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);
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 vol_args = memdup_user(arg, sizeof(*vol_args));
1461 if (IS_ERR(vol_args)) {
1462 ret = PTR_ERR(vol_args);
1466 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1468 sizestr = vol_args->name;
1469 devstr = strchr(sizestr, ':');
1471 sizestr = devstr + 1;
1473 devstr = vol_args->name;
1474 ret = kstrtoull(devstr, 10, &devid);
1481 btrfs_info(fs_info, "resizing devid %llu", devid);
1484 device = btrfs_find_device(fs_info, devid, NULL, NULL);
1486 btrfs_info(fs_info, "resizer unable to find device %llu",
1492 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1494 "resizer unable to apply on readonly device %llu",
1500 if (!strcmp(sizestr, "max"))
1501 new_size = device->bdev->bd_inode->i_size;
1503 if (sizestr[0] == '-') {
1506 } else if (sizestr[0] == '+') {
1510 new_size = memparse(sizestr, &retptr);
1511 if (*retptr != '\0' || new_size == 0) {
1517 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1522 old_size = btrfs_device_get_total_bytes(device);
1525 if (new_size > old_size) {
1529 new_size = old_size - new_size;
1530 } else if (mod > 0) {
1531 if (new_size > ULLONG_MAX - old_size) {
1535 new_size = old_size + new_size;
1538 if (new_size < SZ_256M) {
1542 if (new_size > device->bdev->bd_inode->i_size) {
1547 new_size = round_down(new_size, fs_info->sectorsize);
1549 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1550 rcu_str_deref(device->name), new_size);
1552 if (new_size > old_size) {
1553 trans = btrfs_start_transaction(root, 0);
1554 if (IS_ERR(trans)) {
1555 ret = PTR_ERR(trans);
1558 ret = btrfs_grow_device(trans, device, new_size);
1559 btrfs_commit_transaction(trans);
1560 } else if (new_size < old_size) {
1561 ret = btrfs_shrink_device(device, new_size);
1562 } /* equal, nothing need to do */
1567 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
1568 mnt_drop_write_file(file);
1572 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1573 const char *name, unsigned long fd, int subvol,
1574 u64 *transid, bool readonly,
1575 struct btrfs_qgroup_inherit *inherit)
1580 if (!S_ISDIR(file_inode(file)->i_mode))
1583 ret = mnt_want_write_file(file);
1587 namelen = strlen(name);
1588 if (strchr(name, '/')) {
1590 goto out_drop_write;
1593 if (name[0] == '.' &&
1594 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1596 goto out_drop_write;
1600 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1601 NULL, transid, readonly, inherit);
1603 struct fd src = fdget(fd);
1604 struct inode *src_inode;
1607 goto out_drop_write;
1610 src_inode = file_inode(src.file);
1611 if (src_inode->i_sb != file_inode(file)->i_sb) {
1612 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1613 "Snapshot src from another FS");
1615 } else if (!inode_owner_or_capable(src_inode)) {
1617 * Subvolume creation is not restricted, but snapshots
1618 * are limited to own subvolumes only
1622 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1623 BTRFS_I(src_inode)->root,
1624 transid, readonly, inherit);
1629 mnt_drop_write_file(file);
1634 static noinline int btrfs_ioctl_snap_create(struct file *file,
1635 void __user *arg, int subvol)
1637 struct btrfs_ioctl_vol_args *vol_args;
1640 if (!S_ISDIR(file_inode(file)->i_mode))
1643 vol_args = memdup_user(arg, sizeof(*vol_args));
1644 if (IS_ERR(vol_args))
1645 return PTR_ERR(vol_args);
1646 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1648 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1649 vol_args->fd, subvol,
1656 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1657 void __user *arg, int subvol)
1659 struct btrfs_ioctl_vol_args_v2 *vol_args;
1663 bool readonly = false;
1664 struct btrfs_qgroup_inherit *inherit = NULL;
1666 if (!S_ISDIR(file_inode(file)->i_mode))
1669 vol_args = memdup_user(arg, sizeof(*vol_args));
1670 if (IS_ERR(vol_args))
1671 return PTR_ERR(vol_args);
1672 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1674 if (vol_args->flags &
1675 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1676 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1681 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1683 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1685 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1686 if (vol_args->size > PAGE_SIZE) {
1690 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1691 if (IS_ERR(inherit)) {
1692 ret = PTR_ERR(inherit);
1697 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1698 vol_args->fd, subvol, ptr,
1703 if (ptr && copy_to_user(arg +
1704 offsetof(struct btrfs_ioctl_vol_args_v2,
1716 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1719 struct inode *inode = file_inode(file);
1720 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1721 struct btrfs_root *root = BTRFS_I(inode)->root;
1725 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1728 down_read(&fs_info->subvol_sem);
1729 if (btrfs_root_readonly(root))
1730 flags |= BTRFS_SUBVOL_RDONLY;
1731 up_read(&fs_info->subvol_sem);
1733 if (copy_to_user(arg, &flags, sizeof(flags)))
1739 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1742 struct inode *inode = file_inode(file);
1743 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1744 struct btrfs_root *root = BTRFS_I(inode)->root;
1745 struct btrfs_trans_handle *trans;
1750 if (!inode_owner_or_capable(inode))
1753 ret = mnt_want_write_file(file);
1757 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1759 goto out_drop_write;
1762 if (copy_from_user(&flags, arg, sizeof(flags))) {
1764 goto out_drop_write;
1767 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1769 goto out_drop_write;
1772 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1774 goto out_drop_write;
1777 down_write(&fs_info->subvol_sem);
1780 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1783 root_flags = btrfs_root_flags(&root->root_item);
1784 if (flags & BTRFS_SUBVOL_RDONLY) {
1785 btrfs_set_root_flags(&root->root_item,
1786 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1789 * Block RO -> RW transition if this subvolume is involved in
1792 spin_lock(&root->root_item_lock);
1793 if (root->send_in_progress == 0) {
1794 btrfs_set_root_flags(&root->root_item,
1795 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1796 spin_unlock(&root->root_item_lock);
1798 spin_unlock(&root->root_item_lock);
1800 "Attempt to set subvolume %llu read-write during send",
1801 root->root_key.objectid);
1807 trans = btrfs_start_transaction(root, 1);
1808 if (IS_ERR(trans)) {
1809 ret = PTR_ERR(trans);
1813 ret = btrfs_update_root(trans, fs_info->tree_root,
1814 &root->root_key, &root->root_item);
1816 btrfs_end_transaction(trans);
1820 ret = btrfs_commit_transaction(trans);
1824 btrfs_set_root_flags(&root->root_item, root_flags);
1826 up_write(&fs_info->subvol_sem);
1828 mnt_drop_write_file(file);
1833 static noinline int key_in_sk(struct btrfs_key *key,
1834 struct btrfs_ioctl_search_key *sk)
1836 struct btrfs_key test;
1839 test.objectid = sk->min_objectid;
1840 test.type = sk->min_type;
1841 test.offset = sk->min_offset;
1843 ret = btrfs_comp_cpu_keys(key, &test);
1847 test.objectid = sk->max_objectid;
1848 test.type = sk->max_type;
1849 test.offset = sk->max_offset;
1851 ret = btrfs_comp_cpu_keys(key, &test);
1857 static noinline int copy_to_sk(struct btrfs_path *path,
1858 struct btrfs_key *key,
1859 struct btrfs_ioctl_search_key *sk,
1862 unsigned long *sk_offset,
1866 struct extent_buffer *leaf;
1867 struct btrfs_ioctl_search_header sh;
1868 struct btrfs_key test;
1869 unsigned long item_off;
1870 unsigned long item_len;
1876 leaf = path->nodes[0];
1877 slot = path->slots[0];
1878 nritems = btrfs_header_nritems(leaf);
1880 if (btrfs_header_generation(leaf) > sk->max_transid) {
1884 found_transid = btrfs_header_generation(leaf);
1886 for (i = slot; i < nritems; i++) {
1887 item_off = btrfs_item_ptr_offset(leaf, i);
1888 item_len = btrfs_item_size_nr(leaf, i);
1890 btrfs_item_key_to_cpu(leaf, key, i);
1891 if (!key_in_sk(key, sk))
1894 if (sizeof(sh) + item_len > *buf_size) {
1901 * return one empty item back for v1, which does not
1905 *buf_size = sizeof(sh) + item_len;
1910 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1915 sh.objectid = key->objectid;
1916 sh.offset = key->offset;
1917 sh.type = key->type;
1919 sh.transid = found_transid;
1921 /* copy search result header */
1922 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1927 *sk_offset += sizeof(sh);
1930 char __user *up = ubuf + *sk_offset;
1932 if (read_extent_buffer_to_user(leaf, up,
1933 item_off, item_len)) {
1938 *sk_offset += item_len;
1942 if (ret) /* -EOVERFLOW from above */
1945 if (*num_found >= sk->nr_items) {
1952 test.objectid = sk->max_objectid;
1953 test.type = sk->max_type;
1954 test.offset = sk->max_offset;
1955 if (btrfs_comp_cpu_keys(key, &test) >= 0)
1957 else if (key->offset < (u64)-1)
1959 else if (key->type < (u8)-1) {
1962 } else if (key->objectid < (u64)-1) {
1970 * 0: all items from this leaf copied, continue with next
1971 * 1: * more items can be copied, but unused buffer is too small
1972 * * all items were found
1973 * Either way, it will stops the loop which iterates to the next
1975 * -EOVERFLOW: item was to large for buffer
1976 * -EFAULT: could not copy extent buffer back to userspace
1981 static noinline int search_ioctl(struct inode *inode,
1982 struct btrfs_ioctl_search_key *sk,
1986 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
1987 struct btrfs_root *root;
1988 struct btrfs_key key;
1989 struct btrfs_path *path;
1992 unsigned long sk_offset = 0;
1994 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
1995 *buf_size = sizeof(struct btrfs_ioctl_search_header);
1999 path = btrfs_alloc_path();
2003 if (sk->tree_id == 0) {
2004 /* search the root of the inode that was passed */
2005 root = BTRFS_I(inode)->root;
2007 key.objectid = sk->tree_id;
2008 key.type = BTRFS_ROOT_ITEM_KEY;
2009 key.offset = (u64)-1;
2010 root = btrfs_read_fs_root_no_name(info, &key);
2012 btrfs_free_path(path);
2017 key.objectid = sk->min_objectid;
2018 key.type = sk->min_type;
2019 key.offset = sk->min_offset;
2022 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2028 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2029 &sk_offset, &num_found);
2030 btrfs_release_path(path);
2038 sk->nr_items = num_found;
2039 btrfs_free_path(path);
2043 static noinline int btrfs_ioctl_tree_search(struct file *file,
2046 struct btrfs_ioctl_search_args __user *uargs;
2047 struct btrfs_ioctl_search_key sk;
2048 struct inode *inode;
2052 if (!capable(CAP_SYS_ADMIN))
2055 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2057 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2060 buf_size = sizeof(uargs->buf);
2062 inode = file_inode(file);
2063 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2066 * In the origin implementation an overflow is handled by returning a
2067 * search header with a len of zero, so reset ret.
2069 if (ret == -EOVERFLOW)
2072 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2077 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2080 struct btrfs_ioctl_search_args_v2 __user *uarg;
2081 struct btrfs_ioctl_search_args_v2 args;
2082 struct inode *inode;
2085 const size_t buf_limit = SZ_16M;
2087 if (!capable(CAP_SYS_ADMIN))
2090 /* copy search header and buffer size */
2091 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2092 if (copy_from_user(&args, uarg, sizeof(args)))
2095 buf_size = args.buf_size;
2097 /* limit result size to 16MB */
2098 if (buf_size > buf_limit)
2099 buf_size = buf_limit;
2101 inode = file_inode(file);
2102 ret = search_ioctl(inode, &args.key, &buf_size,
2103 (char __user *)(&uarg->buf[0]));
2104 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2106 else if (ret == -EOVERFLOW &&
2107 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2114 * Search INODE_REFs to identify path name of 'dirid' directory
2115 * in a 'tree_id' tree. and sets path name to 'name'.
2117 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2118 u64 tree_id, u64 dirid, char *name)
2120 struct btrfs_root *root;
2121 struct btrfs_key key;
2127 struct btrfs_inode_ref *iref;
2128 struct extent_buffer *l;
2129 struct btrfs_path *path;
2131 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2136 path = btrfs_alloc_path();
2140 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2142 key.objectid = tree_id;
2143 key.type = BTRFS_ROOT_ITEM_KEY;
2144 key.offset = (u64)-1;
2145 root = btrfs_read_fs_root_no_name(info, &key);
2147 btrfs_err(info, "could not find root %llu", tree_id);
2152 key.objectid = dirid;
2153 key.type = BTRFS_INODE_REF_KEY;
2154 key.offset = (u64)-1;
2157 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2161 ret = btrfs_previous_item(root, path, dirid,
2162 BTRFS_INODE_REF_KEY);
2172 slot = path->slots[0];
2173 btrfs_item_key_to_cpu(l, &key, slot);
2175 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2176 len = btrfs_inode_ref_name_len(l, iref);
2178 total_len += len + 1;
2180 ret = -ENAMETOOLONG;
2185 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2187 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2190 btrfs_release_path(path);
2191 key.objectid = key.offset;
2192 key.offset = (u64)-1;
2193 dirid = key.objectid;
2195 memmove(name, ptr, total_len);
2196 name[total_len] = '\0';
2199 btrfs_free_path(path);
2203 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2206 struct btrfs_ioctl_ino_lookup_args *args;
2207 struct inode *inode;
2210 args = memdup_user(argp, sizeof(*args));
2212 return PTR_ERR(args);
2214 inode = file_inode(file);
2217 * Unprivileged query to obtain the containing subvolume root id. The
2218 * path is reset so it's consistent with btrfs_search_path_in_tree.
2220 if (args->treeid == 0)
2221 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2223 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2228 if (!capable(CAP_SYS_ADMIN)) {
2233 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2234 args->treeid, args->objectid,
2238 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2245 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2248 struct dentry *parent = file->f_path.dentry;
2249 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2250 struct dentry *dentry;
2251 struct inode *dir = d_inode(parent);
2252 struct inode *inode;
2253 struct btrfs_root *root = BTRFS_I(dir)->root;
2254 struct btrfs_root *dest = NULL;
2255 struct btrfs_ioctl_vol_args *vol_args;
2259 if (!S_ISDIR(dir->i_mode))
2262 vol_args = memdup_user(arg, sizeof(*vol_args));
2263 if (IS_ERR(vol_args))
2264 return PTR_ERR(vol_args);
2266 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2267 namelen = strlen(vol_args->name);
2268 if (strchr(vol_args->name, '/') ||
2269 strncmp(vol_args->name, "..", namelen) == 0) {
2274 err = mnt_want_write_file(file);
2279 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2281 goto out_drop_write;
2282 dentry = lookup_one_len(vol_args->name, parent, namelen);
2283 if (IS_ERR(dentry)) {
2284 err = PTR_ERR(dentry);
2285 goto out_unlock_dir;
2288 if (d_really_is_negative(dentry)) {
2293 inode = d_inode(dentry);
2294 dest = BTRFS_I(inode)->root;
2295 if (!capable(CAP_SYS_ADMIN)) {
2297 * Regular user. Only allow this with a special mount
2298 * option, when the user has write+exec access to the
2299 * subvol root, and when rmdir(2) would have been
2302 * Note that this is _not_ check that the subvol is
2303 * empty or doesn't contain data that we wouldn't
2304 * otherwise be able to delete.
2306 * Users who want to delete empty subvols should try
2310 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2314 * Do not allow deletion if the parent dir is the same
2315 * as the dir to be deleted. That means the ioctl
2316 * must be called on the dentry referencing the root
2317 * of the subvol, not a random directory contained
2324 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2329 /* check if subvolume may be deleted by a user */
2330 err = btrfs_may_delete(dir, dentry, 1);
2334 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2340 err = btrfs_delete_subvolume(dir, dentry);
2341 inode_unlock(inode);
2350 mnt_drop_write_file(file);
2356 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2358 struct inode *inode = file_inode(file);
2359 struct btrfs_root *root = BTRFS_I(inode)->root;
2360 struct btrfs_ioctl_defrag_range_args *range;
2363 ret = mnt_want_write_file(file);
2367 if (btrfs_root_readonly(root)) {
2372 switch (inode->i_mode & S_IFMT) {
2374 if (!capable(CAP_SYS_ADMIN)) {
2378 ret = btrfs_defrag_root(root);
2381 if (!(file->f_mode & FMODE_WRITE)) {
2386 range = kzalloc(sizeof(*range), GFP_KERNEL);
2393 if (copy_from_user(range, argp,
2399 /* compression requires us to start the IO */
2400 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2401 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2402 range->extent_thresh = (u32)-1;
2405 /* the rest are all set to zero by kzalloc */
2406 range->len = (u64)-1;
2408 ret = btrfs_defrag_file(file_inode(file), file,
2409 range, BTRFS_OLDEST_GENERATION, 0);
2418 mnt_drop_write_file(file);
2422 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2424 struct btrfs_ioctl_vol_args *vol_args;
2427 if (!capable(CAP_SYS_ADMIN))
2430 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
2431 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2433 vol_args = memdup_user(arg, sizeof(*vol_args));
2434 if (IS_ERR(vol_args)) {
2435 ret = PTR_ERR(vol_args);
2439 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2440 ret = btrfs_init_new_device(fs_info, vol_args->name);
2443 btrfs_info(fs_info, "disk added %s", vol_args->name);
2447 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2451 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2453 struct inode *inode = file_inode(file);
2454 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2455 struct btrfs_ioctl_vol_args_v2 *vol_args;
2458 if (!capable(CAP_SYS_ADMIN))
2461 ret = mnt_want_write_file(file);
2465 vol_args = memdup_user(arg, sizeof(*vol_args));
2466 if (IS_ERR(vol_args)) {
2467 ret = PTR_ERR(vol_args);
2471 /* Check for compatibility reject unknown flags */
2472 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED)
2475 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2476 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2480 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2481 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
2483 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2484 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2486 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2489 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2490 btrfs_info(fs_info, "device deleted: id %llu",
2493 btrfs_info(fs_info, "device deleted: %s",
2499 mnt_drop_write_file(file);
2503 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2505 struct inode *inode = file_inode(file);
2506 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2507 struct btrfs_ioctl_vol_args *vol_args;
2510 if (!capable(CAP_SYS_ADMIN))
2513 ret = mnt_want_write_file(file);
2517 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2518 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2519 goto out_drop_write;
2522 vol_args = memdup_user(arg, sizeof(*vol_args));
2523 if (IS_ERR(vol_args)) {
2524 ret = PTR_ERR(vol_args);
2528 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2529 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2532 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2535 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2537 mnt_drop_write_file(file);
2542 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2545 struct btrfs_ioctl_fs_info_args *fi_args;
2546 struct btrfs_device *device;
2547 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2550 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2555 fi_args->num_devices = fs_devices->num_devices;
2557 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2558 if (device->devid > fi_args->max_id)
2559 fi_args->max_id = device->devid;
2563 memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid));
2564 fi_args->nodesize = fs_info->nodesize;
2565 fi_args->sectorsize = fs_info->sectorsize;
2566 fi_args->clone_alignment = fs_info->sectorsize;
2568 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2575 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2578 struct btrfs_ioctl_dev_info_args *di_args;
2579 struct btrfs_device *dev;
2581 char *s_uuid = NULL;
2583 di_args = memdup_user(arg, sizeof(*di_args));
2584 if (IS_ERR(di_args))
2585 return PTR_ERR(di_args);
2587 if (!btrfs_is_empty_uuid(di_args->uuid))
2588 s_uuid = di_args->uuid;
2591 dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL);
2598 di_args->devid = dev->devid;
2599 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2600 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2601 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2603 struct rcu_string *name;
2605 name = rcu_dereference(dev->name);
2606 strncpy(di_args->path, name->str, sizeof(di_args->path) - 1);
2607 di_args->path[sizeof(di_args->path) - 1] = 0;
2609 di_args->path[0] = '\0';
2614 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2621 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2625 page = grab_cache_page(inode->i_mapping, index);
2627 return ERR_PTR(-ENOMEM);
2629 if (!PageUptodate(page)) {
2632 ret = btrfs_readpage(NULL, page);
2634 return ERR_PTR(ret);
2636 if (!PageUptodate(page)) {
2639 return ERR_PTR(-EIO);
2641 if (page->mapping != inode->i_mapping) {
2644 return ERR_PTR(-EAGAIN);
2651 static int gather_extent_pages(struct inode *inode, struct page **pages,
2652 int num_pages, u64 off)
2655 pgoff_t index = off >> PAGE_SHIFT;
2657 for (i = 0; i < num_pages; i++) {
2659 pages[i] = extent_same_get_page(inode, index + i);
2660 if (IS_ERR(pages[i])) {
2661 int err = PTR_ERR(pages[i]);
2672 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2673 bool retry_range_locking)
2676 * Do any pending delalloc/csum calculations on inode, one way or
2677 * another, and lock file content.
2678 * The locking order is:
2681 * 2) range in the inode's io tree
2684 struct btrfs_ordered_extent *ordered;
2685 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2686 ordered = btrfs_lookup_first_ordered_extent(inode,
2689 ordered->file_offset + ordered->len <= off ||
2690 ordered->file_offset >= off + len) &&
2691 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2692 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2694 btrfs_put_ordered_extent(ordered);
2697 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2699 btrfs_put_ordered_extent(ordered);
2700 if (!retry_range_locking)
2702 btrfs_wait_ordered_range(inode, off, len);
2707 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2709 inode_unlock(inode1);
2710 inode_unlock(inode2);
2713 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2715 if (inode1 < inode2)
2716 swap(inode1, inode2);
2718 inode_lock_nested(inode1, I_MUTEX_PARENT);
2719 inode_lock_nested(inode2, I_MUTEX_CHILD);
2722 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2723 struct inode *inode2, u64 loff2, u64 len)
2725 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2726 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2729 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2730 struct inode *inode2, u64 loff2, u64 len,
2731 bool retry_range_locking)
2735 if (inode1 < inode2) {
2736 swap(inode1, inode2);
2739 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2742 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2744 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2751 struct page **src_pages;
2752 struct page **dst_pages;
2755 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2760 for (i = 0; i < cmp->num_pages; i++) {
2761 pg = cmp->src_pages[i];
2766 pg = cmp->dst_pages[i];
2772 kfree(cmp->src_pages);
2773 kfree(cmp->dst_pages);
2776 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2777 struct inode *dst, u64 dst_loff,
2778 u64 len, struct cmp_pages *cmp)
2781 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
2782 struct page **src_pgarr, **dst_pgarr;
2785 * We must gather up all the pages before we initiate our
2786 * extent locking. We use an array for the page pointers. Size
2787 * of the array is bounded by len, which is in turn bounded by
2788 * BTRFS_MAX_DEDUPE_LEN.
2790 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2791 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2792 if (!src_pgarr || !dst_pgarr) {
2797 cmp->num_pages = num_pages;
2798 cmp->src_pages = src_pgarr;
2799 cmp->dst_pages = dst_pgarr;
2802 * If deduping ranges in the same inode, locking rules make it mandatory
2803 * to always lock pages in ascending order to avoid deadlocks with
2804 * concurrent tasks (such as starting writeback/delalloc).
2806 if (src == dst && dst_loff < loff) {
2807 swap(src_pgarr, dst_pgarr);
2808 swap(loff, dst_loff);
2811 ret = gather_extent_pages(src, src_pgarr, cmp->num_pages, loff);
2815 ret = gather_extent_pages(dst, dst_pgarr, cmp->num_pages, dst_loff);
2819 btrfs_cmp_data_free(cmp);
2823 static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp)
2827 struct page *src_page, *dst_page;
2828 unsigned int cmp_len = PAGE_SIZE;
2829 void *addr, *dst_addr;
2833 if (len < PAGE_SIZE)
2836 BUG_ON(i >= cmp->num_pages);
2838 src_page = cmp->src_pages[i];
2839 dst_page = cmp->dst_pages[i];
2840 ASSERT(PageLocked(src_page));
2841 ASSERT(PageLocked(dst_page));
2843 addr = kmap_atomic(src_page);
2844 dst_addr = kmap_atomic(dst_page);
2846 flush_dcache_page(src_page);
2847 flush_dcache_page(dst_page);
2849 if (memcmp(addr, dst_addr, cmp_len))
2852 kunmap_atomic(addr);
2853 kunmap_atomic(dst_addr);
2865 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
2869 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2871 if (off + olen > inode->i_size || off + olen < off)
2874 /* if we extend to eof, continue to block boundary */
2875 if (off + len == inode->i_size)
2876 *plen = len = ALIGN(inode->i_size, bs) - off;
2878 /* Check that we are block aligned - btrfs_clone() requires this */
2879 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2885 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
2886 struct inode *dst, u64 dst_loff)
2890 struct cmp_pages cmp;
2891 bool same_inode = (src == dst);
2892 u64 same_lock_start = 0;
2893 u64 same_lock_len = 0;
2901 btrfs_double_inode_lock(src, dst);
2903 ret = extent_same_check_offsets(src, loff, &len, olen);
2907 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
2913 * Single inode case wants the same checks, except we
2914 * don't want our length pushed out past i_size as
2915 * comparing that data range makes no sense.
2917 * extent_same_check_offsets() will do this for an
2918 * unaligned length at i_size, so catch it here and
2919 * reject the request.
2921 * This effectively means we require aligned extents
2922 * for the single-inode case, whereas the other cases
2923 * allow an unaligned length so long as it ends at
2931 /* Check for overlapping ranges */
2932 if (dst_loff + len > loff && dst_loff < loff + len) {
2937 same_lock_start = min_t(u64, loff, dst_loff);
2938 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
2941 /* don't make the dst file partly checksummed */
2942 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2943 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2949 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
2954 ret = lock_extent_range(src, same_lock_start, same_lock_len,
2957 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
2960 * If one of the inodes has dirty pages in the respective range or
2961 * ordered extents, we need to flush dellaloc and wait for all ordered
2962 * extents in the range. We must unlock the pages and the ranges in the
2963 * io trees to avoid deadlocks when flushing delalloc (requires locking
2964 * pages) and when waiting for ordered extents to complete (they require
2967 if (ret == -EAGAIN) {
2969 * Ranges in the io trees already unlocked. Now unlock all
2970 * pages before waiting for all IO to complete.
2972 btrfs_cmp_data_free(&cmp);
2974 btrfs_wait_ordered_range(src, same_lock_start,
2977 btrfs_wait_ordered_range(src, loff, len);
2978 btrfs_wait_ordered_range(dst, dst_loff, len);
2984 /* ranges in the io trees already unlocked */
2985 btrfs_cmp_data_free(&cmp);
2989 /* pass original length for comparison so we stay within i_size */
2990 ret = btrfs_cmp_data(olen, &cmp);
2992 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
2995 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
2996 same_lock_start + same_lock_len - 1);
2998 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3000 btrfs_cmp_data_free(&cmp);
3005 btrfs_double_inode_unlock(src, dst);
3010 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3012 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3013 struct file *dst_file, u64 dst_loff)
3015 struct inode *src = file_inode(src_file);
3016 struct inode *dst = file_inode(dst_file);
3017 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3020 if (olen > BTRFS_MAX_DEDUPE_LEN)
3021 olen = BTRFS_MAX_DEDUPE_LEN;
3023 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
3025 * Btrfs does not support blocksize < page_size. As a
3026 * result, btrfs_cmp_data() won't correctly handle
3027 * this situation without an update.
3032 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3038 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3039 struct inode *inode,
3045 struct btrfs_root *root = BTRFS_I(inode)->root;
3048 inode_inc_iversion(inode);
3049 if (!no_time_update)
3050 inode->i_mtime = inode->i_ctime = current_time(inode);
3052 * We round up to the block size at eof when determining which
3053 * extents to clone above, but shouldn't round up the file size.
3055 if (endoff > destoff + olen)
3056 endoff = destoff + olen;
3057 if (endoff > inode->i_size)
3058 btrfs_i_size_write(BTRFS_I(inode), endoff);
3060 ret = btrfs_update_inode(trans, root, inode);
3062 btrfs_abort_transaction(trans, ret);
3063 btrfs_end_transaction(trans);
3066 ret = btrfs_end_transaction(trans);
3071 static void clone_update_extent_map(struct btrfs_inode *inode,
3072 const struct btrfs_trans_handle *trans,
3073 const struct btrfs_path *path,
3074 const u64 hole_offset,
3077 struct extent_map_tree *em_tree = &inode->extent_tree;
3078 struct extent_map *em;
3081 em = alloc_extent_map();
3083 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3088 struct btrfs_file_extent_item *fi;
3090 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3091 struct btrfs_file_extent_item);
3092 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3093 em->generation = -1;
3094 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3095 BTRFS_FILE_EXTENT_INLINE)
3096 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3097 &inode->runtime_flags);
3099 em->start = hole_offset;
3101 em->ram_bytes = em->len;
3102 em->orig_start = hole_offset;
3103 em->block_start = EXTENT_MAP_HOLE;
3105 em->orig_block_len = 0;
3106 em->compress_type = BTRFS_COMPRESS_NONE;
3107 em->generation = trans->transid;
3111 write_lock(&em_tree->lock);
3112 ret = add_extent_mapping(em_tree, em, 1);
3113 write_unlock(&em_tree->lock);
3114 if (ret != -EEXIST) {
3115 free_extent_map(em);
3118 btrfs_drop_extent_cache(inode, em->start,
3119 em->start + em->len - 1, 0);
3123 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3127 * Make sure we do not end up inserting an inline extent into a file that has
3128 * already other (non-inline) extents. If a file has an inline extent it can
3129 * not have any other extents and the (single) inline extent must start at the
3130 * file offset 0. Failing to respect these rules will lead to file corruption,
3131 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3133 * We can have extents that have been already written to disk or we can have
3134 * dirty ranges still in delalloc, in which case the extent maps and items are
3135 * created only when we run delalloc, and the delalloc ranges might fall outside
3136 * the range we are currently locking in the inode's io tree. So we check the
3137 * inode's i_size because of that (i_size updates are done while holding the
3138 * i_mutex, which we are holding here).
3139 * We also check to see if the inode has a size not greater than "datal" but has
3140 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3141 * protected against such concurrent fallocate calls by the i_mutex).
3143 * If the file has no extents but a size greater than datal, do not allow the
3144 * copy because we would need turn the inline extent into a non-inline one (even
3145 * with NO_HOLES enabled). If we find our destination inode only has one inline
3146 * extent, just overwrite it with the source inline extent if its size is less
3147 * than the source extent's size, or we could copy the source inline extent's
3148 * data into the destination inode's inline extent if the later is greater then
3151 static int clone_copy_inline_extent(struct inode *dst,
3152 struct btrfs_trans_handle *trans,
3153 struct btrfs_path *path,
3154 struct btrfs_key *new_key,
3155 const u64 drop_start,
3161 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
3162 struct btrfs_root *root = BTRFS_I(dst)->root;
3163 const u64 aligned_end = ALIGN(new_key->offset + datal,
3164 fs_info->sectorsize);
3166 struct btrfs_key key;
3168 if (new_key->offset > 0)
3171 key.objectid = btrfs_ino(BTRFS_I(dst));
3172 key.type = BTRFS_EXTENT_DATA_KEY;
3174 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3177 } else if (ret > 0) {
3178 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3179 ret = btrfs_next_leaf(root, path);
3183 goto copy_inline_extent;
3185 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3186 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3187 key.type == BTRFS_EXTENT_DATA_KEY) {
3188 ASSERT(key.offset > 0);
3191 } else if (i_size_read(dst) <= datal) {
3192 struct btrfs_file_extent_item *ei;
3196 * If the file size is <= datal, make sure there are no other
3197 * extents following (can happen do to an fallocate call with
3198 * the flag FALLOC_FL_KEEP_SIZE).
3200 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3201 struct btrfs_file_extent_item);
3203 * If it's an inline extent, it can not have other extents
3206 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3207 BTRFS_FILE_EXTENT_INLINE)
3208 goto copy_inline_extent;
3210 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3211 if (ext_len > aligned_end)
3214 ret = btrfs_next_item(root, path);
3217 } else if (ret == 0) {
3218 btrfs_item_key_to_cpu(path->nodes[0], &key,
3220 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3221 key.type == BTRFS_EXTENT_DATA_KEY)
3228 * We have no extent items, or we have an extent at offset 0 which may
3229 * or may not be inlined. All these cases are dealt the same way.
3231 if (i_size_read(dst) > datal) {
3233 * If the destination inode has an inline extent...
3234 * This would require copying the data from the source inline
3235 * extent into the beginning of the destination's inline extent.
3236 * But this is really complex, both extents can be compressed
3237 * or just one of them, which would require decompressing and
3238 * re-compressing data (which could increase the new compressed
3239 * size, not allowing the compressed data to fit anymore in an
3241 * So just don't support this case for now (it should be rare,
3242 * we are not really saving space when cloning inline extents).
3247 btrfs_release_path(path);
3248 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3251 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3256 const u32 start = btrfs_file_extent_calc_inline_size(0);
3258 memmove(inline_data + start, inline_data + start + skip, datal);
3261 write_extent_buffer(path->nodes[0], inline_data,
3262 btrfs_item_ptr_offset(path->nodes[0],
3265 inode_add_bytes(dst, datal);
3271 * btrfs_clone() - clone a range from inode file to another
3273 * @src: Inode to clone from
3274 * @inode: Inode to clone to
3275 * @off: Offset within source to start clone from
3276 * @olen: Original length, passed by user, of range to clone
3277 * @olen_aligned: Block-aligned value of olen
3278 * @destoff: Offset within @inode to start clone
3279 * @no_time_update: Whether to update mtime/ctime on the target inode
3281 static int btrfs_clone(struct inode *src, struct inode *inode,
3282 const u64 off, const u64 olen, const u64 olen_aligned,
3283 const u64 destoff, int no_time_update)
3285 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3286 struct btrfs_root *root = BTRFS_I(inode)->root;
3287 struct btrfs_path *path = NULL;
3288 struct extent_buffer *leaf;
3289 struct btrfs_trans_handle *trans;
3291 struct btrfs_key key;
3295 const u64 len = olen_aligned;
3296 u64 last_dest_end = destoff;
3299 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3303 path = btrfs_alloc_path();
3309 path->reada = READA_FORWARD;
3311 key.objectid = btrfs_ino(BTRFS_I(src));
3312 key.type = BTRFS_EXTENT_DATA_KEY;
3316 u64 next_key_min_offset = key.offset + 1;
3319 * note the key will change type as we walk through the
3322 path->leave_spinning = 1;
3323 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3328 * First search, if no extent item that starts at offset off was
3329 * found but the previous item is an extent item, it's possible
3330 * it might overlap our target range, therefore process it.
3332 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3333 btrfs_item_key_to_cpu(path->nodes[0], &key,
3334 path->slots[0] - 1);
3335 if (key.type == BTRFS_EXTENT_DATA_KEY)
3339 nritems = btrfs_header_nritems(path->nodes[0]);
3341 if (path->slots[0] >= nritems) {
3342 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3347 nritems = btrfs_header_nritems(path->nodes[0]);
3349 leaf = path->nodes[0];
3350 slot = path->slots[0];
3352 btrfs_item_key_to_cpu(leaf, &key, slot);
3353 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3354 key.objectid != btrfs_ino(BTRFS_I(src)))
3357 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3358 struct btrfs_file_extent_item *extent;
3361 struct btrfs_key new_key;
3362 u64 disko = 0, diskl = 0;
3363 u64 datao = 0, datal = 0;
3367 extent = btrfs_item_ptr(leaf, slot,
3368 struct btrfs_file_extent_item);
3369 comp = btrfs_file_extent_compression(leaf, extent);
3370 type = btrfs_file_extent_type(leaf, extent);
3371 if (type == BTRFS_FILE_EXTENT_REG ||
3372 type == BTRFS_FILE_EXTENT_PREALLOC) {
3373 disko = btrfs_file_extent_disk_bytenr(leaf,
3375 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3377 datao = btrfs_file_extent_offset(leaf, extent);
3378 datal = btrfs_file_extent_num_bytes(leaf,
3380 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3381 /* take upper bound, may be compressed */
3382 datal = btrfs_file_extent_ram_bytes(leaf,
3387 * The first search might have left us at an extent
3388 * item that ends before our target range's start, can
3389 * happen if we have holes and NO_HOLES feature enabled.
3391 if (key.offset + datal <= off) {
3394 } else if (key.offset >= off + len) {
3397 next_key_min_offset = key.offset + datal;
3398 size = btrfs_item_size_nr(leaf, slot);
3399 read_extent_buffer(leaf, buf,
3400 btrfs_item_ptr_offset(leaf, slot),
3403 btrfs_release_path(path);
3404 path->leave_spinning = 0;
3406 memcpy(&new_key, &key, sizeof(new_key));
3407 new_key.objectid = btrfs_ino(BTRFS_I(inode));
3408 if (off <= key.offset)
3409 new_key.offset = key.offset + destoff - off;
3411 new_key.offset = destoff;
3414 * Deal with a hole that doesn't have an extent item
3415 * that represents it (NO_HOLES feature enabled).
3416 * This hole is either in the middle of the cloning
3417 * range or at the beginning (fully overlaps it or
3418 * partially overlaps it).
3420 if (new_key.offset != last_dest_end)
3421 drop_start = last_dest_end;
3423 drop_start = new_key.offset;
3426 * 1 - adjusting old extent (we may have to split it)
3427 * 1 - add new extent
3430 trans = btrfs_start_transaction(root, 3);
3431 if (IS_ERR(trans)) {
3432 ret = PTR_ERR(trans);
3436 if (type == BTRFS_FILE_EXTENT_REG ||
3437 type == BTRFS_FILE_EXTENT_PREALLOC) {
3439 * a | --- range to clone ---| b
3440 * | ------------- extent ------------- |
3443 /* subtract range b */
3444 if (key.offset + datal > off + len)
3445 datal = off + len - key.offset;
3447 /* subtract range a */
3448 if (off > key.offset) {
3449 datao += off - key.offset;
3450 datal -= off - key.offset;
3453 ret = btrfs_drop_extents(trans, root, inode,
3455 new_key.offset + datal,
3458 if (ret != -EOPNOTSUPP)
3459 btrfs_abort_transaction(trans,
3461 btrfs_end_transaction(trans);
3465 ret = btrfs_insert_empty_item(trans, root, path,
3468 btrfs_abort_transaction(trans, ret);
3469 btrfs_end_transaction(trans);
3473 leaf = path->nodes[0];
3474 slot = path->slots[0];
3475 write_extent_buffer(leaf, buf,
3476 btrfs_item_ptr_offset(leaf, slot),
3479 extent = btrfs_item_ptr(leaf, slot,
3480 struct btrfs_file_extent_item);
3482 /* disko == 0 means it's a hole */
3486 btrfs_set_file_extent_offset(leaf, extent,
3488 btrfs_set_file_extent_num_bytes(leaf, extent,
3492 inode_add_bytes(inode, datal);
3493 ret = btrfs_inc_extent_ref(trans,
3496 root->root_key.objectid,
3497 btrfs_ino(BTRFS_I(inode)),
3498 new_key.offset - datao);
3500 btrfs_abort_transaction(trans,
3502 btrfs_end_transaction(trans);
3507 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3511 if (off > key.offset) {
3512 skip = off - key.offset;
3513 new_key.offset += skip;
3516 if (key.offset + datal > off + len)
3517 trim = key.offset + datal - (off + len);
3519 if (comp && (skip || trim)) {
3521 btrfs_end_transaction(trans);
3524 size -= skip + trim;
3525 datal -= skip + trim;
3527 ret = clone_copy_inline_extent(inode,
3534 if (ret != -EOPNOTSUPP)
3535 btrfs_abort_transaction(trans,
3537 btrfs_end_transaction(trans);
3540 leaf = path->nodes[0];
3541 slot = path->slots[0];
3544 /* If we have an implicit hole (NO_HOLES feature). */
3545 if (drop_start < new_key.offset)
3546 clone_update_extent_map(BTRFS_I(inode), trans,
3548 new_key.offset - drop_start);
3550 clone_update_extent_map(BTRFS_I(inode), trans,
3553 btrfs_mark_buffer_dirty(leaf);
3554 btrfs_release_path(path);
3556 last_dest_end = ALIGN(new_key.offset + datal,
3557 fs_info->sectorsize);
3558 ret = clone_finish_inode_update(trans, inode,
3564 if (new_key.offset + datal >= destoff + len)
3567 btrfs_release_path(path);
3568 key.offset = next_key_min_offset;
3570 if (fatal_signal_pending(current)) {
3577 if (last_dest_end < destoff + len) {
3579 * We have an implicit hole (NO_HOLES feature is enabled) that
3580 * fully or partially overlaps our cloning range at its end.
3582 btrfs_release_path(path);
3585 * 1 - remove extent(s)
3588 trans = btrfs_start_transaction(root, 2);
3589 if (IS_ERR(trans)) {
3590 ret = PTR_ERR(trans);
3593 ret = btrfs_drop_extents(trans, root, inode,
3594 last_dest_end, destoff + len, 1);
3596 if (ret != -EOPNOTSUPP)
3597 btrfs_abort_transaction(trans, ret);
3598 btrfs_end_transaction(trans);
3601 clone_update_extent_map(BTRFS_I(inode), trans, NULL,
3603 destoff + len - last_dest_end);
3604 ret = clone_finish_inode_update(trans, inode, destoff + len,
3605 destoff, olen, no_time_update);
3609 btrfs_free_path(path);
3614 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3615 u64 off, u64 olen, u64 destoff)
3617 struct inode *inode = file_inode(file);
3618 struct inode *src = file_inode(file_src);
3619 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3620 struct btrfs_root *root = BTRFS_I(inode)->root;
3623 u64 bs = fs_info->sb->s_blocksize;
3624 int same_inode = src == inode;
3628 * - split compressed inline extents. annoying: we need to
3629 * decompress into destination's address_space (the file offset
3630 * may change, so source mapping won't do), then recompress (or
3631 * otherwise reinsert) a subrange.
3633 * - split destination inode's inline extents. The inline extents can
3634 * be either compressed or non-compressed.
3637 if (btrfs_root_readonly(root))
3640 if (file_src->f_path.mnt != file->f_path.mnt ||
3641 src->i_sb != inode->i_sb)
3644 /* don't make the dst file partly checksummed */
3645 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3646 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3649 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3653 btrfs_double_inode_lock(src, inode);
3658 /* determine range to clone */
3660 if (off + len > src->i_size || off + len < off)
3663 olen = len = src->i_size - off;
3664 /* if we extend to eof, continue to block boundary */
3665 if (off + len == src->i_size)
3666 len = ALIGN(src->i_size, bs) - off;
3673 /* verify the end result is block aligned */
3674 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3675 !IS_ALIGNED(destoff, bs))
3678 /* verify if ranges are overlapped within the same file */
3680 if (destoff + len > off && destoff < off + len)
3684 if (destoff > inode->i_size) {
3685 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3691 * Lock the target range too. Right after we replace the file extent
3692 * items in the fs tree (which now point to the cloned data), we might
3693 * have a worker replace them with extent items relative to a write
3694 * operation that was issued before this clone operation (i.e. confront
3695 * with inode.c:btrfs_finish_ordered_io).
3698 u64 lock_start = min_t(u64, off, destoff);
3699 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3701 ret = lock_extent_range(src, lock_start, lock_len, true);
3703 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3708 /* ranges in the io trees already unlocked */
3712 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3715 u64 lock_start = min_t(u64, off, destoff);
3716 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3718 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3720 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3723 * Truncate page cache pages so that future reads will see the cloned
3724 * data immediately and not the previous data.
3726 truncate_inode_pages_range(&inode->i_data,
3727 round_down(destoff, PAGE_SIZE),
3728 round_up(destoff + len, PAGE_SIZE) - 1);
3731 btrfs_double_inode_unlock(src, inode);
3737 int btrfs_clone_file_range(struct file *src_file, loff_t off,
3738 struct file *dst_file, loff_t destoff, u64 len)
3740 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
3743 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3745 struct inode *inode = file_inode(file);
3746 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3747 struct btrfs_root *root = BTRFS_I(inode)->root;
3748 struct btrfs_root *new_root;
3749 struct btrfs_dir_item *di;
3750 struct btrfs_trans_handle *trans;
3751 struct btrfs_path *path;
3752 struct btrfs_key location;
3753 struct btrfs_disk_key disk_key;
3758 if (!capable(CAP_SYS_ADMIN))
3761 ret = mnt_want_write_file(file);
3765 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3771 objectid = BTRFS_FS_TREE_OBJECTID;
3773 location.objectid = objectid;
3774 location.type = BTRFS_ROOT_ITEM_KEY;
3775 location.offset = (u64)-1;
3777 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
3778 if (IS_ERR(new_root)) {
3779 ret = PTR_ERR(new_root);
3782 if (!is_fstree(new_root->objectid)) {
3787 path = btrfs_alloc_path();
3792 path->leave_spinning = 1;
3794 trans = btrfs_start_transaction(root, 1);
3795 if (IS_ERR(trans)) {
3796 btrfs_free_path(path);
3797 ret = PTR_ERR(trans);
3801 dir_id = btrfs_super_root_dir(fs_info->super_copy);
3802 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3803 dir_id, "default", 7, 1);
3804 if (IS_ERR_OR_NULL(di)) {
3805 btrfs_free_path(path);
3806 btrfs_end_transaction(trans);
3808 "Umm, you don't have the default diritem, this isn't going to work");
3813 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3814 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3815 btrfs_mark_buffer_dirty(path->nodes[0]);
3816 btrfs_free_path(path);
3818 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3819 btrfs_end_transaction(trans);
3821 mnt_drop_write_file(file);
3825 static void get_block_group_info(struct list_head *groups_list,
3826 struct btrfs_ioctl_space_info *space)
3828 struct btrfs_block_group_cache *block_group;
3830 space->total_bytes = 0;
3831 space->used_bytes = 0;
3833 list_for_each_entry(block_group, groups_list, list) {
3834 space->flags = block_group->flags;
3835 space->total_bytes += block_group->key.offset;
3836 space->used_bytes +=
3837 btrfs_block_group_used(&block_group->item);
3841 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3844 struct btrfs_ioctl_space_args space_args;
3845 struct btrfs_ioctl_space_info space;
3846 struct btrfs_ioctl_space_info *dest;
3847 struct btrfs_ioctl_space_info *dest_orig;
3848 struct btrfs_ioctl_space_info __user *user_dest;
3849 struct btrfs_space_info *info;
3850 static const u64 types[] = {
3851 BTRFS_BLOCK_GROUP_DATA,
3852 BTRFS_BLOCK_GROUP_SYSTEM,
3853 BTRFS_BLOCK_GROUP_METADATA,
3854 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3862 if (copy_from_user(&space_args,
3863 (struct btrfs_ioctl_space_args __user *)arg,
3864 sizeof(space_args)))
3867 for (i = 0; i < num_types; i++) {
3868 struct btrfs_space_info *tmp;
3872 list_for_each_entry_rcu(tmp, &fs_info->space_info,
3874 if (tmp->flags == types[i]) {
3884 down_read(&info->groups_sem);
3885 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3886 if (!list_empty(&info->block_groups[c]))
3889 up_read(&info->groups_sem);
3893 * Global block reserve, exported as a space_info
3897 /* space_slots == 0 means they are asking for a count */
3898 if (space_args.space_slots == 0) {
3899 space_args.total_spaces = slot_count;
3903 slot_count = min_t(u64, space_args.space_slots, slot_count);
3905 alloc_size = sizeof(*dest) * slot_count;
3907 /* we generally have at most 6 or so space infos, one for each raid
3908 * level. So, a whole page should be more than enough for everyone
3910 if (alloc_size > PAGE_SIZE)
3913 space_args.total_spaces = 0;
3914 dest = kmalloc(alloc_size, GFP_KERNEL);
3919 /* now we have a buffer to copy into */
3920 for (i = 0; i < num_types; i++) {
3921 struct btrfs_space_info *tmp;
3928 list_for_each_entry_rcu(tmp, &fs_info->space_info,
3930 if (tmp->flags == types[i]) {
3939 down_read(&info->groups_sem);
3940 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3941 if (!list_empty(&info->block_groups[c])) {
3942 get_block_group_info(&info->block_groups[c],
3944 memcpy(dest, &space, sizeof(space));
3946 space_args.total_spaces++;
3952 up_read(&info->groups_sem);
3956 * Add global block reserve
3959 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3961 spin_lock(&block_rsv->lock);
3962 space.total_bytes = block_rsv->size;
3963 space.used_bytes = block_rsv->size - block_rsv->reserved;
3964 spin_unlock(&block_rsv->lock);
3965 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3966 memcpy(dest, &space, sizeof(space));
3967 space_args.total_spaces++;
3970 user_dest = (struct btrfs_ioctl_space_info __user *)
3971 (arg + sizeof(struct btrfs_ioctl_space_args));
3973 if (copy_to_user(user_dest, dest_orig, alloc_size))
3978 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3984 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3987 struct btrfs_trans_handle *trans;
3991 trans = btrfs_attach_transaction_barrier(root);
3992 if (IS_ERR(trans)) {
3993 if (PTR_ERR(trans) != -ENOENT)
3994 return PTR_ERR(trans);
3996 /* No running transaction, don't bother */
3997 transid = root->fs_info->last_trans_committed;
4000 transid = trans->transid;
4001 ret = btrfs_commit_transaction_async(trans, 0);
4003 btrfs_end_transaction(trans);
4008 if (copy_to_user(argp, &transid, sizeof(transid)))
4013 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
4019 if (copy_from_user(&transid, argp, sizeof(transid)))
4022 transid = 0; /* current trans */
4024 return btrfs_wait_for_commit(fs_info, transid);
4027 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4029 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4030 struct btrfs_ioctl_scrub_args *sa;
4033 if (!capable(CAP_SYS_ADMIN))
4036 sa = memdup_user(arg, sizeof(*sa));
4040 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4041 ret = mnt_want_write_file(file);
4046 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4047 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4050 if (copy_to_user(arg, sa, sizeof(*sa)))
4053 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4054 mnt_drop_write_file(file);
4060 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
4062 if (!capable(CAP_SYS_ADMIN))
4065 return btrfs_scrub_cancel(fs_info);
4068 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
4071 struct btrfs_ioctl_scrub_args *sa;
4074 if (!capable(CAP_SYS_ADMIN))
4077 sa = memdup_user(arg, sizeof(*sa));
4081 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
4083 if (copy_to_user(arg, sa, sizeof(*sa)))
4090 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
4093 struct btrfs_ioctl_get_dev_stats *sa;
4096 sa = memdup_user(arg, sizeof(*sa));
4100 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4105 ret = btrfs_get_dev_stats(fs_info, sa);
4107 if (copy_to_user(arg, sa, sizeof(*sa)))
4114 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4117 struct btrfs_ioctl_dev_replace_args *p;
4120 if (!capable(CAP_SYS_ADMIN))
4123 p = memdup_user(arg, sizeof(*p));
4128 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4129 if (sb_rdonly(fs_info->sb)) {
4133 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4134 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4136 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
4137 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4140 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4141 btrfs_dev_replace_status(fs_info, p);
4144 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4145 p->result = btrfs_dev_replace_cancel(fs_info);
4153 if (copy_to_user(arg, p, sizeof(*p)))
4160 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4166 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4167 struct inode_fs_paths *ipath = NULL;
4168 struct btrfs_path *path;
4170 if (!capable(CAP_DAC_READ_SEARCH))
4173 path = btrfs_alloc_path();
4179 ipa = memdup_user(arg, sizeof(*ipa));
4186 size = min_t(u32, ipa->size, 4096);
4187 ipath = init_ipath(size, root, path);
4188 if (IS_ERR(ipath)) {
4189 ret = PTR_ERR(ipath);
4194 ret = paths_from_inode(ipa->inum, ipath);
4198 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4199 rel_ptr = ipath->fspath->val[i] -
4200 (u64)(unsigned long)ipath->fspath->val;
4201 ipath->fspath->val[i] = rel_ptr;
4204 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
4205 ipath->fspath, size);
4212 btrfs_free_path(path);
4219 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4221 struct btrfs_data_container *inodes = ctx;
4222 const size_t c = 3 * sizeof(u64);
4224 if (inodes->bytes_left >= c) {
4225 inodes->bytes_left -= c;
4226 inodes->val[inodes->elem_cnt] = inum;
4227 inodes->val[inodes->elem_cnt + 1] = offset;
4228 inodes->val[inodes->elem_cnt + 2] = root;
4229 inodes->elem_cnt += 3;
4231 inodes->bytes_missing += c - inodes->bytes_left;
4232 inodes->bytes_left = 0;
4233 inodes->elem_missed += 3;
4239 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
4240 void __user *arg, int version)
4244 struct btrfs_ioctl_logical_ino_args *loi;
4245 struct btrfs_data_container *inodes = NULL;
4246 struct btrfs_path *path = NULL;
4249 if (!capable(CAP_SYS_ADMIN))
4252 loi = memdup_user(arg, sizeof(*loi));
4254 return PTR_ERR(loi);
4257 ignore_offset = false;
4258 size = min_t(u32, loi->size, SZ_64K);
4260 /* All reserved bits must be 0 for now */
4261 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
4265 /* Only accept flags we have defined so far */
4266 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
4270 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
4271 size = min_t(u32, loi->size, SZ_16M);
4274 path = btrfs_alloc_path();
4280 inodes = init_data_container(size);
4281 if (IS_ERR(inodes)) {
4282 ret = PTR_ERR(inodes);
4287 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
4288 build_ino_list, inodes, ignore_offset);
4294 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
4300 btrfs_free_path(path);
4308 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
4309 struct btrfs_ioctl_balance_args *bargs)
4311 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4313 bargs->flags = bctl->flags;
4315 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
4316 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4317 if (atomic_read(&fs_info->balance_pause_req))
4318 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4319 if (atomic_read(&fs_info->balance_cancel_req))
4320 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4322 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4323 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4324 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4326 spin_lock(&fs_info->balance_lock);
4327 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4328 spin_unlock(&fs_info->balance_lock);
4331 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4333 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4334 struct btrfs_fs_info *fs_info = root->fs_info;
4335 struct btrfs_ioctl_balance_args *bargs;
4336 struct btrfs_balance_control *bctl;
4337 bool need_unlock; /* for mut. excl. ops lock */
4340 if (!capable(CAP_SYS_ADMIN))
4343 ret = mnt_want_write_file(file);
4348 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4349 mutex_lock(&fs_info->balance_mutex);
4355 * mut. excl. ops lock is locked. Three possibilities:
4356 * (1) some other op is running
4357 * (2) balance is running
4358 * (3) balance is paused -- special case (think resume)
4360 mutex_lock(&fs_info->balance_mutex);
4361 if (fs_info->balance_ctl) {
4362 /* this is either (2) or (3) */
4363 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4364 mutex_unlock(&fs_info->balance_mutex);
4366 * Lock released to allow other waiters to continue,
4367 * we'll reexamine the status again.
4369 mutex_lock(&fs_info->balance_mutex);
4371 if (fs_info->balance_ctl &&
4372 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4374 need_unlock = false;
4378 mutex_unlock(&fs_info->balance_mutex);
4382 mutex_unlock(&fs_info->balance_mutex);
4388 mutex_unlock(&fs_info->balance_mutex);
4389 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4394 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
4397 bargs = memdup_user(arg, sizeof(*bargs));
4398 if (IS_ERR(bargs)) {
4399 ret = PTR_ERR(bargs);
4403 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4404 if (!fs_info->balance_ctl) {
4409 bctl = fs_info->balance_ctl;
4410 spin_lock(&fs_info->balance_lock);
4411 bctl->flags |= BTRFS_BALANCE_RESUME;
4412 spin_unlock(&fs_info->balance_lock);
4420 if (fs_info->balance_ctl) {
4425 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4432 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4433 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4434 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4436 bctl->flags = bargs->flags;
4438 /* balance everything - no filters */
4439 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4442 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4449 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to
4450 * btrfs_balance. bctl is freed in reset_balance_state, or, if
4451 * restriper was paused all the way until unmount, in free_fs_info.
4452 * The flag should be cleared after reset_balance_state.
4454 need_unlock = false;
4456 ret = btrfs_balance(fs_info, bctl, bargs);
4460 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4469 mutex_unlock(&fs_info->balance_mutex);
4471 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4473 mnt_drop_write_file(file);
4477 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4479 if (!capable(CAP_SYS_ADMIN))
4483 case BTRFS_BALANCE_CTL_PAUSE:
4484 return btrfs_pause_balance(fs_info);
4485 case BTRFS_BALANCE_CTL_CANCEL:
4486 return btrfs_cancel_balance(fs_info);
4492 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4495 struct btrfs_ioctl_balance_args *bargs;
4498 if (!capable(CAP_SYS_ADMIN))
4501 mutex_lock(&fs_info->balance_mutex);
4502 if (!fs_info->balance_ctl) {
4507 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4513 btrfs_update_ioctl_balance_args(fs_info, bargs);
4515 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4520 mutex_unlock(&fs_info->balance_mutex);
4524 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4526 struct inode *inode = file_inode(file);
4527 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4528 struct btrfs_ioctl_quota_ctl_args *sa;
4529 struct btrfs_trans_handle *trans = NULL;
4533 if (!capable(CAP_SYS_ADMIN))
4536 ret = mnt_want_write_file(file);
4540 sa = memdup_user(arg, sizeof(*sa));
4546 down_write(&fs_info->subvol_sem);
4547 trans = btrfs_start_transaction(fs_info->tree_root, 2);
4548 if (IS_ERR(trans)) {
4549 ret = PTR_ERR(trans);
4554 case BTRFS_QUOTA_CTL_ENABLE:
4555 ret = btrfs_quota_enable(trans, fs_info);
4557 case BTRFS_QUOTA_CTL_DISABLE:
4558 ret = btrfs_quota_disable(trans, fs_info);
4565 err = btrfs_commit_transaction(trans);
4570 up_write(&fs_info->subvol_sem);
4572 mnt_drop_write_file(file);
4576 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4578 struct inode *inode = file_inode(file);
4579 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4580 struct btrfs_root *root = BTRFS_I(inode)->root;
4581 struct btrfs_ioctl_qgroup_assign_args *sa;
4582 struct btrfs_trans_handle *trans;
4586 if (!capable(CAP_SYS_ADMIN))
4589 ret = mnt_want_write_file(file);
4593 sa = memdup_user(arg, sizeof(*sa));
4599 trans = btrfs_join_transaction(root);
4600 if (IS_ERR(trans)) {
4601 ret = PTR_ERR(trans);
4606 ret = btrfs_add_qgroup_relation(trans, fs_info,
4609 ret = btrfs_del_qgroup_relation(trans, fs_info,
4613 /* update qgroup status and info */
4614 err = btrfs_run_qgroups(trans, fs_info);
4616 btrfs_handle_fs_error(fs_info, err,
4617 "failed to update qgroup status and info");
4618 err = btrfs_end_transaction(trans);
4625 mnt_drop_write_file(file);
4629 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4631 struct inode *inode = file_inode(file);
4632 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4633 struct btrfs_root *root = BTRFS_I(inode)->root;
4634 struct btrfs_ioctl_qgroup_create_args *sa;
4635 struct btrfs_trans_handle *trans;
4639 if (!capable(CAP_SYS_ADMIN))
4642 ret = mnt_want_write_file(file);
4646 sa = memdup_user(arg, sizeof(*sa));
4652 if (!sa->qgroupid) {
4657 trans = btrfs_join_transaction(root);
4658 if (IS_ERR(trans)) {
4659 ret = PTR_ERR(trans);
4664 ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid);
4666 ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid);
4669 err = btrfs_end_transaction(trans);
4676 mnt_drop_write_file(file);
4680 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4682 struct inode *inode = file_inode(file);
4683 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4684 struct btrfs_root *root = BTRFS_I(inode)->root;
4685 struct btrfs_ioctl_qgroup_limit_args *sa;
4686 struct btrfs_trans_handle *trans;
4691 if (!capable(CAP_SYS_ADMIN))
4694 ret = mnt_want_write_file(file);
4698 sa = memdup_user(arg, sizeof(*sa));
4704 trans = btrfs_join_transaction(root);
4705 if (IS_ERR(trans)) {
4706 ret = PTR_ERR(trans);
4710 qgroupid = sa->qgroupid;
4712 /* take the current subvol as qgroup */
4713 qgroupid = root->root_key.objectid;
4716 ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim);
4718 err = btrfs_end_transaction(trans);
4725 mnt_drop_write_file(file);
4729 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4731 struct inode *inode = file_inode(file);
4732 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4733 struct btrfs_ioctl_quota_rescan_args *qsa;
4736 if (!capable(CAP_SYS_ADMIN))
4739 ret = mnt_want_write_file(file);
4743 qsa = memdup_user(arg, sizeof(*qsa));
4754 ret = btrfs_qgroup_rescan(fs_info);
4759 mnt_drop_write_file(file);
4763 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4765 struct inode *inode = file_inode(file);
4766 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4767 struct btrfs_ioctl_quota_rescan_args *qsa;
4770 if (!capable(CAP_SYS_ADMIN))
4773 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
4777 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4779 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
4782 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4789 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4791 struct inode *inode = file_inode(file);
4792 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4794 if (!capable(CAP_SYS_ADMIN))
4797 return btrfs_qgroup_wait_for_completion(fs_info, true);
4800 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4801 struct btrfs_ioctl_received_subvol_args *sa)
4803 struct inode *inode = file_inode(file);
4804 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4805 struct btrfs_root *root = BTRFS_I(inode)->root;
4806 struct btrfs_root_item *root_item = &root->root_item;
4807 struct btrfs_trans_handle *trans;
4808 struct timespec ct = current_time(inode);
4810 int received_uuid_changed;
4812 if (!inode_owner_or_capable(inode))
4815 ret = mnt_want_write_file(file);
4819 down_write(&fs_info->subvol_sem);
4821 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4826 if (btrfs_root_readonly(root)) {
4833 * 2 - uuid items (received uuid + subvol uuid)
4835 trans = btrfs_start_transaction(root, 3);
4836 if (IS_ERR(trans)) {
4837 ret = PTR_ERR(trans);
4842 sa->rtransid = trans->transid;
4843 sa->rtime.sec = ct.tv_sec;
4844 sa->rtime.nsec = ct.tv_nsec;
4846 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4848 if (received_uuid_changed &&
4849 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4850 ret = btrfs_uuid_tree_rem(trans, fs_info,
4851 root_item->received_uuid,
4852 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4853 root->root_key.objectid);
4854 if (ret && ret != -ENOENT) {
4855 btrfs_abort_transaction(trans, ret);
4856 btrfs_end_transaction(trans);
4860 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4861 btrfs_set_root_stransid(root_item, sa->stransid);
4862 btrfs_set_root_rtransid(root_item, sa->rtransid);
4863 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4864 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4865 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4866 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4868 ret = btrfs_update_root(trans, fs_info->tree_root,
4869 &root->root_key, &root->root_item);
4871 btrfs_end_transaction(trans);
4874 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4875 ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid,
4876 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4877 root->root_key.objectid);
4878 if (ret < 0 && ret != -EEXIST) {
4879 btrfs_abort_transaction(trans, ret);
4880 btrfs_end_transaction(trans);
4884 ret = btrfs_commit_transaction(trans);
4886 up_write(&fs_info->subvol_sem);
4887 mnt_drop_write_file(file);
4892 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4895 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4896 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4899 args32 = memdup_user(arg, sizeof(*args32));
4901 return PTR_ERR(args32);
4903 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4909 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4910 args64->stransid = args32->stransid;
4911 args64->rtransid = args32->rtransid;
4912 args64->stime.sec = args32->stime.sec;
4913 args64->stime.nsec = args32->stime.nsec;
4914 args64->rtime.sec = args32->rtime.sec;
4915 args64->rtime.nsec = args32->rtime.nsec;
4916 args64->flags = args32->flags;
4918 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4922 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4923 args32->stransid = args64->stransid;
4924 args32->rtransid = args64->rtransid;
4925 args32->stime.sec = args64->stime.sec;
4926 args32->stime.nsec = args64->stime.nsec;
4927 args32->rtime.sec = args64->rtime.sec;
4928 args32->rtime.nsec = args64->rtime.nsec;
4929 args32->flags = args64->flags;
4931 ret = copy_to_user(arg, args32, sizeof(*args32));
4942 static long btrfs_ioctl_set_received_subvol(struct file *file,
4945 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4948 sa = memdup_user(arg, sizeof(*sa));
4952 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4957 ret = copy_to_user(arg, sa, sizeof(*sa));
4966 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4968 struct inode *inode = file_inode(file);
4969 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4972 char label[BTRFS_LABEL_SIZE];
4974 spin_lock(&fs_info->super_lock);
4975 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4976 spin_unlock(&fs_info->super_lock);
4978 len = strnlen(label, BTRFS_LABEL_SIZE);
4980 if (len == BTRFS_LABEL_SIZE) {
4982 "label is too long, return the first %zu bytes",
4986 ret = copy_to_user(arg, label, len);
4988 return ret ? -EFAULT : 0;
4991 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4993 struct inode *inode = file_inode(file);
4994 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4995 struct btrfs_root *root = BTRFS_I(inode)->root;
4996 struct btrfs_super_block *super_block = fs_info->super_copy;
4997 struct btrfs_trans_handle *trans;
4998 char label[BTRFS_LABEL_SIZE];
5001 if (!capable(CAP_SYS_ADMIN))
5004 if (copy_from_user(label, arg, sizeof(label)))
5007 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5009 "unable to set label with more than %d bytes",
5010 BTRFS_LABEL_SIZE - 1);
5014 ret = mnt_want_write_file(file);
5018 trans = btrfs_start_transaction(root, 0);
5019 if (IS_ERR(trans)) {
5020 ret = PTR_ERR(trans);
5024 spin_lock(&fs_info->super_lock);
5025 strcpy(super_block->label, label);
5026 spin_unlock(&fs_info->super_lock);
5027 ret = btrfs_commit_transaction(trans);
5030 mnt_drop_write_file(file);
5034 #define INIT_FEATURE_FLAGS(suffix) \
5035 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5036 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5037 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5039 int btrfs_ioctl_get_supported_features(void __user *arg)
5041 static const struct btrfs_ioctl_feature_flags features[3] = {
5042 INIT_FEATURE_FLAGS(SUPP),
5043 INIT_FEATURE_FLAGS(SAFE_SET),
5044 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5047 if (copy_to_user(arg, &features, sizeof(features)))
5053 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5055 struct inode *inode = file_inode(file);
5056 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5057 struct btrfs_super_block *super_block = fs_info->super_copy;
5058 struct btrfs_ioctl_feature_flags features;
5060 features.compat_flags = btrfs_super_compat_flags(super_block);
5061 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5062 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5064 if (copy_to_user(arg, &features, sizeof(features)))
5070 static int check_feature_bits(struct btrfs_fs_info *fs_info,
5071 enum btrfs_feature_set set,
5072 u64 change_mask, u64 flags, u64 supported_flags,
5073 u64 safe_set, u64 safe_clear)
5075 const char *type = btrfs_feature_set_names[set];
5077 u64 disallowed, unsupported;
5078 u64 set_mask = flags & change_mask;
5079 u64 clear_mask = ~flags & change_mask;
5081 unsupported = set_mask & ~supported_flags;
5083 names = btrfs_printable_features(set, unsupported);
5086 "this kernel does not support the %s feature bit%s",
5087 names, strchr(names, ',') ? "s" : "");
5091 "this kernel does not support %s bits 0x%llx",
5096 disallowed = set_mask & ~safe_set;
5098 names = btrfs_printable_features(set, disallowed);
5101 "can't set the %s feature bit%s while mounted",
5102 names, strchr(names, ',') ? "s" : "");
5106 "can't set %s bits 0x%llx while mounted",
5111 disallowed = clear_mask & ~safe_clear;
5113 names = btrfs_printable_features(set, disallowed);
5116 "can't clear the %s feature bit%s while mounted",
5117 names, strchr(names, ',') ? "s" : "");
5121 "can't clear %s bits 0x%llx while mounted",
5129 #define check_feature(fs_info, change_mask, flags, mask_base) \
5130 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5131 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5132 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5133 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5135 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5137 struct inode *inode = file_inode(file);
5138 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5139 struct btrfs_root *root = BTRFS_I(inode)->root;
5140 struct btrfs_super_block *super_block = fs_info->super_copy;
5141 struct btrfs_ioctl_feature_flags flags[2];
5142 struct btrfs_trans_handle *trans;
5146 if (!capable(CAP_SYS_ADMIN))
5149 if (copy_from_user(flags, arg, sizeof(flags)))
5153 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5154 !flags[0].incompat_flags)
5157 ret = check_feature(fs_info, flags[0].compat_flags,
5158 flags[1].compat_flags, COMPAT);
5162 ret = check_feature(fs_info, flags[0].compat_ro_flags,
5163 flags[1].compat_ro_flags, COMPAT_RO);
5167 ret = check_feature(fs_info, flags[0].incompat_flags,
5168 flags[1].incompat_flags, INCOMPAT);
5172 ret = mnt_want_write_file(file);
5176 trans = btrfs_start_transaction(root, 0);
5177 if (IS_ERR(trans)) {
5178 ret = PTR_ERR(trans);
5179 goto out_drop_write;
5182 spin_lock(&fs_info->super_lock);
5183 newflags = btrfs_super_compat_flags(super_block);
5184 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5185 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5186 btrfs_set_super_compat_flags(super_block, newflags);
5188 newflags = btrfs_super_compat_ro_flags(super_block);
5189 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5190 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5191 btrfs_set_super_compat_ro_flags(super_block, newflags);
5193 newflags = btrfs_super_incompat_flags(super_block);
5194 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5195 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5196 btrfs_set_super_incompat_flags(super_block, newflags);
5197 spin_unlock(&fs_info->super_lock);
5199 ret = btrfs_commit_transaction(trans);
5201 mnt_drop_write_file(file);
5206 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
5208 struct btrfs_ioctl_send_args *arg;
5212 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5213 struct btrfs_ioctl_send_args_32 args32;
5215 ret = copy_from_user(&args32, argp, sizeof(args32));
5218 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
5221 arg->send_fd = args32.send_fd;
5222 arg->clone_sources_count = args32.clone_sources_count;
5223 arg->clone_sources = compat_ptr(args32.clone_sources);
5224 arg->parent_root = args32.parent_root;
5225 arg->flags = args32.flags;
5226 memcpy(arg->reserved, args32.reserved,
5227 sizeof(args32.reserved));
5232 arg = memdup_user(argp, sizeof(*arg));
5234 return PTR_ERR(arg);
5236 ret = btrfs_ioctl_send(file, arg);
5241 long btrfs_ioctl(struct file *file, unsigned int
5242 cmd, unsigned long arg)
5244 struct inode *inode = file_inode(file);
5245 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5246 struct btrfs_root *root = BTRFS_I(inode)->root;
5247 void __user *argp = (void __user *)arg;
5250 case FS_IOC_GETFLAGS:
5251 return btrfs_ioctl_getflags(file, argp);
5252 case FS_IOC_SETFLAGS:
5253 return btrfs_ioctl_setflags(file, argp);
5254 case FS_IOC_GETVERSION:
5255 return btrfs_ioctl_getversion(file, argp);
5257 return btrfs_ioctl_fitrim(file, argp);
5258 case BTRFS_IOC_SNAP_CREATE:
5259 return btrfs_ioctl_snap_create(file, argp, 0);
5260 case BTRFS_IOC_SNAP_CREATE_V2:
5261 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5262 case BTRFS_IOC_SUBVOL_CREATE:
5263 return btrfs_ioctl_snap_create(file, argp, 1);
5264 case BTRFS_IOC_SUBVOL_CREATE_V2:
5265 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5266 case BTRFS_IOC_SNAP_DESTROY:
5267 return btrfs_ioctl_snap_destroy(file, argp);
5268 case BTRFS_IOC_SUBVOL_GETFLAGS:
5269 return btrfs_ioctl_subvol_getflags(file, argp);
5270 case BTRFS_IOC_SUBVOL_SETFLAGS:
5271 return btrfs_ioctl_subvol_setflags(file, argp);
5272 case BTRFS_IOC_DEFAULT_SUBVOL:
5273 return btrfs_ioctl_default_subvol(file, argp);
5274 case BTRFS_IOC_DEFRAG:
5275 return btrfs_ioctl_defrag(file, NULL);
5276 case BTRFS_IOC_DEFRAG_RANGE:
5277 return btrfs_ioctl_defrag(file, argp);
5278 case BTRFS_IOC_RESIZE:
5279 return btrfs_ioctl_resize(file, argp);
5280 case BTRFS_IOC_ADD_DEV:
5281 return btrfs_ioctl_add_dev(fs_info, argp);
5282 case BTRFS_IOC_RM_DEV:
5283 return btrfs_ioctl_rm_dev(file, argp);
5284 case BTRFS_IOC_RM_DEV_V2:
5285 return btrfs_ioctl_rm_dev_v2(file, argp);
5286 case BTRFS_IOC_FS_INFO:
5287 return btrfs_ioctl_fs_info(fs_info, argp);
5288 case BTRFS_IOC_DEV_INFO:
5289 return btrfs_ioctl_dev_info(fs_info, argp);
5290 case BTRFS_IOC_BALANCE:
5291 return btrfs_ioctl_balance(file, NULL);
5292 case BTRFS_IOC_TREE_SEARCH:
5293 return btrfs_ioctl_tree_search(file, argp);
5294 case BTRFS_IOC_TREE_SEARCH_V2:
5295 return btrfs_ioctl_tree_search_v2(file, argp);
5296 case BTRFS_IOC_INO_LOOKUP:
5297 return btrfs_ioctl_ino_lookup(file, argp);
5298 case BTRFS_IOC_INO_PATHS:
5299 return btrfs_ioctl_ino_to_path(root, argp);
5300 case BTRFS_IOC_LOGICAL_INO:
5301 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
5302 case BTRFS_IOC_LOGICAL_INO_V2:
5303 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
5304 case BTRFS_IOC_SPACE_INFO:
5305 return btrfs_ioctl_space_info(fs_info, argp);
5306 case BTRFS_IOC_SYNC: {
5309 ret = btrfs_start_delalloc_roots(fs_info, -1);
5312 ret = btrfs_sync_fs(inode->i_sb, 1);
5314 * The transaction thread may want to do more work,
5315 * namely it pokes the cleaner kthread that will start
5316 * processing uncleaned subvols.
5318 wake_up_process(fs_info->transaction_kthread);
5321 case BTRFS_IOC_START_SYNC:
5322 return btrfs_ioctl_start_sync(root, argp);
5323 case BTRFS_IOC_WAIT_SYNC:
5324 return btrfs_ioctl_wait_sync(fs_info, argp);
5325 case BTRFS_IOC_SCRUB:
5326 return btrfs_ioctl_scrub(file, argp);
5327 case BTRFS_IOC_SCRUB_CANCEL:
5328 return btrfs_ioctl_scrub_cancel(fs_info);
5329 case BTRFS_IOC_SCRUB_PROGRESS:
5330 return btrfs_ioctl_scrub_progress(fs_info, argp);
5331 case BTRFS_IOC_BALANCE_V2:
5332 return btrfs_ioctl_balance(file, argp);
5333 case BTRFS_IOC_BALANCE_CTL:
5334 return btrfs_ioctl_balance_ctl(fs_info, arg);
5335 case BTRFS_IOC_BALANCE_PROGRESS:
5336 return btrfs_ioctl_balance_progress(fs_info, argp);
5337 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5338 return btrfs_ioctl_set_received_subvol(file, argp);
5340 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5341 return btrfs_ioctl_set_received_subvol_32(file, argp);
5343 case BTRFS_IOC_SEND:
5344 return _btrfs_ioctl_send(file, argp, false);
5345 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5346 case BTRFS_IOC_SEND_32:
5347 return _btrfs_ioctl_send(file, argp, true);
5349 case BTRFS_IOC_GET_DEV_STATS:
5350 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5351 case BTRFS_IOC_QUOTA_CTL:
5352 return btrfs_ioctl_quota_ctl(file, argp);
5353 case BTRFS_IOC_QGROUP_ASSIGN:
5354 return btrfs_ioctl_qgroup_assign(file, argp);
5355 case BTRFS_IOC_QGROUP_CREATE:
5356 return btrfs_ioctl_qgroup_create(file, argp);
5357 case BTRFS_IOC_QGROUP_LIMIT:
5358 return btrfs_ioctl_qgroup_limit(file, argp);
5359 case BTRFS_IOC_QUOTA_RESCAN:
5360 return btrfs_ioctl_quota_rescan(file, argp);
5361 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5362 return btrfs_ioctl_quota_rescan_status(file, argp);
5363 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5364 return btrfs_ioctl_quota_rescan_wait(file, argp);
5365 case BTRFS_IOC_DEV_REPLACE:
5366 return btrfs_ioctl_dev_replace(fs_info, argp);
5367 case BTRFS_IOC_GET_FSLABEL:
5368 return btrfs_ioctl_get_fslabel(file, argp);
5369 case BTRFS_IOC_SET_FSLABEL:
5370 return btrfs_ioctl_set_fslabel(file, argp);
5371 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5372 return btrfs_ioctl_get_supported_features(argp);
5373 case BTRFS_IOC_GET_FEATURES:
5374 return btrfs_ioctl_get_features(file, argp);
5375 case BTRFS_IOC_SET_FEATURES:
5376 return btrfs_ioctl_set_features(file, argp);
5382 #ifdef CONFIG_COMPAT
5383 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5386 * These all access 32-bit values anyway so no further
5387 * handling is necessary.
5390 case FS_IOC32_GETFLAGS:
5391 cmd = FS_IOC_GETFLAGS;
5393 case FS_IOC32_SETFLAGS:
5394 cmd = FS_IOC_SETFLAGS;
5396 case FS_IOC32_GETVERSION:
5397 cmd = FS_IOC_GETVERSION;
5401 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));