1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2007 Oracle. All rights reserved.
6 #include <linux/kernel.h>
8 #include <linux/buffer_head.h>
9 #include <linux/file.h>
11 #include <linux/fsnotify.h>
12 #include <linux/pagemap.h>
13 #include <linux/highmem.h>
14 #include <linux/time.h>
15 #include <linux/init.h>
16 #include <linux/string.h>
17 #include <linux/backing-dev.h>
18 #include <linux/mount.h>
19 #include <linux/mpage.h>
20 #include <linux/namei.h>
21 #include <linux/swap.h>
22 #include <linux/writeback.h>
23 #include <linux/compat.h>
24 #include <linux/bit_spinlock.h>
25 #include <linux/security.h>
26 #include <linux/xattr.h>
28 #include <linux/slab.h>
29 #include <linux/blkdev.h>
30 #include <linux/uuid.h>
31 #include <linux/btrfs.h>
32 #include <linux/uaccess.h>
33 #include <linux/iversion.h>
36 #include "transaction.h"
37 #include "btrfs_inode.h"
38 #include "print-tree.h"
41 #include "inode-map.h"
43 #include "rcu-string.h"
45 #include "dev-replace.h"
50 #include "compression.h"
53 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
54 * structures are incorrect, as the timespec structure from userspace
55 * is 4 bytes too small. We define these alternatives here to teach
56 * the kernel about the 32-bit struct packing.
58 struct btrfs_ioctl_timespec_32 {
61 } __attribute__ ((__packed__));
63 struct btrfs_ioctl_received_subvol_args_32 {
64 char uuid[BTRFS_UUID_SIZE]; /* in */
65 __u64 stransid; /* in */
66 __u64 rtransid; /* out */
67 struct btrfs_ioctl_timespec_32 stime; /* in */
68 struct btrfs_ioctl_timespec_32 rtime; /* out */
70 __u64 reserved[16]; /* in */
71 } __attribute__ ((__packed__));
73 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
74 struct btrfs_ioctl_received_subvol_args_32)
77 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
78 struct btrfs_ioctl_send_args_32 {
79 __s64 send_fd; /* in */
80 __u64 clone_sources_count; /* in */
81 compat_uptr_t clone_sources; /* in */
82 __u64 parent_root; /* in */
84 __u64 reserved[4]; /* in */
85 } __attribute__ ((__packed__));
87 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
88 struct btrfs_ioctl_send_args_32)
91 static int btrfs_clone(struct inode *src, struct inode *inode,
92 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
95 /* Mask out flags that are inappropriate for the given type of inode. */
96 static unsigned int btrfs_mask_flags(umode_t mode, unsigned int flags)
100 else if (S_ISREG(mode))
101 return flags & ~FS_DIRSYNC_FL;
103 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
107 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
109 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
111 unsigned int iflags = 0;
113 if (flags & BTRFS_INODE_SYNC)
114 iflags |= FS_SYNC_FL;
115 if (flags & BTRFS_INODE_IMMUTABLE)
116 iflags |= FS_IMMUTABLE_FL;
117 if (flags & BTRFS_INODE_APPEND)
118 iflags |= FS_APPEND_FL;
119 if (flags & BTRFS_INODE_NODUMP)
120 iflags |= FS_NODUMP_FL;
121 if (flags & BTRFS_INODE_NOATIME)
122 iflags |= FS_NOATIME_FL;
123 if (flags & BTRFS_INODE_DIRSYNC)
124 iflags |= FS_DIRSYNC_FL;
125 if (flags & BTRFS_INODE_NODATACOW)
126 iflags |= FS_NOCOW_FL;
128 if (flags & BTRFS_INODE_NOCOMPRESS)
129 iflags |= FS_NOCOMP_FL;
130 else if (flags & BTRFS_INODE_COMPRESS)
131 iflags |= FS_COMPR_FL;
137 * Update inode->i_flags based on the btrfs internal flags.
139 void btrfs_update_iflags(struct inode *inode)
141 struct btrfs_inode *ip = BTRFS_I(inode);
142 unsigned int new_fl = 0;
144 if (ip->flags & BTRFS_INODE_SYNC)
146 if (ip->flags & BTRFS_INODE_IMMUTABLE)
147 new_fl |= S_IMMUTABLE;
148 if (ip->flags & BTRFS_INODE_APPEND)
150 if (ip->flags & BTRFS_INODE_NOATIME)
152 if (ip->flags & BTRFS_INODE_DIRSYNC)
155 set_mask_bits(&inode->i_flags,
156 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
160 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
162 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
163 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
165 if (copy_to_user(arg, &flags, sizeof(flags)))
170 static int check_flags(unsigned int flags)
172 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
173 FS_NOATIME_FL | FS_NODUMP_FL | \
174 FS_SYNC_FL | FS_DIRSYNC_FL | \
175 FS_NOCOMP_FL | FS_COMPR_FL |
179 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
185 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
187 struct inode *inode = file_inode(file);
188 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
189 struct btrfs_inode *ip = BTRFS_I(inode);
190 struct btrfs_root *root = ip->root;
191 struct btrfs_trans_handle *trans;
192 unsigned int flags, oldflags;
195 unsigned int i_oldflags;
198 if (!inode_owner_or_capable(inode))
201 if (btrfs_root_readonly(root))
204 if (copy_from_user(&flags, arg, sizeof(flags)))
207 ret = check_flags(flags);
211 ret = mnt_want_write_file(file);
217 ip_oldflags = ip->flags;
218 i_oldflags = inode->i_flags;
219 mode = inode->i_mode;
221 flags = btrfs_mask_flags(inode->i_mode, flags);
222 oldflags = btrfs_flags_to_ioctl(ip->flags);
223 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
224 if (!capable(CAP_LINUX_IMMUTABLE)) {
230 if (flags & FS_SYNC_FL)
231 ip->flags |= BTRFS_INODE_SYNC;
233 ip->flags &= ~BTRFS_INODE_SYNC;
234 if (flags & FS_IMMUTABLE_FL)
235 ip->flags |= BTRFS_INODE_IMMUTABLE;
237 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
238 if (flags & FS_APPEND_FL)
239 ip->flags |= BTRFS_INODE_APPEND;
241 ip->flags &= ~BTRFS_INODE_APPEND;
242 if (flags & FS_NODUMP_FL)
243 ip->flags |= BTRFS_INODE_NODUMP;
245 ip->flags &= ~BTRFS_INODE_NODUMP;
246 if (flags & FS_NOATIME_FL)
247 ip->flags |= BTRFS_INODE_NOATIME;
249 ip->flags &= ~BTRFS_INODE_NOATIME;
250 if (flags & FS_DIRSYNC_FL)
251 ip->flags |= BTRFS_INODE_DIRSYNC;
253 ip->flags &= ~BTRFS_INODE_DIRSYNC;
254 if (flags & FS_NOCOW_FL) {
257 * It's safe to turn csums off here, no extents exist.
258 * Otherwise we want the flag to reflect the real COW
259 * status of the file and will not set it.
261 if (inode->i_size == 0)
262 ip->flags |= BTRFS_INODE_NODATACOW
263 | BTRFS_INODE_NODATASUM;
265 ip->flags |= BTRFS_INODE_NODATACOW;
269 * Revert back under same assumptions as above
272 if (inode->i_size == 0)
273 ip->flags &= ~(BTRFS_INODE_NODATACOW
274 | BTRFS_INODE_NODATASUM);
276 ip->flags &= ~BTRFS_INODE_NODATACOW;
281 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
282 * flag may be changed automatically if compression code won't make
285 if (flags & FS_NOCOMP_FL) {
286 ip->flags &= ~BTRFS_INODE_COMPRESS;
287 ip->flags |= BTRFS_INODE_NOCOMPRESS;
289 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
290 if (ret && ret != -ENODATA)
292 } else if (flags & FS_COMPR_FL) {
295 ip->flags |= BTRFS_INODE_COMPRESS;
296 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
298 comp = btrfs_compress_type2str(fs_info->compress_type);
299 if (!comp || comp[0] == 0)
300 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
302 ret = btrfs_set_prop(inode, "btrfs.compression",
303 comp, strlen(comp), 0);
308 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
309 if (ret && ret != -ENODATA)
311 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
314 trans = btrfs_start_transaction(root, 1);
316 ret = PTR_ERR(trans);
320 btrfs_update_iflags(inode);
321 inode_inc_iversion(inode);
322 inode->i_ctime = current_time(inode);
323 ret = btrfs_update_inode(trans, root, inode);
325 btrfs_end_transaction(trans);
328 ip->flags = ip_oldflags;
329 inode->i_flags = i_oldflags;
334 mnt_drop_write_file(file);
338 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
340 struct inode *inode = file_inode(file);
342 return put_user(inode->i_generation, arg);
345 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
347 struct inode *inode = file_inode(file);
348 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
349 struct btrfs_device *device;
350 struct request_queue *q;
351 struct fstrim_range range;
352 u64 minlen = ULLONG_MAX;
354 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
357 if (!capable(CAP_SYS_ADMIN))
361 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
365 q = bdev_get_queue(device->bdev);
366 if (blk_queue_discard(q)) {
368 minlen = min_t(u64, q->limits.discard_granularity,
376 if (copy_from_user(&range, arg, sizeof(range)))
378 if (range.start > total_bytes ||
379 range.len < fs_info->sb->s_blocksize)
382 range.len = min(range.len, total_bytes - range.start);
383 range.minlen = max(range.minlen, minlen);
384 ret = btrfs_trim_fs(fs_info, &range);
388 if (copy_to_user(arg, &range, sizeof(range)))
394 int btrfs_is_empty_uuid(u8 *uuid)
398 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
405 static noinline int create_subvol(struct inode *dir,
406 struct dentry *dentry,
407 const char *name, int namelen,
409 struct btrfs_qgroup_inherit *inherit)
411 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
412 struct btrfs_trans_handle *trans;
413 struct btrfs_key key;
414 struct btrfs_root_item *root_item;
415 struct btrfs_inode_item *inode_item;
416 struct extent_buffer *leaf;
417 struct btrfs_root *root = BTRFS_I(dir)->root;
418 struct btrfs_root *new_root;
419 struct btrfs_block_rsv block_rsv;
420 struct timespec cur_time = current_time(dir);
425 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
430 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
434 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
439 * Don't create subvolume whose level is not zero. Or qgroup will be
440 * screwed up since it assumes subvolume qgroup's level to be 0.
442 if (btrfs_qgroup_level(objectid)) {
447 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
449 * The same as the snapshot creation, please see the comment
450 * of create_snapshot().
452 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
453 8, &qgroup_reserved, false);
457 trans = btrfs_start_transaction(root, 0);
459 ret = PTR_ERR(trans);
460 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
463 trans->block_rsv = &block_rsv;
464 trans->bytes_reserved = block_rsv.size;
466 ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit);
470 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
476 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
477 btrfs_set_header_bytenr(leaf, leaf->start);
478 btrfs_set_header_generation(leaf, trans->transid);
479 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
480 btrfs_set_header_owner(leaf, objectid);
482 write_extent_buffer_fsid(leaf, fs_info->fsid);
483 write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid);
484 btrfs_mark_buffer_dirty(leaf);
486 inode_item = &root_item->inode;
487 btrfs_set_stack_inode_generation(inode_item, 1);
488 btrfs_set_stack_inode_size(inode_item, 3);
489 btrfs_set_stack_inode_nlink(inode_item, 1);
490 btrfs_set_stack_inode_nbytes(inode_item,
492 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
494 btrfs_set_root_flags(root_item, 0);
495 btrfs_set_root_limit(root_item, 0);
496 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
498 btrfs_set_root_bytenr(root_item, leaf->start);
499 btrfs_set_root_generation(root_item, trans->transid);
500 btrfs_set_root_level(root_item, 0);
501 btrfs_set_root_refs(root_item, 1);
502 btrfs_set_root_used(root_item, leaf->len);
503 btrfs_set_root_last_snapshot(root_item, 0);
505 btrfs_set_root_generation_v2(root_item,
506 btrfs_root_generation(root_item));
507 uuid_le_gen(&new_uuid);
508 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
509 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
510 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
511 root_item->ctime = root_item->otime;
512 btrfs_set_root_ctransid(root_item, trans->transid);
513 btrfs_set_root_otransid(root_item, trans->transid);
515 btrfs_tree_unlock(leaf);
516 free_extent_buffer(leaf);
519 btrfs_set_root_dirid(root_item, new_dirid);
521 key.objectid = objectid;
523 key.type = BTRFS_ROOT_ITEM_KEY;
524 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
529 key.offset = (u64)-1;
530 new_root = btrfs_read_fs_root_no_name(fs_info, &key);
531 if (IS_ERR(new_root)) {
532 ret = PTR_ERR(new_root);
533 btrfs_abort_transaction(trans, ret);
537 btrfs_record_root_in_trans(trans, new_root);
539 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
541 /* We potentially lose an unused inode item here */
542 btrfs_abort_transaction(trans, ret);
546 mutex_lock(&new_root->objectid_mutex);
547 new_root->highest_objectid = new_dirid;
548 mutex_unlock(&new_root->objectid_mutex);
551 * insert the directory item
553 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
555 btrfs_abort_transaction(trans, ret);
559 ret = btrfs_insert_dir_item(trans, root,
560 name, namelen, BTRFS_I(dir), &key,
561 BTRFS_FT_DIR, index);
563 btrfs_abort_transaction(trans, ret);
567 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
568 ret = btrfs_update_inode(trans, root, dir);
571 ret = btrfs_add_root_ref(trans, fs_info,
572 objectid, root->root_key.objectid,
573 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
576 ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid,
577 BTRFS_UUID_KEY_SUBVOL, objectid);
579 btrfs_abort_transaction(trans, ret);
583 trans->block_rsv = NULL;
584 trans->bytes_reserved = 0;
585 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
588 *async_transid = trans->transid;
589 err = btrfs_commit_transaction_async(trans, 1);
591 err = btrfs_commit_transaction(trans);
593 err = btrfs_commit_transaction(trans);
599 inode = btrfs_lookup_dentry(dir, dentry);
601 return PTR_ERR(inode);
602 d_instantiate(dentry, inode);
611 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
612 struct dentry *dentry,
613 u64 *async_transid, bool readonly,
614 struct btrfs_qgroup_inherit *inherit)
616 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
618 struct btrfs_pending_snapshot *pending_snapshot;
619 struct btrfs_trans_handle *trans;
622 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
625 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
626 if (!pending_snapshot)
629 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
631 pending_snapshot->path = btrfs_alloc_path();
632 if (!pending_snapshot->root_item || !pending_snapshot->path) {
637 atomic_inc(&root->will_be_snapshotted);
638 smp_mb__after_atomic();
639 /* wait for no snapshot writes */
640 wait_event(root->subv_writers->wait,
641 percpu_counter_sum(&root->subv_writers->counter) == 0);
643 ret = btrfs_start_delalloc_inodes(root, 0);
647 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
649 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
650 BTRFS_BLOCK_RSV_TEMP);
652 * 1 - parent dir inode
655 * 2 - root ref/backref
656 * 1 - root of snapshot
659 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
660 &pending_snapshot->block_rsv, 8,
661 &pending_snapshot->qgroup_reserved,
666 pending_snapshot->dentry = dentry;
667 pending_snapshot->root = root;
668 pending_snapshot->readonly = readonly;
669 pending_snapshot->dir = dir;
670 pending_snapshot->inherit = inherit;
672 trans = btrfs_start_transaction(root, 0);
674 ret = PTR_ERR(trans);
678 spin_lock(&fs_info->trans_lock);
679 list_add(&pending_snapshot->list,
680 &trans->transaction->pending_snapshots);
681 spin_unlock(&fs_info->trans_lock);
683 *async_transid = trans->transid;
684 ret = btrfs_commit_transaction_async(trans, 1);
686 ret = btrfs_commit_transaction(trans);
688 ret = btrfs_commit_transaction(trans);
693 ret = pending_snapshot->error;
697 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
701 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
703 ret = PTR_ERR(inode);
707 d_instantiate(dentry, inode);
710 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
712 if (atomic_dec_and_test(&root->will_be_snapshotted))
713 wake_up_var(&root->will_be_snapshotted);
715 kfree(pending_snapshot->root_item);
716 btrfs_free_path(pending_snapshot->path);
717 kfree(pending_snapshot);
722 /* copy of may_delete in fs/namei.c()
723 * Check whether we can remove a link victim from directory dir, check
724 * whether the type of victim is right.
725 * 1. We can't do it if dir is read-only (done in permission())
726 * 2. We should have write and exec permissions on dir
727 * 3. We can't remove anything from append-only dir
728 * 4. We can't do anything with immutable dir (done in permission())
729 * 5. If the sticky bit on dir is set we should either
730 * a. be owner of dir, or
731 * b. be owner of victim, or
732 * c. have CAP_FOWNER capability
733 * 6. If the victim is append-only or immutable we can't do anything with
734 * links pointing to it.
735 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
736 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
737 * 9. We can't remove a root or mountpoint.
738 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
739 * nfs_async_unlink().
742 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
746 if (d_really_is_negative(victim))
749 BUG_ON(d_inode(victim->d_parent) != dir);
750 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
752 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
757 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
758 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
761 if (!d_is_dir(victim))
765 } else if (d_is_dir(victim))
769 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
774 /* copy of may_create in fs/namei.c() */
775 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
777 if (d_really_is_positive(child))
781 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
785 * Create a new subvolume below @parent. This is largely modeled after
786 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
787 * inside this filesystem so it's quite a bit simpler.
789 static noinline int btrfs_mksubvol(const struct path *parent,
790 const char *name, int namelen,
791 struct btrfs_root *snap_src,
792 u64 *async_transid, bool readonly,
793 struct btrfs_qgroup_inherit *inherit)
795 struct inode *dir = d_inode(parent->dentry);
796 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
797 struct dentry *dentry;
800 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
804 dentry = lookup_one_len(name, parent->dentry, namelen);
805 error = PTR_ERR(dentry);
809 error = btrfs_may_create(dir, dentry);
814 * even if this name doesn't exist, we may get hash collisions.
815 * check for them now when we can safely fail
817 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
823 down_read(&fs_info->subvol_sem);
825 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
829 error = create_snapshot(snap_src, dir, dentry,
830 async_transid, readonly, inherit);
832 error = create_subvol(dir, dentry, name, namelen,
833 async_transid, inherit);
836 fsnotify_mkdir(dir, dentry);
838 up_read(&fs_info->subvol_sem);
847 * When we're defragging a range, we don't want to kick it off again
848 * if it is really just waiting for delalloc to send it down.
849 * If we find a nice big extent or delalloc range for the bytes in the
850 * file you want to defrag, we return 0 to let you know to skip this
853 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
855 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
856 struct extent_map *em = NULL;
857 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
860 read_lock(&em_tree->lock);
861 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
862 read_unlock(&em_tree->lock);
865 end = extent_map_end(em);
867 if (end - offset > thresh)
870 /* if we already have a nice delalloc here, just stop */
872 end = count_range_bits(io_tree, &offset, offset + thresh,
873 thresh, EXTENT_DELALLOC, 1);
880 * helper function to walk through a file and find extents
881 * newer than a specific transid, and smaller than thresh.
883 * This is used by the defragging code to find new and small
886 static int find_new_extents(struct btrfs_root *root,
887 struct inode *inode, u64 newer_than,
888 u64 *off, u32 thresh)
890 struct btrfs_path *path;
891 struct btrfs_key min_key;
892 struct extent_buffer *leaf;
893 struct btrfs_file_extent_item *extent;
896 u64 ino = btrfs_ino(BTRFS_I(inode));
898 path = btrfs_alloc_path();
902 min_key.objectid = ino;
903 min_key.type = BTRFS_EXTENT_DATA_KEY;
904 min_key.offset = *off;
907 ret = btrfs_search_forward(root, &min_key, path, newer_than);
911 if (min_key.objectid != ino)
913 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
916 leaf = path->nodes[0];
917 extent = btrfs_item_ptr(leaf, path->slots[0],
918 struct btrfs_file_extent_item);
920 type = btrfs_file_extent_type(leaf, extent);
921 if (type == BTRFS_FILE_EXTENT_REG &&
922 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
923 check_defrag_in_cache(inode, min_key.offset, thresh)) {
924 *off = min_key.offset;
925 btrfs_free_path(path);
930 if (path->slots[0] < btrfs_header_nritems(leaf)) {
931 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
935 if (min_key.offset == (u64)-1)
939 btrfs_release_path(path);
942 btrfs_free_path(path);
946 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
948 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
949 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
950 struct extent_map *em;
954 * hopefully we have this extent in the tree already, try without
955 * the full extent lock
957 read_lock(&em_tree->lock);
958 em = lookup_extent_mapping(em_tree, start, len);
959 read_unlock(&em_tree->lock);
962 struct extent_state *cached = NULL;
963 u64 end = start + len - 1;
965 /* get the big lock and read metadata off disk */
966 lock_extent_bits(io_tree, start, end, &cached);
967 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0);
968 unlock_extent_cached(io_tree, start, end, &cached);
977 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
979 struct extent_map *next;
982 /* this is the last extent */
983 if (em->start + em->len >= i_size_read(inode))
986 next = defrag_lookup_extent(inode, em->start + em->len);
987 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
989 else if ((em->block_start + em->block_len == next->block_start) &&
990 (em->block_len > SZ_128K && next->block_len > SZ_128K))
993 free_extent_map(next);
997 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
998 u64 *last_len, u64 *skip, u64 *defrag_end,
1001 struct extent_map *em;
1003 bool next_mergeable = true;
1004 bool prev_mergeable = true;
1007 * make sure that once we start defragging an extent, we keep on
1010 if (start < *defrag_end)
1015 em = defrag_lookup_extent(inode, start);
1019 /* this will cover holes, and inline extents */
1020 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1026 prev_mergeable = false;
1028 next_mergeable = defrag_check_next_extent(inode, em);
1030 * we hit a real extent, if it is big or the next extent is not a
1031 * real extent, don't bother defragging it
1033 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1034 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1038 * last_len ends up being a counter of how many bytes we've defragged.
1039 * every time we choose not to defrag an extent, we reset *last_len
1040 * so that the next tiny extent will force a defrag.
1042 * The end result of this is that tiny extents before a single big
1043 * extent will force at least part of that big extent to be defragged.
1046 *defrag_end = extent_map_end(em);
1049 *skip = extent_map_end(em);
1053 free_extent_map(em);
1058 * it doesn't do much good to defrag one or two pages
1059 * at a time. This pulls in a nice chunk of pages
1060 * to COW and defrag.
1062 * It also makes sure the delalloc code has enough
1063 * dirty data to avoid making new small extents as part
1066 * It's a good idea to start RA on this range
1067 * before calling this.
1069 static int cluster_pages_for_defrag(struct inode *inode,
1070 struct page **pages,
1071 unsigned long start_index,
1072 unsigned long num_pages)
1074 unsigned long file_end;
1075 u64 isize = i_size_read(inode);
1082 struct btrfs_ordered_extent *ordered;
1083 struct extent_state *cached_state = NULL;
1084 struct extent_io_tree *tree;
1085 struct extent_changeset *data_reserved = NULL;
1086 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1088 file_end = (isize - 1) >> PAGE_SHIFT;
1089 if (!isize || start_index > file_end)
1092 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1094 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
1095 start_index << PAGE_SHIFT,
1096 page_cnt << PAGE_SHIFT);
1100 tree = &BTRFS_I(inode)->io_tree;
1102 /* step one, lock all the pages */
1103 for (i = 0; i < page_cnt; i++) {
1106 page = find_or_create_page(inode->i_mapping,
1107 start_index + i, mask);
1111 page_start = page_offset(page);
1112 page_end = page_start + PAGE_SIZE - 1;
1114 lock_extent_bits(tree, page_start, page_end,
1116 ordered = btrfs_lookup_ordered_extent(inode,
1118 unlock_extent_cached(tree, page_start, page_end,
1124 btrfs_start_ordered_extent(inode, ordered, 1);
1125 btrfs_put_ordered_extent(ordered);
1128 * we unlocked the page above, so we need check if
1129 * it was released or not.
1131 if (page->mapping != inode->i_mapping) {
1138 if (!PageUptodate(page)) {
1139 btrfs_readpage(NULL, page);
1141 if (!PageUptodate(page)) {
1149 if (page->mapping != inode->i_mapping) {
1161 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1165 * so now we have a nice long stream of locked
1166 * and up to date pages, lets wait on them
1168 for (i = 0; i < i_done; i++)
1169 wait_on_page_writeback(pages[i]);
1171 page_start = page_offset(pages[0]);
1172 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1174 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1175 page_start, page_end - 1, &cached_state);
1176 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1177 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1178 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1181 if (i_done != page_cnt) {
1182 spin_lock(&BTRFS_I(inode)->lock);
1183 BTRFS_I(inode)->outstanding_extents++;
1184 spin_unlock(&BTRFS_I(inode)->lock);
1185 btrfs_delalloc_release_space(inode, data_reserved,
1186 start_index << PAGE_SHIFT,
1187 (page_cnt - i_done) << PAGE_SHIFT, true);
1191 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1194 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1195 page_start, page_end - 1, &cached_state);
1197 for (i = 0; i < i_done; i++) {
1198 clear_page_dirty_for_io(pages[i]);
1199 ClearPageChecked(pages[i]);
1200 set_page_extent_mapped(pages[i]);
1201 set_page_dirty(pages[i]);
1202 unlock_page(pages[i]);
1205 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1207 extent_changeset_free(data_reserved);
1210 for (i = 0; i < i_done; i++) {
1211 unlock_page(pages[i]);
1214 btrfs_delalloc_release_space(inode, data_reserved,
1215 start_index << PAGE_SHIFT,
1216 page_cnt << PAGE_SHIFT, true);
1217 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1219 extent_changeset_free(data_reserved);
1224 int btrfs_defrag_file(struct inode *inode, struct file *file,
1225 struct btrfs_ioctl_defrag_range_args *range,
1226 u64 newer_than, unsigned long max_to_defrag)
1228 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1229 struct btrfs_root *root = BTRFS_I(inode)->root;
1230 struct file_ra_state *ra = NULL;
1231 unsigned long last_index;
1232 u64 isize = i_size_read(inode);
1236 u64 newer_off = range->start;
1238 unsigned long ra_index = 0;
1240 int defrag_count = 0;
1241 int compress_type = BTRFS_COMPRESS_ZLIB;
1242 u32 extent_thresh = range->extent_thresh;
1243 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1244 unsigned long cluster = max_cluster;
1245 u64 new_align = ~((u64)SZ_128K - 1);
1246 struct page **pages = NULL;
1247 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1252 if (range->start >= isize)
1256 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1258 if (range->compress_type)
1259 compress_type = range->compress_type;
1262 if (extent_thresh == 0)
1263 extent_thresh = SZ_256K;
1266 * If we were not given a file, allocate a readahead context. As
1267 * readahead is just an optimization, defrag will work without it so
1268 * we don't error out.
1271 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1273 file_ra_state_init(ra, inode->i_mapping);
1278 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1284 /* find the last page to defrag */
1285 if (range->start + range->len > range->start) {
1286 last_index = min_t(u64, isize - 1,
1287 range->start + range->len - 1) >> PAGE_SHIFT;
1289 last_index = (isize - 1) >> PAGE_SHIFT;
1293 ret = find_new_extents(root, inode, newer_than,
1294 &newer_off, SZ_64K);
1296 range->start = newer_off;
1298 * we always align our defrag to help keep
1299 * the extents in the file evenly spaced
1301 i = (newer_off & new_align) >> PAGE_SHIFT;
1305 i = range->start >> PAGE_SHIFT;
1308 max_to_defrag = last_index - i + 1;
1311 * make writeback starts from i, so the defrag range can be
1312 * written sequentially.
1314 if (i < inode->i_mapping->writeback_index)
1315 inode->i_mapping->writeback_index = i;
1317 while (i <= last_index && defrag_count < max_to_defrag &&
1318 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1320 * make sure we stop running if someone unmounts
1323 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1326 if (btrfs_defrag_cancelled(fs_info)) {
1327 btrfs_debug(fs_info, "defrag_file cancelled");
1332 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1333 extent_thresh, &last_len, &skip,
1334 &defrag_end, do_compress)){
1337 * the should_defrag function tells us how much to skip
1338 * bump our counter by the suggested amount
1340 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1341 i = max(i + 1, next);
1346 cluster = (PAGE_ALIGN(defrag_end) >>
1348 cluster = min(cluster, max_cluster);
1350 cluster = max_cluster;
1353 if (i + cluster > ra_index) {
1354 ra_index = max(i, ra_index);
1356 page_cache_sync_readahead(inode->i_mapping, ra,
1357 file, ra_index, cluster);
1358 ra_index += cluster;
1363 BTRFS_I(inode)->defrag_compress = compress_type;
1364 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1366 inode_unlock(inode);
1370 defrag_count += ret;
1371 balance_dirty_pages_ratelimited(inode->i_mapping);
1372 inode_unlock(inode);
1375 if (newer_off == (u64)-1)
1381 newer_off = max(newer_off + 1,
1382 (u64)i << PAGE_SHIFT);
1384 ret = find_new_extents(root, inode, newer_than,
1385 &newer_off, SZ_64K);
1387 range->start = newer_off;
1388 i = (newer_off & new_align) >> PAGE_SHIFT;
1395 last_len += ret << PAGE_SHIFT;
1403 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1404 filemap_flush(inode->i_mapping);
1405 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1406 &BTRFS_I(inode)->runtime_flags))
1407 filemap_flush(inode->i_mapping);
1410 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1411 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1412 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1413 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1421 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1422 inode_unlock(inode);
1430 static noinline int btrfs_ioctl_resize(struct file *file,
1433 struct inode *inode = file_inode(file);
1434 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1438 struct btrfs_root *root = BTRFS_I(inode)->root;
1439 struct btrfs_ioctl_vol_args *vol_args;
1440 struct btrfs_trans_handle *trans;
1441 struct btrfs_device *device = NULL;
1444 char *devstr = NULL;
1448 if (!capable(CAP_SYS_ADMIN))
1451 ret = mnt_want_write_file(file);
1455 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
1456 mnt_drop_write_file(file);
1457 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1460 mutex_lock(&fs_info->volume_mutex);
1461 vol_args = memdup_user(arg, sizeof(*vol_args));
1462 if (IS_ERR(vol_args)) {
1463 ret = PTR_ERR(vol_args);
1467 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1469 sizestr = vol_args->name;
1470 devstr = strchr(sizestr, ':');
1472 sizestr = devstr + 1;
1474 devstr = vol_args->name;
1475 ret = kstrtoull(devstr, 10, &devid);
1482 btrfs_info(fs_info, "resizing devid %llu", devid);
1485 device = btrfs_find_device(fs_info, devid, NULL, NULL);
1487 btrfs_info(fs_info, "resizer unable to find device %llu",
1493 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1495 "resizer unable to apply on readonly device %llu",
1501 if (!strcmp(sizestr, "max"))
1502 new_size = device->bdev->bd_inode->i_size;
1504 if (sizestr[0] == '-') {
1507 } else if (sizestr[0] == '+') {
1511 new_size = memparse(sizestr, &retptr);
1512 if (*retptr != '\0' || new_size == 0) {
1518 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1523 old_size = btrfs_device_get_total_bytes(device);
1526 if (new_size > old_size) {
1530 new_size = old_size - new_size;
1531 } else if (mod > 0) {
1532 if (new_size > ULLONG_MAX - old_size) {
1536 new_size = old_size + new_size;
1539 if (new_size < SZ_256M) {
1543 if (new_size > device->bdev->bd_inode->i_size) {
1548 new_size = round_down(new_size, fs_info->sectorsize);
1550 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1551 rcu_str_deref(device->name), new_size);
1553 if (new_size > old_size) {
1554 trans = btrfs_start_transaction(root, 0);
1555 if (IS_ERR(trans)) {
1556 ret = PTR_ERR(trans);
1559 ret = btrfs_grow_device(trans, device, new_size);
1560 btrfs_commit_transaction(trans);
1561 } else if (new_size < old_size) {
1562 ret = btrfs_shrink_device(device, new_size);
1563 } /* equal, nothing need to do */
1568 mutex_unlock(&fs_info->volume_mutex);
1569 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
1570 mnt_drop_write_file(file);
1574 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1575 const char *name, unsigned long fd, int subvol,
1576 u64 *transid, bool readonly,
1577 struct btrfs_qgroup_inherit *inherit)
1582 if (!S_ISDIR(file_inode(file)->i_mode))
1585 ret = mnt_want_write_file(file);
1589 namelen = strlen(name);
1590 if (strchr(name, '/')) {
1592 goto out_drop_write;
1595 if (name[0] == '.' &&
1596 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1598 goto out_drop_write;
1602 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1603 NULL, transid, readonly, inherit);
1605 struct fd src = fdget(fd);
1606 struct inode *src_inode;
1609 goto out_drop_write;
1612 src_inode = file_inode(src.file);
1613 if (src_inode->i_sb != file_inode(file)->i_sb) {
1614 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1615 "Snapshot src from another FS");
1617 } else if (!inode_owner_or_capable(src_inode)) {
1619 * Subvolume creation is not restricted, but snapshots
1620 * are limited to own subvolumes only
1624 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1625 BTRFS_I(src_inode)->root,
1626 transid, readonly, inherit);
1631 mnt_drop_write_file(file);
1636 static noinline int btrfs_ioctl_snap_create(struct file *file,
1637 void __user *arg, int subvol)
1639 struct btrfs_ioctl_vol_args *vol_args;
1642 if (!S_ISDIR(file_inode(file)->i_mode))
1645 vol_args = memdup_user(arg, sizeof(*vol_args));
1646 if (IS_ERR(vol_args))
1647 return PTR_ERR(vol_args);
1648 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1650 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1651 vol_args->fd, subvol,
1658 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1659 void __user *arg, int subvol)
1661 struct btrfs_ioctl_vol_args_v2 *vol_args;
1665 bool readonly = false;
1666 struct btrfs_qgroup_inherit *inherit = NULL;
1668 if (!S_ISDIR(file_inode(file)->i_mode))
1671 vol_args = memdup_user(arg, sizeof(*vol_args));
1672 if (IS_ERR(vol_args))
1673 return PTR_ERR(vol_args);
1674 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1676 if (vol_args->flags &
1677 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1678 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1683 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1685 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1687 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1688 if (vol_args->size > PAGE_SIZE) {
1692 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1693 if (IS_ERR(inherit)) {
1694 ret = PTR_ERR(inherit);
1699 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1700 vol_args->fd, subvol, ptr,
1705 if (ptr && copy_to_user(arg +
1706 offsetof(struct btrfs_ioctl_vol_args_v2,
1718 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1721 struct inode *inode = file_inode(file);
1722 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1723 struct btrfs_root *root = BTRFS_I(inode)->root;
1727 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1730 down_read(&fs_info->subvol_sem);
1731 if (btrfs_root_readonly(root))
1732 flags |= BTRFS_SUBVOL_RDONLY;
1733 up_read(&fs_info->subvol_sem);
1735 if (copy_to_user(arg, &flags, sizeof(flags)))
1741 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1744 struct inode *inode = file_inode(file);
1745 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1746 struct btrfs_root *root = BTRFS_I(inode)->root;
1747 struct btrfs_trans_handle *trans;
1752 if (!inode_owner_or_capable(inode))
1755 ret = mnt_want_write_file(file);
1759 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1761 goto out_drop_write;
1764 if (copy_from_user(&flags, arg, sizeof(flags))) {
1766 goto out_drop_write;
1769 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1771 goto out_drop_write;
1774 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1776 goto out_drop_write;
1779 down_write(&fs_info->subvol_sem);
1782 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1785 root_flags = btrfs_root_flags(&root->root_item);
1786 if (flags & BTRFS_SUBVOL_RDONLY) {
1787 btrfs_set_root_flags(&root->root_item,
1788 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1791 * Block RO -> RW transition if this subvolume is involved in
1794 spin_lock(&root->root_item_lock);
1795 if (root->send_in_progress == 0) {
1796 btrfs_set_root_flags(&root->root_item,
1797 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1798 spin_unlock(&root->root_item_lock);
1800 spin_unlock(&root->root_item_lock);
1802 "Attempt to set subvolume %llu read-write during send",
1803 root->root_key.objectid);
1809 trans = btrfs_start_transaction(root, 1);
1810 if (IS_ERR(trans)) {
1811 ret = PTR_ERR(trans);
1815 ret = btrfs_update_root(trans, fs_info->tree_root,
1816 &root->root_key, &root->root_item);
1818 btrfs_end_transaction(trans);
1822 ret = btrfs_commit_transaction(trans);
1826 btrfs_set_root_flags(&root->root_item, root_flags);
1828 up_write(&fs_info->subvol_sem);
1830 mnt_drop_write_file(file);
1836 * helper to check if the subvolume references other subvolumes
1838 static noinline int may_destroy_subvol(struct btrfs_root *root)
1840 struct btrfs_fs_info *fs_info = root->fs_info;
1841 struct btrfs_path *path;
1842 struct btrfs_dir_item *di;
1843 struct btrfs_key key;
1847 path = btrfs_alloc_path();
1851 /* Make sure this root isn't set as the default subvol */
1852 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1853 di = btrfs_lookup_dir_item(NULL, fs_info->tree_root, path,
1854 dir_id, "default", 7, 0);
1855 if (di && !IS_ERR(di)) {
1856 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1857 if (key.objectid == root->root_key.objectid) {
1860 "deleting default subvolume %llu is not allowed",
1864 btrfs_release_path(path);
1867 key.objectid = root->root_key.objectid;
1868 key.type = BTRFS_ROOT_REF_KEY;
1869 key.offset = (u64)-1;
1871 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
1877 if (path->slots[0] > 0) {
1879 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1880 if (key.objectid == root->root_key.objectid &&
1881 key.type == BTRFS_ROOT_REF_KEY)
1885 btrfs_free_path(path);
1889 static noinline int key_in_sk(struct btrfs_key *key,
1890 struct btrfs_ioctl_search_key *sk)
1892 struct btrfs_key test;
1895 test.objectid = sk->min_objectid;
1896 test.type = sk->min_type;
1897 test.offset = sk->min_offset;
1899 ret = btrfs_comp_cpu_keys(key, &test);
1903 test.objectid = sk->max_objectid;
1904 test.type = sk->max_type;
1905 test.offset = sk->max_offset;
1907 ret = btrfs_comp_cpu_keys(key, &test);
1913 static noinline int copy_to_sk(struct btrfs_path *path,
1914 struct btrfs_key *key,
1915 struct btrfs_ioctl_search_key *sk,
1918 unsigned long *sk_offset,
1922 struct extent_buffer *leaf;
1923 struct btrfs_ioctl_search_header sh;
1924 struct btrfs_key test;
1925 unsigned long item_off;
1926 unsigned long item_len;
1932 leaf = path->nodes[0];
1933 slot = path->slots[0];
1934 nritems = btrfs_header_nritems(leaf);
1936 if (btrfs_header_generation(leaf) > sk->max_transid) {
1940 found_transid = btrfs_header_generation(leaf);
1942 for (i = slot; i < nritems; i++) {
1943 item_off = btrfs_item_ptr_offset(leaf, i);
1944 item_len = btrfs_item_size_nr(leaf, i);
1946 btrfs_item_key_to_cpu(leaf, key, i);
1947 if (!key_in_sk(key, sk))
1950 if (sizeof(sh) + item_len > *buf_size) {
1957 * return one empty item back for v1, which does not
1961 *buf_size = sizeof(sh) + item_len;
1966 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1971 sh.objectid = key->objectid;
1972 sh.offset = key->offset;
1973 sh.type = key->type;
1975 sh.transid = found_transid;
1977 /* copy search result header */
1978 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1983 *sk_offset += sizeof(sh);
1986 char __user *up = ubuf + *sk_offset;
1988 if (read_extent_buffer_to_user(leaf, up,
1989 item_off, item_len)) {
1994 *sk_offset += item_len;
1998 if (ret) /* -EOVERFLOW from above */
2001 if (*num_found >= sk->nr_items) {
2008 test.objectid = sk->max_objectid;
2009 test.type = sk->max_type;
2010 test.offset = sk->max_offset;
2011 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2013 else if (key->offset < (u64)-1)
2015 else if (key->type < (u8)-1) {
2018 } else if (key->objectid < (u64)-1) {
2026 * 0: all items from this leaf copied, continue with next
2027 * 1: * more items can be copied, but unused buffer is too small
2028 * * all items were found
2029 * Either way, it will stops the loop which iterates to the next
2031 * -EOVERFLOW: item was to large for buffer
2032 * -EFAULT: could not copy extent buffer back to userspace
2037 static noinline int search_ioctl(struct inode *inode,
2038 struct btrfs_ioctl_search_key *sk,
2042 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2043 struct btrfs_root *root;
2044 struct btrfs_key key;
2045 struct btrfs_path *path;
2048 unsigned long sk_offset = 0;
2050 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2051 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2055 path = btrfs_alloc_path();
2059 if (sk->tree_id == 0) {
2060 /* search the root of the inode that was passed */
2061 root = BTRFS_I(inode)->root;
2063 key.objectid = sk->tree_id;
2064 key.type = BTRFS_ROOT_ITEM_KEY;
2065 key.offset = (u64)-1;
2066 root = btrfs_read_fs_root_no_name(info, &key);
2068 btrfs_free_path(path);
2073 key.objectid = sk->min_objectid;
2074 key.type = sk->min_type;
2075 key.offset = sk->min_offset;
2078 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2084 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2085 &sk_offset, &num_found);
2086 btrfs_release_path(path);
2094 sk->nr_items = num_found;
2095 btrfs_free_path(path);
2099 static noinline int btrfs_ioctl_tree_search(struct file *file,
2102 struct btrfs_ioctl_search_args __user *uargs;
2103 struct btrfs_ioctl_search_key sk;
2104 struct inode *inode;
2108 if (!capable(CAP_SYS_ADMIN))
2111 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2113 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2116 buf_size = sizeof(uargs->buf);
2118 inode = file_inode(file);
2119 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2122 * In the origin implementation an overflow is handled by returning a
2123 * search header with a len of zero, so reset ret.
2125 if (ret == -EOVERFLOW)
2128 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2133 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2136 struct btrfs_ioctl_search_args_v2 __user *uarg;
2137 struct btrfs_ioctl_search_args_v2 args;
2138 struct inode *inode;
2141 const size_t buf_limit = SZ_16M;
2143 if (!capable(CAP_SYS_ADMIN))
2146 /* copy search header and buffer size */
2147 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2148 if (copy_from_user(&args, uarg, sizeof(args)))
2151 buf_size = args.buf_size;
2153 /* limit result size to 16MB */
2154 if (buf_size > buf_limit)
2155 buf_size = buf_limit;
2157 inode = file_inode(file);
2158 ret = search_ioctl(inode, &args.key, &buf_size,
2159 (char __user *)(&uarg->buf[0]));
2160 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2162 else if (ret == -EOVERFLOW &&
2163 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2170 * Search INODE_REFs to identify path name of 'dirid' directory
2171 * in a 'tree_id' tree. and sets path name to 'name'.
2173 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2174 u64 tree_id, u64 dirid, char *name)
2176 struct btrfs_root *root;
2177 struct btrfs_key key;
2183 struct btrfs_inode_ref *iref;
2184 struct extent_buffer *l;
2185 struct btrfs_path *path;
2187 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2192 path = btrfs_alloc_path();
2196 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2198 key.objectid = tree_id;
2199 key.type = BTRFS_ROOT_ITEM_KEY;
2200 key.offset = (u64)-1;
2201 root = btrfs_read_fs_root_no_name(info, &key);
2203 btrfs_err(info, "could not find root %llu", tree_id);
2208 key.objectid = dirid;
2209 key.type = BTRFS_INODE_REF_KEY;
2210 key.offset = (u64)-1;
2213 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2217 ret = btrfs_previous_item(root, path, dirid,
2218 BTRFS_INODE_REF_KEY);
2228 slot = path->slots[0];
2229 btrfs_item_key_to_cpu(l, &key, slot);
2231 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2232 len = btrfs_inode_ref_name_len(l, iref);
2234 total_len += len + 1;
2236 ret = -ENAMETOOLONG;
2241 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2243 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2246 btrfs_release_path(path);
2247 key.objectid = key.offset;
2248 key.offset = (u64)-1;
2249 dirid = key.objectid;
2251 memmove(name, ptr, total_len);
2252 name[total_len] = '\0';
2255 btrfs_free_path(path);
2259 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2262 struct btrfs_ioctl_ino_lookup_args *args;
2263 struct inode *inode;
2266 args = memdup_user(argp, sizeof(*args));
2268 return PTR_ERR(args);
2270 inode = file_inode(file);
2273 * Unprivileged query to obtain the containing subvolume root id. The
2274 * path is reset so it's consistent with btrfs_search_path_in_tree.
2276 if (args->treeid == 0)
2277 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2279 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2284 if (!capable(CAP_SYS_ADMIN)) {
2289 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2290 args->treeid, args->objectid,
2294 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2301 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2304 struct dentry *parent = file->f_path.dentry;
2305 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2306 struct dentry *dentry;
2307 struct inode *dir = d_inode(parent);
2308 struct inode *inode;
2309 struct btrfs_root *root = BTRFS_I(dir)->root;
2310 struct btrfs_root *dest = NULL;
2311 struct btrfs_ioctl_vol_args *vol_args;
2312 struct btrfs_trans_handle *trans;
2313 struct btrfs_block_rsv block_rsv;
2315 u64 qgroup_reserved;
2320 if (!S_ISDIR(dir->i_mode))
2323 vol_args = memdup_user(arg, sizeof(*vol_args));
2324 if (IS_ERR(vol_args))
2325 return PTR_ERR(vol_args);
2327 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2328 namelen = strlen(vol_args->name);
2329 if (strchr(vol_args->name, '/') ||
2330 strncmp(vol_args->name, "..", namelen) == 0) {
2335 err = mnt_want_write_file(file);
2340 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2342 goto out_drop_write;
2343 dentry = lookup_one_len(vol_args->name, parent, namelen);
2344 if (IS_ERR(dentry)) {
2345 err = PTR_ERR(dentry);
2346 goto out_unlock_dir;
2349 if (d_really_is_negative(dentry)) {
2354 inode = d_inode(dentry);
2355 dest = BTRFS_I(inode)->root;
2356 if (!capable(CAP_SYS_ADMIN)) {
2358 * Regular user. Only allow this with a special mount
2359 * option, when the user has write+exec access to the
2360 * subvol root, and when rmdir(2) would have been
2363 * Note that this is _not_ check that the subvol is
2364 * empty or doesn't contain data that we wouldn't
2365 * otherwise be able to delete.
2367 * Users who want to delete empty subvols should try
2371 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2375 * Do not allow deletion if the parent dir is the same
2376 * as the dir to be deleted. That means the ioctl
2377 * must be called on the dentry referencing the root
2378 * of the subvol, not a random directory contained
2385 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2390 /* check if subvolume may be deleted by a user */
2391 err = btrfs_may_delete(dir, dentry, 1);
2395 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2403 * Don't allow to delete a subvolume with send in progress. This is
2404 * inside the i_mutex so the error handling that has to drop the bit
2405 * again is not run concurrently.
2407 spin_lock(&dest->root_item_lock);
2408 root_flags = btrfs_root_flags(&dest->root_item);
2409 if (dest->send_in_progress == 0) {
2410 btrfs_set_root_flags(&dest->root_item,
2411 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2412 spin_unlock(&dest->root_item_lock);
2414 spin_unlock(&dest->root_item_lock);
2416 "Attempt to delete subvolume %llu during send",
2417 dest->root_key.objectid);
2419 goto out_unlock_inode;
2422 down_write(&fs_info->subvol_sem);
2424 err = may_destroy_subvol(dest);
2428 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2430 * One for dir inode, two for dir entries, two for root
2433 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2434 5, &qgroup_reserved, true);
2438 trans = btrfs_start_transaction(root, 0);
2439 if (IS_ERR(trans)) {
2440 err = PTR_ERR(trans);
2443 trans->block_rsv = &block_rsv;
2444 trans->bytes_reserved = block_rsv.size;
2446 btrfs_record_snapshot_destroy(trans, BTRFS_I(dir));
2448 ret = btrfs_unlink_subvol(trans, root, dir,
2449 dest->root_key.objectid,
2450 dentry->d_name.name,
2451 dentry->d_name.len);
2454 btrfs_abort_transaction(trans, ret);
2458 btrfs_record_root_in_trans(trans, dest);
2460 memset(&dest->root_item.drop_progress, 0,
2461 sizeof(dest->root_item.drop_progress));
2462 dest->root_item.drop_level = 0;
2463 btrfs_set_root_refs(&dest->root_item, 0);
2465 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2466 ret = btrfs_insert_orphan_item(trans,
2468 dest->root_key.objectid);
2470 btrfs_abort_transaction(trans, ret);
2476 ret = btrfs_uuid_tree_rem(trans, fs_info, dest->root_item.uuid,
2477 BTRFS_UUID_KEY_SUBVOL,
2478 dest->root_key.objectid);
2479 if (ret && ret != -ENOENT) {
2480 btrfs_abort_transaction(trans, ret);
2484 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2485 ret = btrfs_uuid_tree_rem(trans, fs_info,
2486 dest->root_item.received_uuid,
2487 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2488 dest->root_key.objectid);
2489 if (ret && ret != -ENOENT) {
2490 btrfs_abort_transaction(trans, ret);
2497 trans->block_rsv = NULL;
2498 trans->bytes_reserved = 0;
2499 ret = btrfs_end_transaction(trans);
2502 inode->i_flags |= S_DEAD;
2504 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
2506 up_write(&fs_info->subvol_sem);
2508 spin_lock(&dest->root_item_lock);
2509 root_flags = btrfs_root_flags(&dest->root_item);
2510 btrfs_set_root_flags(&dest->root_item,
2511 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2512 spin_unlock(&dest->root_item_lock);
2515 inode_unlock(inode);
2517 d_invalidate(dentry);
2518 btrfs_invalidate_inodes(dest);
2520 ASSERT(dest->send_in_progress == 0);
2523 if (dest->ino_cache_inode) {
2524 iput(dest->ino_cache_inode);
2525 dest->ino_cache_inode = NULL;
2533 mnt_drop_write_file(file);
2539 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2541 struct inode *inode = file_inode(file);
2542 struct btrfs_root *root = BTRFS_I(inode)->root;
2543 struct btrfs_ioctl_defrag_range_args *range;
2546 ret = mnt_want_write_file(file);
2550 if (btrfs_root_readonly(root)) {
2555 switch (inode->i_mode & S_IFMT) {
2557 if (!capable(CAP_SYS_ADMIN)) {
2561 ret = btrfs_defrag_root(root);
2564 if (!(file->f_mode & FMODE_WRITE)) {
2569 range = kzalloc(sizeof(*range), GFP_KERNEL);
2576 if (copy_from_user(range, argp,
2582 /* compression requires us to start the IO */
2583 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2584 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2585 range->extent_thresh = (u32)-1;
2588 /* the rest are all set to zero by kzalloc */
2589 range->len = (u64)-1;
2591 ret = btrfs_defrag_file(file_inode(file), file,
2592 range, BTRFS_OLDEST_GENERATION, 0);
2601 mnt_drop_write_file(file);
2605 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2607 struct btrfs_ioctl_vol_args *vol_args;
2610 if (!capable(CAP_SYS_ADMIN))
2613 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
2614 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2616 mutex_lock(&fs_info->volume_mutex);
2617 vol_args = memdup_user(arg, sizeof(*vol_args));
2618 if (IS_ERR(vol_args)) {
2619 ret = PTR_ERR(vol_args);
2623 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2624 ret = btrfs_init_new_device(fs_info, vol_args->name);
2627 btrfs_info(fs_info, "disk added %s", vol_args->name);
2631 mutex_unlock(&fs_info->volume_mutex);
2632 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2636 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2638 struct inode *inode = file_inode(file);
2639 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2640 struct btrfs_ioctl_vol_args_v2 *vol_args;
2643 if (!capable(CAP_SYS_ADMIN))
2646 ret = mnt_want_write_file(file);
2650 vol_args = memdup_user(arg, sizeof(*vol_args));
2651 if (IS_ERR(vol_args)) {
2652 ret = PTR_ERR(vol_args);
2656 /* Check for compatibility reject unknown flags */
2657 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED)
2660 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2661 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2665 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2666 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
2668 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2669 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2671 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2674 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2675 btrfs_info(fs_info, "device deleted: id %llu",
2678 btrfs_info(fs_info, "device deleted: %s",
2684 mnt_drop_write_file(file);
2688 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2690 struct inode *inode = file_inode(file);
2691 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2692 struct btrfs_ioctl_vol_args *vol_args;
2695 if (!capable(CAP_SYS_ADMIN))
2698 ret = mnt_want_write_file(file);
2702 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2703 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2704 goto out_drop_write;
2707 vol_args = memdup_user(arg, sizeof(*vol_args));
2708 if (IS_ERR(vol_args)) {
2709 ret = PTR_ERR(vol_args);
2713 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2714 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2717 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2720 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2722 mnt_drop_write_file(file);
2727 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2730 struct btrfs_ioctl_fs_info_args *fi_args;
2731 struct btrfs_device *device;
2732 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2735 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2740 fi_args->num_devices = fs_devices->num_devices;
2742 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2743 if (device->devid > fi_args->max_id)
2744 fi_args->max_id = device->devid;
2748 memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid));
2749 fi_args->nodesize = fs_info->nodesize;
2750 fi_args->sectorsize = fs_info->sectorsize;
2751 fi_args->clone_alignment = fs_info->sectorsize;
2753 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2760 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2763 struct btrfs_ioctl_dev_info_args *di_args;
2764 struct btrfs_device *dev;
2766 char *s_uuid = NULL;
2768 di_args = memdup_user(arg, sizeof(*di_args));
2769 if (IS_ERR(di_args))
2770 return PTR_ERR(di_args);
2772 if (!btrfs_is_empty_uuid(di_args->uuid))
2773 s_uuid = di_args->uuid;
2776 dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL);
2783 di_args->devid = dev->devid;
2784 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2785 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2786 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2788 struct rcu_string *name;
2790 name = rcu_dereference(dev->name);
2791 strncpy(di_args->path, name->str, sizeof(di_args->path) - 1);
2792 di_args->path[sizeof(di_args->path) - 1] = 0;
2794 di_args->path[0] = '\0';
2799 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2806 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2810 page = grab_cache_page(inode->i_mapping, index);
2812 return ERR_PTR(-ENOMEM);
2814 if (!PageUptodate(page)) {
2817 ret = btrfs_readpage(NULL, page);
2819 return ERR_PTR(ret);
2821 if (!PageUptodate(page)) {
2824 return ERR_PTR(-EIO);
2826 if (page->mapping != inode->i_mapping) {
2829 return ERR_PTR(-EAGAIN);
2836 static int gather_extent_pages(struct inode *inode, struct page **pages,
2837 int num_pages, u64 off)
2840 pgoff_t index = off >> PAGE_SHIFT;
2842 for (i = 0; i < num_pages; i++) {
2844 pages[i] = extent_same_get_page(inode, index + i);
2845 if (IS_ERR(pages[i])) {
2846 int err = PTR_ERR(pages[i]);
2857 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2858 bool retry_range_locking)
2861 * Do any pending delalloc/csum calculations on inode, one way or
2862 * another, and lock file content.
2863 * The locking order is:
2866 * 2) range in the inode's io tree
2869 struct btrfs_ordered_extent *ordered;
2870 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2871 ordered = btrfs_lookup_first_ordered_extent(inode,
2874 ordered->file_offset + ordered->len <= off ||
2875 ordered->file_offset >= off + len) &&
2876 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2877 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2879 btrfs_put_ordered_extent(ordered);
2882 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2884 btrfs_put_ordered_extent(ordered);
2885 if (!retry_range_locking)
2887 btrfs_wait_ordered_range(inode, off, len);
2892 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2894 inode_unlock(inode1);
2895 inode_unlock(inode2);
2898 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2900 if (inode1 < inode2)
2901 swap(inode1, inode2);
2903 inode_lock_nested(inode1, I_MUTEX_PARENT);
2904 inode_lock_nested(inode2, I_MUTEX_CHILD);
2907 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2908 struct inode *inode2, u64 loff2, u64 len)
2910 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2911 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2914 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2915 struct inode *inode2, u64 loff2, u64 len,
2916 bool retry_range_locking)
2920 if (inode1 < inode2) {
2921 swap(inode1, inode2);
2924 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2927 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2929 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2936 struct page **src_pages;
2937 struct page **dst_pages;
2940 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2945 for (i = 0; i < cmp->num_pages; i++) {
2946 pg = cmp->src_pages[i];
2951 pg = cmp->dst_pages[i];
2957 kfree(cmp->src_pages);
2958 kfree(cmp->dst_pages);
2961 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2962 struct inode *dst, u64 dst_loff,
2963 u64 len, struct cmp_pages *cmp)
2966 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
2967 struct page **src_pgarr, **dst_pgarr;
2970 * We must gather up all the pages before we initiate our
2971 * extent locking. We use an array for the page pointers. Size
2972 * of the array is bounded by len, which is in turn bounded by
2973 * BTRFS_MAX_DEDUPE_LEN.
2975 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2976 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2977 if (!src_pgarr || !dst_pgarr) {
2982 cmp->num_pages = num_pages;
2983 cmp->src_pages = src_pgarr;
2984 cmp->dst_pages = dst_pgarr;
2987 * If deduping ranges in the same inode, locking rules make it mandatory
2988 * to always lock pages in ascending order to avoid deadlocks with
2989 * concurrent tasks (such as starting writeback/delalloc).
2991 if (src == dst && dst_loff < loff) {
2992 swap(src_pgarr, dst_pgarr);
2993 swap(loff, dst_loff);
2996 ret = gather_extent_pages(src, src_pgarr, cmp->num_pages, loff);
3000 ret = gather_extent_pages(dst, dst_pgarr, cmp->num_pages, dst_loff);
3004 btrfs_cmp_data_free(cmp);
3008 static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp)
3012 struct page *src_page, *dst_page;
3013 unsigned int cmp_len = PAGE_SIZE;
3014 void *addr, *dst_addr;
3018 if (len < PAGE_SIZE)
3021 BUG_ON(i >= cmp->num_pages);
3023 src_page = cmp->src_pages[i];
3024 dst_page = cmp->dst_pages[i];
3025 ASSERT(PageLocked(src_page));
3026 ASSERT(PageLocked(dst_page));
3028 addr = kmap_atomic(src_page);
3029 dst_addr = kmap_atomic(dst_page);
3031 flush_dcache_page(src_page);
3032 flush_dcache_page(dst_page);
3034 if (memcmp(addr, dst_addr, cmp_len))
3037 kunmap_atomic(addr);
3038 kunmap_atomic(dst_addr);
3050 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3054 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3056 if (off + olen > inode->i_size || off + olen < off)
3059 /* if we extend to eof, continue to block boundary */
3060 if (off + len == inode->i_size)
3061 *plen = len = ALIGN(inode->i_size, bs) - off;
3063 /* Check that we are block aligned - btrfs_clone() requires this */
3064 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3070 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3071 struct inode *dst, u64 dst_loff)
3075 struct cmp_pages cmp;
3076 bool same_inode = (src == dst);
3077 u64 same_lock_start = 0;
3078 u64 same_lock_len = 0;
3086 btrfs_double_inode_lock(src, dst);
3088 ret = extent_same_check_offsets(src, loff, &len, olen);
3092 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3098 * Single inode case wants the same checks, except we
3099 * don't want our length pushed out past i_size as
3100 * comparing that data range makes no sense.
3102 * extent_same_check_offsets() will do this for an
3103 * unaligned length at i_size, so catch it here and
3104 * reject the request.
3106 * This effectively means we require aligned extents
3107 * for the single-inode case, whereas the other cases
3108 * allow an unaligned length so long as it ends at
3116 /* Check for overlapping ranges */
3117 if (dst_loff + len > loff && dst_loff < loff + len) {
3122 same_lock_start = min_t(u64, loff, dst_loff);
3123 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3126 /* don't make the dst file partly checksummed */
3127 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3128 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3134 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3139 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3142 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3145 * If one of the inodes has dirty pages in the respective range or
3146 * ordered extents, we need to flush dellaloc and wait for all ordered
3147 * extents in the range. We must unlock the pages and the ranges in the
3148 * io trees to avoid deadlocks when flushing delalloc (requires locking
3149 * pages) and when waiting for ordered extents to complete (they require
3152 if (ret == -EAGAIN) {
3154 * Ranges in the io trees already unlocked. Now unlock all
3155 * pages before waiting for all IO to complete.
3157 btrfs_cmp_data_free(&cmp);
3159 btrfs_wait_ordered_range(src, same_lock_start,
3162 btrfs_wait_ordered_range(src, loff, len);
3163 btrfs_wait_ordered_range(dst, dst_loff, len);
3169 /* ranges in the io trees already unlocked */
3170 btrfs_cmp_data_free(&cmp);
3174 /* pass original length for comparison so we stay within i_size */
3175 ret = btrfs_cmp_data(olen, &cmp);
3177 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3180 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3181 same_lock_start + same_lock_len - 1);
3183 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3185 btrfs_cmp_data_free(&cmp);
3190 btrfs_double_inode_unlock(src, dst);
3195 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3197 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3198 struct file *dst_file, u64 dst_loff)
3200 struct inode *src = file_inode(src_file);
3201 struct inode *dst = file_inode(dst_file);
3202 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3205 if (olen > BTRFS_MAX_DEDUPE_LEN)
3206 olen = BTRFS_MAX_DEDUPE_LEN;
3208 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
3210 * Btrfs does not support blocksize < page_size. As a
3211 * result, btrfs_cmp_data() won't correctly handle
3212 * this situation without an update.
3217 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3223 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3224 struct inode *inode,
3230 struct btrfs_root *root = BTRFS_I(inode)->root;
3233 inode_inc_iversion(inode);
3234 if (!no_time_update)
3235 inode->i_mtime = inode->i_ctime = current_time(inode);
3237 * We round up to the block size at eof when determining which
3238 * extents to clone above, but shouldn't round up the file size.
3240 if (endoff > destoff + olen)
3241 endoff = destoff + olen;
3242 if (endoff > inode->i_size)
3243 btrfs_i_size_write(BTRFS_I(inode), endoff);
3245 ret = btrfs_update_inode(trans, root, inode);
3247 btrfs_abort_transaction(trans, ret);
3248 btrfs_end_transaction(trans);
3251 ret = btrfs_end_transaction(trans);
3256 static void clone_update_extent_map(struct btrfs_inode *inode,
3257 const struct btrfs_trans_handle *trans,
3258 const struct btrfs_path *path,
3259 const u64 hole_offset,
3262 struct extent_map_tree *em_tree = &inode->extent_tree;
3263 struct extent_map *em;
3266 em = alloc_extent_map();
3268 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3273 struct btrfs_file_extent_item *fi;
3275 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3276 struct btrfs_file_extent_item);
3277 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3278 em->generation = -1;
3279 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3280 BTRFS_FILE_EXTENT_INLINE)
3281 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3282 &inode->runtime_flags);
3284 em->start = hole_offset;
3286 em->ram_bytes = em->len;
3287 em->orig_start = hole_offset;
3288 em->block_start = EXTENT_MAP_HOLE;
3290 em->orig_block_len = 0;
3291 em->compress_type = BTRFS_COMPRESS_NONE;
3292 em->generation = trans->transid;
3296 write_lock(&em_tree->lock);
3297 ret = add_extent_mapping(em_tree, em, 1);
3298 write_unlock(&em_tree->lock);
3299 if (ret != -EEXIST) {
3300 free_extent_map(em);
3303 btrfs_drop_extent_cache(inode, em->start,
3304 em->start + em->len - 1, 0);
3308 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3312 * Make sure we do not end up inserting an inline extent into a file that has
3313 * already other (non-inline) extents. If a file has an inline extent it can
3314 * not have any other extents and the (single) inline extent must start at the
3315 * file offset 0. Failing to respect these rules will lead to file corruption,
3316 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3318 * We can have extents that have been already written to disk or we can have
3319 * dirty ranges still in delalloc, in which case the extent maps and items are
3320 * created only when we run delalloc, and the delalloc ranges might fall outside
3321 * the range we are currently locking in the inode's io tree. So we check the
3322 * inode's i_size because of that (i_size updates are done while holding the
3323 * i_mutex, which we are holding here).
3324 * We also check to see if the inode has a size not greater than "datal" but has
3325 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3326 * protected against such concurrent fallocate calls by the i_mutex).
3328 * If the file has no extents but a size greater than datal, do not allow the
3329 * copy because we would need turn the inline extent into a non-inline one (even
3330 * with NO_HOLES enabled). If we find our destination inode only has one inline
3331 * extent, just overwrite it with the source inline extent if its size is less
3332 * than the source extent's size, or we could copy the source inline extent's
3333 * data into the destination inode's inline extent if the later is greater then
3336 static int clone_copy_inline_extent(struct inode *dst,
3337 struct btrfs_trans_handle *trans,
3338 struct btrfs_path *path,
3339 struct btrfs_key *new_key,
3340 const u64 drop_start,
3346 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
3347 struct btrfs_root *root = BTRFS_I(dst)->root;
3348 const u64 aligned_end = ALIGN(new_key->offset + datal,
3349 fs_info->sectorsize);
3351 struct btrfs_key key;
3353 if (new_key->offset > 0)
3356 key.objectid = btrfs_ino(BTRFS_I(dst));
3357 key.type = BTRFS_EXTENT_DATA_KEY;
3359 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3362 } else if (ret > 0) {
3363 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3364 ret = btrfs_next_leaf(root, path);
3368 goto copy_inline_extent;
3370 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3371 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3372 key.type == BTRFS_EXTENT_DATA_KEY) {
3373 ASSERT(key.offset > 0);
3376 } else if (i_size_read(dst) <= datal) {
3377 struct btrfs_file_extent_item *ei;
3381 * If the file size is <= datal, make sure there are no other
3382 * extents following (can happen do to an fallocate call with
3383 * the flag FALLOC_FL_KEEP_SIZE).
3385 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3386 struct btrfs_file_extent_item);
3388 * If it's an inline extent, it can not have other extents
3391 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3392 BTRFS_FILE_EXTENT_INLINE)
3393 goto copy_inline_extent;
3395 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3396 if (ext_len > aligned_end)
3399 ret = btrfs_next_item(root, path);
3402 } else if (ret == 0) {
3403 btrfs_item_key_to_cpu(path->nodes[0], &key,
3405 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3406 key.type == BTRFS_EXTENT_DATA_KEY)
3413 * We have no extent items, or we have an extent at offset 0 which may
3414 * or may not be inlined. All these cases are dealt the same way.
3416 if (i_size_read(dst) > datal) {
3418 * If the destination inode has an inline extent...
3419 * This would require copying the data from the source inline
3420 * extent into the beginning of the destination's inline extent.
3421 * But this is really complex, both extents can be compressed
3422 * or just one of them, which would require decompressing and
3423 * re-compressing data (which could increase the new compressed
3424 * size, not allowing the compressed data to fit anymore in an
3426 * So just don't support this case for now (it should be rare,
3427 * we are not really saving space when cloning inline extents).
3432 btrfs_release_path(path);
3433 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3436 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3441 const u32 start = btrfs_file_extent_calc_inline_size(0);
3443 memmove(inline_data + start, inline_data + start + skip, datal);
3446 write_extent_buffer(path->nodes[0], inline_data,
3447 btrfs_item_ptr_offset(path->nodes[0],
3450 inode_add_bytes(dst, datal);
3456 * btrfs_clone() - clone a range from inode file to another
3458 * @src: Inode to clone from
3459 * @inode: Inode to clone to
3460 * @off: Offset within source to start clone from
3461 * @olen: Original length, passed by user, of range to clone
3462 * @olen_aligned: Block-aligned value of olen
3463 * @destoff: Offset within @inode to start clone
3464 * @no_time_update: Whether to update mtime/ctime on the target inode
3466 static int btrfs_clone(struct inode *src, struct inode *inode,
3467 const u64 off, const u64 olen, const u64 olen_aligned,
3468 const u64 destoff, int no_time_update)
3470 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3471 struct btrfs_root *root = BTRFS_I(inode)->root;
3472 struct btrfs_path *path = NULL;
3473 struct extent_buffer *leaf;
3474 struct btrfs_trans_handle *trans;
3476 struct btrfs_key key;
3480 const u64 len = olen_aligned;
3481 u64 last_dest_end = destoff;
3484 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3488 path = btrfs_alloc_path();
3494 path->reada = READA_FORWARD;
3496 key.objectid = btrfs_ino(BTRFS_I(src));
3497 key.type = BTRFS_EXTENT_DATA_KEY;
3501 u64 next_key_min_offset = key.offset + 1;
3504 * note the key will change type as we walk through the
3507 path->leave_spinning = 1;
3508 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3513 * First search, if no extent item that starts at offset off was
3514 * found but the previous item is an extent item, it's possible
3515 * it might overlap our target range, therefore process it.
3517 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3518 btrfs_item_key_to_cpu(path->nodes[0], &key,
3519 path->slots[0] - 1);
3520 if (key.type == BTRFS_EXTENT_DATA_KEY)
3524 nritems = btrfs_header_nritems(path->nodes[0]);
3526 if (path->slots[0] >= nritems) {
3527 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3532 nritems = btrfs_header_nritems(path->nodes[0]);
3534 leaf = path->nodes[0];
3535 slot = path->slots[0];
3537 btrfs_item_key_to_cpu(leaf, &key, slot);
3538 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3539 key.objectid != btrfs_ino(BTRFS_I(src)))
3542 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3543 struct btrfs_file_extent_item *extent;
3546 struct btrfs_key new_key;
3547 u64 disko = 0, diskl = 0;
3548 u64 datao = 0, datal = 0;
3552 extent = btrfs_item_ptr(leaf, slot,
3553 struct btrfs_file_extent_item);
3554 comp = btrfs_file_extent_compression(leaf, extent);
3555 type = btrfs_file_extent_type(leaf, extent);
3556 if (type == BTRFS_FILE_EXTENT_REG ||
3557 type == BTRFS_FILE_EXTENT_PREALLOC) {
3558 disko = btrfs_file_extent_disk_bytenr(leaf,
3560 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3562 datao = btrfs_file_extent_offset(leaf, extent);
3563 datal = btrfs_file_extent_num_bytes(leaf,
3565 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3566 /* take upper bound, may be compressed */
3567 datal = btrfs_file_extent_ram_bytes(leaf,
3572 * The first search might have left us at an extent
3573 * item that ends before our target range's start, can
3574 * happen if we have holes and NO_HOLES feature enabled.
3576 if (key.offset + datal <= off) {
3579 } else if (key.offset >= off + len) {
3582 next_key_min_offset = key.offset + datal;
3583 size = btrfs_item_size_nr(leaf, slot);
3584 read_extent_buffer(leaf, buf,
3585 btrfs_item_ptr_offset(leaf, slot),
3588 btrfs_release_path(path);
3589 path->leave_spinning = 0;
3591 memcpy(&new_key, &key, sizeof(new_key));
3592 new_key.objectid = btrfs_ino(BTRFS_I(inode));
3593 if (off <= key.offset)
3594 new_key.offset = key.offset + destoff - off;
3596 new_key.offset = destoff;
3599 * Deal with a hole that doesn't have an extent item
3600 * that represents it (NO_HOLES feature enabled).
3601 * This hole is either in the middle of the cloning
3602 * range or at the beginning (fully overlaps it or
3603 * partially overlaps it).
3605 if (new_key.offset != last_dest_end)
3606 drop_start = last_dest_end;
3608 drop_start = new_key.offset;
3611 * 1 - adjusting old extent (we may have to split it)
3612 * 1 - add new extent
3615 trans = btrfs_start_transaction(root, 3);
3616 if (IS_ERR(trans)) {
3617 ret = PTR_ERR(trans);
3621 if (type == BTRFS_FILE_EXTENT_REG ||
3622 type == BTRFS_FILE_EXTENT_PREALLOC) {
3624 * a | --- range to clone ---| b
3625 * | ------------- extent ------------- |
3628 /* subtract range b */
3629 if (key.offset + datal > off + len)
3630 datal = off + len - key.offset;
3632 /* subtract range a */
3633 if (off > key.offset) {
3634 datao += off - key.offset;
3635 datal -= off - key.offset;
3638 ret = btrfs_drop_extents(trans, root, inode,
3640 new_key.offset + datal,
3643 if (ret != -EOPNOTSUPP)
3644 btrfs_abort_transaction(trans,
3646 btrfs_end_transaction(trans);
3650 ret = btrfs_insert_empty_item(trans, root, path,
3653 btrfs_abort_transaction(trans, ret);
3654 btrfs_end_transaction(trans);
3658 leaf = path->nodes[0];
3659 slot = path->slots[0];
3660 write_extent_buffer(leaf, buf,
3661 btrfs_item_ptr_offset(leaf, slot),
3664 extent = btrfs_item_ptr(leaf, slot,
3665 struct btrfs_file_extent_item);
3667 /* disko == 0 means it's a hole */
3671 btrfs_set_file_extent_offset(leaf, extent,
3673 btrfs_set_file_extent_num_bytes(leaf, extent,
3677 inode_add_bytes(inode, datal);
3678 ret = btrfs_inc_extent_ref(trans,
3681 root->root_key.objectid,
3682 btrfs_ino(BTRFS_I(inode)),
3683 new_key.offset - datao);
3685 btrfs_abort_transaction(trans,
3687 btrfs_end_transaction(trans);
3692 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3696 if (off > key.offset) {
3697 skip = off - key.offset;
3698 new_key.offset += skip;
3701 if (key.offset + datal > off + len)
3702 trim = key.offset + datal - (off + len);
3704 if (comp && (skip || trim)) {
3706 btrfs_end_transaction(trans);
3709 size -= skip + trim;
3710 datal -= skip + trim;
3712 ret = clone_copy_inline_extent(inode,
3719 if (ret != -EOPNOTSUPP)
3720 btrfs_abort_transaction(trans,
3722 btrfs_end_transaction(trans);
3725 leaf = path->nodes[0];
3726 slot = path->slots[0];
3729 /* If we have an implicit hole (NO_HOLES feature). */
3730 if (drop_start < new_key.offset)
3731 clone_update_extent_map(BTRFS_I(inode), trans,
3733 new_key.offset - drop_start);
3735 clone_update_extent_map(BTRFS_I(inode), trans,
3738 btrfs_mark_buffer_dirty(leaf);
3739 btrfs_release_path(path);
3741 last_dest_end = ALIGN(new_key.offset + datal,
3742 fs_info->sectorsize);
3743 ret = clone_finish_inode_update(trans, inode,
3749 if (new_key.offset + datal >= destoff + len)
3752 btrfs_release_path(path);
3753 key.offset = next_key_min_offset;
3755 if (fatal_signal_pending(current)) {
3762 if (last_dest_end < destoff + len) {
3764 * We have an implicit hole (NO_HOLES feature is enabled) that
3765 * fully or partially overlaps our cloning range at its end.
3767 btrfs_release_path(path);
3770 * 1 - remove extent(s)
3773 trans = btrfs_start_transaction(root, 2);
3774 if (IS_ERR(trans)) {
3775 ret = PTR_ERR(trans);
3778 ret = btrfs_drop_extents(trans, root, inode,
3779 last_dest_end, destoff + len, 1);
3781 if (ret != -EOPNOTSUPP)
3782 btrfs_abort_transaction(trans, ret);
3783 btrfs_end_transaction(trans);
3786 clone_update_extent_map(BTRFS_I(inode), trans, NULL,
3788 destoff + len - last_dest_end);
3789 ret = clone_finish_inode_update(trans, inode, destoff + len,
3790 destoff, olen, no_time_update);
3794 btrfs_free_path(path);
3799 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3800 u64 off, u64 olen, u64 destoff)
3802 struct inode *inode = file_inode(file);
3803 struct inode *src = file_inode(file_src);
3804 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3805 struct btrfs_root *root = BTRFS_I(inode)->root;
3808 u64 bs = fs_info->sb->s_blocksize;
3809 int same_inode = src == inode;
3813 * - split compressed inline extents. annoying: we need to
3814 * decompress into destination's address_space (the file offset
3815 * may change, so source mapping won't do), then recompress (or
3816 * otherwise reinsert) a subrange.
3818 * - split destination inode's inline extents. The inline extents can
3819 * be either compressed or non-compressed.
3822 if (btrfs_root_readonly(root))
3825 if (file_src->f_path.mnt != file->f_path.mnt ||
3826 src->i_sb != inode->i_sb)
3829 /* don't make the dst file partly checksummed */
3830 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3831 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3834 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3838 btrfs_double_inode_lock(src, inode);
3843 /* determine range to clone */
3845 if (off + len > src->i_size || off + len < off)
3848 olen = len = src->i_size - off;
3849 /* if we extend to eof, continue to block boundary */
3850 if (off + len == src->i_size)
3851 len = ALIGN(src->i_size, bs) - off;
3858 /* verify the end result is block aligned */
3859 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3860 !IS_ALIGNED(destoff, bs))
3863 /* verify if ranges are overlapped within the same file */
3865 if (destoff + len > off && destoff < off + len)
3869 if (destoff > inode->i_size) {
3870 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3876 * Lock the target range too. Right after we replace the file extent
3877 * items in the fs tree (which now point to the cloned data), we might
3878 * have a worker replace them with extent items relative to a write
3879 * operation that was issued before this clone operation (i.e. confront
3880 * with inode.c:btrfs_finish_ordered_io).
3883 u64 lock_start = min_t(u64, off, destoff);
3884 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3886 ret = lock_extent_range(src, lock_start, lock_len, true);
3888 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3893 /* ranges in the io trees already unlocked */
3897 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3900 u64 lock_start = min_t(u64, off, destoff);
3901 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3903 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3905 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3908 * Truncate page cache pages so that future reads will see the cloned
3909 * data immediately and not the previous data.
3911 truncate_inode_pages_range(&inode->i_data,
3912 round_down(destoff, PAGE_SIZE),
3913 round_up(destoff + len, PAGE_SIZE) - 1);
3916 btrfs_double_inode_unlock(src, inode);
3922 int btrfs_clone_file_range(struct file *src_file, loff_t off,
3923 struct file *dst_file, loff_t destoff, u64 len)
3925 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
3928 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3930 struct inode *inode = file_inode(file);
3931 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3932 struct btrfs_root *root = BTRFS_I(inode)->root;
3933 struct btrfs_root *new_root;
3934 struct btrfs_dir_item *di;
3935 struct btrfs_trans_handle *trans;
3936 struct btrfs_path *path;
3937 struct btrfs_key location;
3938 struct btrfs_disk_key disk_key;
3943 if (!capable(CAP_SYS_ADMIN))
3946 ret = mnt_want_write_file(file);
3950 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3956 objectid = BTRFS_FS_TREE_OBJECTID;
3958 location.objectid = objectid;
3959 location.type = BTRFS_ROOT_ITEM_KEY;
3960 location.offset = (u64)-1;
3962 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
3963 if (IS_ERR(new_root)) {
3964 ret = PTR_ERR(new_root);
3967 if (!is_fstree(new_root->objectid)) {
3972 path = btrfs_alloc_path();
3977 path->leave_spinning = 1;
3979 trans = btrfs_start_transaction(root, 1);
3980 if (IS_ERR(trans)) {
3981 btrfs_free_path(path);
3982 ret = PTR_ERR(trans);
3986 dir_id = btrfs_super_root_dir(fs_info->super_copy);
3987 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3988 dir_id, "default", 7, 1);
3989 if (IS_ERR_OR_NULL(di)) {
3990 btrfs_free_path(path);
3991 btrfs_end_transaction(trans);
3993 "Umm, you don't have the default diritem, this isn't going to work");
3998 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3999 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4000 btrfs_mark_buffer_dirty(path->nodes[0]);
4001 btrfs_free_path(path);
4003 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
4004 btrfs_end_transaction(trans);
4006 mnt_drop_write_file(file);
4010 void btrfs_get_block_group_info(struct list_head *groups_list,
4011 struct btrfs_ioctl_space_info *space)
4013 struct btrfs_block_group_cache *block_group;
4015 space->total_bytes = 0;
4016 space->used_bytes = 0;
4018 list_for_each_entry(block_group, groups_list, list) {
4019 space->flags = block_group->flags;
4020 space->total_bytes += block_group->key.offset;
4021 space->used_bytes +=
4022 btrfs_block_group_used(&block_group->item);
4026 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
4029 struct btrfs_ioctl_space_args space_args;
4030 struct btrfs_ioctl_space_info space;
4031 struct btrfs_ioctl_space_info *dest;
4032 struct btrfs_ioctl_space_info *dest_orig;
4033 struct btrfs_ioctl_space_info __user *user_dest;
4034 struct btrfs_space_info *info;
4035 static const u64 types[] = {
4036 BTRFS_BLOCK_GROUP_DATA,
4037 BTRFS_BLOCK_GROUP_SYSTEM,
4038 BTRFS_BLOCK_GROUP_METADATA,
4039 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
4047 if (copy_from_user(&space_args,
4048 (struct btrfs_ioctl_space_args __user *)arg,
4049 sizeof(space_args)))
4052 for (i = 0; i < num_types; i++) {
4053 struct btrfs_space_info *tmp;
4057 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4059 if (tmp->flags == types[i]) {
4069 down_read(&info->groups_sem);
4070 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4071 if (!list_empty(&info->block_groups[c]))
4074 up_read(&info->groups_sem);
4078 * Global block reserve, exported as a space_info
4082 /* space_slots == 0 means they are asking for a count */
4083 if (space_args.space_slots == 0) {
4084 space_args.total_spaces = slot_count;
4088 slot_count = min_t(u64, space_args.space_slots, slot_count);
4090 alloc_size = sizeof(*dest) * slot_count;
4092 /* we generally have at most 6 or so space infos, one for each raid
4093 * level. So, a whole page should be more than enough for everyone
4095 if (alloc_size > PAGE_SIZE)
4098 space_args.total_spaces = 0;
4099 dest = kmalloc(alloc_size, GFP_KERNEL);
4104 /* now we have a buffer to copy into */
4105 for (i = 0; i < num_types; i++) {
4106 struct btrfs_space_info *tmp;
4113 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4115 if (tmp->flags == types[i]) {
4124 down_read(&info->groups_sem);
4125 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4126 if (!list_empty(&info->block_groups[c])) {
4127 btrfs_get_block_group_info(
4128 &info->block_groups[c], &space);
4129 memcpy(dest, &space, sizeof(space));
4131 space_args.total_spaces++;
4137 up_read(&info->groups_sem);
4141 * Add global block reserve
4144 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
4146 spin_lock(&block_rsv->lock);
4147 space.total_bytes = block_rsv->size;
4148 space.used_bytes = block_rsv->size - block_rsv->reserved;
4149 spin_unlock(&block_rsv->lock);
4150 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4151 memcpy(dest, &space, sizeof(space));
4152 space_args.total_spaces++;
4155 user_dest = (struct btrfs_ioctl_space_info __user *)
4156 (arg + sizeof(struct btrfs_ioctl_space_args));
4158 if (copy_to_user(user_dest, dest_orig, alloc_size))
4163 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4169 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4172 struct btrfs_trans_handle *trans;
4176 trans = btrfs_attach_transaction_barrier(root);
4177 if (IS_ERR(trans)) {
4178 if (PTR_ERR(trans) != -ENOENT)
4179 return PTR_ERR(trans);
4181 /* No running transaction, don't bother */
4182 transid = root->fs_info->last_trans_committed;
4185 transid = trans->transid;
4186 ret = btrfs_commit_transaction_async(trans, 0);
4188 btrfs_end_transaction(trans);
4193 if (copy_to_user(argp, &transid, sizeof(transid)))
4198 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
4204 if (copy_from_user(&transid, argp, sizeof(transid)))
4207 transid = 0; /* current trans */
4209 return btrfs_wait_for_commit(fs_info, transid);
4212 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4214 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4215 struct btrfs_ioctl_scrub_args *sa;
4218 if (!capable(CAP_SYS_ADMIN))
4221 sa = memdup_user(arg, sizeof(*sa));
4225 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4226 ret = mnt_want_write_file(file);
4231 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4232 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4235 if (copy_to_user(arg, sa, sizeof(*sa)))
4238 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4239 mnt_drop_write_file(file);
4245 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
4247 if (!capable(CAP_SYS_ADMIN))
4250 return btrfs_scrub_cancel(fs_info);
4253 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
4256 struct btrfs_ioctl_scrub_args *sa;
4259 if (!capable(CAP_SYS_ADMIN))
4262 sa = memdup_user(arg, sizeof(*sa));
4266 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
4268 if (copy_to_user(arg, sa, sizeof(*sa)))
4275 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
4278 struct btrfs_ioctl_get_dev_stats *sa;
4281 sa = memdup_user(arg, sizeof(*sa));
4285 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4290 ret = btrfs_get_dev_stats(fs_info, sa);
4292 if (copy_to_user(arg, sa, sizeof(*sa)))
4299 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4302 struct btrfs_ioctl_dev_replace_args *p;
4305 if (!capable(CAP_SYS_ADMIN))
4308 p = memdup_user(arg, sizeof(*p));
4313 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4314 if (sb_rdonly(fs_info->sb)) {
4318 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4319 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4321 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
4322 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4325 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4326 btrfs_dev_replace_status(fs_info, p);
4329 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4330 p->result = btrfs_dev_replace_cancel(fs_info);
4338 if (copy_to_user(arg, p, sizeof(*p)))
4345 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4351 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4352 struct inode_fs_paths *ipath = NULL;
4353 struct btrfs_path *path;
4355 if (!capable(CAP_DAC_READ_SEARCH))
4358 path = btrfs_alloc_path();
4364 ipa = memdup_user(arg, sizeof(*ipa));
4371 size = min_t(u32, ipa->size, 4096);
4372 ipath = init_ipath(size, root, path);
4373 if (IS_ERR(ipath)) {
4374 ret = PTR_ERR(ipath);
4379 ret = paths_from_inode(ipa->inum, ipath);
4383 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4384 rel_ptr = ipath->fspath->val[i] -
4385 (u64)(unsigned long)ipath->fspath->val;
4386 ipath->fspath->val[i] = rel_ptr;
4389 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
4390 ipath->fspath, size);
4397 btrfs_free_path(path);
4404 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4406 struct btrfs_data_container *inodes = ctx;
4407 const size_t c = 3 * sizeof(u64);
4409 if (inodes->bytes_left >= c) {
4410 inodes->bytes_left -= c;
4411 inodes->val[inodes->elem_cnt] = inum;
4412 inodes->val[inodes->elem_cnt + 1] = offset;
4413 inodes->val[inodes->elem_cnt + 2] = root;
4414 inodes->elem_cnt += 3;
4416 inodes->bytes_missing += c - inodes->bytes_left;
4417 inodes->bytes_left = 0;
4418 inodes->elem_missed += 3;
4424 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
4425 void __user *arg, int version)
4429 struct btrfs_ioctl_logical_ino_args *loi;
4430 struct btrfs_data_container *inodes = NULL;
4431 struct btrfs_path *path = NULL;
4434 if (!capable(CAP_SYS_ADMIN))
4437 loi = memdup_user(arg, sizeof(*loi));
4439 return PTR_ERR(loi);
4442 ignore_offset = false;
4443 size = min_t(u32, loi->size, SZ_64K);
4445 /* All reserved bits must be 0 for now */
4446 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
4450 /* Only accept flags we have defined so far */
4451 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
4455 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
4456 size = min_t(u32, loi->size, SZ_16M);
4459 path = btrfs_alloc_path();
4465 inodes = init_data_container(size);
4466 if (IS_ERR(inodes)) {
4467 ret = PTR_ERR(inodes);
4472 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
4473 build_ino_list, inodes, ignore_offset);
4479 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
4485 btrfs_free_path(path);
4493 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4494 struct btrfs_ioctl_balance_args *bargs)
4496 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4498 bargs->flags = bctl->flags;
4500 if (atomic_read(&fs_info->balance_running))
4501 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4502 if (atomic_read(&fs_info->balance_pause_req))
4503 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4504 if (atomic_read(&fs_info->balance_cancel_req))
4505 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4507 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4508 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4509 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4512 spin_lock(&fs_info->balance_lock);
4513 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4514 spin_unlock(&fs_info->balance_lock);
4516 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4520 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4522 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4523 struct btrfs_fs_info *fs_info = root->fs_info;
4524 struct btrfs_ioctl_balance_args *bargs;
4525 struct btrfs_balance_control *bctl;
4526 bool need_unlock; /* for mut. excl. ops lock */
4529 if (!capable(CAP_SYS_ADMIN))
4532 ret = mnt_want_write_file(file);
4537 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4538 mutex_lock(&fs_info->volume_mutex);
4539 mutex_lock(&fs_info->balance_mutex);
4545 * mut. excl. ops lock is locked. Three possibilities:
4546 * (1) some other op is running
4547 * (2) balance is running
4548 * (3) balance is paused -- special case (think resume)
4550 mutex_lock(&fs_info->balance_mutex);
4551 if (fs_info->balance_ctl) {
4552 /* this is either (2) or (3) */
4553 if (!atomic_read(&fs_info->balance_running)) {
4554 mutex_unlock(&fs_info->balance_mutex);
4555 if (!mutex_trylock(&fs_info->volume_mutex))
4557 mutex_lock(&fs_info->balance_mutex);
4559 if (fs_info->balance_ctl &&
4560 !atomic_read(&fs_info->balance_running)) {
4562 need_unlock = false;
4566 mutex_unlock(&fs_info->balance_mutex);
4567 mutex_unlock(&fs_info->volume_mutex);
4571 mutex_unlock(&fs_info->balance_mutex);
4577 mutex_unlock(&fs_info->balance_mutex);
4578 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4583 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
4586 bargs = memdup_user(arg, sizeof(*bargs));
4587 if (IS_ERR(bargs)) {
4588 ret = PTR_ERR(bargs);
4592 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4593 if (!fs_info->balance_ctl) {
4598 bctl = fs_info->balance_ctl;
4599 spin_lock(&fs_info->balance_lock);
4600 bctl->flags |= BTRFS_BALANCE_RESUME;
4601 spin_unlock(&fs_info->balance_lock);
4609 if (fs_info->balance_ctl) {
4614 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4620 bctl->fs_info = fs_info;
4622 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4623 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4624 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4626 bctl->flags = bargs->flags;
4628 /* balance everything - no filters */
4629 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4632 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4639 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP
4640 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4641 * or, if restriper was paused all the way until unmount, in
4642 * free_fs_info. The flag is cleared in __cancel_balance.
4644 need_unlock = false;
4646 ret = btrfs_balance(bctl, bargs);
4650 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4659 mutex_unlock(&fs_info->balance_mutex);
4660 mutex_unlock(&fs_info->volume_mutex);
4662 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4664 mnt_drop_write_file(file);
4668 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4670 if (!capable(CAP_SYS_ADMIN))
4674 case BTRFS_BALANCE_CTL_PAUSE:
4675 return btrfs_pause_balance(fs_info);
4676 case BTRFS_BALANCE_CTL_CANCEL:
4677 return btrfs_cancel_balance(fs_info);
4683 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4686 struct btrfs_ioctl_balance_args *bargs;
4689 if (!capable(CAP_SYS_ADMIN))
4692 mutex_lock(&fs_info->balance_mutex);
4693 if (!fs_info->balance_ctl) {
4698 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4704 update_ioctl_balance_args(fs_info, 1, bargs);
4706 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4711 mutex_unlock(&fs_info->balance_mutex);
4715 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4717 struct inode *inode = file_inode(file);
4718 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4719 struct btrfs_ioctl_quota_ctl_args *sa;
4720 struct btrfs_trans_handle *trans = NULL;
4724 if (!capable(CAP_SYS_ADMIN))
4727 ret = mnt_want_write_file(file);
4731 sa = memdup_user(arg, sizeof(*sa));
4737 down_write(&fs_info->subvol_sem);
4738 trans = btrfs_start_transaction(fs_info->tree_root, 2);
4739 if (IS_ERR(trans)) {
4740 ret = PTR_ERR(trans);
4745 case BTRFS_QUOTA_CTL_ENABLE:
4746 ret = btrfs_quota_enable(trans, fs_info);
4748 case BTRFS_QUOTA_CTL_DISABLE:
4749 ret = btrfs_quota_disable(trans, fs_info);
4756 err = btrfs_commit_transaction(trans);
4761 up_write(&fs_info->subvol_sem);
4763 mnt_drop_write_file(file);
4767 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4769 struct inode *inode = file_inode(file);
4770 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4771 struct btrfs_root *root = BTRFS_I(inode)->root;
4772 struct btrfs_ioctl_qgroup_assign_args *sa;
4773 struct btrfs_trans_handle *trans;
4777 if (!capable(CAP_SYS_ADMIN))
4780 ret = mnt_want_write_file(file);
4784 sa = memdup_user(arg, sizeof(*sa));
4790 trans = btrfs_join_transaction(root);
4791 if (IS_ERR(trans)) {
4792 ret = PTR_ERR(trans);
4797 ret = btrfs_add_qgroup_relation(trans, fs_info,
4800 ret = btrfs_del_qgroup_relation(trans, fs_info,
4804 /* update qgroup status and info */
4805 err = btrfs_run_qgroups(trans, fs_info);
4807 btrfs_handle_fs_error(fs_info, err,
4808 "failed to update qgroup status and info");
4809 err = btrfs_end_transaction(trans);
4816 mnt_drop_write_file(file);
4820 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4822 struct inode *inode = file_inode(file);
4823 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4824 struct btrfs_root *root = BTRFS_I(inode)->root;
4825 struct btrfs_ioctl_qgroup_create_args *sa;
4826 struct btrfs_trans_handle *trans;
4830 if (!capable(CAP_SYS_ADMIN))
4833 ret = mnt_want_write_file(file);
4837 sa = memdup_user(arg, sizeof(*sa));
4843 if (!sa->qgroupid) {
4848 trans = btrfs_join_transaction(root);
4849 if (IS_ERR(trans)) {
4850 ret = PTR_ERR(trans);
4855 ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid);
4857 ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid);
4860 err = btrfs_end_transaction(trans);
4867 mnt_drop_write_file(file);
4871 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4873 struct inode *inode = file_inode(file);
4874 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4875 struct btrfs_root *root = BTRFS_I(inode)->root;
4876 struct btrfs_ioctl_qgroup_limit_args *sa;
4877 struct btrfs_trans_handle *trans;
4882 if (!capable(CAP_SYS_ADMIN))
4885 ret = mnt_want_write_file(file);
4889 sa = memdup_user(arg, sizeof(*sa));
4895 trans = btrfs_join_transaction(root);
4896 if (IS_ERR(trans)) {
4897 ret = PTR_ERR(trans);
4901 qgroupid = sa->qgroupid;
4903 /* take the current subvol as qgroup */
4904 qgroupid = root->root_key.objectid;
4907 ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim);
4909 err = btrfs_end_transaction(trans);
4916 mnt_drop_write_file(file);
4920 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4922 struct inode *inode = file_inode(file);
4923 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4924 struct btrfs_ioctl_quota_rescan_args *qsa;
4927 if (!capable(CAP_SYS_ADMIN))
4930 ret = mnt_want_write_file(file);
4934 qsa = memdup_user(arg, sizeof(*qsa));
4945 ret = btrfs_qgroup_rescan(fs_info);
4950 mnt_drop_write_file(file);
4954 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4956 struct inode *inode = file_inode(file);
4957 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4958 struct btrfs_ioctl_quota_rescan_args *qsa;
4961 if (!capable(CAP_SYS_ADMIN))
4964 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
4968 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4970 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
4973 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4980 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4982 struct inode *inode = file_inode(file);
4983 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4985 if (!capable(CAP_SYS_ADMIN))
4988 return btrfs_qgroup_wait_for_completion(fs_info, true);
4991 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4992 struct btrfs_ioctl_received_subvol_args *sa)
4994 struct inode *inode = file_inode(file);
4995 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4996 struct btrfs_root *root = BTRFS_I(inode)->root;
4997 struct btrfs_root_item *root_item = &root->root_item;
4998 struct btrfs_trans_handle *trans;
4999 struct timespec ct = current_time(inode);
5001 int received_uuid_changed;
5003 if (!inode_owner_or_capable(inode))
5006 ret = mnt_want_write_file(file);
5010 down_write(&fs_info->subvol_sem);
5012 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
5017 if (btrfs_root_readonly(root)) {
5024 * 2 - uuid items (received uuid + subvol uuid)
5026 trans = btrfs_start_transaction(root, 3);
5027 if (IS_ERR(trans)) {
5028 ret = PTR_ERR(trans);
5033 sa->rtransid = trans->transid;
5034 sa->rtime.sec = ct.tv_sec;
5035 sa->rtime.nsec = ct.tv_nsec;
5037 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5039 if (received_uuid_changed &&
5040 !btrfs_is_empty_uuid(root_item->received_uuid)) {
5041 ret = btrfs_uuid_tree_rem(trans, fs_info,
5042 root_item->received_uuid,
5043 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5044 root->root_key.objectid);
5045 if (ret && ret != -ENOENT) {
5046 btrfs_abort_transaction(trans, ret);
5047 btrfs_end_transaction(trans);
5051 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5052 btrfs_set_root_stransid(root_item, sa->stransid);
5053 btrfs_set_root_rtransid(root_item, sa->rtransid);
5054 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5055 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5056 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5057 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5059 ret = btrfs_update_root(trans, fs_info->tree_root,
5060 &root->root_key, &root->root_item);
5062 btrfs_end_transaction(trans);
5065 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5066 ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid,
5067 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5068 root->root_key.objectid);
5069 if (ret < 0 && ret != -EEXIST) {
5070 btrfs_abort_transaction(trans, ret);
5071 btrfs_end_transaction(trans);
5075 ret = btrfs_commit_transaction(trans);
5077 up_write(&fs_info->subvol_sem);
5078 mnt_drop_write_file(file);
5083 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5086 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5087 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5090 args32 = memdup_user(arg, sizeof(*args32));
5092 return PTR_ERR(args32);
5094 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5100 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5101 args64->stransid = args32->stransid;
5102 args64->rtransid = args32->rtransid;
5103 args64->stime.sec = args32->stime.sec;
5104 args64->stime.nsec = args32->stime.nsec;
5105 args64->rtime.sec = args32->rtime.sec;
5106 args64->rtime.nsec = args32->rtime.nsec;
5107 args64->flags = args32->flags;
5109 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5113 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5114 args32->stransid = args64->stransid;
5115 args32->rtransid = args64->rtransid;
5116 args32->stime.sec = args64->stime.sec;
5117 args32->stime.nsec = args64->stime.nsec;
5118 args32->rtime.sec = args64->rtime.sec;
5119 args32->rtime.nsec = args64->rtime.nsec;
5120 args32->flags = args64->flags;
5122 ret = copy_to_user(arg, args32, sizeof(*args32));
5133 static long btrfs_ioctl_set_received_subvol(struct file *file,
5136 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5139 sa = memdup_user(arg, sizeof(*sa));
5143 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5148 ret = copy_to_user(arg, sa, sizeof(*sa));
5157 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5159 struct inode *inode = file_inode(file);
5160 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5163 char label[BTRFS_LABEL_SIZE];
5165 spin_lock(&fs_info->super_lock);
5166 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5167 spin_unlock(&fs_info->super_lock);
5169 len = strnlen(label, BTRFS_LABEL_SIZE);
5171 if (len == BTRFS_LABEL_SIZE) {
5173 "label is too long, return the first %zu bytes",
5177 ret = copy_to_user(arg, label, len);
5179 return ret ? -EFAULT : 0;
5182 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5184 struct inode *inode = file_inode(file);
5185 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5186 struct btrfs_root *root = BTRFS_I(inode)->root;
5187 struct btrfs_super_block *super_block = fs_info->super_copy;
5188 struct btrfs_trans_handle *trans;
5189 char label[BTRFS_LABEL_SIZE];
5192 if (!capable(CAP_SYS_ADMIN))
5195 if (copy_from_user(label, arg, sizeof(label)))
5198 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5200 "unable to set label with more than %d bytes",
5201 BTRFS_LABEL_SIZE - 1);
5205 ret = mnt_want_write_file(file);
5209 trans = btrfs_start_transaction(root, 0);
5210 if (IS_ERR(trans)) {
5211 ret = PTR_ERR(trans);
5215 spin_lock(&fs_info->super_lock);
5216 strcpy(super_block->label, label);
5217 spin_unlock(&fs_info->super_lock);
5218 ret = btrfs_commit_transaction(trans);
5221 mnt_drop_write_file(file);
5225 #define INIT_FEATURE_FLAGS(suffix) \
5226 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5227 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5228 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5230 int btrfs_ioctl_get_supported_features(void __user *arg)
5232 static const struct btrfs_ioctl_feature_flags features[3] = {
5233 INIT_FEATURE_FLAGS(SUPP),
5234 INIT_FEATURE_FLAGS(SAFE_SET),
5235 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5238 if (copy_to_user(arg, &features, sizeof(features)))
5244 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5246 struct inode *inode = file_inode(file);
5247 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5248 struct btrfs_super_block *super_block = fs_info->super_copy;
5249 struct btrfs_ioctl_feature_flags features;
5251 features.compat_flags = btrfs_super_compat_flags(super_block);
5252 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5253 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5255 if (copy_to_user(arg, &features, sizeof(features)))
5261 static int check_feature_bits(struct btrfs_fs_info *fs_info,
5262 enum btrfs_feature_set set,
5263 u64 change_mask, u64 flags, u64 supported_flags,
5264 u64 safe_set, u64 safe_clear)
5266 const char *type = btrfs_feature_set_names[set];
5268 u64 disallowed, unsupported;
5269 u64 set_mask = flags & change_mask;
5270 u64 clear_mask = ~flags & change_mask;
5272 unsupported = set_mask & ~supported_flags;
5274 names = btrfs_printable_features(set, unsupported);
5277 "this kernel does not support the %s feature bit%s",
5278 names, strchr(names, ',') ? "s" : "");
5282 "this kernel does not support %s bits 0x%llx",
5287 disallowed = set_mask & ~safe_set;
5289 names = btrfs_printable_features(set, disallowed);
5292 "can't set the %s feature bit%s while mounted",
5293 names, strchr(names, ',') ? "s" : "");
5297 "can't set %s bits 0x%llx while mounted",
5302 disallowed = clear_mask & ~safe_clear;
5304 names = btrfs_printable_features(set, disallowed);
5307 "can't clear the %s feature bit%s while mounted",
5308 names, strchr(names, ',') ? "s" : "");
5312 "can't clear %s bits 0x%llx while mounted",
5320 #define check_feature(fs_info, change_mask, flags, mask_base) \
5321 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5322 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5323 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5324 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5326 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5328 struct inode *inode = file_inode(file);
5329 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5330 struct btrfs_root *root = BTRFS_I(inode)->root;
5331 struct btrfs_super_block *super_block = fs_info->super_copy;
5332 struct btrfs_ioctl_feature_flags flags[2];
5333 struct btrfs_trans_handle *trans;
5337 if (!capable(CAP_SYS_ADMIN))
5340 if (copy_from_user(flags, arg, sizeof(flags)))
5344 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5345 !flags[0].incompat_flags)
5348 ret = check_feature(fs_info, flags[0].compat_flags,
5349 flags[1].compat_flags, COMPAT);
5353 ret = check_feature(fs_info, flags[0].compat_ro_flags,
5354 flags[1].compat_ro_flags, COMPAT_RO);
5358 ret = check_feature(fs_info, flags[0].incompat_flags,
5359 flags[1].incompat_flags, INCOMPAT);
5363 ret = mnt_want_write_file(file);
5367 trans = btrfs_start_transaction(root, 0);
5368 if (IS_ERR(trans)) {
5369 ret = PTR_ERR(trans);
5370 goto out_drop_write;
5373 spin_lock(&fs_info->super_lock);
5374 newflags = btrfs_super_compat_flags(super_block);
5375 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5376 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5377 btrfs_set_super_compat_flags(super_block, newflags);
5379 newflags = btrfs_super_compat_ro_flags(super_block);
5380 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5381 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5382 btrfs_set_super_compat_ro_flags(super_block, newflags);
5384 newflags = btrfs_super_incompat_flags(super_block);
5385 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5386 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5387 btrfs_set_super_incompat_flags(super_block, newflags);
5388 spin_unlock(&fs_info->super_lock);
5390 ret = btrfs_commit_transaction(trans);
5392 mnt_drop_write_file(file);
5397 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
5399 struct btrfs_ioctl_send_args *arg;
5403 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5404 struct btrfs_ioctl_send_args_32 args32;
5406 ret = copy_from_user(&args32, argp, sizeof(args32));
5409 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
5412 arg->send_fd = args32.send_fd;
5413 arg->clone_sources_count = args32.clone_sources_count;
5414 arg->clone_sources = compat_ptr(args32.clone_sources);
5415 arg->parent_root = args32.parent_root;
5416 arg->flags = args32.flags;
5417 memcpy(arg->reserved, args32.reserved,
5418 sizeof(args32.reserved));
5423 arg = memdup_user(argp, sizeof(*arg));
5425 return PTR_ERR(arg);
5427 ret = btrfs_ioctl_send(file, arg);
5432 long btrfs_ioctl(struct file *file, unsigned int
5433 cmd, unsigned long arg)
5435 struct inode *inode = file_inode(file);
5436 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5437 struct btrfs_root *root = BTRFS_I(inode)->root;
5438 void __user *argp = (void __user *)arg;
5441 case FS_IOC_GETFLAGS:
5442 return btrfs_ioctl_getflags(file, argp);
5443 case FS_IOC_SETFLAGS:
5444 return btrfs_ioctl_setflags(file, argp);
5445 case FS_IOC_GETVERSION:
5446 return btrfs_ioctl_getversion(file, argp);
5448 return btrfs_ioctl_fitrim(file, argp);
5449 case BTRFS_IOC_SNAP_CREATE:
5450 return btrfs_ioctl_snap_create(file, argp, 0);
5451 case BTRFS_IOC_SNAP_CREATE_V2:
5452 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5453 case BTRFS_IOC_SUBVOL_CREATE:
5454 return btrfs_ioctl_snap_create(file, argp, 1);
5455 case BTRFS_IOC_SUBVOL_CREATE_V2:
5456 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5457 case BTRFS_IOC_SNAP_DESTROY:
5458 return btrfs_ioctl_snap_destroy(file, argp);
5459 case BTRFS_IOC_SUBVOL_GETFLAGS:
5460 return btrfs_ioctl_subvol_getflags(file, argp);
5461 case BTRFS_IOC_SUBVOL_SETFLAGS:
5462 return btrfs_ioctl_subvol_setflags(file, argp);
5463 case BTRFS_IOC_DEFAULT_SUBVOL:
5464 return btrfs_ioctl_default_subvol(file, argp);
5465 case BTRFS_IOC_DEFRAG:
5466 return btrfs_ioctl_defrag(file, NULL);
5467 case BTRFS_IOC_DEFRAG_RANGE:
5468 return btrfs_ioctl_defrag(file, argp);
5469 case BTRFS_IOC_RESIZE:
5470 return btrfs_ioctl_resize(file, argp);
5471 case BTRFS_IOC_ADD_DEV:
5472 return btrfs_ioctl_add_dev(fs_info, argp);
5473 case BTRFS_IOC_RM_DEV:
5474 return btrfs_ioctl_rm_dev(file, argp);
5475 case BTRFS_IOC_RM_DEV_V2:
5476 return btrfs_ioctl_rm_dev_v2(file, argp);
5477 case BTRFS_IOC_FS_INFO:
5478 return btrfs_ioctl_fs_info(fs_info, argp);
5479 case BTRFS_IOC_DEV_INFO:
5480 return btrfs_ioctl_dev_info(fs_info, argp);
5481 case BTRFS_IOC_BALANCE:
5482 return btrfs_ioctl_balance(file, NULL);
5483 case BTRFS_IOC_TREE_SEARCH:
5484 return btrfs_ioctl_tree_search(file, argp);
5485 case BTRFS_IOC_TREE_SEARCH_V2:
5486 return btrfs_ioctl_tree_search_v2(file, argp);
5487 case BTRFS_IOC_INO_LOOKUP:
5488 return btrfs_ioctl_ino_lookup(file, argp);
5489 case BTRFS_IOC_INO_PATHS:
5490 return btrfs_ioctl_ino_to_path(root, argp);
5491 case BTRFS_IOC_LOGICAL_INO:
5492 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
5493 case BTRFS_IOC_LOGICAL_INO_V2:
5494 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
5495 case BTRFS_IOC_SPACE_INFO:
5496 return btrfs_ioctl_space_info(fs_info, argp);
5497 case BTRFS_IOC_SYNC: {
5500 ret = btrfs_start_delalloc_roots(fs_info, 0, -1);
5503 ret = btrfs_sync_fs(inode->i_sb, 1);
5505 * The transaction thread may want to do more work,
5506 * namely it pokes the cleaner kthread that will start
5507 * processing uncleaned subvols.
5509 wake_up_process(fs_info->transaction_kthread);
5512 case BTRFS_IOC_START_SYNC:
5513 return btrfs_ioctl_start_sync(root, argp);
5514 case BTRFS_IOC_WAIT_SYNC:
5515 return btrfs_ioctl_wait_sync(fs_info, argp);
5516 case BTRFS_IOC_SCRUB:
5517 return btrfs_ioctl_scrub(file, argp);
5518 case BTRFS_IOC_SCRUB_CANCEL:
5519 return btrfs_ioctl_scrub_cancel(fs_info);
5520 case BTRFS_IOC_SCRUB_PROGRESS:
5521 return btrfs_ioctl_scrub_progress(fs_info, argp);
5522 case BTRFS_IOC_BALANCE_V2:
5523 return btrfs_ioctl_balance(file, argp);
5524 case BTRFS_IOC_BALANCE_CTL:
5525 return btrfs_ioctl_balance_ctl(fs_info, arg);
5526 case BTRFS_IOC_BALANCE_PROGRESS:
5527 return btrfs_ioctl_balance_progress(fs_info, argp);
5528 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5529 return btrfs_ioctl_set_received_subvol(file, argp);
5531 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5532 return btrfs_ioctl_set_received_subvol_32(file, argp);
5534 case BTRFS_IOC_SEND:
5535 return _btrfs_ioctl_send(file, argp, false);
5536 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5537 case BTRFS_IOC_SEND_32:
5538 return _btrfs_ioctl_send(file, argp, true);
5540 case BTRFS_IOC_GET_DEV_STATS:
5541 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5542 case BTRFS_IOC_QUOTA_CTL:
5543 return btrfs_ioctl_quota_ctl(file, argp);
5544 case BTRFS_IOC_QGROUP_ASSIGN:
5545 return btrfs_ioctl_qgroup_assign(file, argp);
5546 case BTRFS_IOC_QGROUP_CREATE:
5547 return btrfs_ioctl_qgroup_create(file, argp);
5548 case BTRFS_IOC_QGROUP_LIMIT:
5549 return btrfs_ioctl_qgroup_limit(file, argp);
5550 case BTRFS_IOC_QUOTA_RESCAN:
5551 return btrfs_ioctl_quota_rescan(file, argp);
5552 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5553 return btrfs_ioctl_quota_rescan_status(file, argp);
5554 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5555 return btrfs_ioctl_quota_rescan_wait(file, argp);
5556 case BTRFS_IOC_DEV_REPLACE:
5557 return btrfs_ioctl_dev_replace(fs_info, argp);
5558 case BTRFS_IOC_GET_FSLABEL:
5559 return btrfs_ioctl_get_fslabel(file, argp);
5560 case BTRFS_IOC_SET_FSLABEL:
5561 return btrfs_ioctl_set_fslabel(file, argp);
5562 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5563 return btrfs_ioctl_get_supported_features(argp);
5564 case BTRFS_IOC_GET_FEATURES:
5565 return btrfs_ioctl_get_features(file, argp);
5566 case BTRFS_IOC_SET_FEATURES:
5567 return btrfs_ioctl_set_features(file, argp);
5573 #ifdef CONFIG_COMPAT
5574 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5577 * These all access 32-bit values anyway so no further
5578 * handling is necessary.
5581 case FS_IOC32_GETFLAGS:
5582 cmd = FS_IOC_GETFLAGS;
5584 case FS_IOC32_SETFLAGS:
5585 cmd = FS_IOC_SETFLAGS;
5587 case FS_IOC32_GETVERSION:
5588 cmd = FS_IOC_GETVERSION;
5592 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));