1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2007 Oracle. All rights reserved.
6 #include <linux/kernel.h>
8 #include <linux/file.h>
10 #include <linux/fsnotify.h>
11 #include <linux/pagemap.h>
12 #include <linux/highmem.h>
13 #include <linux/time.h>
14 #include <linux/string.h>
15 #include <linux/backing-dev.h>
16 #include <linux/mount.h>
17 #include <linux/namei.h>
18 #include <linux/writeback.h>
19 #include <linux/compat.h>
20 #include <linux/security.h>
21 #include <linux/xattr.h>
23 #include <linux/slab.h>
24 #include <linux/blkdev.h>
25 #include <linux/uuid.h>
26 #include <linux/btrfs.h>
27 #include <linux/uaccess.h>
28 #include <linux/iversion.h>
31 #include "transaction.h"
32 #include "btrfs_inode.h"
33 #include "print-tree.h"
36 #include "inode-map.h"
38 #include "rcu-string.h"
40 #include "dev-replace.h"
45 #include "compression.h"
46 #include "space-info.h"
47 #include "delalloc-space.h"
48 #include "block-group.h"
51 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
52 * structures are incorrect, as the timespec structure from userspace
53 * is 4 bytes too small. We define these alternatives here to teach
54 * the kernel about the 32-bit struct packing.
56 struct btrfs_ioctl_timespec_32 {
59 } __attribute__ ((__packed__));
61 struct btrfs_ioctl_received_subvol_args_32 {
62 char uuid[BTRFS_UUID_SIZE]; /* in */
63 __u64 stransid; /* in */
64 __u64 rtransid; /* out */
65 struct btrfs_ioctl_timespec_32 stime; /* in */
66 struct btrfs_ioctl_timespec_32 rtime; /* out */
68 __u64 reserved[16]; /* in */
69 } __attribute__ ((__packed__));
71 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
72 struct btrfs_ioctl_received_subvol_args_32)
75 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
76 struct btrfs_ioctl_send_args_32 {
77 __s64 send_fd; /* in */
78 __u64 clone_sources_count; /* in */
79 compat_uptr_t clone_sources; /* in */
80 __u64 parent_root; /* in */
82 __u64 reserved[4]; /* in */
83 } __attribute__ ((__packed__));
85 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
86 struct btrfs_ioctl_send_args_32)
89 static int btrfs_clone(struct inode *src, struct inode *inode,
90 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
93 /* Mask out flags that are inappropriate for the given type of inode. */
94 static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
97 if (S_ISDIR(inode->i_mode))
99 else if (S_ISREG(inode->i_mode))
100 return flags & ~FS_DIRSYNC_FL;
102 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
106 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
109 static unsigned int btrfs_inode_flags_to_fsflags(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_sync_inode_flags_to_i_flags(struct inode *inode)
141 struct btrfs_inode *binode = BTRFS_I(inode);
142 unsigned int new_fl = 0;
144 if (binode->flags & BTRFS_INODE_SYNC)
146 if (binode->flags & BTRFS_INODE_IMMUTABLE)
147 new_fl |= S_IMMUTABLE;
148 if (binode->flags & BTRFS_INODE_APPEND)
150 if (binode->flags & BTRFS_INODE_NOATIME)
152 if (binode->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 *binode = BTRFS_I(file_inode(file));
163 unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags);
165 if (copy_to_user(arg, &flags, sizeof(flags)))
170 /* Check if @flags are a supported and valid set of FS_*_FL flags */
171 static int check_fsflags(unsigned int flags)
173 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
174 FS_NOATIME_FL | FS_NODUMP_FL | \
175 FS_SYNC_FL | FS_DIRSYNC_FL | \
176 FS_NOCOMP_FL | FS_COMPR_FL |
180 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
186 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
188 struct inode *inode = file_inode(file);
189 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
190 struct btrfs_inode *binode = BTRFS_I(inode);
191 struct btrfs_root *root = binode->root;
192 struct btrfs_trans_handle *trans;
193 unsigned int fsflags, old_fsflags;
195 const char *comp = NULL;
196 u32 binode_flags = binode->flags;
198 if (!inode_owner_or_capable(inode))
201 if (btrfs_root_readonly(root))
204 if (copy_from_user(&fsflags, arg, sizeof(fsflags)))
207 ret = check_fsflags(fsflags);
211 ret = mnt_want_write_file(file);
217 fsflags = btrfs_mask_fsflags_for_type(inode, fsflags);
218 old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags);
219 ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
223 if (fsflags & FS_SYNC_FL)
224 binode_flags |= BTRFS_INODE_SYNC;
226 binode_flags &= ~BTRFS_INODE_SYNC;
227 if (fsflags & FS_IMMUTABLE_FL)
228 binode_flags |= BTRFS_INODE_IMMUTABLE;
230 binode_flags &= ~BTRFS_INODE_IMMUTABLE;
231 if (fsflags & FS_APPEND_FL)
232 binode_flags |= BTRFS_INODE_APPEND;
234 binode_flags &= ~BTRFS_INODE_APPEND;
235 if (fsflags & FS_NODUMP_FL)
236 binode_flags |= BTRFS_INODE_NODUMP;
238 binode_flags &= ~BTRFS_INODE_NODUMP;
239 if (fsflags & FS_NOATIME_FL)
240 binode_flags |= BTRFS_INODE_NOATIME;
242 binode_flags &= ~BTRFS_INODE_NOATIME;
243 if (fsflags & FS_DIRSYNC_FL)
244 binode_flags |= BTRFS_INODE_DIRSYNC;
246 binode_flags &= ~BTRFS_INODE_DIRSYNC;
247 if (fsflags & FS_NOCOW_FL) {
248 if (S_ISREG(inode->i_mode)) {
250 * It's safe to turn csums off here, no extents exist.
251 * Otherwise we want the flag to reflect the real COW
252 * status of the file and will not set it.
254 if (inode->i_size == 0)
255 binode_flags |= BTRFS_INODE_NODATACOW |
256 BTRFS_INODE_NODATASUM;
258 binode_flags |= BTRFS_INODE_NODATACOW;
262 * Revert back under same assumptions as above
264 if (S_ISREG(inode->i_mode)) {
265 if (inode->i_size == 0)
266 binode_flags &= ~(BTRFS_INODE_NODATACOW |
267 BTRFS_INODE_NODATASUM);
269 binode_flags &= ~BTRFS_INODE_NODATACOW;
274 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
275 * flag may be changed automatically if compression code won't make
278 if (fsflags & FS_NOCOMP_FL) {
279 binode_flags &= ~BTRFS_INODE_COMPRESS;
280 binode_flags |= BTRFS_INODE_NOCOMPRESS;
281 } else if (fsflags & FS_COMPR_FL) {
283 if (IS_SWAPFILE(inode)) {
288 binode_flags |= BTRFS_INODE_COMPRESS;
289 binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
291 comp = btrfs_compress_type2str(fs_info->compress_type);
292 if (!comp || comp[0] == 0)
293 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
295 binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
302 trans = btrfs_start_transaction(root, 3);
304 ret = PTR_ERR(trans);
309 ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
312 btrfs_abort_transaction(trans, ret);
316 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
318 if (ret && ret != -ENODATA) {
319 btrfs_abort_transaction(trans, ret);
324 binode->flags = binode_flags;
325 btrfs_sync_inode_flags_to_i_flags(inode);
326 inode_inc_iversion(inode);
327 inode->i_ctime = current_time(inode);
328 ret = btrfs_update_inode(trans, root, inode);
331 btrfs_end_transaction(trans);
334 mnt_drop_write_file(file);
339 * Translate btrfs internal inode flags to xflags as expected by the
340 * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are
343 static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags)
345 unsigned int xflags = 0;
347 if (flags & BTRFS_INODE_APPEND)
348 xflags |= FS_XFLAG_APPEND;
349 if (flags & BTRFS_INODE_IMMUTABLE)
350 xflags |= FS_XFLAG_IMMUTABLE;
351 if (flags & BTRFS_INODE_NOATIME)
352 xflags |= FS_XFLAG_NOATIME;
353 if (flags & BTRFS_INODE_NODUMP)
354 xflags |= FS_XFLAG_NODUMP;
355 if (flags & BTRFS_INODE_SYNC)
356 xflags |= FS_XFLAG_SYNC;
361 /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */
362 static int check_xflags(unsigned int flags)
364 if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME |
365 FS_XFLAG_NODUMP | FS_XFLAG_SYNC))
371 * Set the xflags from the internal inode flags. The remaining items of fsxattr
374 static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg)
376 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
379 simple_fill_fsxattr(&fa, btrfs_inode_flags_to_xflags(binode->flags));
380 if (copy_to_user(arg, &fa, sizeof(fa)))
386 static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg)
388 struct inode *inode = file_inode(file);
389 struct btrfs_inode *binode = BTRFS_I(inode);
390 struct btrfs_root *root = binode->root;
391 struct btrfs_trans_handle *trans;
392 struct fsxattr fa, old_fa;
394 unsigned old_i_flags;
397 if (!inode_owner_or_capable(inode))
400 if (btrfs_root_readonly(root))
403 if (copy_from_user(&fa, arg, sizeof(fa)))
406 ret = check_xflags(fa.fsx_xflags);
410 if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0)
413 ret = mnt_want_write_file(file);
419 old_flags = binode->flags;
420 old_i_flags = inode->i_flags;
422 simple_fill_fsxattr(&old_fa,
423 btrfs_inode_flags_to_xflags(binode->flags));
424 ret = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
428 if (fa.fsx_xflags & FS_XFLAG_SYNC)
429 binode->flags |= BTRFS_INODE_SYNC;
431 binode->flags &= ~BTRFS_INODE_SYNC;
432 if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE)
433 binode->flags |= BTRFS_INODE_IMMUTABLE;
435 binode->flags &= ~BTRFS_INODE_IMMUTABLE;
436 if (fa.fsx_xflags & FS_XFLAG_APPEND)
437 binode->flags |= BTRFS_INODE_APPEND;
439 binode->flags &= ~BTRFS_INODE_APPEND;
440 if (fa.fsx_xflags & FS_XFLAG_NODUMP)
441 binode->flags |= BTRFS_INODE_NODUMP;
443 binode->flags &= ~BTRFS_INODE_NODUMP;
444 if (fa.fsx_xflags & FS_XFLAG_NOATIME)
445 binode->flags |= BTRFS_INODE_NOATIME;
447 binode->flags &= ~BTRFS_INODE_NOATIME;
449 /* 1 item for the inode */
450 trans = btrfs_start_transaction(root, 1);
452 ret = PTR_ERR(trans);
456 btrfs_sync_inode_flags_to_i_flags(inode);
457 inode_inc_iversion(inode);
458 inode->i_ctime = current_time(inode);
459 ret = btrfs_update_inode(trans, root, inode);
461 btrfs_end_transaction(trans);
465 binode->flags = old_flags;
466 inode->i_flags = old_i_flags;
470 mnt_drop_write_file(file);
475 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
477 struct inode *inode = file_inode(file);
479 return put_user(inode->i_generation, arg);
482 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
485 struct btrfs_device *device;
486 struct request_queue *q;
487 struct fstrim_range range;
488 u64 minlen = ULLONG_MAX;
492 if (!capable(CAP_SYS_ADMIN))
496 * If the fs is mounted with nologreplay, which requires it to be
497 * mounted in RO mode as well, we can not allow discard on free space
498 * inside block groups, because log trees refer to extents that are not
499 * pinned in a block group's free space cache (pinning the extents is
500 * precisely the first phase of replaying a log tree).
502 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
506 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
510 q = bdev_get_queue(device->bdev);
511 if (blk_queue_discard(q)) {
513 minlen = min_t(u64, q->limits.discard_granularity,
521 if (copy_from_user(&range, arg, sizeof(range)))
525 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
526 * block group is in the logical address space, which can be any
527 * sectorsize aligned bytenr in the range [0, U64_MAX].
529 if (range.len < fs_info->sb->s_blocksize)
532 range.minlen = max(range.minlen, minlen);
533 ret = btrfs_trim_fs(fs_info, &range);
537 if (copy_to_user(arg, &range, sizeof(range)))
543 int __pure btrfs_is_empty_uuid(u8 *uuid)
547 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
554 static noinline int create_subvol(struct inode *dir,
555 struct dentry *dentry,
556 const char *name, int namelen,
558 struct btrfs_qgroup_inherit *inherit)
560 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
561 struct btrfs_trans_handle *trans;
562 struct btrfs_key key;
563 struct btrfs_root_item *root_item;
564 struct btrfs_inode_item *inode_item;
565 struct extent_buffer *leaf;
566 struct btrfs_root *root = BTRFS_I(dir)->root;
567 struct btrfs_root *new_root;
568 struct btrfs_block_rsv block_rsv;
569 struct timespec64 cur_time = current_time(dir);
574 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
578 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
582 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
587 * Don't create subvolume whose level is not zero. Or qgroup will be
588 * screwed up since it assumes subvolume qgroup's level to be 0.
590 if (btrfs_qgroup_level(objectid)) {
595 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
597 * The same as the snapshot creation, please see the comment
598 * of create_snapshot().
600 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false);
604 trans = btrfs_start_transaction(root, 0);
606 ret = PTR_ERR(trans);
607 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
610 trans->block_rsv = &block_rsv;
611 trans->bytes_reserved = block_rsv.size;
613 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
617 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
623 btrfs_mark_buffer_dirty(leaf);
625 inode_item = &root_item->inode;
626 btrfs_set_stack_inode_generation(inode_item, 1);
627 btrfs_set_stack_inode_size(inode_item, 3);
628 btrfs_set_stack_inode_nlink(inode_item, 1);
629 btrfs_set_stack_inode_nbytes(inode_item,
631 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
633 btrfs_set_root_flags(root_item, 0);
634 btrfs_set_root_limit(root_item, 0);
635 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
637 btrfs_set_root_bytenr(root_item, leaf->start);
638 btrfs_set_root_generation(root_item, trans->transid);
639 btrfs_set_root_level(root_item, 0);
640 btrfs_set_root_refs(root_item, 1);
641 btrfs_set_root_used(root_item, leaf->len);
642 btrfs_set_root_last_snapshot(root_item, 0);
644 btrfs_set_root_generation_v2(root_item,
645 btrfs_root_generation(root_item));
646 uuid_le_gen(&new_uuid);
647 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
648 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
649 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
650 root_item->ctime = root_item->otime;
651 btrfs_set_root_ctransid(root_item, trans->transid);
652 btrfs_set_root_otransid(root_item, trans->transid);
654 btrfs_tree_unlock(leaf);
655 free_extent_buffer(leaf);
658 btrfs_set_root_dirid(root_item, new_dirid);
660 key.objectid = objectid;
662 key.type = BTRFS_ROOT_ITEM_KEY;
663 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
668 key.offset = (u64)-1;
669 new_root = btrfs_read_fs_root_no_name(fs_info, &key);
670 if (IS_ERR(new_root)) {
671 ret = PTR_ERR(new_root);
672 btrfs_abort_transaction(trans, ret);
676 btrfs_record_root_in_trans(trans, new_root);
678 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
680 /* We potentially lose an unused inode item here */
681 btrfs_abort_transaction(trans, ret);
685 mutex_lock(&new_root->objectid_mutex);
686 new_root->highest_objectid = new_dirid;
687 mutex_unlock(&new_root->objectid_mutex);
690 * insert the directory item
692 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
694 btrfs_abort_transaction(trans, ret);
698 ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
699 BTRFS_FT_DIR, index);
701 btrfs_abort_transaction(trans, ret);
705 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
706 ret = btrfs_update_inode(trans, root, dir);
708 btrfs_abort_transaction(trans, ret);
712 ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
713 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
715 btrfs_abort_transaction(trans, ret);
719 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
720 BTRFS_UUID_KEY_SUBVOL, objectid);
722 btrfs_abort_transaction(trans, ret);
726 trans->block_rsv = NULL;
727 trans->bytes_reserved = 0;
728 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
731 *async_transid = trans->transid;
732 err = btrfs_commit_transaction_async(trans, 1);
734 err = btrfs_commit_transaction(trans);
736 err = btrfs_commit_transaction(trans);
742 inode = btrfs_lookup_dentry(dir, dentry);
744 return PTR_ERR(inode);
745 d_instantiate(dentry, inode);
754 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
755 struct dentry *dentry,
756 u64 *async_transid, bool readonly,
757 struct btrfs_qgroup_inherit *inherit)
759 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
761 struct btrfs_pending_snapshot *pending_snapshot;
762 struct btrfs_trans_handle *trans;
764 bool snapshot_force_cow = false;
766 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
769 if (atomic_read(&root->nr_swapfiles)) {
771 "cannot snapshot subvolume with active swapfile");
775 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
776 if (!pending_snapshot)
779 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
781 pending_snapshot->path = btrfs_alloc_path();
782 if (!pending_snapshot->root_item || !pending_snapshot->path) {
788 * Force new buffered writes to reserve space even when NOCOW is
789 * possible. This is to avoid later writeback (running dealloc) to
790 * fallback to COW mode and unexpectedly fail with ENOSPC.
792 atomic_inc(&root->will_be_snapshotted);
793 smp_mb__after_atomic();
794 /* wait for no snapshot writes */
795 wait_event(root->subv_writers->wait,
796 percpu_counter_sum(&root->subv_writers->counter) == 0);
798 ret = btrfs_start_delalloc_snapshot(root);
803 * All previous writes have started writeback in NOCOW mode, so now
804 * we force future writes to fallback to COW mode during snapshot
807 atomic_inc(&root->snapshot_force_cow);
808 snapshot_force_cow = true;
810 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
812 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
813 BTRFS_BLOCK_RSV_TEMP);
815 * 1 - parent dir inode
818 * 2 - root ref/backref
819 * 1 - root of snapshot
822 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
823 &pending_snapshot->block_rsv, 8,
828 pending_snapshot->dentry = dentry;
829 pending_snapshot->root = root;
830 pending_snapshot->readonly = readonly;
831 pending_snapshot->dir = dir;
832 pending_snapshot->inherit = inherit;
834 trans = btrfs_start_transaction(root, 0);
836 ret = PTR_ERR(trans);
840 spin_lock(&fs_info->trans_lock);
841 list_add(&pending_snapshot->list,
842 &trans->transaction->pending_snapshots);
843 spin_unlock(&fs_info->trans_lock);
845 *async_transid = trans->transid;
846 ret = btrfs_commit_transaction_async(trans, 1);
848 ret = btrfs_commit_transaction(trans);
850 ret = btrfs_commit_transaction(trans);
855 ret = pending_snapshot->error;
859 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
863 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
865 ret = PTR_ERR(inode);
869 d_instantiate(dentry, inode);
872 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
874 if (snapshot_force_cow)
875 atomic_dec(&root->snapshot_force_cow);
876 if (atomic_dec_and_test(&root->will_be_snapshotted))
877 wake_up_var(&root->will_be_snapshotted);
879 kfree(pending_snapshot->root_item);
880 btrfs_free_path(pending_snapshot->path);
881 kfree(pending_snapshot);
886 /* copy of may_delete in fs/namei.c()
887 * Check whether we can remove a link victim from directory dir, check
888 * whether the type of victim is right.
889 * 1. We can't do it if dir is read-only (done in permission())
890 * 2. We should have write and exec permissions on dir
891 * 3. We can't remove anything from append-only dir
892 * 4. We can't do anything with immutable dir (done in permission())
893 * 5. If the sticky bit on dir is set we should either
894 * a. be owner of dir, or
895 * b. be owner of victim, or
896 * c. have CAP_FOWNER capability
897 * 6. If the victim is append-only or immutable we can't do anything with
898 * links pointing to it.
899 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
900 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
901 * 9. We can't remove a root or mountpoint.
902 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
903 * nfs_async_unlink().
906 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
910 if (d_really_is_negative(victim))
913 BUG_ON(d_inode(victim->d_parent) != dir);
914 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
916 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
921 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
922 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
925 if (!d_is_dir(victim))
929 } else if (d_is_dir(victim))
933 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
938 /* copy of may_create in fs/namei.c() */
939 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
941 if (d_really_is_positive(child))
945 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
949 * Create a new subvolume below @parent. This is largely modeled after
950 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
951 * inside this filesystem so it's quite a bit simpler.
953 static noinline int btrfs_mksubvol(const struct path *parent,
954 const char *name, int namelen,
955 struct btrfs_root *snap_src,
956 u64 *async_transid, bool readonly,
957 struct btrfs_qgroup_inherit *inherit)
959 struct inode *dir = d_inode(parent->dentry);
960 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
961 struct dentry *dentry;
964 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
968 dentry = lookup_one_len(name, parent->dentry, namelen);
969 error = PTR_ERR(dentry);
973 error = btrfs_may_create(dir, dentry);
978 * even if this name doesn't exist, we may get hash collisions.
979 * check for them now when we can safely fail
981 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
987 down_read(&fs_info->subvol_sem);
989 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
993 error = create_snapshot(snap_src, dir, dentry,
994 async_transid, readonly, inherit);
996 error = create_subvol(dir, dentry, name, namelen,
997 async_transid, inherit);
1000 fsnotify_mkdir(dir, dentry);
1002 up_read(&fs_info->subvol_sem);
1011 * When we're defragging a range, we don't want to kick it off again
1012 * if it is really just waiting for delalloc to send it down.
1013 * If we find a nice big extent or delalloc range for the bytes in the
1014 * file you want to defrag, we return 0 to let you know to skip this
1017 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
1019 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1020 struct extent_map *em = NULL;
1021 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1024 read_lock(&em_tree->lock);
1025 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
1026 read_unlock(&em_tree->lock);
1029 end = extent_map_end(em);
1030 free_extent_map(em);
1031 if (end - offset > thresh)
1034 /* if we already have a nice delalloc here, just stop */
1036 end = count_range_bits(io_tree, &offset, offset + thresh,
1037 thresh, EXTENT_DELALLOC, 1);
1044 * helper function to walk through a file and find extents
1045 * newer than a specific transid, and smaller than thresh.
1047 * This is used by the defragging code to find new and small
1050 static int find_new_extents(struct btrfs_root *root,
1051 struct inode *inode, u64 newer_than,
1052 u64 *off, u32 thresh)
1054 struct btrfs_path *path;
1055 struct btrfs_key min_key;
1056 struct extent_buffer *leaf;
1057 struct btrfs_file_extent_item *extent;
1060 u64 ino = btrfs_ino(BTRFS_I(inode));
1062 path = btrfs_alloc_path();
1066 min_key.objectid = ino;
1067 min_key.type = BTRFS_EXTENT_DATA_KEY;
1068 min_key.offset = *off;
1071 ret = btrfs_search_forward(root, &min_key, path, newer_than);
1075 if (min_key.objectid != ino)
1077 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1080 leaf = path->nodes[0];
1081 extent = btrfs_item_ptr(leaf, path->slots[0],
1082 struct btrfs_file_extent_item);
1084 type = btrfs_file_extent_type(leaf, extent);
1085 if (type == BTRFS_FILE_EXTENT_REG &&
1086 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1087 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1088 *off = min_key.offset;
1089 btrfs_free_path(path);
1094 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1095 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1099 if (min_key.offset == (u64)-1)
1103 btrfs_release_path(path);
1106 btrfs_free_path(path);
1110 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1112 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1113 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1114 struct extent_map *em;
1115 u64 len = PAGE_SIZE;
1118 * hopefully we have this extent in the tree already, try without
1119 * the full extent lock
1121 read_lock(&em_tree->lock);
1122 em = lookup_extent_mapping(em_tree, start, len);
1123 read_unlock(&em_tree->lock);
1126 struct extent_state *cached = NULL;
1127 u64 end = start + len - 1;
1129 /* get the big lock and read metadata off disk */
1130 lock_extent_bits(io_tree, start, end, &cached);
1131 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len);
1132 unlock_extent_cached(io_tree, start, end, &cached);
1141 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1143 struct extent_map *next;
1146 /* this is the last extent */
1147 if (em->start + em->len >= i_size_read(inode))
1150 next = defrag_lookup_extent(inode, em->start + em->len);
1151 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1153 else if ((em->block_start + em->block_len == next->block_start) &&
1154 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1157 free_extent_map(next);
1161 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1162 u64 *last_len, u64 *skip, u64 *defrag_end,
1165 struct extent_map *em;
1167 bool next_mergeable = true;
1168 bool prev_mergeable = true;
1171 * make sure that once we start defragging an extent, we keep on
1174 if (start < *defrag_end)
1179 em = defrag_lookup_extent(inode, start);
1183 /* this will cover holes, and inline extents */
1184 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1190 prev_mergeable = false;
1192 next_mergeable = defrag_check_next_extent(inode, em);
1194 * we hit a real extent, if it is big or the next extent is not a
1195 * real extent, don't bother defragging it
1197 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1198 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1202 * last_len ends up being a counter of how many bytes we've defragged.
1203 * every time we choose not to defrag an extent, we reset *last_len
1204 * so that the next tiny extent will force a defrag.
1206 * The end result of this is that tiny extents before a single big
1207 * extent will force at least part of that big extent to be defragged.
1210 *defrag_end = extent_map_end(em);
1213 *skip = extent_map_end(em);
1217 free_extent_map(em);
1222 * it doesn't do much good to defrag one or two pages
1223 * at a time. This pulls in a nice chunk of pages
1224 * to COW and defrag.
1226 * It also makes sure the delalloc code has enough
1227 * dirty data to avoid making new small extents as part
1230 * It's a good idea to start RA on this range
1231 * before calling this.
1233 static int cluster_pages_for_defrag(struct inode *inode,
1234 struct page **pages,
1235 unsigned long start_index,
1236 unsigned long num_pages)
1238 unsigned long file_end;
1239 u64 isize = i_size_read(inode);
1246 struct btrfs_ordered_extent *ordered;
1247 struct extent_state *cached_state = NULL;
1248 struct extent_io_tree *tree;
1249 struct extent_changeset *data_reserved = NULL;
1250 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1252 file_end = (isize - 1) >> PAGE_SHIFT;
1253 if (!isize || start_index > file_end)
1256 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1258 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
1259 start_index << PAGE_SHIFT,
1260 page_cnt << PAGE_SHIFT);
1264 tree = &BTRFS_I(inode)->io_tree;
1266 /* step one, lock all the pages */
1267 for (i = 0; i < page_cnt; i++) {
1270 page = find_or_create_page(inode->i_mapping,
1271 start_index + i, mask);
1275 page_start = page_offset(page);
1276 page_end = page_start + PAGE_SIZE - 1;
1278 lock_extent_bits(tree, page_start, page_end,
1280 ordered = btrfs_lookup_ordered_extent(inode,
1282 unlock_extent_cached(tree, page_start, page_end,
1288 btrfs_start_ordered_extent(inode, ordered, 1);
1289 btrfs_put_ordered_extent(ordered);
1292 * we unlocked the page above, so we need check if
1293 * it was released or not.
1295 if (page->mapping != inode->i_mapping) {
1302 if (!PageUptodate(page)) {
1303 btrfs_readpage(NULL, page);
1305 if (!PageUptodate(page)) {
1313 if (page->mapping != inode->i_mapping) {
1325 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1329 * so now we have a nice long stream of locked
1330 * and up to date pages, lets wait on them
1332 for (i = 0; i < i_done; i++)
1333 wait_on_page_writeback(pages[i]);
1335 page_start = page_offset(pages[0]);
1336 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1338 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1339 page_start, page_end - 1, &cached_state);
1340 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1341 page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
1342 EXTENT_DEFRAG, 0, 0, &cached_state);
1344 if (i_done != page_cnt) {
1345 spin_lock(&BTRFS_I(inode)->lock);
1346 btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
1347 spin_unlock(&BTRFS_I(inode)->lock);
1348 btrfs_delalloc_release_space(inode, data_reserved,
1349 start_index << PAGE_SHIFT,
1350 (page_cnt - i_done) << PAGE_SHIFT, true);
1354 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1357 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1358 page_start, page_end - 1, &cached_state);
1360 for (i = 0; i < i_done; i++) {
1361 clear_page_dirty_for_io(pages[i]);
1362 ClearPageChecked(pages[i]);
1363 set_page_extent_mapped(pages[i]);
1364 set_page_dirty(pages[i]);
1365 unlock_page(pages[i]);
1368 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1369 extent_changeset_free(data_reserved);
1372 for (i = 0; i < i_done; i++) {
1373 unlock_page(pages[i]);
1376 btrfs_delalloc_release_space(inode, data_reserved,
1377 start_index << PAGE_SHIFT,
1378 page_cnt << PAGE_SHIFT, true);
1379 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1380 extent_changeset_free(data_reserved);
1385 int btrfs_defrag_file(struct inode *inode, struct file *file,
1386 struct btrfs_ioctl_defrag_range_args *range,
1387 u64 newer_than, unsigned long max_to_defrag)
1389 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1390 struct btrfs_root *root = BTRFS_I(inode)->root;
1391 struct file_ra_state *ra = NULL;
1392 unsigned long last_index;
1393 u64 isize = i_size_read(inode);
1397 u64 newer_off = range->start;
1399 unsigned long ra_index = 0;
1401 int defrag_count = 0;
1402 int compress_type = BTRFS_COMPRESS_ZLIB;
1403 u32 extent_thresh = range->extent_thresh;
1404 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1405 unsigned long cluster = max_cluster;
1406 u64 new_align = ~((u64)SZ_128K - 1);
1407 struct page **pages = NULL;
1408 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1413 if (range->start >= isize)
1417 if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES)
1419 if (range->compress_type)
1420 compress_type = range->compress_type;
1423 if (extent_thresh == 0)
1424 extent_thresh = SZ_256K;
1427 * If we were not given a file, allocate a readahead context. As
1428 * readahead is just an optimization, defrag will work without it so
1429 * we don't error out.
1432 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1434 file_ra_state_init(ra, inode->i_mapping);
1439 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1445 /* find the last page to defrag */
1446 if (range->start + range->len > range->start) {
1447 last_index = min_t(u64, isize - 1,
1448 range->start + range->len - 1) >> PAGE_SHIFT;
1450 last_index = (isize - 1) >> PAGE_SHIFT;
1454 ret = find_new_extents(root, inode, newer_than,
1455 &newer_off, SZ_64K);
1457 range->start = newer_off;
1459 * we always align our defrag to help keep
1460 * the extents in the file evenly spaced
1462 i = (newer_off & new_align) >> PAGE_SHIFT;
1466 i = range->start >> PAGE_SHIFT;
1469 max_to_defrag = last_index - i + 1;
1472 * make writeback starts from i, so the defrag range can be
1473 * written sequentially.
1475 if (i < inode->i_mapping->writeback_index)
1476 inode->i_mapping->writeback_index = i;
1478 while (i <= last_index && defrag_count < max_to_defrag &&
1479 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1481 * make sure we stop running if someone unmounts
1484 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1487 if (btrfs_defrag_cancelled(fs_info)) {
1488 btrfs_debug(fs_info, "defrag_file cancelled");
1493 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1494 extent_thresh, &last_len, &skip,
1495 &defrag_end, do_compress)){
1498 * the should_defrag function tells us how much to skip
1499 * bump our counter by the suggested amount
1501 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1502 i = max(i + 1, next);
1507 cluster = (PAGE_ALIGN(defrag_end) >>
1509 cluster = min(cluster, max_cluster);
1511 cluster = max_cluster;
1514 if (i + cluster > ra_index) {
1515 ra_index = max(i, ra_index);
1517 page_cache_sync_readahead(inode->i_mapping, ra,
1518 file, ra_index, cluster);
1519 ra_index += cluster;
1523 if (IS_SWAPFILE(inode)) {
1527 BTRFS_I(inode)->defrag_compress = compress_type;
1528 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1531 inode_unlock(inode);
1535 defrag_count += ret;
1536 balance_dirty_pages_ratelimited(inode->i_mapping);
1537 inode_unlock(inode);
1540 if (newer_off == (u64)-1)
1546 newer_off = max(newer_off + 1,
1547 (u64)i << PAGE_SHIFT);
1549 ret = find_new_extents(root, inode, newer_than,
1550 &newer_off, SZ_64K);
1552 range->start = newer_off;
1553 i = (newer_off & new_align) >> PAGE_SHIFT;
1560 last_len += ret << PAGE_SHIFT;
1568 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1569 filemap_flush(inode->i_mapping);
1570 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1571 &BTRFS_I(inode)->runtime_flags))
1572 filemap_flush(inode->i_mapping);
1575 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1576 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1577 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1578 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1586 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1587 inode_unlock(inode);
1595 static noinline int btrfs_ioctl_resize(struct file *file,
1598 struct inode *inode = file_inode(file);
1599 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1603 struct btrfs_root *root = BTRFS_I(inode)->root;
1604 struct btrfs_ioctl_vol_args *vol_args;
1605 struct btrfs_trans_handle *trans;
1606 struct btrfs_device *device = NULL;
1609 char *devstr = NULL;
1613 if (!capable(CAP_SYS_ADMIN))
1616 ret = mnt_want_write_file(file);
1620 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
1621 mnt_drop_write_file(file);
1622 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1625 vol_args = memdup_user(arg, sizeof(*vol_args));
1626 if (IS_ERR(vol_args)) {
1627 ret = PTR_ERR(vol_args);
1631 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1633 sizestr = vol_args->name;
1634 devstr = strchr(sizestr, ':');
1636 sizestr = devstr + 1;
1638 devstr = vol_args->name;
1639 ret = kstrtoull(devstr, 10, &devid);
1646 btrfs_info(fs_info, "resizing devid %llu", devid);
1649 device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true);
1651 btrfs_info(fs_info, "resizer unable to find device %llu",
1657 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1659 "resizer unable to apply on readonly device %llu",
1665 if (!strcmp(sizestr, "max"))
1666 new_size = device->bdev->bd_inode->i_size;
1668 if (sizestr[0] == '-') {
1671 } else if (sizestr[0] == '+') {
1675 new_size = memparse(sizestr, &retptr);
1676 if (*retptr != '\0' || new_size == 0) {
1682 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1687 old_size = btrfs_device_get_total_bytes(device);
1690 if (new_size > old_size) {
1694 new_size = old_size - new_size;
1695 } else if (mod > 0) {
1696 if (new_size > ULLONG_MAX - old_size) {
1700 new_size = old_size + new_size;
1703 if (new_size < SZ_256M) {
1707 if (new_size > device->bdev->bd_inode->i_size) {
1712 new_size = round_down(new_size, fs_info->sectorsize);
1714 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1715 rcu_str_deref(device->name), new_size);
1717 if (new_size > old_size) {
1718 trans = btrfs_start_transaction(root, 0);
1719 if (IS_ERR(trans)) {
1720 ret = PTR_ERR(trans);
1723 ret = btrfs_grow_device(trans, device, new_size);
1724 btrfs_commit_transaction(trans);
1725 } else if (new_size < old_size) {
1726 ret = btrfs_shrink_device(device, new_size);
1727 } /* equal, nothing need to do */
1732 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
1733 mnt_drop_write_file(file);
1737 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1738 const char *name, unsigned long fd, int subvol,
1739 u64 *transid, bool readonly,
1740 struct btrfs_qgroup_inherit *inherit)
1745 if (!S_ISDIR(file_inode(file)->i_mode))
1748 ret = mnt_want_write_file(file);
1752 namelen = strlen(name);
1753 if (strchr(name, '/')) {
1755 goto out_drop_write;
1758 if (name[0] == '.' &&
1759 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1761 goto out_drop_write;
1765 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1766 NULL, transid, readonly, inherit);
1768 struct fd src = fdget(fd);
1769 struct inode *src_inode;
1772 goto out_drop_write;
1775 src_inode = file_inode(src.file);
1776 if (src_inode->i_sb != file_inode(file)->i_sb) {
1777 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1778 "Snapshot src from another FS");
1780 } else if (!inode_owner_or_capable(src_inode)) {
1782 * Subvolume creation is not restricted, but snapshots
1783 * are limited to own subvolumes only
1787 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1788 BTRFS_I(src_inode)->root,
1789 transid, readonly, inherit);
1794 mnt_drop_write_file(file);
1799 static noinline int btrfs_ioctl_snap_create(struct file *file,
1800 void __user *arg, int subvol)
1802 struct btrfs_ioctl_vol_args *vol_args;
1805 if (!S_ISDIR(file_inode(file)->i_mode))
1808 vol_args = memdup_user(arg, sizeof(*vol_args));
1809 if (IS_ERR(vol_args))
1810 return PTR_ERR(vol_args);
1811 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1813 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1814 vol_args->fd, subvol,
1821 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1822 void __user *arg, int subvol)
1824 struct btrfs_ioctl_vol_args_v2 *vol_args;
1828 bool readonly = false;
1829 struct btrfs_qgroup_inherit *inherit = NULL;
1831 if (!S_ISDIR(file_inode(file)->i_mode))
1834 vol_args = memdup_user(arg, sizeof(*vol_args));
1835 if (IS_ERR(vol_args))
1836 return PTR_ERR(vol_args);
1837 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1839 if (vol_args->flags &
1840 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1841 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1846 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1847 struct inode *inode = file_inode(file);
1848 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1851 "SNAP_CREATE_V2 ioctl with CREATE_ASYNC is deprecated and will be removed in kernel 5.7");
1855 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1857 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1858 if (vol_args->size > PAGE_SIZE) {
1862 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1863 if (IS_ERR(inherit)) {
1864 ret = PTR_ERR(inherit);
1869 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1870 vol_args->fd, subvol, ptr,
1875 if (ptr && copy_to_user(arg +
1876 offsetof(struct btrfs_ioctl_vol_args_v2,
1888 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1891 struct inode *inode = file_inode(file);
1892 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1893 struct btrfs_root *root = BTRFS_I(inode)->root;
1897 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1900 down_read(&fs_info->subvol_sem);
1901 if (btrfs_root_readonly(root))
1902 flags |= BTRFS_SUBVOL_RDONLY;
1903 up_read(&fs_info->subvol_sem);
1905 if (copy_to_user(arg, &flags, sizeof(flags)))
1911 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1914 struct inode *inode = file_inode(file);
1915 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1916 struct btrfs_root *root = BTRFS_I(inode)->root;
1917 struct btrfs_trans_handle *trans;
1922 if (!inode_owner_or_capable(inode))
1925 ret = mnt_want_write_file(file);
1929 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1931 goto out_drop_write;
1934 if (copy_from_user(&flags, arg, sizeof(flags))) {
1936 goto out_drop_write;
1939 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1941 goto out_drop_write;
1944 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1946 goto out_drop_write;
1949 down_write(&fs_info->subvol_sem);
1952 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1955 root_flags = btrfs_root_flags(&root->root_item);
1956 if (flags & BTRFS_SUBVOL_RDONLY) {
1957 btrfs_set_root_flags(&root->root_item,
1958 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1961 * Block RO -> RW transition if this subvolume is involved in
1964 spin_lock(&root->root_item_lock);
1965 if (root->send_in_progress == 0) {
1966 btrfs_set_root_flags(&root->root_item,
1967 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1968 spin_unlock(&root->root_item_lock);
1970 spin_unlock(&root->root_item_lock);
1972 "Attempt to set subvolume %llu read-write during send",
1973 root->root_key.objectid);
1979 trans = btrfs_start_transaction(root, 1);
1980 if (IS_ERR(trans)) {
1981 ret = PTR_ERR(trans);
1985 ret = btrfs_update_root(trans, fs_info->tree_root,
1986 &root->root_key, &root->root_item);
1988 btrfs_end_transaction(trans);
1992 ret = btrfs_commit_transaction(trans);
1996 btrfs_set_root_flags(&root->root_item, root_flags);
1998 up_write(&fs_info->subvol_sem);
2000 mnt_drop_write_file(file);
2005 static noinline int key_in_sk(struct btrfs_key *key,
2006 struct btrfs_ioctl_search_key *sk)
2008 struct btrfs_key test;
2011 test.objectid = sk->min_objectid;
2012 test.type = sk->min_type;
2013 test.offset = sk->min_offset;
2015 ret = btrfs_comp_cpu_keys(key, &test);
2019 test.objectid = sk->max_objectid;
2020 test.type = sk->max_type;
2021 test.offset = sk->max_offset;
2023 ret = btrfs_comp_cpu_keys(key, &test);
2029 static noinline int copy_to_sk(struct btrfs_path *path,
2030 struct btrfs_key *key,
2031 struct btrfs_ioctl_search_key *sk,
2034 unsigned long *sk_offset,
2038 struct extent_buffer *leaf;
2039 struct btrfs_ioctl_search_header sh;
2040 struct btrfs_key test;
2041 unsigned long item_off;
2042 unsigned long item_len;
2048 leaf = path->nodes[0];
2049 slot = path->slots[0];
2050 nritems = btrfs_header_nritems(leaf);
2052 if (btrfs_header_generation(leaf) > sk->max_transid) {
2056 found_transid = btrfs_header_generation(leaf);
2058 for (i = slot; i < nritems; i++) {
2059 item_off = btrfs_item_ptr_offset(leaf, i);
2060 item_len = btrfs_item_size_nr(leaf, i);
2062 btrfs_item_key_to_cpu(leaf, key, i);
2063 if (!key_in_sk(key, sk))
2066 if (sizeof(sh) + item_len > *buf_size) {
2073 * return one empty item back for v1, which does not
2077 *buf_size = sizeof(sh) + item_len;
2082 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2087 sh.objectid = key->objectid;
2088 sh.offset = key->offset;
2089 sh.type = key->type;
2091 sh.transid = found_transid;
2093 /* copy search result header */
2094 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2099 *sk_offset += sizeof(sh);
2102 char __user *up = ubuf + *sk_offset;
2104 if (read_extent_buffer_to_user(leaf, up,
2105 item_off, item_len)) {
2110 *sk_offset += item_len;
2114 if (ret) /* -EOVERFLOW from above */
2117 if (*num_found >= sk->nr_items) {
2124 test.objectid = sk->max_objectid;
2125 test.type = sk->max_type;
2126 test.offset = sk->max_offset;
2127 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2129 else if (key->offset < (u64)-1)
2131 else if (key->type < (u8)-1) {
2134 } else if (key->objectid < (u64)-1) {
2142 * 0: all items from this leaf copied, continue with next
2143 * 1: * more items can be copied, but unused buffer is too small
2144 * * all items were found
2145 * Either way, it will stops the loop which iterates to the next
2147 * -EOVERFLOW: item was to large for buffer
2148 * -EFAULT: could not copy extent buffer back to userspace
2153 static noinline int search_ioctl(struct inode *inode,
2154 struct btrfs_ioctl_search_key *sk,
2158 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2159 struct btrfs_root *root;
2160 struct btrfs_key key;
2161 struct btrfs_path *path;
2164 unsigned long sk_offset = 0;
2166 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2167 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2171 path = btrfs_alloc_path();
2175 if (sk->tree_id == 0) {
2176 /* search the root of the inode that was passed */
2177 root = BTRFS_I(inode)->root;
2179 key.objectid = sk->tree_id;
2180 key.type = BTRFS_ROOT_ITEM_KEY;
2181 key.offset = (u64)-1;
2182 root = btrfs_read_fs_root_no_name(info, &key);
2184 btrfs_free_path(path);
2185 return PTR_ERR(root);
2189 key.objectid = sk->min_objectid;
2190 key.type = sk->min_type;
2191 key.offset = sk->min_offset;
2194 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2200 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2201 &sk_offset, &num_found);
2202 btrfs_release_path(path);
2210 sk->nr_items = num_found;
2211 btrfs_free_path(path);
2215 static noinline int btrfs_ioctl_tree_search(struct file *file,
2218 struct btrfs_ioctl_search_args __user *uargs;
2219 struct btrfs_ioctl_search_key sk;
2220 struct inode *inode;
2224 if (!capable(CAP_SYS_ADMIN))
2227 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2229 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2232 buf_size = sizeof(uargs->buf);
2234 inode = file_inode(file);
2235 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2238 * In the origin implementation an overflow is handled by returning a
2239 * search header with a len of zero, so reset ret.
2241 if (ret == -EOVERFLOW)
2244 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2249 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2252 struct btrfs_ioctl_search_args_v2 __user *uarg;
2253 struct btrfs_ioctl_search_args_v2 args;
2254 struct inode *inode;
2257 const size_t buf_limit = SZ_16M;
2259 if (!capable(CAP_SYS_ADMIN))
2262 /* copy search header and buffer size */
2263 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2264 if (copy_from_user(&args, uarg, sizeof(args)))
2267 buf_size = args.buf_size;
2269 /* limit result size to 16MB */
2270 if (buf_size > buf_limit)
2271 buf_size = buf_limit;
2273 inode = file_inode(file);
2274 ret = search_ioctl(inode, &args.key, &buf_size,
2275 (char __user *)(&uarg->buf[0]));
2276 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2278 else if (ret == -EOVERFLOW &&
2279 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2286 * Search INODE_REFs to identify path name of 'dirid' directory
2287 * in a 'tree_id' tree. and sets path name to 'name'.
2289 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2290 u64 tree_id, u64 dirid, char *name)
2292 struct btrfs_root *root;
2293 struct btrfs_key key;
2299 struct btrfs_inode_ref *iref;
2300 struct extent_buffer *l;
2301 struct btrfs_path *path;
2303 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2308 path = btrfs_alloc_path();
2312 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2314 key.objectid = tree_id;
2315 key.type = BTRFS_ROOT_ITEM_KEY;
2316 key.offset = (u64)-1;
2317 root = btrfs_read_fs_root_no_name(info, &key);
2319 ret = PTR_ERR(root);
2323 key.objectid = dirid;
2324 key.type = BTRFS_INODE_REF_KEY;
2325 key.offset = (u64)-1;
2328 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2332 ret = btrfs_previous_item(root, path, dirid,
2333 BTRFS_INODE_REF_KEY);
2343 slot = path->slots[0];
2344 btrfs_item_key_to_cpu(l, &key, slot);
2346 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2347 len = btrfs_inode_ref_name_len(l, iref);
2349 total_len += len + 1;
2351 ret = -ENAMETOOLONG;
2356 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2358 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2361 btrfs_release_path(path);
2362 key.objectid = key.offset;
2363 key.offset = (u64)-1;
2364 dirid = key.objectid;
2366 memmove(name, ptr, total_len);
2367 name[total_len] = '\0';
2370 btrfs_free_path(path);
2374 static int btrfs_search_path_in_tree_user(struct inode *inode,
2375 struct btrfs_ioctl_ino_lookup_user_args *args)
2377 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2378 struct super_block *sb = inode->i_sb;
2379 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2380 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2381 u64 dirid = args->dirid;
2382 unsigned long item_off;
2383 unsigned long item_len;
2384 struct btrfs_inode_ref *iref;
2385 struct btrfs_root_ref *rref;
2386 struct btrfs_root *root;
2387 struct btrfs_path *path;
2388 struct btrfs_key key, key2;
2389 struct extent_buffer *leaf;
2390 struct inode *temp_inode;
2397 path = btrfs_alloc_path();
2402 * If the bottom subvolume does not exist directly under upper_limit,
2403 * construct the path in from the bottom up.
2405 if (dirid != upper_limit.objectid) {
2406 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2408 key.objectid = treeid;
2409 key.type = BTRFS_ROOT_ITEM_KEY;
2410 key.offset = (u64)-1;
2411 root = btrfs_read_fs_root_no_name(fs_info, &key);
2413 ret = PTR_ERR(root);
2417 key.objectid = dirid;
2418 key.type = BTRFS_INODE_REF_KEY;
2419 key.offset = (u64)-1;
2421 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2424 } else if (ret > 0) {
2425 ret = btrfs_previous_item(root, path, dirid,
2426 BTRFS_INODE_REF_KEY);
2429 } else if (ret > 0) {
2435 leaf = path->nodes[0];
2436 slot = path->slots[0];
2437 btrfs_item_key_to_cpu(leaf, &key, slot);
2439 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2440 len = btrfs_inode_ref_name_len(leaf, iref);
2442 total_len += len + 1;
2443 if (ptr < args->path) {
2444 ret = -ENAMETOOLONG;
2449 read_extent_buffer(leaf, ptr,
2450 (unsigned long)(iref + 1), len);
2452 /* Check the read+exec permission of this directory */
2453 ret = btrfs_previous_item(root, path, dirid,
2454 BTRFS_INODE_ITEM_KEY);
2457 } else if (ret > 0) {
2462 leaf = path->nodes[0];
2463 slot = path->slots[0];
2464 btrfs_item_key_to_cpu(leaf, &key2, slot);
2465 if (key2.objectid != dirid) {
2470 temp_inode = btrfs_iget(sb, &key2, root);
2471 if (IS_ERR(temp_inode)) {
2472 ret = PTR_ERR(temp_inode);
2475 ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC);
2482 if (key.offset == upper_limit.objectid)
2484 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2489 btrfs_release_path(path);
2490 key.objectid = key.offset;
2491 key.offset = (u64)-1;
2492 dirid = key.objectid;
2495 memmove(args->path, ptr, total_len);
2496 args->path[total_len] = '\0';
2497 btrfs_release_path(path);
2500 /* Get the bottom subvolume's name from ROOT_REF */
2501 root = fs_info->tree_root;
2502 key.objectid = treeid;
2503 key.type = BTRFS_ROOT_REF_KEY;
2504 key.offset = args->treeid;
2505 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2508 } else if (ret > 0) {
2513 leaf = path->nodes[0];
2514 slot = path->slots[0];
2515 btrfs_item_key_to_cpu(leaf, &key, slot);
2517 item_off = btrfs_item_ptr_offset(leaf, slot);
2518 item_len = btrfs_item_size_nr(leaf, slot);
2519 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2520 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2521 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2526 /* Copy subvolume's name */
2527 item_off += sizeof(struct btrfs_root_ref);
2528 item_len -= sizeof(struct btrfs_root_ref);
2529 read_extent_buffer(leaf, args->name, item_off, item_len);
2530 args->name[item_len] = 0;
2533 btrfs_free_path(path);
2537 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2540 struct btrfs_ioctl_ino_lookup_args *args;
2541 struct inode *inode;
2544 args = memdup_user(argp, sizeof(*args));
2546 return PTR_ERR(args);
2548 inode = file_inode(file);
2551 * Unprivileged query to obtain the containing subvolume root id. The
2552 * path is reset so it's consistent with btrfs_search_path_in_tree.
2554 if (args->treeid == 0)
2555 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2557 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2562 if (!capable(CAP_SYS_ADMIN)) {
2567 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2568 args->treeid, args->objectid,
2572 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2580 * Version of ino_lookup ioctl (unprivileged)
2582 * The main differences from ino_lookup ioctl are:
2584 * 1. Read + Exec permission will be checked using inode_permission() during
2585 * path construction. -EACCES will be returned in case of failure.
2586 * 2. Path construction will be stopped at the inode number which corresponds
2587 * to the fd with which this ioctl is called. If constructed path does not
2588 * exist under fd's inode, -EACCES will be returned.
2589 * 3. The name of bottom subvolume is also searched and filled.
2591 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2593 struct btrfs_ioctl_ino_lookup_user_args *args;
2594 struct inode *inode;
2597 args = memdup_user(argp, sizeof(*args));
2599 return PTR_ERR(args);
2601 inode = file_inode(file);
2603 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2604 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2606 * The subvolume does not exist under fd with which this is
2613 ret = btrfs_search_path_in_tree_user(inode, args);
2615 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2622 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2623 static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2625 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2626 struct btrfs_fs_info *fs_info;
2627 struct btrfs_root *root;
2628 struct btrfs_path *path;
2629 struct btrfs_key key;
2630 struct btrfs_root_item *root_item;
2631 struct btrfs_root_ref *rref;
2632 struct extent_buffer *leaf;
2633 unsigned long item_off;
2634 unsigned long item_len;
2635 struct inode *inode;
2639 path = btrfs_alloc_path();
2643 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2645 btrfs_free_path(path);
2649 inode = file_inode(file);
2650 fs_info = BTRFS_I(inode)->root->fs_info;
2652 /* Get root_item of inode's subvolume */
2653 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2654 key.type = BTRFS_ROOT_ITEM_KEY;
2655 key.offset = (u64)-1;
2656 root = btrfs_read_fs_root_no_name(fs_info, &key);
2658 ret = PTR_ERR(root);
2661 root_item = &root->root_item;
2663 subvol_info->treeid = key.objectid;
2665 subvol_info->generation = btrfs_root_generation(root_item);
2666 subvol_info->flags = btrfs_root_flags(root_item);
2668 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2669 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2671 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2674 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2675 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2676 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2678 subvol_info->otransid = btrfs_root_otransid(root_item);
2679 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2680 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2682 subvol_info->stransid = btrfs_root_stransid(root_item);
2683 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2684 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2686 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2687 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2688 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2690 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2691 /* Search root tree for ROOT_BACKREF of this subvolume */
2692 root = fs_info->tree_root;
2694 key.type = BTRFS_ROOT_BACKREF_KEY;
2696 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2699 } else if (path->slots[0] >=
2700 btrfs_header_nritems(path->nodes[0])) {
2701 ret = btrfs_next_leaf(root, path);
2704 } else if (ret > 0) {
2710 leaf = path->nodes[0];
2711 slot = path->slots[0];
2712 btrfs_item_key_to_cpu(leaf, &key, slot);
2713 if (key.objectid == subvol_info->treeid &&
2714 key.type == BTRFS_ROOT_BACKREF_KEY) {
2715 subvol_info->parent_id = key.offset;
2717 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2718 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2720 item_off = btrfs_item_ptr_offset(leaf, slot)
2721 + sizeof(struct btrfs_root_ref);
2722 item_len = btrfs_item_size_nr(leaf, slot)
2723 - sizeof(struct btrfs_root_ref);
2724 read_extent_buffer(leaf, subvol_info->name,
2725 item_off, item_len);
2732 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2736 btrfs_free_path(path);
2737 kzfree(subvol_info);
2742 * Return ROOT_REF information of the subvolume containing this inode
2743 * except the subvolume name.
2745 static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2747 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2748 struct btrfs_root_ref *rref;
2749 struct btrfs_root *root;
2750 struct btrfs_path *path;
2751 struct btrfs_key key;
2752 struct extent_buffer *leaf;
2753 struct inode *inode;
2759 path = btrfs_alloc_path();
2763 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2764 if (IS_ERR(rootrefs)) {
2765 btrfs_free_path(path);
2766 return PTR_ERR(rootrefs);
2769 inode = file_inode(file);
2770 root = BTRFS_I(inode)->root->fs_info->tree_root;
2771 objectid = BTRFS_I(inode)->root->root_key.objectid;
2773 key.objectid = objectid;
2774 key.type = BTRFS_ROOT_REF_KEY;
2775 key.offset = rootrefs->min_treeid;
2778 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2781 } else if (path->slots[0] >=
2782 btrfs_header_nritems(path->nodes[0])) {
2783 ret = btrfs_next_leaf(root, path);
2786 } else if (ret > 0) {
2792 leaf = path->nodes[0];
2793 slot = path->slots[0];
2795 btrfs_item_key_to_cpu(leaf, &key, slot);
2796 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2801 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2806 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2807 rootrefs->rootref[found].treeid = key.offset;
2808 rootrefs->rootref[found].dirid =
2809 btrfs_root_ref_dirid(leaf, rref);
2812 ret = btrfs_next_item(root, path);
2815 } else if (ret > 0) {
2822 if (!ret || ret == -EOVERFLOW) {
2823 rootrefs->num_items = found;
2824 /* update min_treeid for next search */
2826 rootrefs->min_treeid =
2827 rootrefs->rootref[found - 1].treeid + 1;
2828 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2833 btrfs_free_path(path);
2838 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2841 struct dentry *parent = file->f_path.dentry;
2842 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2843 struct dentry *dentry;
2844 struct inode *dir = d_inode(parent);
2845 struct inode *inode;
2846 struct btrfs_root *root = BTRFS_I(dir)->root;
2847 struct btrfs_root *dest = NULL;
2848 struct btrfs_ioctl_vol_args *vol_args;
2852 if (!S_ISDIR(dir->i_mode))
2855 vol_args = memdup_user(arg, sizeof(*vol_args));
2856 if (IS_ERR(vol_args))
2857 return PTR_ERR(vol_args);
2859 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2860 namelen = strlen(vol_args->name);
2861 if (strchr(vol_args->name, '/') ||
2862 strncmp(vol_args->name, "..", namelen) == 0) {
2867 err = mnt_want_write_file(file);
2872 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2874 goto out_drop_write;
2875 dentry = lookup_one_len(vol_args->name, parent, namelen);
2876 if (IS_ERR(dentry)) {
2877 err = PTR_ERR(dentry);
2878 goto out_unlock_dir;
2881 if (d_really_is_negative(dentry)) {
2886 inode = d_inode(dentry);
2887 dest = BTRFS_I(inode)->root;
2888 if (!capable(CAP_SYS_ADMIN)) {
2890 * Regular user. Only allow this with a special mount
2891 * option, when the user has write+exec access to the
2892 * subvol root, and when rmdir(2) would have been
2895 * Note that this is _not_ check that the subvol is
2896 * empty or doesn't contain data that we wouldn't
2897 * otherwise be able to delete.
2899 * Users who want to delete empty subvols should try
2903 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2907 * Do not allow deletion if the parent dir is the same
2908 * as the dir to be deleted. That means the ioctl
2909 * must be called on the dentry referencing the root
2910 * of the subvol, not a random directory contained
2917 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2922 /* check if subvolume may be deleted by a user */
2923 err = btrfs_may_delete(dir, dentry, 1);
2927 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2933 err = btrfs_delete_subvolume(dir, dentry);
2934 inode_unlock(inode);
2936 fsnotify_rmdir(dir, dentry);
2945 mnt_drop_write_file(file);
2951 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2953 struct inode *inode = file_inode(file);
2954 struct btrfs_root *root = BTRFS_I(inode)->root;
2955 struct btrfs_ioctl_defrag_range_args *range;
2958 ret = mnt_want_write_file(file);
2962 if (btrfs_root_readonly(root)) {
2967 switch (inode->i_mode & S_IFMT) {
2969 if (!capable(CAP_SYS_ADMIN)) {
2973 ret = btrfs_defrag_root(root);
2977 * Note that this does not check the file descriptor for write
2978 * access. This prevents defragmenting executables that are
2979 * running and allows defrag on files open in read-only mode.
2981 if (!capable(CAP_SYS_ADMIN) &&
2982 inode_permission(inode, MAY_WRITE)) {
2987 range = kzalloc(sizeof(*range), GFP_KERNEL);
2994 if (copy_from_user(range, argp,
3000 /* compression requires us to start the IO */
3001 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3002 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
3003 range->extent_thresh = (u32)-1;
3006 /* the rest are all set to zero by kzalloc */
3007 range->len = (u64)-1;
3009 ret = btrfs_defrag_file(file_inode(file), file,
3010 range, BTRFS_OLDEST_GENERATION, 0);
3019 mnt_drop_write_file(file);
3023 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
3025 struct btrfs_ioctl_vol_args *vol_args;
3028 if (!capable(CAP_SYS_ADMIN))
3031 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
3032 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3034 vol_args = memdup_user(arg, sizeof(*vol_args));
3035 if (IS_ERR(vol_args)) {
3036 ret = PTR_ERR(vol_args);
3040 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3041 ret = btrfs_init_new_device(fs_info, vol_args->name);
3044 btrfs_info(fs_info, "disk added %s", vol_args->name);
3048 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3052 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
3054 struct inode *inode = file_inode(file);
3055 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3056 struct btrfs_ioctl_vol_args_v2 *vol_args;
3059 if (!capable(CAP_SYS_ADMIN))
3062 ret = mnt_want_write_file(file);
3066 vol_args = memdup_user(arg, sizeof(*vol_args));
3067 if (IS_ERR(vol_args)) {
3068 ret = PTR_ERR(vol_args);
3072 /* Check for compatibility reject unknown flags */
3073 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) {
3078 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3079 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3083 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
3084 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
3086 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3087 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3089 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3092 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3093 btrfs_info(fs_info, "device deleted: id %llu",
3096 btrfs_info(fs_info, "device deleted: %s",
3102 mnt_drop_write_file(file);
3106 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3108 struct inode *inode = file_inode(file);
3109 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3110 struct btrfs_ioctl_vol_args *vol_args;
3113 if (!capable(CAP_SYS_ADMIN))
3116 ret = mnt_want_write_file(file);
3120 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3121 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3122 goto out_drop_write;
3125 vol_args = memdup_user(arg, sizeof(*vol_args));
3126 if (IS_ERR(vol_args)) {
3127 ret = PTR_ERR(vol_args);
3131 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3132 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3135 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3138 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3140 mnt_drop_write_file(file);
3145 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3148 struct btrfs_ioctl_fs_info_args *fi_args;
3149 struct btrfs_device *device;
3150 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3153 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
3158 fi_args->num_devices = fs_devices->num_devices;
3160 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3161 if (device->devid > fi_args->max_id)
3162 fi_args->max_id = device->devid;
3166 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
3167 fi_args->nodesize = fs_info->nodesize;
3168 fi_args->sectorsize = fs_info->sectorsize;
3169 fi_args->clone_alignment = fs_info->sectorsize;
3171 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3178 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3181 struct btrfs_ioctl_dev_info_args *di_args;
3182 struct btrfs_device *dev;
3184 char *s_uuid = NULL;
3186 di_args = memdup_user(arg, sizeof(*di_args));
3187 if (IS_ERR(di_args))
3188 return PTR_ERR(di_args);
3190 if (!btrfs_is_empty_uuid(di_args->uuid))
3191 s_uuid = di_args->uuid;
3194 dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid,
3202 di_args->devid = dev->devid;
3203 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3204 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3205 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3207 strncpy(di_args->path, rcu_str_deref(dev->name),
3208 sizeof(di_args->path) - 1);
3209 di_args->path[sizeof(di_args->path) - 1] = 0;
3211 di_args->path[0] = '\0';
3216 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3223 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
3224 struct inode *inode2, u64 loff2, u64 len)
3226 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
3227 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
3230 static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
3231 struct inode *inode2, u64 loff2, u64 len)
3233 if (inode1 < inode2) {
3234 swap(inode1, inode2);
3236 } else if (inode1 == inode2 && loff2 < loff1) {
3239 lock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
3240 lock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
3243 static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 len,
3244 struct inode *dst, u64 dst_loff)
3246 const u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3250 * Lock destination range to serialize with concurrent readpages() and
3251 * source range to serialize with relocation.
3253 btrfs_double_extent_lock(src, loff, dst, dst_loff, len);
3254 ret = btrfs_clone(src, dst, loff, len, ALIGN(len, bs), dst_loff, 1);
3255 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3260 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3262 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3263 struct inode *dst, u64 dst_loff)
3266 u64 i, tail_len, chunk_count;
3267 struct btrfs_root *root_dst = BTRFS_I(dst)->root;
3269 spin_lock(&root_dst->root_item_lock);
3270 if (root_dst->send_in_progress) {
3271 btrfs_warn_rl(root_dst->fs_info,
3272 "cannot deduplicate to root %llu while send operations are using it (%d in progress)",
3273 root_dst->root_key.objectid,
3274 root_dst->send_in_progress);
3275 spin_unlock(&root_dst->root_item_lock);
3278 root_dst->dedupe_in_progress++;
3279 spin_unlock(&root_dst->root_item_lock);
3281 tail_len = olen % BTRFS_MAX_DEDUPE_LEN;
3282 chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN);
3284 for (i = 0; i < chunk_count; i++) {
3285 ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN,
3290 loff += BTRFS_MAX_DEDUPE_LEN;
3291 dst_loff += BTRFS_MAX_DEDUPE_LEN;
3295 ret = btrfs_extent_same_range(src, loff, tail_len, dst,
3298 spin_lock(&root_dst->root_item_lock);
3299 root_dst->dedupe_in_progress--;
3300 spin_unlock(&root_dst->root_item_lock);
3305 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3306 struct inode *inode,
3312 struct btrfs_root *root = BTRFS_I(inode)->root;
3315 inode_inc_iversion(inode);
3316 if (!no_time_update)
3317 inode->i_mtime = inode->i_ctime = current_time(inode);
3319 * We round up to the block size at eof when determining which
3320 * extents to clone above, but shouldn't round up the file size.
3322 if (endoff > destoff + olen)
3323 endoff = destoff + olen;
3324 if (endoff > inode->i_size)
3325 btrfs_i_size_write(BTRFS_I(inode), endoff);
3327 ret = btrfs_update_inode(trans, root, inode);
3329 btrfs_abort_transaction(trans, ret);
3330 btrfs_end_transaction(trans);
3333 ret = btrfs_end_transaction(trans);
3339 * Make sure we do not end up inserting an inline extent into a file that has
3340 * already other (non-inline) extents. If a file has an inline extent it can
3341 * not have any other extents and the (single) inline extent must start at the
3342 * file offset 0. Failing to respect these rules will lead to file corruption,
3343 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3345 * We can have extents that have been already written to disk or we can have
3346 * dirty ranges still in delalloc, in which case the extent maps and items are
3347 * created only when we run delalloc, and the delalloc ranges might fall outside
3348 * the range we are currently locking in the inode's io tree. So we check the
3349 * inode's i_size because of that (i_size updates are done while holding the
3350 * i_mutex, which we are holding here).
3351 * We also check to see if the inode has a size not greater than "datal" but has
3352 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3353 * protected against such concurrent fallocate calls by the i_mutex).
3355 * If the file has no extents but a size greater than datal, do not allow the
3356 * copy because we would need turn the inline extent into a non-inline one (even
3357 * with NO_HOLES enabled). If we find our destination inode only has one inline
3358 * extent, just overwrite it with the source inline extent if its size is less
3359 * than the source extent's size, or we could copy the source inline extent's
3360 * data into the destination inode's inline extent if the later is greater then
3363 static int clone_copy_inline_extent(struct inode *dst,
3364 struct btrfs_trans_handle *trans,
3365 struct btrfs_path *path,
3366 struct btrfs_key *new_key,
3367 const u64 drop_start,
3373 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
3374 struct btrfs_root *root = BTRFS_I(dst)->root;
3375 const u64 aligned_end = ALIGN(new_key->offset + datal,
3376 fs_info->sectorsize);
3378 struct btrfs_key key;
3380 if (new_key->offset > 0)
3383 key.objectid = btrfs_ino(BTRFS_I(dst));
3384 key.type = BTRFS_EXTENT_DATA_KEY;
3386 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3389 } else if (ret > 0) {
3390 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3391 ret = btrfs_next_leaf(root, path);
3395 goto copy_inline_extent;
3397 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3398 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3399 key.type == BTRFS_EXTENT_DATA_KEY) {
3400 ASSERT(key.offset > 0);
3403 } else if (i_size_read(dst) <= datal) {
3404 struct btrfs_file_extent_item *ei;
3408 * If the file size is <= datal, make sure there are no other
3409 * extents following (can happen do to an fallocate call with
3410 * the flag FALLOC_FL_KEEP_SIZE).
3412 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3413 struct btrfs_file_extent_item);
3415 * If it's an inline extent, it can not have other extents
3418 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3419 BTRFS_FILE_EXTENT_INLINE)
3420 goto copy_inline_extent;
3422 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3423 if (ext_len > aligned_end)
3426 ret = btrfs_next_item(root, path);
3429 } else if (ret == 0) {
3430 btrfs_item_key_to_cpu(path->nodes[0], &key,
3432 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3433 key.type == BTRFS_EXTENT_DATA_KEY)
3440 * We have no extent items, or we have an extent at offset 0 which may
3441 * or may not be inlined. All these cases are dealt the same way.
3443 if (i_size_read(dst) > datal) {
3445 * If the destination inode has an inline extent...
3446 * This would require copying the data from the source inline
3447 * extent into the beginning of the destination's inline extent.
3448 * But this is really complex, both extents can be compressed
3449 * or just one of them, which would require decompressing and
3450 * re-compressing data (which could increase the new compressed
3451 * size, not allowing the compressed data to fit anymore in an
3453 * So just don't support this case for now (it should be rare,
3454 * we are not really saving space when cloning inline extents).
3459 btrfs_release_path(path);
3460 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3463 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3468 const u32 start = btrfs_file_extent_calc_inline_size(0);
3470 memmove(inline_data + start, inline_data + start + skip, datal);
3473 write_extent_buffer(path->nodes[0], inline_data,
3474 btrfs_item_ptr_offset(path->nodes[0],
3477 inode_add_bytes(dst, datal);
3478 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(dst)->runtime_flags);
3484 * btrfs_clone() - clone a range from inode file to another
3486 * @src: Inode to clone from
3487 * @inode: Inode to clone to
3488 * @off: Offset within source to start clone from
3489 * @olen: Original length, passed by user, of range to clone
3490 * @olen_aligned: Block-aligned value of olen
3491 * @destoff: Offset within @inode to start clone
3492 * @no_time_update: Whether to update mtime/ctime on the target inode
3494 static int btrfs_clone(struct inode *src, struct inode *inode,
3495 const u64 off, const u64 olen, const u64 olen_aligned,
3496 const u64 destoff, int no_time_update)
3498 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3499 struct btrfs_root *root = BTRFS_I(inode)->root;
3500 struct btrfs_path *path = NULL;
3501 struct extent_buffer *leaf;
3502 struct btrfs_trans_handle *trans;
3504 struct btrfs_key key;
3508 const u64 len = olen_aligned;
3509 u64 last_dest_end = destoff;
3512 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3516 path = btrfs_alloc_path();
3522 path->reada = READA_FORWARD;
3524 key.objectid = btrfs_ino(BTRFS_I(src));
3525 key.type = BTRFS_EXTENT_DATA_KEY;
3529 u64 next_key_min_offset = key.offset + 1;
3530 struct btrfs_file_extent_item *extent;
3533 struct btrfs_key new_key;
3534 u64 disko = 0, diskl = 0;
3535 u64 datao = 0, datal = 0;
3540 * note the key will change type as we walk through the
3543 path->leave_spinning = 1;
3544 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3549 * First search, if no extent item that starts at offset off was
3550 * found but the previous item is an extent item, it's possible
3551 * it might overlap our target range, therefore process it.
3553 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3554 btrfs_item_key_to_cpu(path->nodes[0], &key,
3555 path->slots[0] - 1);
3556 if (key.type == BTRFS_EXTENT_DATA_KEY)
3560 nritems = btrfs_header_nritems(path->nodes[0]);
3562 if (path->slots[0] >= nritems) {
3563 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3568 nritems = btrfs_header_nritems(path->nodes[0]);
3570 leaf = path->nodes[0];
3571 slot = path->slots[0];
3573 btrfs_item_key_to_cpu(leaf, &key, slot);
3574 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3575 key.objectid != btrfs_ino(BTRFS_I(src)))
3578 ASSERT(key.type == BTRFS_EXTENT_DATA_KEY);
3580 extent = btrfs_item_ptr(leaf, slot,
3581 struct btrfs_file_extent_item);
3582 comp = btrfs_file_extent_compression(leaf, extent);
3583 type = btrfs_file_extent_type(leaf, extent);
3584 if (type == BTRFS_FILE_EXTENT_REG ||
3585 type == BTRFS_FILE_EXTENT_PREALLOC) {
3586 disko = btrfs_file_extent_disk_bytenr(leaf, extent);
3587 diskl = btrfs_file_extent_disk_num_bytes(leaf, extent);
3588 datao = btrfs_file_extent_offset(leaf, extent);
3589 datal = btrfs_file_extent_num_bytes(leaf, extent);
3590 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3591 /* Take upper bound, may be compressed */
3592 datal = btrfs_file_extent_ram_bytes(leaf, extent);
3596 * The first search might have left us at an extent item that
3597 * ends before our target range's start, can happen if we have
3598 * holes and NO_HOLES feature enabled.
3600 if (key.offset + datal <= off) {
3603 } else if (key.offset >= off + len) {
3606 next_key_min_offset = key.offset + datal;
3607 size = btrfs_item_size_nr(leaf, slot);
3608 read_extent_buffer(leaf, buf, btrfs_item_ptr_offset(leaf, slot),
3611 btrfs_release_path(path);
3612 path->leave_spinning = 0;
3614 memcpy(&new_key, &key, sizeof(new_key));
3615 new_key.objectid = btrfs_ino(BTRFS_I(inode));
3616 if (off <= key.offset)
3617 new_key.offset = key.offset + destoff - off;
3619 new_key.offset = destoff;
3622 * Deal with a hole that doesn't have an extent item that
3623 * represents it (NO_HOLES feature enabled).
3624 * This hole is either in the middle of the cloning range or at
3625 * the beginning (fully overlaps it or partially overlaps it).
3627 if (new_key.offset != last_dest_end)
3628 drop_start = last_dest_end;
3630 drop_start = new_key.offset;
3632 if (type == BTRFS_FILE_EXTENT_REG ||
3633 type == BTRFS_FILE_EXTENT_PREALLOC) {
3634 struct btrfs_clone_extent_info clone_info;
3637 * a | --- range to clone ---| b
3638 * | ------------- extent ------------- |
3641 /* Subtract range b */
3642 if (key.offset + datal > off + len)
3643 datal = off + len - key.offset;
3645 /* Subtract range a */
3646 if (off > key.offset) {
3647 datao += off - key.offset;
3648 datal -= off - key.offset;
3651 clone_info.disk_offset = disko;
3652 clone_info.disk_len = diskl;
3653 clone_info.data_offset = datao;
3654 clone_info.data_len = datal;
3655 clone_info.file_offset = new_key.offset;
3656 clone_info.extent_buf = buf;
3657 clone_info.item_size = size;
3658 ret = btrfs_punch_hole_range(inode, path,
3660 new_key.offset + datal - 1,
3661 &clone_info, &trans);
3664 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3668 if (off > key.offset) {
3669 skip = off - key.offset;
3670 new_key.offset += skip;
3673 if (key.offset + datal > off + len)
3674 trim = key.offset + datal - (off + len);
3676 if (comp && (skip || trim)) {
3680 size -= skip + trim;
3681 datal -= skip + trim;
3684 * If our extent is inline, we know we will drop or
3685 * adjust at most 1 extent item in the destination root.
3687 * 1 - adjusting old extent (we may have to split it)
3688 * 1 - add new extent
3691 trans = btrfs_start_transaction(root, 3);
3692 if (IS_ERR(trans)) {
3693 ret = PTR_ERR(trans);
3697 ret = clone_copy_inline_extent(inode, trans, path,
3698 &new_key, drop_start,
3699 datal, skip, size, buf);
3701 if (ret != -EOPNOTSUPP)
3702 btrfs_abort_transaction(trans, ret);
3703 btrfs_end_transaction(trans);
3708 btrfs_release_path(path);
3710 last_dest_end = ALIGN(new_key.offset + datal,
3711 fs_info->sectorsize);
3712 ret = clone_finish_inode_update(trans, inode, last_dest_end,
3713 destoff, olen, no_time_update);
3716 if (new_key.offset + datal >= destoff + len)
3719 btrfs_release_path(path);
3720 key.offset = next_key_min_offset;
3722 if (fatal_signal_pending(current)) {
3729 if (last_dest_end < destoff + len) {
3731 * We have an implicit hole that fully or partially overlaps our
3732 * cloning range at its end. This means that we either have the
3733 * NO_HOLES feature enabled or the implicit hole happened due to
3734 * mixing buffered and direct IO writes against this file.
3736 btrfs_release_path(path);
3737 path->leave_spinning = 0;
3739 ret = btrfs_punch_hole_range(inode, path,
3740 last_dest_end, destoff + len - 1,
3745 ret = clone_finish_inode_update(trans, inode, destoff + len,
3746 destoff, olen, no_time_update);
3750 btrfs_free_path(path);
3755 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3756 u64 off, u64 olen, u64 destoff)
3758 struct inode *inode = file_inode(file);
3759 struct inode *src = file_inode(file_src);
3760 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3763 u64 bs = fs_info->sb->s_blocksize;
3767 * - split compressed inline extents. annoying: we need to
3768 * decompress into destination's address_space (the file offset
3769 * may change, so source mapping won't do), then recompress (or
3770 * otherwise reinsert) a subrange.
3772 * - split destination inode's inline extents. The inline extents can
3773 * be either compressed or non-compressed.
3777 * VFS's generic_remap_file_range_prep() protects us from cloning the
3778 * eof block into the middle of a file, which would result in corruption
3779 * if the file size is not blocksize aligned. So we don't need to check
3780 * for that case here.
3782 if (off + len == src->i_size)
3783 len = ALIGN(src->i_size, bs) - off;
3785 if (destoff > inode->i_size) {
3786 const u64 wb_start = ALIGN_DOWN(inode->i_size, bs);
3788 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3792 * We may have truncated the last block if the inode's size is
3793 * not sector size aligned, so we need to wait for writeback to
3794 * complete before proceeding further, otherwise we can race
3795 * with cloning and attempt to increment a reference to an
3796 * extent that no longer exists (writeback completed right after
3797 * we found the previous extent covering eof and before we
3798 * attempted to increment its reference count).
3800 ret = btrfs_wait_ordered_range(inode, wb_start,
3801 destoff - wb_start);
3807 * Lock destination range to serialize with concurrent readpages() and
3808 * source range to serialize with relocation.
3810 btrfs_double_extent_lock(src, off, inode, destoff, len);
3811 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3812 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3814 * Truncate page cache pages so that future reads will see the cloned
3815 * data immediately and not the previous data.
3817 truncate_inode_pages_range(&inode->i_data,
3818 round_down(destoff, PAGE_SIZE),
3819 round_up(destoff + len, PAGE_SIZE) - 1);
3824 static int btrfs_remap_file_range_prep(struct file *file_in, loff_t pos_in,
3825 struct file *file_out, loff_t pos_out,
3826 loff_t *len, unsigned int remap_flags)
3828 struct inode *inode_in = file_inode(file_in);
3829 struct inode *inode_out = file_inode(file_out);
3830 u64 bs = BTRFS_I(inode_out)->root->fs_info->sb->s_blocksize;
3831 bool same_inode = inode_out == inode_in;
3835 if (!(remap_flags & REMAP_FILE_DEDUP)) {
3836 struct btrfs_root *root_out = BTRFS_I(inode_out)->root;
3838 if (btrfs_root_readonly(root_out))
3841 if (file_in->f_path.mnt != file_out->f_path.mnt ||
3842 inode_in->i_sb != inode_out->i_sb)
3846 /* don't make the dst file partly checksummed */
3847 if ((BTRFS_I(inode_in)->flags & BTRFS_INODE_NODATASUM) !=
3848 (BTRFS_I(inode_out)->flags & BTRFS_INODE_NODATASUM)) {
3853 * Now that the inodes are locked, we need to start writeback ourselves
3854 * and can not rely on the writeback from the VFS's generic helper
3855 * generic_remap_file_range_prep() because:
3857 * 1) For compression we must call filemap_fdatawrite_range() range
3858 * twice (btrfs_fdatawrite_range() does it for us), and the generic
3859 * helper only calls it once;
3861 * 2) filemap_fdatawrite_range(), called by the generic helper only
3862 * waits for the writeback to complete, i.e. for IO to be done, and
3863 * not for the ordered extents to complete. We need to wait for them
3864 * to complete so that new file extent items are in the fs tree.
3866 if (*len == 0 && !(remap_flags & REMAP_FILE_DEDUP))
3867 wb_len = ALIGN(inode_in->i_size, bs) - ALIGN_DOWN(pos_in, bs);
3869 wb_len = ALIGN(*len, bs);
3872 * Since we don't lock ranges, wait for ongoing lockless dio writes (as
3873 * any in progress could create its ordered extents after we wait for
3874 * existing ordered extents below).
3876 inode_dio_wait(inode_in);
3878 inode_dio_wait(inode_out);
3881 * Workaround to make sure NOCOW buffered write reach disk as NOCOW.
3883 * Btrfs' back references do not have a block level granularity, they
3884 * work at the whole extent level.
3885 * NOCOW buffered write without data space reserved may not be able
3886 * to fall back to CoW due to lack of data space, thus could cause
3889 * Here we take a shortcut by flushing the whole inode, so that all
3890 * nocow write should reach disk as nocow before we increase the
3891 * reference of the extent. We could do better by only flushing NOCOW
3892 * data, but that needs extra accounting.
3894 * Also we don't need to check ASYNC_EXTENT, as async extent will be
3895 * CoWed anyway, not affecting nocow part.
3897 ret = filemap_flush(inode_in->i_mapping);
3901 ret = btrfs_wait_ordered_range(inode_in, ALIGN_DOWN(pos_in, bs),
3905 ret = btrfs_wait_ordered_range(inode_out, ALIGN_DOWN(pos_out, bs),
3910 return generic_remap_file_range_prep(file_in, pos_in, file_out, pos_out,
3914 loff_t btrfs_remap_file_range(struct file *src_file, loff_t off,
3915 struct file *dst_file, loff_t destoff, loff_t len,
3916 unsigned int remap_flags)
3918 struct inode *src_inode = file_inode(src_file);
3919 struct inode *dst_inode = file_inode(dst_file);
3920 bool same_inode = dst_inode == src_inode;
3923 if (remap_flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_ADVISORY))
3927 inode_lock(src_inode);
3929 lock_two_nondirectories(src_inode, dst_inode);
3931 ret = btrfs_remap_file_range_prep(src_file, off, dst_file, destoff,
3933 if (ret < 0 || len == 0)
3936 if (remap_flags & REMAP_FILE_DEDUP)
3937 ret = btrfs_extent_same(src_inode, off, len, dst_inode, destoff);
3939 ret = btrfs_clone_files(dst_file, src_file, off, len, destoff);
3943 inode_unlock(src_inode);
3945 unlock_two_nondirectories(src_inode, dst_inode);
3947 return ret < 0 ? ret : len;
3950 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3952 struct inode *inode = file_inode(file);
3953 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3954 struct btrfs_root *root = BTRFS_I(inode)->root;
3955 struct btrfs_root *new_root;
3956 struct btrfs_dir_item *di;
3957 struct btrfs_trans_handle *trans;
3958 struct btrfs_path *path;
3959 struct btrfs_key location;
3960 struct btrfs_disk_key disk_key;
3965 if (!capable(CAP_SYS_ADMIN))
3968 ret = mnt_want_write_file(file);
3972 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3978 objectid = BTRFS_FS_TREE_OBJECTID;
3980 location.objectid = objectid;
3981 location.type = BTRFS_ROOT_ITEM_KEY;
3982 location.offset = (u64)-1;
3984 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
3985 if (IS_ERR(new_root)) {
3986 ret = PTR_ERR(new_root);
3989 if (!is_fstree(new_root->root_key.objectid)) {
3994 path = btrfs_alloc_path();
3999 path->leave_spinning = 1;
4001 trans = btrfs_start_transaction(root, 1);
4002 if (IS_ERR(trans)) {
4003 btrfs_free_path(path);
4004 ret = PTR_ERR(trans);
4008 dir_id = btrfs_super_root_dir(fs_info->super_copy);
4009 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
4010 dir_id, "default", 7, 1);
4011 if (IS_ERR_OR_NULL(di)) {
4012 btrfs_free_path(path);
4013 btrfs_end_transaction(trans);
4015 "Umm, you don't have the default diritem, this isn't going to work");
4020 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4021 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4022 btrfs_mark_buffer_dirty(path->nodes[0]);
4023 btrfs_free_path(path);
4025 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
4026 btrfs_end_transaction(trans);
4028 mnt_drop_write_file(file);
4032 static void get_block_group_info(struct list_head *groups_list,
4033 struct btrfs_ioctl_space_info *space)
4035 struct btrfs_block_group *block_group;
4037 space->total_bytes = 0;
4038 space->used_bytes = 0;
4040 list_for_each_entry(block_group, groups_list, list) {
4041 space->flags = block_group->flags;
4042 space->total_bytes += block_group->length;
4043 space->used_bytes += block_group->used;
4047 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
4050 struct btrfs_ioctl_space_args space_args;
4051 struct btrfs_ioctl_space_info space;
4052 struct btrfs_ioctl_space_info *dest;
4053 struct btrfs_ioctl_space_info *dest_orig;
4054 struct btrfs_ioctl_space_info __user *user_dest;
4055 struct btrfs_space_info *info;
4056 static const u64 types[] = {
4057 BTRFS_BLOCK_GROUP_DATA,
4058 BTRFS_BLOCK_GROUP_SYSTEM,
4059 BTRFS_BLOCK_GROUP_METADATA,
4060 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
4068 if (copy_from_user(&space_args,
4069 (struct btrfs_ioctl_space_args __user *)arg,
4070 sizeof(space_args)))
4073 for (i = 0; i < num_types; i++) {
4074 struct btrfs_space_info *tmp;
4078 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4080 if (tmp->flags == types[i]) {
4090 down_read(&info->groups_sem);
4091 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4092 if (!list_empty(&info->block_groups[c]))
4095 up_read(&info->groups_sem);
4099 * Global block reserve, exported as a space_info
4103 /* space_slots == 0 means they are asking for a count */
4104 if (space_args.space_slots == 0) {
4105 space_args.total_spaces = slot_count;
4109 slot_count = min_t(u64, space_args.space_slots, slot_count);
4111 alloc_size = sizeof(*dest) * slot_count;
4113 /* we generally have at most 6 or so space infos, one for each raid
4114 * level. So, a whole page should be more than enough for everyone
4116 if (alloc_size > PAGE_SIZE)
4119 space_args.total_spaces = 0;
4120 dest = kmalloc(alloc_size, GFP_KERNEL);
4125 /* now we have a buffer to copy into */
4126 for (i = 0; i < num_types; i++) {
4127 struct btrfs_space_info *tmp;
4134 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4136 if (tmp->flags == types[i]) {
4145 down_read(&info->groups_sem);
4146 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4147 if (!list_empty(&info->block_groups[c])) {
4148 get_block_group_info(&info->block_groups[c],
4150 memcpy(dest, &space, sizeof(space));
4152 space_args.total_spaces++;
4158 up_read(&info->groups_sem);
4162 * Add global block reserve
4165 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
4167 spin_lock(&block_rsv->lock);
4168 space.total_bytes = block_rsv->size;
4169 space.used_bytes = block_rsv->size - block_rsv->reserved;
4170 spin_unlock(&block_rsv->lock);
4171 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4172 memcpy(dest, &space, sizeof(space));
4173 space_args.total_spaces++;
4176 user_dest = (struct btrfs_ioctl_space_info __user *)
4177 (arg + sizeof(struct btrfs_ioctl_space_args));
4179 if (copy_to_user(user_dest, dest_orig, alloc_size))
4184 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4190 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4193 struct btrfs_trans_handle *trans;
4197 trans = btrfs_attach_transaction_barrier(root);
4198 if (IS_ERR(trans)) {
4199 if (PTR_ERR(trans) != -ENOENT)
4200 return PTR_ERR(trans);
4202 /* No running transaction, don't bother */
4203 transid = root->fs_info->last_trans_committed;
4206 transid = trans->transid;
4207 ret = btrfs_commit_transaction_async(trans, 0);
4209 btrfs_end_transaction(trans);
4214 if (copy_to_user(argp, &transid, sizeof(transid)))
4219 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
4225 if (copy_from_user(&transid, argp, sizeof(transid)))
4228 transid = 0; /* current trans */
4230 return btrfs_wait_for_commit(fs_info, transid);
4233 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4235 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4236 struct btrfs_ioctl_scrub_args *sa;
4239 if (!capable(CAP_SYS_ADMIN))
4242 sa = memdup_user(arg, sizeof(*sa));
4246 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4247 ret = mnt_want_write_file(file);
4252 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4253 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4257 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
4258 * error. This is important as it allows user space to know how much
4259 * progress scrub has done. For example, if scrub is canceled we get
4260 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
4261 * space. Later user space can inspect the progress from the structure
4262 * btrfs_ioctl_scrub_args and resume scrub from where it left off
4263 * previously (btrfs-progs does this).
4264 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
4265 * then return -EFAULT to signal the structure was not copied or it may
4266 * be corrupt and unreliable due to a partial copy.
4268 if (copy_to_user(arg, sa, sizeof(*sa)))
4271 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4272 mnt_drop_write_file(file);
4278 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
4280 if (!capable(CAP_SYS_ADMIN))
4283 return btrfs_scrub_cancel(fs_info);
4286 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
4289 struct btrfs_ioctl_scrub_args *sa;
4292 if (!capable(CAP_SYS_ADMIN))
4295 sa = memdup_user(arg, sizeof(*sa));
4299 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
4301 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
4308 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
4311 struct btrfs_ioctl_get_dev_stats *sa;
4314 sa = memdup_user(arg, sizeof(*sa));
4318 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4323 ret = btrfs_get_dev_stats(fs_info, sa);
4325 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
4332 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4335 struct btrfs_ioctl_dev_replace_args *p;
4338 if (!capable(CAP_SYS_ADMIN))
4341 p = memdup_user(arg, sizeof(*p));
4346 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4347 if (sb_rdonly(fs_info->sb)) {
4351 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4352 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4354 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
4355 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4358 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4359 btrfs_dev_replace_status(fs_info, p);
4362 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4363 p->result = btrfs_dev_replace_cancel(fs_info);
4371 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
4378 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4384 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4385 struct inode_fs_paths *ipath = NULL;
4386 struct btrfs_path *path;
4388 if (!capable(CAP_DAC_READ_SEARCH))
4391 path = btrfs_alloc_path();
4397 ipa = memdup_user(arg, sizeof(*ipa));
4404 size = min_t(u32, ipa->size, 4096);
4405 ipath = init_ipath(size, root, path);
4406 if (IS_ERR(ipath)) {
4407 ret = PTR_ERR(ipath);
4412 ret = paths_from_inode(ipa->inum, ipath);
4416 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4417 rel_ptr = ipath->fspath->val[i] -
4418 (u64)(unsigned long)ipath->fspath->val;
4419 ipath->fspath->val[i] = rel_ptr;
4422 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
4423 ipath->fspath, size);
4430 btrfs_free_path(path);
4437 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4439 struct btrfs_data_container *inodes = ctx;
4440 const size_t c = 3 * sizeof(u64);
4442 if (inodes->bytes_left >= c) {
4443 inodes->bytes_left -= c;
4444 inodes->val[inodes->elem_cnt] = inum;
4445 inodes->val[inodes->elem_cnt + 1] = offset;
4446 inodes->val[inodes->elem_cnt + 2] = root;
4447 inodes->elem_cnt += 3;
4449 inodes->bytes_missing += c - inodes->bytes_left;
4450 inodes->bytes_left = 0;
4451 inodes->elem_missed += 3;
4457 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
4458 void __user *arg, int version)
4462 struct btrfs_ioctl_logical_ino_args *loi;
4463 struct btrfs_data_container *inodes = NULL;
4464 struct btrfs_path *path = NULL;
4467 if (!capable(CAP_SYS_ADMIN))
4470 loi = memdup_user(arg, sizeof(*loi));
4472 return PTR_ERR(loi);
4475 ignore_offset = false;
4476 size = min_t(u32, loi->size, SZ_64K);
4478 /* All reserved bits must be 0 for now */
4479 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
4483 /* Only accept flags we have defined so far */
4484 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
4488 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
4489 size = min_t(u32, loi->size, SZ_16M);
4492 path = btrfs_alloc_path();
4498 inodes = init_data_container(size);
4499 if (IS_ERR(inodes)) {
4500 ret = PTR_ERR(inodes);
4505 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
4506 build_ino_list, inodes, ignore_offset);
4512 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
4518 btrfs_free_path(path);
4526 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
4527 struct btrfs_ioctl_balance_args *bargs)
4529 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4531 bargs->flags = bctl->flags;
4533 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
4534 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4535 if (atomic_read(&fs_info->balance_pause_req))
4536 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4537 if (atomic_read(&fs_info->balance_cancel_req))
4538 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4540 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4541 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4542 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4544 spin_lock(&fs_info->balance_lock);
4545 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4546 spin_unlock(&fs_info->balance_lock);
4549 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4551 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4552 struct btrfs_fs_info *fs_info = root->fs_info;
4553 struct btrfs_ioctl_balance_args *bargs;
4554 struct btrfs_balance_control *bctl;
4555 bool need_unlock; /* for mut. excl. ops lock */
4558 if (!capable(CAP_SYS_ADMIN))
4561 ret = mnt_want_write_file(file);
4566 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4567 mutex_lock(&fs_info->balance_mutex);
4573 * mut. excl. ops lock is locked. Three possibilities:
4574 * (1) some other op is running
4575 * (2) balance is running
4576 * (3) balance is paused -- special case (think resume)
4578 mutex_lock(&fs_info->balance_mutex);
4579 if (fs_info->balance_ctl) {
4580 /* this is either (2) or (3) */
4581 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4582 mutex_unlock(&fs_info->balance_mutex);
4584 * Lock released to allow other waiters to continue,
4585 * we'll reexamine the status again.
4587 mutex_lock(&fs_info->balance_mutex);
4589 if (fs_info->balance_ctl &&
4590 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4592 need_unlock = false;
4596 mutex_unlock(&fs_info->balance_mutex);
4600 mutex_unlock(&fs_info->balance_mutex);
4606 mutex_unlock(&fs_info->balance_mutex);
4607 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4612 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
4615 bargs = memdup_user(arg, sizeof(*bargs));
4616 if (IS_ERR(bargs)) {
4617 ret = PTR_ERR(bargs);
4621 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4622 if (!fs_info->balance_ctl) {
4627 bctl = fs_info->balance_ctl;
4628 spin_lock(&fs_info->balance_lock);
4629 bctl->flags |= BTRFS_BALANCE_RESUME;
4630 spin_unlock(&fs_info->balance_lock);
4638 if (fs_info->balance_ctl) {
4643 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4650 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4651 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4652 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4654 bctl->flags = bargs->flags;
4656 /* balance everything - no filters */
4657 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4660 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4667 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to
4668 * btrfs_balance. bctl is freed in reset_balance_state, or, if
4669 * restriper was paused all the way until unmount, in free_fs_info.
4670 * The flag should be cleared after reset_balance_state.
4672 need_unlock = false;
4674 ret = btrfs_balance(fs_info, bctl, bargs);
4677 if ((ret == 0 || ret == -ECANCELED) && arg) {
4678 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4687 mutex_unlock(&fs_info->balance_mutex);
4689 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4691 mnt_drop_write_file(file);
4695 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4697 if (!capable(CAP_SYS_ADMIN))
4701 case BTRFS_BALANCE_CTL_PAUSE:
4702 return btrfs_pause_balance(fs_info);
4703 case BTRFS_BALANCE_CTL_CANCEL:
4704 return btrfs_cancel_balance(fs_info);
4710 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4713 struct btrfs_ioctl_balance_args *bargs;
4716 if (!capable(CAP_SYS_ADMIN))
4719 mutex_lock(&fs_info->balance_mutex);
4720 if (!fs_info->balance_ctl) {
4725 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4731 btrfs_update_ioctl_balance_args(fs_info, bargs);
4733 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4738 mutex_unlock(&fs_info->balance_mutex);
4742 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4744 struct inode *inode = file_inode(file);
4745 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4746 struct btrfs_ioctl_quota_ctl_args *sa;
4749 if (!capable(CAP_SYS_ADMIN))
4752 ret = mnt_want_write_file(file);
4756 sa = memdup_user(arg, sizeof(*sa));
4762 down_write(&fs_info->subvol_sem);
4765 case BTRFS_QUOTA_CTL_ENABLE:
4766 ret = btrfs_quota_enable(fs_info);
4768 case BTRFS_QUOTA_CTL_DISABLE:
4769 ret = btrfs_quota_disable(fs_info);
4777 up_write(&fs_info->subvol_sem);
4779 mnt_drop_write_file(file);
4783 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4785 struct inode *inode = file_inode(file);
4786 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4787 struct btrfs_root *root = BTRFS_I(inode)->root;
4788 struct btrfs_ioctl_qgroup_assign_args *sa;
4789 struct btrfs_trans_handle *trans;
4793 if (!capable(CAP_SYS_ADMIN))
4796 ret = mnt_want_write_file(file);
4800 sa = memdup_user(arg, sizeof(*sa));
4806 trans = btrfs_join_transaction(root);
4807 if (IS_ERR(trans)) {
4808 ret = PTR_ERR(trans);
4813 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
4815 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
4818 /* update qgroup status and info */
4819 err = btrfs_run_qgroups(trans);
4821 btrfs_handle_fs_error(fs_info, err,
4822 "failed to update qgroup status and info");
4823 err = btrfs_end_transaction(trans);
4830 mnt_drop_write_file(file);
4834 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4836 struct inode *inode = file_inode(file);
4837 struct btrfs_root *root = BTRFS_I(inode)->root;
4838 struct btrfs_ioctl_qgroup_create_args *sa;
4839 struct btrfs_trans_handle *trans;
4843 if (!capable(CAP_SYS_ADMIN))
4846 ret = mnt_want_write_file(file);
4850 sa = memdup_user(arg, sizeof(*sa));
4856 if (!sa->qgroupid) {
4861 trans = btrfs_join_transaction(root);
4862 if (IS_ERR(trans)) {
4863 ret = PTR_ERR(trans);
4868 ret = btrfs_create_qgroup(trans, sa->qgroupid);
4870 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
4873 err = btrfs_end_transaction(trans);
4880 mnt_drop_write_file(file);
4884 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4886 struct inode *inode = file_inode(file);
4887 struct btrfs_root *root = BTRFS_I(inode)->root;
4888 struct btrfs_ioctl_qgroup_limit_args *sa;
4889 struct btrfs_trans_handle *trans;
4894 if (!capable(CAP_SYS_ADMIN))
4897 ret = mnt_want_write_file(file);
4901 sa = memdup_user(arg, sizeof(*sa));
4907 trans = btrfs_join_transaction(root);
4908 if (IS_ERR(trans)) {
4909 ret = PTR_ERR(trans);
4913 qgroupid = sa->qgroupid;
4915 /* take the current subvol as qgroup */
4916 qgroupid = root->root_key.objectid;
4919 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
4921 err = btrfs_end_transaction(trans);
4928 mnt_drop_write_file(file);
4932 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4934 struct inode *inode = file_inode(file);
4935 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4936 struct btrfs_ioctl_quota_rescan_args *qsa;
4939 if (!capable(CAP_SYS_ADMIN))
4942 ret = mnt_want_write_file(file);
4946 qsa = memdup_user(arg, sizeof(*qsa));
4957 ret = btrfs_qgroup_rescan(fs_info);
4962 mnt_drop_write_file(file);
4966 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4969 struct btrfs_ioctl_quota_rescan_args *qsa;
4972 if (!capable(CAP_SYS_ADMIN))
4975 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
4979 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4981 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
4984 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4991 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4994 if (!capable(CAP_SYS_ADMIN))
4997 return btrfs_qgroup_wait_for_completion(fs_info, true);
5000 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5001 struct btrfs_ioctl_received_subvol_args *sa)
5003 struct inode *inode = file_inode(file);
5004 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5005 struct btrfs_root *root = BTRFS_I(inode)->root;
5006 struct btrfs_root_item *root_item = &root->root_item;
5007 struct btrfs_trans_handle *trans;
5008 struct timespec64 ct = current_time(inode);
5010 int received_uuid_changed;
5012 if (!inode_owner_or_capable(inode))
5015 ret = mnt_want_write_file(file);
5019 down_write(&fs_info->subvol_sem);
5021 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
5026 if (btrfs_root_readonly(root)) {
5033 * 2 - uuid items (received uuid + subvol uuid)
5035 trans = btrfs_start_transaction(root, 3);
5036 if (IS_ERR(trans)) {
5037 ret = PTR_ERR(trans);
5042 sa->rtransid = trans->transid;
5043 sa->rtime.sec = ct.tv_sec;
5044 sa->rtime.nsec = ct.tv_nsec;
5046 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5048 if (received_uuid_changed &&
5049 !btrfs_is_empty_uuid(root_item->received_uuid)) {
5050 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
5051 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5052 root->root_key.objectid);
5053 if (ret && ret != -ENOENT) {
5054 btrfs_abort_transaction(trans, ret);
5055 btrfs_end_transaction(trans);
5059 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5060 btrfs_set_root_stransid(root_item, sa->stransid);
5061 btrfs_set_root_rtransid(root_item, sa->rtransid);
5062 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5063 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5064 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5065 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5067 ret = btrfs_update_root(trans, fs_info->tree_root,
5068 &root->root_key, &root->root_item);
5070 btrfs_end_transaction(trans);
5073 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5074 ret = btrfs_uuid_tree_add(trans, sa->uuid,
5075 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5076 root->root_key.objectid);
5077 if (ret < 0 && ret != -EEXIST) {
5078 btrfs_abort_transaction(trans, ret);
5079 btrfs_end_transaction(trans);
5083 ret = btrfs_commit_transaction(trans);
5085 up_write(&fs_info->subvol_sem);
5086 mnt_drop_write_file(file);
5091 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5094 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5095 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5098 args32 = memdup_user(arg, sizeof(*args32));
5100 return PTR_ERR(args32);
5102 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5108 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5109 args64->stransid = args32->stransid;
5110 args64->rtransid = args32->rtransid;
5111 args64->stime.sec = args32->stime.sec;
5112 args64->stime.nsec = args32->stime.nsec;
5113 args64->rtime.sec = args32->rtime.sec;
5114 args64->rtime.nsec = args32->rtime.nsec;
5115 args64->flags = args32->flags;
5117 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5121 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5122 args32->stransid = args64->stransid;
5123 args32->rtransid = args64->rtransid;
5124 args32->stime.sec = args64->stime.sec;
5125 args32->stime.nsec = args64->stime.nsec;
5126 args32->rtime.sec = args64->rtime.sec;
5127 args32->rtime.nsec = args64->rtime.nsec;
5128 args32->flags = args64->flags;
5130 ret = copy_to_user(arg, args32, sizeof(*args32));
5141 static long btrfs_ioctl_set_received_subvol(struct file *file,
5144 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5147 sa = memdup_user(arg, sizeof(*sa));
5151 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5156 ret = copy_to_user(arg, sa, sizeof(*sa));
5165 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
5170 char label[BTRFS_LABEL_SIZE];
5172 spin_lock(&fs_info->super_lock);
5173 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5174 spin_unlock(&fs_info->super_lock);
5176 len = strnlen(label, BTRFS_LABEL_SIZE);
5178 if (len == BTRFS_LABEL_SIZE) {
5180 "label is too long, return the first %zu bytes",
5184 ret = copy_to_user(arg, label, len);
5186 return ret ? -EFAULT : 0;
5189 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5191 struct inode *inode = file_inode(file);
5192 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5193 struct btrfs_root *root = BTRFS_I(inode)->root;
5194 struct btrfs_super_block *super_block = fs_info->super_copy;
5195 struct btrfs_trans_handle *trans;
5196 char label[BTRFS_LABEL_SIZE];
5199 if (!capable(CAP_SYS_ADMIN))
5202 if (copy_from_user(label, arg, sizeof(label)))
5205 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5207 "unable to set label with more than %d bytes",
5208 BTRFS_LABEL_SIZE - 1);
5212 ret = mnt_want_write_file(file);
5216 trans = btrfs_start_transaction(root, 0);
5217 if (IS_ERR(trans)) {
5218 ret = PTR_ERR(trans);
5222 spin_lock(&fs_info->super_lock);
5223 strcpy(super_block->label, label);
5224 spin_unlock(&fs_info->super_lock);
5225 ret = btrfs_commit_transaction(trans);
5228 mnt_drop_write_file(file);
5232 #define INIT_FEATURE_FLAGS(suffix) \
5233 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5234 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5235 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5237 int btrfs_ioctl_get_supported_features(void __user *arg)
5239 static const struct btrfs_ioctl_feature_flags features[3] = {
5240 INIT_FEATURE_FLAGS(SUPP),
5241 INIT_FEATURE_FLAGS(SAFE_SET),
5242 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5245 if (copy_to_user(arg, &features, sizeof(features)))
5251 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
5254 struct btrfs_super_block *super_block = fs_info->super_copy;
5255 struct btrfs_ioctl_feature_flags features;
5257 features.compat_flags = btrfs_super_compat_flags(super_block);
5258 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5259 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5261 if (copy_to_user(arg, &features, sizeof(features)))
5267 static int check_feature_bits(struct btrfs_fs_info *fs_info,
5268 enum btrfs_feature_set set,
5269 u64 change_mask, u64 flags, u64 supported_flags,
5270 u64 safe_set, u64 safe_clear)
5272 const char *type = btrfs_feature_set_name(set);
5274 u64 disallowed, unsupported;
5275 u64 set_mask = flags & change_mask;
5276 u64 clear_mask = ~flags & change_mask;
5278 unsupported = set_mask & ~supported_flags;
5280 names = btrfs_printable_features(set, unsupported);
5283 "this kernel does not support the %s feature bit%s",
5284 names, strchr(names, ',') ? "s" : "");
5288 "this kernel does not support %s bits 0x%llx",
5293 disallowed = set_mask & ~safe_set;
5295 names = btrfs_printable_features(set, disallowed);
5298 "can't set the %s feature bit%s while mounted",
5299 names, strchr(names, ',') ? "s" : "");
5303 "can't set %s bits 0x%llx while mounted",
5308 disallowed = clear_mask & ~safe_clear;
5310 names = btrfs_printable_features(set, disallowed);
5313 "can't clear the %s feature bit%s while mounted",
5314 names, strchr(names, ',') ? "s" : "");
5318 "can't clear %s bits 0x%llx while mounted",
5326 #define check_feature(fs_info, change_mask, flags, mask_base) \
5327 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5328 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5329 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5330 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5332 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5334 struct inode *inode = file_inode(file);
5335 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5336 struct btrfs_root *root = BTRFS_I(inode)->root;
5337 struct btrfs_super_block *super_block = fs_info->super_copy;
5338 struct btrfs_ioctl_feature_flags flags[2];
5339 struct btrfs_trans_handle *trans;
5343 if (!capable(CAP_SYS_ADMIN))
5346 if (copy_from_user(flags, arg, sizeof(flags)))
5350 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5351 !flags[0].incompat_flags)
5354 ret = check_feature(fs_info, flags[0].compat_flags,
5355 flags[1].compat_flags, COMPAT);
5359 ret = check_feature(fs_info, flags[0].compat_ro_flags,
5360 flags[1].compat_ro_flags, COMPAT_RO);
5364 ret = check_feature(fs_info, flags[0].incompat_flags,
5365 flags[1].incompat_flags, INCOMPAT);
5369 ret = mnt_want_write_file(file);
5373 trans = btrfs_start_transaction(root, 0);
5374 if (IS_ERR(trans)) {
5375 ret = PTR_ERR(trans);
5376 goto out_drop_write;
5379 spin_lock(&fs_info->super_lock);
5380 newflags = btrfs_super_compat_flags(super_block);
5381 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5382 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5383 btrfs_set_super_compat_flags(super_block, newflags);
5385 newflags = btrfs_super_compat_ro_flags(super_block);
5386 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5387 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5388 btrfs_set_super_compat_ro_flags(super_block, newflags);
5390 newflags = btrfs_super_incompat_flags(super_block);
5391 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5392 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5393 btrfs_set_super_incompat_flags(super_block, newflags);
5394 spin_unlock(&fs_info->super_lock);
5396 ret = btrfs_commit_transaction(trans);
5398 mnt_drop_write_file(file);
5403 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
5405 struct btrfs_ioctl_send_args *arg;
5409 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5410 struct btrfs_ioctl_send_args_32 args32;
5412 ret = copy_from_user(&args32, argp, sizeof(args32));
5415 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
5418 arg->send_fd = args32.send_fd;
5419 arg->clone_sources_count = args32.clone_sources_count;
5420 arg->clone_sources = compat_ptr(args32.clone_sources);
5421 arg->parent_root = args32.parent_root;
5422 arg->flags = args32.flags;
5423 memcpy(arg->reserved, args32.reserved,
5424 sizeof(args32.reserved));
5429 arg = memdup_user(argp, sizeof(*arg));
5431 return PTR_ERR(arg);
5433 ret = btrfs_ioctl_send(file, arg);
5438 long btrfs_ioctl(struct file *file, unsigned int
5439 cmd, unsigned long arg)
5441 struct inode *inode = file_inode(file);
5442 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5443 struct btrfs_root *root = BTRFS_I(inode)->root;
5444 void __user *argp = (void __user *)arg;
5447 case FS_IOC_GETFLAGS:
5448 return btrfs_ioctl_getflags(file, argp);
5449 case FS_IOC_SETFLAGS:
5450 return btrfs_ioctl_setflags(file, argp);
5451 case FS_IOC_GETVERSION:
5452 return btrfs_ioctl_getversion(file, argp);
5453 case FS_IOC_GETFSLABEL:
5454 return btrfs_ioctl_get_fslabel(fs_info, argp);
5455 case FS_IOC_SETFSLABEL:
5456 return btrfs_ioctl_set_fslabel(file, argp);
5458 return btrfs_ioctl_fitrim(fs_info, argp);
5459 case BTRFS_IOC_SNAP_CREATE:
5460 return btrfs_ioctl_snap_create(file, argp, 0);
5461 case BTRFS_IOC_SNAP_CREATE_V2:
5462 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5463 case BTRFS_IOC_SUBVOL_CREATE:
5464 return btrfs_ioctl_snap_create(file, argp, 1);
5465 case BTRFS_IOC_SUBVOL_CREATE_V2:
5466 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5467 case BTRFS_IOC_SNAP_DESTROY:
5468 return btrfs_ioctl_snap_destroy(file, argp);
5469 case BTRFS_IOC_SUBVOL_GETFLAGS:
5470 return btrfs_ioctl_subvol_getflags(file, argp);
5471 case BTRFS_IOC_SUBVOL_SETFLAGS:
5472 return btrfs_ioctl_subvol_setflags(file, argp);
5473 case BTRFS_IOC_DEFAULT_SUBVOL:
5474 return btrfs_ioctl_default_subvol(file, argp);
5475 case BTRFS_IOC_DEFRAG:
5476 return btrfs_ioctl_defrag(file, NULL);
5477 case BTRFS_IOC_DEFRAG_RANGE:
5478 return btrfs_ioctl_defrag(file, argp);
5479 case BTRFS_IOC_RESIZE:
5480 return btrfs_ioctl_resize(file, argp);
5481 case BTRFS_IOC_ADD_DEV:
5482 return btrfs_ioctl_add_dev(fs_info, argp);
5483 case BTRFS_IOC_RM_DEV:
5484 return btrfs_ioctl_rm_dev(file, argp);
5485 case BTRFS_IOC_RM_DEV_V2:
5486 return btrfs_ioctl_rm_dev_v2(file, argp);
5487 case BTRFS_IOC_FS_INFO:
5488 return btrfs_ioctl_fs_info(fs_info, argp);
5489 case BTRFS_IOC_DEV_INFO:
5490 return btrfs_ioctl_dev_info(fs_info, argp);
5491 case BTRFS_IOC_BALANCE:
5492 return btrfs_ioctl_balance(file, NULL);
5493 case BTRFS_IOC_TREE_SEARCH:
5494 return btrfs_ioctl_tree_search(file, argp);
5495 case BTRFS_IOC_TREE_SEARCH_V2:
5496 return btrfs_ioctl_tree_search_v2(file, argp);
5497 case BTRFS_IOC_INO_LOOKUP:
5498 return btrfs_ioctl_ino_lookup(file, argp);
5499 case BTRFS_IOC_INO_PATHS:
5500 return btrfs_ioctl_ino_to_path(root, argp);
5501 case BTRFS_IOC_LOGICAL_INO:
5502 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
5503 case BTRFS_IOC_LOGICAL_INO_V2:
5504 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
5505 case BTRFS_IOC_SPACE_INFO:
5506 return btrfs_ioctl_space_info(fs_info, argp);
5507 case BTRFS_IOC_SYNC: {
5510 ret = btrfs_start_delalloc_roots(fs_info, -1);
5513 ret = btrfs_sync_fs(inode->i_sb, 1);
5515 * The transaction thread may want to do more work,
5516 * namely it pokes the cleaner kthread that will start
5517 * processing uncleaned subvols.
5519 wake_up_process(fs_info->transaction_kthread);
5522 case BTRFS_IOC_START_SYNC:
5523 return btrfs_ioctl_start_sync(root, argp);
5524 case BTRFS_IOC_WAIT_SYNC:
5525 return btrfs_ioctl_wait_sync(fs_info, argp);
5526 case BTRFS_IOC_SCRUB:
5527 return btrfs_ioctl_scrub(file, argp);
5528 case BTRFS_IOC_SCRUB_CANCEL:
5529 return btrfs_ioctl_scrub_cancel(fs_info);
5530 case BTRFS_IOC_SCRUB_PROGRESS:
5531 return btrfs_ioctl_scrub_progress(fs_info, argp);
5532 case BTRFS_IOC_BALANCE_V2:
5533 return btrfs_ioctl_balance(file, argp);
5534 case BTRFS_IOC_BALANCE_CTL:
5535 return btrfs_ioctl_balance_ctl(fs_info, arg);
5536 case BTRFS_IOC_BALANCE_PROGRESS:
5537 return btrfs_ioctl_balance_progress(fs_info, argp);
5538 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5539 return btrfs_ioctl_set_received_subvol(file, argp);
5541 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5542 return btrfs_ioctl_set_received_subvol_32(file, argp);
5544 case BTRFS_IOC_SEND:
5545 return _btrfs_ioctl_send(file, argp, false);
5546 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5547 case BTRFS_IOC_SEND_32:
5548 return _btrfs_ioctl_send(file, argp, true);
5550 case BTRFS_IOC_GET_DEV_STATS:
5551 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5552 case BTRFS_IOC_QUOTA_CTL:
5553 return btrfs_ioctl_quota_ctl(file, argp);
5554 case BTRFS_IOC_QGROUP_ASSIGN:
5555 return btrfs_ioctl_qgroup_assign(file, argp);
5556 case BTRFS_IOC_QGROUP_CREATE:
5557 return btrfs_ioctl_qgroup_create(file, argp);
5558 case BTRFS_IOC_QGROUP_LIMIT:
5559 return btrfs_ioctl_qgroup_limit(file, argp);
5560 case BTRFS_IOC_QUOTA_RESCAN:
5561 return btrfs_ioctl_quota_rescan(file, argp);
5562 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5563 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
5564 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5565 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
5566 case BTRFS_IOC_DEV_REPLACE:
5567 return btrfs_ioctl_dev_replace(fs_info, argp);
5568 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5569 return btrfs_ioctl_get_supported_features(argp);
5570 case BTRFS_IOC_GET_FEATURES:
5571 return btrfs_ioctl_get_features(fs_info, argp);
5572 case BTRFS_IOC_SET_FEATURES:
5573 return btrfs_ioctl_set_features(file, argp);
5574 case FS_IOC_FSGETXATTR:
5575 return btrfs_ioctl_fsgetxattr(file, argp);
5576 case FS_IOC_FSSETXATTR:
5577 return btrfs_ioctl_fssetxattr(file, argp);
5578 case BTRFS_IOC_GET_SUBVOL_INFO:
5579 return btrfs_ioctl_get_subvol_info(file, argp);
5580 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5581 return btrfs_ioctl_get_subvol_rootref(file, argp);
5582 case BTRFS_IOC_INO_LOOKUP_USER:
5583 return btrfs_ioctl_ino_lookup_user(file, argp);
5589 #ifdef CONFIG_COMPAT
5590 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5593 * These all access 32-bit values anyway so no further
5594 * handling is necessary.
5597 case FS_IOC32_GETFLAGS:
5598 cmd = FS_IOC_GETFLAGS;
5600 case FS_IOC32_SETFLAGS:
5601 cmd = FS_IOC_SETFLAGS;
5603 case FS_IOC32_GETVERSION:
5604 cmd = FS_IOC_GETVERSION;
5608 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));