,
1226 start, page_cnt << PAGE_SHIFT);
1230 tree = &BTRFS_I(inode)->io_tree;
1232 /* step one, lock all the pages */
1233 for (i = 0; i < page_cnt; i++) {
1236 page = find_or_create_page(inode->i_mapping,
1237 start_index + i, mask);
1241 ret = set_page_extent_mapped(page);
1248 page_start = page_offset(page);
1249 page_end = page_start + PAGE_SIZE - 1;
1251 lock_extent_bits(tree, page_start, page_end,
1253 ordered = btrfs_lookup_ordered_extent(BTRFS_I(inode),
1255 unlock_extent_cached(tree, page_start, page_end,
1261 btrfs_start_ordered_extent(ordered, 1);
1262 btrfs_put_ordered_extent(ordered);
1265 * we unlocked the page above, so we need check if
1266 * it was released or not.
1268 if (page->mapping != inode->i_mapping) {
1275 if (!PageUptodate(page)) {
1276 btrfs_readpage(NULL, page);
1278 if (!PageUptodate(page)) {
1286 if (page->mapping != inode->i_mapping) {
1298 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1302 * so now we have a nice long stream of locked
1303 * and up to date pages, lets wait on them
1305 for (i = 0; i < i_done; i++)
1306 wait_on_page_writeback(pages[i]);
1308 page_start = page_offset(pages[0]);
1309 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1311 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1312 page_start, page_end - 1, &cached_state);
1315 * When defragmenting we skip ranges that have holes or inline extents,
1316 * (check should_defrag_range()), to avoid unnecessary IO and wasting
1317 * space. At btrfs_defrag_file(), we check if a range should be defragged
1318 * before locking the inode and then, if it should, we trigger a sync
1319 * page cache readahead - we lock the inode only after that to avoid
1320 * blocking for too long other tasks that possibly want to operate on
1321 * other file ranges. But before we were able to get the inode lock,
1322 * some other task may have punched a hole in the range, or we may have
1323 * now an inline extent, in which case we should not defrag. So check
1324 * for that here, where we have the inode and the range locked, and bail
1325 * out if that happened.
1327 search_start = page_start;
1328 while (search_start < page_end) {
1329 struct extent_map *em;
1331 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, search_start,
1332 page_end - search_start);
1335 goto out_unlock_range;
1337 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1338 free_extent_map(em);
1339 /* Ok, 0 means we did not defrag anything */
1341 goto out_unlock_range;
1343 search_start = extent_map_end(em);
1344 free_extent_map(em);
1347 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1348 page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
1349 EXTENT_DEFRAG, 0, 0, &cached_state);
1351 if (i_done != page_cnt) {
1352 spin_lock(&BTRFS_I(inode)->lock);
1353 btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
1354 spin_unlock(&BTRFS_I(inode)->lock);
1355 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1356 start, (page_cnt - i_done) << PAGE_SHIFT, true);
1360 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1363 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1364 page_start, page_end - 1, &cached_state);
1366 for (i = 0; i < i_done; i++) {
1367 clear_page_dirty_for_io(pages[i]);
1368 ClearPageChecked(pages[i]);
1369 set_page_dirty(pages[i]);
1370 unlock_page(pages[i]);
1373 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1374 extent_changeset_free(data_reserved);
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 unlock_page(pages[i]);
1385 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1386 start, page_cnt << PAGE_SHIFT, true);
1387 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1388 extent_changeset_free(data_reserved);
1393 int btrfs_defrag_file(struct inode *inode, struct file *file,
1394 struct btrfs_ioctl_defrag_range_args *range,
1395 u64 newer_than, unsigned long max_to_defrag)
1397 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1398 struct btrfs_root *root = BTRFS_I(inode)->root;
1399 struct file_ra_state *ra = NULL;
1400 unsigned long last_index;
1401 u64 isize = i_size_read(inode);
1405 u64 newer_off = range->start;
1407 unsigned long ra_index = 0;
1409 int defrag_count = 0;
1410 int compress_type = BTRFS_COMPRESS_ZLIB;
1411 u32 extent_thresh = range->extent_thresh;
1412 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1413 unsigned long cluster = max_cluster;
1414 u64 new_align = ~((u64)SZ_128K - 1);
1415 struct page **pages = NULL;
1416 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1421 if (range->start >= isize)
1425 if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES)
1427 if (range->compress_type)
1428 compress_type = range->compress_type;
1431 if (extent_thresh == 0)
1432 extent_thresh = SZ_256K;
1435 * If we were not given a file, allocate a readahead context. As
1436 * readahead is just an optimization, defrag will work without it so
1437 * we don't error out.
1440 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1442 file_ra_state_init(ra, inode->i_mapping);
1447 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1453 /* find the last page to defrag */
1454 if (range->start + range->len > range->start) {
1455 last_index = min_t(u64, isize - 1,
1456 range->start + range->len - 1) >> PAGE_SHIFT;
1458 last_index = (isize - 1) >> PAGE_SHIFT;
1462 ret = find_new_extents(root, inode, newer_than,
1463 &newer_off, SZ_64K);
1465 range->start = newer_off;
1467 * we always align our defrag to help keep
1468 * the extents in the file evenly spaced
1470 i = (newer_off & new_align) >> PAGE_SHIFT;
1474 i = range->start >> PAGE_SHIFT;
1477 max_to_defrag = last_index - i + 1;
1480 * make writeback starts from i, so the defrag range can be
1481 * written sequentially.
1483 if (i < inode->i_mapping->writeback_index)
1484 inode->i_mapping->writeback_index = i;
1486 while (i <= last_index && defrag_count < max_to_defrag &&
1487 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1489 * make sure we stop running if someone unmounts
1492 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1495 if (btrfs_defrag_cancelled(fs_info)) {
1496 btrfs_debug(fs_info, "defrag_file cancelled");
1501 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1502 extent_thresh, &last_len, &skip,
1503 &defrag_end, do_compress)){
1506 * the should_defrag function tells us how much to skip
1507 * bump our counter by the suggested amount
1509 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1510 i = max(i + 1, next);
1515 cluster = (PAGE_ALIGN(defrag_end) >>
1517 cluster = min(cluster, max_cluster);
1519 cluster = max_cluster;
1522 if (i + cluster > ra_index) {
1523 ra_index = max(i, ra_index);
1525 page_cache_sync_readahead(inode->i_mapping, ra,
1526 file, ra_index, cluster);
1527 ra_index += cluster;
1530 btrfs_inode_lock(inode, 0);
1531 if (IS_SWAPFILE(inode)) {
1535 BTRFS_I(inode)->defrag_compress = compress_type;
1536 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1539 btrfs_inode_unlock(inode, 0);
1543 defrag_count += ret;
1544 balance_dirty_pages_ratelimited(inode->i_mapping);
1545 btrfs_inode_unlock(inode, 0);
1548 if (newer_off == (u64)-1)
1554 newer_off = max(newer_off + 1,
1555 (u64)i << PAGE_SHIFT);
1557 ret = find_new_extents(root, inode, newer_than,
1558 &newer_off, SZ_64K);
1560 range->start = newer_off;
1561 i = (newer_off & new_align) >> PAGE_SHIFT;
1568 last_len += ret << PAGE_SHIFT;
1578 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1579 filemap_flush(inode->i_mapping);
1580 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1581 &BTRFS_I(inode)->runtime_flags))
1582 filemap_flush(inode->i_mapping);
1585 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1586 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1587 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1588 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1593 btrfs_inode_lock(inode, 0);
1594 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1595 btrfs_inode_unlock(inode, 0);
1604 * Try to start exclusive operation @type or cancel it if it's running.
1607 * 0 - normal mode, newly claimed op started
1608 * >0 - normal mode, something else is running,
1609 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1610 * ECANCELED - cancel mode, successful cancel
1611 * ENOTCONN - cancel mode, operation not running anymore
1613 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1614 enum btrfs_exclusive_operation type, bool cancel)
1617 /* Start normal op */
1618 if (!btrfs_exclop_start(fs_info, type))
1619 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1620 /* Exclusive operation is now claimed */
1624 /* Cancel running op */
1625 if (btrfs_exclop_start_try_lock(fs_info, type)) {
1627 * This blocks any exclop finish from setting it to NONE, so we
1628 * request cancellation. Either it runs and we will wait for it,
1629 * or it has finished and no waiting will happen.
1631 atomic_inc(&fs_info->reloc_cancel_req);
1632 btrfs_exclop_start_unlock(fs_info);
1634 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1635 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1636 TASK_INTERRUPTIBLE);
1641 /* Something else is running or none */
1645 static noinline int btrfs_ioctl_resize(struct file *file,
1648 struct inode *inode = file_inode(file);
1649 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1653 struct btrfs_root *root = BTRFS_I(inode)->root;
1654 struct btrfs_ioctl_vol_args *vol_args;
1655 struct btrfs_trans_handle *trans;
1656 struct btrfs_device *device = NULL;
1659 char *devstr = NULL;
1664 if (!capable(CAP_SYS_ADMIN))
1667 ret = mnt_want_write_file(file);
1672 * Read the arguments before checking exclusivity to be able to
1673 * distinguish regular resize and cancel
1675 vol_args = memdup_user(arg, sizeof(*vol_args));
1676 if (IS_ERR(vol_args)) {
1677 ret = PTR_ERR(vol_args);
1680 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1681 sizestr = vol_args->name;
1682 cancel = (strcmp("cancel", sizestr) == 0);
1683 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1686 /* Exclusive operation is now claimed */
1688 devstr = strchr(sizestr, ':');
1690 sizestr = devstr + 1;
1692 devstr = vol_args->name;
1693 ret = kstrtoull(devstr, 10, &devid);
1700 btrfs_info(fs_info, "resizing devid %llu", devid);
1703 device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL);
1705 btrfs_info(fs_info, "resizer unable to find device %llu",
1711 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1713 "resizer unable to apply on readonly device %llu",
1719 if (!strcmp(sizestr, "max"))
1720 new_size = device->bdev->bd_inode->i_size;
1722 if (sizestr[0] == '-') {
1725 } else if (sizestr[0] == '+') {
1729 new_size = memparse(sizestr, &retptr);
1730 if (*retptr != '\0' || new_size == 0) {
1736 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1741 old_size = btrfs_device_get_total_bytes(device);
1744 if (new_size > old_size) {
1748 new_size = old_size - new_size;
1749 } else if (mod > 0) {
1750 if (new_size > ULLONG_MAX - old_size) {
1754 new_size = old_size + new_size;
1757 if (new_size < SZ_256M) {
1761 if (new_size > device->bdev->bd_inode->i_size) {
1766 new_size = round_down(new_size, fs_info->sectorsize);
1768 if (new_size > old_size) {
1769 trans = btrfs_start_transaction(root, 0);
1770 if (IS_ERR(trans)) {
1771 ret = PTR_ERR(trans);
1774 ret = btrfs_grow_device(trans, device, new_size);
1775 btrfs_commit_transaction(trans);
1776 } else if (new_size < old_size) {
1777 ret = btrfs_shrink_device(device, new_size);
1778 } /* equal, nothing need to do */
1780 if (ret == 0 && new_size != old_size)
1781 btrfs_info_in_rcu(fs_info,
1782 "resize device %s (devid %llu) from %llu to %llu",
1783 rcu_str_deref(device->name), device->devid,
1784 old_size, new_size);
1786 btrfs_exclop_finish(fs_info);
1790 mnt_drop_write_file(file);
1794 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1795 const char *name, unsigned long fd, int subvol,
1797 struct btrfs_qgroup_inherit *inherit)
1802 if (!S_ISDIR(file_inode(file)->i_mode))
1805 ret = mnt_want_write_file(file);
1809 namelen = strlen(name);
1810 if (strchr(name, '/')) {
1812 goto out_drop_write;
1815 if (name[0] == '.' &&
1816 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1818 goto out_drop_write;
1822 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1823 NULL, readonly, inherit);
1825 struct fd src = fdget(fd);
1826 struct inode *src_inode;
1829 goto out_drop_write;
1832 src_inode = file_inode(src.file);
1833 if (src_inode->i_sb != file_inode(file)->i_sb) {
1834 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1835 "Snapshot src from another FS");
1837 } else if (!inode_owner_or_capable(&init_user_ns, src_inode)) {
1839 * Subvolume creation is not restricted, but snapshots
1840 * are limited to own subvolumes only
1844 ret = btrfs_mksnapshot(&file->f_path, name, namelen,
1845 BTRFS_I(src_inode)->root,
1851 mnt_drop_write_file(file);
1856 static noinline int btrfs_ioctl_snap_create(struct file *file,
1857 void __user *arg, int subvol)
1859 struct btrfs_ioctl_vol_args *vol_args;
1862 if (!S_ISDIR(file_inode(file)->i_mode))
1865 vol_args = memdup_user(arg, sizeof(*vol_args));
1866 if (IS_ERR(vol_args))
1867 return PTR_ERR(vol_args);
1868 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1870 ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1871 subvol, false, NULL);
1877 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1878 void __user *arg, int subvol)
1880 struct btrfs_ioctl_vol_args_v2 *vol_args;
1882 bool readonly = false;
1883 struct btrfs_qgroup_inherit *inherit = NULL;
1885 if (!S_ISDIR(file_inode(file)->i_mode))
1888 vol_args = memdup_user(arg, sizeof(*vol_args));
1889 if (IS_ERR(vol_args))
1890 return PTR_ERR(vol_args);
1891 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1893 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1898 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1900 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1903 if (vol_args->size < sizeof(*inherit) ||
1904 vol_args->size > PAGE_SIZE) {
1908 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1909 if (IS_ERR(inherit)) {
1910 ret = PTR_ERR(inherit);
1914 if (inherit->num_qgroups > PAGE_SIZE ||
1915 inherit->num_ref_copies > PAGE_SIZE ||
1916 inherit->num_excl_copies > PAGE_SIZE) {
1921 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
1922 2 * inherit->num_excl_copies;
1923 if (vol_args->size != struct_size(inherit, qgroups, nums)) {
1929 ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1930 subvol, readonly, inherit);
1940 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1943 struct inode *inode = file_inode(file);
1944 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1945 struct btrfs_root *root = BTRFS_I(inode)->root;
1949 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1952 down_read(&fs_info->subvol_sem);
1953 if (btrfs_root_readonly(root))
1954 flags |= BTRFS_SUBVOL_RDONLY;
1955 up_read(&fs_info->subvol_sem);
1957 if (copy_to_user(arg, &flags, sizeof(flags)))
1963 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1966 struct inode *inode = file_inode(file);
1967 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1968 struct btrfs_root *root = BTRFS_I(inode)->root;
1969 struct btrfs_trans_handle *trans;
1974 if (!inode_owner_or_capable(&init_user_ns, inode))
1977 ret = mnt_want_write_file(file);
1981 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1983 goto out_drop_write;
1986 if (copy_from_user(&flags, arg, sizeof(flags))) {
1988 goto out_drop_write;
1991 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1993 goto out_drop_write;
1996 down_write(&fs_info->subvol_sem);
1999 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
2002 root_flags = btrfs_root_flags(&root->root_item);
2003 if (flags & BTRFS_SUBVOL_RDONLY) {
2004 btrfs_set_root_flags(&root->root_item,
2005 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2008 * Block RO -> RW transition if this subvolume is involved in
2011 spin_lock(&root->root_item_lock);
2012 if (root->send_in_progress == 0) {
2013 btrfs_set_root_flags(&root->root_item,
2014 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2015 spin_unlock(&root->root_item_lock);
2017 spin_unlock(&root->root_item_lock);
2019 "Attempt to set subvolume %llu read-write during send",
2020 root->root_key.objectid);
2026 trans = btrfs_start_transaction(root, 1);
2027 if (IS_ERR(trans)) {
2028 ret = PTR_ERR(trans);
2032 ret = btrfs_update_root(trans, fs_info->tree_root,
2033 &root->root_key, &root->root_item);
2035 btrfs_end_transaction(trans);
2039 ret = btrfs_commit_transaction(trans);
2043 btrfs_set_root_flags(&root->root_item, root_flags);
2045 up_write(&fs_info->subvol_sem);
2047 mnt_drop_write_file(file);
2052 static noinline int key_in_sk(struct btrfs_key *key,
2053 struct btrfs_ioctl_search_key *sk)
2055 struct btrfs_key test;
2058 test.objectid = sk->min_objectid;
2059 test.type = sk->min_type;
2060 test.offset = sk->min_offset;
2062 ret = btrfs_comp_cpu_keys(key, &test);
2066 test.objectid = sk->max_objectid;
2067 test.type = sk->max_type;
2068 test.offset = sk->max_offset;
2070 ret = btrfs_comp_cpu_keys(key, &test);
2076 static noinline int copy_to_sk(struct btrfs_path *path,
2077 struct btrfs_key *key,
2078 struct btrfs_ioctl_search_key *sk,
2081 unsigned long *sk_offset,
2085 struct extent_buffer *leaf;
2086 struct btrfs_ioctl_search_header sh;
2087 struct btrfs_key test;
2088 unsigned long item_off;
2089 unsigned long item_len;
2095 leaf = path->nodes[0];
2096 slot = path->slots[0];
2097 nritems = btrfs_header_nritems(leaf);
2099 if (btrfs_header_generation(leaf) > sk->max_transid) {
2103 found_transid = btrfs_header_generation(leaf);
2105 for (i = slot; i < nritems; i++) {
2106 item_off = btrfs_item_ptr_offset(leaf, i);
2107 item_len = btrfs_item_size_nr(leaf, i);
2109 btrfs_item_key_to_cpu(leaf, key, i);
2110 if (!key_in_sk(key, sk))
2113 if (sizeof(sh) + item_len > *buf_size) {
2120 * return one empty item back for v1, which does not
2124 *buf_size = sizeof(sh) + item_len;
2129 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2134 sh.objectid = key->objectid;
2135 sh.offset = key->offset;
2136 sh.type = key->type;
2138 sh.transid = found_transid;
2141 * Copy search result header. If we fault then loop again so we
2142 * can fault in the pages and -EFAULT there if there's a
2143 * problem. Otherwise we'll fault and then copy the buffer in
2144 * properly this next time through
2146 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
2151 *sk_offset += sizeof(sh);
2154 char __user *up = ubuf + *sk_offset;
2156 * Copy the item, same behavior as above, but reset the
2157 * * sk_offset so we copy the full thing again.
2159 if (read_extent_buffer_to_user_nofault(leaf, up,
2160 item_off, item_len)) {
2162 *sk_offset -= sizeof(sh);
2166 *sk_offset += item_len;
2170 if (ret) /* -EOVERFLOW from above */
2173 if (*num_found >= sk->nr_items) {
2180 test.objectid = sk->max_objectid;
2181 test.type = sk->max_type;
2182 test.offset = sk->max_offset;
2183 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2185 else if (key->offset < (u64)-1)
2187 else if (key->type < (u8)-1) {
2190 } else if (key->objectid < (u64)-1) {
2198 * 0: all items from this leaf copied, continue with next
2199 * 1: * more items can be copied, but unused buffer is too small
2200 * * all items were found
2201 * Either way, it will stops the loop which iterates to the next
2203 * -EOVERFLOW: item was to large for buffer
2204 * -EFAULT: could not copy extent buffer back to userspace
2209 static noinline int search_ioctl(struct inode *inode,
2210 struct btrfs_ioctl_search_key *sk,
2214 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2215 struct btrfs_root *root;
2216 struct btrfs_key key;
2217 struct btrfs_path *path;
2220 unsigned long sk_offset = 0;
2222 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2223 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2227 path = btrfs_alloc_path();
2231 if (sk->tree_id == 0) {
2232 /* search the root of the inode that was passed */
2233 root = btrfs_grab_root(BTRFS_I(inode)->root);
2235 root = btrfs_get_fs_root(info, sk->tree_id, true);
2237 btrfs_free_path(path);
2238 return PTR_ERR(root);
2242 key.objectid = sk->min_objectid;
2243 key.type = sk->min_type;
2244 key.offset = sk->min_offset;
2247 ret = fault_in_pages_writeable(ubuf + sk_offset,
2248 *buf_size - sk_offset);
2252 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2258 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2259 &sk_offset, &num_found);
2260 btrfs_release_path(path);
2268 sk->nr_items = num_found;
2269 btrfs_put_root(root);
2270 btrfs_free_path(path);
2274 static noinline int btrfs_ioctl_tree_search(struct file *file,
2277 struct btrfs_ioctl_search_args __user *uargs;
2278 struct btrfs_ioctl_search_key sk;
2279 struct inode *inode;
2283 if (!capable(CAP_SYS_ADMIN))
2286 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2288 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2291 buf_size = sizeof(uargs->buf);
2293 inode = file_inode(file);
2294 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2297 * In the origin implementation an overflow is handled by returning a
2298 * search header with a len of zero, so reset ret.
2300 if (ret == -EOVERFLOW)
2303 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2308 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2311 struct btrfs_ioctl_search_args_v2 __user *uarg;
2312 struct btrfs_ioctl_search_args_v2 args;
2313 struct inode *inode;
2316 const size_t buf_limit = SZ_16M;
2318 if (!capable(CAP_SYS_ADMIN))
2321 /* copy search header and buffer size */
2322 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2323 if (copy_from_user(&args, uarg, sizeof(args)))
2326 buf_size = args.buf_size;
2328 /* limit result size to 16MB */
2329 if (buf_size > buf_limit)
2330 buf_size = buf_limit;
2332 inode = file_inode(file);
2333 ret = search_ioctl(inode, &args.key, &buf_size,
2334 (char __user *)(&uarg->buf[0]));
2335 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2337 else if (ret == -EOVERFLOW &&
2338 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2345 * Search INODE_REFs to identify path name of 'dirid' directory
2346 * in a 'tree_id' tree. and sets path name to 'name'.
2348 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2349 u64 tree_id, u64 dirid, char *name)
2351 struct btrfs_root *root;
2352 struct btrfs_key key;
2358 struct btrfs_inode_ref *iref;
2359 struct extent_buffer *l;
2360 struct btrfs_path *path;
2362 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2367 path = btrfs_alloc_path();
2371 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2373 root = btrfs_get_fs_root(info, tree_id, true);
2375 ret = PTR_ERR(root);
2380 key.objectid = dirid;
2381 key.type = BTRFS_INODE_REF_KEY;
2382 key.offset = (u64)-1;
2385 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2389 ret = btrfs_previous_item(root, path, dirid,
2390 BTRFS_INODE_REF_KEY);
2400 slot = path->slots[0];
2401 btrfs_item_key_to_cpu(l, &key, slot);
2403 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2404 len = btrfs_inode_ref_name_len(l, iref);
2406 total_len += len + 1;
2408 ret = -ENAMETOOLONG;
2413 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2415 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2418 btrfs_release_path(path);
2419 key.objectid = key.offset;
2420 key.offset = (u64)-1;
2421 dirid = key.objectid;
2423 memmove(name, ptr, total_len);
2424 name[total_len] = '\0';
2427 btrfs_put_root(root);
2428 btrfs_free_path(path);
2432 static int btrfs_search_path_in_tree_user(struct inode *inode,
2433 struct btrfs_ioctl_ino_lookup_user_args *args)
2435 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2436 struct super_block *sb = inode->i_sb;
2437 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2438 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2439 u64 dirid = args->dirid;
2440 unsigned long item_off;
2441 unsigned long item_len;
2442 struct btrfs_inode_ref *iref;
2443 struct btrfs_root_ref *rref;
2444 struct btrfs_root *root = NULL;
2445 struct btrfs_path *path;
2446 struct btrfs_key key, key2;
2447 struct extent_buffer *leaf;
2448 struct inode *temp_inode;
2455 path = btrfs_alloc_path();
2460 * If the bottom subvolume does not exist directly under upper_limit,
2461 * construct the path in from the bottom up.
2463 if (dirid != upper_limit.objectid) {
2464 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2466 root = btrfs_get_fs_root(fs_info, treeid, true);
2468 ret = PTR_ERR(root);
2472 key.objectid = dirid;
2473 key.type = BTRFS_INODE_REF_KEY;
2474 key.offset = (u64)-1;
2476 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2479 } else if (ret > 0) {
2480 ret = btrfs_previous_item(root, path, dirid,
2481 BTRFS_INODE_REF_KEY);
2484 } else if (ret > 0) {
2490 leaf = path->nodes[0];
2491 slot = path->slots[0];
2492 btrfs_item_key_to_cpu(leaf, &key, slot);
2494 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2495 len = btrfs_inode_ref_name_len(leaf, iref);
2497 total_len += len + 1;
2498 if (ptr < args->path) {
2499 ret = -ENAMETOOLONG;
2504 read_extent_buffer(leaf, ptr,
2505 (unsigned long)(iref + 1), len);
2507 /* Check the read+exec permission of this directory */
2508 ret = btrfs_previous_item(root, path, dirid,
2509 BTRFS_INODE_ITEM_KEY);
2512 } else if (ret > 0) {
2517 leaf = path->nodes[0];
2518 slot = path->slots[0];
2519 btrfs_item_key_to_cpu(leaf, &key2, slot);
2520 if (key2.objectid != dirid) {
2525 temp_inode = btrfs_iget(sb, key2.objectid, root);
2526 if (IS_ERR(temp_inode)) {
2527 ret = PTR_ERR(temp_inode);
2530 ret = inode_permission(&init_user_ns, temp_inode,
2531 MAY_READ | MAY_EXEC);
2538 if (key.offset == upper_limit.objectid)
2540 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2545 btrfs_release_path(path);
2546 key.objectid = key.offset;
2547 key.offset = (u64)-1;
2548 dirid = key.objectid;
2551 memmove(args->path, ptr, total_len);
2552 args->path[total_len] = '\0';
2553 btrfs_put_root(root);
2555 btrfs_release_path(path);
2558 /* Get the bottom subvolume's name from ROOT_REF */
2559 key.objectid = treeid;
2560 key.type = BTRFS_ROOT_REF_KEY;
2561 key.offset = args->treeid;
2562 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2565 } else if (ret > 0) {
2570 leaf = path->nodes[0];
2571 slot = path->slots[0];
2572 btrfs_item_key_to_cpu(leaf, &key, slot);
2574 item_off = btrfs_item_ptr_offset(leaf, slot);
2575 item_len = btrfs_item_size_nr(leaf, slot);
2576 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2577 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2578 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2583 /* Copy subvolume's name */
2584 item_off += sizeof(struct btrfs_root_ref);
2585 item_len -= sizeof(struct btrfs_root_ref);
2586 read_extent_buffer(leaf, args->name, item_off, item_len);
2587 args->name[item_len] = 0;
2590 btrfs_put_root(root);
2592 btrfs_free_path(path);
2596 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2599 struct btrfs_ioctl_ino_lookup_args *args;
2600 struct inode *inode;
2603 args = memdup_user(argp, sizeof(*args));
2605 return PTR_ERR(args);
2607 inode = file_inode(file);
2610 * Unprivileged query to obtain the containing subvolume root id. The
2611 * path is reset so it's consistent with btrfs_search_path_in_tree.
2613 if (args->treeid == 0)
2614 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2616 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2621 if (!capable(CAP_SYS_ADMIN)) {
2626 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2627 args->treeid, args->objectid,
2631 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2639 * Version of ino_lookup ioctl (unprivileged)
2641 * The main differences from ino_lookup ioctl are:
2643 * 1. Read + Exec permission will be checked using inode_permission() during
2644 * path construction. -EACCES will be returned in case of failure.
2645 * 2. Path construction will be stopped at the inode number which corresponds
2646 * to the fd with which this ioctl is called. If constructed path does not
2647 * exist under fd's inode, -EACCES will be returned.
2648 * 3. The name of bottom subvolume is also searched and filled.
2650 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2652 struct btrfs_ioctl_ino_lookup_user_args *args;
2653 struct inode *inode;
2656 args = memdup_user(argp, sizeof(*args));
2658 return PTR_ERR(args);
2660 inode = file_inode(file);
2662 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2663 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2665 * The subvolume does not exist under fd with which this is
2672 ret = btrfs_search_path_in_tree_user(inode, args);
2674 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2681 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2682 static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2684 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2685 struct btrfs_fs_info *fs_info;
2686 struct btrfs_root *root;
2687 struct btrfs_path *path;
2688 struct btrfs_key key;
2689 struct btrfs_root_item *root_item;
2690 struct btrfs_root_ref *rref;
2691 struct extent_buffer *leaf;
2692 unsigned long item_off;
2693 unsigned long item_len;
2694 struct inode *inode;
2698 path = btrfs_alloc_path();
2702 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2704 btrfs_free_path(path);
2708 inode = file_inode(file);
2709 fs_info = BTRFS_I(inode)->root->fs_info;
2711 /* Get root_item of inode's subvolume */
2712 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2713 root = btrfs_get_fs_root(fs_info, key.objectid, true);
2715 ret = PTR_ERR(root);
2718 root_item = &root->root_item;
2720 subvol_info->treeid = key.objectid;
2722 subvol_info->generation = btrfs_root_generation(root_item);
2723 subvol_info->flags = btrfs_root_flags(root_item);
2725 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2726 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2728 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2731 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2732 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2733 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2735 subvol_info->otransid = btrfs_root_otransid(root_item);
2736 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2737 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2739 subvol_info->stransid = btrfs_root_stransid(root_item);
2740 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2741 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2743 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2744 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2745 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2747 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2748 /* Search root tree for ROOT_BACKREF of this subvolume */
2749 key.type = BTRFS_ROOT_BACKREF_KEY;
2751 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2754 } else if (path->slots[0] >=
2755 btrfs_header_nritems(path->nodes[0])) {
2756 ret = btrfs_next_leaf(fs_info->tree_root, path);
2759 } else if (ret > 0) {
2765 leaf = path->nodes[0];
2766 slot = path->slots[0];
2767 btrfs_item_key_to_cpu(leaf, &key, slot);
2768 if (key.objectid == subvol_info->treeid &&
2769 key.type == BTRFS_ROOT_BACKREF_KEY) {
2770 subvol_info->parent_id = key.offset;
2772 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2773 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2775 item_off = btrfs_item_ptr_offset(leaf, slot)
2776 + sizeof(struct btrfs_root_ref);
2777 item_len = btrfs_item_size_nr(leaf, slot)
2778 - sizeof(struct btrfs_root_ref);
2779 read_extent_buffer(leaf, subvol_info->name,
2780 item_off, item_len);
2787 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2791 btrfs_put_root(root);
2793 btrfs_free_path(path);
2799 * Return ROOT_REF information of the subvolume containing this inode
2800 * except the subvolume name.
2802 static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2804 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2805 struct btrfs_root_ref *rref;
2806 struct btrfs_root *root;
2807 struct btrfs_path *path;
2808 struct btrfs_key key;
2809 struct extent_buffer *leaf;
2810 struct inode *inode;
2816 path = btrfs_alloc_path();
2820 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2821 if (IS_ERR(rootrefs)) {
2822 btrfs_free_path(path);
2823 return PTR_ERR(rootrefs);
2826 inode = file_inode(file);
2827 root = BTRFS_I(inode)->root->fs_info->tree_root;
2828 objectid = BTRFS_I(inode)->root->root_key.objectid;
2830 key.objectid = objectid;
2831 key.type = BTRFS_ROOT_REF_KEY;
2832 key.offset = rootrefs->min_treeid;
2835 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2838 } else if (path->slots[0] >=
2839 btrfs_header_nritems(path->nodes[0])) {
2840 ret = btrfs_next_leaf(root, path);
2843 } else if (ret > 0) {
2849 leaf = path->nodes[0];
2850 slot = path->slots[0];
2852 btrfs_item_key_to_cpu(leaf, &key, slot);
2853 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2858 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2863 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2864 rootrefs->rootref[found].treeid = key.offset;
2865 rootrefs->rootref[found].dirid =
2866 btrfs_root_ref_dirid(leaf, rref);
2869 ret = btrfs_next_item(root, path);
2872 } else if (ret > 0) {
2879 if (!ret || ret == -EOVERFLOW) {
2880 rootrefs->num_items = found;
2881 /* update min_treeid for next search */
2883 rootrefs->min_treeid =
2884 rootrefs->rootref[found - 1].treeid + 1;
2885 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2890 btrfs_free_path(path);
2895 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2899 struct dentry *parent = file->f_path.dentry;
2900 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2901 struct dentry *dentry;
2902 struct inode *dir = d_inode(parent);
2903 struct inode *inode;
2904 struct btrfs_root *root = BTRFS_I(dir)->root;
2905 struct btrfs_root *dest = NULL;
2906 struct btrfs_ioctl_vol_args *vol_args = NULL;
2907 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2908 char *subvol_name, *subvol_name_ptr = NULL;
2911 bool destroy_parent = false;
2914 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2915 if (IS_ERR(vol_args2))
2916 return PTR_ERR(vol_args2);
2918 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2924 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2925 * name, same as v1 currently does.
2927 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2928 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2929 subvol_name = vol_args2->name;
2931 err = mnt_want_write_file(file);
2935 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2940 err = mnt_want_write_file(file);
2944 dentry = btrfs_get_dentry(fs_info->sb,
2945 BTRFS_FIRST_FREE_OBJECTID,
2946 vol_args2->subvolid, 0, 0);
2947 if (IS_ERR(dentry)) {
2948 err = PTR_ERR(dentry);
2949 goto out_drop_write;
2953 * Change the default parent since the subvolume being
2954 * deleted can be outside of the current mount point.
2956 parent = btrfs_get_parent(dentry);
2959 * At this point dentry->d_name can point to '/' if the
2960 * subvolume we want to destroy is outsite of the
2961 * current mount point, so we need to release the
2962 * current dentry and execute the lookup to return a new
2963 * one with ->d_name pointing to the
2964 * <mount point>/subvol_name.
2967 if (IS_ERR(parent)) {
2968 err = PTR_ERR(parent);
2969 goto out_drop_write;
2971 dir = d_inode(parent);
2974 * If v2 was used with SPEC_BY_ID, a new parent was
2975 * allocated since the subvolume can be outside of the
2976 * current mount point. Later on we need to release this
2977 * new parent dentry.
2979 destroy_parent = true;
2981 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2982 fs_info, vol_args2->subvolid);
2983 if (IS_ERR(subvol_name_ptr)) {
2984 err = PTR_ERR(subvol_name_ptr);
2987 /* subvol_name_ptr is already nul terminated */
2988 subvol_name = (char *)kbasename(subvol_name_ptr);
2991 vol_args = memdup_user(arg, sizeof(*vol_args));
2992 if (IS_ERR(vol_args))
2993 return PTR_ERR(vol_args);
2995 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
2996 subvol_name = vol_args->name;
2998 err = mnt_want_write_file(file);
3003 subvol_namelen = strlen(subvol_name);
3005 if (strchr(subvol_name, '/') ||
3006 strncmp(subvol_name, "..", subvol_namelen) == 0) {
3008 goto free_subvol_name;
3011 if (!S_ISDIR(dir->i_mode)) {
3013 goto free_subvol_name;
3016 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
3018 goto free_subvol_name;
3019 dentry = lookup_one_len(subvol_name, parent, subvol_namelen);
3020 if (IS_ERR(dentry)) {
3021 err = PTR_ERR(dentry);
3022 goto out_unlock_dir;
3025 if (d_really_is_negative(dentry)) {
3030 inode = d_inode(dentry);
3031 dest = BTRFS_I(inode)->root;
3032 if (!capable(CAP_SYS_ADMIN)) {
3034 * Regular user. Only allow this with a special mount
3035 * option, when the user has write+exec access to the
3036 * subvol root, and when rmdir(2) would have been
3039 * Note that this is _not_ check that the subvol is
3040 * empty or doesn't contain data that we wouldn't
3041 * otherwise be able to delete.
3043 * Users who want to delete empty subvols should try
3047 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
3051 * Do not allow deletion if the parent dir is the same
3052 * as the dir to be deleted. That means the ioctl
3053 * must be called on the dentry referencing the root
3054 * of the subvol, not a random directory contained
3061 err = inode_permission(&init_user_ns, inode,
3062 MAY_WRITE | MAY_EXEC);
3067 /* check if subvolume may be deleted by a user */
3068 err = btrfs_may_delete(dir, dentry, 1);
3072 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3077 btrfs_inode_lock(inode, 0);
3078 err = btrfs_delete_subvolume(dir, dentry);
3079 btrfs_inode_unlock(inode, 0);
3081 fsnotify_rmdir(dir, dentry);
3088 btrfs_inode_unlock(dir, 0);
3090 kfree(subvol_name_ptr);
3095 mnt_drop_write_file(file);
3102 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
3104 struct inode *inode = file_inode(file);
3105 struct btrfs_root *root = BTRFS_I(inode)->root;
3106 struct btrfs_ioctl_defrag_range_args *range;
3109 ret = mnt_want_write_file(file);
3113 if (btrfs_root_readonly(root)) {
3118 switch (inode->i_mode & S_IFMT) {
3120 if (!capable(CAP_SYS_ADMIN)) {
3124 ret = btrfs_defrag_root(root);
3128 * Note that this does not check the file descriptor for write
3129 * access. This prevents defragmenting executables that are
3130 * running and allows defrag on files open in read-only mode.
3132 if (!capable(CAP_SYS_ADMIN) &&
3133 inode_permission(&init_user_ns, inode, MAY_WRITE)) {
3138 range = kzalloc(sizeof(*range), GFP_KERNEL);
3145 if (copy_from_user(range, argp,
3151 /* compression requires us to start the IO */
3152 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3153 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
3154 range->extent_thresh = (u32)-1;
3157 /* the rest are all set to zero by kzalloc */
3158 range->len = (u64)-1;
3160 ret = btrfs_defrag_file(file_inode(file), file,
3161 range, BTRFS_OLDEST_GENERATION, 0);
3170 mnt_drop_write_file(file);
3174 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
3176 struct btrfs_ioctl_vol_args *vol_args;
3179 if (!capable(CAP_SYS_ADMIN))
3182 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD))
3183 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3185 vol_args = memdup_user(arg, sizeof(*vol_args));
3186 if (IS_ERR(vol_args)) {
3187 ret = PTR_ERR(vol_args);
3191 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3192 ret = btrfs_init_new_device(fs_info, vol_args->name);
3195 btrfs_info(fs_info, "disk added %s", vol_args->name);
3199 btrfs_exclop_finish(fs_info);
3203 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
3205 struct inode *inode = file_inode(file);
3206 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3207 struct btrfs_ioctl_vol_args_v2 *vol_args;
3209 bool cancel = false;
3211 if (!capable(CAP_SYS_ADMIN))
3214 ret = mnt_want_write_file(file);
3218 vol_args = memdup_user(arg, sizeof(*vol_args));
3219 if (IS_ERR(vol_args)) {
3220 ret = PTR_ERR(vol_args);
3224 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
3228 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3229 if (!(vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) &&
3230 strcmp("cancel", vol_args->name) == 0)
3233 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3237 /* Exclusive operation is now claimed */
3239 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3240 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
3242 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3244 btrfs_exclop_finish(fs_info);
3247 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3248 btrfs_info(fs_info, "device deleted: id %llu",
3251 btrfs_info(fs_info, "device deleted: %s",
3257 mnt_drop_write_file(file);
3261 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3263 struct inode *inode = file_inode(file);
3264 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3265 struct btrfs_ioctl_vol_args *vol_args;
3269 if (!capable(CAP_SYS_ADMIN))
3272 ret = mnt_want_write_file(file);
3276 vol_args = memdup_user(arg, sizeof(*vol_args));
3277 if (IS_ERR(vol_args)) {
3278 ret = PTR_ERR(vol_args);
3279 goto out_drop_write;
3281 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3282 cancel = (strcmp("cancel", vol_args->name) == 0);
3284 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3287 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3289 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3290 btrfs_exclop_finish(fs_info);
3295 mnt_drop_write_file(file);
3300 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3303 struct btrfs_ioctl_fs_info_args *fi_args;
3304 struct btrfs_device *device;
3305 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3309 fi_args = memdup_user(arg, sizeof(*fi_args));
3310 if (IS_ERR(fi_args))
3311 return PTR_ERR(fi_args);
3313 flags_in = fi_args->flags;
3314 memset(fi_args, 0, sizeof(*fi_args));
3317 fi_args->num_devices = fs_devices->num_devices;
3319 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3320 if (device->devid > fi_args->max_id)
3321 fi_args->max_id = device->devid;
3325 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
3326 fi_args->nodesize = fs_info->nodesize;
3327 fi_args->sectorsize = fs_info->sectorsize;
3328 fi_args->clone_alignment = fs_info->sectorsize;
3330 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
3331 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
3332 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
3333 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
3336 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
3337 fi_args->generation = fs_info->generation;
3338 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
3341 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
3342 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
3343 sizeof(fi_args->metadata_uuid));
3344 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
3347 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3354 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3357 struct btrfs_ioctl_dev_info_args *di_args;
3358 struct btrfs_device *dev;
3360 char *s_uuid = NULL;
3362 di_args = memdup_user(arg, sizeof(*di_args));
3363 if (IS_ERR(di_args))
3364 return PTR_ERR(di_args);
3366 if (!btrfs_is_empty_uuid(di_args->uuid))
3367 s_uuid = di_args->uuid;
3370 dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid,
3378 di_args->devid = dev->devid;
3379 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3380 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3381 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3383 strncpy(di_args->path, rcu_str_deref(dev->name),
3384 sizeof(di_args->path) - 1);
3385 di_args->path[sizeof(di_args->path) - 1] = 0;
3387 di_args->path[0] = '\0';
3392 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3399 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3401 struct inode *inode = file_inode(file);
3402 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3403 struct btrfs_root *root = BTRFS_I(inode)->root;
3404 struct btrfs_root *new_root;
3405 struct btrfs_dir_item *di;
3406 struct btrfs_trans_handle *trans;
3407 struct btrfs_path *path = NULL;
3408 struct btrfs_disk_key disk_key;
3413 if (!capable(CAP_SYS_ADMIN))
3416 ret = mnt_want_write_file(file);
3420 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3426 objectid = BTRFS_FS_TREE_OBJECTID;
3428 new_root = btrfs_get_fs_root(fs_info, objectid, true);
3429 if (IS_ERR(new_root)) {
3430 ret = PTR_ERR(new_root);
3433 if (!is_fstree(new_root->root_key.objectid)) {
3438 path = btrfs_alloc_path();
3444 trans = btrfs_start_transaction(root, 1);
3445 if (IS_ERR(trans)) {
3446 ret = PTR_ERR(trans);
3450 dir_id = btrfs_super_root_dir(fs_info->super_copy);
3451 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3452 dir_id, "default", 7, 1);
3453 if (IS_ERR_OR_NULL(di)) {
3454 btrfs_release_path(path);
3455 btrfs_end_transaction(trans);
3457 "Umm, you don't have the default diritem, this isn't going to work");
3462 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3463 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3464 btrfs_mark_buffer_dirty(path->nodes[0]);
3465 btrfs_release_path(path);
3467 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3468 btrfs_end_transaction(trans);
3470 btrfs_put_root(new_root);
3471 btrfs_free_path(path);
3473 mnt_drop_write_file(file);
3477 static void get_block_group_info(struct list_head *groups_list,
3478 struct btrfs_ioctl_space_info *space)
3480 struct btrfs_block_group *block_group;
3482 space->total_bytes = 0;
3483 space->used_bytes = 0;
3485 list_for_each_entry(block_group, groups_list, list) {
3486 space->flags = block_group->flags;
3487 space->total_bytes += block_group->length;
3488 space->used_bytes += block_group->used;
3492 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3495 struct btrfs_ioctl_space_args space_args;
3496 struct btrfs_ioctl_space_info space;
3497 struct btrfs_ioctl_space_info *dest;
3498 struct btrfs_ioctl_space_info *dest_orig;
3499 struct btrfs_ioctl_space_info __user *user_dest;
3500 struct btrfs_space_info *info;
3501 static const u64 types[] = {
3502 BTRFS_BLOCK_GROUP_DATA,
3503 BTRFS_BLOCK_GROUP_SYSTEM,
3504 BTRFS_BLOCK_GROUP_METADATA,
3505 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3513 if (copy_from_user(&space_args,
3514 (struct btrfs_ioctl_space_args __user *)arg,
3515 sizeof(space_args)))
3518 for (i = 0; i < num_types; i++) {
3519 struct btrfs_space_info *tmp;
3522 list_for_each_entry(tmp, &fs_info->space_info, list) {
3523 if (tmp->flags == types[i]) {
3532 down_read(&info->groups_sem);
3533 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3534 if (!list_empty(&info->block_groups[c]))
3537 up_read(&info->groups_sem);
3541 * Global block reserve, exported as a space_info
3545 /* space_slots == 0 means they are asking for a count */
3546 if (space_args.space_slots == 0) {
3547 space_args.total_spaces = slot_count;
3551 slot_count = min_t(u64, space_args.space_slots, slot_count);
3553 alloc_size = sizeof(*dest) * slot_count;
3555 /* we generally have at most 6 or so space infos, one for each raid
3556 * level. So, a whole page should be more than enough for everyone
3558 if (alloc_size > PAGE_SIZE)
3561 space_args.total_spaces = 0;
3562 dest = kmalloc(alloc_size, GFP_KERNEL);
3567 /* now we have a buffer to copy into */
3568 for (i = 0; i < num_types; i++) {
3569 struct btrfs_space_info *tmp;
3575 list_for_each_entry(tmp, &fs_info->space_info, list) {
3576 if (tmp->flags == types[i]) {
3584 down_read(&info->groups_sem);
3585 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3586 if (!list_empty(&info->block_groups[c])) {
3587 get_block_group_info(&info->block_groups[c],
3589 memcpy(dest, &space, sizeof(space));
3591 space_args.total_spaces++;
3597 up_read(&info->groups_sem);
3601 * Add global block reserve
3604 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3606 spin_lock(&block_rsv->lock);
3607 space.total_bytes = block_rsv->size;
3608 space.used_bytes = block_rsv->size - block_rsv->reserved;
3609 spin_unlock(&block_rsv->lock);
3610 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3611 memcpy(dest, &space, sizeof(space));
3612 space_args.total_spaces++;
3615 user_dest = (struct btrfs_ioctl_space_info __user *)
3616 (arg + sizeof(struct btrfs_ioctl_space_args));
3618 if (copy_to_user(user_dest, dest_orig, alloc_size))
3623 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3629 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3632 struct btrfs_trans_handle *trans;
3636 trans = btrfs_attach_transaction_barrier(root);
3637 if (IS_ERR(trans)) {
3638 if (PTR_ERR(trans) != -ENOENT)
3639 return PTR_ERR(trans);
3641 /* No running transaction, don't bother */
3642 transid = root->fs_info->last_trans_committed;
3645 transid = trans->transid;
3646 ret = btrfs_commit_transaction_async(trans);
3648 btrfs_end_transaction(trans);
3653 if (copy_to_user(argp, &transid, sizeof(transid)))
3658 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3664 if (copy_from_user(&transid, argp, sizeof(transid)))
3667 transid = 0; /* current trans */
3669 return btrfs_wait_for_commit(fs_info, transid);
3672 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3674 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3675 struct btrfs_ioctl_scrub_args *sa;
3678 if (!capable(CAP_SYS_ADMIN))
3681 sa = memdup_user(arg, sizeof(*sa));
3685 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3686 ret = mnt_want_write_file(file);
3691 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3692 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3696 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3697 * error. This is important as it allows user space to know how much
3698 * progress scrub has done. For example, if scrub is canceled we get
3699 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3700 * space. Later user space can inspect the progress from the structure
3701 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3702 * previously (btrfs-progs does this).
3703 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3704 * then return -EFAULT to signal the structure was not copied or it may
3705 * be corrupt and unreliable due to a partial copy.
3707 if (copy_to_user(arg, sa, sizeof(*sa)))
3710 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3711 mnt_drop_write_file(file);
3717 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3719 if (!capable(CAP_SYS_ADMIN))
3722 return btrfs_scrub_cancel(fs_info);
3725 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3728 struct btrfs_ioctl_scrub_args *sa;
3731 if (!capable(CAP_SYS_ADMIN))
3734 sa = memdup_user(arg, sizeof(*sa));
3738 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3740 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3747 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3750 struct btrfs_ioctl_get_dev_stats *sa;
3753 sa = memdup_user(arg, sizeof(*sa));
3757 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3762 ret = btrfs_get_dev_stats(fs_info, sa);
3764 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3771 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3774 struct btrfs_ioctl_dev_replace_args *p;
3777 if (!capable(CAP_SYS_ADMIN))
3780 p = memdup_user(arg, sizeof(*p));
3785 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3786 if (sb_rdonly(fs_info->sb)) {
3790 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3791 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3793 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3794 btrfs_exclop_finish(fs_info);
3797 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3798 btrfs_dev_replace_status(fs_info, p);
3801 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3802 p->result = btrfs_dev_replace_cancel(fs_info);
3810 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3817 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3823 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3824 struct inode_fs_paths *ipath = NULL;
3825 struct btrfs_path *path;
3827 if (!capable(CAP_DAC_READ_SEARCH))
3830 path = btrfs_alloc_path();
3836 ipa = memdup_user(arg, sizeof(*ipa));
3843 size = min_t(u32, ipa->size, 4096);
3844 ipath = init_ipath(size, root, path);
3845 if (IS_ERR(ipath)) {
3846 ret = PTR_ERR(ipath);
3851 ret = paths_from_inode(ipa->inum, ipath);
3855 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3856 rel_ptr = ipath->fspath->val[i] -
3857 (u64)(unsigned long)ipath->fspath->val;
3858 ipath->fspath->val[i] = rel_ptr;
3861 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3862 ipath->fspath, size);
3869 btrfs_free_path(path);
3876 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3878 struct btrfs_data_container *inodes = ctx;
3879 const size_t c = 3 * sizeof(u64);
3881 if (inodes->bytes_left >= c) {
3882 inodes->bytes_left -= c;
3883 inodes->val[inodes->elem_cnt] = inum;
3884 inodes->val[inodes->elem_cnt + 1] = offset;
3885 inodes->val[inodes->elem_cnt + 2] = root;
3886 inodes->elem_cnt += 3;
3888 inodes->bytes_missing += c - inodes->bytes_left;
3889 inodes->bytes_left = 0;
3890 inodes->elem_missed += 3;
3896 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3897 void __user *arg, int version)
3901 struct btrfs_ioctl_logical_ino_args *loi;
3902 struct btrfs_data_container *inodes = NULL;
3903 struct btrfs_path *path = NULL;
3906 if (!capable(CAP_SYS_ADMIN))
3909 loi = memdup_user(arg, sizeof(*loi));
3911 return PTR_ERR(loi);
3914 ignore_offset = false;
3915 size = min_t(u32, loi->size, SZ_64K);
3917 /* All reserved bits must be 0 for now */
3918 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3922 /* Only accept flags we have defined so far */
3923 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3927 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3928 size = min_t(u32, loi->size, SZ_16M);
3931 path = btrfs_alloc_path();
3937 inodes = init_data_container(size);
3938 if (IS_ERR(inodes)) {
3939 ret = PTR_ERR(inodes);
3944 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3945 build_ino_list, inodes, ignore_offset);
3951 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3957 btrfs_free_path(path);
3965 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3966 struct btrfs_ioctl_balance_args *bargs)
3968 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3970 bargs->flags = bctl->flags;
3972 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3973 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3974 if (atomic_read(&fs_info->balance_pause_req))
3975 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3976 if (atomic_read(&fs_info->balance_cancel_req))
3977 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3979 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3980 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3981 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3983 spin_lock(&fs_info->balance_lock);
3984 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3985 spin_unlock(&fs_info->balance_lock);
3988 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3990 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3991 struct btrfs_fs_info *fs_info = root->fs_info;
3992 struct btrfs_ioctl_balance_args *bargs;
3993 struct btrfs_balance_control *bctl;
3994 bool need_unlock; /* for mut. excl. ops lock */
3997 if (!capable(CAP_SYS_ADMIN))
4000 ret = mnt_want_write_file(file);
4005 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
4006 mutex_lock(&fs_info->balance_mutex);
4012 * mut. excl. ops lock is locked. Three possibilities:
4013 * (1) some other op is running
4014 * (2) balance is running
4015 * (3) balance is paused -- special case (think resume)
4017 mutex_lock(&fs_info->balance_mutex);
4018 if (fs_info->balance_ctl) {
4019 /* this is either (2) or (3) */
4020 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4021 mutex_unlock(&fs_info->balance_mutex);
4023 * Lock released to allow other waiters to continue,
4024 * we'll reexamine the status again.
4026 mutex_lock(&fs_info->balance_mutex);
4028 if (fs_info->balance_ctl &&
4029 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4031 need_unlock = false;
4035 mutex_unlock(&fs_info->balance_mutex);
4039 mutex_unlock(&fs_info->balance_mutex);
4045 mutex_unlock(&fs_info->balance_mutex);
4046 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4053 bargs = memdup_user(arg, sizeof(*bargs));
4054 if (IS_ERR(bargs)) {
4055 ret = PTR_ERR(bargs);
4059 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4060 if (!fs_info->balance_ctl) {
4065 bctl = fs_info->balance_ctl;
4066 spin_lock(&fs_info->balance_lock);
4067 bctl->flags |= BTRFS_BALANCE_RESUME;
4068 spin_unlock(&fs_info->balance_lock);
4076 if (fs_info->balance_ctl) {
4081 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4088 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4089 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4090 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4092 bctl->flags = bargs->flags;
4094 /* balance everything - no filters */
4095 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4098 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4105 * Ownership of bctl and exclusive operation goes to btrfs_balance.
4106 * bctl is freed in reset_balance_state, or, if restriper was paused
4107 * all the way until unmount, in free_fs_info. The flag should be
4108 * cleared after reset_balance_state.
4110 need_unlock = false;
4112 ret = btrfs_balance(fs_info, bctl, bargs);
4115 if ((ret == 0 || ret == -ECANCELED) && arg) {
4116 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4125 mutex_unlock(&fs_info->balance_mutex);
4127 btrfs_exclop_finish(fs_info);
4129 mnt_drop_write_file(file);
4133 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4135 if (!capable(CAP_SYS_ADMIN))
4139 case BTRFS_BALANCE_CTL_PAUSE:
4140 return btrfs_pause_balance(fs_info);
4141 case BTRFS_BALANCE_CTL_CANCEL:
4142 return btrfs_cancel_balance(fs_info);
4148 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4151 struct btrfs_ioctl_balance_args *bargs;
4154 if (!capable(CAP_SYS_ADMIN))
4157 mutex_lock(&fs_info->balance_mutex);
4158 if (!fs_info->balance_ctl) {
4163 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4169 btrfs_update_ioctl_balance_args(fs_info, bargs);
4171 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4176 mutex_unlock(&fs_info->balance_mutex);
4180 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4182 struct inode *inode = file_inode(file);
4183 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4184 struct btrfs_ioctl_quota_ctl_args *sa;
4187 if (!capable(CAP_SYS_ADMIN))
4190 ret = mnt_want_write_file(file);
4194 sa = memdup_user(arg, sizeof(*sa));
4200 down_write(&fs_info->subvol_sem);
4203 case BTRFS_QUOTA_CTL_ENABLE:
4204 ret = btrfs_quota_enable(fs_info);
4206 case BTRFS_QUOTA_CTL_DISABLE:
4207 ret = btrfs_quota_disable(fs_info);
4215 up_write(&fs_info->subvol_sem);
4217 mnt_drop_write_file(file);
4221 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4223 struct inode *inode = file_inode(file);
4224 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4225 struct btrfs_root *root = BTRFS_I(inode)->root;
4226 struct btrfs_ioctl_qgroup_assign_args *sa;
4227 struct btrfs_trans_handle *trans;
4231 if (!capable(CAP_SYS_ADMIN))
4234 ret = mnt_want_write_file(file);
4238 sa = memdup_user(arg, sizeof(*sa));
4244 trans = btrfs_join_transaction(root);
4245 if (IS_ERR(trans)) {
4246 ret = PTR_ERR(trans);
4251 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
4253 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
4256 /* update qgroup status and info */
4257 err = btrfs_run_qgroups(trans);
4259 btrfs_handle_fs_error(fs_info, err,
4260 "failed to update qgroup status and info");
4261 err = btrfs_end_transaction(trans);
4268 mnt_drop_write_file(file);
4272 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4274 struct inode *inode = file_inode(file);
4275 struct btrfs_root *root = BTRFS_I(inode)->root;
4276 struct btrfs_ioctl_qgroup_create_args *sa;
4277 struct btrfs_trans_handle *trans;
4281 if (!capable(CAP_SYS_ADMIN))
4284 ret = mnt_want_write_file(file);
4288 sa = memdup_user(arg, sizeof(*sa));
4294 if (!sa->qgroupid) {
4299 trans = btrfs_join_transaction(root);
4300 if (IS_ERR(trans)) {
4301 ret = PTR_ERR(trans);
4306 ret = btrfs_create_qgroup(trans, sa->qgroupid);
4308 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
4311 err = btrfs_end_transaction(trans);
4318 mnt_drop_write_file(file);
4322 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4324 struct inode *inode = file_inode(file);
4325 struct btrfs_root *root = BTRFS_I(inode)->root;
4326 struct btrfs_ioctl_qgroup_limit_args *sa;
4327 struct btrfs_trans_handle *trans;
4332 if (!capable(CAP_SYS_ADMIN))
4335 ret = mnt_want_write_file(file);
4339 sa = memdup_user(arg, sizeof(*sa));
4345 trans = btrfs_join_transaction(root);
4346 if (IS_ERR(trans)) {
4347 ret = PTR_ERR(trans);
4351 qgroupid = sa->qgroupid;
4353 /* take the current subvol as qgroup */
4354 qgroupid = root->root_key.objectid;
4357 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
4359 err = btrfs_end_transaction(trans);
4366 mnt_drop_write_file(file);
4370 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4372 struct inode *inode = file_inode(file);
4373 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4374 struct btrfs_ioctl_quota_rescan_args *qsa;
4377 if (!capable(CAP_SYS_ADMIN))
4380 ret = mnt_want_write_file(file);
4384 qsa = memdup_user(arg, sizeof(*qsa));
4395 ret = btrfs_qgroup_rescan(fs_info);
4400 mnt_drop_write_file(file);
4404 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4407 struct btrfs_ioctl_quota_rescan_args *qsa;
4410 if (!capable(CAP_SYS_ADMIN))
4413 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
4417 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4419 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
4422 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4429 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4432 if (!capable(CAP_SYS_ADMIN))
4435 return btrfs_qgroup_wait_for_completion(fs_info, true);
4438 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4439 struct btrfs_ioctl_received_subvol_args *sa)
4441 struct inode *inode = file_inode(file);
4442 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4443 struct btrfs_root *root = BTRFS_I(inode)->root;
4444 struct btrfs_root_item *root_item = &root->root_item;
4445 struct btrfs_trans_handle *trans;
4446 struct timespec64 ct = current_time(inode);
4448 int received_uuid_changed;
4450 if (!inode_owner_or_capable(&init_user_ns, inode))
4453 ret = mnt_want_write_file(file);
4457 down_write(&fs_info->subvol_sem);
4459 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4464 if (btrfs_root_readonly(root)) {
4471 * 2 - uuid items (received uuid + subvol uuid)
4473 trans = btrfs_start_transaction(root, 3);
4474 if (IS_ERR(trans)) {
4475 ret = PTR_ERR(trans);
4480 sa->rtransid = trans->transid;
4481 sa->rtime.sec = ct.tv_sec;
4482 sa->rtime.nsec = ct.tv_nsec;
4484 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4486 if (received_uuid_changed &&
4487 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4488 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4489 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4490 root->root_key.objectid);
4491 if (ret && ret != -ENOENT) {
4492 btrfs_abort_transaction(trans, ret);
4493 btrfs_end_transaction(trans);
4497 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4498 btrfs_set_root_stransid(root_item, sa->stransid);
4499 btrfs_set_root_rtransid(root_item, sa->rtransid);
4500 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4501 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4502 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4503 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4505 ret = btrfs_update_root(trans, fs_info->tree_root,
4506 &root->root_key, &root->root_item);
4508 btrfs_end_transaction(trans);
4511 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4512 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4513 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4514 root->root_key.objectid);
4515 if (ret < 0 && ret != -EEXIST) {
4516 btrfs_abort_transaction(trans, ret);
4517 btrfs_end_transaction(trans);
4521 ret = btrfs_commit_transaction(trans);
4523 up_write(&fs_info->subvol_sem);
4524 mnt_drop_write_file(file);
4529 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4532 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4533 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4536 args32 = memdup_user(arg, sizeof(*args32));
4538 return PTR_ERR(args32);
4540 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4546 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4547 args64->stransid = args32->stransid;
4548 args64->rtransid = args32->rtransid;
4549 args64->stime.sec = args32->stime.sec;
4550 args64->stime.nsec = args32->stime.nsec;
4551 args64->rtime.sec = args32->rtime.sec;
4552 args64->rtime.nsec = args32->rtime.nsec;
4553 args64->flags = args32->flags;
4555 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4559 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4560 args32->stransid = args64->stransid;
4561 args32->rtransid = args64->rtransid;
4562 args32->stime.sec = args64->stime.sec;
4563 args32->stime.nsec = args64->stime.nsec;
4564 args32->rtime.sec = args64->rtime.sec;
4565 args32->rtime.nsec = args64->rtime.nsec;
4566 args32->flags = args64->flags;
4568 ret = copy_to_user(arg, args32, sizeof(*args32));
4579 static long btrfs_ioctl_set_received_subvol(struct file *file,
4582 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4585 sa = memdup_user(arg, sizeof(*sa));
4589 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4594 ret = copy_to_user(arg, sa, sizeof(*sa));
4603 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4608 char label[BTRFS_LABEL_SIZE];
4610 spin_lock(&fs_info->super_lock);
4611 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4612 spin_unlock(&fs_info->super_lock);
4614 len = strnlen(label, BTRFS_LABEL_SIZE);
4616 if (len == BTRFS_LABEL_SIZE) {
4618 "label is too long, return the first %zu bytes",
4622 ret = copy_to_user(arg, label, len);
4624 return ret ? -EFAULT : 0;
4627 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4629 struct inode *inode = file_inode(file);
4630 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4631 struct btrfs_root *root = BTRFS_I(inode)->root;
4632 struct btrfs_super_block *super_block = fs_info->super_copy;
4633 struct btrfs_trans_handle *trans;
4634 char label[BTRFS_LABEL_SIZE];
4637 if (!capable(CAP_SYS_ADMIN))
4640 if (copy_from_user(label, arg, sizeof(label)))
4643 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4645 "unable to set label with more than %d bytes",
4646 BTRFS_LABEL_SIZE - 1);
4650 ret = mnt_want_write_file(file);
4654 trans = btrfs_start_transaction(root, 0);
4655 if (IS_ERR(trans)) {
4656 ret = PTR_ERR(trans);
4660 spin_lock(&fs_info->super_lock);
4661 strcpy(super_block->label, label);
4662 spin_unlock(&fs_info->super_lock);
4663 ret = btrfs_commit_transaction(trans);
4666 mnt_drop_write_file(file);
4670 #define INIT_FEATURE_FLAGS(suffix) \
4671 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4672 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4673 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4675 int btrfs_ioctl_get_supported_features(void __user *arg)
4677 static const struct btrfs_ioctl_feature_flags features[3] = {
4678 INIT_FEATURE_FLAGS(SUPP),
4679 INIT_FEATURE_FLAGS(SAFE_SET),
4680 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4683 if (copy_to_user(arg, &features, sizeof(features)))
4689 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4692 struct btrfs_super_block *super_block = fs_info->super_copy;
4693 struct btrfs_ioctl_feature_flags features;
4695 features.compat_flags = btrfs_super_compat_flags(super_block);
4696 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4697 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4699 if (copy_to_user(arg, &features, sizeof(features)))
4705 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4706 enum btrfs_feature_set set,
4707 u64 change_mask, u64 flags, u64 supported_flags,
4708 u64 safe_set, u64 safe_clear)
4710 const char *type = btrfs_feature_set_name(set);
4712 u64 disallowed, unsupported;
4713 u64 set_mask = flags & change_mask;
4714 u64 clear_mask = ~flags & change_mask;
4716 unsupported = set_mask & ~supported_flags;
4718 names = btrfs_printable_features(set, unsupported);
4721 "this kernel does not support the %s feature bit%s",
4722 names, strchr(names, ',') ? "s" : "");
4726 "this kernel does not support %s bits 0x%llx",
4731 disallowed = set_mask & ~safe_set;
4733 names = btrfs_printable_features(set, disallowed);
4736 "can't set the %s feature bit%s while mounted",
4737 names, strchr(names, ',') ? "s" : "");
4741 "can't set %s bits 0x%llx while mounted",
4746 disallowed = clear_mask & ~safe_clear;
4748 names = btrfs_printable_features(set, disallowed);
4751 "can't clear the %s feature bit%s while mounted",
4752 names, strchr(names, ',') ? "s" : "");
4756 "can't clear %s bits 0x%llx while mounted",
4764 #define check_feature(fs_info, change_mask, flags, mask_base) \
4765 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
4766 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4767 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4768 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4770 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4772 struct inode *inode = file_inode(file);
4773 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4774 struct btrfs_root *root = BTRFS_I(inode)->root;
4775 struct btrfs_super_block *super_block = fs_info->super_copy;
4776 struct btrfs_ioctl_feature_flags flags[2];
4777 struct btrfs_trans_handle *trans;
4781 if (!capable(CAP_SYS_ADMIN))
4784 if (copy_from_user(flags, arg, sizeof(flags)))
4788 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4789 !flags[0].incompat_flags)
4792 ret = check_feature(fs_info, flags[0].compat_flags,
4793 flags[1].compat_flags, COMPAT);
4797 ret = check_feature(fs_info, flags[0].compat_ro_flags,
4798 flags[1].compat_ro_flags, COMPAT_RO);
4802 ret = check_feature(fs_info, flags[0].incompat_flags,
4803 flags[1].incompat_flags, INCOMPAT);
4807 ret = mnt_want_write_file(file);
4811 trans = btrfs_start_transaction(root, 0);
4812 if (IS_ERR(trans)) {
4813 ret = PTR_ERR(trans);
4814 goto out_drop_write;
4817 spin_lock(&fs_info->super_lock);
4818 newflags = btrfs_super_compat_flags(super_block);
4819 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4820 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4821 btrfs_set_super_compat_flags(super_block, newflags);
4823 newflags = btrfs_super_compat_ro_flags(super_block);
4824 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4825 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4826 btrfs_set_super_compat_ro_flags(super_block, newflags);
4828 newflags = btrfs_super_incompat_flags(super_block);
4829 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4830 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4831 btrfs_set_super_incompat_flags(super_block, newflags);
4832 spin_unlock(&fs_info->super_lock);
4834 ret = btrfs_commit_transaction(trans);
4836 mnt_drop_write_file(file);
4841 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
4843 struct btrfs_ioctl_send_args *arg;
4847 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4848 struct btrfs_ioctl_send_args_32 args32;
4850 ret = copy_from_user(&args32, argp, sizeof(args32));
4853 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4856 arg->send_fd = args32.send_fd;
4857 arg->clone_sources_count = args32.clone_sources_count;
4858 arg->clone_sources = compat_ptr(args32.clone_sources);
4859 arg->parent_root = args32.parent_root;
4860 arg->flags = args32.flags;
4861 memcpy(arg->reserved, args32.reserved,
4862 sizeof(args32.reserved));
4867 arg = memdup_user(argp, sizeof(*arg));
4869 return PTR_ERR(arg);
4871 ret = btrfs_ioctl_send(file, arg);
4876 long btrfs_ioctl(struct file *file, unsigned int
4877 cmd, unsigned long arg)
4879 struct inode *inode = file_inode(file);
4880 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4881 struct btrfs_root *root = BTRFS_I(inode)->root;
4882 void __user *argp = (void __user *)arg;
4885 case FS_IOC_GETVERSION:
4886 return btrfs_ioctl_getversion(file, argp);
4887 case FS_IOC_GETFSLABEL:
4888 return btrfs_ioctl_get_fslabel(fs_info, argp);
4889 case FS_IOC_SETFSLABEL:
4890 return btrfs_ioctl_set_fslabel(file, argp);
4892 return btrfs_ioctl_fitrim(fs_info, argp);
4893 case BTRFS_IOC_SNAP_CREATE:
4894 return btrfs_ioctl_snap_create(file, argp, 0);
4895 case BTRFS_IOC_SNAP_CREATE_V2:
4896 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4897 case BTRFS_IOC_SUBVOL_CREATE:
4898 return btrfs_ioctl_snap_create(file, argp, 1);
4899 case BTRFS_IOC_SUBVOL_CREATE_V2:
4900 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4901 case BTRFS_IOC_SNAP_DESTROY:
4902 return btrfs_ioctl_snap_destroy(file, argp, false);
4903 case BTRFS_IOC_SNAP_DESTROY_V2:
4904 return btrfs_ioctl_snap_destroy(file, argp, true);
4905 case BTRFS_IOC_SUBVOL_GETFLAGS:
4906 return btrfs_ioctl_subvol_getflags(file, argp);
4907 case BTRFS_IOC_SUBVOL_SETFLAGS:
4908 return btrfs_ioctl_subvol_setflags(file, argp);
4909 case BTRFS_IOC_DEFAULT_SUBVOL:
4910 return btrfs_ioctl_default_subvol(file, argp);
4911 case BTRFS_IOC_DEFRAG:
4912 return btrfs_ioctl_defrag(file, NULL);
4913 case BTRFS_IOC_DEFRAG_RANGE:
4914 return btrfs_ioctl_defrag(file, argp);
4915 case BTRFS_IOC_RESIZE:
4916 return btrfs_ioctl_resize(file, argp);
4917 case BTRFS_IOC_ADD_DEV:
4918 return btrfs_ioctl_add_dev(fs_info, argp);
4919 case BTRFS_IOC_RM_DEV:
4920 return btrfs_ioctl_rm_dev(file, argp);
4921 case BTRFS_IOC_RM_DEV_V2:
4922 return btrfs_ioctl_rm_dev_v2(file, argp);
4923 case BTRFS_IOC_FS_INFO:
4924 return btrfs_ioctl_fs_info(fs_info, argp);
4925 case BTRFS_IOC_DEV_INFO:
4926 return btrfs_ioctl_dev_info(fs_info, argp);
4927 case BTRFS_IOC_BALANCE:
4928 return btrfs_ioctl_balance(file, NULL);
4929 case BTRFS_IOC_TREE_SEARCH:
4930 return btrfs_ioctl_tree_search(file, argp);
4931 case BTRFS_IOC_TREE_SEARCH_V2:
4932 return btrfs_ioctl_tree_search_v2(file, argp);
4933 case BTRFS_IOC_INO_LOOKUP:
4934 return btrfs_ioctl_ino_lookup(file, argp);
4935 case BTRFS_IOC_INO_PATHS:
4936 return btrfs_ioctl_ino_to_path(root, argp);
4937 case BTRFS_IOC_LOGICAL_INO:
4938 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4939 case BTRFS_IOC_LOGICAL_INO_V2:
4940 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4941 case BTRFS_IOC_SPACE_INFO:
4942 return btrfs_ioctl_space_info(fs_info, argp);
4943 case BTRFS_IOC_SYNC: {
4946 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4949 ret = btrfs_sync_fs(inode->i_sb, 1);
4951 * The transaction thread may want to do more work,
4952 * namely it pokes the cleaner kthread that will start
4953 * processing uncleaned subvols.
4955 wake_up_process(fs_info->transaction_kthread);
4958 case BTRFS_IOC_START_SYNC:
4959 return btrfs_ioctl_start_sync(root, argp);
4960 case BTRFS_IOC_WAIT_SYNC:
4961 return btrfs_ioctl_wait_sync(fs_info, argp);
4962 case BTRFS_IOC_SCRUB:
4963 return btrfs_ioctl_scrub(file, argp);
4964 case BTRFS_IOC_SCRUB_CANCEL:
4965 return btrfs_ioctl_scrub_cancel(fs_info);
4966 case BTRFS_IOC_SCRUB_PROGRESS:
4967 return btrfs_ioctl_scrub_progress(fs_info, argp);
4968 case BTRFS_IOC_BALANCE_V2:
4969 return btrfs_ioctl_balance(file, argp);
4970 case BTRFS_IOC_BALANCE_CTL:
4971 return btrfs_ioctl_balance_ctl(fs_info, arg);
4972 case BTRFS_IOC_BALANCE_PROGRESS:
4973 return btrfs_ioctl_balance_progress(fs_info, argp);
4974 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4975 return btrfs_ioctl_set_received_subvol(file, argp);
4977 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4978 return btrfs_ioctl_set_received_subvol_32(file, argp);
4980 case BTRFS_IOC_SEND:
4981 return _btrfs_ioctl_send(file, argp, false);
4982 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4983 case BTRFS_IOC_SEND_32:
4984 return _btrfs_ioctl_send(file, argp, true);
4986 case BTRFS_IOC_GET_DEV_STATS:
4987 return btrfs_ioctl_get_dev_stats(fs_info, argp);
4988 case BTRFS_IOC_QUOTA_CTL:
4989 return btrfs_ioctl_quota_ctl(file, argp);
4990 case BTRFS_IOC_QGROUP_ASSIGN:
4991 return btrfs_ioctl_qgroup_assign(file, argp);
4992 case BTRFS_IOC_QGROUP_CREATE:
4993 return btrfs_ioctl_qgroup_create(file, argp);
4994 case BTRFS_IOC_QGROUP_LIMIT:
4995 return btrfs_ioctl_qgroup_limit(file, argp);
4996 case BTRFS_IOC_QUOTA_RESCAN:
4997 return btrfs_ioctl_quota_rescan(file, argp);
4998 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4999 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
5000 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5001 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
5002 case BTRFS_IOC_DEV_REPLACE:
5003 return btrfs_ioctl_dev_replace(fs_info, argp);
5004 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5005 return btrfs_ioctl_get_supported_features(argp);
5006 case BTRFS_IOC_GET_FEATURES:
5007 return btrfs_ioctl_get_features(fs_info, argp);
5008 case BTRFS_IOC_SET_FEATURES:
5009 return btrfs_ioctl_set_features(file, argp);
5010 case BTRFS_IOC_GET_SUBVOL_INFO:
5011 return btrfs_ioctl_get_subvol_info(file, argp);
5012 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5013 return btrfs_ioctl_get_subvol_rootref(file, argp);
5014 case BTRFS_IOC_INO_LOOKUP_USER:
5015 return btrfs_ioctl_ino_lookup_user(file, argp);
5021 #ifdef CONFIG_COMPAT
5022 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5025 * These all access 32-bit values anyway so no further
5026 * handling is necessary.
5029 case FS_IOC32_GETVERSION:
5030 cmd = FS_IOC_GETVERSION;
5034 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));