2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/compat.h>
37 #include <linux/bit_spinlock.h>
38 #include <linux/security.h>
39 #include <linux/xattr.h>
41 #include <linux/slab.h>
42 #include <linux/blkdev.h>
43 #include <linux/uuid.h>
44 #include <linux/btrfs.h>
45 #include <linux/uaccess.h>
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "print-tree.h"
53 #include "inode-map.h"
55 #include "rcu-string.h"
57 #include "dev-replace.h"
62 #include "compression.h"
65 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
66 * structures are incorrect, as the timespec structure from userspace
67 * is 4 bytes too small. We define these alternatives here to teach
68 * the kernel about the 32-bit struct packing.
70 struct btrfs_ioctl_timespec_32 {
73 } __attribute__ ((__packed__));
75 struct btrfs_ioctl_received_subvol_args_32 {
76 char uuid[BTRFS_UUID_SIZE]; /* in */
77 __u64 stransid; /* in */
78 __u64 rtransid; /* out */
79 struct btrfs_ioctl_timespec_32 stime; /* in */
80 struct btrfs_ioctl_timespec_32 rtime; /* out */
82 __u64 reserved[16]; /* in */
83 } __attribute__ ((__packed__));
85 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
86 struct btrfs_ioctl_received_subvol_args_32)
90 static int btrfs_clone(struct inode *src, struct inode *inode,
91 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
94 /* Mask out flags that are inappropriate for the given type of inode. */
95 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
99 else if (S_ISREG(mode))
100 return flags & ~FS_DIRSYNC_FL;
102 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
106 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
108 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
110 unsigned int iflags = 0;
112 if (flags & BTRFS_INODE_SYNC)
113 iflags |= FS_SYNC_FL;
114 if (flags & BTRFS_INODE_IMMUTABLE)
115 iflags |= FS_IMMUTABLE_FL;
116 if (flags & BTRFS_INODE_APPEND)
117 iflags |= FS_APPEND_FL;
118 if (flags & BTRFS_INODE_NODUMP)
119 iflags |= FS_NODUMP_FL;
120 if (flags & BTRFS_INODE_NOATIME)
121 iflags |= FS_NOATIME_FL;
122 if (flags & BTRFS_INODE_DIRSYNC)
123 iflags |= FS_DIRSYNC_FL;
124 if (flags & BTRFS_INODE_NODATACOW)
125 iflags |= FS_NOCOW_FL;
127 if (flags & BTRFS_INODE_NOCOMPRESS)
128 iflags |= FS_NOCOMP_FL;
129 else if (flags & BTRFS_INODE_COMPRESS)
130 iflags |= FS_COMPR_FL;
136 * Update inode->i_flags based on the btrfs internal flags.
138 void btrfs_update_iflags(struct inode *inode)
140 struct btrfs_inode *ip = BTRFS_I(inode);
141 unsigned int new_fl = 0;
143 if (ip->flags & BTRFS_INODE_SYNC)
145 if (ip->flags & BTRFS_INODE_IMMUTABLE)
146 new_fl |= S_IMMUTABLE;
147 if (ip->flags & BTRFS_INODE_APPEND)
149 if (ip->flags & BTRFS_INODE_NOATIME)
151 if (ip->flags & BTRFS_INODE_DIRSYNC)
154 set_mask_bits(&inode->i_flags,
155 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
159 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
161 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
162 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
164 if (copy_to_user(arg, &flags, sizeof(flags)))
169 static int check_flags(unsigned int flags)
171 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
172 FS_NOATIME_FL | FS_NODUMP_FL | \
173 FS_SYNC_FL | FS_DIRSYNC_FL | \
174 FS_NOCOMP_FL | FS_COMPR_FL |
178 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
184 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
186 struct inode *inode = file_inode(file);
187 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
188 struct btrfs_inode *ip = BTRFS_I(inode);
189 struct btrfs_root *root = ip->root;
190 struct btrfs_trans_handle *trans;
191 unsigned int flags, oldflags;
194 unsigned int i_oldflags;
197 if (!inode_owner_or_capable(inode))
200 if (btrfs_root_readonly(root))
203 if (copy_from_user(&flags, arg, sizeof(flags)))
206 ret = check_flags(flags);
210 ret = mnt_want_write_file(file);
216 ip_oldflags = ip->flags;
217 i_oldflags = inode->i_flags;
218 mode = inode->i_mode;
220 flags = btrfs_mask_flags(inode->i_mode, flags);
221 oldflags = btrfs_flags_to_ioctl(ip->flags);
222 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
223 if (!capable(CAP_LINUX_IMMUTABLE)) {
229 if (flags & FS_SYNC_FL)
230 ip->flags |= BTRFS_INODE_SYNC;
232 ip->flags &= ~BTRFS_INODE_SYNC;
233 if (flags & FS_IMMUTABLE_FL)
234 ip->flags |= BTRFS_INODE_IMMUTABLE;
236 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
237 if (flags & FS_APPEND_FL)
238 ip->flags |= BTRFS_INODE_APPEND;
240 ip->flags &= ~BTRFS_INODE_APPEND;
241 if (flags & FS_NODUMP_FL)
242 ip->flags |= BTRFS_INODE_NODUMP;
244 ip->flags &= ~BTRFS_INODE_NODUMP;
245 if (flags & FS_NOATIME_FL)
246 ip->flags |= BTRFS_INODE_NOATIME;
248 ip->flags &= ~BTRFS_INODE_NOATIME;
249 if (flags & FS_DIRSYNC_FL)
250 ip->flags |= BTRFS_INODE_DIRSYNC;
252 ip->flags &= ~BTRFS_INODE_DIRSYNC;
253 if (flags & FS_NOCOW_FL) {
256 * It's safe to turn csums off here, no extents exist.
257 * Otherwise we want the flag to reflect the real COW
258 * status of the file and will not set it.
260 if (inode->i_size == 0)
261 ip->flags |= BTRFS_INODE_NODATACOW
262 | BTRFS_INODE_NODATASUM;
264 ip->flags |= BTRFS_INODE_NODATACOW;
268 * Revert back under same assumptions as above
271 if (inode->i_size == 0)
272 ip->flags &= ~(BTRFS_INODE_NODATACOW
273 | BTRFS_INODE_NODATASUM);
275 ip->flags &= ~BTRFS_INODE_NODATACOW;
280 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
281 * flag may be changed automatically if compression code won't make
284 if (flags & FS_NOCOMP_FL) {
285 ip->flags &= ~BTRFS_INODE_COMPRESS;
286 ip->flags |= BTRFS_INODE_NOCOMPRESS;
288 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
289 if (ret && ret != -ENODATA)
291 } else if (flags & FS_COMPR_FL) {
294 ip->flags |= BTRFS_INODE_COMPRESS;
295 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
297 if (fs_info->compress_type == BTRFS_COMPRESS_LZO)
299 else if (fs_info->compress_type == BTRFS_COMPRESS_ZLIB)
303 ret = btrfs_set_prop(inode, "btrfs.compression",
304 comp, strlen(comp), 0);
309 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
310 if (ret && ret != -ENODATA)
312 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
315 trans = btrfs_start_transaction(root, 1);
317 ret = PTR_ERR(trans);
321 btrfs_update_iflags(inode);
322 inode_inc_iversion(inode);
323 inode->i_ctime = current_time(inode);
324 ret = btrfs_update_inode(trans, root, inode);
326 btrfs_end_transaction(trans);
329 ip->flags = ip_oldflags;
330 inode->i_flags = i_oldflags;
335 mnt_drop_write_file(file);
339 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
341 struct inode *inode = file_inode(file);
343 return put_user(inode->i_generation, arg);
346 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
348 struct inode *inode = file_inode(file);
349 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
350 struct btrfs_device *device;
351 struct request_queue *q;
352 struct fstrim_range range;
353 u64 minlen = ULLONG_MAX;
355 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
358 if (!capable(CAP_SYS_ADMIN))
362 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
366 q = bdev_get_queue(device->bdev);
367 if (blk_queue_discard(q)) {
369 minlen = min_t(u64, q->limits.discard_granularity,
377 if (copy_from_user(&range, arg, sizeof(range)))
379 if (range.start > total_bytes ||
380 range.len < fs_info->sb->s_blocksize)
383 range.len = min(range.len, total_bytes - range.start);
384 range.minlen = max(range.minlen, minlen);
385 ret = btrfs_trim_fs(fs_info, &range);
389 if (copy_to_user(arg, &range, sizeof(range)))
395 int btrfs_is_empty_uuid(u8 *uuid)
399 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
406 static noinline int create_subvol(struct inode *dir,
407 struct dentry *dentry,
408 const char *name, int namelen,
410 struct btrfs_qgroup_inherit *inherit)
412 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
413 struct btrfs_trans_handle *trans;
414 struct btrfs_key key;
415 struct btrfs_root_item *root_item;
416 struct btrfs_inode_item *inode_item;
417 struct extent_buffer *leaf;
418 struct btrfs_root *root = BTRFS_I(dir)->root;
419 struct btrfs_root *new_root;
420 struct btrfs_block_rsv block_rsv;
421 struct timespec cur_time = current_time(dir);
426 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
431 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
435 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
440 * Don't create subvolume whose level is not zero. Or qgroup will be
441 * screwed up since it assumes subvolume qgroup's level to be 0.
443 if (btrfs_qgroup_level(objectid)) {
448 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
450 * The same as the snapshot creation, please see the comment
451 * of create_snapshot().
453 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
454 8, &qgroup_reserved, false);
458 trans = btrfs_start_transaction(root, 0);
460 ret = PTR_ERR(trans);
461 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
464 trans->block_rsv = &block_rsv;
465 trans->bytes_reserved = block_rsv.size;
467 ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit);
471 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
477 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
478 btrfs_set_header_bytenr(leaf, leaf->start);
479 btrfs_set_header_generation(leaf, trans->transid);
480 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
481 btrfs_set_header_owner(leaf, objectid);
483 write_extent_buffer_fsid(leaf, fs_info->fsid);
484 write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid);
485 btrfs_mark_buffer_dirty(leaf);
487 inode_item = &root_item->inode;
488 btrfs_set_stack_inode_generation(inode_item, 1);
489 btrfs_set_stack_inode_size(inode_item, 3);
490 btrfs_set_stack_inode_nlink(inode_item, 1);
491 btrfs_set_stack_inode_nbytes(inode_item,
493 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
495 btrfs_set_root_flags(root_item, 0);
496 btrfs_set_root_limit(root_item, 0);
497 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
499 btrfs_set_root_bytenr(root_item, leaf->start);
500 btrfs_set_root_generation(root_item, trans->transid);
501 btrfs_set_root_level(root_item, 0);
502 btrfs_set_root_refs(root_item, 1);
503 btrfs_set_root_used(root_item, leaf->len);
504 btrfs_set_root_last_snapshot(root_item, 0);
506 btrfs_set_root_generation_v2(root_item,
507 btrfs_root_generation(root_item));
508 uuid_le_gen(&new_uuid);
509 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
510 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
511 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
512 root_item->ctime = root_item->otime;
513 btrfs_set_root_ctransid(root_item, trans->transid);
514 btrfs_set_root_otransid(root_item, trans->transid);
516 btrfs_tree_unlock(leaf);
517 free_extent_buffer(leaf);
520 btrfs_set_root_dirid(root_item, new_dirid);
522 key.objectid = objectid;
524 key.type = BTRFS_ROOT_ITEM_KEY;
525 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
530 key.offset = (u64)-1;
531 new_root = btrfs_read_fs_root_no_name(fs_info, &key);
532 if (IS_ERR(new_root)) {
533 ret = PTR_ERR(new_root);
534 btrfs_abort_transaction(trans, ret);
538 btrfs_record_root_in_trans(trans, new_root);
540 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
542 /* We potentially lose an unused inode item here */
543 btrfs_abort_transaction(trans, ret);
547 mutex_lock(&new_root->objectid_mutex);
548 new_root->highest_objectid = new_dirid;
549 mutex_unlock(&new_root->objectid_mutex);
552 * insert the directory item
554 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
556 btrfs_abort_transaction(trans, ret);
560 ret = btrfs_insert_dir_item(trans, root,
561 name, namelen, BTRFS_I(dir), &key,
562 BTRFS_FT_DIR, index);
564 btrfs_abort_transaction(trans, ret);
568 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
569 ret = btrfs_update_inode(trans, root, dir);
572 ret = btrfs_add_root_ref(trans, fs_info,
573 objectid, root->root_key.objectid,
574 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
577 ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid,
578 BTRFS_UUID_KEY_SUBVOL, objectid);
580 btrfs_abort_transaction(trans, ret);
584 trans->block_rsv = NULL;
585 trans->bytes_reserved = 0;
586 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
589 *async_transid = trans->transid;
590 err = btrfs_commit_transaction_async(trans, 1);
592 err = btrfs_commit_transaction(trans);
594 err = btrfs_commit_transaction(trans);
600 inode = btrfs_lookup_dentry(dir, dentry);
602 return PTR_ERR(inode);
603 d_instantiate(dentry, inode);
612 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
613 struct dentry *dentry,
614 u64 *async_transid, bool readonly,
615 struct btrfs_qgroup_inherit *inherit)
617 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
619 struct btrfs_pending_snapshot *pending_snapshot;
620 struct btrfs_trans_handle *trans;
623 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
626 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
627 if (!pending_snapshot)
630 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
632 pending_snapshot->path = btrfs_alloc_path();
633 if (!pending_snapshot->root_item || !pending_snapshot->path) {
638 atomic_inc(&root->will_be_snapshotted);
639 smp_mb__after_atomic();
640 /* wait for no snapshot writes */
641 wait_event(root->subv_writers->wait,
642 percpu_counter_sum(&root->subv_writers->counter) == 0);
644 ret = btrfs_start_delalloc_inodes(root, 0);
648 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
650 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
651 BTRFS_BLOCK_RSV_TEMP);
653 * 1 - parent dir inode
656 * 2 - root ref/backref
657 * 1 - root of snapshot
660 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
661 &pending_snapshot->block_rsv, 8,
662 &pending_snapshot->qgroup_reserved,
667 pending_snapshot->dentry = dentry;
668 pending_snapshot->root = root;
669 pending_snapshot->readonly = readonly;
670 pending_snapshot->dir = dir;
671 pending_snapshot->inherit = inherit;
673 trans = btrfs_start_transaction(root, 0);
675 ret = PTR_ERR(trans);
679 spin_lock(&fs_info->trans_lock);
680 list_add(&pending_snapshot->list,
681 &trans->transaction->pending_snapshots);
682 spin_unlock(&fs_info->trans_lock);
684 *async_transid = trans->transid;
685 ret = btrfs_commit_transaction_async(trans, 1);
687 ret = btrfs_commit_transaction(trans);
689 ret = btrfs_commit_transaction(trans);
694 ret = pending_snapshot->error;
698 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
702 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
704 ret = PTR_ERR(inode);
708 d_instantiate(dentry, inode);
711 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
713 if (atomic_dec_and_test(&root->will_be_snapshotted))
714 wake_up_atomic_t(&root->will_be_snapshotted);
716 kfree(pending_snapshot->root_item);
717 btrfs_free_path(pending_snapshot->path);
718 kfree(pending_snapshot);
723 /* copy of may_delete in fs/namei.c()
724 * Check whether we can remove a link victim from directory dir, check
725 * whether the type of victim is right.
726 * 1. We can't do it if dir is read-only (done in permission())
727 * 2. We should have write and exec permissions on dir
728 * 3. We can't remove anything from append-only dir
729 * 4. We can't do anything with immutable dir (done in permission())
730 * 5. If the sticky bit on dir is set we should either
731 * a. be owner of dir, or
732 * b. be owner of victim, or
733 * c. have CAP_FOWNER capability
734 * 6. If the victim is append-only or immutable we can't do anything with
735 * links pointing to it.
736 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
737 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
738 * 9. We can't remove a root or mountpoint.
739 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
740 * nfs_async_unlink().
743 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
747 if (d_really_is_negative(victim))
750 BUG_ON(d_inode(victim->d_parent) != dir);
751 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
753 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
758 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
759 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
762 if (!d_is_dir(victim))
766 } else if (d_is_dir(victim))
770 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
775 /* copy of may_create in fs/namei.c() */
776 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
778 if (d_really_is_positive(child))
782 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
786 * Create a new subvolume below @parent. This is largely modeled after
787 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
788 * inside this filesystem so it's quite a bit simpler.
790 static noinline int btrfs_mksubvol(const struct path *parent,
791 const char *name, int namelen,
792 struct btrfs_root *snap_src,
793 u64 *async_transid, bool readonly,
794 struct btrfs_qgroup_inherit *inherit)
796 struct inode *dir = d_inode(parent->dentry);
797 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
798 struct dentry *dentry;
801 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
805 dentry = lookup_one_len(name, parent->dentry, namelen);
806 error = PTR_ERR(dentry);
810 error = btrfs_may_create(dir, dentry);
815 * even if this name doesn't exist, we may get hash collisions.
816 * check for them now when we can safely fail
818 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
824 down_read(&fs_info->subvol_sem);
826 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
830 error = create_snapshot(snap_src, dir, dentry,
831 async_transid, readonly, inherit);
833 error = create_subvol(dir, dentry, name, namelen,
834 async_transid, inherit);
837 fsnotify_mkdir(dir, dentry);
839 up_read(&fs_info->subvol_sem);
848 * When we're defragging a range, we don't want to kick it off again
849 * if it is really just waiting for delalloc to send it down.
850 * If we find a nice big extent or delalloc range for the bytes in the
851 * file you want to defrag, we return 0 to let you know to skip this
854 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
856 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
857 struct extent_map *em = NULL;
858 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
861 read_lock(&em_tree->lock);
862 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
863 read_unlock(&em_tree->lock);
866 end = extent_map_end(em);
868 if (end - offset > thresh)
871 /* if we already have a nice delalloc here, just stop */
873 end = count_range_bits(io_tree, &offset, offset + thresh,
874 thresh, EXTENT_DELALLOC, 1);
881 * helper function to walk through a file and find extents
882 * newer than a specific transid, and smaller than thresh.
884 * This is used by the defragging code to find new and small
887 static int find_new_extents(struct btrfs_root *root,
888 struct inode *inode, u64 newer_than,
889 u64 *off, u32 thresh)
891 struct btrfs_path *path;
892 struct btrfs_key min_key;
893 struct extent_buffer *leaf;
894 struct btrfs_file_extent_item *extent;
897 u64 ino = btrfs_ino(BTRFS_I(inode));
899 path = btrfs_alloc_path();
903 min_key.objectid = ino;
904 min_key.type = BTRFS_EXTENT_DATA_KEY;
905 min_key.offset = *off;
908 ret = btrfs_search_forward(root, &min_key, path, newer_than);
912 if (min_key.objectid != ino)
914 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
917 leaf = path->nodes[0];
918 extent = btrfs_item_ptr(leaf, path->slots[0],
919 struct btrfs_file_extent_item);
921 type = btrfs_file_extent_type(leaf, extent);
922 if (type == BTRFS_FILE_EXTENT_REG &&
923 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
924 check_defrag_in_cache(inode, min_key.offset, thresh)) {
925 *off = min_key.offset;
926 btrfs_free_path(path);
931 if (path->slots[0] < btrfs_header_nritems(leaf)) {
932 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
936 if (min_key.offset == (u64)-1)
940 btrfs_release_path(path);
943 btrfs_free_path(path);
947 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
949 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
950 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
951 struct extent_map *em;
955 * hopefully we have this extent in the tree already, try without
956 * the full extent lock
958 read_lock(&em_tree->lock);
959 em = lookup_extent_mapping(em_tree, start, len);
960 read_unlock(&em_tree->lock);
963 struct extent_state *cached = NULL;
964 u64 end = start + len - 1;
966 /* get the big lock and read metadata off disk */
967 lock_extent_bits(io_tree, start, end, &cached);
968 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0);
969 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
978 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
980 struct extent_map *next;
983 /* this is the last extent */
984 if (em->start + em->len >= i_size_read(inode))
987 next = defrag_lookup_extent(inode, em->start + em->len);
988 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
990 else if ((em->block_start + em->block_len == next->block_start) &&
991 (em->block_len > SZ_128K && next->block_len > SZ_128K))
994 free_extent_map(next);
998 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
999 u64 *last_len, u64 *skip, u64 *defrag_end,
1002 struct extent_map *em;
1004 bool next_mergeable = true;
1005 bool prev_mergeable = true;
1008 * make sure that once we start defragging an extent, we keep on
1011 if (start < *defrag_end)
1016 em = defrag_lookup_extent(inode, start);
1020 /* this will cover holes, and inline extents */
1021 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1027 prev_mergeable = false;
1029 next_mergeable = defrag_check_next_extent(inode, em);
1031 * we hit a real extent, if it is big or the next extent is not a
1032 * real extent, don't bother defragging it
1034 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1035 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1039 * last_len ends up being a counter of how many bytes we've defragged.
1040 * every time we choose not to defrag an extent, we reset *last_len
1041 * so that the next tiny extent will force a defrag.
1043 * The end result of this is that tiny extents before a single big
1044 * extent will force at least part of that big extent to be defragged.
1047 *defrag_end = extent_map_end(em);
1050 *skip = extent_map_end(em);
1054 free_extent_map(em);
1059 * it doesn't do much good to defrag one or two pages
1060 * at a time. This pulls in a nice chunk of pages
1061 * to COW and defrag.
1063 * It also makes sure the delalloc code has enough
1064 * dirty data to avoid making new small extents as part
1067 * It's a good idea to start RA on this range
1068 * before calling this.
1070 static int cluster_pages_for_defrag(struct inode *inode,
1071 struct page **pages,
1072 unsigned long start_index,
1073 unsigned long num_pages)
1075 unsigned long file_end;
1076 u64 isize = i_size_read(inode);
1083 struct btrfs_ordered_extent *ordered;
1084 struct extent_state *cached_state = NULL;
1085 struct extent_io_tree *tree;
1086 struct extent_changeset *data_reserved = NULL;
1087 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1089 file_end = (isize - 1) >> PAGE_SHIFT;
1090 if (!isize || start_index > file_end)
1093 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1095 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
1096 start_index << PAGE_SHIFT,
1097 page_cnt << PAGE_SHIFT);
1101 tree = &BTRFS_I(inode)->io_tree;
1103 /* step one, lock all the pages */
1104 for (i = 0; i < page_cnt; i++) {
1107 page = find_or_create_page(inode->i_mapping,
1108 start_index + i, mask);
1112 page_start = page_offset(page);
1113 page_end = page_start + PAGE_SIZE - 1;
1115 lock_extent_bits(tree, page_start, page_end,
1117 ordered = btrfs_lookup_ordered_extent(inode,
1119 unlock_extent_cached(tree, page_start, page_end,
1120 &cached_state, GFP_NOFS);
1125 btrfs_start_ordered_extent(inode, ordered, 1);
1126 btrfs_put_ordered_extent(ordered);
1129 * we unlocked the page above, so we need check if
1130 * it was released or not.
1132 if (page->mapping != inode->i_mapping) {
1139 if (!PageUptodate(page)) {
1140 btrfs_readpage(NULL, page);
1142 if (!PageUptodate(page)) {
1150 if (page->mapping != inode->i_mapping) {
1162 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1166 * so now we have a nice long stream of locked
1167 * and up to date pages, lets wait on them
1169 for (i = 0; i < i_done; i++)
1170 wait_on_page_writeback(pages[i]);
1172 page_start = page_offset(pages[0]);
1173 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1175 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1176 page_start, page_end - 1, &cached_state);
1177 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1178 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1179 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1180 &cached_state, GFP_NOFS);
1182 if (i_done != page_cnt) {
1183 spin_lock(&BTRFS_I(inode)->lock);
1184 BTRFS_I(inode)->outstanding_extents++;
1185 spin_unlock(&BTRFS_I(inode)->lock);
1186 btrfs_delalloc_release_space(inode, data_reserved,
1187 start_index << PAGE_SHIFT,
1188 (page_cnt - i_done) << PAGE_SHIFT);
1192 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1195 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1196 page_start, page_end - 1, &cached_state,
1199 for (i = 0; i < i_done; i++) {
1200 clear_page_dirty_for_io(pages[i]);
1201 ClearPageChecked(pages[i]);
1202 set_page_extent_mapped(pages[i]);
1203 set_page_dirty(pages[i]);
1204 unlock_page(pages[i]);
1207 extent_changeset_free(data_reserved);
1210 for (i = 0; i < i_done; i++) {
1211 unlock_page(pages[i]);
1214 btrfs_delalloc_release_space(inode, data_reserved,
1215 start_index << PAGE_SHIFT,
1216 page_cnt << PAGE_SHIFT);
1217 extent_changeset_free(data_reserved);
1222 int btrfs_defrag_file(struct inode *inode, struct file *file,
1223 struct btrfs_ioctl_defrag_range_args *range,
1224 u64 newer_than, unsigned long max_to_defrag)
1226 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1227 struct btrfs_root *root = BTRFS_I(inode)->root;
1228 struct file_ra_state *ra = NULL;
1229 unsigned long last_index;
1230 u64 isize = i_size_read(inode);
1234 u64 newer_off = range->start;
1236 unsigned long ra_index = 0;
1238 int defrag_count = 0;
1239 int compress_type = BTRFS_COMPRESS_ZLIB;
1240 u32 extent_thresh = range->extent_thresh;
1241 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1242 unsigned long cluster = max_cluster;
1243 u64 new_align = ~((u64)SZ_128K - 1);
1244 struct page **pages = NULL;
1245 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1250 if (range->start >= isize)
1254 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1256 if (range->compress_type)
1257 compress_type = range->compress_type;
1260 if (extent_thresh == 0)
1261 extent_thresh = SZ_256K;
1264 * If we were not given a file, allocate a readahead context. As
1265 * readahead is just an optimization, defrag will work without it so
1266 * we don't error out.
1269 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1271 file_ra_state_init(ra, inode->i_mapping);
1276 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1282 /* find the last page to defrag */
1283 if (range->start + range->len > range->start) {
1284 last_index = min_t(u64, isize - 1,
1285 range->start + range->len - 1) >> PAGE_SHIFT;
1287 last_index = (isize - 1) >> PAGE_SHIFT;
1291 ret = find_new_extents(root, inode, newer_than,
1292 &newer_off, SZ_64K);
1294 range->start = newer_off;
1296 * we always align our defrag to help keep
1297 * the extents in the file evenly spaced
1299 i = (newer_off & new_align) >> PAGE_SHIFT;
1303 i = range->start >> PAGE_SHIFT;
1306 max_to_defrag = last_index - i + 1;
1309 * make writeback starts from i, so the defrag range can be
1310 * written sequentially.
1312 if (i < inode->i_mapping->writeback_index)
1313 inode->i_mapping->writeback_index = i;
1315 while (i <= last_index && defrag_count < max_to_defrag &&
1316 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1318 * make sure we stop running if someone unmounts
1321 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1324 if (btrfs_defrag_cancelled(fs_info)) {
1325 btrfs_debug(fs_info, "defrag_file cancelled");
1330 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1331 extent_thresh, &last_len, &skip,
1332 &defrag_end, do_compress)){
1335 * the should_defrag function tells us how much to skip
1336 * bump our counter by the suggested amount
1338 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1339 i = max(i + 1, next);
1344 cluster = (PAGE_ALIGN(defrag_end) >>
1346 cluster = min(cluster, max_cluster);
1348 cluster = max_cluster;
1351 if (i + cluster > ra_index) {
1352 ra_index = max(i, ra_index);
1354 page_cache_sync_readahead(inode->i_mapping, ra,
1355 file, ra_index, cluster);
1356 ra_index += cluster;
1361 BTRFS_I(inode)->defrag_compress = compress_type;
1362 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1364 inode_unlock(inode);
1368 defrag_count += ret;
1369 balance_dirty_pages_ratelimited(inode->i_mapping);
1370 inode_unlock(inode);
1373 if (newer_off == (u64)-1)
1379 newer_off = max(newer_off + 1,
1380 (u64)i << PAGE_SHIFT);
1382 ret = find_new_extents(root, inode, newer_than,
1383 &newer_off, SZ_64K);
1385 range->start = newer_off;
1386 i = (newer_off & new_align) >> PAGE_SHIFT;
1393 last_len += ret << PAGE_SHIFT;
1401 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1402 filemap_flush(inode->i_mapping);
1403 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1404 &BTRFS_I(inode)->runtime_flags))
1405 filemap_flush(inode->i_mapping);
1409 /* the filemap_flush will queue IO into the worker threads, but
1410 * we have to make sure the IO is actually started and that
1411 * ordered extents get created before we return
1413 atomic_inc(&fs_info->async_submit_draining);
1414 while (atomic_read(&fs_info->nr_async_submits) ||
1415 atomic_read(&fs_info->async_delalloc_pages)) {
1416 wait_event(fs_info->async_submit_wait,
1417 (atomic_read(&fs_info->nr_async_submits) == 0 &&
1418 atomic_read(&fs_info->async_delalloc_pages) == 0));
1420 atomic_dec(&fs_info->async_submit_draining);
1423 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1424 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1425 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1426 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1434 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1435 inode_unlock(inode);
1443 static noinline int btrfs_ioctl_resize(struct file *file,
1446 struct inode *inode = file_inode(file);
1447 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1451 struct btrfs_root *root = BTRFS_I(inode)->root;
1452 struct btrfs_ioctl_vol_args *vol_args;
1453 struct btrfs_trans_handle *trans;
1454 struct btrfs_device *device = NULL;
1457 char *devstr = NULL;
1461 if (!capable(CAP_SYS_ADMIN))
1464 ret = mnt_want_write_file(file);
1468 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
1469 mnt_drop_write_file(file);
1470 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1473 mutex_lock(&fs_info->volume_mutex);
1474 vol_args = memdup_user(arg, sizeof(*vol_args));
1475 if (IS_ERR(vol_args)) {
1476 ret = PTR_ERR(vol_args);
1480 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1482 sizestr = vol_args->name;
1483 devstr = strchr(sizestr, ':');
1485 sizestr = devstr + 1;
1487 devstr = vol_args->name;
1488 ret = kstrtoull(devstr, 10, &devid);
1495 btrfs_info(fs_info, "resizing devid %llu", devid);
1498 device = btrfs_find_device(fs_info, devid, NULL, NULL);
1500 btrfs_info(fs_info, "resizer unable to find device %llu",
1506 if (!device->writeable) {
1508 "resizer unable to apply on readonly device %llu",
1514 if (!strcmp(sizestr, "max"))
1515 new_size = device->bdev->bd_inode->i_size;
1517 if (sizestr[0] == '-') {
1520 } else if (sizestr[0] == '+') {
1524 new_size = memparse(sizestr, &retptr);
1525 if (*retptr != '\0' || new_size == 0) {
1531 if (device->is_tgtdev_for_dev_replace) {
1536 old_size = btrfs_device_get_total_bytes(device);
1539 if (new_size > old_size) {
1543 new_size = old_size - new_size;
1544 } else if (mod > 0) {
1545 if (new_size > ULLONG_MAX - old_size) {
1549 new_size = old_size + new_size;
1552 if (new_size < SZ_256M) {
1556 if (new_size > device->bdev->bd_inode->i_size) {
1561 new_size = round_down(new_size, fs_info->sectorsize);
1563 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1564 rcu_str_deref(device->name), new_size);
1566 if (new_size > old_size) {
1567 trans = btrfs_start_transaction(root, 0);
1568 if (IS_ERR(trans)) {
1569 ret = PTR_ERR(trans);
1572 ret = btrfs_grow_device(trans, device, new_size);
1573 btrfs_commit_transaction(trans);
1574 } else if (new_size < old_size) {
1575 ret = btrfs_shrink_device(device, new_size);
1576 } /* equal, nothing need to do */
1581 mutex_unlock(&fs_info->volume_mutex);
1582 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
1583 mnt_drop_write_file(file);
1587 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1588 const char *name, unsigned long fd, int subvol,
1589 u64 *transid, bool readonly,
1590 struct btrfs_qgroup_inherit *inherit)
1595 if (!S_ISDIR(file_inode(file)->i_mode))
1598 ret = mnt_want_write_file(file);
1602 namelen = strlen(name);
1603 if (strchr(name, '/')) {
1605 goto out_drop_write;
1608 if (name[0] == '.' &&
1609 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1611 goto out_drop_write;
1615 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1616 NULL, transid, readonly, inherit);
1618 struct fd src = fdget(fd);
1619 struct inode *src_inode;
1622 goto out_drop_write;
1625 src_inode = file_inode(src.file);
1626 if (src_inode->i_sb != file_inode(file)->i_sb) {
1627 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1628 "Snapshot src from another FS");
1630 } else if (!inode_owner_or_capable(src_inode)) {
1632 * Subvolume creation is not restricted, but snapshots
1633 * are limited to own subvolumes only
1637 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1638 BTRFS_I(src_inode)->root,
1639 transid, readonly, inherit);
1644 mnt_drop_write_file(file);
1649 static noinline int btrfs_ioctl_snap_create(struct file *file,
1650 void __user *arg, int subvol)
1652 struct btrfs_ioctl_vol_args *vol_args;
1655 if (!S_ISDIR(file_inode(file)->i_mode))
1658 vol_args = memdup_user(arg, sizeof(*vol_args));
1659 if (IS_ERR(vol_args))
1660 return PTR_ERR(vol_args);
1661 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1663 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1664 vol_args->fd, subvol,
1671 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1672 void __user *arg, int subvol)
1674 struct btrfs_ioctl_vol_args_v2 *vol_args;
1678 bool readonly = false;
1679 struct btrfs_qgroup_inherit *inherit = NULL;
1681 if (!S_ISDIR(file_inode(file)->i_mode))
1684 vol_args = memdup_user(arg, sizeof(*vol_args));
1685 if (IS_ERR(vol_args))
1686 return PTR_ERR(vol_args);
1687 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1689 if (vol_args->flags &
1690 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1691 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1696 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1698 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1700 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1701 if (vol_args->size > PAGE_SIZE) {
1705 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1706 if (IS_ERR(inherit)) {
1707 ret = PTR_ERR(inherit);
1712 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1713 vol_args->fd, subvol, ptr,
1718 if (ptr && copy_to_user(arg +
1719 offsetof(struct btrfs_ioctl_vol_args_v2,
1731 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1734 struct inode *inode = file_inode(file);
1735 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1736 struct btrfs_root *root = BTRFS_I(inode)->root;
1740 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1743 down_read(&fs_info->subvol_sem);
1744 if (btrfs_root_readonly(root))
1745 flags |= BTRFS_SUBVOL_RDONLY;
1746 up_read(&fs_info->subvol_sem);
1748 if (copy_to_user(arg, &flags, sizeof(flags)))
1754 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1757 struct inode *inode = file_inode(file);
1758 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1759 struct btrfs_root *root = BTRFS_I(inode)->root;
1760 struct btrfs_trans_handle *trans;
1765 if (!inode_owner_or_capable(inode))
1768 ret = mnt_want_write_file(file);
1772 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1774 goto out_drop_write;
1777 if (copy_from_user(&flags, arg, sizeof(flags))) {
1779 goto out_drop_write;
1782 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1784 goto out_drop_write;
1787 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1789 goto out_drop_write;
1792 down_write(&fs_info->subvol_sem);
1795 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1798 root_flags = btrfs_root_flags(&root->root_item);
1799 if (flags & BTRFS_SUBVOL_RDONLY) {
1800 btrfs_set_root_flags(&root->root_item,
1801 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1804 * Block RO -> RW transition if this subvolume is involved in
1807 spin_lock(&root->root_item_lock);
1808 if (root->send_in_progress == 0) {
1809 btrfs_set_root_flags(&root->root_item,
1810 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1811 spin_unlock(&root->root_item_lock);
1813 spin_unlock(&root->root_item_lock);
1815 "Attempt to set subvolume %llu read-write during send",
1816 root->root_key.objectid);
1822 trans = btrfs_start_transaction(root, 1);
1823 if (IS_ERR(trans)) {
1824 ret = PTR_ERR(trans);
1828 ret = btrfs_update_root(trans, fs_info->tree_root,
1829 &root->root_key, &root->root_item);
1831 btrfs_end_transaction(trans);
1835 ret = btrfs_commit_transaction(trans);
1839 btrfs_set_root_flags(&root->root_item, root_flags);
1841 up_write(&fs_info->subvol_sem);
1843 mnt_drop_write_file(file);
1849 * helper to check if the subvolume references other subvolumes
1851 static noinline int may_destroy_subvol(struct btrfs_root *root)
1853 struct btrfs_fs_info *fs_info = root->fs_info;
1854 struct btrfs_path *path;
1855 struct btrfs_dir_item *di;
1856 struct btrfs_key key;
1860 path = btrfs_alloc_path();
1864 /* Make sure this root isn't set as the default subvol */
1865 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1866 di = btrfs_lookup_dir_item(NULL, fs_info->tree_root, path,
1867 dir_id, "default", 7, 0);
1868 if (di && !IS_ERR(di)) {
1869 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1870 if (key.objectid == root->root_key.objectid) {
1873 "deleting default subvolume %llu is not allowed",
1877 btrfs_release_path(path);
1880 key.objectid = root->root_key.objectid;
1881 key.type = BTRFS_ROOT_REF_KEY;
1882 key.offset = (u64)-1;
1884 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
1890 if (path->slots[0] > 0) {
1892 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1893 if (key.objectid == root->root_key.objectid &&
1894 key.type == BTRFS_ROOT_REF_KEY)
1898 btrfs_free_path(path);
1902 static noinline int key_in_sk(struct btrfs_key *key,
1903 struct btrfs_ioctl_search_key *sk)
1905 struct btrfs_key test;
1908 test.objectid = sk->min_objectid;
1909 test.type = sk->min_type;
1910 test.offset = sk->min_offset;
1912 ret = btrfs_comp_cpu_keys(key, &test);
1916 test.objectid = sk->max_objectid;
1917 test.type = sk->max_type;
1918 test.offset = sk->max_offset;
1920 ret = btrfs_comp_cpu_keys(key, &test);
1926 static noinline int copy_to_sk(struct btrfs_path *path,
1927 struct btrfs_key *key,
1928 struct btrfs_ioctl_search_key *sk,
1931 unsigned long *sk_offset,
1935 struct extent_buffer *leaf;
1936 struct btrfs_ioctl_search_header sh;
1937 struct btrfs_key test;
1938 unsigned long item_off;
1939 unsigned long item_len;
1945 leaf = path->nodes[0];
1946 slot = path->slots[0];
1947 nritems = btrfs_header_nritems(leaf);
1949 if (btrfs_header_generation(leaf) > sk->max_transid) {
1953 found_transid = btrfs_header_generation(leaf);
1955 for (i = slot; i < nritems; i++) {
1956 item_off = btrfs_item_ptr_offset(leaf, i);
1957 item_len = btrfs_item_size_nr(leaf, i);
1959 btrfs_item_key_to_cpu(leaf, key, i);
1960 if (!key_in_sk(key, sk))
1963 if (sizeof(sh) + item_len > *buf_size) {
1970 * return one empty item back for v1, which does not
1974 *buf_size = sizeof(sh) + item_len;
1979 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1984 sh.objectid = key->objectid;
1985 sh.offset = key->offset;
1986 sh.type = key->type;
1988 sh.transid = found_transid;
1990 /* copy search result header */
1991 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1996 *sk_offset += sizeof(sh);
1999 char __user *up = ubuf + *sk_offset;
2001 if (read_extent_buffer_to_user(leaf, up,
2002 item_off, item_len)) {
2007 *sk_offset += item_len;
2011 if (ret) /* -EOVERFLOW from above */
2014 if (*num_found >= sk->nr_items) {
2021 test.objectid = sk->max_objectid;
2022 test.type = sk->max_type;
2023 test.offset = sk->max_offset;
2024 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2026 else if (key->offset < (u64)-1)
2028 else if (key->type < (u8)-1) {
2031 } else if (key->objectid < (u64)-1) {
2039 * 0: all items from this leaf copied, continue with next
2040 * 1: * more items can be copied, but unused buffer is too small
2041 * * all items were found
2042 * Either way, it will stops the loop which iterates to the next
2044 * -EOVERFLOW: item was to large for buffer
2045 * -EFAULT: could not copy extent buffer back to userspace
2050 static noinline int search_ioctl(struct inode *inode,
2051 struct btrfs_ioctl_search_key *sk,
2055 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2056 struct btrfs_root *root;
2057 struct btrfs_key key;
2058 struct btrfs_path *path;
2061 unsigned long sk_offset = 0;
2063 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2064 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2068 path = btrfs_alloc_path();
2072 if (sk->tree_id == 0) {
2073 /* search the root of the inode that was passed */
2074 root = BTRFS_I(inode)->root;
2076 key.objectid = sk->tree_id;
2077 key.type = BTRFS_ROOT_ITEM_KEY;
2078 key.offset = (u64)-1;
2079 root = btrfs_read_fs_root_no_name(info, &key);
2081 btrfs_free_path(path);
2086 key.objectid = sk->min_objectid;
2087 key.type = sk->min_type;
2088 key.offset = sk->min_offset;
2091 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2097 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2098 &sk_offset, &num_found);
2099 btrfs_release_path(path);
2107 sk->nr_items = num_found;
2108 btrfs_free_path(path);
2112 static noinline int btrfs_ioctl_tree_search(struct file *file,
2115 struct btrfs_ioctl_search_args __user *uargs;
2116 struct btrfs_ioctl_search_key sk;
2117 struct inode *inode;
2121 if (!capable(CAP_SYS_ADMIN))
2124 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2126 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2129 buf_size = sizeof(uargs->buf);
2131 inode = file_inode(file);
2132 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2135 * In the origin implementation an overflow is handled by returning a
2136 * search header with a len of zero, so reset ret.
2138 if (ret == -EOVERFLOW)
2141 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2146 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2149 struct btrfs_ioctl_search_args_v2 __user *uarg;
2150 struct btrfs_ioctl_search_args_v2 args;
2151 struct inode *inode;
2154 const size_t buf_limit = SZ_16M;
2156 if (!capable(CAP_SYS_ADMIN))
2159 /* copy search header and buffer size */
2160 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2161 if (copy_from_user(&args, uarg, sizeof(args)))
2164 buf_size = args.buf_size;
2166 /* limit result size to 16MB */
2167 if (buf_size > buf_limit)
2168 buf_size = buf_limit;
2170 inode = file_inode(file);
2171 ret = search_ioctl(inode, &args.key, &buf_size,
2172 (char __user *)(&uarg->buf[0]));
2173 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2175 else if (ret == -EOVERFLOW &&
2176 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2183 * Search INODE_REFs to identify path name of 'dirid' directory
2184 * in a 'tree_id' tree. and sets path name to 'name'.
2186 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2187 u64 tree_id, u64 dirid, char *name)
2189 struct btrfs_root *root;
2190 struct btrfs_key key;
2196 struct btrfs_inode_ref *iref;
2197 struct extent_buffer *l;
2198 struct btrfs_path *path;
2200 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2205 path = btrfs_alloc_path();
2209 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2211 key.objectid = tree_id;
2212 key.type = BTRFS_ROOT_ITEM_KEY;
2213 key.offset = (u64)-1;
2214 root = btrfs_read_fs_root_no_name(info, &key);
2216 btrfs_err(info, "could not find root %llu", tree_id);
2221 key.objectid = dirid;
2222 key.type = BTRFS_INODE_REF_KEY;
2223 key.offset = (u64)-1;
2226 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2230 ret = btrfs_previous_item(root, path, dirid,
2231 BTRFS_INODE_REF_KEY);
2241 slot = path->slots[0];
2242 btrfs_item_key_to_cpu(l, &key, slot);
2244 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2245 len = btrfs_inode_ref_name_len(l, iref);
2247 total_len += len + 1;
2249 ret = -ENAMETOOLONG;
2254 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2256 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2259 btrfs_release_path(path);
2260 key.objectid = key.offset;
2261 key.offset = (u64)-1;
2262 dirid = key.objectid;
2264 memmove(name, ptr, total_len);
2265 name[total_len] = '\0';
2268 btrfs_free_path(path);
2272 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2275 struct btrfs_ioctl_ino_lookup_args *args;
2276 struct inode *inode;
2279 args = memdup_user(argp, sizeof(*args));
2281 return PTR_ERR(args);
2283 inode = file_inode(file);
2286 * Unprivileged query to obtain the containing subvolume root id. The
2287 * path is reset so it's consistent with btrfs_search_path_in_tree.
2289 if (args->treeid == 0)
2290 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2292 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2297 if (!capable(CAP_SYS_ADMIN)) {
2302 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2303 args->treeid, args->objectid,
2307 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2314 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2317 struct dentry *parent = file->f_path.dentry;
2318 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2319 struct dentry *dentry;
2320 struct inode *dir = d_inode(parent);
2321 struct inode *inode;
2322 struct btrfs_root *root = BTRFS_I(dir)->root;
2323 struct btrfs_root *dest = NULL;
2324 struct btrfs_ioctl_vol_args *vol_args;
2325 struct btrfs_trans_handle *trans;
2326 struct btrfs_block_rsv block_rsv;
2328 u64 qgroup_reserved;
2333 if (!S_ISDIR(dir->i_mode))
2336 vol_args = memdup_user(arg, sizeof(*vol_args));
2337 if (IS_ERR(vol_args))
2338 return PTR_ERR(vol_args);
2340 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2341 namelen = strlen(vol_args->name);
2342 if (strchr(vol_args->name, '/') ||
2343 strncmp(vol_args->name, "..", namelen) == 0) {
2348 err = mnt_want_write_file(file);
2353 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2355 goto out_drop_write;
2356 dentry = lookup_one_len(vol_args->name, parent, namelen);
2357 if (IS_ERR(dentry)) {
2358 err = PTR_ERR(dentry);
2359 goto out_unlock_dir;
2362 if (d_really_is_negative(dentry)) {
2367 inode = d_inode(dentry);
2368 dest = BTRFS_I(inode)->root;
2369 if (!capable(CAP_SYS_ADMIN)) {
2371 * Regular user. Only allow this with a special mount
2372 * option, when the user has write+exec access to the
2373 * subvol root, and when rmdir(2) would have been
2376 * Note that this is _not_ check that the subvol is
2377 * empty or doesn't contain data that we wouldn't
2378 * otherwise be able to delete.
2380 * Users who want to delete empty subvols should try
2384 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2388 * Do not allow deletion if the parent dir is the same
2389 * as the dir to be deleted. That means the ioctl
2390 * must be called on the dentry referencing the root
2391 * of the subvol, not a random directory contained
2398 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2403 /* check if subvolume may be deleted by a user */
2404 err = btrfs_may_delete(dir, dentry, 1);
2408 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2416 * Don't allow to delete a subvolume with send in progress. This is
2417 * inside the i_mutex so the error handling that has to drop the bit
2418 * again is not run concurrently.
2420 spin_lock(&dest->root_item_lock);
2421 root_flags = btrfs_root_flags(&dest->root_item);
2422 if (dest->send_in_progress == 0) {
2423 btrfs_set_root_flags(&dest->root_item,
2424 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2425 spin_unlock(&dest->root_item_lock);
2427 spin_unlock(&dest->root_item_lock);
2429 "Attempt to delete subvolume %llu during send",
2430 dest->root_key.objectid);
2432 goto out_unlock_inode;
2435 down_write(&fs_info->subvol_sem);
2437 err = may_destroy_subvol(dest);
2441 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2443 * One for dir inode, two for dir entries, two for root
2446 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2447 5, &qgroup_reserved, true);
2451 trans = btrfs_start_transaction(root, 0);
2452 if (IS_ERR(trans)) {
2453 err = PTR_ERR(trans);
2456 trans->block_rsv = &block_rsv;
2457 trans->bytes_reserved = block_rsv.size;
2459 btrfs_record_snapshot_destroy(trans, BTRFS_I(dir));
2461 ret = btrfs_unlink_subvol(trans, root, dir,
2462 dest->root_key.objectid,
2463 dentry->d_name.name,
2464 dentry->d_name.len);
2467 btrfs_abort_transaction(trans, ret);
2471 btrfs_record_root_in_trans(trans, dest);
2473 memset(&dest->root_item.drop_progress, 0,
2474 sizeof(dest->root_item.drop_progress));
2475 dest->root_item.drop_level = 0;
2476 btrfs_set_root_refs(&dest->root_item, 0);
2478 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2479 ret = btrfs_insert_orphan_item(trans,
2481 dest->root_key.objectid);
2483 btrfs_abort_transaction(trans, ret);
2489 ret = btrfs_uuid_tree_rem(trans, fs_info, dest->root_item.uuid,
2490 BTRFS_UUID_KEY_SUBVOL,
2491 dest->root_key.objectid);
2492 if (ret && ret != -ENOENT) {
2493 btrfs_abort_transaction(trans, ret);
2497 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2498 ret = btrfs_uuid_tree_rem(trans, fs_info,
2499 dest->root_item.received_uuid,
2500 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2501 dest->root_key.objectid);
2502 if (ret && ret != -ENOENT) {
2503 btrfs_abort_transaction(trans, ret);
2510 trans->block_rsv = NULL;
2511 trans->bytes_reserved = 0;
2512 ret = btrfs_end_transaction(trans);
2515 inode->i_flags |= S_DEAD;
2517 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
2519 up_write(&fs_info->subvol_sem);
2521 spin_lock(&dest->root_item_lock);
2522 root_flags = btrfs_root_flags(&dest->root_item);
2523 btrfs_set_root_flags(&dest->root_item,
2524 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2525 spin_unlock(&dest->root_item_lock);
2528 inode_unlock(inode);
2530 d_invalidate(dentry);
2531 btrfs_invalidate_inodes(dest);
2533 ASSERT(dest->send_in_progress == 0);
2536 if (dest->ino_cache_inode) {
2537 iput(dest->ino_cache_inode);
2538 dest->ino_cache_inode = NULL;
2546 mnt_drop_write_file(file);
2552 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2554 struct inode *inode = file_inode(file);
2555 struct btrfs_root *root = BTRFS_I(inode)->root;
2556 struct btrfs_ioctl_defrag_range_args *range;
2559 ret = mnt_want_write_file(file);
2563 if (btrfs_root_readonly(root)) {
2568 switch (inode->i_mode & S_IFMT) {
2570 if (!capable(CAP_SYS_ADMIN)) {
2574 ret = btrfs_defrag_root(root);
2577 if (!(file->f_mode & FMODE_WRITE)) {
2582 range = kzalloc(sizeof(*range), GFP_KERNEL);
2589 if (copy_from_user(range, argp,
2595 /* compression requires us to start the IO */
2596 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2597 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2598 range->extent_thresh = (u32)-1;
2601 /* the rest are all set to zero by kzalloc */
2602 range->len = (u64)-1;
2604 ret = btrfs_defrag_file(file_inode(file), file,
2614 mnt_drop_write_file(file);
2618 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2620 struct btrfs_ioctl_vol_args *vol_args;
2623 if (!capable(CAP_SYS_ADMIN))
2626 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
2627 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2629 mutex_lock(&fs_info->volume_mutex);
2630 vol_args = memdup_user(arg, sizeof(*vol_args));
2631 if (IS_ERR(vol_args)) {
2632 ret = PTR_ERR(vol_args);
2636 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2637 ret = btrfs_init_new_device(fs_info, vol_args->name);
2640 btrfs_info(fs_info, "disk added %s", vol_args->name);
2644 mutex_unlock(&fs_info->volume_mutex);
2645 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2649 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2651 struct inode *inode = file_inode(file);
2652 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2653 struct btrfs_ioctl_vol_args_v2 *vol_args;
2656 if (!capable(CAP_SYS_ADMIN))
2659 ret = mnt_want_write_file(file);
2663 vol_args = memdup_user(arg, sizeof(*vol_args));
2664 if (IS_ERR(vol_args)) {
2665 ret = PTR_ERR(vol_args);
2669 /* Check for compatibility reject unknown flags */
2670 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED)
2673 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2674 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2678 mutex_lock(&fs_info->volume_mutex);
2679 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2680 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
2682 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2683 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2685 mutex_unlock(&fs_info->volume_mutex);
2686 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2689 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2690 btrfs_info(fs_info, "device deleted: id %llu",
2693 btrfs_info(fs_info, "device deleted: %s",
2699 mnt_drop_write_file(file);
2703 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2705 struct inode *inode = file_inode(file);
2706 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2707 struct btrfs_ioctl_vol_args *vol_args;
2710 if (!capable(CAP_SYS_ADMIN))
2713 ret = mnt_want_write_file(file);
2717 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2718 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2719 goto out_drop_write;
2722 vol_args = memdup_user(arg, sizeof(*vol_args));
2723 if (IS_ERR(vol_args)) {
2724 ret = PTR_ERR(vol_args);
2728 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2729 mutex_lock(&fs_info->volume_mutex);
2730 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2731 mutex_unlock(&fs_info->volume_mutex);
2734 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2737 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2739 mnt_drop_write_file(file);
2744 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2747 struct btrfs_ioctl_fs_info_args *fi_args;
2748 struct btrfs_device *device;
2749 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2752 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2756 mutex_lock(&fs_devices->device_list_mutex);
2757 fi_args->num_devices = fs_devices->num_devices;
2758 memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid));
2760 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2761 if (device->devid > fi_args->max_id)
2762 fi_args->max_id = device->devid;
2764 mutex_unlock(&fs_devices->device_list_mutex);
2766 fi_args->nodesize = fs_info->nodesize;
2767 fi_args->sectorsize = fs_info->sectorsize;
2768 fi_args->clone_alignment = fs_info->sectorsize;
2770 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2777 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2780 struct btrfs_ioctl_dev_info_args *di_args;
2781 struct btrfs_device *dev;
2782 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2784 char *s_uuid = NULL;
2786 di_args = memdup_user(arg, sizeof(*di_args));
2787 if (IS_ERR(di_args))
2788 return PTR_ERR(di_args);
2790 if (!btrfs_is_empty_uuid(di_args->uuid))
2791 s_uuid = di_args->uuid;
2793 mutex_lock(&fs_devices->device_list_mutex);
2794 dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL);
2801 di_args->devid = dev->devid;
2802 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2803 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2804 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2806 struct rcu_string *name;
2809 name = rcu_dereference(dev->name);
2810 strncpy(di_args->path, name->str, sizeof(di_args->path));
2812 di_args->path[sizeof(di_args->path) - 1] = 0;
2814 di_args->path[0] = '\0';
2818 mutex_unlock(&fs_devices->device_list_mutex);
2819 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2826 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2830 page = grab_cache_page(inode->i_mapping, index);
2832 return ERR_PTR(-ENOMEM);
2834 if (!PageUptodate(page)) {
2837 ret = btrfs_readpage(NULL, page);
2839 return ERR_PTR(ret);
2841 if (!PageUptodate(page)) {
2844 return ERR_PTR(-EIO);
2846 if (page->mapping != inode->i_mapping) {
2849 return ERR_PTR(-EAGAIN);
2856 static int gather_extent_pages(struct inode *inode, struct page **pages,
2857 int num_pages, u64 off)
2860 pgoff_t index = off >> PAGE_SHIFT;
2862 for (i = 0; i < num_pages; i++) {
2864 pages[i] = extent_same_get_page(inode, index + i);
2865 if (IS_ERR(pages[i])) {
2866 int err = PTR_ERR(pages[i]);
2877 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2878 bool retry_range_locking)
2881 * Do any pending delalloc/csum calculations on inode, one way or
2882 * another, and lock file content.
2883 * The locking order is:
2886 * 2) range in the inode's io tree
2889 struct btrfs_ordered_extent *ordered;
2890 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2891 ordered = btrfs_lookup_first_ordered_extent(inode,
2894 ordered->file_offset + ordered->len <= off ||
2895 ordered->file_offset >= off + len) &&
2896 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2897 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2899 btrfs_put_ordered_extent(ordered);
2902 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2904 btrfs_put_ordered_extent(ordered);
2905 if (!retry_range_locking)
2907 btrfs_wait_ordered_range(inode, off, len);
2912 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2914 inode_unlock(inode1);
2915 inode_unlock(inode2);
2918 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2920 if (inode1 < inode2)
2921 swap(inode1, inode2);
2923 inode_lock_nested(inode1, I_MUTEX_PARENT);
2924 inode_lock_nested(inode2, I_MUTEX_CHILD);
2927 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2928 struct inode *inode2, u64 loff2, u64 len)
2930 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2931 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2934 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2935 struct inode *inode2, u64 loff2, u64 len,
2936 bool retry_range_locking)
2940 if (inode1 < inode2) {
2941 swap(inode1, inode2);
2944 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2947 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2949 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2956 struct page **src_pages;
2957 struct page **dst_pages;
2960 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2965 for (i = 0; i < cmp->num_pages; i++) {
2966 pg = cmp->src_pages[i];
2971 pg = cmp->dst_pages[i];
2977 kfree(cmp->src_pages);
2978 kfree(cmp->dst_pages);
2981 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2982 struct inode *dst, u64 dst_loff,
2983 u64 len, struct cmp_pages *cmp)
2986 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
2987 struct page **src_pgarr, **dst_pgarr;
2990 * We must gather up all the pages before we initiate our
2991 * extent locking. We use an array for the page pointers. Size
2992 * of the array is bounded by len, which is in turn bounded by
2993 * BTRFS_MAX_DEDUPE_LEN.
2995 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2996 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2997 if (!src_pgarr || !dst_pgarr) {
3002 cmp->num_pages = num_pages;
3003 cmp->src_pages = src_pgarr;
3004 cmp->dst_pages = dst_pgarr;
3007 * If deduping ranges in the same inode, locking rules make it mandatory
3008 * to always lock pages in ascending order to avoid deadlocks with
3009 * concurrent tasks (such as starting writeback/delalloc).
3011 if (src == dst && dst_loff < loff) {
3012 swap(src_pgarr, dst_pgarr);
3013 swap(loff, dst_loff);
3016 ret = gather_extent_pages(src, src_pgarr, cmp->num_pages, loff);
3020 ret = gather_extent_pages(dst, dst_pgarr, cmp->num_pages, dst_loff);
3024 btrfs_cmp_data_free(cmp);
3028 static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp)
3032 struct page *src_page, *dst_page;
3033 unsigned int cmp_len = PAGE_SIZE;
3034 void *addr, *dst_addr;
3038 if (len < PAGE_SIZE)
3041 BUG_ON(i >= cmp->num_pages);
3043 src_page = cmp->src_pages[i];
3044 dst_page = cmp->dst_pages[i];
3045 ASSERT(PageLocked(src_page));
3046 ASSERT(PageLocked(dst_page));
3048 addr = kmap_atomic(src_page);
3049 dst_addr = kmap_atomic(dst_page);
3051 flush_dcache_page(src_page);
3052 flush_dcache_page(dst_page);
3054 if (memcmp(addr, dst_addr, cmp_len))
3057 kunmap_atomic(addr);
3058 kunmap_atomic(dst_addr);
3070 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3074 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3076 if (off + olen > inode->i_size || off + olen < off)
3079 /* if we extend to eof, continue to block boundary */
3080 if (off + len == inode->i_size)
3081 *plen = len = ALIGN(inode->i_size, bs) - off;
3083 /* Check that we are block aligned - btrfs_clone() requires this */
3084 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3090 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3091 struct inode *dst, u64 dst_loff)
3095 struct cmp_pages cmp;
3096 bool same_inode = (src == dst);
3097 u64 same_lock_start = 0;
3098 u64 same_lock_len = 0;
3106 btrfs_double_inode_lock(src, dst);
3108 ret = extent_same_check_offsets(src, loff, &len, olen);
3112 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3118 * Single inode case wants the same checks, except we
3119 * don't want our length pushed out past i_size as
3120 * comparing that data range makes no sense.
3122 * extent_same_check_offsets() will do this for an
3123 * unaligned length at i_size, so catch it here and
3124 * reject the request.
3126 * This effectively means we require aligned extents
3127 * for the single-inode case, whereas the other cases
3128 * allow an unaligned length so long as it ends at
3136 /* Check for overlapping ranges */
3137 if (dst_loff + len > loff && dst_loff < loff + len) {
3142 same_lock_start = min_t(u64, loff, dst_loff);
3143 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3146 /* don't make the dst file partly checksummed */
3147 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3148 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3154 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3159 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3162 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3165 * If one of the inodes has dirty pages in the respective range or
3166 * ordered extents, we need to flush dellaloc and wait for all ordered
3167 * extents in the range. We must unlock the pages and the ranges in the
3168 * io trees to avoid deadlocks when flushing delalloc (requires locking
3169 * pages) and when waiting for ordered extents to complete (they require
3172 if (ret == -EAGAIN) {
3174 * Ranges in the io trees already unlocked. Now unlock all
3175 * pages before waiting for all IO to complete.
3177 btrfs_cmp_data_free(&cmp);
3179 btrfs_wait_ordered_range(src, same_lock_start,
3182 btrfs_wait_ordered_range(src, loff, len);
3183 btrfs_wait_ordered_range(dst, dst_loff, len);
3189 /* ranges in the io trees already unlocked */
3190 btrfs_cmp_data_free(&cmp);
3194 /* pass original length for comparison so we stay within i_size */
3195 ret = btrfs_cmp_data(olen, &cmp);
3197 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3200 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3201 same_lock_start + same_lock_len - 1);
3203 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3205 btrfs_cmp_data_free(&cmp);
3210 btrfs_double_inode_unlock(src, dst);
3215 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3217 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3218 struct file *dst_file, u64 dst_loff)
3220 struct inode *src = file_inode(src_file);
3221 struct inode *dst = file_inode(dst_file);
3222 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3225 if (olen > BTRFS_MAX_DEDUPE_LEN)
3226 olen = BTRFS_MAX_DEDUPE_LEN;
3228 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
3230 * Btrfs does not support blocksize < page_size. As a
3231 * result, btrfs_cmp_data() won't correctly handle
3232 * this situation without an update.
3237 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3243 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3244 struct inode *inode,
3250 struct btrfs_root *root = BTRFS_I(inode)->root;
3253 inode_inc_iversion(inode);
3254 if (!no_time_update)
3255 inode->i_mtime = inode->i_ctime = current_time(inode);
3257 * We round up to the block size at eof when determining which
3258 * extents to clone above, but shouldn't round up the file size.
3260 if (endoff > destoff + olen)
3261 endoff = destoff + olen;
3262 if (endoff > inode->i_size)
3263 btrfs_i_size_write(BTRFS_I(inode), endoff);
3265 ret = btrfs_update_inode(trans, root, inode);
3267 btrfs_abort_transaction(trans, ret);
3268 btrfs_end_transaction(trans);
3271 ret = btrfs_end_transaction(trans);
3276 static void clone_update_extent_map(struct btrfs_inode *inode,
3277 const struct btrfs_trans_handle *trans,
3278 const struct btrfs_path *path,
3279 const u64 hole_offset,
3282 struct extent_map_tree *em_tree = &inode->extent_tree;
3283 struct extent_map *em;
3286 em = alloc_extent_map();
3288 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3293 struct btrfs_file_extent_item *fi;
3295 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3296 struct btrfs_file_extent_item);
3297 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3298 em->generation = -1;
3299 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3300 BTRFS_FILE_EXTENT_INLINE)
3301 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3302 &inode->runtime_flags);
3304 em->start = hole_offset;
3306 em->ram_bytes = em->len;
3307 em->orig_start = hole_offset;
3308 em->block_start = EXTENT_MAP_HOLE;
3310 em->orig_block_len = 0;
3311 em->compress_type = BTRFS_COMPRESS_NONE;
3312 em->generation = trans->transid;
3316 write_lock(&em_tree->lock);
3317 ret = add_extent_mapping(em_tree, em, 1);
3318 write_unlock(&em_tree->lock);
3319 if (ret != -EEXIST) {
3320 free_extent_map(em);
3323 btrfs_drop_extent_cache(inode, em->start,
3324 em->start + em->len - 1, 0);
3328 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3332 * Make sure we do not end up inserting an inline extent into a file that has
3333 * already other (non-inline) extents. If a file has an inline extent it can
3334 * not have any other extents and the (single) inline extent must start at the
3335 * file offset 0. Failing to respect these rules will lead to file corruption,
3336 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3338 * We can have extents that have been already written to disk or we can have
3339 * dirty ranges still in delalloc, in which case the extent maps and items are
3340 * created only when we run delalloc, and the delalloc ranges might fall outside
3341 * the range we are currently locking in the inode's io tree. So we check the
3342 * inode's i_size because of that (i_size updates are done while holding the
3343 * i_mutex, which we are holding here).
3344 * We also check to see if the inode has a size not greater than "datal" but has
3345 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3346 * protected against such concurrent fallocate calls by the i_mutex).
3348 * If the file has no extents but a size greater than datal, do not allow the
3349 * copy because we would need turn the inline extent into a non-inline one (even
3350 * with NO_HOLES enabled). If we find our destination inode only has one inline
3351 * extent, just overwrite it with the source inline extent if its size is less
3352 * than the source extent's size, or we could copy the source inline extent's
3353 * data into the destination inode's inline extent if the later is greater then
3356 static int clone_copy_inline_extent(struct inode *dst,
3357 struct btrfs_trans_handle *trans,
3358 struct btrfs_path *path,
3359 struct btrfs_key *new_key,
3360 const u64 drop_start,
3366 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
3367 struct btrfs_root *root = BTRFS_I(dst)->root;
3368 const u64 aligned_end = ALIGN(new_key->offset + datal,
3369 fs_info->sectorsize);
3371 struct btrfs_key key;
3373 if (new_key->offset > 0)
3376 key.objectid = btrfs_ino(BTRFS_I(dst));
3377 key.type = BTRFS_EXTENT_DATA_KEY;
3379 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3382 } else if (ret > 0) {
3383 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3384 ret = btrfs_next_leaf(root, path);
3388 goto copy_inline_extent;
3390 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3391 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3392 key.type == BTRFS_EXTENT_DATA_KEY) {
3393 ASSERT(key.offset > 0);
3396 } else if (i_size_read(dst) <= datal) {
3397 struct btrfs_file_extent_item *ei;
3401 * If the file size is <= datal, make sure there are no other
3402 * extents following (can happen do to an fallocate call with
3403 * the flag FALLOC_FL_KEEP_SIZE).
3405 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3406 struct btrfs_file_extent_item);
3408 * If it's an inline extent, it can not have other extents
3411 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3412 BTRFS_FILE_EXTENT_INLINE)
3413 goto copy_inline_extent;
3415 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3416 if (ext_len > aligned_end)
3419 ret = btrfs_next_item(root, path);
3422 } else if (ret == 0) {
3423 btrfs_item_key_to_cpu(path->nodes[0], &key,
3425 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3426 key.type == BTRFS_EXTENT_DATA_KEY)
3433 * We have no extent items, or we have an extent at offset 0 which may
3434 * or may not be inlined. All these cases are dealt the same way.
3436 if (i_size_read(dst) > datal) {
3438 * If the destination inode has an inline extent...
3439 * This would require copying the data from the source inline
3440 * extent into the beginning of the destination's inline extent.
3441 * But this is really complex, both extents can be compressed
3442 * or just one of them, which would require decompressing and
3443 * re-compressing data (which could increase the new compressed
3444 * size, not allowing the compressed data to fit anymore in an
3446 * So just don't support this case for now (it should be rare,
3447 * we are not really saving space when cloning inline extents).
3452 btrfs_release_path(path);
3453 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3456 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3461 const u32 start = btrfs_file_extent_calc_inline_size(0);
3463 memmove(inline_data + start, inline_data + start + skip, datal);
3466 write_extent_buffer(path->nodes[0], inline_data,
3467 btrfs_item_ptr_offset(path->nodes[0],
3470 inode_add_bytes(dst, datal);
3476 * btrfs_clone() - clone a range from inode file to another
3478 * @src: Inode to clone from
3479 * @inode: Inode to clone to
3480 * @off: Offset within source to start clone from
3481 * @olen: Original length, passed by user, of range to clone
3482 * @olen_aligned: Block-aligned value of olen
3483 * @destoff: Offset within @inode to start clone
3484 * @no_time_update: Whether to update mtime/ctime on the target inode
3486 static int btrfs_clone(struct inode *src, struct inode *inode,
3487 const u64 off, const u64 olen, const u64 olen_aligned,
3488 const u64 destoff, int no_time_update)
3490 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3491 struct btrfs_root *root = BTRFS_I(inode)->root;
3492 struct btrfs_path *path = NULL;
3493 struct extent_buffer *leaf;
3494 struct btrfs_trans_handle *trans;
3496 struct btrfs_key key;
3500 const u64 len = olen_aligned;
3501 u64 last_dest_end = destoff;
3504 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3508 path = btrfs_alloc_path();
3514 path->reada = READA_FORWARD;
3516 key.objectid = btrfs_ino(BTRFS_I(src));
3517 key.type = BTRFS_EXTENT_DATA_KEY;
3521 u64 next_key_min_offset = key.offset + 1;
3524 * note the key will change type as we walk through the
3527 path->leave_spinning = 1;
3528 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3533 * First search, if no extent item that starts at offset off was
3534 * found but the previous item is an extent item, it's possible
3535 * it might overlap our target range, therefore process it.
3537 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3538 btrfs_item_key_to_cpu(path->nodes[0], &key,
3539 path->slots[0] - 1);
3540 if (key.type == BTRFS_EXTENT_DATA_KEY)
3544 nritems = btrfs_header_nritems(path->nodes[0]);
3546 if (path->slots[0] >= nritems) {
3547 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3552 nritems = btrfs_header_nritems(path->nodes[0]);
3554 leaf = path->nodes[0];
3555 slot = path->slots[0];
3557 btrfs_item_key_to_cpu(leaf, &key, slot);
3558 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3559 key.objectid != btrfs_ino(BTRFS_I(src)))
3562 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3563 struct btrfs_file_extent_item *extent;
3566 struct btrfs_key new_key;
3567 u64 disko = 0, diskl = 0;
3568 u64 datao = 0, datal = 0;
3572 extent = btrfs_item_ptr(leaf, slot,
3573 struct btrfs_file_extent_item);
3574 comp = btrfs_file_extent_compression(leaf, extent);
3575 type = btrfs_file_extent_type(leaf, extent);
3576 if (type == BTRFS_FILE_EXTENT_REG ||
3577 type == BTRFS_FILE_EXTENT_PREALLOC) {
3578 disko = btrfs_file_extent_disk_bytenr(leaf,
3580 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3582 datao = btrfs_file_extent_offset(leaf, extent);
3583 datal = btrfs_file_extent_num_bytes(leaf,
3585 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3586 /* take upper bound, may be compressed */
3587 datal = btrfs_file_extent_ram_bytes(leaf,
3592 * The first search might have left us at an extent
3593 * item that ends before our target range's start, can
3594 * happen if we have holes and NO_HOLES feature enabled.
3596 if (key.offset + datal <= off) {
3599 } else if (key.offset >= off + len) {
3602 next_key_min_offset = key.offset + datal;
3603 size = btrfs_item_size_nr(leaf, slot);
3604 read_extent_buffer(leaf, buf,
3605 btrfs_item_ptr_offset(leaf, slot),
3608 btrfs_release_path(path);
3609 path->leave_spinning = 0;
3611 memcpy(&new_key, &key, sizeof(new_key));
3612 new_key.objectid = btrfs_ino(BTRFS_I(inode));
3613 if (off <= key.offset)
3614 new_key.offset = key.offset + destoff - off;
3616 new_key.offset = destoff;
3619 * Deal with a hole that doesn't have an extent item
3620 * that represents it (NO_HOLES feature enabled).
3621 * This hole is either in the middle of the cloning
3622 * range or at the beginning (fully overlaps it or
3623 * partially overlaps it).
3625 if (new_key.offset != last_dest_end)
3626 drop_start = last_dest_end;
3628 drop_start = new_key.offset;
3631 * 1 - adjusting old extent (we may have to split it)
3632 * 1 - add new extent
3635 trans = btrfs_start_transaction(root, 3);
3636 if (IS_ERR(trans)) {
3637 ret = PTR_ERR(trans);
3641 if (type == BTRFS_FILE_EXTENT_REG ||
3642 type == BTRFS_FILE_EXTENT_PREALLOC) {
3644 * a | --- range to clone ---| b
3645 * | ------------- extent ------------- |
3648 /* subtract range b */
3649 if (key.offset + datal > off + len)
3650 datal = off + len - key.offset;
3652 /* subtract range a */
3653 if (off > key.offset) {
3654 datao += off - key.offset;
3655 datal -= off - key.offset;
3658 ret = btrfs_drop_extents(trans, root, inode,
3660 new_key.offset + datal,
3663 if (ret != -EOPNOTSUPP)
3664 btrfs_abort_transaction(trans,
3666 btrfs_end_transaction(trans);
3670 ret = btrfs_insert_empty_item(trans, root, path,
3673 btrfs_abort_transaction(trans, ret);
3674 btrfs_end_transaction(trans);
3678 leaf = path->nodes[0];
3679 slot = path->slots[0];
3680 write_extent_buffer(leaf, buf,
3681 btrfs_item_ptr_offset(leaf, slot),
3684 extent = btrfs_item_ptr(leaf, slot,
3685 struct btrfs_file_extent_item);
3687 /* disko == 0 means it's a hole */
3691 btrfs_set_file_extent_offset(leaf, extent,
3693 btrfs_set_file_extent_num_bytes(leaf, extent,
3697 inode_add_bytes(inode, datal);
3698 ret = btrfs_inc_extent_ref(trans,
3701 root->root_key.objectid,
3702 btrfs_ino(BTRFS_I(inode)),
3703 new_key.offset - datao);
3705 btrfs_abort_transaction(trans,
3707 btrfs_end_transaction(trans);
3712 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3716 if (off > key.offset) {
3717 skip = off - key.offset;
3718 new_key.offset += skip;
3721 if (key.offset + datal > off + len)
3722 trim = key.offset + datal - (off + len);
3724 if (comp && (skip || trim)) {
3726 btrfs_end_transaction(trans);
3729 size -= skip + trim;
3730 datal -= skip + trim;
3732 ret = clone_copy_inline_extent(inode,
3739 if (ret != -EOPNOTSUPP)
3740 btrfs_abort_transaction(trans,
3742 btrfs_end_transaction(trans);
3745 leaf = path->nodes[0];
3746 slot = path->slots[0];
3749 /* If we have an implicit hole (NO_HOLES feature). */
3750 if (drop_start < new_key.offset)
3751 clone_update_extent_map(BTRFS_I(inode), trans,
3753 new_key.offset - drop_start);
3755 clone_update_extent_map(BTRFS_I(inode), trans,
3758 btrfs_mark_buffer_dirty(leaf);
3759 btrfs_release_path(path);
3761 last_dest_end = ALIGN(new_key.offset + datal,
3762 fs_info->sectorsize);
3763 ret = clone_finish_inode_update(trans, inode,
3769 if (new_key.offset + datal >= destoff + len)
3772 btrfs_release_path(path);
3773 key.offset = next_key_min_offset;
3775 if (fatal_signal_pending(current)) {
3782 if (last_dest_end < destoff + len) {
3784 * We have an implicit hole (NO_HOLES feature is enabled) that
3785 * fully or partially overlaps our cloning range at its end.
3787 btrfs_release_path(path);
3790 * 1 - remove extent(s)
3793 trans = btrfs_start_transaction(root, 2);
3794 if (IS_ERR(trans)) {
3795 ret = PTR_ERR(trans);
3798 ret = btrfs_drop_extents(trans, root, inode,
3799 last_dest_end, destoff + len, 1);
3801 if (ret != -EOPNOTSUPP)
3802 btrfs_abort_transaction(trans, ret);
3803 btrfs_end_transaction(trans);
3806 clone_update_extent_map(BTRFS_I(inode), trans, NULL,
3808 destoff + len - last_dest_end);
3809 ret = clone_finish_inode_update(trans, inode, destoff + len,
3810 destoff, olen, no_time_update);
3814 btrfs_free_path(path);
3819 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3820 u64 off, u64 olen, u64 destoff)
3822 struct inode *inode = file_inode(file);
3823 struct inode *src = file_inode(file_src);
3824 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3825 struct btrfs_root *root = BTRFS_I(inode)->root;
3828 u64 bs = fs_info->sb->s_blocksize;
3829 int same_inode = src == inode;
3833 * - split compressed inline extents. annoying: we need to
3834 * decompress into destination's address_space (the file offset
3835 * may change, so source mapping won't do), then recompress (or
3836 * otherwise reinsert) a subrange.
3838 * - split destination inode's inline extents. The inline extents can
3839 * be either compressed or non-compressed.
3842 if (btrfs_root_readonly(root))
3845 if (file_src->f_path.mnt != file->f_path.mnt ||
3846 src->i_sb != inode->i_sb)
3849 /* don't make the dst file partly checksummed */
3850 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3851 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3854 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3858 btrfs_double_inode_lock(src, inode);
3863 /* determine range to clone */
3865 if (off + len > src->i_size || off + len < off)
3868 olen = len = src->i_size - off;
3869 /* if we extend to eof, continue to block boundary */
3870 if (off + len == src->i_size)
3871 len = ALIGN(src->i_size, bs) - off;
3878 /* verify the end result is block aligned */
3879 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3880 !IS_ALIGNED(destoff, bs))
3883 /* verify if ranges are overlapped within the same file */
3885 if (destoff + len > off && destoff < off + len)
3889 if (destoff > inode->i_size) {
3890 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3896 * Lock the target range too. Right after we replace the file extent
3897 * items in the fs tree (which now point to the cloned data), we might
3898 * have a worker replace them with extent items relative to a write
3899 * operation that was issued before this clone operation (i.e. confront
3900 * with inode.c:btrfs_finish_ordered_io).
3903 u64 lock_start = min_t(u64, off, destoff);
3904 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3906 ret = lock_extent_range(src, lock_start, lock_len, true);
3908 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3913 /* ranges in the io trees already unlocked */
3917 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3920 u64 lock_start = min_t(u64, off, destoff);
3921 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3923 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3925 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3928 * Truncate page cache pages so that future reads will see the cloned
3929 * data immediately and not the previous data.
3931 truncate_inode_pages_range(&inode->i_data,
3932 round_down(destoff, PAGE_SIZE),
3933 round_up(destoff + len, PAGE_SIZE) - 1);
3936 btrfs_double_inode_unlock(src, inode);
3942 int btrfs_clone_file_range(struct file *src_file, loff_t off,
3943 struct file *dst_file, loff_t destoff, u64 len)
3945 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
3949 * there are many ways the trans_start and trans_end ioctls can lead
3950 * to deadlocks. They should only be used by applications that
3951 * basically own the machine, and have a very in depth understanding
3952 * of all the possible deadlocks and enospc problems.
3954 static long btrfs_ioctl_trans_start(struct file *file)
3956 struct inode *inode = file_inode(file);
3957 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3958 struct btrfs_root *root = BTRFS_I(inode)->root;
3959 struct btrfs_trans_handle *trans;
3960 struct btrfs_file_private *private;
3962 static bool warned = false;
3965 if (!capable(CAP_SYS_ADMIN))
3970 "Userspace transaction mechanism is considered "
3971 "deprecated and slated to be removed in 4.17. "
3972 "If you have a valid use case please "
3973 "speak up on the mailing list");
3979 private = file->private_data;
3980 if (private && private->trans)
3983 private = kzalloc(sizeof(struct btrfs_file_private),
3987 file->private_data = private;
3991 if (btrfs_root_readonly(root))
3994 ret = mnt_want_write_file(file);
3998 atomic_inc(&fs_info->open_ioctl_trans);
4001 trans = btrfs_start_ioctl_transaction(root);
4005 private->trans = trans;
4009 atomic_dec(&fs_info->open_ioctl_trans);
4010 mnt_drop_write_file(file);
4015 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
4017 struct inode *inode = file_inode(file);
4018 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4019 struct btrfs_root *root = BTRFS_I(inode)->root;
4020 struct btrfs_root *new_root;
4021 struct btrfs_dir_item *di;
4022 struct btrfs_trans_handle *trans;
4023 struct btrfs_path *path;
4024 struct btrfs_key location;
4025 struct btrfs_disk_key disk_key;
4030 if (!capable(CAP_SYS_ADMIN))
4033 ret = mnt_want_write_file(file);
4037 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4043 objectid = BTRFS_FS_TREE_OBJECTID;
4045 location.objectid = objectid;
4046 location.type = BTRFS_ROOT_ITEM_KEY;
4047 location.offset = (u64)-1;
4049 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
4050 if (IS_ERR(new_root)) {
4051 ret = PTR_ERR(new_root);
4054 if (!is_fstree(new_root->objectid)) {
4059 path = btrfs_alloc_path();
4064 path->leave_spinning = 1;
4066 trans = btrfs_start_transaction(root, 1);
4067 if (IS_ERR(trans)) {
4068 btrfs_free_path(path);
4069 ret = PTR_ERR(trans);
4073 dir_id = btrfs_super_root_dir(fs_info->super_copy);
4074 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
4075 dir_id, "default", 7, 1);
4076 if (IS_ERR_OR_NULL(di)) {
4077 btrfs_free_path(path);
4078 btrfs_end_transaction(trans);
4080 "Umm, you don't have the default diritem, this isn't going to work");
4085 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4086 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4087 btrfs_mark_buffer_dirty(path->nodes[0]);
4088 btrfs_free_path(path);
4090 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
4091 btrfs_end_transaction(trans);
4093 mnt_drop_write_file(file);
4097 void btrfs_get_block_group_info(struct list_head *groups_list,
4098 struct btrfs_ioctl_space_info *space)
4100 struct btrfs_block_group_cache *block_group;
4102 space->total_bytes = 0;
4103 space->used_bytes = 0;
4105 list_for_each_entry(block_group, groups_list, list) {
4106 space->flags = block_group->flags;
4107 space->total_bytes += block_group->key.offset;
4108 space->used_bytes +=
4109 btrfs_block_group_used(&block_group->item);
4113 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
4116 struct btrfs_ioctl_space_args space_args;
4117 struct btrfs_ioctl_space_info space;
4118 struct btrfs_ioctl_space_info *dest;
4119 struct btrfs_ioctl_space_info *dest_orig;
4120 struct btrfs_ioctl_space_info __user *user_dest;
4121 struct btrfs_space_info *info;
4122 static const u64 types[] = {
4123 BTRFS_BLOCK_GROUP_DATA,
4124 BTRFS_BLOCK_GROUP_SYSTEM,
4125 BTRFS_BLOCK_GROUP_METADATA,
4126 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
4134 if (copy_from_user(&space_args,
4135 (struct btrfs_ioctl_space_args __user *)arg,
4136 sizeof(space_args)))
4139 for (i = 0; i < num_types; i++) {
4140 struct btrfs_space_info *tmp;
4144 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4146 if (tmp->flags == types[i]) {
4156 down_read(&info->groups_sem);
4157 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4158 if (!list_empty(&info->block_groups[c]))
4161 up_read(&info->groups_sem);
4165 * Global block reserve, exported as a space_info
4169 /* space_slots == 0 means they are asking for a count */
4170 if (space_args.space_slots == 0) {
4171 space_args.total_spaces = slot_count;
4175 slot_count = min_t(u64, space_args.space_slots, slot_count);
4177 alloc_size = sizeof(*dest) * slot_count;
4179 /* we generally have at most 6 or so space infos, one for each raid
4180 * level. So, a whole page should be more than enough for everyone
4182 if (alloc_size > PAGE_SIZE)
4185 space_args.total_spaces = 0;
4186 dest = kmalloc(alloc_size, GFP_KERNEL);
4191 /* now we have a buffer to copy into */
4192 for (i = 0; i < num_types; i++) {
4193 struct btrfs_space_info *tmp;
4200 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4202 if (tmp->flags == types[i]) {
4211 down_read(&info->groups_sem);
4212 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4213 if (!list_empty(&info->block_groups[c])) {
4214 btrfs_get_block_group_info(
4215 &info->block_groups[c], &space);
4216 memcpy(dest, &space, sizeof(space));
4218 space_args.total_spaces++;
4224 up_read(&info->groups_sem);
4228 * Add global block reserve
4231 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
4233 spin_lock(&block_rsv->lock);
4234 space.total_bytes = block_rsv->size;
4235 space.used_bytes = block_rsv->size - block_rsv->reserved;
4236 spin_unlock(&block_rsv->lock);
4237 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4238 memcpy(dest, &space, sizeof(space));
4239 space_args.total_spaces++;
4242 user_dest = (struct btrfs_ioctl_space_info __user *)
4243 (arg + sizeof(struct btrfs_ioctl_space_args));
4245 if (copy_to_user(user_dest, dest_orig, alloc_size))
4250 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4257 * there are many ways the trans_start and trans_end ioctls can lead
4258 * to deadlocks. They should only be used by applications that
4259 * basically own the machine, and have a very in depth understanding
4260 * of all the possible deadlocks and enospc problems.
4262 long btrfs_ioctl_trans_end(struct file *file)
4264 struct inode *inode = file_inode(file);
4265 struct btrfs_root *root = BTRFS_I(inode)->root;
4266 struct btrfs_file_private *private = file->private_data;
4268 if (!private || !private->trans)
4271 btrfs_end_transaction(private->trans);
4272 private->trans = NULL;
4274 atomic_dec(&root->fs_info->open_ioctl_trans);
4276 mnt_drop_write_file(file);
4280 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4283 struct btrfs_trans_handle *trans;
4287 trans = btrfs_attach_transaction_barrier(root);
4288 if (IS_ERR(trans)) {
4289 if (PTR_ERR(trans) != -ENOENT)
4290 return PTR_ERR(trans);
4292 /* No running transaction, don't bother */
4293 transid = root->fs_info->last_trans_committed;
4296 transid = trans->transid;
4297 ret = btrfs_commit_transaction_async(trans, 0);
4299 btrfs_end_transaction(trans);
4304 if (copy_to_user(argp, &transid, sizeof(transid)))
4309 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
4315 if (copy_from_user(&transid, argp, sizeof(transid)))
4318 transid = 0; /* current trans */
4320 return btrfs_wait_for_commit(fs_info, transid);
4323 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4325 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4326 struct btrfs_ioctl_scrub_args *sa;
4329 if (!capable(CAP_SYS_ADMIN))
4332 sa = memdup_user(arg, sizeof(*sa));
4336 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4337 ret = mnt_want_write_file(file);
4342 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4343 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4346 if (copy_to_user(arg, sa, sizeof(*sa)))
4349 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4350 mnt_drop_write_file(file);
4356 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
4358 if (!capable(CAP_SYS_ADMIN))
4361 return btrfs_scrub_cancel(fs_info);
4364 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
4367 struct btrfs_ioctl_scrub_args *sa;
4370 if (!capable(CAP_SYS_ADMIN))
4373 sa = memdup_user(arg, sizeof(*sa));
4377 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
4379 if (copy_to_user(arg, sa, sizeof(*sa)))
4386 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
4389 struct btrfs_ioctl_get_dev_stats *sa;
4392 sa = memdup_user(arg, sizeof(*sa));
4396 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4401 ret = btrfs_get_dev_stats(fs_info, sa);
4403 if (copy_to_user(arg, sa, sizeof(*sa)))
4410 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4413 struct btrfs_ioctl_dev_replace_args *p;
4416 if (!capable(CAP_SYS_ADMIN))
4419 p = memdup_user(arg, sizeof(*p));
4424 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4425 if (sb_rdonly(fs_info->sb)) {
4429 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4430 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4432 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
4433 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4436 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4437 btrfs_dev_replace_status(fs_info, p);
4440 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4441 ret = btrfs_dev_replace_cancel(fs_info, p);
4448 if (copy_to_user(arg, p, sizeof(*p)))
4455 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4461 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4462 struct inode_fs_paths *ipath = NULL;
4463 struct btrfs_path *path;
4465 if (!capable(CAP_DAC_READ_SEARCH))
4468 path = btrfs_alloc_path();
4474 ipa = memdup_user(arg, sizeof(*ipa));
4481 size = min_t(u32, ipa->size, 4096);
4482 ipath = init_ipath(size, root, path);
4483 if (IS_ERR(ipath)) {
4484 ret = PTR_ERR(ipath);
4489 ret = paths_from_inode(ipa->inum, ipath);
4493 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4494 rel_ptr = ipath->fspath->val[i] -
4495 (u64)(unsigned long)ipath->fspath->val;
4496 ipath->fspath->val[i] = rel_ptr;
4499 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
4500 ipath->fspath, size);
4507 btrfs_free_path(path);
4514 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4516 struct btrfs_data_container *inodes = ctx;
4517 const size_t c = 3 * sizeof(u64);
4519 if (inodes->bytes_left >= c) {
4520 inodes->bytes_left -= c;
4521 inodes->val[inodes->elem_cnt] = inum;
4522 inodes->val[inodes->elem_cnt + 1] = offset;
4523 inodes->val[inodes->elem_cnt + 2] = root;
4524 inodes->elem_cnt += 3;
4526 inodes->bytes_missing += c - inodes->bytes_left;
4527 inodes->bytes_left = 0;
4528 inodes->elem_missed += 3;
4534 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
4539 struct btrfs_ioctl_logical_ino_args *loi;
4540 struct btrfs_data_container *inodes = NULL;
4541 struct btrfs_path *path = NULL;
4543 if (!capable(CAP_SYS_ADMIN))
4546 loi = memdup_user(arg, sizeof(*loi));
4548 return PTR_ERR(loi);
4550 path = btrfs_alloc_path();
4556 size = min_t(u32, loi->size, SZ_64K);
4557 inodes = init_data_container(size);
4558 if (IS_ERR(inodes)) {
4559 ret = PTR_ERR(inodes);
4564 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
4565 build_ino_list, inodes);
4571 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
4577 btrfs_free_path(path);
4584 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4585 struct btrfs_ioctl_balance_args *bargs)
4587 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4589 bargs->flags = bctl->flags;
4591 if (atomic_read(&fs_info->balance_running))
4592 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4593 if (atomic_read(&fs_info->balance_pause_req))
4594 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4595 if (atomic_read(&fs_info->balance_cancel_req))
4596 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4598 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4599 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4600 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4603 spin_lock(&fs_info->balance_lock);
4604 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4605 spin_unlock(&fs_info->balance_lock);
4607 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4611 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4613 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4614 struct btrfs_fs_info *fs_info = root->fs_info;
4615 struct btrfs_ioctl_balance_args *bargs;
4616 struct btrfs_balance_control *bctl;
4617 bool need_unlock; /* for mut. excl. ops lock */
4620 if (!capable(CAP_SYS_ADMIN))
4623 ret = mnt_want_write_file(file);
4628 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4629 mutex_lock(&fs_info->volume_mutex);
4630 mutex_lock(&fs_info->balance_mutex);
4636 * mut. excl. ops lock is locked. Three possibilities:
4637 * (1) some other op is running
4638 * (2) balance is running
4639 * (3) balance is paused -- special case (think resume)
4641 mutex_lock(&fs_info->balance_mutex);
4642 if (fs_info->balance_ctl) {
4643 /* this is either (2) or (3) */
4644 if (!atomic_read(&fs_info->balance_running)) {
4645 mutex_unlock(&fs_info->balance_mutex);
4646 if (!mutex_trylock(&fs_info->volume_mutex))
4648 mutex_lock(&fs_info->balance_mutex);
4650 if (fs_info->balance_ctl &&
4651 !atomic_read(&fs_info->balance_running)) {
4653 need_unlock = false;
4657 mutex_unlock(&fs_info->balance_mutex);
4658 mutex_unlock(&fs_info->volume_mutex);
4662 mutex_unlock(&fs_info->balance_mutex);
4668 mutex_unlock(&fs_info->balance_mutex);
4669 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4674 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
4677 bargs = memdup_user(arg, sizeof(*bargs));
4678 if (IS_ERR(bargs)) {
4679 ret = PTR_ERR(bargs);
4683 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4684 if (!fs_info->balance_ctl) {
4689 bctl = fs_info->balance_ctl;
4690 spin_lock(&fs_info->balance_lock);
4691 bctl->flags |= BTRFS_BALANCE_RESUME;
4692 spin_unlock(&fs_info->balance_lock);
4700 if (fs_info->balance_ctl) {
4705 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4711 bctl->fs_info = fs_info;
4713 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4714 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4715 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4717 bctl->flags = bargs->flags;
4719 /* balance everything - no filters */
4720 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4723 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4730 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP
4731 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4732 * or, if restriper was paused all the way until unmount, in
4733 * free_fs_info. The flag is cleared in __cancel_balance.
4735 need_unlock = false;
4737 ret = btrfs_balance(bctl, bargs);
4741 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4750 mutex_unlock(&fs_info->balance_mutex);
4751 mutex_unlock(&fs_info->volume_mutex);
4753 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4755 mnt_drop_write_file(file);
4759 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4761 if (!capable(CAP_SYS_ADMIN))
4765 case BTRFS_BALANCE_CTL_PAUSE:
4766 return btrfs_pause_balance(fs_info);
4767 case BTRFS_BALANCE_CTL_CANCEL:
4768 return btrfs_cancel_balance(fs_info);
4774 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4777 struct btrfs_ioctl_balance_args *bargs;
4780 if (!capable(CAP_SYS_ADMIN))
4783 mutex_lock(&fs_info->balance_mutex);
4784 if (!fs_info->balance_ctl) {
4789 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4795 update_ioctl_balance_args(fs_info, 1, bargs);
4797 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4802 mutex_unlock(&fs_info->balance_mutex);
4806 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4808 struct inode *inode = file_inode(file);
4809 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4810 struct btrfs_ioctl_quota_ctl_args *sa;
4811 struct btrfs_trans_handle *trans = NULL;
4815 if (!capable(CAP_SYS_ADMIN))
4818 ret = mnt_want_write_file(file);
4822 sa = memdup_user(arg, sizeof(*sa));
4828 down_write(&fs_info->subvol_sem);
4829 trans = btrfs_start_transaction(fs_info->tree_root, 2);
4830 if (IS_ERR(trans)) {
4831 ret = PTR_ERR(trans);
4836 case BTRFS_QUOTA_CTL_ENABLE:
4837 ret = btrfs_quota_enable(trans, fs_info);
4839 case BTRFS_QUOTA_CTL_DISABLE:
4840 ret = btrfs_quota_disable(trans, fs_info);
4847 err = btrfs_commit_transaction(trans);
4852 up_write(&fs_info->subvol_sem);
4854 mnt_drop_write_file(file);
4858 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4860 struct inode *inode = file_inode(file);
4861 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4862 struct btrfs_root *root = BTRFS_I(inode)->root;
4863 struct btrfs_ioctl_qgroup_assign_args *sa;
4864 struct btrfs_trans_handle *trans;
4868 if (!capable(CAP_SYS_ADMIN))
4871 ret = mnt_want_write_file(file);
4875 sa = memdup_user(arg, sizeof(*sa));
4881 trans = btrfs_join_transaction(root);
4882 if (IS_ERR(trans)) {
4883 ret = PTR_ERR(trans);
4888 ret = btrfs_add_qgroup_relation(trans, fs_info,
4891 ret = btrfs_del_qgroup_relation(trans, fs_info,
4895 /* update qgroup status and info */
4896 err = btrfs_run_qgroups(trans, fs_info);
4898 btrfs_handle_fs_error(fs_info, err,
4899 "failed to update qgroup status and info");
4900 err = btrfs_end_transaction(trans);
4907 mnt_drop_write_file(file);
4911 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4913 struct inode *inode = file_inode(file);
4914 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4915 struct btrfs_root *root = BTRFS_I(inode)->root;
4916 struct btrfs_ioctl_qgroup_create_args *sa;
4917 struct btrfs_trans_handle *trans;
4921 if (!capable(CAP_SYS_ADMIN))
4924 ret = mnt_want_write_file(file);
4928 sa = memdup_user(arg, sizeof(*sa));
4934 if (!sa->qgroupid) {
4939 trans = btrfs_join_transaction(root);
4940 if (IS_ERR(trans)) {
4941 ret = PTR_ERR(trans);
4946 ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid);
4948 ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid);
4951 err = btrfs_end_transaction(trans);
4958 mnt_drop_write_file(file);
4962 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4964 struct inode *inode = file_inode(file);
4965 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4966 struct btrfs_root *root = BTRFS_I(inode)->root;
4967 struct btrfs_ioctl_qgroup_limit_args *sa;
4968 struct btrfs_trans_handle *trans;
4973 if (!capable(CAP_SYS_ADMIN))
4976 ret = mnt_want_write_file(file);
4980 sa = memdup_user(arg, sizeof(*sa));
4986 trans = btrfs_join_transaction(root);
4987 if (IS_ERR(trans)) {
4988 ret = PTR_ERR(trans);
4992 qgroupid = sa->qgroupid;
4994 /* take the current subvol as qgroup */
4995 qgroupid = root->root_key.objectid;
4998 ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim);
5000 err = btrfs_end_transaction(trans);
5007 mnt_drop_write_file(file);
5011 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
5013 struct inode *inode = file_inode(file);
5014 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5015 struct btrfs_ioctl_quota_rescan_args *qsa;
5018 if (!capable(CAP_SYS_ADMIN))
5021 ret = mnt_want_write_file(file);
5025 qsa = memdup_user(arg, sizeof(*qsa));
5036 ret = btrfs_qgroup_rescan(fs_info);
5041 mnt_drop_write_file(file);
5045 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5047 struct inode *inode = file_inode(file);
5048 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5049 struct btrfs_ioctl_quota_rescan_args *qsa;
5052 if (!capable(CAP_SYS_ADMIN))
5055 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
5059 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5061 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
5064 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5071 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5073 struct inode *inode = file_inode(file);
5074 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5076 if (!capable(CAP_SYS_ADMIN))
5079 return btrfs_qgroup_wait_for_completion(fs_info, true);
5082 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5083 struct btrfs_ioctl_received_subvol_args *sa)
5085 struct inode *inode = file_inode(file);
5086 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5087 struct btrfs_root *root = BTRFS_I(inode)->root;
5088 struct btrfs_root_item *root_item = &root->root_item;
5089 struct btrfs_trans_handle *trans;
5090 struct timespec ct = current_time(inode);
5092 int received_uuid_changed;
5094 if (!inode_owner_or_capable(inode))
5097 ret = mnt_want_write_file(file);
5101 down_write(&fs_info->subvol_sem);
5103 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
5108 if (btrfs_root_readonly(root)) {
5115 * 2 - uuid items (received uuid + subvol uuid)
5117 trans = btrfs_start_transaction(root, 3);
5118 if (IS_ERR(trans)) {
5119 ret = PTR_ERR(trans);
5124 sa->rtransid = trans->transid;
5125 sa->rtime.sec = ct.tv_sec;
5126 sa->rtime.nsec = ct.tv_nsec;
5128 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5130 if (received_uuid_changed &&
5131 !btrfs_is_empty_uuid(root_item->received_uuid))
5132 btrfs_uuid_tree_rem(trans, fs_info, root_item->received_uuid,
5133 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5134 root->root_key.objectid);
5135 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5136 btrfs_set_root_stransid(root_item, sa->stransid);
5137 btrfs_set_root_rtransid(root_item, sa->rtransid);
5138 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5139 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5140 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5141 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5143 ret = btrfs_update_root(trans, fs_info->tree_root,
5144 &root->root_key, &root->root_item);
5146 btrfs_end_transaction(trans);
5149 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5150 ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid,
5151 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5152 root->root_key.objectid);
5153 if (ret < 0 && ret != -EEXIST) {
5154 btrfs_abort_transaction(trans, ret);
5158 ret = btrfs_commit_transaction(trans);
5160 btrfs_abort_transaction(trans, ret);
5165 up_write(&fs_info->subvol_sem);
5166 mnt_drop_write_file(file);
5171 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5174 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5175 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5178 args32 = memdup_user(arg, sizeof(*args32));
5180 return PTR_ERR(args32);
5182 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5188 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5189 args64->stransid = args32->stransid;
5190 args64->rtransid = args32->rtransid;
5191 args64->stime.sec = args32->stime.sec;
5192 args64->stime.nsec = args32->stime.nsec;
5193 args64->rtime.sec = args32->rtime.sec;
5194 args64->rtime.nsec = args32->rtime.nsec;
5195 args64->flags = args32->flags;
5197 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5201 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5202 args32->stransid = args64->stransid;
5203 args32->rtransid = args64->rtransid;
5204 args32->stime.sec = args64->stime.sec;
5205 args32->stime.nsec = args64->stime.nsec;
5206 args32->rtime.sec = args64->rtime.sec;
5207 args32->rtime.nsec = args64->rtime.nsec;
5208 args32->flags = args64->flags;
5210 ret = copy_to_user(arg, args32, sizeof(*args32));
5221 static long btrfs_ioctl_set_received_subvol(struct file *file,
5224 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5227 sa = memdup_user(arg, sizeof(*sa));
5231 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5236 ret = copy_to_user(arg, sa, sizeof(*sa));
5245 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5247 struct inode *inode = file_inode(file);
5248 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5251 char label[BTRFS_LABEL_SIZE];
5253 spin_lock(&fs_info->super_lock);
5254 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5255 spin_unlock(&fs_info->super_lock);
5257 len = strnlen(label, BTRFS_LABEL_SIZE);
5259 if (len == BTRFS_LABEL_SIZE) {
5261 "label is too long, return the first %zu bytes",
5265 ret = copy_to_user(arg, label, len);
5267 return ret ? -EFAULT : 0;
5270 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5272 struct inode *inode = file_inode(file);
5273 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5274 struct btrfs_root *root = BTRFS_I(inode)->root;
5275 struct btrfs_super_block *super_block = fs_info->super_copy;
5276 struct btrfs_trans_handle *trans;
5277 char label[BTRFS_LABEL_SIZE];
5280 if (!capable(CAP_SYS_ADMIN))
5283 if (copy_from_user(label, arg, sizeof(label)))
5286 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5288 "unable to set label with more than %d bytes",
5289 BTRFS_LABEL_SIZE - 1);
5293 ret = mnt_want_write_file(file);
5297 trans = btrfs_start_transaction(root, 0);
5298 if (IS_ERR(trans)) {
5299 ret = PTR_ERR(trans);
5303 spin_lock(&fs_info->super_lock);
5304 strcpy(super_block->label, label);
5305 spin_unlock(&fs_info->super_lock);
5306 ret = btrfs_commit_transaction(trans);
5309 mnt_drop_write_file(file);
5313 #define INIT_FEATURE_FLAGS(suffix) \
5314 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5315 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5316 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5318 int btrfs_ioctl_get_supported_features(void __user *arg)
5320 static const struct btrfs_ioctl_feature_flags features[3] = {
5321 INIT_FEATURE_FLAGS(SUPP),
5322 INIT_FEATURE_FLAGS(SAFE_SET),
5323 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5326 if (copy_to_user(arg, &features, sizeof(features)))
5332 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5334 struct inode *inode = file_inode(file);
5335 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5336 struct btrfs_super_block *super_block = fs_info->super_copy;
5337 struct btrfs_ioctl_feature_flags features;
5339 features.compat_flags = btrfs_super_compat_flags(super_block);
5340 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5341 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5343 if (copy_to_user(arg, &features, sizeof(features)))
5349 static int check_feature_bits(struct btrfs_fs_info *fs_info,
5350 enum btrfs_feature_set set,
5351 u64 change_mask, u64 flags, u64 supported_flags,
5352 u64 safe_set, u64 safe_clear)
5354 const char *type = btrfs_feature_set_names[set];
5356 u64 disallowed, unsupported;
5357 u64 set_mask = flags & change_mask;
5358 u64 clear_mask = ~flags & change_mask;
5360 unsupported = set_mask & ~supported_flags;
5362 names = btrfs_printable_features(set, unsupported);
5365 "this kernel does not support the %s feature bit%s",
5366 names, strchr(names, ',') ? "s" : "");
5370 "this kernel does not support %s bits 0x%llx",
5375 disallowed = set_mask & ~safe_set;
5377 names = btrfs_printable_features(set, disallowed);
5380 "can't set the %s feature bit%s while mounted",
5381 names, strchr(names, ',') ? "s" : "");
5385 "can't set %s bits 0x%llx while mounted",
5390 disallowed = clear_mask & ~safe_clear;
5392 names = btrfs_printable_features(set, disallowed);
5395 "can't clear the %s feature bit%s while mounted",
5396 names, strchr(names, ',') ? "s" : "");
5400 "can't clear %s bits 0x%llx while mounted",
5408 #define check_feature(fs_info, change_mask, flags, mask_base) \
5409 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5410 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5411 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5412 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5414 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5416 struct inode *inode = file_inode(file);
5417 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5418 struct btrfs_root *root = BTRFS_I(inode)->root;
5419 struct btrfs_super_block *super_block = fs_info->super_copy;
5420 struct btrfs_ioctl_feature_flags flags[2];
5421 struct btrfs_trans_handle *trans;
5425 if (!capable(CAP_SYS_ADMIN))
5428 if (copy_from_user(flags, arg, sizeof(flags)))
5432 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5433 !flags[0].incompat_flags)
5436 ret = check_feature(fs_info, flags[0].compat_flags,
5437 flags[1].compat_flags, COMPAT);
5441 ret = check_feature(fs_info, flags[0].compat_ro_flags,
5442 flags[1].compat_ro_flags, COMPAT_RO);
5446 ret = check_feature(fs_info, flags[0].incompat_flags,
5447 flags[1].incompat_flags, INCOMPAT);
5451 ret = mnt_want_write_file(file);
5455 trans = btrfs_start_transaction(root, 0);
5456 if (IS_ERR(trans)) {
5457 ret = PTR_ERR(trans);
5458 goto out_drop_write;
5461 spin_lock(&fs_info->super_lock);
5462 newflags = btrfs_super_compat_flags(super_block);
5463 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5464 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5465 btrfs_set_super_compat_flags(super_block, newflags);
5467 newflags = btrfs_super_compat_ro_flags(super_block);
5468 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5469 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5470 btrfs_set_super_compat_ro_flags(super_block, newflags);
5472 newflags = btrfs_super_incompat_flags(super_block);
5473 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5474 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5475 btrfs_set_super_incompat_flags(super_block, newflags);
5476 spin_unlock(&fs_info->super_lock);
5478 ret = btrfs_commit_transaction(trans);
5480 mnt_drop_write_file(file);
5485 long btrfs_ioctl(struct file *file, unsigned int
5486 cmd, unsigned long arg)
5488 struct inode *inode = file_inode(file);
5489 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5490 struct btrfs_root *root = BTRFS_I(inode)->root;
5491 void __user *argp = (void __user *)arg;
5494 case FS_IOC_GETFLAGS:
5495 return btrfs_ioctl_getflags(file, argp);
5496 case FS_IOC_SETFLAGS:
5497 return btrfs_ioctl_setflags(file, argp);
5498 case FS_IOC_GETVERSION:
5499 return btrfs_ioctl_getversion(file, argp);
5501 return btrfs_ioctl_fitrim(file, argp);
5502 case BTRFS_IOC_SNAP_CREATE:
5503 return btrfs_ioctl_snap_create(file, argp, 0);
5504 case BTRFS_IOC_SNAP_CREATE_V2:
5505 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5506 case BTRFS_IOC_SUBVOL_CREATE:
5507 return btrfs_ioctl_snap_create(file, argp, 1);
5508 case BTRFS_IOC_SUBVOL_CREATE_V2:
5509 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5510 case BTRFS_IOC_SNAP_DESTROY:
5511 return btrfs_ioctl_snap_destroy(file, argp);
5512 case BTRFS_IOC_SUBVOL_GETFLAGS:
5513 return btrfs_ioctl_subvol_getflags(file, argp);
5514 case BTRFS_IOC_SUBVOL_SETFLAGS:
5515 return btrfs_ioctl_subvol_setflags(file, argp);
5516 case BTRFS_IOC_DEFAULT_SUBVOL:
5517 return btrfs_ioctl_default_subvol(file, argp);
5518 case BTRFS_IOC_DEFRAG:
5519 return btrfs_ioctl_defrag(file, NULL);
5520 case BTRFS_IOC_DEFRAG_RANGE:
5521 return btrfs_ioctl_defrag(file, argp);
5522 case BTRFS_IOC_RESIZE:
5523 return btrfs_ioctl_resize(file, argp);
5524 case BTRFS_IOC_ADD_DEV:
5525 return btrfs_ioctl_add_dev(fs_info, argp);
5526 case BTRFS_IOC_RM_DEV:
5527 return btrfs_ioctl_rm_dev(file, argp);
5528 case BTRFS_IOC_RM_DEV_V2:
5529 return btrfs_ioctl_rm_dev_v2(file, argp);
5530 case BTRFS_IOC_FS_INFO:
5531 return btrfs_ioctl_fs_info(fs_info, argp);
5532 case BTRFS_IOC_DEV_INFO:
5533 return btrfs_ioctl_dev_info(fs_info, argp);
5534 case BTRFS_IOC_BALANCE:
5535 return btrfs_ioctl_balance(file, NULL);
5536 case BTRFS_IOC_TRANS_START:
5537 return btrfs_ioctl_trans_start(file);
5538 case BTRFS_IOC_TRANS_END:
5539 return btrfs_ioctl_trans_end(file);
5540 case BTRFS_IOC_TREE_SEARCH:
5541 return btrfs_ioctl_tree_search(file, argp);
5542 case BTRFS_IOC_TREE_SEARCH_V2:
5543 return btrfs_ioctl_tree_search_v2(file, argp);
5544 case BTRFS_IOC_INO_LOOKUP:
5545 return btrfs_ioctl_ino_lookup(file, argp);
5546 case BTRFS_IOC_INO_PATHS:
5547 return btrfs_ioctl_ino_to_path(root, argp);
5548 case BTRFS_IOC_LOGICAL_INO:
5549 return btrfs_ioctl_logical_to_ino(fs_info, argp);
5550 case BTRFS_IOC_SPACE_INFO:
5551 return btrfs_ioctl_space_info(fs_info, argp);
5552 case BTRFS_IOC_SYNC: {
5555 ret = btrfs_start_delalloc_roots(fs_info, 0, -1);
5558 ret = btrfs_sync_fs(inode->i_sb, 1);
5560 * The transaction thread may want to do more work,
5561 * namely it pokes the cleaner kthread that will start
5562 * processing uncleaned subvols.
5564 wake_up_process(fs_info->transaction_kthread);
5567 case BTRFS_IOC_START_SYNC:
5568 return btrfs_ioctl_start_sync(root, argp);
5569 case BTRFS_IOC_WAIT_SYNC:
5570 return btrfs_ioctl_wait_sync(fs_info, argp);
5571 case BTRFS_IOC_SCRUB:
5572 return btrfs_ioctl_scrub(file, argp);
5573 case BTRFS_IOC_SCRUB_CANCEL:
5574 return btrfs_ioctl_scrub_cancel(fs_info);
5575 case BTRFS_IOC_SCRUB_PROGRESS:
5576 return btrfs_ioctl_scrub_progress(fs_info, argp);
5577 case BTRFS_IOC_BALANCE_V2:
5578 return btrfs_ioctl_balance(file, argp);
5579 case BTRFS_IOC_BALANCE_CTL:
5580 return btrfs_ioctl_balance_ctl(fs_info, arg);
5581 case BTRFS_IOC_BALANCE_PROGRESS:
5582 return btrfs_ioctl_balance_progress(fs_info, argp);
5583 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5584 return btrfs_ioctl_set_received_subvol(file, argp);
5586 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5587 return btrfs_ioctl_set_received_subvol_32(file, argp);
5589 case BTRFS_IOC_SEND:
5590 return btrfs_ioctl_send(file, argp);
5591 case BTRFS_IOC_GET_DEV_STATS:
5592 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5593 case BTRFS_IOC_QUOTA_CTL:
5594 return btrfs_ioctl_quota_ctl(file, argp);
5595 case BTRFS_IOC_QGROUP_ASSIGN:
5596 return btrfs_ioctl_qgroup_assign(file, argp);
5597 case BTRFS_IOC_QGROUP_CREATE:
5598 return btrfs_ioctl_qgroup_create(file, argp);
5599 case BTRFS_IOC_QGROUP_LIMIT:
5600 return btrfs_ioctl_qgroup_limit(file, argp);
5601 case BTRFS_IOC_QUOTA_RESCAN:
5602 return btrfs_ioctl_quota_rescan(file, argp);
5603 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5604 return btrfs_ioctl_quota_rescan_status(file, argp);
5605 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5606 return btrfs_ioctl_quota_rescan_wait(file, argp);
5607 case BTRFS_IOC_DEV_REPLACE:
5608 return btrfs_ioctl_dev_replace(fs_info, argp);
5609 case BTRFS_IOC_GET_FSLABEL:
5610 return btrfs_ioctl_get_fslabel(file, argp);
5611 case BTRFS_IOC_SET_FSLABEL:
5612 return btrfs_ioctl_set_fslabel(file, argp);
5613 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5614 return btrfs_ioctl_get_supported_features(argp);
5615 case BTRFS_IOC_GET_FEATURES:
5616 return btrfs_ioctl_get_features(file, argp);
5617 case BTRFS_IOC_SET_FEATURES:
5618 return btrfs_ioctl_set_features(file, argp);
5624 #ifdef CONFIG_COMPAT
5625 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5628 * These all access 32-bit values anyway so no further
5629 * handling is necessary.
5632 case FS_IOC32_GETFLAGS:
5633 cmd = FS_IOC_GETFLAGS;
5635 case FS_IOC32_SETFLAGS:
5636 cmd = FS_IOC_SETFLAGS;
5638 case FS_IOC32_GETVERSION:
5639 cmd = FS_IOC_GETVERSION;
5643 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));