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>
29 #include <linux/fileattr.h>
30 #include <linux/fsverity.h>
31 #include <linux/sched/xacct.h>
35 #include "transaction.h"
36 #include "btrfs_inode.h"
37 #include "print-tree.h"
41 #include "rcu-string.h"
43 #include "dev-replace.h"
48 #include "compression.h"
49 #include "space-info.h"
50 #include "delalloc-space.h"
51 #include "block-group.h"
55 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
56 * structures are incorrect, as the timespec structure from userspace
57 * is 4 bytes too small. We define these alternatives here to teach
58 * the kernel about the 32-bit struct packing.
60 struct btrfs_ioctl_timespec_32 {
63 } __attribute__ ((__packed__));
65 struct btrfs_ioctl_received_subvol_args_32 {
66 char uuid[BTRFS_UUID_SIZE]; /* in */
67 __u64 stransid; /* in */
68 __u64 rtransid; /* out */
69 struct btrfs_ioctl_timespec_32 stime; /* in */
70 struct btrfs_ioctl_timespec_32 rtime; /* out */
72 __u64 reserved[16]; /* in */
73 } __attribute__ ((__packed__));
75 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
76 struct btrfs_ioctl_received_subvol_args_32)
79 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
80 struct btrfs_ioctl_send_args_32 {
81 __s64 send_fd; /* in */
82 __u64 clone_sources_count; /* in */
83 compat_uptr_t clone_sources; /* in */
84 __u64 parent_root; /* in */
86 __u32 version; /* in */
87 __u8 reserved[28]; /* in */
88 } __attribute__ ((__packed__));
90 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
91 struct btrfs_ioctl_send_args_32)
93 struct btrfs_ioctl_encoded_io_args_32 {
95 compat_ulong_t iovcnt;
100 __u64 unencoded_offset;
106 #define BTRFS_IOC_ENCODED_READ_32 _IOR(BTRFS_IOCTL_MAGIC, 64, \
107 struct btrfs_ioctl_encoded_io_args_32)
108 #define BTRFS_IOC_ENCODED_WRITE_32 _IOW(BTRFS_IOCTL_MAGIC, 64, \
109 struct btrfs_ioctl_encoded_io_args_32)
112 /* Mask out flags that are inappropriate for the given type of inode. */
113 static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
116 if (S_ISDIR(inode->i_mode))
118 else if (S_ISREG(inode->i_mode))
119 return flags & ~FS_DIRSYNC_FL;
121 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
125 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
128 static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode)
130 unsigned int iflags = 0;
131 u32 flags = binode->flags;
132 u32 ro_flags = binode->ro_flags;
134 if (flags & BTRFS_INODE_SYNC)
135 iflags |= FS_SYNC_FL;
136 if (flags & BTRFS_INODE_IMMUTABLE)
137 iflags |= FS_IMMUTABLE_FL;
138 if (flags & BTRFS_INODE_APPEND)
139 iflags |= FS_APPEND_FL;
140 if (flags & BTRFS_INODE_NODUMP)
141 iflags |= FS_NODUMP_FL;
142 if (flags & BTRFS_INODE_NOATIME)
143 iflags |= FS_NOATIME_FL;
144 if (flags & BTRFS_INODE_DIRSYNC)
145 iflags |= FS_DIRSYNC_FL;
146 if (flags & BTRFS_INODE_NODATACOW)
147 iflags |= FS_NOCOW_FL;
148 if (ro_flags & BTRFS_INODE_RO_VERITY)
149 iflags |= FS_VERITY_FL;
151 if (flags & BTRFS_INODE_NOCOMPRESS)
152 iflags |= FS_NOCOMP_FL;
153 else if (flags & BTRFS_INODE_COMPRESS)
154 iflags |= FS_COMPR_FL;
160 * Update inode->i_flags based on the btrfs internal flags.
162 void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
164 struct btrfs_inode *binode = BTRFS_I(inode);
165 unsigned int new_fl = 0;
167 if (binode->flags & BTRFS_INODE_SYNC)
169 if (binode->flags & BTRFS_INODE_IMMUTABLE)
170 new_fl |= S_IMMUTABLE;
171 if (binode->flags & BTRFS_INODE_APPEND)
173 if (binode->flags & BTRFS_INODE_NOATIME)
175 if (binode->flags & BTRFS_INODE_DIRSYNC)
177 if (binode->ro_flags & BTRFS_INODE_RO_VERITY)
180 set_mask_bits(&inode->i_flags,
181 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC |
186 * Check if @flags are a supported and valid set of FS_*_FL flags and that
187 * the old and new flags are not conflicting
189 static int check_fsflags(unsigned int old_flags, unsigned int flags)
191 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
192 FS_NOATIME_FL | FS_NODUMP_FL | \
193 FS_SYNC_FL | FS_DIRSYNC_FL | \
194 FS_NOCOMP_FL | FS_COMPR_FL |
198 /* COMPR and NOCOMP on new/old are valid */
199 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
202 if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
205 /* NOCOW and compression options are mutually exclusive */
206 if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
208 if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
214 static int check_fsflags_compatible(struct btrfs_fs_info *fs_info,
217 if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL))
224 * Set flags/xflags from the internal inode flags. The remaining items of
225 * fsxattr are zeroed.
227 int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
229 struct btrfs_inode *binode = BTRFS_I(d_inode(dentry));
231 fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode));
235 int btrfs_fileattr_set(struct user_namespace *mnt_userns,
236 struct dentry *dentry, struct fileattr *fa)
238 struct inode *inode = d_inode(dentry);
239 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
240 struct btrfs_inode *binode = BTRFS_I(inode);
241 struct btrfs_root *root = binode->root;
242 struct btrfs_trans_handle *trans;
243 unsigned int fsflags, old_fsflags;
245 const char *comp = NULL;
248 if (btrfs_root_readonly(root))
251 if (fileattr_has_fsx(fa))
254 fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags);
255 old_fsflags = btrfs_inode_flags_to_fsflags(binode);
256 ret = check_fsflags(old_fsflags, fsflags);
260 ret = check_fsflags_compatible(fs_info, fsflags);
264 binode_flags = binode->flags;
265 if (fsflags & FS_SYNC_FL)
266 binode_flags |= BTRFS_INODE_SYNC;
268 binode_flags &= ~BTRFS_INODE_SYNC;
269 if (fsflags & FS_IMMUTABLE_FL)
270 binode_flags |= BTRFS_INODE_IMMUTABLE;
272 binode_flags &= ~BTRFS_INODE_IMMUTABLE;
273 if (fsflags & FS_APPEND_FL)
274 binode_flags |= BTRFS_INODE_APPEND;
276 binode_flags &= ~BTRFS_INODE_APPEND;
277 if (fsflags & FS_NODUMP_FL)
278 binode_flags |= BTRFS_INODE_NODUMP;
280 binode_flags &= ~BTRFS_INODE_NODUMP;
281 if (fsflags & FS_NOATIME_FL)
282 binode_flags |= BTRFS_INODE_NOATIME;
284 binode_flags &= ~BTRFS_INODE_NOATIME;
286 /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */
287 if (!fa->flags_valid) {
288 /* 1 item for the inode */
289 trans = btrfs_start_transaction(root, 1);
291 return PTR_ERR(trans);
295 if (fsflags & FS_DIRSYNC_FL)
296 binode_flags |= BTRFS_INODE_DIRSYNC;
298 binode_flags &= ~BTRFS_INODE_DIRSYNC;
299 if (fsflags & FS_NOCOW_FL) {
300 if (S_ISREG(inode->i_mode)) {
302 * It's safe to turn csums off here, no extents exist.
303 * Otherwise we want the flag to reflect the real COW
304 * status of the file and will not set it.
306 if (inode->i_size == 0)
307 binode_flags |= BTRFS_INODE_NODATACOW |
308 BTRFS_INODE_NODATASUM;
310 binode_flags |= BTRFS_INODE_NODATACOW;
314 * Revert back under same assumptions as above
316 if (S_ISREG(inode->i_mode)) {
317 if (inode->i_size == 0)
318 binode_flags &= ~(BTRFS_INODE_NODATACOW |
319 BTRFS_INODE_NODATASUM);
321 binode_flags &= ~BTRFS_INODE_NODATACOW;
326 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
327 * flag may be changed automatically if compression code won't make
330 if (fsflags & FS_NOCOMP_FL) {
331 binode_flags &= ~BTRFS_INODE_COMPRESS;
332 binode_flags |= BTRFS_INODE_NOCOMPRESS;
333 } else if (fsflags & FS_COMPR_FL) {
335 if (IS_SWAPFILE(inode))
338 binode_flags |= BTRFS_INODE_COMPRESS;
339 binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
341 comp = btrfs_compress_type2str(fs_info->compress_type);
342 if (!comp || comp[0] == 0)
343 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
345 binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
352 trans = btrfs_start_transaction(root, 3);
354 return PTR_ERR(trans);
357 ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
360 btrfs_abort_transaction(trans, ret);
364 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
366 if (ret && ret != -ENODATA) {
367 btrfs_abort_transaction(trans, ret);
373 binode->flags = binode_flags;
374 btrfs_sync_inode_flags_to_i_flags(inode);
375 inode_inc_iversion(inode);
376 inode->i_ctime = current_time(inode);
377 ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
380 btrfs_end_transaction(trans);
385 * Start exclusive operation @type, return true on success
387 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
388 enum btrfs_exclusive_operation type)
392 spin_lock(&fs_info->super_lock);
393 if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) {
394 fs_info->exclusive_operation = type;
397 spin_unlock(&fs_info->super_lock);
403 * Conditionally allow to enter the exclusive operation in case it's compatible
404 * with the running one. This must be paired with btrfs_exclop_start_unlock and
405 * btrfs_exclop_finish.
408 * - the same type is already running
409 * - when trying to add a device and balance has been paused
410 * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller
411 * must check the condition first that would allow none -> @type
413 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
414 enum btrfs_exclusive_operation type)
416 spin_lock(&fs_info->super_lock);
417 if (fs_info->exclusive_operation == type ||
418 (fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED &&
419 type == BTRFS_EXCLOP_DEV_ADD))
422 spin_unlock(&fs_info->super_lock);
426 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info)
428 spin_unlock(&fs_info->super_lock);
431 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info)
433 spin_lock(&fs_info->super_lock);
434 WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE);
435 spin_unlock(&fs_info->super_lock);
436 sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation");
439 void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
440 enum btrfs_exclusive_operation op)
443 case BTRFS_EXCLOP_BALANCE_PAUSED:
444 spin_lock(&fs_info->super_lock);
445 ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE ||
446 fs_info->exclusive_operation == BTRFS_EXCLOP_DEV_ADD ||
447 fs_info->exclusive_operation == BTRFS_EXCLOP_NONE ||
448 fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
449 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED;
450 spin_unlock(&fs_info->super_lock);
452 case BTRFS_EXCLOP_BALANCE:
453 spin_lock(&fs_info->super_lock);
454 ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
455 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE;
456 spin_unlock(&fs_info->super_lock);
460 "invalid exclop balance operation %d requested", op);
464 static int btrfs_ioctl_getversion(struct inode *inode, int __user *arg)
466 return put_user(inode->i_generation, arg);
469 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
472 struct btrfs_device *device;
473 struct fstrim_range range;
474 u64 minlen = ULLONG_MAX;
478 if (!capable(CAP_SYS_ADMIN))
482 * btrfs_trim_block_group() depends on space cache, which is not
483 * available in zoned filesystem. So, disallow fitrim on a zoned
484 * filesystem for now.
486 if (btrfs_is_zoned(fs_info))
490 * If the fs is mounted with nologreplay, which requires it to be
491 * mounted in RO mode as well, we can not allow discard on free space
492 * inside block groups, because log trees refer to extents that are not
493 * pinned in a block group's free space cache (pinning the extents is
494 * precisely the first phase of replaying a log tree).
496 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
500 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
502 if (!device->bdev || !bdev_max_discard_sectors(device->bdev))
505 minlen = min_t(u64, bdev_discard_granularity(device->bdev),
512 if (copy_from_user(&range, arg, sizeof(range)))
516 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
517 * block group is in the logical address space, which can be any
518 * sectorsize aligned bytenr in the range [0, U64_MAX].
520 if (range.len < fs_info->sb->s_blocksize)
523 range.minlen = max(range.minlen, minlen);
524 ret = btrfs_trim_fs(fs_info, &range);
528 if (copy_to_user(arg, &range, sizeof(range)))
534 int __pure btrfs_is_empty_uuid(u8 *uuid)
538 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
546 * Calculate the number of transaction items to reserve for creating a subvolume
547 * or snapshot, not including the inode, directory entries, or parent directory.
549 static unsigned int create_subvol_num_items(struct btrfs_qgroup_inherit *inherit)
552 * 1 to add root block
555 * 1 to add root backref
557 * 1 to add qgroup info
558 * 1 to add qgroup limit
560 * Ideally the last two would only be accounted if qgroups are enabled,
561 * but that can change between now and the time we would insert them.
563 unsigned int num_items = 7;
566 /* 2 to add qgroup relations for each inherited qgroup */
567 num_items += 2 * inherit->num_qgroups;
572 static noinline int create_subvol(struct user_namespace *mnt_userns,
573 struct inode *dir, struct dentry *dentry,
574 struct btrfs_qgroup_inherit *inherit)
576 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
577 struct btrfs_trans_handle *trans;
578 struct btrfs_key key;
579 struct btrfs_root_item *root_item;
580 struct btrfs_inode_item *inode_item;
581 struct extent_buffer *leaf;
582 struct btrfs_root *root = BTRFS_I(dir)->root;
583 struct btrfs_root *new_root;
584 struct btrfs_block_rsv block_rsv;
585 struct timespec64 cur_time = current_time(dir);
586 struct btrfs_new_inode_args new_inode_args = {
591 unsigned int trans_num_items;
596 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
600 ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
605 * Don't create subvolume whose level is not zero. Or qgroup will be
606 * screwed up since it assumes subvolume qgroup's level to be 0.
608 if (btrfs_qgroup_level(objectid)) {
613 ret = get_anon_bdev(&anon_dev);
617 new_inode_args.inode = btrfs_new_subvol_inode(mnt_userns, dir);
618 if (!new_inode_args.inode) {
622 ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items);
625 trans_num_items += create_subvol_num_items(inherit);
627 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
628 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
629 trans_num_items, false);
631 goto out_new_inode_args;
633 trans = btrfs_start_transaction(root, 0);
635 ret = PTR_ERR(trans);
636 btrfs_subvolume_release_metadata(root, &block_rsv);
637 goto out_new_inode_args;
639 trans->block_rsv = &block_rsv;
640 trans->bytes_reserved = block_rsv.size;
642 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
646 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
647 BTRFS_NESTING_NORMAL);
653 btrfs_mark_buffer_dirty(leaf);
655 inode_item = &root_item->inode;
656 btrfs_set_stack_inode_generation(inode_item, 1);
657 btrfs_set_stack_inode_size(inode_item, 3);
658 btrfs_set_stack_inode_nlink(inode_item, 1);
659 btrfs_set_stack_inode_nbytes(inode_item,
661 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
663 btrfs_set_root_flags(root_item, 0);
664 btrfs_set_root_limit(root_item, 0);
665 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
667 btrfs_set_root_bytenr(root_item, leaf->start);
668 btrfs_set_root_generation(root_item, trans->transid);
669 btrfs_set_root_level(root_item, 0);
670 btrfs_set_root_refs(root_item, 1);
671 btrfs_set_root_used(root_item, leaf->len);
672 btrfs_set_root_last_snapshot(root_item, 0);
674 btrfs_set_root_generation_v2(root_item,
675 btrfs_root_generation(root_item));
676 generate_random_guid(root_item->uuid);
677 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
678 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
679 root_item->ctime = root_item->otime;
680 btrfs_set_root_ctransid(root_item, trans->transid);
681 btrfs_set_root_otransid(root_item, trans->transid);
683 btrfs_tree_unlock(leaf);
685 btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
687 key.objectid = objectid;
689 key.type = BTRFS_ROOT_ITEM_KEY;
690 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
694 * Since we don't abort the transaction in this case, free the
695 * tree block so that we don't leak space and leave the
696 * filesystem in an inconsistent state (an extent item in the
697 * extent tree with a backreference for a root that does not
700 btrfs_tree_lock(leaf);
701 btrfs_clean_tree_block(leaf);
702 btrfs_tree_unlock(leaf);
703 btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
704 free_extent_buffer(leaf);
708 free_extent_buffer(leaf);
711 new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
712 if (IS_ERR(new_root)) {
713 ret = PTR_ERR(new_root);
714 btrfs_abort_transaction(trans, ret);
717 /* anon_dev is owned by new_root now. */
719 BTRFS_I(new_inode_args.inode)->root = new_root;
720 /* ... and new_root is owned by new_inode_args.inode now. */
722 ret = btrfs_record_root_in_trans(trans, new_root);
724 btrfs_abort_transaction(trans, ret);
728 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
729 BTRFS_UUID_KEY_SUBVOL, objectid);
731 btrfs_abort_transaction(trans, ret);
735 ret = btrfs_create_new_inode(trans, &new_inode_args);
737 btrfs_abort_transaction(trans, ret);
741 d_instantiate_new(dentry, new_inode_args.inode);
742 new_inode_args.inode = NULL;
745 trans->block_rsv = NULL;
746 trans->bytes_reserved = 0;
747 btrfs_subvolume_release_metadata(root, &block_rsv);
750 btrfs_end_transaction(trans);
752 ret = btrfs_commit_transaction(trans);
754 btrfs_new_inode_args_destroy(&new_inode_args);
756 iput(new_inode_args.inode);
759 free_anon_bdev(anon_dev);
765 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
766 struct dentry *dentry, bool readonly,
767 struct btrfs_qgroup_inherit *inherit)
769 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
771 struct btrfs_pending_snapshot *pending_snapshot;
772 unsigned int trans_num_items;
773 struct btrfs_trans_handle *trans;
776 /* We do not support snapshotting right now. */
777 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
779 "extent tree v2 doesn't support snapshotting yet");
783 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
786 if (atomic_read(&root->nr_swapfiles)) {
788 "cannot snapshot subvolume with active swapfile");
792 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
793 if (!pending_snapshot)
796 ret = get_anon_bdev(&pending_snapshot->anon_dev);
799 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
801 pending_snapshot->path = btrfs_alloc_path();
802 if (!pending_snapshot->root_item || !pending_snapshot->path) {
807 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
808 BTRFS_BLOCK_RSV_TEMP);
812 * 1 to update parent inode item
814 trans_num_items = create_subvol_num_items(inherit) + 3;
815 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
816 &pending_snapshot->block_rsv,
817 trans_num_items, false);
821 pending_snapshot->dentry = dentry;
822 pending_snapshot->root = root;
823 pending_snapshot->readonly = readonly;
824 pending_snapshot->dir = dir;
825 pending_snapshot->inherit = inherit;
827 trans = btrfs_start_transaction(root, 0);
829 ret = PTR_ERR(trans);
833 trans->pending_snapshot = pending_snapshot;
835 ret = btrfs_commit_transaction(trans);
839 ret = pending_snapshot->error;
843 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
847 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
849 ret = PTR_ERR(inode);
853 d_instantiate(dentry, inode);
855 pending_snapshot->anon_dev = 0;
857 /* Prevent double freeing of anon_dev */
858 if (ret && pending_snapshot->snap)
859 pending_snapshot->snap->anon_dev = 0;
860 btrfs_put_root(pending_snapshot->snap);
861 btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
863 if (pending_snapshot->anon_dev)
864 free_anon_bdev(pending_snapshot->anon_dev);
865 kfree(pending_snapshot->root_item);
866 btrfs_free_path(pending_snapshot->path);
867 kfree(pending_snapshot);
872 /* copy of may_delete in fs/namei.c()
873 * Check whether we can remove a link victim from directory dir, check
874 * whether the type of victim is right.
875 * 1. We can't do it if dir is read-only (done in permission())
876 * 2. We should have write and exec permissions on dir
877 * 3. We can't remove anything from append-only dir
878 * 4. We can't do anything with immutable dir (done in permission())
879 * 5. If the sticky bit on dir is set we should either
880 * a. be owner of dir, or
881 * b. be owner of victim, or
882 * c. have CAP_FOWNER capability
883 * 6. If the victim is append-only or immutable we can't do anything with
884 * links pointing to it.
885 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
886 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
887 * 9. We can't remove a root or mountpoint.
888 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
889 * nfs_async_unlink().
892 static int btrfs_may_delete(struct user_namespace *mnt_userns,
893 struct inode *dir, struct dentry *victim, int isdir)
897 if (d_really_is_negative(victim))
900 BUG_ON(d_inode(victim->d_parent) != dir);
901 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
903 error = inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC);
908 if (check_sticky(mnt_userns, dir, d_inode(victim)) ||
909 IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
910 IS_SWAPFILE(d_inode(victim)))
913 if (!d_is_dir(victim))
917 } else if (d_is_dir(victim))
921 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
926 /* copy of may_create in fs/namei.c() */
927 static inline int btrfs_may_create(struct user_namespace *mnt_userns,
928 struct inode *dir, struct dentry *child)
930 if (d_really_is_positive(child))
934 if (!fsuidgid_has_mapping(dir->i_sb, mnt_userns))
936 return inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC);
940 * Create a new subvolume below @parent. This is largely modeled after
941 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
942 * inside this filesystem so it's quite a bit simpler.
944 static noinline int btrfs_mksubvol(const struct path *parent,
945 struct user_namespace *mnt_userns,
946 const char *name, int namelen,
947 struct btrfs_root *snap_src,
949 struct btrfs_qgroup_inherit *inherit)
951 struct inode *dir = d_inode(parent->dentry);
952 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
953 struct dentry *dentry;
956 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
960 dentry = lookup_one(mnt_userns, name, parent->dentry, namelen);
961 error = PTR_ERR(dentry);
965 error = btrfs_may_create(mnt_userns, dir, dentry);
970 * even if this name doesn't exist, we may get hash collisions.
971 * check for them now when we can safely fail
973 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
979 down_read(&fs_info->subvol_sem);
981 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
985 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
987 error = create_subvol(mnt_userns, dir, dentry, inherit);
990 fsnotify_mkdir(dir, dentry);
992 up_read(&fs_info->subvol_sem);
996 btrfs_inode_unlock(dir, 0);
1000 static noinline int btrfs_mksnapshot(const struct path *parent,
1001 struct user_namespace *mnt_userns,
1002 const char *name, int namelen,
1003 struct btrfs_root *root,
1005 struct btrfs_qgroup_inherit *inherit)
1008 bool snapshot_force_cow = false;
1011 * Force new buffered writes to reserve space even when NOCOW is
1012 * possible. This is to avoid later writeback (running dealloc) to
1013 * fallback to COW mode and unexpectedly fail with ENOSPC.
1015 btrfs_drew_read_lock(&root->snapshot_lock);
1017 ret = btrfs_start_delalloc_snapshot(root, false);
1022 * All previous writes have started writeback in NOCOW mode, so now
1023 * we force future writes to fallback to COW mode during snapshot
1026 atomic_inc(&root->snapshot_force_cow);
1027 snapshot_force_cow = true;
1029 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
1031 ret = btrfs_mksubvol(parent, mnt_userns, name, namelen,
1032 root, readonly, inherit);
1034 if (snapshot_force_cow)
1035 atomic_dec(&root->snapshot_force_cow);
1036 btrfs_drew_read_unlock(&root->snapshot_lock);
1041 * Defrag specific helper to get an extent map.
1043 * Differences between this and btrfs_get_extent() are:
1045 * - No extent_map will be added to inode->extent_tree
1046 * To reduce memory usage in the long run.
1048 * - Extra optimization to skip file extents older than @newer_than
1049 * By using btrfs_search_forward() we can skip entire file ranges that
1050 * have extents created in past transactions, because btrfs_search_forward()
1051 * will not visit leaves and nodes with a generation smaller than given
1052 * minimal generation threshold (@newer_than).
1054 * Return valid em if we find a file extent matching the requirement.
1055 * Return NULL if we can not find a file extent matching the requirement.
1057 * Return ERR_PTR() for error.
1059 static struct extent_map *defrag_get_extent(struct btrfs_inode *inode,
1060 u64 start, u64 newer_than)
1062 struct btrfs_root *root = inode->root;
1063 struct btrfs_file_extent_item *fi;
1064 struct btrfs_path path = { 0 };
1065 struct extent_map *em;
1066 struct btrfs_key key;
1067 u64 ino = btrfs_ino(inode);
1070 em = alloc_extent_map();
1077 key.type = BTRFS_EXTENT_DATA_KEY;
1081 ret = btrfs_search_forward(root, &key, &path, newer_than);
1084 /* Can't find anything newer */
1088 ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
1092 if (path.slots[0] >= btrfs_header_nritems(path.nodes[0])) {
1094 * If btrfs_search_slot() makes path to point beyond nritems,
1095 * we should not have an empty leaf, as this inode must at
1096 * least have its INODE_ITEM.
1098 ASSERT(btrfs_header_nritems(path.nodes[0]));
1099 path.slots[0] = btrfs_header_nritems(path.nodes[0]) - 1;
1101 btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]);
1102 /* Perfect match, no need to go one slot back */
1103 if (key.objectid == ino && key.type == BTRFS_EXTENT_DATA_KEY &&
1104 key.offset == start)
1107 /* We didn't find a perfect match, needs to go one slot back */
1108 if (path.slots[0] > 0) {
1109 btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]);
1110 if (key.objectid == ino && key.type == BTRFS_EXTENT_DATA_KEY)
1115 /* Iterate through the path to find a file extent covering @start */
1119 if (path.slots[0] >= btrfs_header_nritems(path.nodes[0]))
1122 btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]);
1125 * We may go one slot back to INODE_REF/XATTR item, then
1126 * need to go forward until we reach an EXTENT_DATA.
1127 * But we should still has the correct ino as key.objectid.
1129 if (WARN_ON(key.objectid < ino) || key.type < BTRFS_EXTENT_DATA_KEY)
1132 /* It's beyond our target range, definitely not extent found */
1133 if (key.objectid > ino || key.type > BTRFS_EXTENT_DATA_KEY)
1137 * | |<- File extent ->|
1140 * This means there is a hole between start and key.offset.
1142 if (key.offset > start) {
1144 em->orig_start = start;
1145 em->block_start = EXTENT_MAP_HOLE;
1146 em->len = key.offset - start;
1150 fi = btrfs_item_ptr(path.nodes[0], path.slots[0],
1151 struct btrfs_file_extent_item);
1152 extent_end = btrfs_file_extent_end(&path);
1155 * |<- file extent ->| |
1158 * We haven't reached start, search next slot.
1160 if (extent_end <= start)
1163 /* Now this extent covers @start, convert it to em */
1164 btrfs_extent_item_to_extent_map(inode, &path, fi, false, em);
1167 ret = btrfs_next_item(root, &path);
1173 btrfs_release_path(&path);
1177 btrfs_release_path(&path);
1178 free_extent_map(em);
1182 btrfs_release_path(&path);
1183 free_extent_map(em);
1184 return ERR_PTR(ret);
1187 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start,
1188 u64 newer_than, bool locked)
1190 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1191 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1192 struct extent_map *em;
1193 const u32 sectorsize = BTRFS_I(inode)->root->fs_info->sectorsize;
1196 * hopefully we have this extent in the tree already, try without
1197 * the full extent lock
1199 read_lock(&em_tree->lock);
1200 em = lookup_extent_mapping(em_tree, start, sectorsize);
1201 read_unlock(&em_tree->lock);
1204 * We can get a merged extent, in that case, we need to re-search
1205 * tree to get the original em for defrag.
1207 * If @newer_than is 0 or em::generation < newer_than, we can trust
1208 * this em, as either we don't care about the generation, or the
1209 * merged extent map will be rejected anyway.
1211 if (em && test_bit(EXTENT_FLAG_MERGED, &em->flags) &&
1212 newer_than && em->generation >= newer_than) {
1213 free_extent_map(em);
1218 struct extent_state *cached = NULL;
1219 u64 end = start + sectorsize - 1;
1221 /* get the big lock and read metadata off disk */
1223 lock_extent(io_tree, start, end, &cached);
1224 em = defrag_get_extent(BTRFS_I(inode), start, newer_than);
1226 unlock_extent(io_tree, start, end, &cached);
1235 static u32 get_extent_max_capacity(const struct btrfs_fs_info *fs_info,
1236 const struct extent_map *em)
1238 if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
1239 return BTRFS_MAX_COMPRESSED;
1240 return fs_info->max_extent_size;
1243 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em,
1244 u32 extent_thresh, u64 newer_than, bool locked)
1246 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1247 struct extent_map *next;
1250 /* this is the last extent */
1251 if (em->start + em->len >= i_size_read(inode))
1255 * Here we need to pass @newer_then when checking the next extent, or
1256 * we will hit a case we mark current extent for defrag, but the next
1257 * one will not be a target.
1258 * This will just cause extra IO without really reducing the fragments.
1260 next = defrag_lookup_extent(inode, em->start + em->len, newer_than, locked);
1261 /* No more em or hole */
1262 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1264 if (test_bit(EXTENT_FLAG_PREALLOC, &next->flags))
1267 * If the next extent is at its max capacity, defragging current extent
1268 * makes no sense, as the total number of extents won't change.
1270 if (next->len >= get_extent_max_capacity(fs_info, em))
1272 /* Skip older extent */
1273 if (next->generation < newer_than)
1275 /* Also check extent size */
1276 if (next->len >= extent_thresh)
1281 free_extent_map(next);
1286 * Prepare one page to be defragged.
1290 * - Returned page is locked and has been set up properly.
1291 * - No ordered extent exists in the page.
1292 * - The page is uptodate.
1294 * NOTE: Caller should also wait for page writeback after the cluster is
1295 * prepared, here we don't do writeback wait for each page.
1297 static struct page *defrag_prepare_one_page(struct btrfs_inode *inode,
1300 struct address_space *mapping = inode->vfs_inode.i_mapping;
1301 gfp_t mask = btrfs_alloc_write_mask(mapping);
1302 u64 page_start = (u64)index << PAGE_SHIFT;
1303 u64 page_end = page_start + PAGE_SIZE - 1;
1304 struct extent_state *cached_state = NULL;
1309 page = find_or_create_page(mapping, index, mask);
1311 return ERR_PTR(-ENOMEM);
1314 * Since we can defragment files opened read-only, we can encounter
1315 * transparent huge pages here (see CONFIG_READ_ONLY_THP_FOR_FS). We
1316 * can't do I/O using huge pages yet, so return an error for now.
1317 * Filesystem transparent huge pages are typically only used for
1318 * executables that explicitly enable them, so this isn't very
1321 if (PageCompound(page)) {
1324 return ERR_PTR(-ETXTBSY);
1327 ret = set_page_extent_mapped(page);
1331 return ERR_PTR(ret);
1334 /* Wait for any existing ordered extent in the range */
1336 struct btrfs_ordered_extent *ordered;
1338 lock_extent(&inode->io_tree, page_start, page_end, &cached_state);
1339 ordered = btrfs_lookup_ordered_range(inode, page_start, PAGE_SIZE);
1340 unlock_extent(&inode->io_tree, page_start, page_end,
1346 btrfs_start_ordered_extent(ordered, 1);
1347 btrfs_put_ordered_extent(ordered);
1350 * We unlocked the page above, so we need check if it was
1353 if (page->mapping != mapping || !PagePrivate(page)) {
1361 * Now the page range has no ordered extent any more. Read the page to
1364 if (!PageUptodate(page)) {
1365 btrfs_read_folio(NULL, page_folio(page));
1367 if (page->mapping != mapping || !PagePrivate(page)) {
1372 if (!PageUptodate(page)) {
1375 return ERR_PTR(-EIO);
1381 struct defrag_target_range {
1382 struct list_head list;
1388 * Collect all valid target extents.
1390 * @start: file offset to lookup
1391 * @len: length to lookup
1392 * @extent_thresh: file extent size threshold, any extent size >= this value
1394 * @newer_than: only defrag extents newer than this value
1395 * @do_compress: whether the defrag is doing compression
1396 * if true, @extent_thresh will be ignored and all regular
1397 * file extents meeting @newer_than will be targets.
1398 * @locked: if the range has already held extent lock
1399 * @target_list: list of targets file extents
1401 static int defrag_collect_targets(struct btrfs_inode *inode,
1402 u64 start, u64 len, u32 extent_thresh,
1403 u64 newer_than, bool do_compress,
1404 bool locked, struct list_head *target_list,
1405 u64 *last_scanned_ret)
1407 struct btrfs_fs_info *fs_info = inode->root->fs_info;
1408 bool last_is_target = false;
1412 while (cur < start + len) {
1413 struct extent_map *em;
1414 struct defrag_target_range *new;
1415 bool next_mergeable = true;
1418 last_is_target = false;
1419 em = defrag_lookup_extent(&inode->vfs_inode, cur,
1420 newer_than, locked);
1425 * If the file extent is an inlined one, we may still want to
1426 * defrag it (fallthrough) if it will cause a regular extent.
1427 * This is for users who want to convert inline extents to
1428 * regular ones through max_inline= mount option.
1430 if (em->block_start == EXTENT_MAP_INLINE &&
1431 em->len <= inode->root->fs_info->max_inline)
1434 /* Skip hole/delalloc/preallocated extents */
1435 if (em->block_start == EXTENT_MAP_HOLE ||
1436 em->block_start == EXTENT_MAP_DELALLOC ||
1437 test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
1440 /* Skip older extent */
1441 if (em->generation < newer_than)
1444 /* This em is under writeback, no need to defrag */
1445 if (em->generation == (u64)-1)
1449 * Our start offset might be in the middle of an existing extent
1450 * map, so take that into account.
1452 range_len = em->len - (cur - em->start);
1454 * If this range of the extent map is already flagged for delalloc,
1457 * 1) We could deadlock later, when trying to reserve space for
1458 * delalloc, because in case we can't immediately reserve space
1459 * the flusher can start delalloc and wait for the respective
1460 * ordered extents to complete. The deadlock would happen
1461 * because we do the space reservation while holding the range
1462 * locked, and starting writeback, or finishing an ordered
1463 * extent, requires locking the range;
1465 * 2) If there's delalloc there, it means there's dirty pages for
1466 * which writeback has not started yet (we clean the delalloc
1467 * flag when starting writeback and after creating an ordered
1468 * extent). If we mark pages in an adjacent range for defrag,
1469 * then we will have a larger contiguous range for delalloc,
1470 * very likely resulting in a larger extent after writeback is
1471 * triggered (except in a case of free space fragmentation).
1473 if (test_range_bit(&inode->io_tree, cur, cur + range_len - 1,
1474 EXTENT_DELALLOC, 0, NULL))
1478 * For do_compress case, we want to compress all valid file
1479 * extents, thus no @extent_thresh or mergeable check.
1484 /* Skip too large extent */
1485 if (range_len >= extent_thresh)
1489 * Skip extents already at its max capacity, this is mostly for
1490 * compressed extents, which max cap is only 128K.
1492 if (em->len >= get_extent_max_capacity(fs_info, em))
1496 * Normally there are no more extents after an inline one, thus
1497 * @next_mergeable will normally be false and not defragged.
1498 * So if an inline extent passed all above checks, just add it
1499 * for defrag, and be converted to regular extents.
1501 if (em->block_start == EXTENT_MAP_INLINE)
1504 next_mergeable = defrag_check_next_extent(&inode->vfs_inode, em,
1505 extent_thresh, newer_than, locked);
1506 if (!next_mergeable) {
1507 struct defrag_target_range *last;
1509 /* Empty target list, no way to merge with last entry */
1510 if (list_empty(target_list))
1512 last = list_entry(target_list->prev,
1513 struct defrag_target_range, list);
1514 /* Not mergeable with last entry */
1515 if (last->start + last->len != cur)
1518 /* Mergeable, fall through to add it to @target_list. */
1522 last_is_target = true;
1523 range_len = min(extent_map_end(em), start + len) - cur;
1525 * This one is a good target, check if it can be merged into
1526 * last range of the target list.
1528 if (!list_empty(target_list)) {
1529 struct defrag_target_range *last;
1531 last = list_entry(target_list->prev,
1532 struct defrag_target_range, list);
1533 ASSERT(last->start + last->len <= cur);
1534 if (last->start + last->len == cur) {
1535 /* Mergeable, enlarge the last entry */
1536 last->len += range_len;
1539 /* Fall through to allocate a new entry */
1542 /* Allocate new defrag_target_range */
1543 new = kmalloc(sizeof(*new), GFP_NOFS);
1545 free_extent_map(em);
1550 new->len = range_len;
1551 list_add_tail(&new->list, target_list);
1554 cur = extent_map_end(em);
1555 free_extent_map(em);
1558 struct defrag_target_range *entry;
1559 struct defrag_target_range *tmp;
1561 list_for_each_entry_safe(entry, tmp, target_list, list) {
1562 list_del_init(&entry->list);
1566 if (!ret && last_scanned_ret) {
1568 * If the last extent is not a target, the caller can skip to
1569 * the end of that extent.
1570 * Otherwise, we can only go the end of the specified range.
1572 if (!last_is_target)
1573 *last_scanned_ret = max(cur, *last_scanned_ret);
1575 *last_scanned_ret = max(start + len, *last_scanned_ret);
1580 #define CLUSTER_SIZE (SZ_256K)
1581 static_assert(IS_ALIGNED(CLUSTER_SIZE, PAGE_SIZE));
1584 * Defrag one contiguous target range.
1586 * @inode: target inode
1587 * @target: target range to defrag
1588 * @pages: locked pages covering the defrag range
1589 * @nr_pages: number of locked pages
1591 * Caller should ensure:
1593 * - Pages are prepared
1594 * Pages should be locked, no ordered extent in the pages range,
1597 * - Extent bits are locked
1599 static int defrag_one_locked_target(struct btrfs_inode *inode,
1600 struct defrag_target_range *target,
1601 struct page **pages, int nr_pages,
1602 struct extent_state **cached_state)
1604 struct btrfs_fs_info *fs_info = inode->root->fs_info;
1605 struct extent_changeset *data_reserved = NULL;
1606 const u64 start = target->start;
1607 const u64 len = target->len;
1608 unsigned long last_index = (start + len - 1) >> PAGE_SHIFT;
1609 unsigned long start_index = start >> PAGE_SHIFT;
1610 unsigned long first_index = page_index(pages[0]);
1614 ASSERT(last_index - first_index + 1 <= nr_pages);
1616 ret = btrfs_delalloc_reserve_space(inode, &data_reserved, start, len);
1619 clear_extent_bit(&inode->io_tree, start, start + len - 1,
1620 EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
1621 EXTENT_DEFRAG, cached_state);
1622 set_extent_defrag(&inode->io_tree, start, start + len - 1, cached_state);
1624 /* Update the page status */
1625 for (i = start_index - first_index; i <= last_index - first_index; i++) {
1626 ClearPageChecked(pages[i]);
1627 btrfs_page_clamp_set_dirty(fs_info, pages[i], start, len);
1629 btrfs_delalloc_release_extents(inode, len);
1630 extent_changeset_free(data_reserved);
1635 static int defrag_one_range(struct btrfs_inode *inode, u64 start, u32 len,
1636 u32 extent_thresh, u64 newer_than, bool do_compress,
1637 u64 *last_scanned_ret)
1639 struct extent_state *cached_state = NULL;
1640 struct defrag_target_range *entry;
1641 struct defrag_target_range *tmp;
1642 LIST_HEAD(target_list);
1643 struct page **pages;
1644 const u32 sectorsize = inode->root->fs_info->sectorsize;
1645 u64 last_index = (start + len - 1) >> PAGE_SHIFT;
1646 u64 start_index = start >> PAGE_SHIFT;
1647 unsigned int nr_pages = last_index - start_index + 1;
1651 ASSERT(nr_pages <= CLUSTER_SIZE / PAGE_SIZE);
1652 ASSERT(IS_ALIGNED(start, sectorsize) && IS_ALIGNED(len, sectorsize));
1654 pages = kcalloc(nr_pages, sizeof(struct page *), GFP_NOFS);
1658 /* Prepare all pages */
1659 for (i = 0; i < nr_pages; i++) {
1660 pages[i] = defrag_prepare_one_page(inode, start_index + i);
1661 if (IS_ERR(pages[i])) {
1662 ret = PTR_ERR(pages[i]);
1667 for (i = 0; i < nr_pages; i++)
1668 wait_on_page_writeback(pages[i]);
1670 /* Lock the pages range */
1671 lock_extent(&inode->io_tree, start_index << PAGE_SHIFT,
1672 (last_index << PAGE_SHIFT) + PAGE_SIZE - 1,
1675 * Now we have a consistent view about the extent map, re-check
1676 * which range really needs to be defragged.
1678 * And this time we have extent locked already, pass @locked = true
1679 * so that we won't relock the extent range and cause deadlock.
1681 ret = defrag_collect_targets(inode, start, len, extent_thresh,
1682 newer_than, do_compress, true,
1683 &target_list, last_scanned_ret);
1687 list_for_each_entry(entry, &target_list, list) {
1688 ret = defrag_one_locked_target(inode, entry, pages, nr_pages,
1694 list_for_each_entry_safe(entry, tmp, &target_list, list) {
1695 list_del_init(&entry->list);
1699 unlock_extent(&inode->io_tree, start_index << PAGE_SHIFT,
1700 (last_index << PAGE_SHIFT) + PAGE_SIZE - 1,
1703 for (i = 0; i < nr_pages; i++) {
1705 unlock_page(pages[i]);
1713 static int defrag_one_cluster(struct btrfs_inode *inode,
1714 struct file_ra_state *ra,
1715 u64 start, u32 len, u32 extent_thresh,
1716 u64 newer_than, bool do_compress,
1717 unsigned long *sectors_defragged,
1718 unsigned long max_sectors,
1719 u64 *last_scanned_ret)
1721 const u32 sectorsize = inode->root->fs_info->sectorsize;
1722 struct defrag_target_range *entry;
1723 struct defrag_target_range *tmp;
1724 LIST_HEAD(target_list);
1727 ret = defrag_collect_targets(inode, start, len, extent_thresh,
1728 newer_than, do_compress, false,
1729 &target_list, NULL);
1733 list_for_each_entry(entry, &target_list, list) {
1734 u32 range_len = entry->len;
1736 /* Reached or beyond the limit */
1737 if (max_sectors && *sectors_defragged >= max_sectors) {
1743 range_len = min_t(u32, range_len,
1744 (max_sectors - *sectors_defragged) * sectorsize);
1747 * If defrag_one_range() has updated last_scanned_ret,
1748 * our range may already be invalid (e.g. hole punched).
1749 * Skip if our range is before last_scanned_ret, as there is
1750 * no need to defrag the range anymore.
1752 if (entry->start + range_len <= *last_scanned_ret)
1756 page_cache_sync_readahead(inode->vfs_inode.i_mapping,
1757 ra, NULL, entry->start >> PAGE_SHIFT,
1758 ((entry->start + range_len - 1) >> PAGE_SHIFT) -
1759 (entry->start >> PAGE_SHIFT) + 1);
1761 * Here we may not defrag any range if holes are punched before
1762 * we locked the pages.
1763 * But that's fine, it only affects the @sectors_defragged
1766 ret = defrag_one_range(inode, entry->start, range_len,
1767 extent_thresh, newer_than, do_compress,
1771 *sectors_defragged += range_len >>
1772 inode->root->fs_info->sectorsize_bits;
1775 list_for_each_entry_safe(entry, tmp, &target_list, list) {
1776 list_del_init(&entry->list);
1780 *last_scanned_ret = max(*last_scanned_ret, start + len);
1785 * Entry point to file defragmentation.
1787 * @inode: inode to be defragged
1788 * @ra: readahead state (can be NUL)
1789 * @range: defrag options including range and flags
1790 * @newer_than: minimum transid to defrag
1791 * @max_to_defrag: max number of sectors to be defragged, if 0, the whole inode
1792 * will be defragged.
1794 * Return <0 for error.
1795 * Return >=0 for the number of sectors defragged, and range->start will be updated
1796 * to indicate the file offset where next defrag should be started at.
1797 * (Mostly for autodefrag, which sets @max_to_defrag thus we may exit early without
1798 * defragging all the range).
1800 int btrfs_defrag_file(struct inode *inode, struct file_ra_state *ra,
1801 struct btrfs_ioctl_defrag_range_args *range,
1802 u64 newer_than, unsigned long max_to_defrag)
1804 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1805 unsigned long sectors_defragged = 0;
1806 u64 isize = i_size_read(inode);
1809 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1810 bool ra_allocated = false;
1811 int compress_type = BTRFS_COMPRESS_ZLIB;
1813 u32 extent_thresh = range->extent_thresh;
1814 pgoff_t start_index;
1819 if (range->start >= isize)
1823 if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES)
1825 if (range->compress_type)
1826 compress_type = range->compress_type;
1829 if (extent_thresh == 0)
1830 extent_thresh = SZ_256K;
1832 if (range->start + range->len > range->start) {
1833 /* Got a specific range */
1834 last_byte = min(isize, range->start + range->len);
1836 /* Defrag until file end */
1840 /* Align the range */
1841 cur = round_down(range->start, fs_info->sectorsize);
1842 last_byte = round_up(last_byte, fs_info->sectorsize) - 1;
1845 * If we were not given a ra, allocate a readahead context. As
1846 * readahead is just an optimization, defrag will work without it so
1847 * we don't error out.
1850 ra_allocated = true;
1851 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1853 file_ra_state_init(ra, inode->i_mapping);
1857 * Make writeback start from the beginning of the range, so that the
1858 * defrag range can be written sequentially.
1860 start_index = cur >> PAGE_SHIFT;
1861 if (start_index < inode->i_mapping->writeback_index)
1862 inode->i_mapping->writeback_index = start_index;
1864 while (cur < last_byte) {
1865 const unsigned long prev_sectors_defragged = sectors_defragged;
1866 u64 last_scanned = cur;
1869 if (btrfs_defrag_cancelled(fs_info)) {
1874 /* We want the cluster end at page boundary when possible */
1875 cluster_end = (((cur >> PAGE_SHIFT) +
1876 (SZ_256K >> PAGE_SHIFT)) << PAGE_SHIFT) - 1;
1877 cluster_end = min(cluster_end, last_byte);
1879 btrfs_inode_lock(inode, 0);
1880 if (IS_SWAPFILE(inode)) {
1882 btrfs_inode_unlock(inode, 0);
1885 if (!(inode->i_sb->s_flags & SB_ACTIVE)) {
1886 btrfs_inode_unlock(inode, 0);
1890 BTRFS_I(inode)->defrag_compress = compress_type;
1891 ret = defrag_one_cluster(BTRFS_I(inode), ra, cur,
1892 cluster_end + 1 - cur, extent_thresh,
1893 newer_than, do_compress, §ors_defragged,
1894 max_to_defrag, &last_scanned);
1896 if (sectors_defragged > prev_sectors_defragged)
1897 balance_dirty_pages_ratelimited(inode->i_mapping);
1899 btrfs_inode_unlock(inode, 0);
1902 cur = max(cluster_end + 1, last_scanned);
1913 * Update range.start for autodefrag, this will indicate where to start
1917 if (sectors_defragged) {
1919 * We have defragged some sectors, for compression case they
1920 * need to be written back immediately.
1922 if (range->flags & BTRFS_DEFRAG_RANGE_START_IO) {
1923 filemap_flush(inode->i_mapping);
1924 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1925 &BTRFS_I(inode)->runtime_flags))
1926 filemap_flush(inode->i_mapping);
1928 if (range->compress_type == BTRFS_COMPRESS_LZO)
1929 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1930 else if (range->compress_type == BTRFS_COMPRESS_ZSTD)
1931 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1932 ret = sectors_defragged;
1935 btrfs_inode_lock(inode, 0);
1936 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1937 btrfs_inode_unlock(inode, 0);
1943 * Try to start exclusive operation @type or cancel it if it's running.
1946 * 0 - normal mode, newly claimed op started
1947 * >0 - normal mode, something else is running,
1948 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1949 * ECANCELED - cancel mode, successful cancel
1950 * ENOTCONN - cancel mode, operation not running anymore
1952 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1953 enum btrfs_exclusive_operation type, bool cancel)
1956 /* Start normal op */
1957 if (!btrfs_exclop_start(fs_info, type))
1958 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1959 /* Exclusive operation is now claimed */
1963 /* Cancel running op */
1964 if (btrfs_exclop_start_try_lock(fs_info, type)) {
1966 * This blocks any exclop finish from setting it to NONE, so we
1967 * request cancellation. Either it runs and we will wait for it,
1968 * or it has finished and no waiting will happen.
1970 atomic_inc(&fs_info->reloc_cancel_req);
1971 btrfs_exclop_start_unlock(fs_info);
1973 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1974 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1975 TASK_INTERRUPTIBLE);
1980 /* Something else is running or none */
1984 static noinline int btrfs_ioctl_resize(struct file *file,
1987 BTRFS_DEV_LOOKUP_ARGS(args);
1988 struct inode *inode = file_inode(file);
1989 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1993 struct btrfs_root *root = BTRFS_I(inode)->root;
1994 struct btrfs_ioctl_vol_args *vol_args;
1995 struct btrfs_trans_handle *trans;
1996 struct btrfs_device *device = NULL;
1999 char *devstr = NULL;
2004 if (!capable(CAP_SYS_ADMIN))
2007 ret = mnt_want_write_file(file);
2012 * Read the arguments before checking exclusivity to be able to
2013 * distinguish regular resize and cancel
2015 vol_args = memdup_user(arg, sizeof(*vol_args));
2016 if (IS_ERR(vol_args)) {
2017 ret = PTR_ERR(vol_args);
2020 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2021 sizestr = vol_args->name;
2022 cancel = (strcmp("cancel", sizestr) == 0);
2023 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
2026 /* Exclusive operation is now claimed */
2028 devstr = strchr(sizestr, ':');
2030 sizestr = devstr + 1;
2032 devstr = vol_args->name;
2033 ret = kstrtoull(devstr, 10, &devid);
2040 btrfs_info(fs_info, "resizing devid %llu", devid);
2044 device = btrfs_find_device(fs_info->fs_devices, &args);
2046 btrfs_info(fs_info, "resizer unable to find device %llu",
2052 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
2054 "resizer unable to apply on readonly device %llu",
2060 if (!strcmp(sizestr, "max"))
2061 new_size = bdev_nr_bytes(device->bdev);
2063 if (sizestr[0] == '-') {
2066 } else if (sizestr[0] == '+') {
2070 new_size = memparse(sizestr, &retptr);
2071 if (*retptr != '\0' || new_size == 0) {
2077 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
2082 old_size = btrfs_device_get_total_bytes(device);
2085 if (new_size > old_size) {
2089 new_size = old_size - new_size;
2090 } else if (mod > 0) {
2091 if (new_size > ULLONG_MAX - old_size) {
2095 new_size = old_size + new_size;
2098 if (new_size < SZ_256M) {
2102 if (new_size > bdev_nr_bytes(device->bdev)) {
2107 new_size = round_down(new_size, fs_info->sectorsize);
2109 if (new_size > old_size) {
2110 trans = btrfs_start_transaction(root, 0);
2111 if (IS_ERR(trans)) {
2112 ret = PTR_ERR(trans);
2115 ret = btrfs_grow_device(trans, device, new_size);
2116 btrfs_commit_transaction(trans);
2117 } else if (new_size < old_size) {
2118 ret = btrfs_shrink_device(device, new_size);
2119 } /* equal, nothing need to do */
2121 if (ret == 0 && new_size != old_size)
2122 btrfs_info_in_rcu(fs_info,
2123 "resize device %s (devid %llu) from %llu to %llu",
2124 rcu_str_deref(device->name), device->devid,
2125 old_size, new_size);
2127 btrfs_exclop_finish(fs_info);
2131 mnt_drop_write_file(file);
2135 static noinline int __btrfs_ioctl_snap_create(struct file *file,
2136 struct user_namespace *mnt_userns,
2137 const char *name, unsigned long fd, int subvol,
2139 struct btrfs_qgroup_inherit *inherit)
2144 if (!S_ISDIR(file_inode(file)->i_mode))
2147 ret = mnt_want_write_file(file);
2151 namelen = strlen(name);
2152 if (strchr(name, '/')) {
2154 goto out_drop_write;
2157 if (name[0] == '.' &&
2158 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
2160 goto out_drop_write;
2164 ret = btrfs_mksubvol(&file->f_path, mnt_userns, name,
2165 namelen, NULL, readonly, inherit);
2167 struct fd src = fdget(fd);
2168 struct inode *src_inode;
2171 goto out_drop_write;
2174 src_inode = file_inode(src.file);
2175 if (src_inode->i_sb != file_inode(file)->i_sb) {
2176 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
2177 "Snapshot src from another FS");
2179 } else if (!inode_owner_or_capable(mnt_userns, src_inode)) {
2181 * Subvolume creation is not restricted, but snapshots
2182 * are limited to own subvolumes only
2186 ret = btrfs_mksnapshot(&file->f_path, mnt_userns,
2188 BTRFS_I(src_inode)->root,
2194 mnt_drop_write_file(file);
2199 static noinline int btrfs_ioctl_snap_create(struct file *file,
2200 void __user *arg, int subvol)
2202 struct btrfs_ioctl_vol_args *vol_args;
2205 if (!S_ISDIR(file_inode(file)->i_mode))
2208 vol_args = memdup_user(arg, sizeof(*vol_args));
2209 if (IS_ERR(vol_args))
2210 return PTR_ERR(vol_args);
2211 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2213 ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file),
2214 vol_args->name, vol_args->fd, subvol,
2221 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
2222 void __user *arg, int subvol)
2224 struct btrfs_ioctl_vol_args_v2 *vol_args;
2226 bool readonly = false;
2227 struct btrfs_qgroup_inherit *inherit = NULL;
2229 if (!S_ISDIR(file_inode(file)->i_mode))
2232 vol_args = memdup_user(arg, sizeof(*vol_args));
2233 if (IS_ERR(vol_args))
2234 return PTR_ERR(vol_args);
2235 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2237 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
2242 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
2244 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
2247 if (vol_args->size < sizeof(*inherit) ||
2248 vol_args->size > PAGE_SIZE) {
2252 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
2253 if (IS_ERR(inherit)) {
2254 ret = PTR_ERR(inherit);
2258 if (inherit->num_qgroups > PAGE_SIZE ||
2259 inherit->num_ref_copies > PAGE_SIZE ||
2260 inherit->num_excl_copies > PAGE_SIZE) {
2265 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
2266 2 * inherit->num_excl_copies;
2267 if (vol_args->size != struct_size(inherit, qgroups, nums)) {
2273 ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file),
2274 vol_args->name, vol_args->fd, subvol,
2285 static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode,
2288 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2289 struct btrfs_root *root = BTRFS_I(inode)->root;
2293 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
2296 down_read(&fs_info->subvol_sem);
2297 if (btrfs_root_readonly(root))
2298 flags |= BTRFS_SUBVOL_RDONLY;
2299 up_read(&fs_info->subvol_sem);
2301 if (copy_to_user(arg, &flags, sizeof(flags)))
2307 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
2310 struct inode *inode = file_inode(file);
2311 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2312 struct btrfs_root *root = BTRFS_I(inode)->root;
2313 struct btrfs_trans_handle *trans;
2318 if (!inode_owner_or_capable(file_mnt_user_ns(file), inode))
2321 ret = mnt_want_write_file(file);
2325 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2327 goto out_drop_write;
2330 if (copy_from_user(&flags, arg, sizeof(flags))) {
2332 goto out_drop_write;
2335 if (flags & ~BTRFS_SUBVOL_RDONLY) {
2337 goto out_drop_write;
2340 down_write(&fs_info->subvol_sem);
2343 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
2346 root_flags = btrfs_root_flags(&root->root_item);
2347 if (flags & BTRFS_SUBVOL_RDONLY) {
2348 btrfs_set_root_flags(&root->root_item,
2349 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2352 * Block RO -> RW transition if this subvolume is involved in
2355 spin_lock(&root->root_item_lock);
2356 if (root->send_in_progress == 0) {
2357 btrfs_set_root_flags(&root->root_item,
2358 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2359 spin_unlock(&root->root_item_lock);
2361 spin_unlock(&root->root_item_lock);
2363 "Attempt to set subvolume %llu read-write during send",
2364 root->root_key.objectid);
2370 trans = btrfs_start_transaction(root, 1);
2371 if (IS_ERR(trans)) {
2372 ret = PTR_ERR(trans);
2376 ret = btrfs_update_root(trans, fs_info->tree_root,
2377 &root->root_key, &root->root_item);
2379 btrfs_end_transaction(trans);
2383 ret = btrfs_commit_transaction(trans);
2387 btrfs_set_root_flags(&root->root_item, root_flags);
2389 up_write(&fs_info->subvol_sem);
2391 mnt_drop_write_file(file);
2396 static noinline int key_in_sk(struct btrfs_key *key,
2397 struct btrfs_ioctl_search_key *sk)
2399 struct btrfs_key test;
2402 test.objectid = sk->min_objectid;
2403 test.type = sk->min_type;
2404 test.offset = sk->min_offset;
2406 ret = btrfs_comp_cpu_keys(key, &test);
2410 test.objectid = sk->max_objectid;
2411 test.type = sk->max_type;
2412 test.offset = sk->max_offset;
2414 ret = btrfs_comp_cpu_keys(key, &test);
2420 static noinline int copy_to_sk(struct btrfs_path *path,
2421 struct btrfs_key *key,
2422 struct btrfs_ioctl_search_key *sk,
2425 unsigned long *sk_offset,
2429 struct extent_buffer *leaf;
2430 struct btrfs_ioctl_search_header sh;
2431 struct btrfs_key test;
2432 unsigned long item_off;
2433 unsigned long item_len;
2439 leaf = path->nodes[0];
2440 slot = path->slots[0];
2441 nritems = btrfs_header_nritems(leaf);
2443 if (btrfs_header_generation(leaf) > sk->max_transid) {
2447 found_transid = btrfs_header_generation(leaf);
2449 for (i = slot; i < nritems; i++) {
2450 item_off = btrfs_item_ptr_offset(leaf, i);
2451 item_len = btrfs_item_size(leaf, i);
2453 btrfs_item_key_to_cpu(leaf, key, i);
2454 if (!key_in_sk(key, sk))
2457 if (sizeof(sh) + item_len > *buf_size) {
2464 * return one empty item back for v1, which does not
2468 *buf_size = sizeof(sh) + item_len;
2473 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2478 sh.objectid = key->objectid;
2479 sh.offset = key->offset;
2480 sh.type = key->type;
2482 sh.transid = found_transid;
2485 * Copy search result header. If we fault then loop again so we
2486 * can fault in the pages and -EFAULT there if there's a
2487 * problem. Otherwise we'll fault and then copy the buffer in
2488 * properly this next time through
2490 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
2495 *sk_offset += sizeof(sh);
2498 char __user *up = ubuf + *sk_offset;
2500 * Copy the item, same behavior as above, but reset the
2501 * * sk_offset so we copy the full thing again.
2503 if (read_extent_buffer_to_user_nofault(leaf, up,
2504 item_off, item_len)) {
2506 *sk_offset -= sizeof(sh);
2510 *sk_offset += item_len;
2514 if (ret) /* -EOVERFLOW from above */
2517 if (*num_found >= sk->nr_items) {
2524 test.objectid = sk->max_objectid;
2525 test.type = sk->max_type;
2526 test.offset = sk->max_offset;
2527 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2529 else if (key->offset < (u64)-1)
2531 else if (key->type < (u8)-1) {
2534 } else if (key->objectid < (u64)-1) {
2542 * 0: all items from this leaf copied, continue with next
2543 * 1: * more items can be copied, but unused buffer is too small
2544 * * all items were found
2545 * Either way, it will stops the loop which iterates to the next
2547 * -EOVERFLOW: item was to large for buffer
2548 * -EFAULT: could not copy extent buffer back to userspace
2553 static noinline int search_ioctl(struct inode *inode,
2554 struct btrfs_ioctl_search_key *sk,
2558 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2559 struct btrfs_root *root;
2560 struct btrfs_key key;
2561 struct btrfs_path *path;
2564 unsigned long sk_offset = 0;
2566 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2567 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2571 path = btrfs_alloc_path();
2575 if (sk->tree_id == 0) {
2576 /* search the root of the inode that was passed */
2577 root = btrfs_grab_root(BTRFS_I(inode)->root);
2579 root = btrfs_get_fs_root(info, sk->tree_id, true);
2581 btrfs_free_path(path);
2582 return PTR_ERR(root);
2586 key.objectid = sk->min_objectid;
2587 key.type = sk->min_type;
2588 key.offset = sk->min_offset;
2593 * Ensure that the whole user buffer is faulted in at sub-page
2594 * granularity, otherwise the loop may live-lock.
2596 if (fault_in_subpage_writeable(ubuf + sk_offset,
2597 *buf_size - sk_offset))
2600 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2606 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2607 &sk_offset, &num_found);
2608 btrfs_release_path(path);
2616 sk->nr_items = num_found;
2617 btrfs_put_root(root);
2618 btrfs_free_path(path);
2622 static noinline int btrfs_ioctl_tree_search(struct inode *inode,
2625 struct btrfs_ioctl_search_args __user *uargs = argp;
2626 struct btrfs_ioctl_search_key sk;
2630 if (!capable(CAP_SYS_ADMIN))
2633 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2636 buf_size = sizeof(uargs->buf);
2638 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2641 * In the origin implementation an overflow is handled by returning a
2642 * search header with a len of zero, so reset ret.
2644 if (ret == -EOVERFLOW)
2647 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2652 static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode,
2655 struct btrfs_ioctl_search_args_v2 __user *uarg = argp;
2656 struct btrfs_ioctl_search_args_v2 args;
2659 const size_t buf_limit = SZ_16M;
2661 if (!capable(CAP_SYS_ADMIN))
2664 /* copy search header and buffer size */
2665 if (copy_from_user(&args, uarg, sizeof(args)))
2668 buf_size = args.buf_size;
2670 /* limit result size to 16MB */
2671 if (buf_size > buf_limit)
2672 buf_size = buf_limit;
2674 ret = search_ioctl(inode, &args.key, &buf_size,
2675 (char __user *)(&uarg->buf[0]));
2676 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2678 else if (ret == -EOVERFLOW &&
2679 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2686 * Search INODE_REFs to identify path name of 'dirid' directory
2687 * in a 'tree_id' tree. and sets path name to 'name'.
2689 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2690 u64 tree_id, u64 dirid, char *name)
2692 struct btrfs_root *root;
2693 struct btrfs_key key;
2699 struct btrfs_inode_ref *iref;
2700 struct extent_buffer *l;
2701 struct btrfs_path *path;
2703 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2708 path = btrfs_alloc_path();
2712 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2714 root = btrfs_get_fs_root(info, tree_id, true);
2716 ret = PTR_ERR(root);
2721 key.objectid = dirid;
2722 key.type = BTRFS_INODE_REF_KEY;
2723 key.offset = (u64)-1;
2726 ret = btrfs_search_backwards(root, &key, path);
2735 slot = path->slots[0];
2737 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2738 len = btrfs_inode_ref_name_len(l, iref);
2740 total_len += len + 1;
2742 ret = -ENAMETOOLONG;
2747 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2749 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2752 btrfs_release_path(path);
2753 key.objectid = key.offset;
2754 key.offset = (u64)-1;
2755 dirid = key.objectid;
2757 memmove(name, ptr, total_len);
2758 name[total_len] = '\0';
2761 btrfs_put_root(root);
2762 btrfs_free_path(path);
2766 static int btrfs_search_path_in_tree_user(struct user_namespace *mnt_userns,
2767 struct inode *inode,
2768 struct btrfs_ioctl_ino_lookup_user_args *args)
2770 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2771 struct super_block *sb = inode->i_sb;
2772 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2773 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2774 u64 dirid = args->dirid;
2775 unsigned long item_off;
2776 unsigned long item_len;
2777 struct btrfs_inode_ref *iref;
2778 struct btrfs_root_ref *rref;
2779 struct btrfs_root *root = NULL;
2780 struct btrfs_path *path;
2781 struct btrfs_key key, key2;
2782 struct extent_buffer *leaf;
2783 struct inode *temp_inode;
2790 path = btrfs_alloc_path();
2795 * If the bottom subvolume does not exist directly under upper_limit,
2796 * construct the path in from the bottom up.
2798 if (dirid != upper_limit.objectid) {
2799 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2801 root = btrfs_get_fs_root(fs_info, treeid, true);
2803 ret = PTR_ERR(root);
2807 key.objectid = dirid;
2808 key.type = BTRFS_INODE_REF_KEY;
2809 key.offset = (u64)-1;
2811 ret = btrfs_search_backwards(root, &key, path);
2819 leaf = path->nodes[0];
2820 slot = path->slots[0];
2822 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2823 len = btrfs_inode_ref_name_len(leaf, iref);
2825 total_len += len + 1;
2826 if (ptr < args->path) {
2827 ret = -ENAMETOOLONG;
2832 read_extent_buffer(leaf, ptr,
2833 (unsigned long)(iref + 1), len);
2835 /* Check the read+exec permission of this directory */
2836 ret = btrfs_previous_item(root, path, dirid,
2837 BTRFS_INODE_ITEM_KEY);
2840 } else if (ret > 0) {
2845 leaf = path->nodes[0];
2846 slot = path->slots[0];
2847 btrfs_item_key_to_cpu(leaf, &key2, slot);
2848 if (key2.objectid != dirid) {
2853 temp_inode = btrfs_iget(sb, key2.objectid, root);
2854 if (IS_ERR(temp_inode)) {
2855 ret = PTR_ERR(temp_inode);
2858 ret = inode_permission(mnt_userns, temp_inode,
2859 MAY_READ | MAY_EXEC);
2866 if (key.offset == upper_limit.objectid)
2868 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2873 btrfs_release_path(path);
2874 key.objectid = key.offset;
2875 key.offset = (u64)-1;
2876 dirid = key.objectid;
2879 memmove(args->path, ptr, total_len);
2880 args->path[total_len] = '\0';
2881 btrfs_put_root(root);
2883 btrfs_release_path(path);
2886 /* Get the bottom subvolume's name from ROOT_REF */
2887 key.objectid = treeid;
2888 key.type = BTRFS_ROOT_REF_KEY;
2889 key.offset = args->treeid;
2890 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2893 } else if (ret > 0) {
2898 leaf = path->nodes[0];
2899 slot = path->slots[0];
2900 btrfs_item_key_to_cpu(leaf, &key, slot);
2902 item_off = btrfs_item_ptr_offset(leaf, slot);
2903 item_len = btrfs_item_size(leaf, slot);
2904 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2905 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2906 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2911 /* Copy subvolume's name */
2912 item_off += sizeof(struct btrfs_root_ref);
2913 item_len -= sizeof(struct btrfs_root_ref);
2914 read_extent_buffer(leaf, args->name, item_off, item_len);
2915 args->name[item_len] = 0;
2918 btrfs_put_root(root);
2920 btrfs_free_path(path);
2924 static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root,
2927 struct btrfs_ioctl_ino_lookup_args *args;
2930 args = memdup_user(argp, sizeof(*args));
2932 return PTR_ERR(args);
2935 * Unprivileged query to obtain the containing subvolume root id. The
2936 * path is reset so it's consistent with btrfs_search_path_in_tree.
2938 if (args->treeid == 0)
2939 args->treeid = root->root_key.objectid;
2941 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2946 if (!capable(CAP_SYS_ADMIN)) {
2951 ret = btrfs_search_path_in_tree(root->fs_info,
2952 args->treeid, args->objectid,
2956 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2964 * Version of ino_lookup ioctl (unprivileged)
2966 * The main differences from ino_lookup ioctl are:
2968 * 1. Read + Exec permission will be checked using inode_permission() during
2969 * path construction. -EACCES will be returned in case of failure.
2970 * 2. Path construction will be stopped at the inode number which corresponds
2971 * to the fd with which this ioctl is called. If constructed path does not
2972 * exist under fd's inode, -EACCES will be returned.
2973 * 3. The name of bottom subvolume is also searched and filled.
2975 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2977 struct btrfs_ioctl_ino_lookup_user_args *args;
2978 struct inode *inode;
2981 args = memdup_user(argp, sizeof(*args));
2983 return PTR_ERR(args);
2985 inode = file_inode(file);
2987 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2988 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2990 * The subvolume does not exist under fd with which this is
2997 ret = btrfs_search_path_in_tree_user(file_mnt_user_ns(file), inode, args);
2999 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
3006 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
3007 static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp)
3009 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
3010 struct btrfs_fs_info *fs_info;
3011 struct btrfs_root *root;
3012 struct btrfs_path *path;
3013 struct btrfs_key key;
3014 struct btrfs_root_item *root_item;
3015 struct btrfs_root_ref *rref;
3016 struct extent_buffer *leaf;
3017 unsigned long item_off;
3018 unsigned long item_len;
3022 path = btrfs_alloc_path();
3026 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
3028 btrfs_free_path(path);
3032 fs_info = BTRFS_I(inode)->root->fs_info;
3034 /* Get root_item of inode's subvolume */
3035 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
3036 root = btrfs_get_fs_root(fs_info, key.objectid, true);
3038 ret = PTR_ERR(root);
3041 root_item = &root->root_item;
3043 subvol_info->treeid = key.objectid;
3045 subvol_info->generation = btrfs_root_generation(root_item);
3046 subvol_info->flags = btrfs_root_flags(root_item);
3048 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
3049 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
3051 memcpy(subvol_info->received_uuid, root_item->received_uuid,
3054 subvol_info->ctransid = btrfs_root_ctransid(root_item);
3055 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
3056 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
3058 subvol_info->otransid = btrfs_root_otransid(root_item);
3059 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
3060 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
3062 subvol_info->stransid = btrfs_root_stransid(root_item);
3063 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
3064 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
3066 subvol_info->rtransid = btrfs_root_rtransid(root_item);
3067 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
3068 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
3070 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
3071 /* Search root tree for ROOT_BACKREF of this subvolume */
3072 key.type = BTRFS_ROOT_BACKREF_KEY;
3074 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
3077 } else if (path->slots[0] >=
3078 btrfs_header_nritems(path->nodes[0])) {
3079 ret = btrfs_next_leaf(fs_info->tree_root, path);
3082 } else if (ret > 0) {
3088 leaf = path->nodes[0];
3089 slot = path->slots[0];
3090 btrfs_item_key_to_cpu(leaf, &key, slot);
3091 if (key.objectid == subvol_info->treeid &&
3092 key.type == BTRFS_ROOT_BACKREF_KEY) {
3093 subvol_info->parent_id = key.offset;
3095 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
3096 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
3098 item_off = btrfs_item_ptr_offset(leaf, slot)
3099 + sizeof(struct btrfs_root_ref);
3100 item_len = btrfs_item_size(leaf, slot)
3101 - sizeof(struct btrfs_root_ref);
3102 read_extent_buffer(leaf, subvol_info->name,
3103 item_off, item_len);
3110 btrfs_free_path(path);
3112 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
3116 btrfs_put_root(root);
3118 btrfs_free_path(path);
3124 * Return ROOT_REF information of the subvolume containing this inode
3125 * except the subvolume name.
3127 static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root,
3130 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
3131 struct btrfs_root_ref *rref;
3132 struct btrfs_path *path;
3133 struct btrfs_key key;
3134 struct extent_buffer *leaf;
3140 path = btrfs_alloc_path();
3144 rootrefs = memdup_user(argp, sizeof(*rootrefs));
3145 if (IS_ERR(rootrefs)) {
3146 btrfs_free_path(path);
3147 return PTR_ERR(rootrefs);
3150 objectid = root->root_key.objectid;
3151 key.objectid = objectid;
3152 key.type = BTRFS_ROOT_REF_KEY;
3153 key.offset = rootrefs->min_treeid;
3156 root = root->fs_info->tree_root;
3157 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3160 } else if (path->slots[0] >=
3161 btrfs_header_nritems(path->nodes[0])) {
3162 ret = btrfs_next_leaf(root, path);
3165 } else if (ret > 0) {
3171 leaf = path->nodes[0];
3172 slot = path->slots[0];
3174 btrfs_item_key_to_cpu(leaf, &key, slot);
3175 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
3180 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
3185 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
3186 rootrefs->rootref[found].treeid = key.offset;
3187 rootrefs->rootref[found].dirid =
3188 btrfs_root_ref_dirid(leaf, rref);
3191 ret = btrfs_next_item(root, path);
3194 } else if (ret > 0) {
3201 btrfs_free_path(path);
3203 if (!ret || ret == -EOVERFLOW) {
3204 rootrefs->num_items = found;
3205 /* update min_treeid for next search */
3207 rootrefs->min_treeid =
3208 rootrefs->rootref[found - 1].treeid + 1;
3209 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
3218 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
3222 struct dentry *parent = file->f_path.dentry;
3223 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
3224 struct dentry *dentry;
3225 struct inode *dir = d_inode(parent);
3226 struct inode *inode;
3227 struct btrfs_root *root = BTRFS_I(dir)->root;
3228 struct btrfs_root *dest = NULL;
3229 struct btrfs_ioctl_vol_args *vol_args = NULL;
3230 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
3231 struct user_namespace *mnt_userns = file_mnt_user_ns(file);
3232 char *subvol_name, *subvol_name_ptr = NULL;
3235 bool destroy_parent = false;
3237 /* We don't support snapshots with extent tree v2 yet. */
3238 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3240 "extent tree v2 doesn't support snapshot deletion yet");
3245 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
3246 if (IS_ERR(vol_args2))
3247 return PTR_ERR(vol_args2);
3249 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
3255 * If SPEC_BY_ID is not set, we are looking for the subvolume by
3256 * name, same as v1 currently does.
3258 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
3259 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
3260 subvol_name = vol_args2->name;
3262 err = mnt_want_write_file(file);
3266 struct inode *old_dir;
3268 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
3273 err = mnt_want_write_file(file);
3277 dentry = btrfs_get_dentry(fs_info->sb,
3278 BTRFS_FIRST_FREE_OBJECTID,
3279 vol_args2->subvolid, 0, 0);
3280 if (IS_ERR(dentry)) {
3281 err = PTR_ERR(dentry);
3282 goto out_drop_write;
3286 * Change the default parent since the subvolume being
3287 * deleted can be outside of the current mount point.
3289 parent = btrfs_get_parent(dentry);
3292 * At this point dentry->d_name can point to '/' if the
3293 * subvolume we want to destroy is outsite of the
3294 * current mount point, so we need to release the
3295 * current dentry and execute the lookup to return a new
3296 * one with ->d_name pointing to the
3297 * <mount point>/subvol_name.
3300 if (IS_ERR(parent)) {
3301 err = PTR_ERR(parent);
3302 goto out_drop_write;
3305 dir = d_inode(parent);
3308 * If v2 was used with SPEC_BY_ID, a new parent was
3309 * allocated since the subvolume can be outside of the
3310 * current mount point. Later on we need to release this
3311 * new parent dentry.
3313 destroy_parent = true;
3316 * On idmapped mounts, deletion via subvolid is
3317 * restricted to subvolumes that are immediate
3318 * ancestors of the inode referenced by the file
3319 * descriptor in the ioctl. Otherwise the idmapping
3320 * could potentially be abused to delete subvolumes
3321 * anywhere in the filesystem the user wouldn't be able
3322 * to delete without an idmapped mount.
3324 if (old_dir != dir && mnt_userns != &init_user_ns) {
3329 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
3330 fs_info, vol_args2->subvolid);
3331 if (IS_ERR(subvol_name_ptr)) {
3332 err = PTR_ERR(subvol_name_ptr);
3335 /* subvol_name_ptr is already nul terminated */
3336 subvol_name = (char *)kbasename(subvol_name_ptr);
3339 vol_args = memdup_user(arg, sizeof(*vol_args));
3340 if (IS_ERR(vol_args))
3341 return PTR_ERR(vol_args);
3343 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
3344 subvol_name = vol_args->name;
3346 err = mnt_want_write_file(file);
3351 subvol_namelen = strlen(subvol_name);
3353 if (strchr(subvol_name, '/') ||
3354 strncmp(subvol_name, "..", subvol_namelen) == 0) {
3356 goto free_subvol_name;
3359 if (!S_ISDIR(dir->i_mode)) {
3361 goto free_subvol_name;
3364 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
3366 goto free_subvol_name;
3367 dentry = lookup_one(mnt_userns, subvol_name, parent, subvol_namelen);
3368 if (IS_ERR(dentry)) {
3369 err = PTR_ERR(dentry);
3370 goto out_unlock_dir;
3373 if (d_really_is_negative(dentry)) {
3378 inode = d_inode(dentry);
3379 dest = BTRFS_I(inode)->root;
3380 if (!capable(CAP_SYS_ADMIN)) {
3382 * Regular user. Only allow this with a special mount
3383 * option, when the user has write+exec access to the
3384 * subvol root, and when rmdir(2) would have been
3387 * Note that this is _not_ check that the subvol is
3388 * empty or doesn't contain data that we wouldn't
3389 * otherwise be able to delete.
3391 * Users who want to delete empty subvols should try
3395 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
3399 * Do not allow deletion if the parent dir is the same
3400 * as the dir to be deleted. That means the ioctl
3401 * must be called on the dentry referencing the root
3402 * of the subvol, not a random directory contained
3409 err = inode_permission(mnt_userns, inode, MAY_WRITE | MAY_EXEC);
3414 /* check if subvolume may be deleted by a user */
3415 err = btrfs_may_delete(mnt_userns, dir, dentry, 1);
3419 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3424 btrfs_inode_lock(inode, 0);
3425 err = btrfs_delete_subvolume(dir, dentry);
3426 btrfs_inode_unlock(inode, 0);
3428 d_delete_notify(dir, dentry);
3433 btrfs_inode_unlock(dir, 0);
3435 kfree(subvol_name_ptr);
3440 mnt_drop_write_file(file);
3447 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
3449 struct inode *inode = file_inode(file);
3450 struct btrfs_root *root = BTRFS_I(inode)->root;
3451 struct btrfs_ioctl_defrag_range_args range = {0};
3454 ret = mnt_want_write_file(file);
3458 if (btrfs_root_readonly(root)) {
3463 switch (inode->i_mode & S_IFMT) {
3465 if (!capable(CAP_SYS_ADMIN)) {
3469 ret = btrfs_defrag_root(root);
3473 * Note that this does not check the file descriptor for write
3474 * access. This prevents defragmenting executables that are
3475 * running and allows defrag on files open in read-only mode.
3477 if (!capable(CAP_SYS_ADMIN) &&
3478 inode_permission(&init_user_ns, inode, MAY_WRITE)) {
3484 if (copy_from_user(&range, argp, sizeof(range))) {
3488 /* compression requires us to start the IO */
3489 if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3490 range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
3491 range.extent_thresh = (u32)-1;
3494 /* the rest are all set to zero by kzalloc */
3495 range.len = (u64)-1;
3497 ret = btrfs_defrag_file(file_inode(file), &file->f_ra,
3498 &range, BTRFS_OLDEST_GENERATION, 0);
3506 mnt_drop_write_file(file);
3510 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
3512 struct btrfs_ioctl_vol_args *vol_args;
3513 bool restore_op = false;
3516 if (!capable(CAP_SYS_ADMIN))
3519 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3520 btrfs_err(fs_info, "device add not supported on extent tree v2 yet");
3524 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) {
3525 if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD))
3526 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3529 * We can do the device add because we have a paused balanced,
3530 * change the exclusive op type and remember we should bring
3531 * back the paused balance
3533 fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD;
3534 btrfs_exclop_start_unlock(fs_info);
3538 vol_args = memdup_user(arg, sizeof(*vol_args));
3539 if (IS_ERR(vol_args)) {
3540 ret = PTR_ERR(vol_args);
3544 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3545 ret = btrfs_init_new_device(fs_info, vol_args->name);
3548 btrfs_info(fs_info, "disk added %s", vol_args->name);
3553 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
3555 btrfs_exclop_finish(fs_info);
3559 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
3561 BTRFS_DEV_LOOKUP_ARGS(args);
3562 struct inode *inode = file_inode(file);
3563 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3564 struct btrfs_ioctl_vol_args_v2 *vol_args;
3565 struct block_device *bdev = NULL;
3568 bool cancel = false;
3570 if (!capable(CAP_SYS_ADMIN))
3573 vol_args = memdup_user(arg, sizeof(*vol_args));
3574 if (IS_ERR(vol_args))
3575 return PTR_ERR(vol_args);
3577 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
3582 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3583 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
3584 args.devid = vol_args->devid;
3585 } else if (!strcmp("cancel", vol_args->name)) {
3588 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
3593 ret = mnt_want_write_file(file);
3597 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3602 /* Exclusive operation is now claimed */
3603 ret = btrfs_rm_device(fs_info, &args, &bdev, &mode);
3605 btrfs_exclop_finish(fs_info);
3608 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3609 btrfs_info(fs_info, "device deleted: id %llu",
3612 btrfs_info(fs_info, "device deleted: %s",
3616 mnt_drop_write_file(file);
3618 blkdev_put(bdev, mode);
3620 btrfs_put_dev_args_from_path(&args);
3625 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3627 BTRFS_DEV_LOOKUP_ARGS(args);
3628 struct inode *inode = file_inode(file);
3629 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3630 struct btrfs_ioctl_vol_args *vol_args;
3631 struct block_device *bdev = NULL;
3634 bool cancel = false;
3636 if (!capable(CAP_SYS_ADMIN))
3639 vol_args = memdup_user(arg, sizeof(*vol_args));
3640 if (IS_ERR(vol_args))
3641 return PTR_ERR(vol_args);
3643 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3644 if (!strcmp("cancel", vol_args->name)) {
3647 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
3652 ret = mnt_want_write_file(file);
3656 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3659 ret = btrfs_rm_device(fs_info, &args, &bdev, &mode);
3661 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3662 btrfs_exclop_finish(fs_info);
3665 mnt_drop_write_file(file);
3667 blkdev_put(bdev, mode);
3669 btrfs_put_dev_args_from_path(&args);
3674 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3677 struct btrfs_ioctl_fs_info_args *fi_args;
3678 struct btrfs_device *device;
3679 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3683 fi_args = memdup_user(arg, sizeof(*fi_args));
3684 if (IS_ERR(fi_args))
3685 return PTR_ERR(fi_args);
3687 flags_in = fi_args->flags;
3688 memset(fi_args, 0, sizeof(*fi_args));
3691 fi_args->num_devices = fs_devices->num_devices;
3693 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3694 if (device->devid > fi_args->max_id)
3695 fi_args->max_id = device->devid;
3699 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
3700 fi_args->nodesize = fs_info->nodesize;
3701 fi_args->sectorsize = fs_info->sectorsize;
3702 fi_args->clone_alignment = fs_info->sectorsize;
3704 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
3705 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
3706 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
3707 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
3710 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
3711 fi_args->generation = fs_info->generation;
3712 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
3715 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
3716 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
3717 sizeof(fi_args->metadata_uuid));
3718 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
3721 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3728 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3731 BTRFS_DEV_LOOKUP_ARGS(args);
3732 struct btrfs_ioctl_dev_info_args *di_args;
3733 struct btrfs_device *dev;
3736 di_args = memdup_user(arg, sizeof(*di_args));
3737 if (IS_ERR(di_args))
3738 return PTR_ERR(di_args);
3740 args.devid = di_args->devid;
3741 if (!btrfs_is_empty_uuid(di_args->uuid))
3742 args.uuid = di_args->uuid;
3745 dev = btrfs_find_device(fs_info->fs_devices, &args);
3751 di_args->devid = dev->devid;
3752 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3753 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3754 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3756 strscpy(di_args->path, rcu_str_deref(dev->name), sizeof(di_args->path));
3758 di_args->path[0] = '\0';
3762 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3769 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3771 struct inode *inode = file_inode(file);
3772 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3773 struct btrfs_root *root = BTRFS_I(inode)->root;
3774 struct btrfs_root *new_root;
3775 struct btrfs_dir_item *di;
3776 struct btrfs_trans_handle *trans;
3777 struct btrfs_path *path = NULL;
3778 struct btrfs_disk_key disk_key;
3783 if (!capable(CAP_SYS_ADMIN))
3786 ret = mnt_want_write_file(file);
3790 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3796 objectid = BTRFS_FS_TREE_OBJECTID;
3798 new_root = btrfs_get_fs_root(fs_info, objectid, true);
3799 if (IS_ERR(new_root)) {
3800 ret = PTR_ERR(new_root);
3803 if (!is_fstree(new_root->root_key.objectid)) {
3808 path = btrfs_alloc_path();
3814 trans = btrfs_start_transaction(root, 1);
3815 if (IS_ERR(trans)) {
3816 ret = PTR_ERR(trans);
3820 dir_id = btrfs_super_root_dir(fs_info->super_copy);
3821 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3822 dir_id, "default", 7, 1);
3823 if (IS_ERR_OR_NULL(di)) {
3824 btrfs_release_path(path);
3825 btrfs_end_transaction(trans);
3827 "Umm, you don't have the default diritem, this isn't going to work");
3832 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3833 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3834 btrfs_mark_buffer_dirty(path->nodes[0]);
3835 btrfs_release_path(path);
3837 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3838 btrfs_end_transaction(trans);
3840 btrfs_put_root(new_root);
3841 btrfs_free_path(path);
3843 mnt_drop_write_file(file);
3847 static void get_block_group_info(struct list_head *groups_list,
3848 struct btrfs_ioctl_space_info *space)
3850 struct btrfs_block_group *block_group;
3852 space->total_bytes = 0;
3853 space->used_bytes = 0;
3855 list_for_each_entry(block_group, groups_list, list) {
3856 space->flags = block_group->flags;
3857 space->total_bytes += block_group->length;
3858 space->used_bytes += block_group->used;
3862 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3865 struct btrfs_ioctl_space_args space_args;
3866 struct btrfs_ioctl_space_info space;
3867 struct btrfs_ioctl_space_info *dest;
3868 struct btrfs_ioctl_space_info *dest_orig;
3869 struct btrfs_ioctl_space_info __user *user_dest;
3870 struct btrfs_space_info *info;
3871 static const u64 types[] = {
3872 BTRFS_BLOCK_GROUP_DATA,
3873 BTRFS_BLOCK_GROUP_SYSTEM,
3874 BTRFS_BLOCK_GROUP_METADATA,
3875 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3883 if (copy_from_user(&space_args,
3884 (struct btrfs_ioctl_space_args __user *)arg,
3885 sizeof(space_args)))
3888 for (i = 0; i < num_types; i++) {
3889 struct btrfs_space_info *tmp;
3892 list_for_each_entry(tmp, &fs_info->space_info, list) {
3893 if (tmp->flags == types[i]) {
3902 down_read(&info->groups_sem);
3903 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3904 if (!list_empty(&info->block_groups[c]))
3907 up_read(&info->groups_sem);
3911 * Global block reserve, exported as a space_info
3915 /* space_slots == 0 means they are asking for a count */
3916 if (space_args.space_slots == 0) {
3917 space_args.total_spaces = slot_count;
3921 slot_count = min_t(u64, space_args.space_slots, slot_count);
3923 alloc_size = sizeof(*dest) * slot_count;
3925 /* we generally have at most 6 or so space infos, one for each raid
3926 * level. So, a whole page should be more than enough for everyone
3928 if (alloc_size > PAGE_SIZE)
3931 space_args.total_spaces = 0;
3932 dest = kmalloc(alloc_size, GFP_KERNEL);
3937 /* now we have a buffer to copy into */
3938 for (i = 0; i < num_types; i++) {
3939 struct btrfs_space_info *tmp;
3945 list_for_each_entry(tmp, &fs_info->space_info, list) {
3946 if (tmp->flags == types[i]) {
3954 down_read(&info->groups_sem);
3955 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3956 if (!list_empty(&info->block_groups[c])) {
3957 get_block_group_info(&info->block_groups[c],
3959 memcpy(dest, &space, sizeof(space));
3961 space_args.total_spaces++;
3967 up_read(&info->groups_sem);
3971 * Add global block reserve
3974 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3976 spin_lock(&block_rsv->lock);
3977 space.total_bytes = block_rsv->size;
3978 space.used_bytes = block_rsv->size - block_rsv->reserved;
3979 spin_unlock(&block_rsv->lock);
3980 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3981 memcpy(dest, &space, sizeof(space));
3982 space_args.total_spaces++;
3985 user_dest = (struct btrfs_ioctl_space_info __user *)
3986 (arg + sizeof(struct btrfs_ioctl_space_args));
3988 if (copy_to_user(user_dest, dest_orig, alloc_size))
3993 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3999 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4002 struct btrfs_trans_handle *trans;
4005 trans = btrfs_attach_transaction_barrier(root);
4006 if (IS_ERR(trans)) {
4007 if (PTR_ERR(trans) != -ENOENT)
4008 return PTR_ERR(trans);
4010 /* No running transaction, don't bother */
4011 transid = root->fs_info->last_trans_committed;
4014 transid = trans->transid;
4015 btrfs_commit_transaction_async(trans);
4018 if (copy_to_user(argp, &transid, sizeof(transid)))
4023 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
4029 if (copy_from_user(&transid, argp, sizeof(transid)))
4032 transid = 0; /* current trans */
4034 return btrfs_wait_for_commit(fs_info, transid);
4037 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4039 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4040 struct btrfs_ioctl_scrub_args *sa;
4043 if (!capable(CAP_SYS_ADMIN))
4046 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
4047 btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet");
4051 sa = memdup_user(arg, sizeof(*sa));
4055 if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) {
4060 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4061 ret = mnt_want_write_file(file);
4066 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4067 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4071 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
4072 * error. This is important as it allows user space to know how much
4073 * progress scrub has done. For example, if scrub is canceled we get
4074 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
4075 * space. Later user space can inspect the progress from the structure
4076 * btrfs_ioctl_scrub_args and resume scrub from where it left off
4077 * previously (btrfs-progs does this).
4078 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
4079 * then return -EFAULT to signal the structure was not copied or it may
4080 * be corrupt and unreliable due to a partial copy.
4082 if (copy_to_user(arg, sa, sizeof(*sa)))
4085 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4086 mnt_drop_write_file(file);
4092 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
4094 if (!capable(CAP_SYS_ADMIN))
4097 return btrfs_scrub_cancel(fs_info);
4100 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
4103 struct btrfs_ioctl_scrub_args *sa;
4106 if (!capable(CAP_SYS_ADMIN))
4109 sa = memdup_user(arg, sizeof(*sa));
4113 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
4115 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
4122 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
4125 struct btrfs_ioctl_get_dev_stats *sa;
4128 sa = memdup_user(arg, sizeof(*sa));
4132 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4137 ret = btrfs_get_dev_stats(fs_info, sa);
4139 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
4146 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4149 struct btrfs_ioctl_dev_replace_args *p;
4152 if (!capable(CAP_SYS_ADMIN))
4155 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
4156 btrfs_err(fs_info, "device replace not supported on extent tree v2 yet");
4160 p = memdup_user(arg, sizeof(*p));
4165 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4166 if (sb_rdonly(fs_info->sb)) {
4170 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
4171 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4173 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
4174 btrfs_exclop_finish(fs_info);
4177 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4178 btrfs_dev_replace_status(fs_info, p);
4181 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4182 p->result = btrfs_dev_replace_cancel(fs_info);
4190 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
4197 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4203 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4204 struct inode_fs_paths *ipath = NULL;
4205 struct btrfs_path *path;
4207 if (!capable(CAP_DAC_READ_SEARCH))
4210 path = btrfs_alloc_path();
4216 ipa = memdup_user(arg, sizeof(*ipa));
4223 size = min_t(u32, ipa->size, 4096);
4224 ipath = init_ipath(size, root, path);
4225 if (IS_ERR(ipath)) {
4226 ret = PTR_ERR(ipath);
4231 ret = paths_from_inode(ipa->inum, ipath);
4235 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4236 rel_ptr = ipath->fspath->val[i] -
4237 (u64)(unsigned long)ipath->fspath->val;
4238 ipath->fspath->val[i] = rel_ptr;
4241 btrfs_free_path(path);
4243 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
4244 ipath->fspath, size);
4251 btrfs_free_path(path);
4258 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
4259 void __user *arg, int version)
4263 struct btrfs_ioctl_logical_ino_args *loi;
4264 struct btrfs_data_container *inodes = NULL;
4265 struct btrfs_path *path = NULL;
4268 if (!capable(CAP_SYS_ADMIN))
4271 loi = memdup_user(arg, sizeof(*loi));
4273 return PTR_ERR(loi);
4276 ignore_offset = false;
4277 size = min_t(u32, loi->size, SZ_64K);
4279 /* All reserved bits must be 0 for now */
4280 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
4284 /* Only accept flags we have defined so far */
4285 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
4289 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
4290 size = min_t(u32, loi->size, SZ_16M);
4293 inodes = init_data_container(size);
4294 if (IS_ERR(inodes)) {
4295 ret = PTR_ERR(inodes);
4299 path = btrfs_alloc_path();
4304 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
4305 inodes, ignore_offset);
4306 btrfs_free_path(path);
4312 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
4325 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
4326 struct btrfs_ioctl_balance_args *bargs)
4328 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4330 bargs->flags = bctl->flags;
4332 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
4333 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4334 if (atomic_read(&fs_info->balance_pause_req))
4335 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4336 if (atomic_read(&fs_info->balance_cancel_req))
4337 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4339 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4340 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4341 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4343 spin_lock(&fs_info->balance_lock);
4344 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4345 spin_unlock(&fs_info->balance_lock);
4349 * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as
4352 * @fs_info: the filesystem
4353 * @excl_acquired: ptr to boolean value which is set to false in case balance
4356 * Return 0 on success in which case both fs_info::balance is acquired as well
4357 * as exclusive ops are blocked. In case of failure return an error code.
4359 static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired)
4364 * Exclusive operation is locked. Three possibilities:
4365 * (1) some other op is running
4366 * (2) balance is running
4367 * (3) balance is paused -- special case (think resume)
4370 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
4371 *excl_acquired = true;
4372 mutex_lock(&fs_info->balance_mutex);
4376 mutex_lock(&fs_info->balance_mutex);
4377 if (fs_info->balance_ctl) {
4378 /* This is either (2) or (3) */
4379 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4385 mutex_unlock(&fs_info->balance_mutex);
4387 * Lock released to allow other waiters to
4388 * continue, we'll reexamine the status again.
4390 mutex_lock(&fs_info->balance_mutex);
4392 if (fs_info->balance_ctl &&
4393 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4395 *excl_acquired = false;
4401 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4405 mutex_unlock(&fs_info->balance_mutex);
4409 mutex_unlock(&fs_info->balance_mutex);
4410 *excl_acquired = false;
4414 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4416 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4417 struct btrfs_fs_info *fs_info = root->fs_info;
4418 struct btrfs_ioctl_balance_args *bargs;
4419 struct btrfs_balance_control *bctl;
4420 bool need_unlock = true;
4423 if (!capable(CAP_SYS_ADMIN))
4426 ret = mnt_want_write_file(file);
4430 bargs = memdup_user(arg, sizeof(*bargs));
4431 if (IS_ERR(bargs)) {
4432 ret = PTR_ERR(bargs);
4437 ret = btrfs_try_lock_balance(fs_info, &need_unlock);
4441 lockdep_assert_held(&fs_info->balance_mutex);
4443 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4444 if (!fs_info->balance_ctl) {
4449 bctl = fs_info->balance_ctl;
4450 spin_lock(&fs_info->balance_lock);
4451 bctl->flags |= BTRFS_BALANCE_RESUME;
4452 spin_unlock(&fs_info->balance_lock);
4453 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE);
4458 if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4463 if (fs_info->balance_ctl) {
4468 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4474 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4475 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4476 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4478 bctl->flags = bargs->flags;
4481 * Ownership of bctl and exclusive operation goes to btrfs_balance.
4482 * bctl is freed in reset_balance_state, or, if restriper was paused
4483 * all the way until unmount, in free_fs_info. The flag should be
4484 * cleared after reset_balance_state.
4486 need_unlock = false;
4488 ret = btrfs_balance(fs_info, bctl, bargs);
4491 if (ret == 0 || ret == -ECANCELED) {
4492 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4498 mutex_unlock(&fs_info->balance_mutex);
4500 btrfs_exclop_finish(fs_info);
4502 mnt_drop_write_file(file);
4507 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4509 if (!capable(CAP_SYS_ADMIN))
4513 case BTRFS_BALANCE_CTL_PAUSE:
4514 return btrfs_pause_balance(fs_info);
4515 case BTRFS_BALANCE_CTL_CANCEL:
4516 return btrfs_cancel_balance(fs_info);
4522 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4525 struct btrfs_ioctl_balance_args *bargs;
4528 if (!capable(CAP_SYS_ADMIN))
4531 mutex_lock(&fs_info->balance_mutex);
4532 if (!fs_info->balance_ctl) {
4537 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4543 btrfs_update_ioctl_balance_args(fs_info, bargs);
4545 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4550 mutex_unlock(&fs_info->balance_mutex);
4554 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4556 struct inode *inode = file_inode(file);
4557 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4558 struct btrfs_ioctl_quota_ctl_args *sa;
4561 if (!capable(CAP_SYS_ADMIN))
4564 ret = mnt_want_write_file(file);
4568 sa = memdup_user(arg, sizeof(*sa));
4574 down_write(&fs_info->subvol_sem);
4577 case BTRFS_QUOTA_CTL_ENABLE:
4578 ret = btrfs_quota_enable(fs_info);
4580 case BTRFS_QUOTA_CTL_DISABLE:
4581 ret = btrfs_quota_disable(fs_info);
4589 up_write(&fs_info->subvol_sem);
4591 mnt_drop_write_file(file);
4595 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4597 struct inode *inode = file_inode(file);
4598 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4599 struct btrfs_root *root = BTRFS_I(inode)->root;
4600 struct btrfs_ioctl_qgroup_assign_args *sa;
4601 struct btrfs_trans_handle *trans;
4605 if (!capable(CAP_SYS_ADMIN))
4608 ret = mnt_want_write_file(file);
4612 sa = memdup_user(arg, sizeof(*sa));
4618 trans = btrfs_join_transaction(root);
4619 if (IS_ERR(trans)) {
4620 ret = PTR_ERR(trans);
4625 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
4627 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
4630 /* update qgroup status and info */
4631 mutex_lock(&fs_info->qgroup_ioctl_lock);
4632 err = btrfs_run_qgroups(trans);
4633 mutex_unlock(&fs_info->qgroup_ioctl_lock);
4635 btrfs_handle_fs_error(fs_info, err,
4636 "failed to update qgroup status and info");
4637 err = btrfs_end_transaction(trans);
4644 mnt_drop_write_file(file);
4648 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4650 struct inode *inode = file_inode(file);
4651 struct btrfs_root *root = BTRFS_I(inode)->root;
4652 struct btrfs_ioctl_qgroup_create_args *sa;
4653 struct btrfs_trans_handle *trans;
4657 if (!capable(CAP_SYS_ADMIN))
4660 ret = mnt_want_write_file(file);
4664 sa = memdup_user(arg, sizeof(*sa));
4670 if (!sa->qgroupid) {
4675 trans = btrfs_join_transaction(root);
4676 if (IS_ERR(trans)) {
4677 ret = PTR_ERR(trans);
4682 ret = btrfs_create_qgroup(trans, sa->qgroupid);
4684 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
4687 err = btrfs_end_transaction(trans);
4694 mnt_drop_write_file(file);
4698 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4700 struct inode *inode = file_inode(file);
4701 struct btrfs_root *root = BTRFS_I(inode)->root;
4702 struct btrfs_ioctl_qgroup_limit_args *sa;
4703 struct btrfs_trans_handle *trans;
4708 if (!capable(CAP_SYS_ADMIN))
4711 ret = mnt_want_write_file(file);
4715 sa = memdup_user(arg, sizeof(*sa));
4721 trans = btrfs_join_transaction(root);
4722 if (IS_ERR(trans)) {
4723 ret = PTR_ERR(trans);
4727 qgroupid = sa->qgroupid;
4729 /* take the current subvol as qgroup */
4730 qgroupid = root->root_key.objectid;
4733 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
4735 err = btrfs_end_transaction(trans);
4742 mnt_drop_write_file(file);
4746 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4748 struct inode *inode = file_inode(file);
4749 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4750 struct btrfs_ioctl_quota_rescan_args *qsa;
4753 if (!capable(CAP_SYS_ADMIN))
4756 ret = mnt_want_write_file(file);
4760 qsa = memdup_user(arg, sizeof(*qsa));
4771 ret = btrfs_qgroup_rescan(fs_info);
4776 mnt_drop_write_file(file);
4780 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4783 struct btrfs_ioctl_quota_rescan_args qsa = {0};
4785 if (!capable(CAP_SYS_ADMIN))
4788 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4790 qsa.progress = fs_info->qgroup_rescan_progress.objectid;
4793 if (copy_to_user(arg, &qsa, sizeof(qsa)))
4799 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4802 if (!capable(CAP_SYS_ADMIN))
4805 return btrfs_qgroup_wait_for_completion(fs_info, true);
4808 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4809 struct user_namespace *mnt_userns,
4810 struct btrfs_ioctl_received_subvol_args *sa)
4812 struct inode *inode = file_inode(file);
4813 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4814 struct btrfs_root *root = BTRFS_I(inode)->root;
4815 struct btrfs_root_item *root_item = &root->root_item;
4816 struct btrfs_trans_handle *trans;
4817 struct timespec64 ct = current_time(inode);
4819 int received_uuid_changed;
4821 if (!inode_owner_or_capable(mnt_userns, inode))
4824 ret = mnt_want_write_file(file);
4828 down_write(&fs_info->subvol_sem);
4830 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4835 if (btrfs_root_readonly(root)) {
4842 * 2 - uuid items (received uuid + subvol uuid)
4844 trans = btrfs_start_transaction(root, 3);
4845 if (IS_ERR(trans)) {
4846 ret = PTR_ERR(trans);
4851 sa->rtransid = trans->transid;
4852 sa->rtime.sec = ct.tv_sec;
4853 sa->rtime.nsec = ct.tv_nsec;
4855 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4857 if (received_uuid_changed &&
4858 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4859 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4860 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4861 root->root_key.objectid);
4862 if (ret && ret != -ENOENT) {
4863 btrfs_abort_transaction(trans, ret);
4864 btrfs_end_transaction(trans);
4868 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4869 btrfs_set_root_stransid(root_item, sa->stransid);
4870 btrfs_set_root_rtransid(root_item, sa->rtransid);
4871 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4872 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4873 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4874 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4876 ret = btrfs_update_root(trans, fs_info->tree_root,
4877 &root->root_key, &root->root_item);
4879 btrfs_end_transaction(trans);
4882 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4883 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4884 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4885 root->root_key.objectid);
4886 if (ret < 0 && ret != -EEXIST) {
4887 btrfs_abort_transaction(trans, ret);
4888 btrfs_end_transaction(trans);
4892 ret = btrfs_commit_transaction(trans);
4894 up_write(&fs_info->subvol_sem);
4895 mnt_drop_write_file(file);
4900 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4903 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4904 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4907 args32 = memdup_user(arg, sizeof(*args32));
4909 return PTR_ERR(args32);
4911 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4917 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4918 args64->stransid = args32->stransid;
4919 args64->rtransid = args32->rtransid;
4920 args64->stime.sec = args32->stime.sec;
4921 args64->stime.nsec = args32->stime.nsec;
4922 args64->rtime.sec = args32->rtime.sec;
4923 args64->rtime.nsec = args32->rtime.nsec;
4924 args64->flags = args32->flags;
4926 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), args64);
4930 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4931 args32->stransid = args64->stransid;
4932 args32->rtransid = args64->rtransid;
4933 args32->stime.sec = args64->stime.sec;
4934 args32->stime.nsec = args64->stime.nsec;
4935 args32->rtime.sec = args64->rtime.sec;
4936 args32->rtime.nsec = args64->rtime.nsec;
4937 args32->flags = args64->flags;
4939 ret = copy_to_user(arg, args32, sizeof(*args32));
4950 static long btrfs_ioctl_set_received_subvol(struct file *file,
4953 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4956 sa = memdup_user(arg, sizeof(*sa));
4960 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), sa);
4965 ret = copy_to_user(arg, sa, sizeof(*sa));
4974 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4979 char label[BTRFS_LABEL_SIZE];
4981 spin_lock(&fs_info->super_lock);
4982 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4983 spin_unlock(&fs_info->super_lock);
4985 len = strnlen(label, BTRFS_LABEL_SIZE);
4987 if (len == BTRFS_LABEL_SIZE) {
4989 "label is too long, return the first %zu bytes",
4993 ret = copy_to_user(arg, label, len);
4995 return ret ? -EFAULT : 0;
4998 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5000 struct inode *inode = file_inode(file);
5001 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5002 struct btrfs_root *root = BTRFS_I(inode)->root;
5003 struct btrfs_super_block *super_block = fs_info->super_copy;
5004 struct btrfs_trans_handle *trans;
5005 char label[BTRFS_LABEL_SIZE];
5008 if (!capable(CAP_SYS_ADMIN))
5011 if (copy_from_user(label, arg, sizeof(label)))
5014 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5016 "unable to set label with more than %d bytes",
5017 BTRFS_LABEL_SIZE - 1);
5021 ret = mnt_want_write_file(file);
5025 trans = btrfs_start_transaction(root, 0);
5026 if (IS_ERR(trans)) {
5027 ret = PTR_ERR(trans);
5031 spin_lock(&fs_info->super_lock);
5032 strcpy(super_block->label, label);
5033 spin_unlock(&fs_info->super_lock);
5034 ret = btrfs_commit_transaction(trans);
5037 mnt_drop_write_file(file);
5041 #define INIT_FEATURE_FLAGS(suffix) \
5042 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5043 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5044 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5046 int btrfs_ioctl_get_supported_features(void __user *arg)
5048 static const struct btrfs_ioctl_feature_flags features[3] = {
5049 INIT_FEATURE_FLAGS(SUPP),
5050 INIT_FEATURE_FLAGS(SAFE_SET),
5051 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5054 if (copy_to_user(arg, &features, sizeof(features)))
5060 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
5063 struct btrfs_super_block *super_block = fs_info->super_copy;
5064 struct btrfs_ioctl_feature_flags features;
5066 features.compat_flags = btrfs_super_compat_flags(super_block);
5067 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5068 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5070 if (copy_to_user(arg, &features, sizeof(features)))
5076 static int check_feature_bits(struct btrfs_fs_info *fs_info,
5077 enum btrfs_feature_set set,
5078 u64 change_mask, u64 flags, u64 supported_flags,
5079 u64 safe_set, u64 safe_clear)
5081 const char *type = btrfs_feature_set_name(set);
5083 u64 disallowed, unsupported;
5084 u64 set_mask = flags & change_mask;
5085 u64 clear_mask = ~flags & change_mask;
5087 unsupported = set_mask & ~supported_flags;
5089 names = btrfs_printable_features(set, unsupported);
5092 "this kernel does not support the %s feature bit%s",
5093 names, strchr(names, ',') ? "s" : "");
5097 "this kernel does not support %s bits 0x%llx",
5102 disallowed = set_mask & ~safe_set;
5104 names = btrfs_printable_features(set, disallowed);
5107 "can't set the %s feature bit%s while mounted",
5108 names, strchr(names, ',') ? "s" : "");
5112 "can't set %s bits 0x%llx while mounted",
5117 disallowed = clear_mask & ~safe_clear;
5119 names = btrfs_printable_features(set, disallowed);
5122 "can't clear the %s feature bit%s while mounted",
5123 names, strchr(names, ',') ? "s" : "");
5127 "can't clear %s bits 0x%llx while mounted",
5135 #define check_feature(fs_info, change_mask, flags, mask_base) \
5136 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5137 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5138 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5139 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5141 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5143 struct inode *inode = file_inode(file);
5144 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5145 struct btrfs_root *root = BTRFS_I(inode)->root;
5146 struct btrfs_super_block *super_block = fs_info->super_copy;
5147 struct btrfs_ioctl_feature_flags flags[2];
5148 struct btrfs_trans_handle *trans;
5152 if (!capable(CAP_SYS_ADMIN))
5155 if (copy_from_user(flags, arg, sizeof(flags)))
5159 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5160 !flags[0].incompat_flags)
5163 ret = check_feature(fs_info, flags[0].compat_flags,
5164 flags[1].compat_flags, COMPAT);
5168 ret = check_feature(fs_info, flags[0].compat_ro_flags,
5169 flags[1].compat_ro_flags, COMPAT_RO);
5173 ret = check_feature(fs_info, flags[0].incompat_flags,
5174 flags[1].incompat_flags, INCOMPAT);
5178 ret = mnt_want_write_file(file);
5182 trans = btrfs_start_transaction(root, 0);
5183 if (IS_ERR(trans)) {
5184 ret = PTR_ERR(trans);
5185 goto out_drop_write;
5188 spin_lock(&fs_info->super_lock);
5189 newflags = btrfs_super_compat_flags(super_block);
5190 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5191 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5192 btrfs_set_super_compat_flags(super_block, newflags);
5194 newflags = btrfs_super_compat_ro_flags(super_block);
5195 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5196 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5197 btrfs_set_super_compat_ro_flags(super_block, newflags);
5199 newflags = btrfs_super_incompat_flags(super_block);
5200 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5201 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5202 btrfs_set_super_incompat_flags(super_block, newflags);
5203 spin_unlock(&fs_info->super_lock);
5205 ret = btrfs_commit_transaction(trans);
5207 mnt_drop_write_file(file);
5212 static int _btrfs_ioctl_send(struct inode *inode, void __user *argp, bool compat)
5214 struct btrfs_ioctl_send_args *arg;
5218 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5219 struct btrfs_ioctl_send_args_32 args32;
5221 ret = copy_from_user(&args32, argp, sizeof(args32));
5224 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
5227 arg->send_fd = args32.send_fd;
5228 arg->clone_sources_count = args32.clone_sources_count;
5229 arg->clone_sources = compat_ptr(args32.clone_sources);
5230 arg->parent_root = args32.parent_root;
5231 arg->flags = args32.flags;
5232 memcpy(arg->reserved, args32.reserved,
5233 sizeof(args32.reserved));
5238 arg = memdup_user(argp, sizeof(*arg));
5240 return PTR_ERR(arg);
5242 ret = btrfs_ioctl_send(inode, arg);
5247 static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,
5250 struct btrfs_ioctl_encoded_io_args args = { 0 };
5251 size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args,
5254 struct iovec iovstack[UIO_FASTIOV];
5255 struct iovec *iov = iovstack;
5256 struct iov_iter iter;
5261 if (!capable(CAP_SYS_ADMIN)) {
5267 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5268 struct btrfs_ioctl_encoded_io_args_32 args32;
5270 copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32,
5272 if (copy_from_user(&args32, argp, copy_end)) {
5276 args.iov = compat_ptr(args32.iov);
5277 args.iovcnt = args32.iovcnt;
5278 args.offset = args32.offset;
5279 args.flags = args32.flags;
5284 copy_end = copy_end_kernel;
5285 if (copy_from_user(&args, argp, copy_end)) {
5290 if (args.flags != 0) {
5295 ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
5300 if (iov_iter_count(&iter) == 0) {
5305 ret = rw_verify_area(READ, file, &pos, args.len);
5309 init_sync_kiocb(&kiocb, file);
5312 ret = btrfs_encoded_read(&kiocb, &iter, &args);
5314 fsnotify_access(file);
5315 if (copy_to_user(argp + copy_end,
5316 (char *)&args + copy_end_kernel,
5317 sizeof(args) - copy_end_kernel))
5325 add_rchar(current, ret);
5330 static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat)
5332 struct btrfs_ioctl_encoded_io_args args;
5333 struct iovec iovstack[UIO_FASTIOV];
5334 struct iovec *iov = iovstack;
5335 struct iov_iter iter;
5340 if (!capable(CAP_SYS_ADMIN)) {
5345 if (!(file->f_mode & FMODE_WRITE)) {
5351 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5352 struct btrfs_ioctl_encoded_io_args_32 args32;
5354 if (copy_from_user(&args32, argp, sizeof(args32))) {
5358 args.iov = compat_ptr(args32.iov);
5359 args.iovcnt = args32.iovcnt;
5360 args.offset = args32.offset;
5361 args.flags = args32.flags;
5362 args.len = args32.len;
5363 args.unencoded_len = args32.unencoded_len;
5364 args.unencoded_offset = args32.unencoded_offset;
5365 args.compression = args32.compression;
5366 args.encryption = args32.encryption;
5367 memcpy(args.reserved, args32.reserved, sizeof(args.reserved));
5372 if (copy_from_user(&args, argp, sizeof(args))) {
5379 if (args.flags != 0)
5381 if (memchr_inv(args.reserved, 0, sizeof(args.reserved)))
5383 if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
5384 args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
5386 if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
5387 args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
5389 if (args.unencoded_offset > args.unencoded_len)
5391 if (args.len > args.unencoded_len - args.unencoded_offset)
5394 ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
5399 file_start_write(file);
5401 if (iov_iter_count(&iter) == 0) {
5406 ret = rw_verify_area(WRITE, file, &pos, args.len);
5410 init_sync_kiocb(&kiocb, file);
5411 ret = kiocb_set_rw_flags(&kiocb, 0);
5416 ret = btrfs_do_write_iter(&kiocb, &iter, &args);
5418 fsnotify_modify(file);
5421 file_end_write(file);
5425 add_wchar(current, ret);
5430 long btrfs_ioctl(struct file *file, unsigned int
5431 cmd, unsigned long arg)
5433 struct inode *inode = file_inode(file);
5434 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5435 struct btrfs_root *root = BTRFS_I(inode)->root;
5436 void __user *argp = (void __user *)arg;
5439 case FS_IOC_GETVERSION:
5440 return btrfs_ioctl_getversion(inode, argp);
5441 case FS_IOC_GETFSLABEL:
5442 return btrfs_ioctl_get_fslabel(fs_info, argp);
5443 case FS_IOC_SETFSLABEL:
5444 return btrfs_ioctl_set_fslabel(file, argp);
5446 return btrfs_ioctl_fitrim(fs_info, argp);
5447 case BTRFS_IOC_SNAP_CREATE:
5448 return btrfs_ioctl_snap_create(file, argp, 0);
5449 case BTRFS_IOC_SNAP_CREATE_V2:
5450 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5451 case BTRFS_IOC_SUBVOL_CREATE:
5452 return btrfs_ioctl_snap_create(file, argp, 1);
5453 case BTRFS_IOC_SUBVOL_CREATE_V2:
5454 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5455 case BTRFS_IOC_SNAP_DESTROY:
5456 return btrfs_ioctl_snap_destroy(file, argp, false);
5457 case BTRFS_IOC_SNAP_DESTROY_V2:
5458 return btrfs_ioctl_snap_destroy(file, argp, true);
5459 case BTRFS_IOC_SUBVOL_GETFLAGS:
5460 return btrfs_ioctl_subvol_getflags(inode, argp);
5461 case BTRFS_IOC_SUBVOL_SETFLAGS:
5462 return btrfs_ioctl_subvol_setflags(file, argp);
5463 case BTRFS_IOC_DEFAULT_SUBVOL:
5464 return btrfs_ioctl_default_subvol(file, argp);
5465 case BTRFS_IOC_DEFRAG:
5466 return btrfs_ioctl_defrag(file, NULL);
5467 case BTRFS_IOC_DEFRAG_RANGE:
5468 return btrfs_ioctl_defrag(file, argp);
5469 case BTRFS_IOC_RESIZE:
5470 return btrfs_ioctl_resize(file, argp);
5471 case BTRFS_IOC_ADD_DEV:
5472 return btrfs_ioctl_add_dev(fs_info, argp);
5473 case BTRFS_IOC_RM_DEV:
5474 return btrfs_ioctl_rm_dev(file, argp);
5475 case BTRFS_IOC_RM_DEV_V2:
5476 return btrfs_ioctl_rm_dev_v2(file, argp);
5477 case BTRFS_IOC_FS_INFO:
5478 return btrfs_ioctl_fs_info(fs_info, argp);
5479 case BTRFS_IOC_DEV_INFO:
5480 return btrfs_ioctl_dev_info(fs_info, argp);
5481 case BTRFS_IOC_TREE_SEARCH:
5482 return btrfs_ioctl_tree_search(inode, argp);
5483 case BTRFS_IOC_TREE_SEARCH_V2:
5484 return btrfs_ioctl_tree_search_v2(inode, argp);
5485 case BTRFS_IOC_INO_LOOKUP:
5486 return btrfs_ioctl_ino_lookup(root, argp);
5487 case BTRFS_IOC_INO_PATHS:
5488 return btrfs_ioctl_ino_to_path(root, argp);
5489 case BTRFS_IOC_LOGICAL_INO:
5490 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
5491 case BTRFS_IOC_LOGICAL_INO_V2:
5492 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
5493 case BTRFS_IOC_SPACE_INFO:
5494 return btrfs_ioctl_space_info(fs_info, argp);
5495 case BTRFS_IOC_SYNC: {
5498 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
5501 ret = btrfs_sync_fs(inode->i_sb, 1);
5503 * The transaction thread may want to do more work,
5504 * namely it pokes the cleaner kthread that will start
5505 * processing uncleaned subvols.
5507 wake_up_process(fs_info->transaction_kthread);
5510 case BTRFS_IOC_START_SYNC:
5511 return btrfs_ioctl_start_sync(root, argp);
5512 case BTRFS_IOC_WAIT_SYNC:
5513 return btrfs_ioctl_wait_sync(fs_info, argp);
5514 case BTRFS_IOC_SCRUB:
5515 return btrfs_ioctl_scrub(file, argp);
5516 case BTRFS_IOC_SCRUB_CANCEL:
5517 return btrfs_ioctl_scrub_cancel(fs_info);
5518 case BTRFS_IOC_SCRUB_PROGRESS:
5519 return btrfs_ioctl_scrub_progress(fs_info, argp);
5520 case BTRFS_IOC_BALANCE_V2:
5521 return btrfs_ioctl_balance(file, argp);
5522 case BTRFS_IOC_BALANCE_CTL:
5523 return btrfs_ioctl_balance_ctl(fs_info, arg);
5524 case BTRFS_IOC_BALANCE_PROGRESS:
5525 return btrfs_ioctl_balance_progress(fs_info, argp);
5526 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5527 return btrfs_ioctl_set_received_subvol(file, argp);
5529 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5530 return btrfs_ioctl_set_received_subvol_32(file, argp);
5532 case BTRFS_IOC_SEND:
5533 return _btrfs_ioctl_send(inode, argp, false);
5534 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5535 case BTRFS_IOC_SEND_32:
5536 return _btrfs_ioctl_send(inode, argp, true);
5538 case BTRFS_IOC_GET_DEV_STATS:
5539 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5540 case BTRFS_IOC_QUOTA_CTL:
5541 return btrfs_ioctl_quota_ctl(file, argp);
5542 case BTRFS_IOC_QGROUP_ASSIGN:
5543 return btrfs_ioctl_qgroup_assign(file, argp);
5544 case BTRFS_IOC_QGROUP_CREATE:
5545 return btrfs_ioctl_qgroup_create(file, argp);
5546 case BTRFS_IOC_QGROUP_LIMIT:
5547 return btrfs_ioctl_qgroup_limit(file, argp);
5548 case BTRFS_IOC_QUOTA_RESCAN:
5549 return btrfs_ioctl_quota_rescan(file, argp);
5550 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5551 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
5552 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5553 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
5554 case BTRFS_IOC_DEV_REPLACE:
5555 return btrfs_ioctl_dev_replace(fs_info, argp);
5556 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5557 return btrfs_ioctl_get_supported_features(argp);
5558 case BTRFS_IOC_GET_FEATURES:
5559 return btrfs_ioctl_get_features(fs_info, argp);
5560 case BTRFS_IOC_SET_FEATURES:
5561 return btrfs_ioctl_set_features(file, argp);
5562 case BTRFS_IOC_GET_SUBVOL_INFO:
5563 return btrfs_ioctl_get_subvol_info(inode, argp);
5564 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5565 return btrfs_ioctl_get_subvol_rootref(root, argp);
5566 case BTRFS_IOC_INO_LOOKUP_USER:
5567 return btrfs_ioctl_ino_lookup_user(file, argp);
5568 case FS_IOC_ENABLE_VERITY:
5569 return fsverity_ioctl_enable(file, (const void __user *)argp);
5570 case FS_IOC_MEASURE_VERITY:
5571 return fsverity_ioctl_measure(file, argp);
5572 case BTRFS_IOC_ENCODED_READ:
5573 return btrfs_ioctl_encoded_read(file, argp, false);
5574 case BTRFS_IOC_ENCODED_WRITE:
5575 return btrfs_ioctl_encoded_write(file, argp, false);
5576 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5577 case BTRFS_IOC_ENCODED_READ_32:
5578 return btrfs_ioctl_encoded_read(file, argp, true);
5579 case BTRFS_IOC_ENCODED_WRITE_32:
5580 return btrfs_ioctl_encoded_write(file, argp, true);
5587 #ifdef CONFIG_COMPAT
5588 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5591 * These all access 32-bit values anyway so no further
5592 * handling is necessary.
5595 case FS_IOC32_GETVERSION:
5596 cmd = FS_IOC_GETVERSION;
5600 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));