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>
46 #include <linux/iversion.h>
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
54 #include "inode-map.h"
56 #include "rcu-string.h"
58 #include "dev-replace.h"
63 #include "compression.h"
66 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
67 * structures are incorrect, as the timespec structure from userspace
68 * is 4 bytes too small. We define these alternatives here to teach
69 * the kernel about the 32-bit struct packing.
71 struct btrfs_ioctl_timespec_32 {
74 } __attribute__ ((__packed__));
76 struct btrfs_ioctl_received_subvol_args_32 {
77 char uuid[BTRFS_UUID_SIZE]; /* in */
78 __u64 stransid; /* in */
79 __u64 rtransid; /* out */
80 struct btrfs_ioctl_timespec_32 stime; /* in */
81 struct btrfs_ioctl_timespec_32 rtime; /* out */
83 __u64 reserved[16]; /* in */
84 } __attribute__ ((__packed__));
86 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
87 struct btrfs_ioctl_received_subvol_args_32)
90 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
91 struct btrfs_ioctl_send_args_32 {
92 __s64 send_fd; /* in */
93 __u64 clone_sources_count; /* in */
94 compat_uptr_t clone_sources; /* in */
95 __u64 parent_root; /* in */
97 __u64 reserved[4]; /* in */
98 } __attribute__ ((__packed__));
100 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
101 struct btrfs_ioctl_send_args_32)
104 static int btrfs_clone(struct inode *src, struct inode *inode,
105 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
108 /* Mask out flags that are inappropriate for the given type of inode. */
109 static unsigned int btrfs_mask_flags(umode_t mode, unsigned int flags)
113 else if (S_ISREG(mode))
114 return flags & ~FS_DIRSYNC_FL;
116 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
120 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
122 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
124 unsigned int iflags = 0;
126 if (flags & BTRFS_INODE_SYNC)
127 iflags |= FS_SYNC_FL;
128 if (flags & BTRFS_INODE_IMMUTABLE)
129 iflags |= FS_IMMUTABLE_FL;
130 if (flags & BTRFS_INODE_APPEND)
131 iflags |= FS_APPEND_FL;
132 if (flags & BTRFS_INODE_NODUMP)
133 iflags |= FS_NODUMP_FL;
134 if (flags & BTRFS_INODE_NOATIME)
135 iflags |= FS_NOATIME_FL;
136 if (flags & BTRFS_INODE_DIRSYNC)
137 iflags |= FS_DIRSYNC_FL;
138 if (flags & BTRFS_INODE_NODATACOW)
139 iflags |= FS_NOCOW_FL;
141 if (flags & BTRFS_INODE_NOCOMPRESS)
142 iflags |= FS_NOCOMP_FL;
143 else if (flags & BTRFS_INODE_COMPRESS)
144 iflags |= FS_COMPR_FL;
150 * Update inode->i_flags based on the btrfs internal flags.
152 void btrfs_update_iflags(struct inode *inode)
154 struct btrfs_inode *ip = BTRFS_I(inode);
155 unsigned int new_fl = 0;
157 if (ip->flags & BTRFS_INODE_SYNC)
159 if (ip->flags & BTRFS_INODE_IMMUTABLE)
160 new_fl |= S_IMMUTABLE;
161 if (ip->flags & BTRFS_INODE_APPEND)
163 if (ip->flags & BTRFS_INODE_NOATIME)
165 if (ip->flags & BTRFS_INODE_DIRSYNC)
168 set_mask_bits(&inode->i_flags,
169 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
173 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
175 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
176 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
178 if (copy_to_user(arg, &flags, sizeof(flags)))
183 static int check_flags(unsigned int flags)
185 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
186 FS_NOATIME_FL | FS_NODUMP_FL | \
187 FS_SYNC_FL | FS_DIRSYNC_FL | \
188 FS_NOCOMP_FL | FS_COMPR_FL |
192 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
198 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
200 struct inode *inode = file_inode(file);
201 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
202 struct btrfs_inode *ip = BTRFS_I(inode);
203 struct btrfs_root *root = ip->root;
204 struct btrfs_trans_handle *trans;
205 unsigned int flags, oldflags;
208 unsigned int i_oldflags;
211 if (!inode_owner_or_capable(inode))
214 if (btrfs_root_readonly(root))
217 if (copy_from_user(&flags, arg, sizeof(flags)))
220 ret = check_flags(flags);
224 ret = mnt_want_write_file(file);
230 ip_oldflags = ip->flags;
231 i_oldflags = inode->i_flags;
232 mode = inode->i_mode;
234 flags = btrfs_mask_flags(inode->i_mode, flags);
235 oldflags = btrfs_flags_to_ioctl(ip->flags);
236 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
237 if (!capable(CAP_LINUX_IMMUTABLE)) {
243 if (flags & FS_SYNC_FL)
244 ip->flags |= BTRFS_INODE_SYNC;
246 ip->flags &= ~BTRFS_INODE_SYNC;
247 if (flags & FS_IMMUTABLE_FL)
248 ip->flags |= BTRFS_INODE_IMMUTABLE;
250 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
251 if (flags & FS_APPEND_FL)
252 ip->flags |= BTRFS_INODE_APPEND;
254 ip->flags &= ~BTRFS_INODE_APPEND;
255 if (flags & FS_NODUMP_FL)
256 ip->flags |= BTRFS_INODE_NODUMP;
258 ip->flags &= ~BTRFS_INODE_NODUMP;
259 if (flags & FS_NOATIME_FL)
260 ip->flags |= BTRFS_INODE_NOATIME;
262 ip->flags &= ~BTRFS_INODE_NOATIME;
263 if (flags & FS_DIRSYNC_FL)
264 ip->flags |= BTRFS_INODE_DIRSYNC;
266 ip->flags &= ~BTRFS_INODE_DIRSYNC;
267 if (flags & FS_NOCOW_FL) {
270 * It's safe to turn csums off here, no extents exist.
271 * Otherwise we want the flag to reflect the real COW
272 * status of the file and will not set it.
274 if (inode->i_size == 0)
275 ip->flags |= BTRFS_INODE_NODATACOW
276 | BTRFS_INODE_NODATASUM;
278 ip->flags |= BTRFS_INODE_NODATACOW;
282 * Revert back under same assumptions as above
285 if (inode->i_size == 0)
286 ip->flags &= ~(BTRFS_INODE_NODATACOW
287 | BTRFS_INODE_NODATASUM);
289 ip->flags &= ~BTRFS_INODE_NODATACOW;
294 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
295 * flag may be changed automatically if compression code won't make
298 if (flags & FS_NOCOMP_FL) {
299 ip->flags &= ~BTRFS_INODE_COMPRESS;
300 ip->flags |= BTRFS_INODE_NOCOMPRESS;
302 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
303 if (ret && ret != -ENODATA)
305 } else if (flags & FS_COMPR_FL) {
308 ip->flags |= BTRFS_INODE_COMPRESS;
309 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
311 comp = btrfs_compress_type2str(fs_info->compress_type);
312 if (!comp || comp[0] == 0)
313 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
315 ret = btrfs_set_prop(inode, "btrfs.compression",
316 comp, strlen(comp), 0);
321 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
322 if (ret && ret != -ENODATA)
324 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
327 trans = btrfs_start_transaction(root, 1);
329 ret = PTR_ERR(trans);
333 btrfs_update_iflags(inode);
334 inode_inc_iversion(inode);
335 inode->i_ctime = current_time(inode);
336 ret = btrfs_update_inode(trans, root, inode);
338 btrfs_end_transaction(trans);
341 ip->flags = ip_oldflags;
342 inode->i_flags = i_oldflags;
347 mnt_drop_write_file(file);
351 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
353 struct inode *inode = file_inode(file);
355 return put_user(inode->i_generation, arg);
358 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
360 struct inode *inode = file_inode(file);
361 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
362 struct btrfs_device *device;
363 struct request_queue *q;
364 struct fstrim_range range;
365 u64 minlen = ULLONG_MAX;
367 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
370 if (!capable(CAP_SYS_ADMIN))
374 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
378 q = bdev_get_queue(device->bdev);
379 if (blk_queue_discard(q)) {
381 minlen = min_t(u64, q->limits.discard_granularity,
389 if (copy_from_user(&range, arg, sizeof(range)))
391 if (range.start > total_bytes ||
392 range.len < fs_info->sb->s_blocksize)
395 range.len = min(range.len, total_bytes - range.start);
396 range.minlen = max(range.minlen, minlen);
397 ret = btrfs_trim_fs(fs_info, &range);
401 if (copy_to_user(arg, &range, sizeof(range)))
407 int btrfs_is_empty_uuid(u8 *uuid)
411 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
418 static noinline int create_subvol(struct inode *dir,
419 struct dentry *dentry,
420 const char *name, int namelen,
422 struct btrfs_qgroup_inherit *inherit)
424 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
425 struct btrfs_trans_handle *trans;
426 struct btrfs_key key;
427 struct btrfs_root_item *root_item;
428 struct btrfs_inode_item *inode_item;
429 struct extent_buffer *leaf;
430 struct btrfs_root *root = BTRFS_I(dir)->root;
431 struct btrfs_root *new_root;
432 struct btrfs_block_rsv block_rsv;
433 struct timespec cur_time = current_time(dir);
438 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
443 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
447 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
452 * Don't create subvolume whose level is not zero. Or qgroup will be
453 * screwed up since it assumes subvolume qgroup's level to be 0.
455 if (btrfs_qgroup_level(objectid)) {
460 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
462 * The same as the snapshot creation, please see the comment
463 * of create_snapshot().
465 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
466 8, &qgroup_reserved, false);
470 trans = btrfs_start_transaction(root, 0);
472 ret = PTR_ERR(trans);
473 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
476 trans->block_rsv = &block_rsv;
477 trans->bytes_reserved = block_rsv.size;
479 ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit);
483 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
489 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
490 btrfs_set_header_bytenr(leaf, leaf->start);
491 btrfs_set_header_generation(leaf, trans->transid);
492 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
493 btrfs_set_header_owner(leaf, objectid);
495 write_extent_buffer_fsid(leaf, fs_info->fsid);
496 write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid);
497 btrfs_mark_buffer_dirty(leaf);
499 inode_item = &root_item->inode;
500 btrfs_set_stack_inode_generation(inode_item, 1);
501 btrfs_set_stack_inode_size(inode_item, 3);
502 btrfs_set_stack_inode_nlink(inode_item, 1);
503 btrfs_set_stack_inode_nbytes(inode_item,
505 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
507 btrfs_set_root_flags(root_item, 0);
508 btrfs_set_root_limit(root_item, 0);
509 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
511 btrfs_set_root_bytenr(root_item, leaf->start);
512 btrfs_set_root_generation(root_item, trans->transid);
513 btrfs_set_root_level(root_item, 0);
514 btrfs_set_root_refs(root_item, 1);
515 btrfs_set_root_used(root_item, leaf->len);
516 btrfs_set_root_last_snapshot(root_item, 0);
518 btrfs_set_root_generation_v2(root_item,
519 btrfs_root_generation(root_item));
520 uuid_le_gen(&new_uuid);
521 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
522 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
523 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
524 root_item->ctime = root_item->otime;
525 btrfs_set_root_ctransid(root_item, trans->transid);
526 btrfs_set_root_otransid(root_item, trans->transid);
528 btrfs_tree_unlock(leaf);
529 free_extent_buffer(leaf);
532 btrfs_set_root_dirid(root_item, new_dirid);
534 key.objectid = objectid;
536 key.type = BTRFS_ROOT_ITEM_KEY;
537 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
542 key.offset = (u64)-1;
543 new_root = btrfs_read_fs_root_no_name(fs_info, &key);
544 if (IS_ERR(new_root)) {
545 ret = PTR_ERR(new_root);
546 btrfs_abort_transaction(trans, ret);
550 btrfs_record_root_in_trans(trans, new_root);
552 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
554 /* We potentially lose an unused inode item here */
555 btrfs_abort_transaction(trans, ret);
559 mutex_lock(&new_root->objectid_mutex);
560 new_root->highest_objectid = new_dirid;
561 mutex_unlock(&new_root->objectid_mutex);
564 * insert the directory item
566 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
568 btrfs_abort_transaction(trans, ret);
572 ret = btrfs_insert_dir_item(trans, root,
573 name, namelen, BTRFS_I(dir), &key,
574 BTRFS_FT_DIR, index);
576 btrfs_abort_transaction(trans, ret);
580 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
581 ret = btrfs_update_inode(trans, root, dir);
584 ret = btrfs_add_root_ref(trans, fs_info,
585 objectid, root->root_key.objectid,
586 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
589 ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid,
590 BTRFS_UUID_KEY_SUBVOL, objectid);
592 btrfs_abort_transaction(trans, ret);
596 trans->block_rsv = NULL;
597 trans->bytes_reserved = 0;
598 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
601 *async_transid = trans->transid;
602 err = btrfs_commit_transaction_async(trans, 1);
604 err = btrfs_commit_transaction(trans);
606 err = btrfs_commit_transaction(trans);
612 inode = btrfs_lookup_dentry(dir, dentry);
614 return PTR_ERR(inode);
615 d_instantiate(dentry, inode);
624 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
625 struct dentry *dentry,
626 u64 *async_transid, bool readonly,
627 struct btrfs_qgroup_inherit *inherit)
629 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
631 struct btrfs_pending_snapshot *pending_snapshot;
632 struct btrfs_trans_handle *trans;
635 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
638 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
639 if (!pending_snapshot)
642 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
644 pending_snapshot->path = btrfs_alloc_path();
645 if (!pending_snapshot->root_item || !pending_snapshot->path) {
650 atomic_inc(&root->will_be_snapshotted);
651 smp_mb__after_atomic();
652 /* wait for no snapshot writes */
653 wait_event(root->subv_writers->wait,
654 percpu_counter_sum(&root->subv_writers->counter) == 0);
656 ret = btrfs_start_delalloc_inodes(root, 0);
660 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
662 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
663 BTRFS_BLOCK_RSV_TEMP);
665 * 1 - parent dir inode
668 * 2 - root ref/backref
669 * 1 - root of snapshot
672 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
673 &pending_snapshot->block_rsv, 8,
674 &pending_snapshot->qgroup_reserved,
679 pending_snapshot->dentry = dentry;
680 pending_snapshot->root = root;
681 pending_snapshot->readonly = readonly;
682 pending_snapshot->dir = dir;
683 pending_snapshot->inherit = inherit;
685 trans = btrfs_start_transaction(root, 0);
687 ret = PTR_ERR(trans);
691 spin_lock(&fs_info->trans_lock);
692 list_add(&pending_snapshot->list,
693 &trans->transaction->pending_snapshots);
694 spin_unlock(&fs_info->trans_lock);
696 *async_transid = trans->transid;
697 ret = btrfs_commit_transaction_async(trans, 1);
699 ret = btrfs_commit_transaction(trans);
701 ret = btrfs_commit_transaction(trans);
706 ret = pending_snapshot->error;
710 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
714 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
716 ret = PTR_ERR(inode);
720 d_instantiate(dentry, inode);
723 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
725 if (atomic_dec_and_test(&root->will_be_snapshotted))
726 wake_up_var(&root->will_be_snapshotted);
728 kfree(pending_snapshot->root_item);
729 btrfs_free_path(pending_snapshot->path);
730 kfree(pending_snapshot);
735 /* copy of may_delete in fs/namei.c()
736 * Check whether we can remove a link victim from directory dir, check
737 * whether the type of victim is right.
738 * 1. We can't do it if dir is read-only (done in permission())
739 * 2. We should have write and exec permissions on dir
740 * 3. We can't remove anything from append-only dir
741 * 4. We can't do anything with immutable dir (done in permission())
742 * 5. If the sticky bit on dir is set we should either
743 * a. be owner of dir, or
744 * b. be owner of victim, or
745 * c. have CAP_FOWNER capability
746 * 6. If the victim is append-only or immutable we can't do anything with
747 * links pointing to it.
748 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
749 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
750 * 9. We can't remove a root or mountpoint.
751 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
752 * nfs_async_unlink().
755 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
759 if (d_really_is_negative(victim))
762 BUG_ON(d_inode(victim->d_parent) != dir);
763 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
765 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
770 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
771 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
774 if (!d_is_dir(victim))
778 } else if (d_is_dir(victim))
782 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
787 /* copy of may_create in fs/namei.c() */
788 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
790 if (d_really_is_positive(child))
794 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
798 * Create a new subvolume below @parent. This is largely modeled after
799 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
800 * inside this filesystem so it's quite a bit simpler.
802 static noinline int btrfs_mksubvol(const struct path *parent,
803 const char *name, int namelen,
804 struct btrfs_root *snap_src,
805 u64 *async_transid, bool readonly,
806 struct btrfs_qgroup_inherit *inherit)
808 struct inode *dir = d_inode(parent->dentry);
809 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
810 struct dentry *dentry;
813 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
817 dentry = lookup_one_len(name, parent->dentry, namelen);
818 error = PTR_ERR(dentry);
822 error = btrfs_may_create(dir, dentry);
827 * even if this name doesn't exist, we may get hash collisions.
828 * check for them now when we can safely fail
830 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
836 down_read(&fs_info->subvol_sem);
838 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
842 error = create_snapshot(snap_src, dir, dentry,
843 async_transid, readonly, inherit);
845 error = create_subvol(dir, dentry, name, namelen,
846 async_transid, inherit);
849 fsnotify_mkdir(dir, dentry);
851 up_read(&fs_info->subvol_sem);
860 * When we're defragging a range, we don't want to kick it off again
861 * if it is really just waiting for delalloc to send it down.
862 * If we find a nice big extent or delalloc range for the bytes in the
863 * file you want to defrag, we return 0 to let you know to skip this
866 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
868 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
869 struct extent_map *em = NULL;
870 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
873 read_lock(&em_tree->lock);
874 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
875 read_unlock(&em_tree->lock);
878 end = extent_map_end(em);
880 if (end - offset > thresh)
883 /* if we already have a nice delalloc here, just stop */
885 end = count_range_bits(io_tree, &offset, offset + thresh,
886 thresh, EXTENT_DELALLOC, 1);
893 * helper function to walk through a file and find extents
894 * newer than a specific transid, and smaller than thresh.
896 * This is used by the defragging code to find new and small
899 static int find_new_extents(struct btrfs_root *root,
900 struct inode *inode, u64 newer_than,
901 u64 *off, u32 thresh)
903 struct btrfs_path *path;
904 struct btrfs_key min_key;
905 struct extent_buffer *leaf;
906 struct btrfs_file_extent_item *extent;
909 u64 ino = btrfs_ino(BTRFS_I(inode));
911 path = btrfs_alloc_path();
915 min_key.objectid = ino;
916 min_key.type = BTRFS_EXTENT_DATA_KEY;
917 min_key.offset = *off;
920 ret = btrfs_search_forward(root, &min_key, path, newer_than);
924 if (min_key.objectid != ino)
926 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
929 leaf = path->nodes[0];
930 extent = btrfs_item_ptr(leaf, path->slots[0],
931 struct btrfs_file_extent_item);
933 type = btrfs_file_extent_type(leaf, extent);
934 if (type == BTRFS_FILE_EXTENT_REG &&
935 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
936 check_defrag_in_cache(inode, min_key.offset, thresh)) {
937 *off = min_key.offset;
938 btrfs_free_path(path);
943 if (path->slots[0] < btrfs_header_nritems(leaf)) {
944 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
948 if (min_key.offset == (u64)-1)
952 btrfs_release_path(path);
955 btrfs_free_path(path);
959 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
961 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
962 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
963 struct extent_map *em;
967 * hopefully we have this extent in the tree already, try without
968 * the full extent lock
970 read_lock(&em_tree->lock);
971 em = lookup_extent_mapping(em_tree, start, len);
972 read_unlock(&em_tree->lock);
975 struct extent_state *cached = NULL;
976 u64 end = start + len - 1;
978 /* get the big lock and read metadata off disk */
979 lock_extent_bits(io_tree, start, end, &cached);
980 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0);
981 unlock_extent_cached(io_tree, start, end, &cached);
990 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
992 struct extent_map *next;
995 /* this is the last extent */
996 if (em->start + em->len >= i_size_read(inode))
999 next = defrag_lookup_extent(inode, em->start + em->len);
1000 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1002 else if ((em->block_start + em->block_len == next->block_start) &&
1003 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1006 free_extent_map(next);
1010 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1011 u64 *last_len, u64 *skip, u64 *defrag_end,
1014 struct extent_map *em;
1016 bool next_mergeable = true;
1017 bool prev_mergeable = true;
1020 * make sure that once we start defragging an extent, we keep on
1023 if (start < *defrag_end)
1028 em = defrag_lookup_extent(inode, start);
1032 /* this will cover holes, and inline extents */
1033 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1039 prev_mergeable = false;
1041 next_mergeable = defrag_check_next_extent(inode, em);
1043 * we hit a real extent, if it is big or the next extent is not a
1044 * real extent, don't bother defragging it
1046 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1047 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1051 * last_len ends up being a counter of how many bytes we've defragged.
1052 * every time we choose not to defrag an extent, we reset *last_len
1053 * so that the next tiny extent will force a defrag.
1055 * The end result of this is that tiny extents before a single big
1056 * extent will force at least part of that big extent to be defragged.
1059 *defrag_end = extent_map_end(em);
1062 *skip = extent_map_end(em);
1066 free_extent_map(em);
1071 * it doesn't do much good to defrag one or two pages
1072 * at a time. This pulls in a nice chunk of pages
1073 * to COW and defrag.
1075 * It also makes sure the delalloc code has enough
1076 * dirty data to avoid making new small extents as part
1079 * It's a good idea to start RA on this range
1080 * before calling this.
1082 static int cluster_pages_for_defrag(struct inode *inode,
1083 struct page **pages,
1084 unsigned long start_index,
1085 unsigned long num_pages)
1087 unsigned long file_end;
1088 u64 isize = i_size_read(inode);
1095 struct btrfs_ordered_extent *ordered;
1096 struct extent_state *cached_state = NULL;
1097 struct extent_io_tree *tree;
1098 struct extent_changeset *data_reserved = NULL;
1099 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1101 file_end = (isize - 1) >> PAGE_SHIFT;
1102 if (!isize || start_index > file_end)
1105 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1107 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
1108 start_index << PAGE_SHIFT,
1109 page_cnt << PAGE_SHIFT);
1113 tree = &BTRFS_I(inode)->io_tree;
1115 /* step one, lock all the pages */
1116 for (i = 0; i < page_cnt; i++) {
1119 page = find_or_create_page(inode->i_mapping,
1120 start_index + i, mask);
1124 page_start = page_offset(page);
1125 page_end = page_start + PAGE_SIZE - 1;
1127 lock_extent_bits(tree, page_start, page_end,
1129 ordered = btrfs_lookup_ordered_extent(inode,
1131 unlock_extent_cached(tree, page_start, page_end,
1137 btrfs_start_ordered_extent(inode, ordered, 1);
1138 btrfs_put_ordered_extent(ordered);
1141 * we unlocked the page above, so we need check if
1142 * it was released or not.
1144 if (page->mapping != inode->i_mapping) {
1151 if (!PageUptodate(page)) {
1152 btrfs_readpage(NULL, page);
1154 if (!PageUptodate(page)) {
1162 if (page->mapping != inode->i_mapping) {
1174 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1178 * so now we have a nice long stream of locked
1179 * and up to date pages, lets wait on them
1181 for (i = 0; i < i_done; i++)
1182 wait_on_page_writeback(pages[i]);
1184 page_start = page_offset(pages[0]);
1185 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1187 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1188 page_start, page_end - 1, &cached_state);
1189 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1190 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1191 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1194 if (i_done != page_cnt) {
1195 spin_lock(&BTRFS_I(inode)->lock);
1196 BTRFS_I(inode)->outstanding_extents++;
1197 spin_unlock(&BTRFS_I(inode)->lock);
1198 btrfs_delalloc_release_space(inode, data_reserved,
1199 start_index << PAGE_SHIFT,
1200 (page_cnt - i_done) << PAGE_SHIFT, true);
1204 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1207 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1208 page_start, page_end - 1, &cached_state);
1210 for (i = 0; i < i_done; i++) {
1211 clear_page_dirty_for_io(pages[i]);
1212 ClearPageChecked(pages[i]);
1213 set_page_extent_mapped(pages[i]);
1214 set_page_dirty(pages[i]);
1215 unlock_page(pages[i]);
1218 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1220 extent_changeset_free(data_reserved);
1223 for (i = 0; i < i_done; i++) {
1224 unlock_page(pages[i]);
1227 btrfs_delalloc_release_space(inode, data_reserved,
1228 start_index << PAGE_SHIFT,
1229 page_cnt << PAGE_SHIFT, true);
1230 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1232 extent_changeset_free(data_reserved);
1237 int btrfs_defrag_file(struct inode *inode, struct file *file,
1238 struct btrfs_ioctl_defrag_range_args *range,
1239 u64 newer_than, unsigned long max_to_defrag)
1241 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1242 struct btrfs_root *root = BTRFS_I(inode)->root;
1243 struct file_ra_state *ra = NULL;
1244 unsigned long last_index;
1245 u64 isize = i_size_read(inode);
1249 u64 newer_off = range->start;
1251 unsigned long ra_index = 0;
1253 int defrag_count = 0;
1254 int compress_type = BTRFS_COMPRESS_ZLIB;
1255 u32 extent_thresh = range->extent_thresh;
1256 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1257 unsigned long cluster = max_cluster;
1258 u64 new_align = ~((u64)SZ_128K - 1);
1259 struct page **pages = NULL;
1260 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1265 if (range->start >= isize)
1269 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1271 if (range->compress_type)
1272 compress_type = range->compress_type;
1275 if (extent_thresh == 0)
1276 extent_thresh = SZ_256K;
1279 * If we were not given a file, allocate a readahead context. As
1280 * readahead is just an optimization, defrag will work without it so
1281 * we don't error out.
1284 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1286 file_ra_state_init(ra, inode->i_mapping);
1291 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1297 /* find the last page to defrag */
1298 if (range->start + range->len > range->start) {
1299 last_index = min_t(u64, isize - 1,
1300 range->start + range->len - 1) >> PAGE_SHIFT;
1302 last_index = (isize - 1) >> PAGE_SHIFT;
1306 ret = find_new_extents(root, inode, newer_than,
1307 &newer_off, SZ_64K);
1309 range->start = newer_off;
1311 * we always align our defrag to help keep
1312 * the extents in the file evenly spaced
1314 i = (newer_off & new_align) >> PAGE_SHIFT;
1318 i = range->start >> PAGE_SHIFT;
1321 max_to_defrag = last_index - i + 1;
1324 * make writeback starts from i, so the defrag range can be
1325 * written sequentially.
1327 if (i < inode->i_mapping->writeback_index)
1328 inode->i_mapping->writeback_index = i;
1330 while (i <= last_index && defrag_count < max_to_defrag &&
1331 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1333 * make sure we stop running if someone unmounts
1336 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1339 if (btrfs_defrag_cancelled(fs_info)) {
1340 btrfs_debug(fs_info, "defrag_file cancelled");
1345 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1346 extent_thresh, &last_len, &skip,
1347 &defrag_end, do_compress)){
1350 * the should_defrag function tells us how much to skip
1351 * bump our counter by the suggested amount
1353 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1354 i = max(i + 1, next);
1359 cluster = (PAGE_ALIGN(defrag_end) >>
1361 cluster = min(cluster, max_cluster);
1363 cluster = max_cluster;
1366 if (i + cluster > ra_index) {
1367 ra_index = max(i, ra_index);
1369 page_cache_sync_readahead(inode->i_mapping, ra,
1370 file, ra_index, cluster);
1371 ra_index += cluster;
1376 BTRFS_I(inode)->defrag_compress = compress_type;
1377 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1379 inode_unlock(inode);
1383 defrag_count += ret;
1384 balance_dirty_pages_ratelimited(inode->i_mapping);
1385 inode_unlock(inode);
1388 if (newer_off == (u64)-1)
1394 newer_off = max(newer_off + 1,
1395 (u64)i << PAGE_SHIFT);
1397 ret = find_new_extents(root, inode, newer_than,
1398 &newer_off, SZ_64K);
1400 range->start = newer_off;
1401 i = (newer_off & new_align) >> PAGE_SHIFT;
1408 last_len += ret << PAGE_SHIFT;
1416 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1417 filemap_flush(inode->i_mapping);
1418 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1419 &BTRFS_I(inode)->runtime_flags))
1420 filemap_flush(inode->i_mapping);
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 (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
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 (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
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 - 1];
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,
2605 range, BTRFS_OLDEST_GENERATION, 0);
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 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2679 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
2681 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2682 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2684 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2687 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2688 btrfs_info(fs_info, "device deleted: id %llu",
2691 btrfs_info(fs_info, "device deleted: %s",
2697 mnt_drop_write_file(file);
2701 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2703 struct inode *inode = file_inode(file);
2704 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2705 struct btrfs_ioctl_vol_args *vol_args;
2708 if (!capable(CAP_SYS_ADMIN))
2711 ret = mnt_want_write_file(file);
2715 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2716 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2717 goto out_drop_write;
2720 vol_args = memdup_user(arg, sizeof(*vol_args));
2721 if (IS_ERR(vol_args)) {
2722 ret = PTR_ERR(vol_args);
2726 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2727 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2730 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2733 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2735 mnt_drop_write_file(file);
2740 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2743 struct btrfs_ioctl_fs_info_args *fi_args;
2744 struct btrfs_device *device;
2745 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2748 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2753 fi_args->num_devices = fs_devices->num_devices;
2755 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2756 if (device->devid > fi_args->max_id)
2757 fi_args->max_id = device->devid;
2761 memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid));
2762 fi_args->nodesize = fs_info->nodesize;
2763 fi_args->sectorsize = fs_info->sectorsize;
2764 fi_args->clone_alignment = fs_info->sectorsize;
2766 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2773 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2776 struct btrfs_ioctl_dev_info_args *di_args;
2777 struct btrfs_device *dev;
2779 char *s_uuid = NULL;
2781 di_args = memdup_user(arg, sizeof(*di_args));
2782 if (IS_ERR(di_args))
2783 return PTR_ERR(di_args);
2785 if (!btrfs_is_empty_uuid(di_args->uuid))
2786 s_uuid = di_args->uuid;
2789 dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL);
2796 di_args->devid = dev->devid;
2797 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2798 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2799 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2801 struct rcu_string *name;
2803 name = rcu_dereference(dev->name);
2804 strncpy(di_args->path, name->str, sizeof(di_args->path) - 1);
2805 di_args->path[sizeof(di_args->path) - 1] = 0;
2807 di_args->path[0] = '\0';
2812 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2819 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2823 page = grab_cache_page(inode->i_mapping, index);
2825 return ERR_PTR(-ENOMEM);
2827 if (!PageUptodate(page)) {
2830 ret = btrfs_readpage(NULL, page);
2832 return ERR_PTR(ret);
2834 if (!PageUptodate(page)) {
2837 return ERR_PTR(-EIO);
2839 if (page->mapping != inode->i_mapping) {
2842 return ERR_PTR(-EAGAIN);
2849 static int gather_extent_pages(struct inode *inode, struct page **pages,
2850 int num_pages, u64 off)
2853 pgoff_t index = off >> PAGE_SHIFT;
2855 for (i = 0; i < num_pages; i++) {
2857 pages[i] = extent_same_get_page(inode, index + i);
2858 if (IS_ERR(pages[i])) {
2859 int err = PTR_ERR(pages[i]);
2870 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2871 bool retry_range_locking)
2874 * Do any pending delalloc/csum calculations on inode, one way or
2875 * another, and lock file content.
2876 * The locking order is:
2879 * 2) range in the inode's io tree
2882 struct btrfs_ordered_extent *ordered;
2883 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2884 ordered = btrfs_lookup_first_ordered_extent(inode,
2887 ordered->file_offset + ordered->len <= off ||
2888 ordered->file_offset >= off + len) &&
2889 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2890 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2892 btrfs_put_ordered_extent(ordered);
2895 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2897 btrfs_put_ordered_extent(ordered);
2898 if (!retry_range_locking)
2900 btrfs_wait_ordered_range(inode, off, len);
2905 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2907 inode_unlock(inode1);
2908 inode_unlock(inode2);
2911 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2913 if (inode1 < inode2)
2914 swap(inode1, inode2);
2916 inode_lock_nested(inode1, I_MUTEX_PARENT);
2917 inode_lock_nested(inode2, I_MUTEX_CHILD);
2920 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2921 struct inode *inode2, u64 loff2, u64 len)
2923 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2924 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2927 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2928 struct inode *inode2, u64 loff2, u64 len,
2929 bool retry_range_locking)
2933 if (inode1 < inode2) {
2934 swap(inode1, inode2);
2937 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2940 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2942 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2949 struct page **src_pages;
2950 struct page **dst_pages;
2953 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2958 for (i = 0; i < cmp->num_pages; i++) {
2959 pg = cmp->src_pages[i];
2964 pg = cmp->dst_pages[i];
2970 kfree(cmp->src_pages);
2971 kfree(cmp->dst_pages);
2974 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2975 struct inode *dst, u64 dst_loff,
2976 u64 len, struct cmp_pages *cmp)
2979 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
2980 struct page **src_pgarr, **dst_pgarr;
2983 * We must gather up all the pages before we initiate our
2984 * extent locking. We use an array for the page pointers. Size
2985 * of the array is bounded by len, which is in turn bounded by
2986 * BTRFS_MAX_DEDUPE_LEN.
2988 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2989 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2990 if (!src_pgarr || !dst_pgarr) {
2995 cmp->num_pages = num_pages;
2996 cmp->src_pages = src_pgarr;
2997 cmp->dst_pages = dst_pgarr;
3000 * If deduping ranges in the same inode, locking rules make it mandatory
3001 * to always lock pages in ascending order to avoid deadlocks with
3002 * concurrent tasks (such as starting writeback/delalloc).
3004 if (src == dst && dst_loff < loff) {
3005 swap(src_pgarr, dst_pgarr);
3006 swap(loff, dst_loff);
3009 ret = gather_extent_pages(src, src_pgarr, cmp->num_pages, loff);
3013 ret = gather_extent_pages(dst, dst_pgarr, cmp->num_pages, dst_loff);
3017 btrfs_cmp_data_free(cmp);
3021 static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp)
3025 struct page *src_page, *dst_page;
3026 unsigned int cmp_len = PAGE_SIZE;
3027 void *addr, *dst_addr;
3031 if (len < PAGE_SIZE)
3034 BUG_ON(i >= cmp->num_pages);
3036 src_page = cmp->src_pages[i];
3037 dst_page = cmp->dst_pages[i];
3038 ASSERT(PageLocked(src_page));
3039 ASSERT(PageLocked(dst_page));
3041 addr = kmap_atomic(src_page);
3042 dst_addr = kmap_atomic(dst_page);
3044 flush_dcache_page(src_page);
3045 flush_dcache_page(dst_page);
3047 if (memcmp(addr, dst_addr, cmp_len))
3050 kunmap_atomic(addr);
3051 kunmap_atomic(dst_addr);
3063 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3067 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3069 if (off + olen > inode->i_size || off + olen < off)
3072 /* if we extend to eof, continue to block boundary */
3073 if (off + len == inode->i_size)
3074 *plen = len = ALIGN(inode->i_size, bs) - off;
3076 /* Check that we are block aligned - btrfs_clone() requires this */
3077 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3083 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3084 struct inode *dst, u64 dst_loff)
3088 struct cmp_pages cmp;
3089 bool same_inode = (src == dst);
3090 u64 same_lock_start = 0;
3091 u64 same_lock_len = 0;
3099 btrfs_double_inode_lock(src, dst);
3101 ret = extent_same_check_offsets(src, loff, &len, olen);
3105 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3111 * Single inode case wants the same checks, except we
3112 * don't want our length pushed out past i_size as
3113 * comparing that data range makes no sense.
3115 * extent_same_check_offsets() will do this for an
3116 * unaligned length at i_size, so catch it here and
3117 * reject the request.
3119 * This effectively means we require aligned extents
3120 * for the single-inode case, whereas the other cases
3121 * allow an unaligned length so long as it ends at
3129 /* Check for overlapping ranges */
3130 if (dst_loff + len > loff && dst_loff < loff + len) {
3135 same_lock_start = min_t(u64, loff, dst_loff);
3136 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3139 /* don't make the dst file partly checksummed */
3140 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3141 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3147 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3152 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3155 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3158 * If one of the inodes has dirty pages in the respective range or
3159 * ordered extents, we need to flush dellaloc and wait for all ordered
3160 * extents in the range. We must unlock the pages and the ranges in the
3161 * io trees to avoid deadlocks when flushing delalloc (requires locking
3162 * pages) and when waiting for ordered extents to complete (they require
3165 if (ret == -EAGAIN) {
3167 * Ranges in the io trees already unlocked. Now unlock all
3168 * pages before waiting for all IO to complete.
3170 btrfs_cmp_data_free(&cmp);
3172 btrfs_wait_ordered_range(src, same_lock_start,
3175 btrfs_wait_ordered_range(src, loff, len);
3176 btrfs_wait_ordered_range(dst, dst_loff, len);
3182 /* ranges in the io trees already unlocked */
3183 btrfs_cmp_data_free(&cmp);
3187 /* pass original length for comparison so we stay within i_size */
3188 ret = btrfs_cmp_data(olen, &cmp);
3190 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3193 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3194 same_lock_start + same_lock_len - 1);
3196 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3198 btrfs_cmp_data_free(&cmp);
3203 btrfs_double_inode_unlock(src, dst);
3208 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3210 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3211 struct file *dst_file, u64 dst_loff)
3213 struct inode *src = file_inode(src_file);
3214 struct inode *dst = file_inode(dst_file);
3215 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3218 if (olen > BTRFS_MAX_DEDUPE_LEN)
3219 olen = BTRFS_MAX_DEDUPE_LEN;
3221 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
3223 * Btrfs does not support blocksize < page_size. As a
3224 * result, btrfs_cmp_data() won't correctly handle
3225 * this situation without an update.
3230 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3236 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3237 struct inode *inode,
3243 struct btrfs_root *root = BTRFS_I(inode)->root;
3246 inode_inc_iversion(inode);
3247 if (!no_time_update)
3248 inode->i_mtime = inode->i_ctime = current_time(inode);
3250 * We round up to the block size at eof when determining which
3251 * extents to clone above, but shouldn't round up the file size.
3253 if (endoff > destoff + olen)
3254 endoff = destoff + olen;
3255 if (endoff > inode->i_size)
3256 btrfs_i_size_write(BTRFS_I(inode), endoff);
3258 ret = btrfs_update_inode(trans, root, inode);
3260 btrfs_abort_transaction(trans, ret);
3261 btrfs_end_transaction(trans);
3264 ret = btrfs_end_transaction(trans);
3269 static void clone_update_extent_map(struct btrfs_inode *inode,
3270 const struct btrfs_trans_handle *trans,
3271 const struct btrfs_path *path,
3272 const u64 hole_offset,
3275 struct extent_map_tree *em_tree = &inode->extent_tree;
3276 struct extent_map *em;
3279 em = alloc_extent_map();
3281 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3286 struct btrfs_file_extent_item *fi;
3288 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3289 struct btrfs_file_extent_item);
3290 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3291 em->generation = -1;
3292 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3293 BTRFS_FILE_EXTENT_INLINE)
3294 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3295 &inode->runtime_flags);
3297 em->start = hole_offset;
3299 em->ram_bytes = em->len;
3300 em->orig_start = hole_offset;
3301 em->block_start = EXTENT_MAP_HOLE;
3303 em->orig_block_len = 0;
3304 em->compress_type = BTRFS_COMPRESS_NONE;
3305 em->generation = trans->transid;
3309 write_lock(&em_tree->lock);
3310 ret = add_extent_mapping(em_tree, em, 1);
3311 write_unlock(&em_tree->lock);
3312 if (ret != -EEXIST) {
3313 free_extent_map(em);
3316 btrfs_drop_extent_cache(inode, em->start,
3317 em->start + em->len - 1, 0);
3321 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3325 * Make sure we do not end up inserting an inline extent into a file that has
3326 * already other (non-inline) extents. If a file has an inline extent it can
3327 * not have any other extents and the (single) inline extent must start at the
3328 * file offset 0. Failing to respect these rules will lead to file corruption,
3329 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3331 * We can have extents that have been already written to disk or we can have
3332 * dirty ranges still in delalloc, in which case the extent maps and items are
3333 * created only when we run delalloc, and the delalloc ranges might fall outside
3334 * the range we are currently locking in the inode's io tree. So we check the
3335 * inode's i_size because of that (i_size updates are done while holding the
3336 * i_mutex, which we are holding here).
3337 * We also check to see if the inode has a size not greater than "datal" but has
3338 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3339 * protected against such concurrent fallocate calls by the i_mutex).
3341 * If the file has no extents but a size greater than datal, do not allow the
3342 * copy because we would need turn the inline extent into a non-inline one (even
3343 * with NO_HOLES enabled). If we find our destination inode only has one inline
3344 * extent, just overwrite it with the source inline extent if its size is less
3345 * than the source extent's size, or we could copy the source inline extent's
3346 * data into the destination inode's inline extent if the later is greater then
3349 static int clone_copy_inline_extent(struct inode *dst,
3350 struct btrfs_trans_handle *trans,
3351 struct btrfs_path *path,
3352 struct btrfs_key *new_key,
3353 const u64 drop_start,
3359 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
3360 struct btrfs_root *root = BTRFS_I(dst)->root;
3361 const u64 aligned_end = ALIGN(new_key->offset + datal,
3362 fs_info->sectorsize);
3364 struct btrfs_key key;
3366 if (new_key->offset > 0)
3369 key.objectid = btrfs_ino(BTRFS_I(dst));
3370 key.type = BTRFS_EXTENT_DATA_KEY;
3372 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3375 } else if (ret > 0) {
3376 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3377 ret = btrfs_next_leaf(root, path);
3381 goto copy_inline_extent;
3383 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3384 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3385 key.type == BTRFS_EXTENT_DATA_KEY) {
3386 ASSERT(key.offset > 0);
3389 } else if (i_size_read(dst) <= datal) {
3390 struct btrfs_file_extent_item *ei;
3394 * If the file size is <= datal, make sure there are no other
3395 * extents following (can happen do to an fallocate call with
3396 * the flag FALLOC_FL_KEEP_SIZE).
3398 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3399 struct btrfs_file_extent_item);
3401 * If it's an inline extent, it can not have other extents
3404 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3405 BTRFS_FILE_EXTENT_INLINE)
3406 goto copy_inline_extent;
3408 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3409 if (ext_len > aligned_end)
3412 ret = btrfs_next_item(root, path);
3415 } else if (ret == 0) {
3416 btrfs_item_key_to_cpu(path->nodes[0], &key,
3418 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3419 key.type == BTRFS_EXTENT_DATA_KEY)
3426 * We have no extent items, or we have an extent at offset 0 which may
3427 * or may not be inlined. All these cases are dealt the same way.
3429 if (i_size_read(dst) > datal) {
3431 * If the destination inode has an inline extent...
3432 * This would require copying the data from the source inline
3433 * extent into the beginning of the destination's inline extent.
3434 * But this is really complex, both extents can be compressed
3435 * or just one of them, which would require decompressing and
3436 * re-compressing data (which could increase the new compressed
3437 * size, not allowing the compressed data to fit anymore in an
3439 * So just don't support this case for now (it should be rare,
3440 * we are not really saving space when cloning inline extents).
3445 btrfs_release_path(path);
3446 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3449 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3454 const u32 start = btrfs_file_extent_calc_inline_size(0);
3456 memmove(inline_data + start, inline_data + start + skip, datal);
3459 write_extent_buffer(path->nodes[0], inline_data,
3460 btrfs_item_ptr_offset(path->nodes[0],
3463 inode_add_bytes(dst, datal);
3469 * btrfs_clone() - clone a range from inode file to another
3471 * @src: Inode to clone from
3472 * @inode: Inode to clone to
3473 * @off: Offset within source to start clone from
3474 * @olen: Original length, passed by user, of range to clone
3475 * @olen_aligned: Block-aligned value of olen
3476 * @destoff: Offset within @inode to start clone
3477 * @no_time_update: Whether to update mtime/ctime on the target inode
3479 static int btrfs_clone(struct inode *src, struct inode *inode,
3480 const u64 off, const u64 olen, const u64 olen_aligned,
3481 const u64 destoff, int no_time_update)
3483 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3484 struct btrfs_root *root = BTRFS_I(inode)->root;
3485 struct btrfs_path *path = NULL;
3486 struct extent_buffer *leaf;
3487 struct btrfs_trans_handle *trans;
3489 struct btrfs_key key;
3493 const u64 len = olen_aligned;
3494 u64 last_dest_end = destoff;
3497 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3501 path = btrfs_alloc_path();
3507 path->reada = READA_FORWARD;
3509 key.objectid = btrfs_ino(BTRFS_I(src));
3510 key.type = BTRFS_EXTENT_DATA_KEY;
3514 u64 next_key_min_offset = key.offset + 1;
3517 * note the key will change type as we walk through the
3520 path->leave_spinning = 1;
3521 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3526 * First search, if no extent item that starts at offset off was
3527 * found but the previous item is an extent item, it's possible
3528 * it might overlap our target range, therefore process it.
3530 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3531 btrfs_item_key_to_cpu(path->nodes[0], &key,
3532 path->slots[0] - 1);
3533 if (key.type == BTRFS_EXTENT_DATA_KEY)
3537 nritems = btrfs_header_nritems(path->nodes[0]);
3539 if (path->slots[0] >= nritems) {
3540 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3545 nritems = btrfs_header_nritems(path->nodes[0]);
3547 leaf = path->nodes[0];
3548 slot = path->slots[0];
3550 btrfs_item_key_to_cpu(leaf, &key, slot);
3551 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3552 key.objectid != btrfs_ino(BTRFS_I(src)))
3555 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3556 struct btrfs_file_extent_item *extent;
3559 struct btrfs_key new_key;
3560 u64 disko = 0, diskl = 0;
3561 u64 datao = 0, datal = 0;
3565 extent = btrfs_item_ptr(leaf, slot,
3566 struct btrfs_file_extent_item);
3567 comp = btrfs_file_extent_compression(leaf, extent);
3568 type = btrfs_file_extent_type(leaf, extent);
3569 if (type == BTRFS_FILE_EXTENT_REG ||
3570 type == BTRFS_FILE_EXTENT_PREALLOC) {
3571 disko = btrfs_file_extent_disk_bytenr(leaf,
3573 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3575 datao = btrfs_file_extent_offset(leaf, extent);
3576 datal = btrfs_file_extent_num_bytes(leaf,
3578 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3579 /* take upper bound, may be compressed */
3580 datal = btrfs_file_extent_ram_bytes(leaf,
3585 * The first search might have left us at an extent
3586 * item that ends before our target range's start, can
3587 * happen if we have holes and NO_HOLES feature enabled.
3589 if (key.offset + datal <= off) {
3592 } else if (key.offset >= off + len) {
3595 next_key_min_offset = key.offset + datal;
3596 size = btrfs_item_size_nr(leaf, slot);
3597 read_extent_buffer(leaf, buf,
3598 btrfs_item_ptr_offset(leaf, slot),
3601 btrfs_release_path(path);
3602 path->leave_spinning = 0;
3604 memcpy(&new_key, &key, sizeof(new_key));
3605 new_key.objectid = btrfs_ino(BTRFS_I(inode));
3606 if (off <= key.offset)
3607 new_key.offset = key.offset + destoff - off;
3609 new_key.offset = destoff;
3612 * Deal with a hole that doesn't have an extent item
3613 * that represents it (NO_HOLES feature enabled).
3614 * This hole is either in the middle of the cloning
3615 * range or at the beginning (fully overlaps it or
3616 * partially overlaps it).
3618 if (new_key.offset != last_dest_end)
3619 drop_start = last_dest_end;
3621 drop_start = new_key.offset;
3624 * 1 - adjusting old extent (we may have to split it)
3625 * 1 - add new extent
3628 trans = btrfs_start_transaction(root, 3);
3629 if (IS_ERR(trans)) {
3630 ret = PTR_ERR(trans);
3634 if (type == BTRFS_FILE_EXTENT_REG ||
3635 type == BTRFS_FILE_EXTENT_PREALLOC) {
3637 * a | --- range to clone ---| b
3638 * | ------------- extent ------------- |
3641 /* subtract range b */
3642 if (key.offset + datal > off + len)
3643 datal = off + len - key.offset;
3645 /* subtract range a */
3646 if (off > key.offset) {
3647 datao += off - key.offset;
3648 datal -= off - key.offset;
3651 ret = btrfs_drop_extents(trans, root, inode,
3653 new_key.offset + datal,
3656 if (ret != -EOPNOTSUPP)
3657 btrfs_abort_transaction(trans,
3659 btrfs_end_transaction(trans);
3663 ret = btrfs_insert_empty_item(trans, root, path,
3666 btrfs_abort_transaction(trans, ret);
3667 btrfs_end_transaction(trans);
3671 leaf = path->nodes[0];
3672 slot = path->slots[0];
3673 write_extent_buffer(leaf, buf,
3674 btrfs_item_ptr_offset(leaf, slot),
3677 extent = btrfs_item_ptr(leaf, slot,
3678 struct btrfs_file_extent_item);
3680 /* disko == 0 means it's a hole */
3684 btrfs_set_file_extent_offset(leaf, extent,
3686 btrfs_set_file_extent_num_bytes(leaf, extent,
3690 inode_add_bytes(inode, datal);
3691 ret = btrfs_inc_extent_ref(trans,
3694 root->root_key.objectid,
3695 btrfs_ino(BTRFS_I(inode)),
3696 new_key.offset - datao);
3698 btrfs_abort_transaction(trans,
3700 btrfs_end_transaction(trans);
3705 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3709 if (off > key.offset) {
3710 skip = off - key.offset;
3711 new_key.offset += skip;
3714 if (key.offset + datal > off + len)
3715 trim = key.offset + datal - (off + len);
3717 if (comp && (skip || trim)) {
3719 btrfs_end_transaction(trans);
3722 size -= skip + trim;
3723 datal -= skip + trim;
3725 ret = clone_copy_inline_extent(inode,
3732 if (ret != -EOPNOTSUPP)
3733 btrfs_abort_transaction(trans,
3735 btrfs_end_transaction(trans);
3738 leaf = path->nodes[0];
3739 slot = path->slots[0];
3742 /* If we have an implicit hole (NO_HOLES feature). */
3743 if (drop_start < new_key.offset)
3744 clone_update_extent_map(BTRFS_I(inode), trans,
3746 new_key.offset - drop_start);
3748 clone_update_extent_map(BTRFS_I(inode), trans,
3751 btrfs_mark_buffer_dirty(leaf);
3752 btrfs_release_path(path);
3754 last_dest_end = ALIGN(new_key.offset + datal,
3755 fs_info->sectorsize);
3756 ret = clone_finish_inode_update(trans, inode,
3762 if (new_key.offset + datal >= destoff + len)
3765 btrfs_release_path(path);
3766 key.offset = next_key_min_offset;
3768 if (fatal_signal_pending(current)) {
3775 if (last_dest_end < destoff + len) {
3777 * We have an implicit hole (NO_HOLES feature is enabled) that
3778 * fully or partially overlaps our cloning range at its end.
3780 btrfs_release_path(path);
3783 * 1 - remove extent(s)
3786 trans = btrfs_start_transaction(root, 2);
3787 if (IS_ERR(trans)) {
3788 ret = PTR_ERR(trans);
3791 ret = btrfs_drop_extents(trans, root, inode,
3792 last_dest_end, destoff + len, 1);
3794 if (ret != -EOPNOTSUPP)
3795 btrfs_abort_transaction(trans, ret);
3796 btrfs_end_transaction(trans);
3799 clone_update_extent_map(BTRFS_I(inode), trans, NULL,
3801 destoff + len - last_dest_end);
3802 ret = clone_finish_inode_update(trans, inode, destoff + len,
3803 destoff, olen, no_time_update);
3807 btrfs_free_path(path);
3812 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3813 u64 off, u64 olen, u64 destoff)
3815 struct inode *inode = file_inode(file);
3816 struct inode *src = file_inode(file_src);
3817 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3818 struct btrfs_root *root = BTRFS_I(inode)->root;
3821 u64 bs = fs_info->sb->s_blocksize;
3822 int same_inode = src == inode;
3826 * - split compressed inline extents. annoying: we need to
3827 * decompress into destination's address_space (the file offset
3828 * may change, so source mapping won't do), then recompress (or
3829 * otherwise reinsert) a subrange.
3831 * - split destination inode's inline extents. The inline extents can
3832 * be either compressed or non-compressed.
3835 if (btrfs_root_readonly(root))
3838 if (file_src->f_path.mnt != file->f_path.mnt ||
3839 src->i_sb != inode->i_sb)
3842 /* don't make the dst file partly checksummed */
3843 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3844 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3847 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3851 btrfs_double_inode_lock(src, inode);
3856 /* determine range to clone */
3858 if (off + len > src->i_size || off + len < off)
3861 olen = len = src->i_size - off;
3862 /* if we extend to eof, continue to block boundary */
3863 if (off + len == src->i_size)
3864 len = ALIGN(src->i_size, bs) - off;
3871 /* verify the end result is block aligned */
3872 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3873 !IS_ALIGNED(destoff, bs))
3876 /* verify if ranges are overlapped within the same file */
3878 if (destoff + len > off && destoff < off + len)
3882 if (destoff > inode->i_size) {
3883 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3889 * Lock the target range too. Right after we replace the file extent
3890 * items in the fs tree (which now point to the cloned data), we might
3891 * have a worker replace them with extent items relative to a write
3892 * operation that was issued before this clone operation (i.e. confront
3893 * with inode.c:btrfs_finish_ordered_io).
3896 u64 lock_start = min_t(u64, off, destoff);
3897 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3899 ret = lock_extent_range(src, lock_start, lock_len, true);
3901 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3906 /* ranges in the io trees already unlocked */
3910 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3913 u64 lock_start = min_t(u64, off, destoff);
3914 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3916 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3918 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3921 * Truncate page cache pages so that future reads will see the cloned
3922 * data immediately and not the previous data.
3924 truncate_inode_pages_range(&inode->i_data,
3925 round_down(destoff, PAGE_SIZE),
3926 round_up(destoff + len, PAGE_SIZE) - 1);
3929 btrfs_double_inode_unlock(src, inode);
3935 int btrfs_clone_file_range(struct file *src_file, loff_t off,
3936 struct file *dst_file, loff_t destoff, u64 len)
3938 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
3941 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3943 struct inode *inode = file_inode(file);
3944 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3945 struct btrfs_root *root = BTRFS_I(inode)->root;
3946 struct btrfs_root *new_root;
3947 struct btrfs_dir_item *di;
3948 struct btrfs_trans_handle *trans;
3949 struct btrfs_path *path;
3950 struct btrfs_key location;
3951 struct btrfs_disk_key disk_key;
3956 if (!capable(CAP_SYS_ADMIN))
3959 ret = mnt_want_write_file(file);
3963 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3969 objectid = BTRFS_FS_TREE_OBJECTID;
3971 location.objectid = objectid;
3972 location.type = BTRFS_ROOT_ITEM_KEY;
3973 location.offset = (u64)-1;
3975 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
3976 if (IS_ERR(new_root)) {
3977 ret = PTR_ERR(new_root);
3980 if (!is_fstree(new_root->objectid)) {
3985 path = btrfs_alloc_path();
3990 path->leave_spinning = 1;
3992 trans = btrfs_start_transaction(root, 1);
3993 if (IS_ERR(trans)) {
3994 btrfs_free_path(path);
3995 ret = PTR_ERR(trans);
3999 dir_id = btrfs_super_root_dir(fs_info->super_copy);
4000 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
4001 dir_id, "default", 7, 1);
4002 if (IS_ERR_OR_NULL(di)) {
4003 btrfs_free_path(path);
4004 btrfs_end_transaction(trans);
4006 "Umm, you don't have the default diritem, this isn't going to work");
4011 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4012 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4013 btrfs_mark_buffer_dirty(path->nodes[0]);
4014 btrfs_free_path(path);
4016 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
4017 btrfs_end_transaction(trans);
4019 mnt_drop_write_file(file);
4023 void btrfs_get_block_group_info(struct list_head *groups_list,
4024 struct btrfs_ioctl_space_info *space)
4026 struct btrfs_block_group_cache *block_group;
4028 space->total_bytes = 0;
4029 space->used_bytes = 0;
4031 list_for_each_entry(block_group, groups_list, list) {
4032 space->flags = block_group->flags;
4033 space->total_bytes += block_group->key.offset;
4034 space->used_bytes +=
4035 btrfs_block_group_used(&block_group->item);
4039 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
4042 struct btrfs_ioctl_space_args space_args;
4043 struct btrfs_ioctl_space_info space;
4044 struct btrfs_ioctl_space_info *dest;
4045 struct btrfs_ioctl_space_info *dest_orig;
4046 struct btrfs_ioctl_space_info __user *user_dest;
4047 struct btrfs_space_info *info;
4048 static const u64 types[] = {
4049 BTRFS_BLOCK_GROUP_DATA,
4050 BTRFS_BLOCK_GROUP_SYSTEM,
4051 BTRFS_BLOCK_GROUP_METADATA,
4052 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
4060 if (copy_from_user(&space_args,
4061 (struct btrfs_ioctl_space_args __user *)arg,
4062 sizeof(space_args)))
4065 for (i = 0; i < num_types; i++) {
4066 struct btrfs_space_info *tmp;
4070 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4072 if (tmp->flags == types[i]) {
4082 down_read(&info->groups_sem);
4083 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4084 if (!list_empty(&info->block_groups[c]))
4087 up_read(&info->groups_sem);
4091 * Global block reserve, exported as a space_info
4095 /* space_slots == 0 means they are asking for a count */
4096 if (space_args.space_slots == 0) {
4097 space_args.total_spaces = slot_count;
4101 slot_count = min_t(u64, space_args.space_slots, slot_count);
4103 alloc_size = sizeof(*dest) * slot_count;
4105 /* we generally have at most 6 or so space infos, one for each raid
4106 * level. So, a whole page should be more than enough for everyone
4108 if (alloc_size > PAGE_SIZE)
4111 space_args.total_spaces = 0;
4112 dest = kmalloc(alloc_size, GFP_KERNEL);
4117 /* now we have a buffer to copy into */
4118 for (i = 0; i < num_types; i++) {
4119 struct btrfs_space_info *tmp;
4126 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4128 if (tmp->flags == types[i]) {
4137 down_read(&info->groups_sem);
4138 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4139 if (!list_empty(&info->block_groups[c])) {
4140 btrfs_get_block_group_info(
4141 &info->block_groups[c], &space);
4142 memcpy(dest, &space, sizeof(space));
4144 space_args.total_spaces++;
4150 up_read(&info->groups_sem);
4154 * Add global block reserve
4157 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
4159 spin_lock(&block_rsv->lock);
4160 space.total_bytes = block_rsv->size;
4161 space.used_bytes = block_rsv->size - block_rsv->reserved;
4162 spin_unlock(&block_rsv->lock);
4163 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4164 memcpy(dest, &space, sizeof(space));
4165 space_args.total_spaces++;
4168 user_dest = (struct btrfs_ioctl_space_info __user *)
4169 (arg + sizeof(struct btrfs_ioctl_space_args));
4171 if (copy_to_user(user_dest, dest_orig, alloc_size))
4176 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4182 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4185 struct btrfs_trans_handle *trans;
4189 trans = btrfs_attach_transaction_barrier(root);
4190 if (IS_ERR(trans)) {
4191 if (PTR_ERR(trans) != -ENOENT)
4192 return PTR_ERR(trans);
4194 /* No running transaction, don't bother */
4195 transid = root->fs_info->last_trans_committed;
4198 transid = trans->transid;
4199 ret = btrfs_commit_transaction_async(trans, 0);
4201 btrfs_end_transaction(trans);
4206 if (copy_to_user(argp, &transid, sizeof(transid)))
4211 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
4217 if (copy_from_user(&transid, argp, sizeof(transid)))
4220 transid = 0; /* current trans */
4222 return btrfs_wait_for_commit(fs_info, transid);
4225 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4227 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4228 struct btrfs_ioctl_scrub_args *sa;
4231 if (!capable(CAP_SYS_ADMIN))
4234 sa = memdup_user(arg, sizeof(*sa));
4238 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4239 ret = mnt_want_write_file(file);
4244 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4245 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4248 if (copy_to_user(arg, sa, sizeof(*sa)))
4251 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4252 mnt_drop_write_file(file);
4258 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
4260 if (!capable(CAP_SYS_ADMIN))
4263 return btrfs_scrub_cancel(fs_info);
4266 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
4269 struct btrfs_ioctl_scrub_args *sa;
4272 if (!capable(CAP_SYS_ADMIN))
4275 sa = memdup_user(arg, sizeof(*sa));
4279 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
4281 if (copy_to_user(arg, sa, sizeof(*sa)))
4288 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
4291 struct btrfs_ioctl_get_dev_stats *sa;
4294 sa = memdup_user(arg, sizeof(*sa));
4298 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4303 ret = btrfs_get_dev_stats(fs_info, sa);
4305 if (copy_to_user(arg, sa, sizeof(*sa)))
4312 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4315 struct btrfs_ioctl_dev_replace_args *p;
4318 if (!capable(CAP_SYS_ADMIN))
4321 p = memdup_user(arg, sizeof(*p));
4326 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4327 if (sb_rdonly(fs_info->sb)) {
4331 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4332 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4334 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
4335 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4338 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4339 btrfs_dev_replace_status(fs_info, p);
4342 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4343 p->result = btrfs_dev_replace_cancel(fs_info);
4351 if (copy_to_user(arg, p, sizeof(*p)))
4358 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4364 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4365 struct inode_fs_paths *ipath = NULL;
4366 struct btrfs_path *path;
4368 if (!capable(CAP_DAC_READ_SEARCH))
4371 path = btrfs_alloc_path();
4377 ipa = memdup_user(arg, sizeof(*ipa));
4384 size = min_t(u32, ipa->size, 4096);
4385 ipath = init_ipath(size, root, path);
4386 if (IS_ERR(ipath)) {
4387 ret = PTR_ERR(ipath);
4392 ret = paths_from_inode(ipa->inum, ipath);
4396 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4397 rel_ptr = ipath->fspath->val[i] -
4398 (u64)(unsigned long)ipath->fspath->val;
4399 ipath->fspath->val[i] = rel_ptr;
4402 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
4403 ipath->fspath, size);
4410 btrfs_free_path(path);
4417 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4419 struct btrfs_data_container *inodes = ctx;
4420 const size_t c = 3 * sizeof(u64);
4422 if (inodes->bytes_left >= c) {
4423 inodes->bytes_left -= c;
4424 inodes->val[inodes->elem_cnt] = inum;
4425 inodes->val[inodes->elem_cnt + 1] = offset;
4426 inodes->val[inodes->elem_cnt + 2] = root;
4427 inodes->elem_cnt += 3;
4429 inodes->bytes_missing += c - inodes->bytes_left;
4430 inodes->bytes_left = 0;
4431 inodes->elem_missed += 3;
4437 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
4438 void __user *arg, int version)
4442 struct btrfs_ioctl_logical_ino_args *loi;
4443 struct btrfs_data_container *inodes = NULL;
4444 struct btrfs_path *path = NULL;
4447 if (!capable(CAP_SYS_ADMIN))
4450 loi = memdup_user(arg, sizeof(*loi));
4452 return PTR_ERR(loi);
4455 ignore_offset = false;
4456 size = min_t(u32, loi->size, SZ_64K);
4458 /* All reserved bits must be 0 for now */
4459 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
4463 /* Only accept flags we have defined so far */
4464 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
4468 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
4469 size = min_t(u32, loi->size, SZ_16M);
4472 path = btrfs_alloc_path();
4478 inodes = init_data_container(size);
4479 if (IS_ERR(inodes)) {
4480 ret = PTR_ERR(inodes);
4485 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
4486 build_ino_list, inodes, ignore_offset);
4492 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
4498 btrfs_free_path(path);
4506 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4507 struct btrfs_ioctl_balance_args *bargs)
4509 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4511 bargs->flags = bctl->flags;
4513 if (atomic_read(&fs_info->balance_running))
4514 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4515 if (atomic_read(&fs_info->balance_pause_req))
4516 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4517 if (atomic_read(&fs_info->balance_cancel_req))
4518 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4520 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4521 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4522 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4525 spin_lock(&fs_info->balance_lock);
4526 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4527 spin_unlock(&fs_info->balance_lock);
4529 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4533 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4535 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4536 struct btrfs_fs_info *fs_info = root->fs_info;
4537 struct btrfs_ioctl_balance_args *bargs;
4538 struct btrfs_balance_control *bctl;
4539 bool need_unlock; /* for mut. excl. ops lock */
4542 if (!capable(CAP_SYS_ADMIN))
4545 ret = mnt_want_write_file(file);
4550 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4551 mutex_lock(&fs_info->volume_mutex);
4552 mutex_lock(&fs_info->balance_mutex);
4558 * mut. excl. ops lock is locked. Three possibilities:
4559 * (1) some other op is running
4560 * (2) balance is running
4561 * (3) balance is paused -- special case (think resume)
4563 mutex_lock(&fs_info->balance_mutex);
4564 if (fs_info->balance_ctl) {
4565 /* this is either (2) or (3) */
4566 if (!atomic_read(&fs_info->balance_running)) {
4567 mutex_unlock(&fs_info->balance_mutex);
4568 if (!mutex_trylock(&fs_info->volume_mutex))
4570 mutex_lock(&fs_info->balance_mutex);
4572 if (fs_info->balance_ctl &&
4573 !atomic_read(&fs_info->balance_running)) {
4575 need_unlock = false;
4579 mutex_unlock(&fs_info->balance_mutex);
4580 mutex_unlock(&fs_info->volume_mutex);
4584 mutex_unlock(&fs_info->balance_mutex);
4590 mutex_unlock(&fs_info->balance_mutex);
4591 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4596 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
4599 bargs = memdup_user(arg, sizeof(*bargs));
4600 if (IS_ERR(bargs)) {
4601 ret = PTR_ERR(bargs);
4605 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4606 if (!fs_info->balance_ctl) {
4611 bctl = fs_info->balance_ctl;
4612 spin_lock(&fs_info->balance_lock);
4613 bctl->flags |= BTRFS_BALANCE_RESUME;
4614 spin_unlock(&fs_info->balance_lock);
4622 if (fs_info->balance_ctl) {
4627 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4633 bctl->fs_info = fs_info;
4635 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4636 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4637 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4639 bctl->flags = bargs->flags;
4641 /* balance everything - no filters */
4642 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4645 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4652 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP
4653 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4654 * or, if restriper was paused all the way until unmount, in
4655 * free_fs_info. The flag is cleared in __cancel_balance.
4657 need_unlock = false;
4659 ret = btrfs_balance(bctl, bargs);
4663 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4672 mutex_unlock(&fs_info->balance_mutex);
4673 mutex_unlock(&fs_info->volume_mutex);
4675 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4677 mnt_drop_write_file(file);
4681 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4683 if (!capable(CAP_SYS_ADMIN))
4687 case BTRFS_BALANCE_CTL_PAUSE:
4688 return btrfs_pause_balance(fs_info);
4689 case BTRFS_BALANCE_CTL_CANCEL:
4690 return btrfs_cancel_balance(fs_info);
4696 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4699 struct btrfs_ioctl_balance_args *bargs;
4702 if (!capable(CAP_SYS_ADMIN))
4705 mutex_lock(&fs_info->balance_mutex);
4706 if (!fs_info->balance_ctl) {
4711 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4717 update_ioctl_balance_args(fs_info, 1, bargs);
4719 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4724 mutex_unlock(&fs_info->balance_mutex);
4728 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4730 struct inode *inode = file_inode(file);
4731 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4732 struct btrfs_ioctl_quota_ctl_args *sa;
4733 struct btrfs_trans_handle *trans = NULL;
4737 if (!capable(CAP_SYS_ADMIN))
4740 ret = mnt_want_write_file(file);
4744 sa = memdup_user(arg, sizeof(*sa));
4750 down_write(&fs_info->subvol_sem);
4751 trans = btrfs_start_transaction(fs_info->tree_root, 2);
4752 if (IS_ERR(trans)) {
4753 ret = PTR_ERR(trans);
4758 case BTRFS_QUOTA_CTL_ENABLE:
4759 ret = btrfs_quota_enable(trans, fs_info);
4761 case BTRFS_QUOTA_CTL_DISABLE:
4762 ret = btrfs_quota_disable(trans, fs_info);
4769 err = btrfs_commit_transaction(trans);
4774 up_write(&fs_info->subvol_sem);
4776 mnt_drop_write_file(file);
4780 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4782 struct inode *inode = file_inode(file);
4783 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4784 struct btrfs_root *root = BTRFS_I(inode)->root;
4785 struct btrfs_ioctl_qgroup_assign_args *sa;
4786 struct btrfs_trans_handle *trans;
4790 if (!capable(CAP_SYS_ADMIN))
4793 ret = mnt_want_write_file(file);
4797 sa = memdup_user(arg, sizeof(*sa));
4803 trans = btrfs_join_transaction(root);
4804 if (IS_ERR(trans)) {
4805 ret = PTR_ERR(trans);
4810 ret = btrfs_add_qgroup_relation(trans, fs_info,
4813 ret = btrfs_del_qgroup_relation(trans, fs_info,
4817 /* update qgroup status and info */
4818 err = btrfs_run_qgroups(trans, fs_info);
4820 btrfs_handle_fs_error(fs_info, err,
4821 "failed to update qgroup status and info");
4822 err = btrfs_end_transaction(trans);
4829 mnt_drop_write_file(file);
4833 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4835 struct inode *inode = file_inode(file);
4836 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4837 struct btrfs_root *root = BTRFS_I(inode)->root;
4838 struct btrfs_ioctl_qgroup_create_args *sa;
4839 struct btrfs_trans_handle *trans;
4843 if (!capable(CAP_SYS_ADMIN))
4846 ret = mnt_want_write_file(file);
4850 sa = memdup_user(arg, sizeof(*sa));
4856 if (!sa->qgroupid) {
4861 trans = btrfs_join_transaction(root);
4862 if (IS_ERR(trans)) {
4863 ret = PTR_ERR(trans);
4868 ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid);
4870 ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid);
4873 err = btrfs_end_transaction(trans);
4880 mnt_drop_write_file(file);
4884 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4886 struct inode *inode = file_inode(file);
4887 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4888 struct btrfs_root *root = BTRFS_I(inode)->root;
4889 struct btrfs_ioctl_qgroup_limit_args *sa;
4890 struct btrfs_trans_handle *trans;
4895 if (!capable(CAP_SYS_ADMIN))
4898 ret = mnt_want_write_file(file);
4902 sa = memdup_user(arg, sizeof(*sa));
4908 trans = btrfs_join_transaction(root);
4909 if (IS_ERR(trans)) {
4910 ret = PTR_ERR(trans);
4914 qgroupid = sa->qgroupid;
4916 /* take the current subvol as qgroup */
4917 qgroupid = root->root_key.objectid;
4920 ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim);
4922 err = btrfs_end_transaction(trans);
4929 mnt_drop_write_file(file);
4933 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4935 struct inode *inode = file_inode(file);
4936 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4937 struct btrfs_ioctl_quota_rescan_args *qsa;
4940 if (!capable(CAP_SYS_ADMIN))
4943 ret = mnt_want_write_file(file);
4947 qsa = memdup_user(arg, sizeof(*qsa));
4958 ret = btrfs_qgroup_rescan(fs_info);
4963 mnt_drop_write_file(file);
4967 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4969 struct inode *inode = file_inode(file);
4970 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4971 struct btrfs_ioctl_quota_rescan_args *qsa;
4974 if (!capable(CAP_SYS_ADMIN))
4977 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
4981 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4983 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
4986 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4993 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4995 struct inode *inode = file_inode(file);
4996 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4998 if (!capable(CAP_SYS_ADMIN))
5001 return btrfs_qgroup_wait_for_completion(fs_info, true);
5004 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5005 struct btrfs_ioctl_received_subvol_args *sa)
5007 struct inode *inode = file_inode(file);
5008 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5009 struct btrfs_root *root = BTRFS_I(inode)->root;
5010 struct btrfs_root_item *root_item = &root->root_item;
5011 struct btrfs_trans_handle *trans;
5012 struct timespec ct = current_time(inode);
5014 int received_uuid_changed;
5016 if (!inode_owner_or_capable(inode))
5019 ret = mnt_want_write_file(file);
5023 down_write(&fs_info->subvol_sem);
5025 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
5030 if (btrfs_root_readonly(root)) {
5037 * 2 - uuid items (received uuid + subvol uuid)
5039 trans = btrfs_start_transaction(root, 3);
5040 if (IS_ERR(trans)) {
5041 ret = PTR_ERR(trans);
5046 sa->rtransid = trans->transid;
5047 sa->rtime.sec = ct.tv_sec;
5048 sa->rtime.nsec = ct.tv_nsec;
5050 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5052 if (received_uuid_changed &&
5053 !btrfs_is_empty_uuid(root_item->received_uuid)) {
5054 ret = btrfs_uuid_tree_rem(trans, fs_info,
5055 root_item->received_uuid,
5056 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5057 root->root_key.objectid);
5058 if (ret && ret != -ENOENT) {
5059 btrfs_abort_transaction(trans, ret);
5060 btrfs_end_transaction(trans);
5064 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5065 btrfs_set_root_stransid(root_item, sa->stransid);
5066 btrfs_set_root_rtransid(root_item, sa->rtransid);
5067 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5068 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5069 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5070 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5072 ret = btrfs_update_root(trans, fs_info->tree_root,
5073 &root->root_key, &root->root_item);
5075 btrfs_end_transaction(trans);
5078 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5079 ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid,
5080 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5081 root->root_key.objectid);
5082 if (ret < 0 && ret != -EEXIST) {
5083 btrfs_abort_transaction(trans, ret);
5084 btrfs_end_transaction(trans);
5088 ret = btrfs_commit_transaction(trans);
5090 up_write(&fs_info->subvol_sem);
5091 mnt_drop_write_file(file);
5096 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5099 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5100 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5103 args32 = memdup_user(arg, sizeof(*args32));
5105 return PTR_ERR(args32);
5107 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5113 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5114 args64->stransid = args32->stransid;
5115 args64->rtransid = args32->rtransid;
5116 args64->stime.sec = args32->stime.sec;
5117 args64->stime.nsec = args32->stime.nsec;
5118 args64->rtime.sec = args32->rtime.sec;
5119 args64->rtime.nsec = args32->rtime.nsec;
5120 args64->flags = args32->flags;
5122 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5126 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5127 args32->stransid = args64->stransid;
5128 args32->rtransid = args64->rtransid;
5129 args32->stime.sec = args64->stime.sec;
5130 args32->stime.nsec = args64->stime.nsec;
5131 args32->rtime.sec = args64->rtime.sec;
5132 args32->rtime.nsec = args64->rtime.nsec;
5133 args32->flags = args64->flags;
5135 ret = copy_to_user(arg, args32, sizeof(*args32));
5146 static long btrfs_ioctl_set_received_subvol(struct file *file,
5149 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5152 sa = memdup_user(arg, sizeof(*sa));
5156 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5161 ret = copy_to_user(arg, sa, sizeof(*sa));
5170 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5172 struct inode *inode = file_inode(file);
5173 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5176 char label[BTRFS_LABEL_SIZE];
5178 spin_lock(&fs_info->super_lock);
5179 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5180 spin_unlock(&fs_info->super_lock);
5182 len = strnlen(label, BTRFS_LABEL_SIZE);
5184 if (len == BTRFS_LABEL_SIZE) {
5186 "label is too long, return the first %zu bytes",
5190 ret = copy_to_user(arg, label, len);
5192 return ret ? -EFAULT : 0;
5195 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5197 struct inode *inode = file_inode(file);
5198 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5199 struct btrfs_root *root = BTRFS_I(inode)->root;
5200 struct btrfs_super_block *super_block = fs_info->super_copy;
5201 struct btrfs_trans_handle *trans;
5202 char label[BTRFS_LABEL_SIZE];
5205 if (!capable(CAP_SYS_ADMIN))
5208 if (copy_from_user(label, arg, sizeof(label)))
5211 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5213 "unable to set label with more than %d bytes",
5214 BTRFS_LABEL_SIZE - 1);
5218 ret = mnt_want_write_file(file);
5222 trans = btrfs_start_transaction(root, 0);
5223 if (IS_ERR(trans)) {
5224 ret = PTR_ERR(trans);
5228 spin_lock(&fs_info->super_lock);
5229 strcpy(super_block->label, label);
5230 spin_unlock(&fs_info->super_lock);
5231 ret = btrfs_commit_transaction(trans);
5234 mnt_drop_write_file(file);
5238 #define INIT_FEATURE_FLAGS(suffix) \
5239 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5240 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5241 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5243 int btrfs_ioctl_get_supported_features(void __user *arg)
5245 static const struct btrfs_ioctl_feature_flags features[3] = {
5246 INIT_FEATURE_FLAGS(SUPP),
5247 INIT_FEATURE_FLAGS(SAFE_SET),
5248 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5251 if (copy_to_user(arg, &features, sizeof(features)))
5257 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5259 struct inode *inode = file_inode(file);
5260 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5261 struct btrfs_super_block *super_block = fs_info->super_copy;
5262 struct btrfs_ioctl_feature_flags features;
5264 features.compat_flags = btrfs_super_compat_flags(super_block);
5265 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5266 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5268 if (copy_to_user(arg, &features, sizeof(features)))
5274 static int check_feature_bits(struct btrfs_fs_info *fs_info,
5275 enum btrfs_feature_set set,
5276 u64 change_mask, u64 flags, u64 supported_flags,
5277 u64 safe_set, u64 safe_clear)
5279 const char *type = btrfs_feature_set_names[set];
5281 u64 disallowed, unsupported;
5282 u64 set_mask = flags & change_mask;
5283 u64 clear_mask = ~flags & change_mask;
5285 unsupported = set_mask & ~supported_flags;
5287 names = btrfs_printable_features(set, unsupported);
5290 "this kernel does not support the %s feature bit%s",
5291 names, strchr(names, ',') ? "s" : "");
5295 "this kernel does not support %s bits 0x%llx",
5300 disallowed = set_mask & ~safe_set;
5302 names = btrfs_printable_features(set, disallowed);
5305 "can't set the %s feature bit%s while mounted",
5306 names, strchr(names, ',') ? "s" : "");
5310 "can't set %s bits 0x%llx while mounted",
5315 disallowed = clear_mask & ~safe_clear;
5317 names = btrfs_printable_features(set, disallowed);
5320 "can't clear the %s feature bit%s while mounted",
5321 names, strchr(names, ',') ? "s" : "");
5325 "can't clear %s bits 0x%llx while mounted",
5333 #define check_feature(fs_info, change_mask, flags, mask_base) \
5334 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5335 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5336 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5337 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5339 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5341 struct inode *inode = file_inode(file);
5342 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5343 struct btrfs_root *root = BTRFS_I(inode)->root;
5344 struct btrfs_super_block *super_block = fs_info->super_copy;
5345 struct btrfs_ioctl_feature_flags flags[2];
5346 struct btrfs_trans_handle *trans;
5350 if (!capable(CAP_SYS_ADMIN))
5353 if (copy_from_user(flags, arg, sizeof(flags)))
5357 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5358 !flags[0].incompat_flags)
5361 ret = check_feature(fs_info, flags[0].compat_flags,
5362 flags[1].compat_flags, COMPAT);
5366 ret = check_feature(fs_info, flags[0].compat_ro_flags,
5367 flags[1].compat_ro_flags, COMPAT_RO);
5371 ret = check_feature(fs_info, flags[0].incompat_flags,
5372 flags[1].incompat_flags, INCOMPAT);
5376 ret = mnt_want_write_file(file);
5380 trans = btrfs_start_transaction(root, 0);
5381 if (IS_ERR(trans)) {
5382 ret = PTR_ERR(trans);
5383 goto out_drop_write;
5386 spin_lock(&fs_info->super_lock);
5387 newflags = btrfs_super_compat_flags(super_block);
5388 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5389 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5390 btrfs_set_super_compat_flags(super_block, newflags);
5392 newflags = btrfs_super_compat_ro_flags(super_block);
5393 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5394 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5395 btrfs_set_super_compat_ro_flags(super_block, newflags);
5397 newflags = btrfs_super_incompat_flags(super_block);
5398 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5399 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5400 btrfs_set_super_incompat_flags(super_block, newflags);
5401 spin_unlock(&fs_info->super_lock);
5403 ret = btrfs_commit_transaction(trans);
5405 mnt_drop_write_file(file);
5410 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
5412 struct btrfs_ioctl_send_args *arg;
5416 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5417 struct btrfs_ioctl_send_args_32 args32;
5419 ret = copy_from_user(&args32, argp, sizeof(args32));
5422 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
5425 arg->send_fd = args32.send_fd;
5426 arg->clone_sources_count = args32.clone_sources_count;
5427 arg->clone_sources = compat_ptr(args32.clone_sources);
5428 arg->parent_root = args32.parent_root;
5429 arg->flags = args32.flags;
5430 memcpy(arg->reserved, args32.reserved,
5431 sizeof(args32.reserved));
5436 arg = memdup_user(argp, sizeof(*arg));
5438 return PTR_ERR(arg);
5440 ret = btrfs_ioctl_send(file, arg);
5445 long btrfs_ioctl(struct file *file, unsigned int
5446 cmd, unsigned long arg)
5448 struct inode *inode = file_inode(file);
5449 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5450 struct btrfs_root *root = BTRFS_I(inode)->root;
5451 void __user *argp = (void __user *)arg;
5454 case FS_IOC_GETFLAGS:
5455 return btrfs_ioctl_getflags(file, argp);
5456 case FS_IOC_SETFLAGS:
5457 return btrfs_ioctl_setflags(file, argp);
5458 case FS_IOC_GETVERSION:
5459 return btrfs_ioctl_getversion(file, argp);
5461 return btrfs_ioctl_fitrim(file, argp);
5462 case BTRFS_IOC_SNAP_CREATE:
5463 return btrfs_ioctl_snap_create(file, argp, 0);
5464 case BTRFS_IOC_SNAP_CREATE_V2:
5465 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5466 case BTRFS_IOC_SUBVOL_CREATE:
5467 return btrfs_ioctl_snap_create(file, argp, 1);
5468 case BTRFS_IOC_SUBVOL_CREATE_V2:
5469 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5470 case BTRFS_IOC_SNAP_DESTROY:
5471 return btrfs_ioctl_snap_destroy(file, argp);
5472 case BTRFS_IOC_SUBVOL_GETFLAGS:
5473 return btrfs_ioctl_subvol_getflags(file, argp);
5474 case BTRFS_IOC_SUBVOL_SETFLAGS:
5475 return btrfs_ioctl_subvol_setflags(file, argp);
5476 case BTRFS_IOC_DEFAULT_SUBVOL:
5477 return btrfs_ioctl_default_subvol(file, argp);
5478 case BTRFS_IOC_DEFRAG:
5479 return btrfs_ioctl_defrag(file, NULL);
5480 case BTRFS_IOC_DEFRAG_RANGE:
5481 return btrfs_ioctl_defrag(file, argp);
5482 case BTRFS_IOC_RESIZE:
5483 return btrfs_ioctl_resize(file, argp);
5484 case BTRFS_IOC_ADD_DEV:
5485 return btrfs_ioctl_add_dev(fs_info, argp);
5486 case BTRFS_IOC_RM_DEV:
5487 return btrfs_ioctl_rm_dev(file, argp);
5488 case BTRFS_IOC_RM_DEV_V2:
5489 return btrfs_ioctl_rm_dev_v2(file, argp);
5490 case BTRFS_IOC_FS_INFO:
5491 return btrfs_ioctl_fs_info(fs_info, argp);
5492 case BTRFS_IOC_DEV_INFO:
5493 return btrfs_ioctl_dev_info(fs_info, argp);
5494 case BTRFS_IOC_BALANCE:
5495 return btrfs_ioctl_balance(file, NULL);
5496 case BTRFS_IOC_TREE_SEARCH:
5497 return btrfs_ioctl_tree_search(file, argp);
5498 case BTRFS_IOC_TREE_SEARCH_V2:
5499 return btrfs_ioctl_tree_search_v2(file, argp);
5500 case BTRFS_IOC_INO_LOOKUP:
5501 return btrfs_ioctl_ino_lookup(file, argp);
5502 case BTRFS_IOC_INO_PATHS:
5503 return btrfs_ioctl_ino_to_path(root, argp);
5504 case BTRFS_IOC_LOGICAL_INO:
5505 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
5506 case BTRFS_IOC_LOGICAL_INO_V2:
5507 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
5508 case BTRFS_IOC_SPACE_INFO:
5509 return btrfs_ioctl_space_info(fs_info, argp);
5510 case BTRFS_IOC_SYNC: {
5513 ret = btrfs_start_delalloc_roots(fs_info, 0, -1);
5516 ret = btrfs_sync_fs(inode->i_sb, 1);
5518 * The transaction thread may want to do more work,
5519 * namely it pokes the cleaner kthread that will start
5520 * processing uncleaned subvols.
5522 wake_up_process(fs_info->transaction_kthread);
5525 case BTRFS_IOC_START_SYNC:
5526 return btrfs_ioctl_start_sync(root, argp);
5527 case BTRFS_IOC_WAIT_SYNC:
5528 return btrfs_ioctl_wait_sync(fs_info, argp);
5529 case BTRFS_IOC_SCRUB:
5530 return btrfs_ioctl_scrub(file, argp);
5531 case BTRFS_IOC_SCRUB_CANCEL:
5532 return btrfs_ioctl_scrub_cancel(fs_info);
5533 case BTRFS_IOC_SCRUB_PROGRESS:
5534 return btrfs_ioctl_scrub_progress(fs_info, argp);
5535 case BTRFS_IOC_BALANCE_V2:
5536 return btrfs_ioctl_balance(file, argp);
5537 case BTRFS_IOC_BALANCE_CTL:
5538 return btrfs_ioctl_balance_ctl(fs_info, arg);
5539 case BTRFS_IOC_BALANCE_PROGRESS:
5540 return btrfs_ioctl_balance_progress(fs_info, argp);
5541 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5542 return btrfs_ioctl_set_received_subvol(file, argp);
5544 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5545 return btrfs_ioctl_set_received_subvol_32(file, argp);
5547 case BTRFS_IOC_SEND:
5548 return _btrfs_ioctl_send(file, argp, false);
5549 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5550 case BTRFS_IOC_SEND_32:
5551 return _btrfs_ioctl_send(file, argp, true);
5553 case BTRFS_IOC_GET_DEV_STATS:
5554 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5555 case BTRFS_IOC_QUOTA_CTL:
5556 return btrfs_ioctl_quota_ctl(file, argp);
5557 case BTRFS_IOC_QGROUP_ASSIGN:
5558 return btrfs_ioctl_qgroup_assign(file, argp);
5559 case BTRFS_IOC_QGROUP_CREATE:
5560 return btrfs_ioctl_qgroup_create(file, argp);
5561 case BTRFS_IOC_QGROUP_LIMIT:
5562 return btrfs_ioctl_qgroup_limit(file, argp);
5563 case BTRFS_IOC_QUOTA_RESCAN:
5564 return btrfs_ioctl_quota_rescan(file, argp);
5565 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5566 return btrfs_ioctl_quota_rescan_status(file, argp);
5567 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5568 return btrfs_ioctl_quota_rescan_wait(file, argp);
5569 case BTRFS_IOC_DEV_REPLACE:
5570 return btrfs_ioctl_dev_replace(fs_info, argp);
5571 case BTRFS_IOC_GET_FSLABEL:
5572 return btrfs_ioctl_get_fslabel(file, argp);
5573 case BTRFS_IOC_SET_FSLABEL:
5574 return btrfs_ioctl_set_fslabel(file, argp);
5575 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5576 return btrfs_ioctl_get_supported_features(argp);
5577 case BTRFS_IOC_GET_FEATURES:
5578 return btrfs_ioctl_get_features(file, argp);
5579 case BTRFS_IOC_SET_FEATURES:
5580 return btrfs_ioctl_set_features(file, argp);
5586 #ifdef CONFIG_COMPAT
5587 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5590 * These all access 32-bit values anyway so no further
5591 * handling is necessary.
5594 case FS_IOC32_GETFLAGS:
5595 cmd = FS_IOC_GETFLAGS;
5597 case FS_IOC32_SETFLAGS:
5598 cmd = FS_IOC_SETFLAGS;
5600 case FS_IOC32_GETVERSION:
5601 cmd = FS_IOC_GETVERSION;
5605 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));