2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/compat.h>
37 #include <linux/bit_spinlock.h>
38 #include <linux/security.h>
39 #include <linux/xattr.h>
41 #include <linux/slab.h>
42 #include <linux/blkdev.h>
43 #include <linux/uuid.h>
44 #include <linux/btrfs.h>
45 #include <linux/uaccess.h>
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "print-tree.h"
53 #include "inode-map.h"
55 #include "rcu-string.h"
57 #include "dev-replace.h"
62 #include "compression.h"
65 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
66 * structures are incorrect, as the timespec structure from userspace
67 * is 4 bytes too small. We define these alternatives here to teach
68 * the kernel about the 32-bit struct packing.
70 struct btrfs_ioctl_timespec_32 {
73 } __attribute__ ((__packed__));
75 struct btrfs_ioctl_received_subvol_args_32 {
76 char uuid[BTRFS_UUID_SIZE]; /* in */
77 __u64 stransid; /* in */
78 __u64 rtransid; /* out */
79 struct btrfs_ioctl_timespec_32 stime; /* in */
80 struct btrfs_ioctl_timespec_32 rtime; /* out */
82 __u64 reserved[16]; /* in */
83 } __attribute__ ((__packed__));
85 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
86 struct btrfs_ioctl_received_subvol_args_32)
90 static int btrfs_clone(struct inode *src, struct inode *inode,
91 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
94 /* Mask out flags that are inappropriate for the given type of inode. */
95 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
99 else if (S_ISREG(mode))
100 return flags & ~FS_DIRSYNC_FL;
102 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
106 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
108 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
110 unsigned int iflags = 0;
112 if (flags & BTRFS_INODE_SYNC)
113 iflags |= FS_SYNC_FL;
114 if (flags & BTRFS_INODE_IMMUTABLE)
115 iflags |= FS_IMMUTABLE_FL;
116 if (flags & BTRFS_INODE_APPEND)
117 iflags |= FS_APPEND_FL;
118 if (flags & BTRFS_INODE_NODUMP)
119 iflags |= FS_NODUMP_FL;
120 if (flags & BTRFS_INODE_NOATIME)
121 iflags |= FS_NOATIME_FL;
122 if (flags & BTRFS_INODE_DIRSYNC)
123 iflags |= FS_DIRSYNC_FL;
124 if (flags & BTRFS_INODE_NODATACOW)
125 iflags |= FS_NOCOW_FL;
127 if (flags & BTRFS_INODE_NOCOMPRESS)
128 iflags |= FS_NOCOMP_FL;
129 else if (flags & BTRFS_INODE_COMPRESS)
130 iflags |= FS_COMPR_FL;
136 * Update inode->i_flags based on the btrfs internal flags.
138 void btrfs_update_iflags(struct inode *inode)
140 struct btrfs_inode *ip = BTRFS_I(inode);
141 unsigned int new_fl = 0;
143 if (ip->flags & BTRFS_INODE_SYNC)
145 if (ip->flags & BTRFS_INODE_IMMUTABLE)
146 new_fl |= S_IMMUTABLE;
147 if (ip->flags & BTRFS_INODE_APPEND)
149 if (ip->flags & BTRFS_INODE_NOATIME)
151 if (ip->flags & BTRFS_INODE_DIRSYNC)
154 set_mask_bits(&inode->i_flags,
155 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
159 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
161 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
162 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
164 if (copy_to_user(arg, &flags, sizeof(flags)))
169 static int check_flags(unsigned int flags)
171 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
172 FS_NOATIME_FL | FS_NODUMP_FL | \
173 FS_SYNC_FL | FS_DIRSYNC_FL | \
174 FS_NOCOMP_FL | FS_COMPR_FL |
178 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
184 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
186 struct inode *inode = file_inode(file);
187 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
188 struct btrfs_inode *ip = BTRFS_I(inode);
189 struct btrfs_root *root = ip->root;
190 struct btrfs_trans_handle *trans;
191 unsigned int flags, oldflags;
194 unsigned int i_oldflags;
197 if (!inode_owner_or_capable(inode))
200 if (btrfs_root_readonly(root))
203 if (copy_from_user(&flags, arg, sizeof(flags)))
206 ret = check_flags(flags);
210 ret = mnt_want_write_file(file);
216 ip_oldflags = ip->flags;
217 i_oldflags = inode->i_flags;
218 mode = inode->i_mode;
220 flags = btrfs_mask_flags(inode->i_mode, flags);
221 oldflags = btrfs_flags_to_ioctl(ip->flags);
222 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
223 if (!capable(CAP_LINUX_IMMUTABLE)) {
229 if (flags & FS_SYNC_FL)
230 ip->flags |= BTRFS_INODE_SYNC;
232 ip->flags &= ~BTRFS_INODE_SYNC;
233 if (flags & FS_IMMUTABLE_FL)
234 ip->flags |= BTRFS_INODE_IMMUTABLE;
236 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
237 if (flags & FS_APPEND_FL)
238 ip->flags |= BTRFS_INODE_APPEND;
240 ip->flags &= ~BTRFS_INODE_APPEND;
241 if (flags & FS_NODUMP_FL)
242 ip->flags |= BTRFS_INODE_NODUMP;
244 ip->flags &= ~BTRFS_INODE_NODUMP;
245 if (flags & FS_NOATIME_FL)
246 ip->flags |= BTRFS_INODE_NOATIME;
248 ip->flags &= ~BTRFS_INODE_NOATIME;
249 if (flags & FS_DIRSYNC_FL)
250 ip->flags |= BTRFS_INODE_DIRSYNC;
252 ip->flags &= ~BTRFS_INODE_DIRSYNC;
253 if (flags & FS_NOCOW_FL) {
256 * It's safe to turn csums off here, no extents exist.
257 * Otherwise we want the flag to reflect the real COW
258 * status of the file and will not set it.
260 if (inode->i_size == 0)
261 ip->flags |= BTRFS_INODE_NODATACOW
262 | BTRFS_INODE_NODATASUM;
264 ip->flags |= BTRFS_INODE_NODATACOW;
268 * Revert back under same assumptions as above
271 if (inode->i_size == 0)
272 ip->flags &= ~(BTRFS_INODE_NODATACOW
273 | BTRFS_INODE_NODATASUM);
275 ip->flags &= ~BTRFS_INODE_NODATACOW;
280 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
281 * flag may be changed automatically if compression code won't make
284 if (flags & FS_NOCOMP_FL) {
285 ip->flags &= ~BTRFS_INODE_COMPRESS;
286 ip->flags |= BTRFS_INODE_NOCOMPRESS;
288 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
289 if (ret && ret != -ENODATA)
291 } else if (flags & FS_COMPR_FL) {
294 ip->flags |= BTRFS_INODE_COMPRESS;
295 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
297 if (fs_info->compress_type == BTRFS_COMPRESS_LZO)
301 ret = btrfs_set_prop(inode, "btrfs.compression",
302 comp, strlen(comp), 0);
307 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
308 if (ret && ret != -ENODATA)
310 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
313 trans = btrfs_start_transaction(root, 1);
315 ret = PTR_ERR(trans);
319 btrfs_update_iflags(inode);
320 inode_inc_iversion(inode);
321 inode->i_ctime = current_time(inode);
322 ret = btrfs_update_inode(trans, root, inode);
324 btrfs_end_transaction(trans);
327 ip->flags = ip_oldflags;
328 inode->i_flags = i_oldflags;
333 mnt_drop_write_file(file);
337 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
339 struct inode *inode = file_inode(file);
341 return put_user(inode->i_generation, arg);
344 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
346 struct inode *inode = file_inode(file);
347 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
348 struct btrfs_device *device;
349 struct request_queue *q;
350 struct fstrim_range range;
351 u64 minlen = ULLONG_MAX;
353 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
356 if (!capable(CAP_SYS_ADMIN))
360 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
364 q = bdev_get_queue(device->bdev);
365 if (blk_queue_discard(q)) {
367 minlen = min_t(u64, q->limits.discard_granularity,
375 if (copy_from_user(&range, arg, sizeof(range)))
377 if (range.start > total_bytes ||
378 range.len < fs_info->sb->s_blocksize)
381 range.len = min(range.len, total_bytes - range.start);
382 range.minlen = max(range.minlen, minlen);
383 ret = btrfs_trim_fs(fs_info, &range);
387 if (copy_to_user(arg, &range, sizeof(range)))
393 int btrfs_is_empty_uuid(u8 *uuid)
397 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
404 static noinline int create_subvol(struct inode *dir,
405 struct dentry *dentry,
406 const char *name, int namelen,
408 struct btrfs_qgroup_inherit *inherit)
410 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
411 struct btrfs_trans_handle *trans;
412 struct btrfs_key key;
413 struct btrfs_root_item *root_item;
414 struct btrfs_inode_item *inode_item;
415 struct extent_buffer *leaf;
416 struct btrfs_root *root = BTRFS_I(dir)->root;
417 struct btrfs_root *new_root;
418 struct btrfs_block_rsv block_rsv;
419 struct timespec cur_time = current_time(dir);
424 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
429 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
433 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
438 * Don't create subvolume whose level is not zero. Or qgroup will be
439 * screwed up since it assumes subvolume qgroup's level to be 0.
441 if (btrfs_qgroup_level(objectid)) {
446 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
448 * The same as the snapshot creation, please see the comment
449 * of create_snapshot().
451 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
452 8, &qgroup_reserved, false);
456 trans = btrfs_start_transaction(root, 0);
458 ret = PTR_ERR(trans);
459 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
462 trans->block_rsv = &block_rsv;
463 trans->bytes_reserved = block_rsv.size;
465 ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit);
469 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
475 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
476 btrfs_set_header_bytenr(leaf, leaf->start);
477 btrfs_set_header_generation(leaf, trans->transid);
478 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
479 btrfs_set_header_owner(leaf, objectid);
481 write_extent_buffer_fsid(leaf, fs_info->fsid);
482 write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid);
483 btrfs_mark_buffer_dirty(leaf);
485 inode_item = &root_item->inode;
486 btrfs_set_stack_inode_generation(inode_item, 1);
487 btrfs_set_stack_inode_size(inode_item, 3);
488 btrfs_set_stack_inode_nlink(inode_item, 1);
489 btrfs_set_stack_inode_nbytes(inode_item,
491 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
493 btrfs_set_root_flags(root_item, 0);
494 btrfs_set_root_limit(root_item, 0);
495 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
497 btrfs_set_root_bytenr(root_item, leaf->start);
498 btrfs_set_root_generation(root_item, trans->transid);
499 btrfs_set_root_level(root_item, 0);
500 btrfs_set_root_refs(root_item, 1);
501 btrfs_set_root_used(root_item, leaf->len);
502 btrfs_set_root_last_snapshot(root_item, 0);
504 btrfs_set_root_generation_v2(root_item,
505 btrfs_root_generation(root_item));
506 uuid_le_gen(&new_uuid);
507 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
508 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
509 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
510 root_item->ctime = root_item->otime;
511 btrfs_set_root_ctransid(root_item, trans->transid);
512 btrfs_set_root_otransid(root_item, trans->transid);
514 btrfs_tree_unlock(leaf);
515 free_extent_buffer(leaf);
518 btrfs_set_root_dirid(root_item, new_dirid);
520 key.objectid = objectid;
522 key.type = BTRFS_ROOT_ITEM_KEY;
523 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
528 key.offset = (u64)-1;
529 new_root = btrfs_read_fs_root_no_name(fs_info, &key);
530 if (IS_ERR(new_root)) {
531 ret = PTR_ERR(new_root);
532 btrfs_abort_transaction(trans, ret);
536 btrfs_record_root_in_trans(trans, new_root);
538 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
540 /* We potentially lose an unused inode item here */
541 btrfs_abort_transaction(trans, ret);
545 mutex_lock(&new_root->objectid_mutex);
546 new_root->highest_objectid = new_dirid;
547 mutex_unlock(&new_root->objectid_mutex);
550 * insert the directory item
552 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
554 btrfs_abort_transaction(trans, ret);
558 ret = btrfs_insert_dir_item(trans, root,
559 name, namelen, BTRFS_I(dir), &key,
560 BTRFS_FT_DIR, index);
562 btrfs_abort_transaction(trans, ret);
566 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
567 ret = btrfs_update_inode(trans, root, dir);
570 ret = btrfs_add_root_ref(trans, fs_info,
571 objectid, root->root_key.objectid,
572 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
575 ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid,
576 BTRFS_UUID_KEY_SUBVOL, objectid);
578 btrfs_abort_transaction(trans, ret);
582 trans->block_rsv = NULL;
583 trans->bytes_reserved = 0;
584 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
587 *async_transid = trans->transid;
588 err = btrfs_commit_transaction_async(trans, 1);
590 err = btrfs_commit_transaction(trans);
592 err = btrfs_commit_transaction(trans);
598 inode = btrfs_lookup_dentry(dir, dentry);
600 return PTR_ERR(inode);
601 d_instantiate(dentry, inode);
610 static void btrfs_wait_for_no_snapshotting_writes(struct btrfs_root *root)
616 prepare_to_wait(&root->subv_writers->wait, &wait,
617 TASK_UNINTERRUPTIBLE);
619 writers = percpu_counter_sum(&root->subv_writers->counter);
623 finish_wait(&root->subv_writers->wait, &wait);
627 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
628 struct dentry *dentry,
629 u64 *async_transid, bool readonly,
630 struct btrfs_qgroup_inherit *inherit)
632 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
634 struct btrfs_pending_snapshot *pending_snapshot;
635 struct btrfs_trans_handle *trans;
638 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
641 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
642 if (!pending_snapshot)
645 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
647 pending_snapshot->path = btrfs_alloc_path();
648 if (!pending_snapshot->root_item || !pending_snapshot->path) {
653 atomic_inc(&root->will_be_snapshotted);
654 smp_mb__after_atomic();
655 btrfs_wait_for_no_snapshotting_writes(root);
657 ret = btrfs_start_delalloc_inodes(root, 0);
661 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
663 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
664 BTRFS_BLOCK_RSV_TEMP);
666 * 1 - parent dir inode
669 * 2 - root ref/backref
670 * 1 - root of snapshot
673 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
674 &pending_snapshot->block_rsv, 8,
675 &pending_snapshot->qgroup_reserved,
680 pending_snapshot->dentry = dentry;
681 pending_snapshot->root = root;
682 pending_snapshot->readonly = readonly;
683 pending_snapshot->dir = dir;
684 pending_snapshot->inherit = inherit;
686 trans = btrfs_start_transaction(root, 0);
688 ret = PTR_ERR(trans);
692 spin_lock(&fs_info->trans_lock);
693 list_add(&pending_snapshot->list,
694 &trans->transaction->pending_snapshots);
695 spin_unlock(&fs_info->trans_lock);
697 *async_transid = trans->transid;
698 ret = btrfs_commit_transaction_async(trans, 1);
700 ret = btrfs_commit_transaction(trans);
702 ret = btrfs_commit_transaction(trans);
707 ret = pending_snapshot->error;
711 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
715 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
717 ret = PTR_ERR(inode);
721 d_instantiate(dentry, inode);
724 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
726 if (atomic_dec_and_test(&root->will_be_snapshotted))
727 wake_up_atomic_t(&root->will_be_snapshotted);
729 kfree(pending_snapshot->root_item);
730 btrfs_free_path(pending_snapshot->path);
731 kfree(pending_snapshot);
736 /* copy of may_delete in fs/namei.c()
737 * Check whether we can remove a link victim from directory dir, check
738 * whether the type of victim is right.
739 * 1. We can't do it if dir is read-only (done in permission())
740 * 2. We should have write and exec permissions on dir
741 * 3. We can't remove anything from append-only dir
742 * 4. We can't do anything with immutable dir (done in permission())
743 * 5. If the sticky bit on dir is set we should either
744 * a. be owner of dir, or
745 * b. be owner of victim, or
746 * c. have CAP_FOWNER capability
747 * 6. If the victim is append-only or immutable we can't do anything with
748 * links pointing to it.
749 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
750 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
751 * 9. We can't remove a root or mountpoint.
752 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
753 * nfs_async_unlink().
756 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
760 if (d_really_is_negative(victim))
763 BUG_ON(d_inode(victim->d_parent) != dir);
764 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
766 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
771 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
772 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
775 if (!d_is_dir(victim))
779 } else if (d_is_dir(victim))
783 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
788 /* copy of may_create in fs/namei.c() */
789 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
791 if (d_really_is_positive(child))
795 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
799 * Create a new subvolume below @parent. This is largely modeled after
800 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
801 * inside this filesystem so it's quite a bit simpler.
803 static noinline int btrfs_mksubvol(const struct path *parent,
804 const char *name, int namelen,
805 struct btrfs_root *snap_src,
806 u64 *async_transid, bool readonly,
807 struct btrfs_qgroup_inherit *inherit)
809 struct inode *dir = d_inode(parent->dentry);
810 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
811 struct dentry *dentry;
814 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
818 dentry = lookup_one_len(name, parent->dentry, namelen);
819 error = PTR_ERR(dentry);
823 error = btrfs_may_create(dir, dentry);
828 * even if this name doesn't exist, we may get hash collisions.
829 * check for them now when we can safely fail
831 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
837 down_read(&fs_info->subvol_sem);
839 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
843 error = create_snapshot(snap_src, dir, dentry,
844 async_transid, readonly, inherit);
846 error = create_subvol(dir, dentry, name, namelen,
847 async_transid, inherit);
850 fsnotify_mkdir(dir, dentry);
852 up_read(&fs_info->subvol_sem);
861 * When we're defragging a range, we don't want to kick it off again
862 * if it is really just waiting for delalloc to send it down.
863 * If we find a nice big extent or delalloc range for the bytes in the
864 * file you want to defrag, we return 0 to let you know to skip this
867 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
869 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
870 struct extent_map *em = NULL;
871 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
874 read_lock(&em_tree->lock);
875 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
876 read_unlock(&em_tree->lock);
879 end = extent_map_end(em);
881 if (end - offset > thresh)
884 /* if we already have a nice delalloc here, just stop */
886 end = count_range_bits(io_tree, &offset, offset + thresh,
887 thresh, EXTENT_DELALLOC, 1);
894 * helper function to walk through a file and find extents
895 * newer than a specific transid, and smaller than thresh.
897 * This is used by the defragging code to find new and small
900 static int find_new_extents(struct btrfs_root *root,
901 struct inode *inode, u64 newer_than,
902 u64 *off, u32 thresh)
904 struct btrfs_path *path;
905 struct btrfs_key min_key;
906 struct extent_buffer *leaf;
907 struct btrfs_file_extent_item *extent;
910 u64 ino = btrfs_ino(BTRFS_I(inode));
912 path = btrfs_alloc_path();
916 min_key.objectid = ino;
917 min_key.type = BTRFS_EXTENT_DATA_KEY;
918 min_key.offset = *off;
921 ret = btrfs_search_forward(root, &min_key, path, newer_than);
925 if (min_key.objectid != ino)
927 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
930 leaf = path->nodes[0];
931 extent = btrfs_item_ptr(leaf, path->slots[0],
932 struct btrfs_file_extent_item);
934 type = btrfs_file_extent_type(leaf, extent);
935 if (type == BTRFS_FILE_EXTENT_REG &&
936 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
937 check_defrag_in_cache(inode, min_key.offset, thresh)) {
938 *off = min_key.offset;
939 btrfs_free_path(path);
944 if (path->slots[0] < btrfs_header_nritems(leaf)) {
945 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
949 if (min_key.offset == (u64)-1)
953 btrfs_release_path(path);
956 btrfs_free_path(path);
960 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
962 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
963 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
964 struct extent_map *em;
968 * hopefully we have this extent in the tree already, try without
969 * the full extent lock
971 read_lock(&em_tree->lock);
972 em = lookup_extent_mapping(em_tree, start, len);
973 read_unlock(&em_tree->lock);
976 struct extent_state *cached = NULL;
977 u64 end = start + len - 1;
979 /* get the big lock and read metadata off disk */
980 lock_extent_bits(io_tree, start, end, &cached);
981 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0);
982 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
991 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
993 struct extent_map *next;
996 /* this is the last extent */
997 if (em->start + em->len >= i_size_read(inode))
1000 next = defrag_lookup_extent(inode, em->start + em->len);
1001 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1003 else if ((em->block_start + em->block_len == next->block_start) &&
1004 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1007 free_extent_map(next);
1011 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1012 u64 *last_len, u64 *skip, u64 *defrag_end,
1015 struct extent_map *em;
1017 bool next_mergeable = true;
1018 bool prev_mergeable = true;
1021 * make sure that once we start defragging an extent, we keep on
1024 if (start < *defrag_end)
1029 em = defrag_lookup_extent(inode, start);
1033 /* this will cover holes, and inline extents */
1034 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1040 prev_mergeable = false;
1042 next_mergeable = defrag_check_next_extent(inode, em);
1044 * we hit a real extent, if it is big or the next extent is not a
1045 * real extent, don't bother defragging it
1047 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1048 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1052 * last_len ends up being a counter of how many bytes we've defragged.
1053 * every time we choose not to defrag an extent, we reset *last_len
1054 * so that the next tiny extent will force a defrag.
1056 * The end result of this is that tiny extents before a single big
1057 * extent will force at least part of that big extent to be defragged.
1060 *defrag_end = extent_map_end(em);
1063 *skip = extent_map_end(em);
1067 free_extent_map(em);
1072 * it doesn't do much good to defrag one or two pages
1073 * at a time. This pulls in a nice chunk of pages
1074 * to COW and defrag.
1076 * It also makes sure the delalloc code has enough
1077 * dirty data to avoid making new small extents as part
1080 * It's a good idea to start RA on this range
1081 * before calling this.
1083 static int cluster_pages_for_defrag(struct inode *inode,
1084 struct page **pages,
1085 unsigned long start_index,
1086 unsigned long num_pages)
1088 unsigned long file_end;
1089 u64 isize = i_size_read(inode);
1096 struct btrfs_ordered_extent *ordered;
1097 struct extent_state *cached_state = NULL;
1098 struct extent_io_tree *tree;
1099 struct extent_changeset *data_reserved = NULL;
1100 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1102 file_end = (isize - 1) >> PAGE_SHIFT;
1103 if (!isize || start_index > file_end)
1106 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1108 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
1109 start_index << PAGE_SHIFT,
1110 page_cnt << PAGE_SHIFT);
1114 tree = &BTRFS_I(inode)->io_tree;
1116 /* step one, lock all the pages */
1117 for (i = 0; i < page_cnt; i++) {
1120 page = find_or_create_page(inode->i_mapping,
1121 start_index + i, mask);
1125 page_start = page_offset(page);
1126 page_end = page_start + PAGE_SIZE - 1;
1128 lock_extent_bits(tree, page_start, page_end,
1130 ordered = btrfs_lookup_ordered_extent(inode,
1132 unlock_extent_cached(tree, page_start, page_end,
1133 &cached_state, GFP_NOFS);
1138 btrfs_start_ordered_extent(inode, ordered, 1);
1139 btrfs_put_ordered_extent(ordered);
1142 * we unlocked the page above, so we need check if
1143 * it was released or not.
1145 if (page->mapping != inode->i_mapping) {
1152 if (!PageUptodate(page)) {
1153 btrfs_readpage(NULL, page);
1155 if (!PageUptodate(page)) {
1163 if (page->mapping != inode->i_mapping) {
1175 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1179 * so now we have a nice long stream of locked
1180 * and up to date pages, lets wait on them
1182 for (i = 0; i < i_done; i++)
1183 wait_on_page_writeback(pages[i]);
1185 page_start = page_offset(pages[0]);
1186 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1188 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1189 page_start, page_end - 1, &cached_state);
1190 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1191 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1192 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1193 &cached_state, GFP_NOFS);
1195 if (i_done != page_cnt) {
1196 spin_lock(&BTRFS_I(inode)->lock);
1197 BTRFS_I(inode)->outstanding_extents++;
1198 spin_unlock(&BTRFS_I(inode)->lock);
1199 btrfs_delalloc_release_space(inode, data_reserved,
1200 start_index << PAGE_SHIFT,
1201 (page_cnt - i_done) << PAGE_SHIFT);
1205 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1208 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1209 page_start, page_end - 1, &cached_state,
1212 for (i = 0; i < i_done; i++) {
1213 clear_page_dirty_for_io(pages[i]);
1214 ClearPageChecked(pages[i]);
1215 set_page_extent_mapped(pages[i]);
1216 set_page_dirty(pages[i]);
1217 unlock_page(pages[i]);
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);
1230 extent_changeset_free(data_reserved);
1235 int btrfs_defrag_file(struct inode *inode, struct file *file,
1236 struct btrfs_ioctl_defrag_range_args *range,
1237 u64 newer_than, unsigned long max_to_defrag)
1239 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1240 struct btrfs_root *root = BTRFS_I(inode)->root;
1241 struct file_ra_state *ra = NULL;
1242 unsigned long last_index;
1243 u64 isize = i_size_read(inode);
1247 u64 newer_off = range->start;
1249 unsigned long ra_index = 0;
1251 int defrag_count = 0;
1252 int compress_type = BTRFS_COMPRESS_ZLIB;
1253 u32 extent_thresh = range->extent_thresh;
1254 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1255 unsigned long cluster = max_cluster;
1256 u64 new_align = ~((u64)SZ_128K - 1);
1257 struct page **pages = NULL;
1262 if (range->start >= isize)
1265 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1266 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1268 if (range->compress_type)
1269 compress_type = range->compress_type;
1272 if (extent_thresh == 0)
1273 extent_thresh = SZ_256K;
1276 * If we were not given a file, allocate a readahead context. As
1277 * readahead is just an optimization, defrag will work without it so
1278 * we don't error out.
1281 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1283 file_ra_state_init(ra, inode->i_mapping);
1288 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1294 /* find the last page to defrag */
1295 if (range->start + range->len > range->start) {
1296 last_index = min_t(u64, isize - 1,
1297 range->start + range->len - 1) >> PAGE_SHIFT;
1299 last_index = (isize - 1) >> PAGE_SHIFT;
1303 ret = find_new_extents(root, inode, newer_than,
1304 &newer_off, SZ_64K);
1306 range->start = newer_off;
1308 * we always align our defrag to help keep
1309 * the extents in the file evenly spaced
1311 i = (newer_off & new_align) >> PAGE_SHIFT;
1315 i = range->start >> PAGE_SHIFT;
1318 max_to_defrag = last_index - i + 1;
1321 * make writeback starts from i, so the defrag range can be
1322 * written sequentially.
1324 if (i < inode->i_mapping->writeback_index)
1325 inode->i_mapping->writeback_index = i;
1327 while (i <= last_index && defrag_count < max_to_defrag &&
1328 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1330 * make sure we stop running if someone unmounts
1333 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1336 if (btrfs_defrag_cancelled(fs_info)) {
1337 btrfs_debug(fs_info, "defrag_file cancelled");
1342 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1343 extent_thresh, &last_len, &skip,
1344 &defrag_end, range->flags &
1345 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1348 * the should_defrag function tells us how much to skip
1349 * bump our counter by the suggested amount
1351 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1352 i = max(i + 1, next);
1357 cluster = (PAGE_ALIGN(defrag_end) >>
1359 cluster = min(cluster, max_cluster);
1361 cluster = max_cluster;
1364 if (i + cluster > ra_index) {
1365 ra_index = max(i, ra_index);
1367 page_cache_sync_readahead(inode->i_mapping, ra,
1368 file, ra_index, cluster);
1369 ra_index += cluster;
1373 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1374 BTRFS_I(inode)->defrag_compress = compress_type;
1375 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1377 inode_unlock(inode);
1381 defrag_count += ret;
1382 balance_dirty_pages_ratelimited(inode->i_mapping);
1383 inode_unlock(inode);
1386 if (newer_off == (u64)-1)
1392 newer_off = max(newer_off + 1,
1393 (u64)i << PAGE_SHIFT);
1395 ret = find_new_extents(root, inode, newer_than,
1396 &newer_off, SZ_64K);
1398 range->start = newer_off;
1399 i = (newer_off & new_align) >> PAGE_SHIFT;
1406 last_len += ret << PAGE_SHIFT;
1414 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1415 filemap_flush(inode->i_mapping);
1416 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1417 &BTRFS_I(inode)->runtime_flags))
1418 filemap_flush(inode->i_mapping);
1421 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1422 /* the filemap_flush will queue IO into the worker threads, but
1423 * we have to make sure the IO is actually started and that
1424 * ordered extents get created before we return
1426 atomic_inc(&fs_info->async_submit_draining);
1427 while (atomic_read(&fs_info->nr_async_submits) ||
1428 atomic_read(&fs_info->async_delalloc_pages)) {
1429 wait_event(fs_info->async_submit_wait,
1430 (atomic_read(&fs_info->nr_async_submits) == 0 &&
1431 atomic_read(&fs_info->async_delalloc_pages) == 0));
1433 atomic_dec(&fs_info->async_submit_draining);
1436 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1437 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1443 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1445 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1446 inode_unlock(inode);
1454 static noinline int btrfs_ioctl_resize(struct file *file,
1457 struct inode *inode = file_inode(file);
1458 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1462 struct btrfs_root *root = BTRFS_I(inode)->root;
1463 struct btrfs_ioctl_vol_args *vol_args;
1464 struct btrfs_trans_handle *trans;
1465 struct btrfs_device *device = NULL;
1468 char *devstr = NULL;
1472 if (!capable(CAP_SYS_ADMIN))
1475 ret = mnt_want_write_file(file);
1479 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
1480 mnt_drop_write_file(file);
1481 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1484 mutex_lock(&fs_info->volume_mutex);
1485 vol_args = memdup_user(arg, sizeof(*vol_args));
1486 if (IS_ERR(vol_args)) {
1487 ret = PTR_ERR(vol_args);
1491 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1493 sizestr = vol_args->name;
1494 devstr = strchr(sizestr, ':');
1496 sizestr = devstr + 1;
1498 devstr = vol_args->name;
1499 ret = kstrtoull(devstr, 10, &devid);
1506 btrfs_info(fs_info, "resizing devid %llu", devid);
1509 device = btrfs_find_device(fs_info, devid, NULL, NULL);
1511 btrfs_info(fs_info, "resizer unable to find device %llu",
1517 if (!device->writeable) {
1519 "resizer unable to apply on readonly device %llu",
1525 if (!strcmp(sizestr, "max"))
1526 new_size = device->bdev->bd_inode->i_size;
1528 if (sizestr[0] == '-') {
1531 } else if (sizestr[0] == '+') {
1535 new_size = memparse(sizestr, &retptr);
1536 if (*retptr != '\0' || new_size == 0) {
1542 if (device->is_tgtdev_for_dev_replace) {
1547 old_size = btrfs_device_get_total_bytes(device);
1550 if (new_size > old_size) {
1554 new_size = old_size - new_size;
1555 } else if (mod > 0) {
1556 if (new_size > ULLONG_MAX - old_size) {
1560 new_size = old_size + new_size;
1563 if (new_size < SZ_256M) {
1567 if (new_size > device->bdev->bd_inode->i_size) {
1572 new_size = round_down(new_size, fs_info->sectorsize);
1574 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1575 rcu_str_deref(device->name), new_size);
1577 if (new_size > old_size) {
1578 trans = btrfs_start_transaction(root, 0);
1579 if (IS_ERR(trans)) {
1580 ret = PTR_ERR(trans);
1583 ret = btrfs_grow_device(trans, device, new_size);
1584 btrfs_commit_transaction(trans);
1585 } else if (new_size < old_size) {
1586 ret = btrfs_shrink_device(device, new_size);
1587 } /* equal, nothing need to do */
1592 mutex_unlock(&fs_info->volume_mutex);
1593 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
1594 mnt_drop_write_file(file);
1598 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1599 const char *name, unsigned long fd, int subvol,
1600 u64 *transid, bool readonly,
1601 struct btrfs_qgroup_inherit *inherit)
1606 if (!S_ISDIR(file_inode(file)->i_mode))
1609 ret = mnt_want_write_file(file);
1613 namelen = strlen(name);
1614 if (strchr(name, '/')) {
1616 goto out_drop_write;
1619 if (name[0] == '.' &&
1620 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1622 goto out_drop_write;
1626 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1627 NULL, transid, readonly, inherit);
1629 struct fd src = fdget(fd);
1630 struct inode *src_inode;
1633 goto out_drop_write;
1636 src_inode = file_inode(src.file);
1637 if (src_inode->i_sb != file_inode(file)->i_sb) {
1638 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1639 "Snapshot src from another FS");
1641 } else if (!inode_owner_or_capable(src_inode)) {
1643 * Subvolume creation is not restricted, but snapshots
1644 * are limited to own subvolumes only
1648 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1649 BTRFS_I(src_inode)->root,
1650 transid, readonly, inherit);
1655 mnt_drop_write_file(file);
1660 static noinline int btrfs_ioctl_snap_create(struct file *file,
1661 void __user *arg, int subvol)
1663 struct btrfs_ioctl_vol_args *vol_args;
1666 if (!S_ISDIR(file_inode(file)->i_mode))
1669 vol_args = memdup_user(arg, sizeof(*vol_args));
1670 if (IS_ERR(vol_args))
1671 return PTR_ERR(vol_args);
1672 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1674 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1675 vol_args->fd, subvol,
1682 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1683 void __user *arg, int subvol)
1685 struct btrfs_ioctl_vol_args_v2 *vol_args;
1689 bool readonly = false;
1690 struct btrfs_qgroup_inherit *inherit = NULL;
1692 if (!S_ISDIR(file_inode(file)->i_mode))
1695 vol_args = memdup_user(arg, sizeof(*vol_args));
1696 if (IS_ERR(vol_args))
1697 return PTR_ERR(vol_args);
1698 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1700 if (vol_args->flags &
1701 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1702 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1707 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1709 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1711 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1712 if (vol_args->size > PAGE_SIZE) {
1716 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1717 if (IS_ERR(inherit)) {
1718 ret = PTR_ERR(inherit);
1723 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1724 vol_args->fd, subvol, ptr,
1729 if (ptr && copy_to_user(arg +
1730 offsetof(struct btrfs_ioctl_vol_args_v2,
1742 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1745 struct inode *inode = file_inode(file);
1746 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1747 struct btrfs_root *root = BTRFS_I(inode)->root;
1751 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1754 down_read(&fs_info->subvol_sem);
1755 if (btrfs_root_readonly(root))
1756 flags |= BTRFS_SUBVOL_RDONLY;
1757 up_read(&fs_info->subvol_sem);
1759 if (copy_to_user(arg, &flags, sizeof(flags)))
1765 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1768 struct inode *inode = file_inode(file);
1769 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1770 struct btrfs_root *root = BTRFS_I(inode)->root;
1771 struct btrfs_trans_handle *trans;
1776 if (!inode_owner_or_capable(inode))
1779 ret = mnt_want_write_file(file);
1783 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1785 goto out_drop_write;
1788 if (copy_from_user(&flags, arg, sizeof(flags))) {
1790 goto out_drop_write;
1793 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1795 goto out_drop_write;
1798 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1800 goto out_drop_write;
1803 down_write(&fs_info->subvol_sem);
1806 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1809 root_flags = btrfs_root_flags(&root->root_item);
1810 if (flags & BTRFS_SUBVOL_RDONLY) {
1811 btrfs_set_root_flags(&root->root_item,
1812 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1815 * Block RO -> RW transition if this subvolume is involved in
1818 spin_lock(&root->root_item_lock);
1819 if (root->send_in_progress == 0) {
1820 btrfs_set_root_flags(&root->root_item,
1821 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1822 spin_unlock(&root->root_item_lock);
1824 spin_unlock(&root->root_item_lock);
1826 "Attempt to set subvolume %llu read-write during send",
1827 root->root_key.objectid);
1833 trans = btrfs_start_transaction(root, 1);
1834 if (IS_ERR(trans)) {
1835 ret = PTR_ERR(trans);
1839 ret = btrfs_update_root(trans, fs_info->tree_root,
1840 &root->root_key, &root->root_item);
1842 btrfs_commit_transaction(trans);
1845 btrfs_set_root_flags(&root->root_item, root_flags);
1847 up_write(&fs_info->subvol_sem);
1849 mnt_drop_write_file(file);
1855 * helper to check if the subvolume references other subvolumes
1857 static noinline int may_destroy_subvol(struct btrfs_root *root)
1859 struct btrfs_fs_info *fs_info = root->fs_info;
1860 struct btrfs_path *path;
1861 struct btrfs_dir_item *di;
1862 struct btrfs_key key;
1866 path = btrfs_alloc_path();
1870 /* Make sure this root isn't set as the default subvol */
1871 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1872 di = btrfs_lookup_dir_item(NULL, fs_info->tree_root, path,
1873 dir_id, "default", 7, 0);
1874 if (di && !IS_ERR(di)) {
1875 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1876 if (key.objectid == root->root_key.objectid) {
1879 "deleting default subvolume %llu is not allowed",
1883 btrfs_release_path(path);
1886 key.objectid = root->root_key.objectid;
1887 key.type = BTRFS_ROOT_REF_KEY;
1888 key.offset = (u64)-1;
1890 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
1896 if (path->slots[0] > 0) {
1898 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1899 if (key.objectid == root->root_key.objectid &&
1900 key.type == BTRFS_ROOT_REF_KEY)
1904 btrfs_free_path(path);
1908 static noinline int key_in_sk(struct btrfs_key *key,
1909 struct btrfs_ioctl_search_key *sk)
1911 struct btrfs_key test;
1914 test.objectid = sk->min_objectid;
1915 test.type = sk->min_type;
1916 test.offset = sk->min_offset;
1918 ret = btrfs_comp_cpu_keys(key, &test);
1922 test.objectid = sk->max_objectid;
1923 test.type = sk->max_type;
1924 test.offset = sk->max_offset;
1926 ret = btrfs_comp_cpu_keys(key, &test);
1932 static noinline int copy_to_sk(struct btrfs_path *path,
1933 struct btrfs_key *key,
1934 struct btrfs_ioctl_search_key *sk,
1937 unsigned long *sk_offset,
1941 struct extent_buffer *leaf;
1942 struct btrfs_ioctl_search_header sh;
1943 struct btrfs_key test;
1944 unsigned long item_off;
1945 unsigned long item_len;
1951 leaf = path->nodes[0];
1952 slot = path->slots[0];
1953 nritems = btrfs_header_nritems(leaf);
1955 if (btrfs_header_generation(leaf) > sk->max_transid) {
1959 found_transid = btrfs_header_generation(leaf);
1961 for (i = slot; i < nritems; i++) {
1962 item_off = btrfs_item_ptr_offset(leaf, i);
1963 item_len = btrfs_item_size_nr(leaf, i);
1965 btrfs_item_key_to_cpu(leaf, key, i);
1966 if (!key_in_sk(key, sk))
1969 if (sizeof(sh) + item_len > *buf_size) {
1976 * return one empty item back for v1, which does not
1980 *buf_size = sizeof(sh) + item_len;
1985 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1990 sh.objectid = key->objectid;
1991 sh.offset = key->offset;
1992 sh.type = key->type;
1994 sh.transid = found_transid;
1996 /* copy search result header */
1997 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2002 *sk_offset += sizeof(sh);
2005 char __user *up = ubuf + *sk_offset;
2007 if (read_extent_buffer_to_user(leaf, up,
2008 item_off, item_len)) {
2013 *sk_offset += item_len;
2017 if (ret) /* -EOVERFLOW from above */
2020 if (*num_found >= sk->nr_items) {
2027 test.objectid = sk->max_objectid;
2028 test.type = sk->max_type;
2029 test.offset = sk->max_offset;
2030 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2032 else if (key->offset < (u64)-1)
2034 else if (key->type < (u8)-1) {
2037 } else if (key->objectid < (u64)-1) {
2045 * 0: all items from this leaf copied, continue with next
2046 * 1: * more items can be copied, but unused buffer is too small
2047 * * all items were found
2048 * Either way, it will stops the loop which iterates to the next
2050 * -EOVERFLOW: item was to large for buffer
2051 * -EFAULT: could not copy extent buffer back to userspace
2056 static noinline int search_ioctl(struct inode *inode,
2057 struct btrfs_ioctl_search_key *sk,
2061 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2062 struct btrfs_root *root;
2063 struct btrfs_key key;
2064 struct btrfs_path *path;
2067 unsigned long sk_offset = 0;
2069 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2070 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2074 path = btrfs_alloc_path();
2078 if (sk->tree_id == 0) {
2079 /* search the root of the inode that was passed */
2080 root = BTRFS_I(inode)->root;
2082 key.objectid = sk->tree_id;
2083 key.type = BTRFS_ROOT_ITEM_KEY;
2084 key.offset = (u64)-1;
2085 root = btrfs_read_fs_root_no_name(info, &key);
2087 btrfs_free_path(path);
2092 key.objectid = sk->min_objectid;
2093 key.type = sk->min_type;
2094 key.offset = sk->min_offset;
2097 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2103 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2104 &sk_offset, &num_found);
2105 btrfs_release_path(path);
2113 sk->nr_items = num_found;
2114 btrfs_free_path(path);
2118 static noinline int btrfs_ioctl_tree_search(struct file *file,
2121 struct btrfs_ioctl_search_args __user *uargs;
2122 struct btrfs_ioctl_search_key sk;
2123 struct inode *inode;
2127 if (!capable(CAP_SYS_ADMIN))
2130 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2132 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2135 buf_size = sizeof(uargs->buf);
2137 inode = file_inode(file);
2138 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2141 * In the origin implementation an overflow is handled by returning a
2142 * search header with a len of zero, so reset ret.
2144 if (ret == -EOVERFLOW)
2147 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2152 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2155 struct btrfs_ioctl_search_args_v2 __user *uarg;
2156 struct btrfs_ioctl_search_args_v2 args;
2157 struct inode *inode;
2160 const size_t buf_limit = SZ_16M;
2162 if (!capable(CAP_SYS_ADMIN))
2165 /* copy search header and buffer size */
2166 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2167 if (copy_from_user(&args, uarg, sizeof(args)))
2170 buf_size = args.buf_size;
2172 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2175 /* limit result size to 16MB */
2176 if (buf_size > buf_limit)
2177 buf_size = buf_limit;
2179 inode = file_inode(file);
2180 ret = search_ioctl(inode, &args.key, &buf_size,
2181 (char *)(&uarg->buf[0]));
2182 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2184 else if (ret == -EOVERFLOW &&
2185 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2192 * Search INODE_REFs to identify path name of 'dirid' directory
2193 * in a 'tree_id' tree. and sets path name to 'name'.
2195 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2196 u64 tree_id, u64 dirid, char *name)
2198 struct btrfs_root *root;
2199 struct btrfs_key key;
2205 struct btrfs_inode_ref *iref;
2206 struct extent_buffer *l;
2207 struct btrfs_path *path;
2209 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2214 path = btrfs_alloc_path();
2218 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2220 key.objectid = tree_id;
2221 key.type = BTRFS_ROOT_ITEM_KEY;
2222 key.offset = (u64)-1;
2223 root = btrfs_read_fs_root_no_name(info, &key);
2225 btrfs_err(info, "could not find root %llu", tree_id);
2230 key.objectid = dirid;
2231 key.type = BTRFS_INODE_REF_KEY;
2232 key.offset = (u64)-1;
2235 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2239 ret = btrfs_previous_item(root, path, dirid,
2240 BTRFS_INODE_REF_KEY);
2250 slot = path->slots[0];
2251 btrfs_item_key_to_cpu(l, &key, slot);
2253 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2254 len = btrfs_inode_ref_name_len(l, iref);
2256 total_len += len + 1;
2258 ret = -ENAMETOOLONG;
2263 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2265 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2268 btrfs_release_path(path);
2269 key.objectid = key.offset;
2270 key.offset = (u64)-1;
2271 dirid = key.objectid;
2273 memmove(name, ptr, total_len);
2274 name[total_len] = '\0';
2277 btrfs_free_path(path);
2281 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2284 struct btrfs_ioctl_ino_lookup_args *args;
2285 struct inode *inode;
2288 args = memdup_user(argp, sizeof(*args));
2290 return PTR_ERR(args);
2292 inode = file_inode(file);
2295 * Unprivileged query to obtain the containing subvolume root id. The
2296 * path is reset so it's consistent with btrfs_search_path_in_tree.
2298 if (args->treeid == 0)
2299 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2301 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2306 if (!capable(CAP_SYS_ADMIN)) {
2311 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2312 args->treeid, args->objectid,
2316 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2323 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2326 struct dentry *parent = file->f_path.dentry;
2327 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2328 struct dentry *dentry;
2329 struct inode *dir = d_inode(parent);
2330 struct inode *inode;
2331 struct btrfs_root *root = BTRFS_I(dir)->root;
2332 struct btrfs_root *dest = NULL;
2333 struct btrfs_ioctl_vol_args *vol_args;
2334 struct btrfs_trans_handle *trans;
2335 struct btrfs_block_rsv block_rsv;
2337 u64 qgroup_reserved;
2342 if (!S_ISDIR(dir->i_mode))
2345 vol_args = memdup_user(arg, sizeof(*vol_args));
2346 if (IS_ERR(vol_args))
2347 return PTR_ERR(vol_args);
2349 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2350 namelen = strlen(vol_args->name);
2351 if (strchr(vol_args->name, '/') ||
2352 strncmp(vol_args->name, "..", namelen) == 0) {
2357 err = mnt_want_write_file(file);
2362 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2364 goto out_drop_write;
2365 dentry = lookup_one_len(vol_args->name, parent, namelen);
2366 if (IS_ERR(dentry)) {
2367 err = PTR_ERR(dentry);
2368 goto out_unlock_dir;
2371 if (d_really_is_negative(dentry)) {
2376 inode = d_inode(dentry);
2377 dest = BTRFS_I(inode)->root;
2378 if (!capable(CAP_SYS_ADMIN)) {
2380 * Regular user. Only allow this with a special mount
2381 * option, when the user has write+exec access to the
2382 * subvol root, and when rmdir(2) would have been
2385 * Note that this is _not_ check that the subvol is
2386 * empty or doesn't contain data that we wouldn't
2387 * otherwise be able to delete.
2389 * Users who want to delete empty subvols should try
2393 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2397 * Do not allow deletion if the parent dir is the same
2398 * as the dir to be deleted. That means the ioctl
2399 * must be called on the dentry referencing the root
2400 * of the subvol, not a random directory contained
2407 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2412 /* check if subvolume may be deleted by a user */
2413 err = btrfs_may_delete(dir, dentry, 1);
2417 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2425 * Don't allow to delete a subvolume with send in progress. This is
2426 * inside the i_mutex so the error handling that has to drop the bit
2427 * again is not run concurrently.
2429 spin_lock(&dest->root_item_lock);
2430 root_flags = btrfs_root_flags(&dest->root_item);
2431 if (dest->send_in_progress == 0) {
2432 btrfs_set_root_flags(&dest->root_item,
2433 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2434 spin_unlock(&dest->root_item_lock);
2436 spin_unlock(&dest->root_item_lock);
2438 "Attempt to delete subvolume %llu during send",
2439 dest->root_key.objectid);
2441 goto out_unlock_inode;
2444 down_write(&fs_info->subvol_sem);
2446 err = may_destroy_subvol(dest);
2450 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2452 * One for dir inode, two for dir entries, two for root
2455 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2456 5, &qgroup_reserved, true);
2460 trans = btrfs_start_transaction(root, 0);
2461 if (IS_ERR(trans)) {
2462 err = PTR_ERR(trans);
2465 trans->block_rsv = &block_rsv;
2466 trans->bytes_reserved = block_rsv.size;
2468 btrfs_record_snapshot_destroy(trans, BTRFS_I(dir));
2470 ret = btrfs_unlink_subvol(trans, root, dir,
2471 dest->root_key.objectid,
2472 dentry->d_name.name,
2473 dentry->d_name.len);
2476 btrfs_abort_transaction(trans, ret);
2480 btrfs_record_root_in_trans(trans, dest);
2482 memset(&dest->root_item.drop_progress, 0,
2483 sizeof(dest->root_item.drop_progress));
2484 dest->root_item.drop_level = 0;
2485 btrfs_set_root_refs(&dest->root_item, 0);
2487 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2488 ret = btrfs_insert_orphan_item(trans,
2490 dest->root_key.objectid);
2492 btrfs_abort_transaction(trans, ret);
2498 ret = btrfs_uuid_tree_rem(trans, fs_info, dest->root_item.uuid,
2499 BTRFS_UUID_KEY_SUBVOL,
2500 dest->root_key.objectid);
2501 if (ret && ret != -ENOENT) {
2502 btrfs_abort_transaction(trans, ret);
2506 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2507 ret = btrfs_uuid_tree_rem(trans, fs_info,
2508 dest->root_item.received_uuid,
2509 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2510 dest->root_key.objectid);
2511 if (ret && ret != -ENOENT) {
2512 btrfs_abort_transaction(trans, ret);
2519 trans->block_rsv = NULL;
2520 trans->bytes_reserved = 0;
2521 ret = btrfs_end_transaction(trans);
2524 inode->i_flags |= S_DEAD;
2526 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
2528 up_write(&fs_info->subvol_sem);
2530 spin_lock(&dest->root_item_lock);
2531 root_flags = btrfs_root_flags(&dest->root_item);
2532 btrfs_set_root_flags(&dest->root_item,
2533 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2534 spin_unlock(&dest->root_item_lock);
2537 inode_unlock(inode);
2539 d_invalidate(dentry);
2540 btrfs_invalidate_inodes(dest);
2542 ASSERT(dest->send_in_progress == 0);
2545 if (dest->ino_cache_inode) {
2546 iput(dest->ino_cache_inode);
2547 dest->ino_cache_inode = NULL;
2555 mnt_drop_write_file(file);
2561 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2563 struct inode *inode = file_inode(file);
2564 struct btrfs_root *root = BTRFS_I(inode)->root;
2565 struct btrfs_ioctl_defrag_range_args *range;
2568 ret = mnt_want_write_file(file);
2572 if (btrfs_root_readonly(root)) {
2577 switch (inode->i_mode & S_IFMT) {
2579 if (!capable(CAP_SYS_ADMIN)) {
2583 ret = btrfs_defrag_root(root);
2586 if (!(file->f_mode & FMODE_WRITE)) {
2591 range = kzalloc(sizeof(*range), GFP_KERNEL);
2598 if (copy_from_user(range, argp,
2604 /* compression requires us to start the IO */
2605 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2606 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2607 range->extent_thresh = (u32)-1;
2610 /* the rest are all set to zero by kzalloc */
2611 range->len = (u64)-1;
2613 ret = btrfs_defrag_file(file_inode(file), file,
2623 mnt_drop_write_file(file);
2627 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2629 struct btrfs_ioctl_vol_args *vol_args;
2632 if (!capable(CAP_SYS_ADMIN))
2635 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
2636 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2638 mutex_lock(&fs_info->volume_mutex);
2639 vol_args = memdup_user(arg, sizeof(*vol_args));
2640 if (IS_ERR(vol_args)) {
2641 ret = PTR_ERR(vol_args);
2645 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2646 ret = btrfs_init_new_device(fs_info, vol_args->name);
2649 btrfs_info(fs_info, "disk added %s", vol_args->name);
2653 mutex_unlock(&fs_info->volume_mutex);
2654 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2658 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2660 struct inode *inode = file_inode(file);
2661 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2662 struct btrfs_ioctl_vol_args_v2 *vol_args;
2665 if (!capable(CAP_SYS_ADMIN))
2668 ret = mnt_want_write_file(file);
2672 vol_args = memdup_user(arg, sizeof(*vol_args));
2673 if (IS_ERR(vol_args)) {
2674 ret = PTR_ERR(vol_args);
2678 /* Check for compatibility reject unknown flags */
2679 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED)
2682 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2683 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2687 mutex_lock(&fs_info->volume_mutex);
2688 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2689 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
2691 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2692 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2694 mutex_unlock(&fs_info->volume_mutex);
2695 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2698 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2699 btrfs_info(fs_info, "device deleted: id %llu",
2702 btrfs_info(fs_info, "device deleted: %s",
2708 mnt_drop_write_file(file);
2712 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2714 struct inode *inode = file_inode(file);
2715 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2716 struct btrfs_ioctl_vol_args *vol_args;
2719 if (!capable(CAP_SYS_ADMIN))
2722 ret = mnt_want_write_file(file);
2726 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2727 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2728 goto out_drop_write;
2731 vol_args = memdup_user(arg, sizeof(*vol_args));
2732 if (IS_ERR(vol_args)) {
2733 ret = PTR_ERR(vol_args);
2737 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2738 mutex_lock(&fs_info->volume_mutex);
2739 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2740 mutex_unlock(&fs_info->volume_mutex);
2743 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2746 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2748 mnt_drop_write_file(file);
2753 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2756 struct btrfs_ioctl_fs_info_args *fi_args;
2757 struct btrfs_device *device;
2758 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2761 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2765 mutex_lock(&fs_devices->device_list_mutex);
2766 fi_args->num_devices = fs_devices->num_devices;
2767 memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid));
2769 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2770 if (device->devid > fi_args->max_id)
2771 fi_args->max_id = device->devid;
2773 mutex_unlock(&fs_devices->device_list_mutex);
2775 fi_args->nodesize = fs_info->super_copy->nodesize;
2776 fi_args->sectorsize = fs_info->super_copy->sectorsize;
2777 fi_args->clone_alignment = fs_info->super_copy->sectorsize;
2779 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2786 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2789 struct btrfs_ioctl_dev_info_args *di_args;
2790 struct btrfs_device *dev;
2791 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2793 char *s_uuid = NULL;
2795 di_args = memdup_user(arg, sizeof(*di_args));
2796 if (IS_ERR(di_args))
2797 return PTR_ERR(di_args);
2799 if (!btrfs_is_empty_uuid(di_args->uuid))
2800 s_uuid = di_args->uuid;
2802 mutex_lock(&fs_devices->device_list_mutex);
2803 dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL);
2810 di_args->devid = dev->devid;
2811 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2812 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2813 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2815 struct rcu_string *name;
2818 name = rcu_dereference(dev->name);
2819 strncpy(di_args->path, name->str, sizeof(di_args->path));
2821 di_args->path[sizeof(di_args->path) - 1] = 0;
2823 di_args->path[0] = '\0';
2827 mutex_unlock(&fs_devices->device_list_mutex);
2828 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2835 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2839 page = grab_cache_page(inode->i_mapping, index);
2841 return ERR_PTR(-ENOMEM);
2843 if (!PageUptodate(page)) {
2846 ret = btrfs_readpage(NULL, page);
2848 return ERR_PTR(ret);
2850 if (!PageUptodate(page)) {
2853 return ERR_PTR(-EIO);
2855 if (page->mapping != inode->i_mapping) {
2858 return ERR_PTR(-EAGAIN);
2865 static int gather_extent_pages(struct inode *inode, struct page **pages,
2866 int num_pages, u64 off)
2869 pgoff_t index = off >> PAGE_SHIFT;
2871 for (i = 0; i < num_pages; i++) {
2873 pages[i] = extent_same_get_page(inode, index + i);
2874 if (IS_ERR(pages[i])) {
2875 int err = PTR_ERR(pages[i]);
2886 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2887 bool retry_range_locking)
2890 * Do any pending delalloc/csum calculations on inode, one way or
2891 * another, and lock file content.
2892 * The locking order is:
2895 * 2) range in the inode's io tree
2898 struct btrfs_ordered_extent *ordered;
2899 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2900 ordered = btrfs_lookup_first_ordered_extent(inode,
2903 ordered->file_offset + ordered->len <= off ||
2904 ordered->file_offset >= off + len) &&
2905 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2906 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2908 btrfs_put_ordered_extent(ordered);
2911 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2913 btrfs_put_ordered_extent(ordered);
2914 if (!retry_range_locking)
2916 btrfs_wait_ordered_range(inode, off, len);
2921 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2923 inode_unlock(inode1);
2924 inode_unlock(inode2);
2927 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2929 if (inode1 < inode2)
2930 swap(inode1, inode2);
2932 inode_lock_nested(inode1, I_MUTEX_PARENT);
2933 inode_lock_nested(inode2, I_MUTEX_CHILD);
2936 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2937 struct inode *inode2, u64 loff2, u64 len)
2939 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2940 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2943 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2944 struct inode *inode2, u64 loff2, u64 len,
2945 bool retry_range_locking)
2949 if (inode1 < inode2) {
2950 swap(inode1, inode2);
2953 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2956 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2958 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2965 struct page **src_pages;
2966 struct page **dst_pages;
2969 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2974 for (i = 0; i < cmp->num_pages; i++) {
2975 pg = cmp->src_pages[i];
2980 pg = cmp->dst_pages[i];
2986 kfree(cmp->src_pages);
2987 kfree(cmp->dst_pages);
2990 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2991 struct inode *dst, u64 dst_loff,
2992 u64 len, struct cmp_pages *cmp)
2995 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
2996 struct page **src_pgarr, **dst_pgarr;
2999 * We must gather up all the pages before we initiate our
3000 * extent locking. We use an array for the page pointers. Size
3001 * of the array is bounded by len, which is in turn bounded by
3002 * BTRFS_MAX_DEDUPE_LEN.
3004 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
3005 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
3006 if (!src_pgarr || !dst_pgarr) {
3011 cmp->num_pages = num_pages;
3012 cmp->src_pages = src_pgarr;
3013 cmp->dst_pages = dst_pgarr;
3016 * If deduping ranges in the same inode, locking rules make it mandatory
3017 * to always lock pages in ascending order to avoid deadlocks with
3018 * concurrent tasks (such as starting writeback/delalloc).
3020 if (src == dst && dst_loff < loff) {
3021 swap(src_pgarr, dst_pgarr);
3022 swap(loff, dst_loff);
3025 ret = gather_extent_pages(src, src_pgarr, cmp->num_pages, loff);
3029 ret = gather_extent_pages(dst, dst_pgarr, cmp->num_pages, dst_loff);
3033 btrfs_cmp_data_free(cmp);
3037 static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp)
3041 struct page *src_page, *dst_page;
3042 unsigned int cmp_len = PAGE_SIZE;
3043 void *addr, *dst_addr;
3047 if (len < PAGE_SIZE)
3050 BUG_ON(i >= cmp->num_pages);
3052 src_page = cmp->src_pages[i];
3053 dst_page = cmp->dst_pages[i];
3054 ASSERT(PageLocked(src_page));
3055 ASSERT(PageLocked(dst_page));
3057 addr = kmap_atomic(src_page);
3058 dst_addr = kmap_atomic(dst_page);
3060 flush_dcache_page(src_page);
3061 flush_dcache_page(dst_page);
3063 if (memcmp(addr, dst_addr, cmp_len))
3066 kunmap_atomic(addr);
3067 kunmap_atomic(dst_addr);
3079 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3083 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3085 if (off + olen > inode->i_size || off + olen < off)
3088 /* if we extend to eof, continue to block boundary */
3089 if (off + len == inode->i_size)
3090 *plen = len = ALIGN(inode->i_size, bs) - off;
3092 /* Check that we are block aligned - btrfs_clone() requires this */
3093 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3099 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3100 struct inode *dst, u64 dst_loff)
3104 struct cmp_pages cmp;
3105 bool same_inode = (src == dst);
3106 u64 same_lock_start = 0;
3107 u64 same_lock_len = 0;
3115 btrfs_double_inode_lock(src, dst);
3117 ret = extent_same_check_offsets(src, loff, &len, olen);
3121 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3127 * Single inode case wants the same checks, except we
3128 * don't want our length pushed out past i_size as
3129 * comparing that data range makes no sense.
3131 * extent_same_check_offsets() will do this for an
3132 * unaligned length at i_size, so catch it here and
3133 * reject the request.
3135 * This effectively means we require aligned extents
3136 * for the single-inode case, whereas the other cases
3137 * allow an unaligned length so long as it ends at
3145 /* Check for overlapping ranges */
3146 if (dst_loff + len > loff && dst_loff < loff + len) {
3151 same_lock_start = min_t(u64, loff, dst_loff);
3152 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3155 /* don't make the dst file partly checksummed */
3156 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3157 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3163 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3168 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3171 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3174 * If one of the inodes has dirty pages in the respective range or
3175 * ordered extents, we need to flush dellaloc and wait for all ordered
3176 * extents in the range. We must unlock the pages and the ranges in the
3177 * io trees to avoid deadlocks when flushing delalloc (requires locking
3178 * pages) and when waiting for ordered extents to complete (they require
3181 if (ret == -EAGAIN) {
3183 * Ranges in the io trees already unlocked. Now unlock all
3184 * pages before waiting for all IO to complete.
3186 btrfs_cmp_data_free(&cmp);
3188 btrfs_wait_ordered_range(src, same_lock_start,
3191 btrfs_wait_ordered_range(src, loff, len);
3192 btrfs_wait_ordered_range(dst, dst_loff, len);
3198 /* ranges in the io trees already unlocked */
3199 btrfs_cmp_data_free(&cmp);
3203 /* pass original length for comparison so we stay within i_size */
3204 ret = btrfs_cmp_data(olen, &cmp);
3206 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3209 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3210 same_lock_start + same_lock_len - 1);
3212 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3214 btrfs_cmp_data_free(&cmp);
3219 btrfs_double_inode_unlock(src, dst);
3224 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3226 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3227 struct file *dst_file, u64 dst_loff)
3229 struct inode *src = file_inode(src_file);
3230 struct inode *dst = file_inode(dst_file);
3231 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3234 if (olen > BTRFS_MAX_DEDUPE_LEN)
3235 olen = BTRFS_MAX_DEDUPE_LEN;
3237 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
3239 * Btrfs does not support blocksize < page_size. As a
3240 * result, btrfs_cmp_data() won't correctly handle
3241 * this situation without an update.
3246 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3252 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3253 struct inode *inode,
3259 struct btrfs_root *root = BTRFS_I(inode)->root;
3262 inode_inc_iversion(inode);
3263 if (!no_time_update)
3264 inode->i_mtime = inode->i_ctime = current_time(inode);
3266 * We round up to the block size at eof when determining which
3267 * extents to clone above, but shouldn't round up the file size.
3269 if (endoff > destoff + olen)
3270 endoff = destoff + olen;
3271 if (endoff > inode->i_size)
3272 btrfs_i_size_write(BTRFS_I(inode), endoff);
3274 ret = btrfs_update_inode(trans, root, inode);
3276 btrfs_abort_transaction(trans, ret);
3277 btrfs_end_transaction(trans);
3280 ret = btrfs_end_transaction(trans);
3285 static void clone_update_extent_map(struct btrfs_inode *inode,
3286 const struct btrfs_trans_handle *trans,
3287 const struct btrfs_path *path,
3288 const u64 hole_offset,
3291 struct extent_map_tree *em_tree = &inode->extent_tree;
3292 struct extent_map *em;
3295 em = alloc_extent_map();
3297 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3302 struct btrfs_file_extent_item *fi;
3304 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3305 struct btrfs_file_extent_item);
3306 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3307 em->generation = -1;
3308 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3309 BTRFS_FILE_EXTENT_INLINE)
3310 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3311 &inode->runtime_flags);
3313 em->start = hole_offset;
3315 em->ram_bytes = em->len;
3316 em->orig_start = hole_offset;
3317 em->block_start = EXTENT_MAP_HOLE;
3319 em->orig_block_len = 0;
3320 em->compress_type = BTRFS_COMPRESS_NONE;
3321 em->generation = trans->transid;
3325 write_lock(&em_tree->lock);
3326 ret = add_extent_mapping(em_tree, em, 1);
3327 write_unlock(&em_tree->lock);
3328 if (ret != -EEXIST) {
3329 free_extent_map(em);
3332 btrfs_drop_extent_cache(inode, em->start,
3333 em->start + em->len - 1, 0);
3337 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3341 * Make sure we do not end up inserting an inline extent into a file that has
3342 * already other (non-inline) extents. If a file has an inline extent it can
3343 * not have any other extents and the (single) inline extent must start at the
3344 * file offset 0. Failing to respect these rules will lead to file corruption,
3345 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3347 * We can have extents that have been already written to disk or we can have
3348 * dirty ranges still in delalloc, in which case the extent maps and items are
3349 * created only when we run delalloc, and the delalloc ranges might fall outside
3350 * the range we are currently locking in the inode's io tree. So we check the
3351 * inode's i_size because of that (i_size updates are done while holding the
3352 * i_mutex, which we are holding here).
3353 * We also check to see if the inode has a size not greater than "datal" but has
3354 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3355 * protected against such concurrent fallocate calls by the i_mutex).
3357 * If the file has no extents but a size greater than datal, do not allow the
3358 * copy because we would need turn the inline extent into a non-inline one (even
3359 * with NO_HOLES enabled). If we find our destination inode only has one inline
3360 * extent, just overwrite it with the source inline extent if its size is less
3361 * than the source extent's size, or we could copy the source inline extent's
3362 * data into the destination inode's inline extent if the later is greater then
3365 static int clone_copy_inline_extent(struct inode *dst,
3366 struct btrfs_trans_handle *trans,
3367 struct btrfs_path *path,
3368 struct btrfs_key *new_key,
3369 const u64 drop_start,
3375 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
3376 struct btrfs_root *root = BTRFS_I(dst)->root;
3377 const u64 aligned_end = ALIGN(new_key->offset + datal,
3378 fs_info->sectorsize);
3380 struct btrfs_key key;
3382 if (new_key->offset > 0)
3385 key.objectid = btrfs_ino(BTRFS_I(dst));
3386 key.type = BTRFS_EXTENT_DATA_KEY;
3388 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3391 } else if (ret > 0) {
3392 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3393 ret = btrfs_next_leaf(root, path);
3397 goto copy_inline_extent;
3399 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3400 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3401 key.type == BTRFS_EXTENT_DATA_KEY) {
3402 ASSERT(key.offset > 0);
3405 } else if (i_size_read(dst) <= datal) {
3406 struct btrfs_file_extent_item *ei;
3410 * If the file size is <= datal, make sure there are no other
3411 * extents following (can happen do to an fallocate call with
3412 * the flag FALLOC_FL_KEEP_SIZE).
3414 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3415 struct btrfs_file_extent_item);
3417 * If it's an inline extent, it can not have other extents
3420 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3421 BTRFS_FILE_EXTENT_INLINE)
3422 goto copy_inline_extent;
3424 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3425 if (ext_len > aligned_end)
3428 ret = btrfs_next_item(root, path);
3431 } else if (ret == 0) {
3432 btrfs_item_key_to_cpu(path->nodes[0], &key,
3434 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3435 key.type == BTRFS_EXTENT_DATA_KEY)
3442 * We have no extent items, or we have an extent at offset 0 which may
3443 * or may not be inlined. All these cases are dealt the same way.
3445 if (i_size_read(dst) > datal) {
3447 * If the destination inode has an inline extent...
3448 * This would require copying the data from the source inline
3449 * extent into the beginning of the destination's inline extent.
3450 * But this is really complex, both extents can be compressed
3451 * or just one of them, which would require decompressing and
3452 * re-compressing data (which could increase the new compressed
3453 * size, not allowing the compressed data to fit anymore in an
3455 * So just don't support this case for now (it should be rare,
3456 * we are not really saving space when cloning inline extents).
3461 btrfs_release_path(path);
3462 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3465 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3470 const u32 start = btrfs_file_extent_calc_inline_size(0);
3472 memmove(inline_data + start, inline_data + start + skip, datal);
3475 write_extent_buffer(path->nodes[0], inline_data,
3476 btrfs_item_ptr_offset(path->nodes[0],
3479 inode_add_bytes(dst, datal);
3485 * btrfs_clone() - clone a range from inode file to another
3487 * @src: Inode to clone from
3488 * @inode: Inode to clone to
3489 * @off: Offset within source to start clone from
3490 * @olen: Original length, passed by user, of range to clone
3491 * @olen_aligned: Block-aligned value of olen
3492 * @destoff: Offset within @inode to start clone
3493 * @no_time_update: Whether to update mtime/ctime on the target inode
3495 static int btrfs_clone(struct inode *src, struct inode *inode,
3496 const u64 off, const u64 olen, const u64 olen_aligned,
3497 const u64 destoff, int no_time_update)
3499 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3500 struct btrfs_root *root = BTRFS_I(inode)->root;
3501 struct btrfs_path *path = NULL;
3502 struct extent_buffer *leaf;
3503 struct btrfs_trans_handle *trans;
3505 struct btrfs_key key;
3509 const u64 len = olen_aligned;
3510 u64 last_dest_end = destoff;
3513 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3517 path = btrfs_alloc_path();
3523 path->reada = READA_FORWARD;
3525 key.objectid = btrfs_ino(BTRFS_I(src));
3526 key.type = BTRFS_EXTENT_DATA_KEY;
3530 u64 next_key_min_offset = key.offset + 1;
3533 * note the key will change type as we walk through the
3536 path->leave_spinning = 1;
3537 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3542 * First search, if no extent item that starts at offset off was
3543 * found but the previous item is an extent item, it's possible
3544 * it might overlap our target range, therefore process it.
3546 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3547 btrfs_item_key_to_cpu(path->nodes[0], &key,
3548 path->slots[0] - 1);
3549 if (key.type == BTRFS_EXTENT_DATA_KEY)
3553 nritems = btrfs_header_nritems(path->nodes[0]);
3555 if (path->slots[0] >= nritems) {
3556 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3561 nritems = btrfs_header_nritems(path->nodes[0]);
3563 leaf = path->nodes[0];
3564 slot = path->slots[0];
3566 btrfs_item_key_to_cpu(leaf, &key, slot);
3567 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3568 key.objectid != btrfs_ino(BTRFS_I(src)))
3571 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3572 struct btrfs_file_extent_item *extent;
3575 struct btrfs_key new_key;
3576 u64 disko = 0, diskl = 0;
3577 u64 datao = 0, datal = 0;
3581 extent = btrfs_item_ptr(leaf, slot,
3582 struct btrfs_file_extent_item);
3583 comp = btrfs_file_extent_compression(leaf, extent);
3584 type = btrfs_file_extent_type(leaf, extent);
3585 if (type == BTRFS_FILE_EXTENT_REG ||
3586 type == BTRFS_FILE_EXTENT_PREALLOC) {
3587 disko = btrfs_file_extent_disk_bytenr(leaf,
3589 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3591 datao = btrfs_file_extent_offset(leaf, extent);
3592 datal = btrfs_file_extent_num_bytes(leaf,
3594 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3595 /* take upper bound, may be compressed */
3596 datal = btrfs_file_extent_ram_bytes(leaf,
3601 * The first search might have left us at an extent
3602 * item that ends before our target range's start, can
3603 * happen if we have holes and NO_HOLES feature enabled.
3605 if (key.offset + datal <= off) {
3608 } else if (key.offset >= off + len) {
3611 next_key_min_offset = key.offset + datal;
3612 size = btrfs_item_size_nr(leaf, slot);
3613 read_extent_buffer(leaf, buf,
3614 btrfs_item_ptr_offset(leaf, slot),
3617 btrfs_release_path(path);
3618 path->leave_spinning = 0;
3620 memcpy(&new_key, &key, sizeof(new_key));
3621 new_key.objectid = btrfs_ino(BTRFS_I(inode));
3622 if (off <= key.offset)
3623 new_key.offset = key.offset + destoff - off;
3625 new_key.offset = destoff;
3628 * Deal with a hole that doesn't have an extent item
3629 * that represents it (NO_HOLES feature enabled).
3630 * This hole is either in the middle of the cloning
3631 * range or at the beginning (fully overlaps it or
3632 * partially overlaps it).
3634 if (new_key.offset != last_dest_end)
3635 drop_start = last_dest_end;
3637 drop_start = new_key.offset;
3640 * 1 - adjusting old extent (we may have to split it)
3641 * 1 - add new extent
3644 trans = btrfs_start_transaction(root, 3);
3645 if (IS_ERR(trans)) {
3646 ret = PTR_ERR(trans);
3650 if (type == BTRFS_FILE_EXTENT_REG ||
3651 type == BTRFS_FILE_EXTENT_PREALLOC) {
3653 * a | --- range to clone ---| b
3654 * | ------------- extent ------------- |
3657 /* subtract range b */
3658 if (key.offset + datal > off + len)
3659 datal = off + len - key.offset;
3661 /* subtract range a */
3662 if (off > key.offset) {
3663 datao += off - key.offset;
3664 datal -= off - key.offset;
3667 ret = btrfs_drop_extents(trans, root, inode,
3669 new_key.offset + datal,
3672 if (ret != -EOPNOTSUPP)
3673 btrfs_abort_transaction(trans,
3675 btrfs_end_transaction(trans);
3679 ret = btrfs_insert_empty_item(trans, root, path,
3682 btrfs_abort_transaction(trans, ret);
3683 btrfs_end_transaction(trans);
3687 leaf = path->nodes[0];
3688 slot = path->slots[0];
3689 write_extent_buffer(leaf, buf,
3690 btrfs_item_ptr_offset(leaf, slot),
3693 extent = btrfs_item_ptr(leaf, slot,
3694 struct btrfs_file_extent_item);
3696 /* disko == 0 means it's a hole */
3700 btrfs_set_file_extent_offset(leaf, extent,
3702 btrfs_set_file_extent_num_bytes(leaf, extent,
3706 inode_add_bytes(inode, datal);
3707 ret = btrfs_inc_extent_ref(trans,
3710 root->root_key.objectid,
3711 btrfs_ino(BTRFS_I(inode)),
3712 new_key.offset - datao);
3714 btrfs_abort_transaction(trans,
3716 btrfs_end_transaction(trans);
3721 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3725 if (off > key.offset) {
3726 skip = off - key.offset;
3727 new_key.offset += skip;
3730 if (key.offset + datal > off + len)
3731 trim = key.offset + datal - (off + len);
3733 if (comp && (skip || trim)) {
3735 btrfs_end_transaction(trans);
3738 size -= skip + trim;
3739 datal -= skip + trim;
3741 ret = clone_copy_inline_extent(inode,
3748 if (ret != -EOPNOTSUPP)
3749 btrfs_abort_transaction(trans,
3751 btrfs_end_transaction(trans);
3754 leaf = path->nodes[0];
3755 slot = path->slots[0];
3758 /* If we have an implicit hole (NO_HOLES feature). */
3759 if (drop_start < new_key.offset)
3760 clone_update_extent_map(BTRFS_I(inode), trans,
3762 new_key.offset - drop_start);
3764 clone_update_extent_map(BTRFS_I(inode), trans,
3767 btrfs_mark_buffer_dirty(leaf);
3768 btrfs_release_path(path);
3770 last_dest_end = ALIGN(new_key.offset + datal,
3771 fs_info->sectorsize);
3772 ret = clone_finish_inode_update(trans, inode,
3778 if (new_key.offset + datal >= destoff + len)
3781 btrfs_release_path(path);
3782 key.offset = next_key_min_offset;
3784 if (fatal_signal_pending(current)) {
3791 if (last_dest_end < destoff + len) {
3793 * We have an implicit hole (NO_HOLES feature is enabled) that
3794 * fully or partially overlaps our cloning range at its end.
3796 btrfs_release_path(path);
3799 * 1 - remove extent(s)
3802 trans = btrfs_start_transaction(root, 2);
3803 if (IS_ERR(trans)) {
3804 ret = PTR_ERR(trans);
3807 ret = btrfs_drop_extents(trans, root, inode,
3808 last_dest_end, destoff + len, 1);
3810 if (ret != -EOPNOTSUPP)
3811 btrfs_abort_transaction(trans, ret);
3812 btrfs_end_transaction(trans);
3815 clone_update_extent_map(BTRFS_I(inode), trans, NULL,
3817 destoff + len - last_dest_end);
3818 ret = clone_finish_inode_update(trans, inode, destoff + len,
3819 destoff, olen, no_time_update);
3823 btrfs_free_path(path);
3828 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3829 u64 off, u64 olen, u64 destoff)
3831 struct inode *inode = file_inode(file);
3832 struct inode *src = file_inode(file_src);
3833 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3834 struct btrfs_root *root = BTRFS_I(inode)->root;
3837 u64 bs = fs_info->sb->s_blocksize;
3838 int same_inode = src == inode;
3842 * - split compressed inline extents. annoying: we need to
3843 * decompress into destination's address_space (the file offset
3844 * may change, so source mapping won't do), then recompress (or
3845 * otherwise reinsert) a subrange.
3847 * - split destination inode's inline extents. The inline extents can
3848 * be either compressed or non-compressed.
3851 if (btrfs_root_readonly(root))
3854 if (file_src->f_path.mnt != file->f_path.mnt ||
3855 src->i_sb != inode->i_sb)
3858 /* don't make the dst file partly checksummed */
3859 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3860 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3863 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3867 btrfs_double_inode_lock(src, inode);
3872 /* determine range to clone */
3874 if (off + len > src->i_size || off + len < off)
3877 olen = len = src->i_size - off;
3878 /* if we extend to eof, continue to block boundary */
3879 if (off + len == src->i_size)
3880 len = ALIGN(src->i_size, bs) - off;
3887 /* verify the end result is block aligned */
3888 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3889 !IS_ALIGNED(destoff, bs))
3892 /* verify if ranges are overlapped within the same file */
3894 if (destoff + len > off && destoff < off + len)
3898 if (destoff > inode->i_size) {
3899 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3905 * Lock the target range too. Right after we replace the file extent
3906 * items in the fs tree (which now point to the cloned data), we might
3907 * have a worker replace them with extent items relative to a write
3908 * operation that was issued before this clone operation (i.e. confront
3909 * with inode.c:btrfs_finish_ordered_io).
3912 u64 lock_start = min_t(u64, off, destoff);
3913 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3915 ret = lock_extent_range(src, lock_start, lock_len, true);
3917 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3922 /* ranges in the io trees already unlocked */
3926 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3929 u64 lock_start = min_t(u64, off, destoff);
3930 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3932 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3934 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3937 * Truncate page cache pages so that future reads will see the cloned
3938 * data immediately and not the previous data.
3940 truncate_inode_pages_range(&inode->i_data,
3941 round_down(destoff, PAGE_SIZE),
3942 round_up(destoff + len, PAGE_SIZE) - 1);
3945 btrfs_double_inode_unlock(src, inode);
3951 int btrfs_clone_file_range(struct file *src_file, loff_t off,
3952 struct file *dst_file, loff_t destoff, u64 len)
3954 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
3958 * there are many ways the trans_start and trans_end ioctls can lead
3959 * to deadlocks. They should only be used by applications that
3960 * basically own the machine, and have a very in depth understanding
3961 * of all the possible deadlocks and enospc problems.
3963 static long btrfs_ioctl_trans_start(struct file *file)
3965 struct inode *inode = file_inode(file);
3966 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3967 struct btrfs_root *root = BTRFS_I(inode)->root;
3968 struct btrfs_trans_handle *trans;
3970 static bool warned = false;
3973 if (!capable(CAP_SYS_ADMIN))
3978 "Userspace transaction mechanism is considered "
3979 "deprecated and slated to be removed in 4.17. "
3980 "If you have a valid use case please "
3981 "speak up on the mailing list");
3987 if (file->private_data)
3991 if (btrfs_root_readonly(root))
3994 ret = mnt_want_write_file(file);
3998 atomic_inc(&fs_info->open_ioctl_trans);
4001 trans = btrfs_start_ioctl_transaction(root);
4005 file->private_data = trans;
4009 atomic_dec(&fs_info->open_ioctl_trans);
4010 mnt_drop_write_file(file);
4015 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
4017 struct inode *inode = file_inode(file);
4018 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4019 struct btrfs_root *root = BTRFS_I(inode)->root;
4020 struct btrfs_root *new_root;
4021 struct btrfs_dir_item *di;
4022 struct btrfs_trans_handle *trans;
4023 struct btrfs_path *path;
4024 struct btrfs_key location;
4025 struct btrfs_disk_key disk_key;
4030 if (!capable(CAP_SYS_ADMIN))
4033 ret = mnt_want_write_file(file);
4037 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4043 objectid = BTRFS_FS_TREE_OBJECTID;
4045 location.objectid = objectid;
4046 location.type = BTRFS_ROOT_ITEM_KEY;
4047 location.offset = (u64)-1;
4049 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
4050 if (IS_ERR(new_root)) {
4051 ret = PTR_ERR(new_root);
4055 path = btrfs_alloc_path();
4060 path->leave_spinning = 1;
4062 trans = btrfs_start_transaction(root, 1);
4063 if (IS_ERR(trans)) {
4064 btrfs_free_path(path);
4065 ret = PTR_ERR(trans);
4069 dir_id = btrfs_super_root_dir(fs_info->super_copy);
4070 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
4071 dir_id, "default", 7, 1);
4072 if (IS_ERR_OR_NULL(di)) {
4073 btrfs_free_path(path);
4074 btrfs_end_transaction(trans);
4076 "Umm, you don't have the default diritem, this isn't going to work");
4081 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4082 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4083 btrfs_mark_buffer_dirty(path->nodes[0]);
4084 btrfs_free_path(path);
4086 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
4087 btrfs_end_transaction(trans);
4089 mnt_drop_write_file(file);
4093 void btrfs_get_block_group_info(struct list_head *groups_list,
4094 struct btrfs_ioctl_space_info *space)
4096 struct btrfs_block_group_cache *block_group;
4098 space->total_bytes = 0;
4099 space->used_bytes = 0;
4101 list_for_each_entry(block_group, groups_list, list) {
4102 space->flags = block_group->flags;
4103 space->total_bytes += block_group->key.offset;
4104 space->used_bytes +=
4105 btrfs_block_group_used(&block_group->item);
4109 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
4112 struct btrfs_ioctl_space_args space_args;
4113 struct btrfs_ioctl_space_info space;
4114 struct btrfs_ioctl_space_info *dest;
4115 struct btrfs_ioctl_space_info *dest_orig;
4116 struct btrfs_ioctl_space_info __user *user_dest;
4117 struct btrfs_space_info *info;
4118 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4119 BTRFS_BLOCK_GROUP_SYSTEM,
4120 BTRFS_BLOCK_GROUP_METADATA,
4121 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4128 if (copy_from_user(&space_args,
4129 (struct btrfs_ioctl_space_args __user *)arg,
4130 sizeof(space_args)))
4133 for (i = 0; i < num_types; i++) {
4134 struct btrfs_space_info *tmp;
4138 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4140 if (tmp->flags == types[i]) {
4150 down_read(&info->groups_sem);
4151 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4152 if (!list_empty(&info->block_groups[c]))
4155 up_read(&info->groups_sem);
4159 * Global block reserve, exported as a space_info
4163 /* space_slots == 0 means they are asking for a count */
4164 if (space_args.space_slots == 0) {
4165 space_args.total_spaces = slot_count;
4169 slot_count = min_t(u64, space_args.space_slots, slot_count);
4171 alloc_size = sizeof(*dest) * slot_count;
4173 /* we generally have at most 6 or so space infos, one for each raid
4174 * level. So, a whole page should be more than enough for everyone
4176 if (alloc_size > PAGE_SIZE)
4179 space_args.total_spaces = 0;
4180 dest = kmalloc(alloc_size, GFP_KERNEL);
4185 /* now we have a buffer to copy into */
4186 for (i = 0; i < num_types; i++) {
4187 struct btrfs_space_info *tmp;
4194 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4196 if (tmp->flags == types[i]) {
4205 down_read(&info->groups_sem);
4206 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4207 if (!list_empty(&info->block_groups[c])) {
4208 btrfs_get_block_group_info(
4209 &info->block_groups[c], &space);
4210 memcpy(dest, &space, sizeof(space));
4212 space_args.total_spaces++;
4218 up_read(&info->groups_sem);
4222 * Add global block reserve
4225 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
4227 spin_lock(&block_rsv->lock);
4228 space.total_bytes = block_rsv->size;
4229 space.used_bytes = block_rsv->size - block_rsv->reserved;
4230 spin_unlock(&block_rsv->lock);
4231 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4232 memcpy(dest, &space, sizeof(space));
4233 space_args.total_spaces++;
4236 user_dest = (struct btrfs_ioctl_space_info __user *)
4237 (arg + sizeof(struct btrfs_ioctl_space_args));
4239 if (copy_to_user(user_dest, dest_orig, alloc_size))
4244 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4251 * there are many ways the trans_start and trans_end ioctls can lead
4252 * to deadlocks. They should only be used by applications that
4253 * basically own the machine, and have a very in depth understanding
4254 * of all the possible deadlocks and enospc problems.
4256 long btrfs_ioctl_trans_end(struct file *file)
4258 struct inode *inode = file_inode(file);
4259 struct btrfs_root *root = BTRFS_I(inode)->root;
4260 struct btrfs_trans_handle *trans;
4262 trans = file->private_data;
4265 file->private_data = NULL;
4267 btrfs_end_transaction(trans);
4269 atomic_dec(&root->fs_info->open_ioctl_trans);
4271 mnt_drop_write_file(file);
4275 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4278 struct btrfs_trans_handle *trans;
4282 trans = btrfs_attach_transaction_barrier(root);
4283 if (IS_ERR(trans)) {
4284 if (PTR_ERR(trans) != -ENOENT)
4285 return PTR_ERR(trans);
4287 /* No running transaction, don't bother */
4288 transid = root->fs_info->last_trans_committed;
4291 transid = trans->transid;
4292 ret = btrfs_commit_transaction_async(trans, 0);
4294 btrfs_end_transaction(trans);
4299 if (copy_to_user(argp, &transid, sizeof(transid)))
4304 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
4310 if (copy_from_user(&transid, argp, sizeof(transid)))
4313 transid = 0; /* current trans */
4315 return btrfs_wait_for_commit(fs_info, transid);
4318 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4320 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4321 struct btrfs_ioctl_scrub_args *sa;
4324 if (!capable(CAP_SYS_ADMIN))
4327 sa = memdup_user(arg, sizeof(*sa));
4331 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4332 ret = mnt_want_write_file(file);
4337 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4338 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4341 if (copy_to_user(arg, sa, sizeof(*sa)))
4344 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4345 mnt_drop_write_file(file);
4351 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
4353 if (!capable(CAP_SYS_ADMIN))
4356 return btrfs_scrub_cancel(fs_info);
4359 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
4362 struct btrfs_ioctl_scrub_args *sa;
4365 if (!capable(CAP_SYS_ADMIN))
4368 sa = memdup_user(arg, sizeof(*sa));
4372 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
4374 if (copy_to_user(arg, sa, sizeof(*sa)))
4381 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
4384 struct btrfs_ioctl_get_dev_stats *sa;
4387 sa = memdup_user(arg, sizeof(*sa));
4391 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4396 ret = btrfs_get_dev_stats(fs_info, sa);
4398 if (copy_to_user(arg, sa, sizeof(*sa)))
4405 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4408 struct btrfs_ioctl_dev_replace_args *p;
4411 if (!capable(CAP_SYS_ADMIN))
4414 p = memdup_user(arg, sizeof(*p));
4419 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4420 if (fs_info->sb->s_flags & MS_RDONLY) {
4424 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4425 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4427 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
4428 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4431 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4432 btrfs_dev_replace_status(fs_info, p);
4435 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4436 ret = btrfs_dev_replace_cancel(fs_info, p);
4443 if (copy_to_user(arg, p, sizeof(*p)))
4450 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4456 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4457 struct inode_fs_paths *ipath = NULL;
4458 struct btrfs_path *path;
4460 if (!capable(CAP_DAC_READ_SEARCH))
4463 path = btrfs_alloc_path();
4469 ipa = memdup_user(arg, sizeof(*ipa));
4476 size = min_t(u32, ipa->size, 4096);
4477 ipath = init_ipath(size, root, path);
4478 if (IS_ERR(ipath)) {
4479 ret = PTR_ERR(ipath);
4484 ret = paths_from_inode(ipa->inum, ipath);
4488 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4489 rel_ptr = ipath->fspath->val[i] -
4490 (u64)(unsigned long)ipath->fspath->val;
4491 ipath->fspath->val[i] = rel_ptr;
4494 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4495 (void *)(unsigned long)ipath->fspath, size);
4502 btrfs_free_path(path);
4509 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4511 struct btrfs_data_container *inodes = ctx;
4512 const size_t c = 3 * sizeof(u64);
4514 if (inodes->bytes_left >= c) {
4515 inodes->bytes_left -= c;
4516 inodes->val[inodes->elem_cnt] = inum;
4517 inodes->val[inodes->elem_cnt + 1] = offset;
4518 inodes->val[inodes->elem_cnt + 2] = root;
4519 inodes->elem_cnt += 3;
4521 inodes->bytes_missing += c - inodes->bytes_left;
4522 inodes->bytes_left = 0;
4523 inodes->elem_missed += 3;
4529 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
4534 struct btrfs_ioctl_logical_ino_args *loi;
4535 struct btrfs_data_container *inodes = NULL;
4536 struct btrfs_path *path = NULL;
4538 if (!capable(CAP_SYS_ADMIN))
4541 loi = memdup_user(arg, sizeof(*loi));
4543 return PTR_ERR(loi);
4545 path = btrfs_alloc_path();
4551 size = min_t(u32, loi->size, SZ_64K);
4552 inodes = init_data_container(size);
4553 if (IS_ERR(inodes)) {
4554 ret = PTR_ERR(inodes);
4559 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
4560 build_ino_list, inodes);
4566 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4567 (void *)(unsigned long)inodes, size);
4572 btrfs_free_path(path);
4579 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4580 struct btrfs_ioctl_balance_args *bargs)
4582 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4584 bargs->flags = bctl->flags;
4586 if (atomic_read(&fs_info->balance_running))
4587 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4588 if (atomic_read(&fs_info->balance_pause_req))
4589 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4590 if (atomic_read(&fs_info->balance_cancel_req))
4591 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4593 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4594 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4595 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4598 spin_lock(&fs_info->balance_lock);
4599 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4600 spin_unlock(&fs_info->balance_lock);
4602 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4606 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4608 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4609 struct btrfs_fs_info *fs_info = root->fs_info;
4610 struct btrfs_ioctl_balance_args *bargs;
4611 struct btrfs_balance_control *bctl;
4612 bool need_unlock; /* for mut. excl. ops lock */
4615 if (!capable(CAP_SYS_ADMIN))
4618 ret = mnt_want_write_file(file);
4623 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4624 mutex_lock(&fs_info->volume_mutex);
4625 mutex_lock(&fs_info->balance_mutex);
4631 * mut. excl. ops lock is locked. Three possibilities:
4632 * (1) some other op is running
4633 * (2) balance is running
4634 * (3) balance is paused -- special case (think resume)
4636 mutex_lock(&fs_info->balance_mutex);
4637 if (fs_info->balance_ctl) {
4638 /* this is either (2) or (3) */
4639 if (!atomic_read(&fs_info->balance_running)) {
4640 mutex_unlock(&fs_info->balance_mutex);
4641 if (!mutex_trylock(&fs_info->volume_mutex))
4643 mutex_lock(&fs_info->balance_mutex);
4645 if (fs_info->balance_ctl &&
4646 !atomic_read(&fs_info->balance_running)) {
4648 need_unlock = false;
4652 mutex_unlock(&fs_info->balance_mutex);
4653 mutex_unlock(&fs_info->volume_mutex);
4657 mutex_unlock(&fs_info->balance_mutex);
4663 mutex_unlock(&fs_info->balance_mutex);
4664 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4669 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
4672 bargs = memdup_user(arg, sizeof(*bargs));
4673 if (IS_ERR(bargs)) {
4674 ret = PTR_ERR(bargs);
4678 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4679 if (!fs_info->balance_ctl) {
4684 bctl = fs_info->balance_ctl;
4685 spin_lock(&fs_info->balance_lock);
4686 bctl->flags |= BTRFS_BALANCE_RESUME;
4687 spin_unlock(&fs_info->balance_lock);
4695 if (fs_info->balance_ctl) {
4700 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4706 bctl->fs_info = fs_info;
4708 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4709 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4710 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4712 bctl->flags = bargs->flags;
4714 /* balance everything - no filters */
4715 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4718 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4725 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP
4726 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4727 * or, if restriper was paused all the way until unmount, in
4728 * free_fs_info. The flag is cleared in __cancel_balance.
4730 need_unlock = false;
4732 ret = btrfs_balance(bctl, bargs);
4736 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4745 mutex_unlock(&fs_info->balance_mutex);
4746 mutex_unlock(&fs_info->volume_mutex);
4748 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4750 mnt_drop_write_file(file);
4754 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4756 if (!capable(CAP_SYS_ADMIN))
4760 case BTRFS_BALANCE_CTL_PAUSE:
4761 return btrfs_pause_balance(fs_info);
4762 case BTRFS_BALANCE_CTL_CANCEL:
4763 return btrfs_cancel_balance(fs_info);
4769 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4772 struct btrfs_ioctl_balance_args *bargs;
4775 if (!capable(CAP_SYS_ADMIN))
4778 mutex_lock(&fs_info->balance_mutex);
4779 if (!fs_info->balance_ctl) {
4784 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4790 update_ioctl_balance_args(fs_info, 1, bargs);
4792 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4797 mutex_unlock(&fs_info->balance_mutex);
4801 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4803 struct inode *inode = file_inode(file);
4804 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4805 struct btrfs_ioctl_quota_ctl_args *sa;
4806 struct btrfs_trans_handle *trans = NULL;
4810 if (!capable(CAP_SYS_ADMIN))
4813 ret = mnt_want_write_file(file);
4817 sa = memdup_user(arg, sizeof(*sa));
4823 down_write(&fs_info->subvol_sem);
4824 trans = btrfs_start_transaction(fs_info->tree_root, 2);
4825 if (IS_ERR(trans)) {
4826 ret = PTR_ERR(trans);
4831 case BTRFS_QUOTA_CTL_ENABLE:
4832 ret = btrfs_quota_enable(trans, fs_info);
4834 case BTRFS_QUOTA_CTL_DISABLE:
4835 ret = btrfs_quota_disable(trans, fs_info);
4842 err = btrfs_commit_transaction(trans);
4847 up_write(&fs_info->subvol_sem);
4849 mnt_drop_write_file(file);
4853 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4855 struct inode *inode = file_inode(file);
4856 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4857 struct btrfs_root *root = BTRFS_I(inode)->root;
4858 struct btrfs_ioctl_qgroup_assign_args *sa;
4859 struct btrfs_trans_handle *trans;
4863 if (!capable(CAP_SYS_ADMIN))
4866 ret = mnt_want_write_file(file);
4870 sa = memdup_user(arg, sizeof(*sa));
4876 trans = btrfs_join_transaction(root);
4877 if (IS_ERR(trans)) {
4878 ret = PTR_ERR(trans);
4883 ret = btrfs_add_qgroup_relation(trans, fs_info,
4886 ret = btrfs_del_qgroup_relation(trans, fs_info,
4890 /* update qgroup status and info */
4891 err = btrfs_run_qgroups(trans, fs_info);
4893 btrfs_handle_fs_error(fs_info, err,
4894 "failed to update qgroup status and info");
4895 err = btrfs_end_transaction(trans);
4902 mnt_drop_write_file(file);
4906 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4908 struct inode *inode = file_inode(file);
4909 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4910 struct btrfs_root *root = BTRFS_I(inode)->root;
4911 struct btrfs_ioctl_qgroup_create_args *sa;
4912 struct btrfs_trans_handle *trans;
4916 if (!capable(CAP_SYS_ADMIN))
4919 ret = mnt_want_write_file(file);
4923 sa = memdup_user(arg, sizeof(*sa));
4929 if (!sa->qgroupid) {
4934 trans = btrfs_join_transaction(root);
4935 if (IS_ERR(trans)) {
4936 ret = PTR_ERR(trans);
4941 ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid);
4943 ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid);
4946 err = btrfs_end_transaction(trans);
4953 mnt_drop_write_file(file);
4957 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4959 struct inode *inode = file_inode(file);
4960 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4961 struct btrfs_root *root = BTRFS_I(inode)->root;
4962 struct btrfs_ioctl_qgroup_limit_args *sa;
4963 struct btrfs_trans_handle *trans;
4968 if (!capable(CAP_SYS_ADMIN))
4971 ret = mnt_want_write_file(file);
4975 sa = memdup_user(arg, sizeof(*sa));
4981 trans = btrfs_join_transaction(root);
4982 if (IS_ERR(trans)) {
4983 ret = PTR_ERR(trans);
4987 qgroupid = sa->qgroupid;
4989 /* take the current subvol as qgroup */
4990 qgroupid = root->root_key.objectid;
4993 ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim);
4995 err = btrfs_end_transaction(trans);
5002 mnt_drop_write_file(file);
5006 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
5008 struct inode *inode = file_inode(file);
5009 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5010 struct btrfs_ioctl_quota_rescan_args *qsa;
5013 if (!capable(CAP_SYS_ADMIN))
5016 ret = mnt_want_write_file(file);
5020 qsa = memdup_user(arg, sizeof(*qsa));
5031 ret = btrfs_qgroup_rescan(fs_info);
5036 mnt_drop_write_file(file);
5040 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5042 struct inode *inode = file_inode(file);
5043 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5044 struct btrfs_ioctl_quota_rescan_args *qsa;
5047 if (!capable(CAP_SYS_ADMIN))
5050 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
5054 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5056 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
5059 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5066 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5068 struct inode *inode = file_inode(file);
5069 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5071 if (!capable(CAP_SYS_ADMIN))
5074 return btrfs_qgroup_wait_for_completion(fs_info, true);
5077 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5078 struct btrfs_ioctl_received_subvol_args *sa)
5080 struct inode *inode = file_inode(file);
5081 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5082 struct btrfs_root *root = BTRFS_I(inode)->root;
5083 struct btrfs_root_item *root_item = &root->root_item;
5084 struct btrfs_trans_handle *trans;
5085 struct timespec ct = current_time(inode);
5087 int received_uuid_changed;
5089 if (!inode_owner_or_capable(inode))
5092 ret = mnt_want_write_file(file);
5096 down_write(&fs_info->subvol_sem);
5098 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
5103 if (btrfs_root_readonly(root)) {
5110 * 2 - uuid items (received uuid + subvol uuid)
5112 trans = btrfs_start_transaction(root, 3);
5113 if (IS_ERR(trans)) {
5114 ret = PTR_ERR(trans);
5119 sa->rtransid = trans->transid;
5120 sa->rtime.sec = ct.tv_sec;
5121 sa->rtime.nsec = ct.tv_nsec;
5123 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5125 if (received_uuid_changed &&
5126 !btrfs_is_empty_uuid(root_item->received_uuid))
5127 btrfs_uuid_tree_rem(trans, fs_info, root_item->received_uuid,
5128 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5129 root->root_key.objectid);
5130 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5131 btrfs_set_root_stransid(root_item, sa->stransid);
5132 btrfs_set_root_rtransid(root_item, sa->rtransid);
5133 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5134 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5135 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5136 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5138 ret = btrfs_update_root(trans, fs_info->tree_root,
5139 &root->root_key, &root->root_item);
5141 btrfs_end_transaction(trans);
5144 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5145 ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid,
5146 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5147 root->root_key.objectid);
5148 if (ret < 0 && ret != -EEXIST) {
5149 btrfs_abort_transaction(trans, ret);
5153 ret = btrfs_commit_transaction(trans);
5155 btrfs_abort_transaction(trans, ret);
5160 up_write(&fs_info->subvol_sem);
5161 mnt_drop_write_file(file);
5166 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5169 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5170 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5173 args32 = memdup_user(arg, sizeof(*args32));
5175 return PTR_ERR(args32);
5177 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5183 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5184 args64->stransid = args32->stransid;
5185 args64->rtransid = args32->rtransid;
5186 args64->stime.sec = args32->stime.sec;
5187 args64->stime.nsec = args32->stime.nsec;
5188 args64->rtime.sec = args32->rtime.sec;
5189 args64->rtime.nsec = args32->rtime.nsec;
5190 args64->flags = args32->flags;
5192 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5196 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5197 args32->stransid = args64->stransid;
5198 args32->rtransid = args64->rtransid;
5199 args32->stime.sec = args64->stime.sec;
5200 args32->stime.nsec = args64->stime.nsec;
5201 args32->rtime.sec = args64->rtime.sec;
5202 args32->rtime.nsec = args64->rtime.nsec;
5203 args32->flags = args64->flags;
5205 ret = copy_to_user(arg, args32, sizeof(*args32));
5216 static long btrfs_ioctl_set_received_subvol(struct file *file,
5219 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5222 sa = memdup_user(arg, sizeof(*sa));
5226 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5231 ret = copy_to_user(arg, sa, sizeof(*sa));
5240 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5242 struct inode *inode = file_inode(file);
5243 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5246 char label[BTRFS_LABEL_SIZE];
5248 spin_lock(&fs_info->super_lock);
5249 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5250 spin_unlock(&fs_info->super_lock);
5252 len = strnlen(label, BTRFS_LABEL_SIZE);
5254 if (len == BTRFS_LABEL_SIZE) {
5256 "label is too long, return the first %zu bytes",
5260 ret = copy_to_user(arg, label, len);
5262 return ret ? -EFAULT : 0;
5265 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5267 struct inode *inode = file_inode(file);
5268 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5269 struct btrfs_root *root = BTRFS_I(inode)->root;
5270 struct btrfs_super_block *super_block = fs_info->super_copy;
5271 struct btrfs_trans_handle *trans;
5272 char label[BTRFS_LABEL_SIZE];
5275 if (!capable(CAP_SYS_ADMIN))
5278 if (copy_from_user(label, arg, sizeof(label)))
5281 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5283 "unable to set label with more than %d bytes",
5284 BTRFS_LABEL_SIZE - 1);
5288 ret = mnt_want_write_file(file);
5292 trans = btrfs_start_transaction(root, 0);
5293 if (IS_ERR(trans)) {
5294 ret = PTR_ERR(trans);
5298 spin_lock(&fs_info->super_lock);
5299 strcpy(super_block->label, label);
5300 spin_unlock(&fs_info->super_lock);
5301 ret = btrfs_commit_transaction(trans);
5304 mnt_drop_write_file(file);
5308 #define INIT_FEATURE_FLAGS(suffix) \
5309 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5310 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5311 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5313 int btrfs_ioctl_get_supported_features(void __user *arg)
5315 static const struct btrfs_ioctl_feature_flags features[3] = {
5316 INIT_FEATURE_FLAGS(SUPP),
5317 INIT_FEATURE_FLAGS(SAFE_SET),
5318 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5321 if (copy_to_user(arg, &features, sizeof(features)))
5327 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5329 struct inode *inode = file_inode(file);
5330 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5331 struct btrfs_super_block *super_block = fs_info->super_copy;
5332 struct btrfs_ioctl_feature_flags features;
5334 features.compat_flags = btrfs_super_compat_flags(super_block);
5335 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5336 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5338 if (copy_to_user(arg, &features, sizeof(features)))
5344 static int check_feature_bits(struct btrfs_fs_info *fs_info,
5345 enum btrfs_feature_set set,
5346 u64 change_mask, u64 flags, u64 supported_flags,
5347 u64 safe_set, u64 safe_clear)
5349 const char *type = btrfs_feature_set_names[set];
5351 u64 disallowed, unsupported;
5352 u64 set_mask = flags & change_mask;
5353 u64 clear_mask = ~flags & change_mask;
5355 unsupported = set_mask & ~supported_flags;
5357 names = btrfs_printable_features(set, unsupported);
5360 "this kernel does not support the %s feature bit%s",
5361 names, strchr(names, ',') ? "s" : "");
5365 "this kernel does not support %s bits 0x%llx",
5370 disallowed = set_mask & ~safe_set;
5372 names = btrfs_printable_features(set, disallowed);
5375 "can't set the %s feature bit%s while mounted",
5376 names, strchr(names, ',') ? "s" : "");
5380 "can't set %s bits 0x%llx while mounted",
5385 disallowed = clear_mask & ~safe_clear;
5387 names = btrfs_printable_features(set, disallowed);
5390 "can't clear the %s feature bit%s while mounted",
5391 names, strchr(names, ',') ? "s" : "");
5395 "can't clear %s bits 0x%llx while mounted",
5403 #define check_feature(fs_info, change_mask, flags, mask_base) \
5404 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5405 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5406 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5407 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5409 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5411 struct inode *inode = file_inode(file);
5412 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5413 struct btrfs_root *root = BTRFS_I(inode)->root;
5414 struct btrfs_super_block *super_block = fs_info->super_copy;
5415 struct btrfs_ioctl_feature_flags flags[2];
5416 struct btrfs_trans_handle *trans;
5420 if (!capable(CAP_SYS_ADMIN))
5423 if (copy_from_user(flags, arg, sizeof(flags)))
5427 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5428 !flags[0].incompat_flags)
5431 ret = check_feature(fs_info, flags[0].compat_flags,
5432 flags[1].compat_flags, COMPAT);
5436 ret = check_feature(fs_info, flags[0].compat_ro_flags,
5437 flags[1].compat_ro_flags, COMPAT_RO);
5441 ret = check_feature(fs_info, flags[0].incompat_flags,
5442 flags[1].incompat_flags, INCOMPAT);
5446 ret = mnt_want_write_file(file);
5450 trans = btrfs_start_transaction(root, 0);
5451 if (IS_ERR(trans)) {
5452 ret = PTR_ERR(trans);
5453 goto out_drop_write;
5456 spin_lock(&fs_info->super_lock);
5457 newflags = btrfs_super_compat_flags(super_block);
5458 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5459 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5460 btrfs_set_super_compat_flags(super_block, newflags);
5462 newflags = btrfs_super_compat_ro_flags(super_block);
5463 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5464 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5465 btrfs_set_super_compat_ro_flags(super_block, newflags);
5467 newflags = btrfs_super_incompat_flags(super_block);
5468 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5469 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5470 btrfs_set_super_incompat_flags(super_block, newflags);
5471 spin_unlock(&fs_info->super_lock);
5473 ret = btrfs_commit_transaction(trans);
5475 mnt_drop_write_file(file);
5480 long btrfs_ioctl(struct file *file, unsigned int
5481 cmd, unsigned long arg)
5483 struct inode *inode = file_inode(file);
5484 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5485 struct btrfs_root *root = BTRFS_I(inode)->root;
5486 void __user *argp = (void __user *)arg;
5489 case FS_IOC_GETFLAGS:
5490 return btrfs_ioctl_getflags(file, argp);
5491 case FS_IOC_SETFLAGS:
5492 return btrfs_ioctl_setflags(file, argp);
5493 case FS_IOC_GETVERSION:
5494 return btrfs_ioctl_getversion(file, argp);
5496 return btrfs_ioctl_fitrim(file, argp);
5497 case BTRFS_IOC_SNAP_CREATE:
5498 return btrfs_ioctl_snap_create(file, argp, 0);
5499 case BTRFS_IOC_SNAP_CREATE_V2:
5500 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5501 case BTRFS_IOC_SUBVOL_CREATE:
5502 return btrfs_ioctl_snap_create(file, argp, 1);
5503 case BTRFS_IOC_SUBVOL_CREATE_V2:
5504 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5505 case BTRFS_IOC_SNAP_DESTROY:
5506 return btrfs_ioctl_snap_destroy(file, argp);
5507 case BTRFS_IOC_SUBVOL_GETFLAGS:
5508 return btrfs_ioctl_subvol_getflags(file, argp);
5509 case BTRFS_IOC_SUBVOL_SETFLAGS:
5510 return btrfs_ioctl_subvol_setflags(file, argp);
5511 case BTRFS_IOC_DEFAULT_SUBVOL:
5512 return btrfs_ioctl_default_subvol(file, argp);
5513 case BTRFS_IOC_DEFRAG:
5514 return btrfs_ioctl_defrag(file, NULL);
5515 case BTRFS_IOC_DEFRAG_RANGE:
5516 return btrfs_ioctl_defrag(file, argp);
5517 case BTRFS_IOC_RESIZE:
5518 return btrfs_ioctl_resize(file, argp);
5519 case BTRFS_IOC_ADD_DEV:
5520 return btrfs_ioctl_add_dev(fs_info, argp);
5521 case BTRFS_IOC_RM_DEV:
5522 return btrfs_ioctl_rm_dev(file, argp);
5523 case BTRFS_IOC_RM_DEV_V2:
5524 return btrfs_ioctl_rm_dev_v2(file, argp);
5525 case BTRFS_IOC_FS_INFO:
5526 return btrfs_ioctl_fs_info(fs_info, argp);
5527 case BTRFS_IOC_DEV_INFO:
5528 return btrfs_ioctl_dev_info(fs_info, argp);
5529 case BTRFS_IOC_BALANCE:
5530 return btrfs_ioctl_balance(file, NULL);
5531 case BTRFS_IOC_TRANS_START:
5532 return btrfs_ioctl_trans_start(file);
5533 case BTRFS_IOC_TRANS_END:
5534 return btrfs_ioctl_trans_end(file);
5535 case BTRFS_IOC_TREE_SEARCH:
5536 return btrfs_ioctl_tree_search(file, argp);
5537 case BTRFS_IOC_TREE_SEARCH_V2:
5538 return btrfs_ioctl_tree_search_v2(file, argp);
5539 case BTRFS_IOC_INO_LOOKUP:
5540 return btrfs_ioctl_ino_lookup(file, argp);
5541 case BTRFS_IOC_INO_PATHS:
5542 return btrfs_ioctl_ino_to_path(root, argp);
5543 case BTRFS_IOC_LOGICAL_INO:
5544 return btrfs_ioctl_logical_to_ino(fs_info, argp);
5545 case BTRFS_IOC_SPACE_INFO:
5546 return btrfs_ioctl_space_info(fs_info, argp);
5547 case BTRFS_IOC_SYNC: {
5550 ret = btrfs_start_delalloc_roots(fs_info, 0, -1);
5553 ret = btrfs_sync_fs(inode->i_sb, 1);
5555 * The transaction thread may want to do more work,
5556 * namely it pokes the cleaner kthread that will start
5557 * processing uncleaned subvols.
5559 wake_up_process(fs_info->transaction_kthread);
5562 case BTRFS_IOC_START_SYNC:
5563 return btrfs_ioctl_start_sync(root, argp);
5564 case BTRFS_IOC_WAIT_SYNC:
5565 return btrfs_ioctl_wait_sync(fs_info, argp);
5566 case BTRFS_IOC_SCRUB:
5567 return btrfs_ioctl_scrub(file, argp);
5568 case BTRFS_IOC_SCRUB_CANCEL:
5569 return btrfs_ioctl_scrub_cancel(fs_info);
5570 case BTRFS_IOC_SCRUB_PROGRESS:
5571 return btrfs_ioctl_scrub_progress(fs_info, argp);
5572 case BTRFS_IOC_BALANCE_V2:
5573 return btrfs_ioctl_balance(file, argp);
5574 case BTRFS_IOC_BALANCE_CTL:
5575 return btrfs_ioctl_balance_ctl(fs_info, arg);
5576 case BTRFS_IOC_BALANCE_PROGRESS:
5577 return btrfs_ioctl_balance_progress(fs_info, argp);
5578 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5579 return btrfs_ioctl_set_received_subvol(file, argp);
5581 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5582 return btrfs_ioctl_set_received_subvol_32(file, argp);
5584 case BTRFS_IOC_SEND:
5585 return btrfs_ioctl_send(file, argp);
5586 case BTRFS_IOC_GET_DEV_STATS:
5587 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5588 case BTRFS_IOC_QUOTA_CTL:
5589 return btrfs_ioctl_quota_ctl(file, argp);
5590 case BTRFS_IOC_QGROUP_ASSIGN:
5591 return btrfs_ioctl_qgroup_assign(file, argp);
5592 case BTRFS_IOC_QGROUP_CREATE:
5593 return btrfs_ioctl_qgroup_create(file, argp);
5594 case BTRFS_IOC_QGROUP_LIMIT:
5595 return btrfs_ioctl_qgroup_limit(file, argp);
5596 case BTRFS_IOC_QUOTA_RESCAN:
5597 return btrfs_ioctl_quota_rescan(file, argp);
5598 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5599 return btrfs_ioctl_quota_rescan_status(file, argp);
5600 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5601 return btrfs_ioctl_quota_rescan_wait(file, argp);
5602 case BTRFS_IOC_DEV_REPLACE:
5603 return btrfs_ioctl_dev_replace(fs_info, argp);
5604 case BTRFS_IOC_GET_FSLABEL:
5605 return btrfs_ioctl_get_fslabel(file, argp);
5606 case BTRFS_IOC_SET_FSLABEL:
5607 return btrfs_ioctl_set_fslabel(file, argp);
5608 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5609 return btrfs_ioctl_get_supported_features(argp);
5610 case BTRFS_IOC_GET_FEATURES:
5611 return btrfs_ioctl_get_features(file, argp);
5612 case BTRFS_IOC_SET_FEATURES:
5613 return btrfs_ioctl_set_features(file, argp);
5619 #ifdef CONFIG_COMPAT
5620 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5623 * These all access 32-bit values anyway so no further
5624 * handling is necessary.
5627 case FS_IOC32_GETFLAGS:
5628 cmd = FS_IOC_GETFLAGS;
5630 case FS_IOC32_SETFLAGS:
5631 cmd = FS_IOC_SETFLAGS;
5633 case FS_IOC32_GETVERSION:
5634 cmd = FS_IOC_GETVERSION;
5638 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));