Loading...
1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
24#include <linux/fsnotify.h>
25#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
30#include <linux/backing-dev.h>
31#include <linux/mount.h>
32#include <linux/mpage.h>
33#include <linux/namei.h>
34#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
39#include <linux/security.h>
40#include <linux/xattr.h>
41#include <linux/vmalloc.h>
42#include <linux/slab.h>
43#include <linux/blkdev.h>
44#include "compat.h"
45#include "ctree.h"
46#include "disk-io.h"
47#include "transaction.h"
48#include "btrfs_inode.h"
49#include "ioctl.h"
50#include "print-tree.h"
51#include "volumes.h"
52#include "locking.h"
53#include "inode-map.h"
54#include "backref.h"
55#include "rcu-string.h"
56
57/* Mask out flags that are inappropriate for the given type of inode. */
58static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
59{
60 if (S_ISDIR(mode))
61 return flags;
62 else if (S_ISREG(mode))
63 return flags & ~FS_DIRSYNC_FL;
64 else
65 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
66}
67
68/*
69 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
70 */
71static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
72{
73 unsigned int iflags = 0;
74
75 if (flags & BTRFS_INODE_SYNC)
76 iflags |= FS_SYNC_FL;
77 if (flags & BTRFS_INODE_IMMUTABLE)
78 iflags |= FS_IMMUTABLE_FL;
79 if (flags & BTRFS_INODE_APPEND)
80 iflags |= FS_APPEND_FL;
81 if (flags & BTRFS_INODE_NODUMP)
82 iflags |= FS_NODUMP_FL;
83 if (flags & BTRFS_INODE_NOATIME)
84 iflags |= FS_NOATIME_FL;
85 if (flags & BTRFS_INODE_DIRSYNC)
86 iflags |= FS_DIRSYNC_FL;
87 if (flags & BTRFS_INODE_NODATACOW)
88 iflags |= FS_NOCOW_FL;
89
90 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
91 iflags |= FS_COMPR_FL;
92 else if (flags & BTRFS_INODE_NOCOMPRESS)
93 iflags |= FS_NOCOMP_FL;
94
95 return iflags;
96}
97
98/*
99 * Update inode->i_flags based on the btrfs internal flags.
100 */
101void btrfs_update_iflags(struct inode *inode)
102{
103 struct btrfs_inode *ip = BTRFS_I(inode);
104
105 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
106
107 if (ip->flags & BTRFS_INODE_SYNC)
108 inode->i_flags |= S_SYNC;
109 if (ip->flags & BTRFS_INODE_IMMUTABLE)
110 inode->i_flags |= S_IMMUTABLE;
111 if (ip->flags & BTRFS_INODE_APPEND)
112 inode->i_flags |= S_APPEND;
113 if (ip->flags & BTRFS_INODE_NOATIME)
114 inode->i_flags |= S_NOATIME;
115 if (ip->flags & BTRFS_INODE_DIRSYNC)
116 inode->i_flags |= S_DIRSYNC;
117}
118
119/*
120 * Inherit flags from the parent inode.
121 *
122 * Currently only the compression flags and the cow flags are inherited.
123 */
124void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
125{
126 unsigned int flags;
127
128 if (!dir)
129 return;
130
131 flags = BTRFS_I(dir)->flags;
132
133 if (flags & BTRFS_INODE_NOCOMPRESS) {
134 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
135 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
136 } else if (flags & BTRFS_INODE_COMPRESS) {
137 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
138 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
139 }
140
141 if (flags & BTRFS_INODE_NODATACOW)
142 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
143
144 btrfs_update_iflags(inode);
145}
146
147static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
148{
149 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
150 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
151
152 if (copy_to_user(arg, &flags, sizeof(flags)))
153 return -EFAULT;
154 return 0;
155}
156
157static int check_flags(unsigned int flags)
158{
159 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
160 FS_NOATIME_FL | FS_NODUMP_FL | \
161 FS_SYNC_FL | FS_DIRSYNC_FL | \
162 FS_NOCOMP_FL | FS_COMPR_FL |
163 FS_NOCOW_FL))
164 return -EOPNOTSUPP;
165
166 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
167 return -EINVAL;
168
169 return 0;
170}
171
172static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
173{
174 struct inode *inode = file->f_path.dentry->d_inode;
175 struct btrfs_inode *ip = BTRFS_I(inode);
176 struct btrfs_root *root = ip->root;
177 struct btrfs_trans_handle *trans;
178 unsigned int flags, oldflags;
179 int ret;
180 u64 ip_oldflags;
181 unsigned int i_oldflags;
182
183 if (btrfs_root_readonly(root))
184 return -EROFS;
185
186 if (copy_from_user(&flags, arg, sizeof(flags)))
187 return -EFAULT;
188
189 ret = check_flags(flags);
190 if (ret)
191 return ret;
192
193 if (!inode_owner_or_capable(inode))
194 return -EACCES;
195
196 mutex_lock(&inode->i_mutex);
197
198 ip_oldflags = ip->flags;
199 i_oldflags = inode->i_flags;
200
201 flags = btrfs_mask_flags(inode->i_mode, flags);
202 oldflags = btrfs_flags_to_ioctl(ip->flags);
203 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
204 if (!capable(CAP_LINUX_IMMUTABLE)) {
205 ret = -EPERM;
206 goto out_unlock;
207 }
208 }
209
210 ret = mnt_want_write_file(file);
211 if (ret)
212 goto out_unlock;
213
214 if (flags & FS_SYNC_FL)
215 ip->flags |= BTRFS_INODE_SYNC;
216 else
217 ip->flags &= ~BTRFS_INODE_SYNC;
218 if (flags & FS_IMMUTABLE_FL)
219 ip->flags |= BTRFS_INODE_IMMUTABLE;
220 else
221 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
222 if (flags & FS_APPEND_FL)
223 ip->flags |= BTRFS_INODE_APPEND;
224 else
225 ip->flags &= ~BTRFS_INODE_APPEND;
226 if (flags & FS_NODUMP_FL)
227 ip->flags |= BTRFS_INODE_NODUMP;
228 else
229 ip->flags &= ~BTRFS_INODE_NODUMP;
230 if (flags & FS_NOATIME_FL)
231 ip->flags |= BTRFS_INODE_NOATIME;
232 else
233 ip->flags &= ~BTRFS_INODE_NOATIME;
234 if (flags & FS_DIRSYNC_FL)
235 ip->flags |= BTRFS_INODE_DIRSYNC;
236 else
237 ip->flags &= ~BTRFS_INODE_DIRSYNC;
238 if (flags & FS_NOCOW_FL)
239 ip->flags |= BTRFS_INODE_NODATACOW;
240 else
241 ip->flags &= ~BTRFS_INODE_NODATACOW;
242
243 /*
244 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
245 * flag may be changed automatically if compression code won't make
246 * things smaller.
247 */
248 if (flags & FS_NOCOMP_FL) {
249 ip->flags &= ~BTRFS_INODE_COMPRESS;
250 ip->flags |= BTRFS_INODE_NOCOMPRESS;
251 } else if (flags & FS_COMPR_FL) {
252 ip->flags |= BTRFS_INODE_COMPRESS;
253 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
254 } else {
255 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
256 }
257
258 trans = btrfs_start_transaction(root, 1);
259 if (IS_ERR(trans)) {
260 ret = PTR_ERR(trans);
261 goto out_drop;
262 }
263
264 btrfs_update_iflags(inode);
265 inode_inc_iversion(inode);
266 inode->i_ctime = CURRENT_TIME;
267 ret = btrfs_update_inode(trans, root, inode);
268
269 btrfs_end_transaction(trans, root);
270 out_drop:
271 if (ret) {
272 ip->flags = ip_oldflags;
273 inode->i_flags = i_oldflags;
274 }
275
276 mnt_drop_write_file(file);
277 out_unlock:
278 mutex_unlock(&inode->i_mutex);
279 return ret;
280}
281
282static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
283{
284 struct inode *inode = file->f_path.dentry->d_inode;
285
286 return put_user(inode->i_generation, arg);
287}
288
289static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
290{
291 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
292 struct btrfs_device *device;
293 struct request_queue *q;
294 struct fstrim_range range;
295 u64 minlen = ULLONG_MAX;
296 u64 num_devices = 0;
297 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
298 int ret;
299
300 if (!capable(CAP_SYS_ADMIN))
301 return -EPERM;
302
303 rcu_read_lock();
304 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
305 dev_list) {
306 if (!device->bdev)
307 continue;
308 q = bdev_get_queue(device->bdev);
309 if (blk_queue_discard(q)) {
310 num_devices++;
311 minlen = min((u64)q->limits.discard_granularity,
312 minlen);
313 }
314 }
315 rcu_read_unlock();
316
317 if (!num_devices)
318 return -EOPNOTSUPP;
319 if (copy_from_user(&range, arg, sizeof(range)))
320 return -EFAULT;
321 if (range.start > total_bytes)
322 return -EINVAL;
323
324 range.len = min(range.len, total_bytes - range.start);
325 range.minlen = max(range.minlen, minlen);
326 ret = btrfs_trim_fs(fs_info->tree_root, &range);
327 if (ret < 0)
328 return ret;
329
330 if (copy_to_user(arg, &range, sizeof(range)))
331 return -EFAULT;
332
333 return 0;
334}
335
336static noinline int create_subvol(struct btrfs_root *root,
337 struct dentry *dentry,
338 char *name, int namelen,
339 u64 *async_transid)
340{
341 struct btrfs_trans_handle *trans;
342 struct btrfs_key key;
343 struct btrfs_root_item root_item;
344 struct btrfs_inode_item *inode_item;
345 struct extent_buffer *leaf;
346 struct btrfs_root *new_root;
347 struct dentry *parent = dentry->d_parent;
348 struct inode *dir;
349 int ret;
350 int err;
351 u64 objectid;
352 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
353 u64 index = 0;
354
355 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
356 if (ret)
357 return ret;
358
359 dir = parent->d_inode;
360
361 /*
362 * 1 - inode item
363 * 2 - refs
364 * 1 - root item
365 * 2 - dir items
366 */
367 trans = btrfs_start_transaction(root, 6);
368 if (IS_ERR(trans))
369 return PTR_ERR(trans);
370
371 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
372 0, objectid, NULL, 0, 0, 0);
373 if (IS_ERR(leaf)) {
374 ret = PTR_ERR(leaf);
375 goto fail;
376 }
377
378 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
379 btrfs_set_header_bytenr(leaf, leaf->start);
380 btrfs_set_header_generation(leaf, trans->transid);
381 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
382 btrfs_set_header_owner(leaf, objectid);
383
384 write_extent_buffer(leaf, root->fs_info->fsid,
385 (unsigned long)btrfs_header_fsid(leaf),
386 BTRFS_FSID_SIZE);
387 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
388 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
389 BTRFS_UUID_SIZE);
390 btrfs_mark_buffer_dirty(leaf);
391
392 inode_item = &root_item.inode;
393 memset(inode_item, 0, sizeof(*inode_item));
394 inode_item->generation = cpu_to_le64(1);
395 inode_item->size = cpu_to_le64(3);
396 inode_item->nlink = cpu_to_le32(1);
397 inode_item->nbytes = cpu_to_le64(root->leafsize);
398 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
399
400 root_item.flags = 0;
401 root_item.byte_limit = 0;
402 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
403
404 btrfs_set_root_bytenr(&root_item, leaf->start);
405 btrfs_set_root_generation(&root_item, trans->transid);
406 btrfs_set_root_level(&root_item, 0);
407 btrfs_set_root_refs(&root_item, 1);
408 btrfs_set_root_used(&root_item, leaf->len);
409 btrfs_set_root_last_snapshot(&root_item, 0);
410
411 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
412 root_item.drop_level = 0;
413
414 btrfs_tree_unlock(leaf);
415 free_extent_buffer(leaf);
416 leaf = NULL;
417
418 btrfs_set_root_dirid(&root_item, new_dirid);
419
420 key.objectid = objectid;
421 key.offset = 0;
422 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
423 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
424 &root_item);
425 if (ret)
426 goto fail;
427
428 key.offset = (u64)-1;
429 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
430 if (IS_ERR(new_root)) {
431 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
432 ret = PTR_ERR(new_root);
433 goto fail;
434 }
435
436 btrfs_record_root_in_trans(trans, new_root);
437
438 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
439 if (ret) {
440 /* We potentially lose an unused inode item here */
441 btrfs_abort_transaction(trans, root, ret);
442 goto fail;
443 }
444
445 /*
446 * insert the directory item
447 */
448 ret = btrfs_set_inode_index(dir, &index);
449 if (ret) {
450 btrfs_abort_transaction(trans, root, ret);
451 goto fail;
452 }
453
454 ret = btrfs_insert_dir_item(trans, root,
455 name, namelen, dir, &key,
456 BTRFS_FT_DIR, index);
457 if (ret) {
458 btrfs_abort_transaction(trans, root, ret);
459 goto fail;
460 }
461
462 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
463 ret = btrfs_update_inode(trans, root, dir);
464 BUG_ON(ret);
465
466 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
467 objectid, root->root_key.objectid,
468 btrfs_ino(dir), index, name, namelen);
469
470 BUG_ON(ret);
471
472 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
473fail:
474 if (async_transid) {
475 *async_transid = trans->transid;
476 err = btrfs_commit_transaction_async(trans, root, 1);
477 } else {
478 err = btrfs_commit_transaction(trans, root);
479 }
480 if (err && !ret)
481 ret = err;
482 return ret;
483}
484
485static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
486 char *name, int namelen, u64 *async_transid,
487 bool readonly)
488{
489 struct inode *inode;
490 struct btrfs_pending_snapshot *pending_snapshot;
491 struct btrfs_trans_handle *trans;
492 int ret;
493
494 if (!root->ref_cows)
495 return -EINVAL;
496
497 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
498 if (!pending_snapshot)
499 return -ENOMEM;
500
501 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
502 pending_snapshot->dentry = dentry;
503 pending_snapshot->root = root;
504 pending_snapshot->readonly = readonly;
505
506 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
507 if (IS_ERR(trans)) {
508 ret = PTR_ERR(trans);
509 goto fail;
510 }
511
512 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
513 BUG_ON(ret);
514
515 spin_lock(&root->fs_info->trans_lock);
516 list_add(&pending_snapshot->list,
517 &trans->transaction->pending_snapshots);
518 spin_unlock(&root->fs_info->trans_lock);
519 if (async_transid) {
520 *async_transid = trans->transid;
521 ret = btrfs_commit_transaction_async(trans,
522 root->fs_info->extent_root, 1);
523 } else {
524 ret = btrfs_commit_transaction(trans,
525 root->fs_info->extent_root);
526 }
527 BUG_ON(ret);
528
529 ret = pending_snapshot->error;
530 if (ret)
531 goto fail;
532
533 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
534 if (ret)
535 goto fail;
536
537 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
538 if (IS_ERR(inode)) {
539 ret = PTR_ERR(inode);
540 goto fail;
541 }
542 BUG_ON(!inode);
543 d_instantiate(dentry, inode);
544 ret = 0;
545fail:
546 kfree(pending_snapshot);
547 return ret;
548}
549
550/* copy of check_sticky in fs/namei.c()
551* It's inline, so penalty for filesystems that don't use sticky bit is
552* minimal.
553*/
554static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
555{
556 uid_t fsuid = current_fsuid();
557
558 if (!(dir->i_mode & S_ISVTX))
559 return 0;
560 if (inode->i_uid == fsuid)
561 return 0;
562 if (dir->i_uid == fsuid)
563 return 0;
564 return !capable(CAP_FOWNER);
565}
566
567/* copy of may_delete in fs/namei.c()
568 * Check whether we can remove a link victim from directory dir, check
569 * whether the type of victim is right.
570 * 1. We can't do it if dir is read-only (done in permission())
571 * 2. We should have write and exec permissions on dir
572 * 3. We can't remove anything from append-only dir
573 * 4. We can't do anything with immutable dir (done in permission())
574 * 5. If the sticky bit on dir is set we should either
575 * a. be owner of dir, or
576 * b. be owner of victim, or
577 * c. have CAP_FOWNER capability
578 * 6. If the victim is append-only or immutable we can't do antyhing with
579 * links pointing to it.
580 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
581 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
582 * 9. We can't remove a root or mountpoint.
583 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
584 * nfs_async_unlink().
585 */
586
587static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
588{
589 int error;
590
591 if (!victim->d_inode)
592 return -ENOENT;
593
594 BUG_ON(victim->d_parent->d_inode != dir);
595 audit_inode_child(victim, dir);
596
597 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
598 if (error)
599 return error;
600 if (IS_APPEND(dir))
601 return -EPERM;
602 if (btrfs_check_sticky(dir, victim->d_inode)||
603 IS_APPEND(victim->d_inode)||
604 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
605 return -EPERM;
606 if (isdir) {
607 if (!S_ISDIR(victim->d_inode->i_mode))
608 return -ENOTDIR;
609 if (IS_ROOT(victim))
610 return -EBUSY;
611 } else if (S_ISDIR(victim->d_inode->i_mode))
612 return -EISDIR;
613 if (IS_DEADDIR(dir))
614 return -ENOENT;
615 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
616 return -EBUSY;
617 return 0;
618}
619
620/* copy of may_create in fs/namei.c() */
621static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
622{
623 if (child->d_inode)
624 return -EEXIST;
625 if (IS_DEADDIR(dir))
626 return -ENOENT;
627 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
628}
629
630/*
631 * Create a new subvolume below @parent. This is largely modeled after
632 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
633 * inside this filesystem so it's quite a bit simpler.
634 */
635static noinline int btrfs_mksubvol(struct path *parent,
636 char *name, int namelen,
637 struct btrfs_root *snap_src,
638 u64 *async_transid, bool readonly)
639{
640 struct inode *dir = parent->dentry->d_inode;
641 struct dentry *dentry;
642 int error;
643
644 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
645
646 dentry = lookup_one_len(name, parent->dentry, namelen);
647 error = PTR_ERR(dentry);
648 if (IS_ERR(dentry))
649 goto out_unlock;
650
651 error = -EEXIST;
652 if (dentry->d_inode)
653 goto out_dput;
654
655 error = mnt_want_write(parent->mnt);
656 if (error)
657 goto out_dput;
658
659 error = btrfs_may_create(dir, dentry);
660 if (error)
661 goto out_drop_write;
662
663 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
664
665 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
666 goto out_up_read;
667
668 if (snap_src) {
669 error = create_snapshot(snap_src, dentry,
670 name, namelen, async_transid, readonly);
671 } else {
672 error = create_subvol(BTRFS_I(dir)->root, dentry,
673 name, namelen, async_transid);
674 }
675 if (!error)
676 fsnotify_mkdir(dir, dentry);
677out_up_read:
678 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
679out_drop_write:
680 mnt_drop_write(parent->mnt);
681out_dput:
682 dput(dentry);
683out_unlock:
684 mutex_unlock(&dir->i_mutex);
685 return error;
686}
687
688/*
689 * When we're defragging a range, we don't want to kick it off again
690 * if it is really just waiting for delalloc to send it down.
691 * If we find a nice big extent or delalloc range for the bytes in the
692 * file you want to defrag, we return 0 to let you know to skip this
693 * part of the file
694 */
695static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
696{
697 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
698 struct extent_map *em = NULL;
699 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
700 u64 end;
701
702 read_lock(&em_tree->lock);
703 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
704 read_unlock(&em_tree->lock);
705
706 if (em) {
707 end = extent_map_end(em);
708 free_extent_map(em);
709 if (end - offset > thresh)
710 return 0;
711 }
712 /* if we already have a nice delalloc here, just stop */
713 thresh /= 2;
714 end = count_range_bits(io_tree, &offset, offset + thresh,
715 thresh, EXTENT_DELALLOC, 1);
716 if (end >= thresh)
717 return 0;
718 return 1;
719}
720
721/*
722 * helper function to walk through a file and find extents
723 * newer than a specific transid, and smaller than thresh.
724 *
725 * This is used by the defragging code to find new and small
726 * extents
727 */
728static int find_new_extents(struct btrfs_root *root,
729 struct inode *inode, u64 newer_than,
730 u64 *off, int thresh)
731{
732 struct btrfs_path *path;
733 struct btrfs_key min_key;
734 struct btrfs_key max_key;
735 struct extent_buffer *leaf;
736 struct btrfs_file_extent_item *extent;
737 int type;
738 int ret;
739 u64 ino = btrfs_ino(inode);
740
741 path = btrfs_alloc_path();
742 if (!path)
743 return -ENOMEM;
744
745 min_key.objectid = ino;
746 min_key.type = BTRFS_EXTENT_DATA_KEY;
747 min_key.offset = *off;
748
749 max_key.objectid = ino;
750 max_key.type = (u8)-1;
751 max_key.offset = (u64)-1;
752
753 path->keep_locks = 1;
754
755 while(1) {
756 ret = btrfs_search_forward(root, &min_key, &max_key,
757 path, 0, newer_than);
758 if (ret != 0)
759 goto none;
760 if (min_key.objectid != ino)
761 goto none;
762 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
763 goto none;
764
765 leaf = path->nodes[0];
766 extent = btrfs_item_ptr(leaf, path->slots[0],
767 struct btrfs_file_extent_item);
768
769 type = btrfs_file_extent_type(leaf, extent);
770 if (type == BTRFS_FILE_EXTENT_REG &&
771 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
772 check_defrag_in_cache(inode, min_key.offset, thresh)) {
773 *off = min_key.offset;
774 btrfs_free_path(path);
775 return 0;
776 }
777
778 if (min_key.offset == (u64)-1)
779 goto none;
780
781 min_key.offset++;
782 btrfs_release_path(path);
783 }
784none:
785 btrfs_free_path(path);
786 return -ENOENT;
787}
788
789static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
790{
791 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
792 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
793 struct extent_map *em;
794 u64 len = PAGE_CACHE_SIZE;
795
796 /*
797 * hopefully we have this extent in the tree already, try without
798 * the full extent lock
799 */
800 read_lock(&em_tree->lock);
801 em = lookup_extent_mapping(em_tree, start, len);
802 read_unlock(&em_tree->lock);
803
804 if (!em) {
805 /* get the big lock and read metadata off disk */
806 lock_extent(io_tree, start, start + len - 1);
807 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
808 unlock_extent(io_tree, start, start + len - 1);
809
810 if (IS_ERR(em))
811 return NULL;
812 }
813
814 return em;
815}
816
817static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
818{
819 struct extent_map *next;
820 bool ret = true;
821
822 /* this is the last extent */
823 if (em->start + em->len >= i_size_read(inode))
824 return false;
825
826 next = defrag_lookup_extent(inode, em->start + em->len);
827 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
828 ret = false;
829
830 free_extent_map(next);
831 return ret;
832}
833
834static int should_defrag_range(struct inode *inode, u64 start, int thresh,
835 u64 *last_len, u64 *skip, u64 *defrag_end)
836{
837 struct extent_map *em;
838 int ret = 1;
839 bool next_mergeable = true;
840
841 /*
842 * make sure that once we start defragging an extent, we keep on
843 * defragging it
844 */
845 if (start < *defrag_end)
846 return 1;
847
848 *skip = 0;
849
850 em = defrag_lookup_extent(inode, start);
851 if (!em)
852 return 0;
853
854 /* this will cover holes, and inline extents */
855 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
856 ret = 0;
857 goto out;
858 }
859
860 next_mergeable = defrag_check_next_extent(inode, em);
861
862 /*
863 * we hit a real extent, if it is big or the next extent is not a
864 * real extent, don't bother defragging it
865 */
866 if ((*last_len == 0 || *last_len >= thresh) &&
867 (em->len >= thresh || !next_mergeable))
868 ret = 0;
869out:
870 /*
871 * last_len ends up being a counter of how many bytes we've defragged.
872 * every time we choose not to defrag an extent, we reset *last_len
873 * so that the next tiny extent will force a defrag.
874 *
875 * The end result of this is that tiny extents before a single big
876 * extent will force at least part of that big extent to be defragged.
877 */
878 if (ret) {
879 *defrag_end = extent_map_end(em);
880 } else {
881 *last_len = 0;
882 *skip = extent_map_end(em);
883 *defrag_end = 0;
884 }
885
886 free_extent_map(em);
887 return ret;
888}
889
890/*
891 * it doesn't do much good to defrag one or two pages
892 * at a time. This pulls in a nice chunk of pages
893 * to COW and defrag.
894 *
895 * It also makes sure the delalloc code has enough
896 * dirty data to avoid making new small extents as part
897 * of the defrag
898 *
899 * It's a good idea to start RA on this range
900 * before calling this.
901 */
902static int cluster_pages_for_defrag(struct inode *inode,
903 struct page **pages,
904 unsigned long start_index,
905 int num_pages)
906{
907 unsigned long file_end;
908 u64 isize = i_size_read(inode);
909 u64 page_start;
910 u64 page_end;
911 u64 page_cnt;
912 int ret;
913 int i;
914 int i_done;
915 struct btrfs_ordered_extent *ordered;
916 struct extent_state *cached_state = NULL;
917 struct extent_io_tree *tree;
918 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
919
920 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
921 if (!isize || start_index > file_end)
922 return 0;
923
924 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
925
926 ret = btrfs_delalloc_reserve_space(inode,
927 page_cnt << PAGE_CACHE_SHIFT);
928 if (ret)
929 return ret;
930 i_done = 0;
931 tree = &BTRFS_I(inode)->io_tree;
932
933 /* step one, lock all the pages */
934 for (i = 0; i < page_cnt; i++) {
935 struct page *page;
936again:
937 page = find_or_create_page(inode->i_mapping,
938 start_index + i, mask);
939 if (!page)
940 break;
941
942 page_start = page_offset(page);
943 page_end = page_start + PAGE_CACHE_SIZE - 1;
944 while (1) {
945 lock_extent(tree, page_start, page_end);
946 ordered = btrfs_lookup_ordered_extent(inode,
947 page_start);
948 unlock_extent(tree, page_start, page_end);
949 if (!ordered)
950 break;
951
952 unlock_page(page);
953 btrfs_start_ordered_extent(inode, ordered, 1);
954 btrfs_put_ordered_extent(ordered);
955 lock_page(page);
956 /*
957 * we unlocked the page above, so we need check if
958 * it was released or not.
959 */
960 if (page->mapping != inode->i_mapping) {
961 unlock_page(page);
962 page_cache_release(page);
963 goto again;
964 }
965 }
966
967 if (!PageUptodate(page)) {
968 btrfs_readpage(NULL, page);
969 lock_page(page);
970 if (!PageUptodate(page)) {
971 unlock_page(page);
972 page_cache_release(page);
973 ret = -EIO;
974 break;
975 }
976 }
977
978 if (page->mapping != inode->i_mapping) {
979 unlock_page(page);
980 page_cache_release(page);
981 goto again;
982 }
983
984 pages[i] = page;
985 i_done++;
986 }
987 if (!i_done || ret)
988 goto out;
989
990 if (!(inode->i_sb->s_flags & MS_ACTIVE))
991 goto out;
992
993 /*
994 * so now we have a nice long stream of locked
995 * and up to date pages, lets wait on them
996 */
997 for (i = 0; i < i_done; i++)
998 wait_on_page_writeback(pages[i]);
999
1000 page_start = page_offset(pages[0]);
1001 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1002
1003 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1004 page_start, page_end - 1, 0, &cached_state);
1005 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1006 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1007 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1008 GFP_NOFS);
1009
1010 if (i_done != page_cnt) {
1011 spin_lock(&BTRFS_I(inode)->lock);
1012 BTRFS_I(inode)->outstanding_extents++;
1013 spin_unlock(&BTRFS_I(inode)->lock);
1014 btrfs_delalloc_release_space(inode,
1015 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1016 }
1017
1018
1019 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1020 &cached_state);
1021
1022 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1023 page_start, page_end - 1, &cached_state,
1024 GFP_NOFS);
1025
1026 for (i = 0; i < i_done; i++) {
1027 clear_page_dirty_for_io(pages[i]);
1028 ClearPageChecked(pages[i]);
1029 set_page_extent_mapped(pages[i]);
1030 set_page_dirty(pages[i]);
1031 unlock_page(pages[i]);
1032 page_cache_release(pages[i]);
1033 }
1034 return i_done;
1035out:
1036 for (i = 0; i < i_done; i++) {
1037 unlock_page(pages[i]);
1038 page_cache_release(pages[i]);
1039 }
1040 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1041 return ret;
1042
1043}
1044
1045int btrfs_defrag_file(struct inode *inode, struct file *file,
1046 struct btrfs_ioctl_defrag_range_args *range,
1047 u64 newer_than, unsigned long max_to_defrag)
1048{
1049 struct btrfs_root *root = BTRFS_I(inode)->root;
1050 struct btrfs_super_block *disk_super;
1051 struct file_ra_state *ra = NULL;
1052 unsigned long last_index;
1053 u64 isize = i_size_read(inode);
1054 u64 features;
1055 u64 last_len = 0;
1056 u64 skip = 0;
1057 u64 defrag_end = 0;
1058 u64 newer_off = range->start;
1059 unsigned long i;
1060 unsigned long ra_index = 0;
1061 int ret;
1062 int defrag_count = 0;
1063 int compress_type = BTRFS_COMPRESS_ZLIB;
1064 int extent_thresh = range->extent_thresh;
1065 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1066 int cluster = max_cluster;
1067 u64 new_align = ~((u64)128 * 1024 - 1);
1068 struct page **pages = NULL;
1069
1070 if (extent_thresh == 0)
1071 extent_thresh = 256 * 1024;
1072
1073 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1074 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1075 return -EINVAL;
1076 if (range->compress_type)
1077 compress_type = range->compress_type;
1078 }
1079
1080 if (isize == 0)
1081 return 0;
1082
1083 /*
1084 * if we were not given a file, allocate a readahead
1085 * context
1086 */
1087 if (!file) {
1088 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1089 if (!ra)
1090 return -ENOMEM;
1091 file_ra_state_init(ra, inode->i_mapping);
1092 } else {
1093 ra = &file->f_ra;
1094 }
1095
1096 pages = kmalloc(sizeof(struct page *) * max_cluster,
1097 GFP_NOFS);
1098 if (!pages) {
1099 ret = -ENOMEM;
1100 goto out_ra;
1101 }
1102
1103 /* find the last page to defrag */
1104 if (range->start + range->len > range->start) {
1105 last_index = min_t(u64, isize - 1,
1106 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1107 } else {
1108 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1109 }
1110
1111 if (newer_than) {
1112 ret = find_new_extents(root, inode, newer_than,
1113 &newer_off, 64 * 1024);
1114 if (!ret) {
1115 range->start = newer_off;
1116 /*
1117 * we always align our defrag to help keep
1118 * the extents in the file evenly spaced
1119 */
1120 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1121 } else
1122 goto out_ra;
1123 } else {
1124 i = range->start >> PAGE_CACHE_SHIFT;
1125 }
1126 if (!max_to_defrag)
1127 max_to_defrag = last_index + 1;
1128
1129 /*
1130 * make writeback starts from i, so the defrag range can be
1131 * written sequentially.
1132 */
1133 if (i < inode->i_mapping->writeback_index)
1134 inode->i_mapping->writeback_index = i;
1135
1136 while (i <= last_index && defrag_count < max_to_defrag &&
1137 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1138 PAGE_CACHE_SHIFT)) {
1139 /*
1140 * make sure we stop running if someone unmounts
1141 * the FS
1142 */
1143 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1144 break;
1145
1146 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1147 extent_thresh, &last_len, &skip,
1148 &defrag_end)) {
1149 unsigned long next;
1150 /*
1151 * the should_defrag function tells us how much to skip
1152 * bump our counter by the suggested amount
1153 */
1154 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1155 i = max(i + 1, next);
1156 continue;
1157 }
1158
1159 if (!newer_than) {
1160 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1161 PAGE_CACHE_SHIFT) - i;
1162 cluster = min(cluster, max_cluster);
1163 } else {
1164 cluster = max_cluster;
1165 }
1166
1167 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1168 BTRFS_I(inode)->force_compress = compress_type;
1169
1170 if (i + cluster > ra_index) {
1171 ra_index = max(i, ra_index);
1172 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1173 cluster);
1174 ra_index += max_cluster;
1175 }
1176
1177 mutex_lock(&inode->i_mutex);
1178 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1179 if (ret < 0) {
1180 mutex_unlock(&inode->i_mutex);
1181 goto out_ra;
1182 }
1183
1184 defrag_count += ret;
1185 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1186 mutex_unlock(&inode->i_mutex);
1187
1188 if (newer_than) {
1189 if (newer_off == (u64)-1)
1190 break;
1191
1192 if (ret > 0)
1193 i += ret;
1194
1195 newer_off = max(newer_off + 1,
1196 (u64)i << PAGE_CACHE_SHIFT);
1197
1198 ret = find_new_extents(root, inode,
1199 newer_than, &newer_off,
1200 64 * 1024);
1201 if (!ret) {
1202 range->start = newer_off;
1203 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1204 } else {
1205 break;
1206 }
1207 } else {
1208 if (ret > 0) {
1209 i += ret;
1210 last_len += ret << PAGE_CACHE_SHIFT;
1211 } else {
1212 i++;
1213 last_len = 0;
1214 }
1215 }
1216 }
1217
1218 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1219 filemap_flush(inode->i_mapping);
1220
1221 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1222 /* the filemap_flush will queue IO into the worker threads, but
1223 * we have to make sure the IO is actually started and that
1224 * ordered extents get created before we return
1225 */
1226 atomic_inc(&root->fs_info->async_submit_draining);
1227 while (atomic_read(&root->fs_info->nr_async_submits) ||
1228 atomic_read(&root->fs_info->async_delalloc_pages)) {
1229 wait_event(root->fs_info->async_submit_wait,
1230 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1231 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1232 }
1233 atomic_dec(&root->fs_info->async_submit_draining);
1234
1235 mutex_lock(&inode->i_mutex);
1236 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1237 mutex_unlock(&inode->i_mutex);
1238 }
1239
1240 disk_super = root->fs_info->super_copy;
1241 features = btrfs_super_incompat_flags(disk_super);
1242 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1243 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1244 btrfs_set_super_incompat_flags(disk_super, features);
1245 }
1246
1247 ret = defrag_count;
1248
1249out_ra:
1250 if (!file)
1251 kfree(ra);
1252 kfree(pages);
1253 return ret;
1254}
1255
1256static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1257 void __user *arg)
1258{
1259 u64 new_size;
1260 u64 old_size;
1261 u64 devid = 1;
1262 struct btrfs_ioctl_vol_args *vol_args;
1263 struct btrfs_trans_handle *trans;
1264 struct btrfs_device *device = NULL;
1265 char *sizestr;
1266 char *devstr = NULL;
1267 int ret = 0;
1268 int mod = 0;
1269
1270 if (root->fs_info->sb->s_flags & MS_RDONLY)
1271 return -EROFS;
1272
1273 if (!capable(CAP_SYS_ADMIN))
1274 return -EPERM;
1275
1276 mutex_lock(&root->fs_info->volume_mutex);
1277 if (root->fs_info->balance_ctl) {
1278 printk(KERN_INFO "btrfs: balance in progress\n");
1279 ret = -EINVAL;
1280 goto out;
1281 }
1282
1283 vol_args = memdup_user(arg, sizeof(*vol_args));
1284 if (IS_ERR(vol_args)) {
1285 ret = PTR_ERR(vol_args);
1286 goto out;
1287 }
1288
1289 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1290
1291 sizestr = vol_args->name;
1292 devstr = strchr(sizestr, ':');
1293 if (devstr) {
1294 char *end;
1295 sizestr = devstr + 1;
1296 *devstr = '\0';
1297 devstr = vol_args->name;
1298 devid = simple_strtoull(devstr, &end, 10);
1299 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1300 (unsigned long long)devid);
1301 }
1302 device = btrfs_find_device(root, devid, NULL, NULL);
1303 if (!device) {
1304 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1305 (unsigned long long)devid);
1306 ret = -EINVAL;
1307 goto out_free;
1308 }
1309 if (device->fs_devices && device->fs_devices->seeding) {
1310 printk(KERN_INFO "btrfs: resizer unable to apply on "
1311 "seeding device %llu\n",
1312 (unsigned long long)devid);
1313 ret = -EINVAL;
1314 goto out_free;
1315 }
1316
1317 if (!strcmp(sizestr, "max"))
1318 new_size = device->bdev->bd_inode->i_size;
1319 else {
1320 if (sizestr[0] == '-') {
1321 mod = -1;
1322 sizestr++;
1323 } else if (sizestr[0] == '+') {
1324 mod = 1;
1325 sizestr++;
1326 }
1327 new_size = memparse(sizestr, NULL);
1328 if (new_size == 0) {
1329 ret = -EINVAL;
1330 goto out_free;
1331 }
1332 }
1333
1334 old_size = device->total_bytes;
1335
1336 if (mod < 0) {
1337 if (new_size > old_size) {
1338 ret = -EINVAL;
1339 goto out_free;
1340 }
1341 new_size = old_size - new_size;
1342 } else if (mod > 0) {
1343 new_size = old_size + new_size;
1344 }
1345
1346 if (new_size < 256 * 1024 * 1024) {
1347 ret = -EINVAL;
1348 goto out_free;
1349 }
1350 if (new_size > device->bdev->bd_inode->i_size) {
1351 ret = -EFBIG;
1352 goto out_free;
1353 }
1354
1355 do_div(new_size, root->sectorsize);
1356 new_size *= root->sectorsize;
1357
1358 printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1359 rcu_str_deref(device->name),
1360 (unsigned long long)new_size);
1361
1362 if (new_size > old_size) {
1363 trans = btrfs_start_transaction(root, 0);
1364 if (IS_ERR(trans)) {
1365 ret = PTR_ERR(trans);
1366 goto out_free;
1367 }
1368 ret = btrfs_grow_device(trans, device, new_size);
1369 btrfs_commit_transaction(trans, root);
1370 } else if (new_size < old_size) {
1371 ret = btrfs_shrink_device(device, new_size);
1372 }
1373
1374out_free:
1375 kfree(vol_args);
1376out:
1377 mutex_unlock(&root->fs_info->volume_mutex);
1378 return ret;
1379}
1380
1381static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1382 char *name,
1383 unsigned long fd,
1384 int subvol,
1385 u64 *transid,
1386 bool readonly)
1387{
1388 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1389 struct file *src_file;
1390 int namelen;
1391 int ret = 0;
1392
1393 if (root->fs_info->sb->s_flags & MS_RDONLY)
1394 return -EROFS;
1395
1396 namelen = strlen(name);
1397 if (strchr(name, '/')) {
1398 ret = -EINVAL;
1399 goto out;
1400 }
1401
1402 if (name[0] == '.' &&
1403 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1404 ret = -EEXIST;
1405 goto out;
1406 }
1407
1408 if (subvol) {
1409 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1410 NULL, transid, readonly);
1411 } else {
1412 struct inode *src_inode;
1413 src_file = fget(fd);
1414 if (!src_file) {
1415 ret = -EINVAL;
1416 goto out;
1417 }
1418
1419 src_inode = src_file->f_path.dentry->d_inode;
1420 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1421 printk(KERN_INFO "btrfs: Snapshot src from "
1422 "another FS\n");
1423 ret = -EINVAL;
1424 fput(src_file);
1425 goto out;
1426 }
1427 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1428 BTRFS_I(src_inode)->root,
1429 transid, readonly);
1430 fput(src_file);
1431 }
1432out:
1433 return ret;
1434}
1435
1436static noinline int btrfs_ioctl_snap_create(struct file *file,
1437 void __user *arg, int subvol)
1438{
1439 struct btrfs_ioctl_vol_args *vol_args;
1440 int ret;
1441
1442 vol_args = memdup_user(arg, sizeof(*vol_args));
1443 if (IS_ERR(vol_args))
1444 return PTR_ERR(vol_args);
1445 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1446
1447 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1448 vol_args->fd, subvol,
1449 NULL, false);
1450
1451 kfree(vol_args);
1452 return ret;
1453}
1454
1455static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1456 void __user *arg, int subvol)
1457{
1458 struct btrfs_ioctl_vol_args_v2 *vol_args;
1459 int ret;
1460 u64 transid = 0;
1461 u64 *ptr = NULL;
1462 bool readonly = false;
1463
1464 vol_args = memdup_user(arg, sizeof(*vol_args));
1465 if (IS_ERR(vol_args))
1466 return PTR_ERR(vol_args);
1467 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1468
1469 if (vol_args->flags &
1470 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1471 ret = -EOPNOTSUPP;
1472 goto out;
1473 }
1474
1475 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1476 ptr = &transid;
1477 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1478 readonly = true;
1479
1480 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1481 vol_args->fd, subvol,
1482 ptr, readonly);
1483
1484 if (ret == 0 && ptr &&
1485 copy_to_user(arg +
1486 offsetof(struct btrfs_ioctl_vol_args_v2,
1487 transid), ptr, sizeof(*ptr)))
1488 ret = -EFAULT;
1489out:
1490 kfree(vol_args);
1491 return ret;
1492}
1493
1494static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1495 void __user *arg)
1496{
1497 struct inode *inode = fdentry(file)->d_inode;
1498 struct btrfs_root *root = BTRFS_I(inode)->root;
1499 int ret = 0;
1500 u64 flags = 0;
1501
1502 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1503 return -EINVAL;
1504
1505 down_read(&root->fs_info->subvol_sem);
1506 if (btrfs_root_readonly(root))
1507 flags |= BTRFS_SUBVOL_RDONLY;
1508 up_read(&root->fs_info->subvol_sem);
1509
1510 if (copy_to_user(arg, &flags, sizeof(flags)))
1511 ret = -EFAULT;
1512
1513 return ret;
1514}
1515
1516static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1517 void __user *arg)
1518{
1519 struct inode *inode = fdentry(file)->d_inode;
1520 struct btrfs_root *root = BTRFS_I(inode)->root;
1521 struct btrfs_trans_handle *trans;
1522 u64 root_flags;
1523 u64 flags;
1524 int ret = 0;
1525
1526 if (root->fs_info->sb->s_flags & MS_RDONLY)
1527 return -EROFS;
1528
1529 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1530 return -EINVAL;
1531
1532 if (copy_from_user(&flags, arg, sizeof(flags)))
1533 return -EFAULT;
1534
1535 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1536 return -EINVAL;
1537
1538 if (flags & ~BTRFS_SUBVOL_RDONLY)
1539 return -EOPNOTSUPP;
1540
1541 if (!inode_owner_or_capable(inode))
1542 return -EACCES;
1543
1544 down_write(&root->fs_info->subvol_sem);
1545
1546 /* nothing to do */
1547 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1548 goto out;
1549
1550 root_flags = btrfs_root_flags(&root->root_item);
1551 if (flags & BTRFS_SUBVOL_RDONLY)
1552 btrfs_set_root_flags(&root->root_item,
1553 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1554 else
1555 btrfs_set_root_flags(&root->root_item,
1556 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1557
1558 trans = btrfs_start_transaction(root, 1);
1559 if (IS_ERR(trans)) {
1560 ret = PTR_ERR(trans);
1561 goto out_reset;
1562 }
1563
1564 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1565 &root->root_key, &root->root_item);
1566
1567 btrfs_commit_transaction(trans, root);
1568out_reset:
1569 if (ret)
1570 btrfs_set_root_flags(&root->root_item, root_flags);
1571out:
1572 up_write(&root->fs_info->subvol_sem);
1573 return ret;
1574}
1575
1576/*
1577 * helper to check if the subvolume references other subvolumes
1578 */
1579static noinline int may_destroy_subvol(struct btrfs_root *root)
1580{
1581 struct btrfs_path *path;
1582 struct btrfs_key key;
1583 int ret;
1584
1585 path = btrfs_alloc_path();
1586 if (!path)
1587 return -ENOMEM;
1588
1589 key.objectid = root->root_key.objectid;
1590 key.type = BTRFS_ROOT_REF_KEY;
1591 key.offset = (u64)-1;
1592
1593 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1594 &key, path, 0, 0);
1595 if (ret < 0)
1596 goto out;
1597 BUG_ON(ret == 0);
1598
1599 ret = 0;
1600 if (path->slots[0] > 0) {
1601 path->slots[0]--;
1602 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1603 if (key.objectid == root->root_key.objectid &&
1604 key.type == BTRFS_ROOT_REF_KEY)
1605 ret = -ENOTEMPTY;
1606 }
1607out:
1608 btrfs_free_path(path);
1609 return ret;
1610}
1611
1612static noinline int key_in_sk(struct btrfs_key *key,
1613 struct btrfs_ioctl_search_key *sk)
1614{
1615 struct btrfs_key test;
1616 int ret;
1617
1618 test.objectid = sk->min_objectid;
1619 test.type = sk->min_type;
1620 test.offset = sk->min_offset;
1621
1622 ret = btrfs_comp_cpu_keys(key, &test);
1623 if (ret < 0)
1624 return 0;
1625
1626 test.objectid = sk->max_objectid;
1627 test.type = sk->max_type;
1628 test.offset = sk->max_offset;
1629
1630 ret = btrfs_comp_cpu_keys(key, &test);
1631 if (ret > 0)
1632 return 0;
1633 return 1;
1634}
1635
1636static noinline int copy_to_sk(struct btrfs_root *root,
1637 struct btrfs_path *path,
1638 struct btrfs_key *key,
1639 struct btrfs_ioctl_search_key *sk,
1640 char *buf,
1641 unsigned long *sk_offset,
1642 int *num_found)
1643{
1644 u64 found_transid;
1645 struct extent_buffer *leaf;
1646 struct btrfs_ioctl_search_header sh;
1647 unsigned long item_off;
1648 unsigned long item_len;
1649 int nritems;
1650 int i;
1651 int slot;
1652 int ret = 0;
1653
1654 leaf = path->nodes[0];
1655 slot = path->slots[0];
1656 nritems = btrfs_header_nritems(leaf);
1657
1658 if (btrfs_header_generation(leaf) > sk->max_transid) {
1659 i = nritems;
1660 goto advance_key;
1661 }
1662 found_transid = btrfs_header_generation(leaf);
1663
1664 for (i = slot; i < nritems; i++) {
1665 item_off = btrfs_item_ptr_offset(leaf, i);
1666 item_len = btrfs_item_size_nr(leaf, i);
1667
1668 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1669 item_len = 0;
1670
1671 if (sizeof(sh) + item_len + *sk_offset >
1672 BTRFS_SEARCH_ARGS_BUFSIZE) {
1673 ret = 1;
1674 goto overflow;
1675 }
1676
1677 btrfs_item_key_to_cpu(leaf, key, i);
1678 if (!key_in_sk(key, sk))
1679 continue;
1680
1681 sh.objectid = key->objectid;
1682 sh.offset = key->offset;
1683 sh.type = key->type;
1684 sh.len = item_len;
1685 sh.transid = found_transid;
1686
1687 /* copy search result header */
1688 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1689 *sk_offset += sizeof(sh);
1690
1691 if (item_len) {
1692 char *p = buf + *sk_offset;
1693 /* copy the item */
1694 read_extent_buffer(leaf, p,
1695 item_off, item_len);
1696 *sk_offset += item_len;
1697 }
1698 (*num_found)++;
1699
1700 if (*num_found >= sk->nr_items)
1701 break;
1702 }
1703advance_key:
1704 ret = 0;
1705 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1706 key->offset++;
1707 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1708 key->offset = 0;
1709 key->type++;
1710 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1711 key->offset = 0;
1712 key->type = 0;
1713 key->objectid++;
1714 } else
1715 ret = 1;
1716overflow:
1717 return ret;
1718}
1719
1720static noinline int search_ioctl(struct inode *inode,
1721 struct btrfs_ioctl_search_args *args)
1722{
1723 struct btrfs_root *root;
1724 struct btrfs_key key;
1725 struct btrfs_key max_key;
1726 struct btrfs_path *path;
1727 struct btrfs_ioctl_search_key *sk = &args->key;
1728 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1729 int ret;
1730 int num_found = 0;
1731 unsigned long sk_offset = 0;
1732
1733 path = btrfs_alloc_path();
1734 if (!path)
1735 return -ENOMEM;
1736
1737 if (sk->tree_id == 0) {
1738 /* search the root of the inode that was passed */
1739 root = BTRFS_I(inode)->root;
1740 } else {
1741 key.objectid = sk->tree_id;
1742 key.type = BTRFS_ROOT_ITEM_KEY;
1743 key.offset = (u64)-1;
1744 root = btrfs_read_fs_root_no_name(info, &key);
1745 if (IS_ERR(root)) {
1746 printk(KERN_ERR "could not find root %llu\n",
1747 sk->tree_id);
1748 btrfs_free_path(path);
1749 return -ENOENT;
1750 }
1751 }
1752
1753 key.objectid = sk->min_objectid;
1754 key.type = sk->min_type;
1755 key.offset = sk->min_offset;
1756
1757 max_key.objectid = sk->max_objectid;
1758 max_key.type = sk->max_type;
1759 max_key.offset = sk->max_offset;
1760
1761 path->keep_locks = 1;
1762
1763 while(1) {
1764 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1765 sk->min_transid);
1766 if (ret != 0) {
1767 if (ret > 0)
1768 ret = 0;
1769 goto err;
1770 }
1771 ret = copy_to_sk(root, path, &key, sk, args->buf,
1772 &sk_offset, &num_found);
1773 btrfs_release_path(path);
1774 if (ret || num_found >= sk->nr_items)
1775 break;
1776
1777 }
1778 ret = 0;
1779err:
1780 sk->nr_items = num_found;
1781 btrfs_free_path(path);
1782 return ret;
1783}
1784
1785static noinline int btrfs_ioctl_tree_search(struct file *file,
1786 void __user *argp)
1787{
1788 struct btrfs_ioctl_search_args *args;
1789 struct inode *inode;
1790 int ret;
1791
1792 if (!capable(CAP_SYS_ADMIN))
1793 return -EPERM;
1794
1795 args = memdup_user(argp, sizeof(*args));
1796 if (IS_ERR(args))
1797 return PTR_ERR(args);
1798
1799 inode = fdentry(file)->d_inode;
1800 ret = search_ioctl(inode, args);
1801 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1802 ret = -EFAULT;
1803 kfree(args);
1804 return ret;
1805}
1806
1807/*
1808 * Search INODE_REFs to identify path name of 'dirid' directory
1809 * in a 'tree_id' tree. and sets path name to 'name'.
1810 */
1811static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1812 u64 tree_id, u64 dirid, char *name)
1813{
1814 struct btrfs_root *root;
1815 struct btrfs_key key;
1816 char *ptr;
1817 int ret = -1;
1818 int slot;
1819 int len;
1820 int total_len = 0;
1821 struct btrfs_inode_ref *iref;
1822 struct extent_buffer *l;
1823 struct btrfs_path *path;
1824
1825 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1826 name[0]='\0';
1827 return 0;
1828 }
1829
1830 path = btrfs_alloc_path();
1831 if (!path)
1832 return -ENOMEM;
1833
1834 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1835
1836 key.objectid = tree_id;
1837 key.type = BTRFS_ROOT_ITEM_KEY;
1838 key.offset = (u64)-1;
1839 root = btrfs_read_fs_root_no_name(info, &key);
1840 if (IS_ERR(root)) {
1841 printk(KERN_ERR "could not find root %llu\n", tree_id);
1842 ret = -ENOENT;
1843 goto out;
1844 }
1845
1846 key.objectid = dirid;
1847 key.type = BTRFS_INODE_REF_KEY;
1848 key.offset = (u64)-1;
1849
1850 while(1) {
1851 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1852 if (ret < 0)
1853 goto out;
1854
1855 l = path->nodes[0];
1856 slot = path->slots[0];
1857 if (ret > 0 && slot > 0)
1858 slot--;
1859 btrfs_item_key_to_cpu(l, &key, slot);
1860
1861 if (ret > 0 && (key.objectid != dirid ||
1862 key.type != BTRFS_INODE_REF_KEY)) {
1863 ret = -ENOENT;
1864 goto out;
1865 }
1866
1867 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1868 len = btrfs_inode_ref_name_len(l, iref);
1869 ptr -= len + 1;
1870 total_len += len + 1;
1871 if (ptr < name)
1872 goto out;
1873
1874 *(ptr + len) = '/';
1875 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1876
1877 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1878 break;
1879
1880 btrfs_release_path(path);
1881 key.objectid = key.offset;
1882 key.offset = (u64)-1;
1883 dirid = key.objectid;
1884 }
1885 if (ptr < name)
1886 goto out;
1887 memmove(name, ptr, total_len);
1888 name[total_len]='\0';
1889 ret = 0;
1890out:
1891 btrfs_free_path(path);
1892 return ret;
1893}
1894
1895static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1896 void __user *argp)
1897{
1898 struct btrfs_ioctl_ino_lookup_args *args;
1899 struct inode *inode;
1900 int ret;
1901
1902 if (!capable(CAP_SYS_ADMIN))
1903 return -EPERM;
1904
1905 args = memdup_user(argp, sizeof(*args));
1906 if (IS_ERR(args))
1907 return PTR_ERR(args);
1908
1909 inode = fdentry(file)->d_inode;
1910
1911 if (args->treeid == 0)
1912 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1913
1914 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1915 args->treeid, args->objectid,
1916 args->name);
1917
1918 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1919 ret = -EFAULT;
1920
1921 kfree(args);
1922 return ret;
1923}
1924
1925static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1926 void __user *arg)
1927{
1928 struct dentry *parent = fdentry(file);
1929 struct dentry *dentry;
1930 struct inode *dir = parent->d_inode;
1931 struct inode *inode;
1932 struct btrfs_root *root = BTRFS_I(dir)->root;
1933 struct btrfs_root *dest = NULL;
1934 struct btrfs_ioctl_vol_args *vol_args;
1935 struct btrfs_trans_handle *trans;
1936 int namelen;
1937 int ret;
1938 int err = 0;
1939
1940 vol_args = memdup_user(arg, sizeof(*vol_args));
1941 if (IS_ERR(vol_args))
1942 return PTR_ERR(vol_args);
1943
1944 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1945 namelen = strlen(vol_args->name);
1946 if (strchr(vol_args->name, '/') ||
1947 strncmp(vol_args->name, "..", namelen) == 0) {
1948 err = -EINVAL;
1949 goto out;
1950 }
1951
1952 err = mnt_want_write_file(file);
1953 if (err)
1954 goto out;
1955
1956 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1957 dentry = lookup_one_len(vol_args->name, parent, namelen);
1958 if (IS_ERR(dentry)) {
1959 err = PTR_ERR(dentry);
1960 goto out_unlock_dir;
1961 }
1962
1963 if (!dentry->d_inode) {
1964 err = -ENOENT;
1965 goto out_dput;
1966 }
1967
1968 inode = dentry->d_inode;
1969 dest = BTRFS_I(inode)->root;
1970 if (!capable(CAP_SYS_ADMIN)){
1971 /*
1972 * Regular user. Only allow this with a special mount
1973 * option, when the user has write+exec access to the
1974 * subvol root, and when rmdir(2) would have been
1975 * allowed.
1976 *
1977 * Note that this is _not_ check that the subvol is
1978 * empty or doesn't contain data that we wouldn't
1979 * otherwise be able to delete.
1980 *
1981 * Users who want to delete empty subvols should try
1982 * rmdir(2).
1983 */
1984 err = -EPERM;
1985 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1986 goto out_dput;
1987
1988 /*
1989 * Do not allow deletion if the parent dir is the same
1990 * as the dir to be deleted. That means the ioctl
1991 * must be called on the dentry referencing the root
1992 * of the subvol, not a random directory contained
1993 * within it.
1994 */
1995 err = -EINVAL;
1996 if (root == dest)
1997 goto out_dput;
1998
1999 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2000 if (err)
2001 goto out_dput;
2002
2003 /* check if subvolume may be deleted by a non-root user */
2004 err = btrfs_may_delete(dir, dentry, 1);
2005 if (err)
2006 goto out_dput;
2007 }
2008
2009 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2010 err = -EINVAL;
2011 goto out_dput;
2012 }
2013
2014 mutex_lock(&inode->i_mutex);
2015 err = d_invalidate(dentry);
2016 if (err)
2017 goto out_unlock;
2018
2019 down_write(&root->fs_info->subvol_sem);
2020
2021 err = may_destroy_subvol(dest);
2022 if (err)
2023 goto out_up_write;
2024
2025 trans = btrfs_start_transaction(root, 0);
2026 if (IS_ERR(trans)) {
2027 err = PTR_ERR(trans);
2028 goto out_up_write;
2029 }
2030 trans->block_rsv = &root->fs_info->global_block_rsv;
2031
2032 ret = btrfs_unlink_subvol(trans, root, dir,
2033 dest->root_key.objectid,
2034 dentry->d_name.name,
2035 dentry->d_name.len);
2036 if (ret) {
2037 err = ret;
2038 btrfs_abort_transaction(trans, root, ret);
2039 goto out_end_trans;
2040 }
2041
2042 btrfs_record_root_in_trans(trans, dest);
2043
2044 memset(&dest->root_item.drop_progress, 0,
2045 sizeof(dest->root_item.drop_progress));
2046 dest->root_item.drop_level = 0;
2047 btrfs_set_root_refs(&dest->root_item, 0);
2048
2049 if (!xchg(&dest->orphan_item_inserted, 1)) {
2050 ret = btrfs_insert_orphan_item(trans,
2051 root->fs_info->tree_root,
2052 dest->root_key.objectid);
2053 if (ret) {
2054 btrfs_abort_transaction(trans, root, ret);
2055 err = ret;
2056 goto out_end_trans;
2057 }
2058 }
2059out_end_trans:
2060 ret = btrfs_end_transaction(trans, root);
2061 if (ret && !err)
2062 err = ret;
2063 inode->i_flags |= S_DEAD;
2064out_up_write:
2065 up_write(&root->fs_info->subvol_sem);
2066out_unlock:
2067 mutex_unlock(&inode->i_mutex);
2068 if (!err) {
2069 shrink_dcache_sb(root->fs_info->sb);
2070 btrfs_invalidate_inodes(dest);
2071 d_delete(dentry);
2072 }
2073out_dput:
2074 dput(dentry);
2075out_unlock_dir:
2076 mutex_unlock(&dir->i_mutex);
2077 mnt_drop_write_file(file);
2078out:
2079 kfree(vol_args);
2080 return err;
2081}
2082
2083static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2084{
2085 struct inode *inode = fdentry(file)->d_inode;
2086 struct btrfs_root *root = BTRFS_I(inode)->root;
2087 struct btrfs_ioctl_defrag_range_args *range;
2088 int ret;
2089
2090 if (btrfs_root_readonly(root))
2091 return -EROFS;
2092
2093 ret = mnt_want_write_file(file);
2094 if (ret)
2095 return ret;
2096
2097 switch (inode->i_mode & S_IFMT) {
2098 case S_IFDIR:
2099 if (!capable(CAP_SYS_ADMIN)) {
2100 ret = -EPERM;
2101 goto out;
2102 }
2103 ret = btrfs_defrag_root(root, 0);
2104 if (ret)
2105 goto out;
2106 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2107 break;
2108 case S_IFREG:
2109 if (!(file->f_mode & FMODE_WRITE)) {
2110 ret = -EINVAL;
2111 goto out;
2112 }
2113
2114 range = kzalloc(sizeof(*range), GFP_KERNEL);
2115 if (!range) {
2116 ret = -ENOMEM;
2117 goto out;
2118 }
2119
2120 if (argp) {
2121 if (copy_from_user(range, argp,
2122 sizeof(*range))) {
2123 ret = -EFAULT;
2124 kfree(range);
2125 goto out;
2126 }
2127 /* compression requires us to start the IO */
2128 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2129 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2130 range->extent_thresh = (u32)-1;
2131 }
2132 } else {
2133 /* the rest are all set to zero by kzalloc */
2134 range->len = (u64)-1;
2135 }
2136 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2137 range, 0, 0);
2138 if (ret > 0)
2139 ret = 0;
2140 kfree(range);
2141 break;
2142 default:
2143 ret = -EINVAL;
2144 }
2145out:
2146 mnt_drop_write_file(file);
2147 return ret;
2148}
2149
2150static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2151{
2152 struct btrfs_ioctl_vol_args *vol_args;
2153 int ret;
2154
2155 if (!capable(CAP_SYS_ADMIN))
2156 return -EPERM;
2157
2158 mutex_lock(&root->fs_info->volume_mutex);
2159 if (root->fs_info->balance_ctl) {
2160 printk(KERN_INFO "btrfs: balance in progress\n");
2161 ret = -EINVAL;
2162 goto out;
2163 }
2164
2165 vol_args = memdup_user(arg, sizeof(*vol_args));
2166 if (IS_ERR(vol_args)) {
2167 ret = PTR_ERR(vol_args);
2168 goto out;
2169 }
2170
2171 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2172 ret = btrfs_init_new_device(root, vol_args->name);
2173
2174 kfree(vol_args);
2175out:
2176 mutex_unlock(&root->fs_info->volume_mutex);
2177 return ret;
2178}
2179
2180static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2181{
2182 struct btrfs_ioctl_vol_args *vol_args;
2183 int ret;
2184
2185 if (!capable(CAP_SYS_ADMIN))
2186 return -EPERM;
2187
2188 if (root->fs_info->sb->s_flags & MS_RDONLY)
2189 return -EROFS;
2190
2191 mutex_lock(&root->fs_info->volume_mutex);
2192 if (root->fs_info->balance_ctl) {
2193 printk(KERN_INFO "btrfs: balance in progress\n");
2194 ret = -EINVAL;
2195 goto out;
2196 }
2197
2198 vol_args = memdup_user(arg, sizeof(*vol_args));
2199 if (IS_ERR(vol_args)) {
2200 ret = PTR_ERR(vol_args);
2201 goto out;
2202 }
2203
2204 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2205 ret = btrfs_rm_device(root, vol_args->name);
2206
2207 kfree(vol_args);
2208out:
2209 mutex_unlock(&root->fs_info->volume_mutex);
2210 return ret;
2211}
2212
2213static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2214{
2215 struct btrfs_ioctl_fs_info_args *fi_args;
2216 struct btrfs_device *device;
2217 struct btrfs_device *next;
2218 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2219 int ret = 0;
2220
2221 if (!capable(CAP_SYS_ADMIN))
2222 return -EPERM;
2223
2224 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2225 if (!fi_args)
2226 return -ENOMEM;
2227
2228 fi_args->num_devices = fs_devices->num_devices;
2229 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2230
2231 mutex_lock(&fs_devices->device_list_mutex);
2232 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2233 if (device->devid > fi_args->max_id)
2234 fi_args->max_id = device->devid;
2235 }
2236 mutex_unlock(&fs_devices->device_list_mutex);
2237
2238 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2239 ret = -EFAULT;
2240
2241 kfree(fi_args);
2242 return ret;
2243}
2244
2245static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2246{
2247 struct btrfs_ioctl_dev_info_args *di_args;
2248 struct btrfs_device *dev;
2249 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2250 int ret = 0;
2251 char *s_uuid = NULL;
2252 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2253
2254 if (!capable(CAP_SYS_ADMIN))
2255 return -EPERM;
2256
2257 di_args = memdup_user(arg, sizeof(*di_args));
2258 if (IS_ERR(di_args))
2259 return PTR_ERR(di_args);
2260
2261 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2262 s_uuid = di_args->uuid;
2263
2264 mutex_lock(&fs_devices->device_list_mutex);
2265 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2266 mutex_unlock(&fs_devices->device_list_mutex);
2267
2268 if (!dev) {
2269 ret = -ENODEV;
2270 goto out;
2271 }
2272
2273 di_args->devid = dev->devid;
2274 di_args->bytes_used = dev->bytes_used;
2275 di_args->total_bytes = dev->total_bytes;
2276 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2277 if (dev->name) {
2278 struct rcu_string *name;
2279
2280 rcu_read_lock();
2281 name = rcu_dereference(dev->name);
2282 strncpy(di_args->path, name->str, sizeof(di_args->path));
2283 rcu_read_unlock();
2284 di_args->path[sizeof(di_args->path) - 1] = 0;
2285 } else {
2286 di_args->path[0] = '\0';
2287 }
2288
2289out:
2290 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2291 ret = -EFAULT;
2292
2293 kfree(di_args);
2294 return ret;
2295}
2296
2297static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2298 u64 off, u64 olen, u64 destoff)
2299{
2300 struct inode *inode = fdentry(file)->d_inode;
2301 struct btrfs_root *root = BTRFS_I(inode)->root;
2302 struct file *src_file;
2303 struct inode *src;
2304 struct btrfs_trans_handle *trans;
2305 struct btrfs_path *path;
2306 struct extent_buffer *leaf;
2307 char *buf;
2308 struct btrfs_key key;
2309 u32 nritems;
2310 int slot;
2311 int ret;
2312 u64 len = olen;
2313 u64 bs = root->fs_info->sb->s_blocksize;
2314 u64 hint_byte;
2315
2316 /*
2317 * TODO:
2318 * - split compressed inline extents. annoying: we need to
2319 * decompress into destination's address_space (the file offset
2320 * may change, so source mapping won't do), then recompress (or
2321 * otherwise reinsert) a subrange.
2322 * - allow ranges within the same file to be cloned (provided
2323 * they don't overlap)?
2324 */
2325
2326 /* the destination must be opened for writing */
2327 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2328 return -EINVAL;
2329
2330 if (btrfs_root_readonly(root))
2331 return -EROFS;
2332
2333 ret = mnt_want_write_file(file);
2334 if (ret)
2335 return ret;
2336
2337 src_file = fget(srcfd);
2338 if (!src_file) {
2339 ret = -EBADF;
2340 goto out_drop_write;
2341 }
2342
2343 src = src_file->f_dentry->d_inode;
2344
2345 ret = -EINVAL;
2346 if (src == inode)
2347 goto out_fput;
2348
2349 /* the src must be open for reading */
2350 if (!(src_file->f_mode & FMODE_READ))
2351 goto out_fput;
2352
2353 /* don't make the dst file partly checksummed */
2354 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2355 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2356 goto out_fput;
2357
2358 ret = -EISDIR;
2359 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2360 goto out_fput;
2361
2362 ret = -EXDEV;
2363 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2364 goto out_fput;
2365
2366 ret = -ENOMEM;
2367 buf = vmalloc(btrfs_level_size(root, 0));
2368 if (!buf)
2369 goto out_fput;
2370
2371 path = btrfs_alloc_path();
2372 if (!path) {
2373 vfree(buf);
2374 goto out_fput;
2375 }
2376 path->reada = 2;
2377
2378 if (inode < src) {
2379 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2380 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2381 } else {
2382 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2383 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2384 }
2385
2386 /* determine range to clone */
2387 ret = -EINVAL;
2388 if (off + len > src->i_size || off + len < off)
2389 goto out_unlock;
2390 if (len == 0)
2391 olen = len = src->i_size - off;
2392 /* if we extend to eof, continue to block boundary */
2393 if (off + len == src->i_size)
2394 len = ALIGN(src->i_size, bs) - off;
2395
2396 /* verify the end result is block aligned */
2397 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2398 !IS_ALIGNED(destoff, bs))
2399 goto out_unlock;
2400
2401 if (destoff > inode->i_size) {
2402 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2403 if (ret)
2404 goto out_unlock;
2405 }
2406
2407 /* truncate page cache pages from target inode range */
2408 truncate_inode_pages_range(&inode->i_data, destoff,
2409 PAGE_CACHE_ALIGN(destoff + len) - 1);
2410
2411 /* do any pending delalloc/csum calc on src, one way or
2412 another, and lock file content */
2413 while (1) {
2414 struct btrfs_ordered_extent *ordered;
2415 lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2416 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2417 if (!ordered &&
2418 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2419 EXTENT_DELALLOC, 0, NULL))
2420 break;
2421 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2422 if (ordered)
2423 btrfs_put_ordered_extent(ordered);
2424 btrfs_wait_ordered_range(src, off, len);
2425 }
2426
2427 /* clone data */
2428 key.objectid = btrfs_ino(src);
2429 key.type = BTRFS_EXTENT_DATA_KEY;
2430 key.offset = 0;
2431
2432 while (1) {
2433 /*
2434 * note the key will change type as we walk through the
2435 * tree.
2436 */
2437 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2438 if (ret < 0)
2439 goto out;
2440
2441 nritems = btrfs_header_nritems(path->nodes[0]);
2442 if (path->slots[0] >= nritems) {
2443 ret = btrfs_next_leaf(root, path);
2444 if (ret < 0)
2445 goto out;
2446 if (ret > 0)
2447 break;
2448 nritems = btrfs_header_nritems(path->nodes[0]);
2449 }
2450 leaf = path->nodes[0];
2451 slot = path->slots[0];
2452
2453 btrfs_item_key_to_cpu(leaf, &key, slot);
2454 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2455 key.objectid != btrfs_ino(src))
2456 break;
2457
2458 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2459 struct btrfs_file_extent_item *extent;
2460 int type;
2461 u32 size;
2462 struct btrfs_key new_key;
2463 u64 disko = 0, diskl = 0;
2464 u64 datao = 0, datal = 0;
2465 u8 comp;
2466 u64 endoff;
2467
2468 size = btrfs_item_size_nr(leaf, slot);
2469 read_extent_buffer(leaf, buf,
2470 btrfs_item_ptr_offset(leaf, slot),
2471 size);
2472
2473 extent = btrfs_item_ptr(leaf, slot,
2474 struct btrfs_file_extent_item);
2475 comp = btrfs_file_extent_compression(leaf, extent);
2476 type = btrfs_file_extent_type(leaf, extent);
2477 if (type == BTRFS_FILE_EXTENT_REG ||
2478 type == BTRFS_FILE_EXTENT_PREALLOC) {
2479 disko = btrfs_file_extent_disk_bytenr(leaf,
2480 extent);
2481 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2482 extent);
2483 datao = btrfs_file_extent_offset(leaf, extent);
2484 datal = btrfs_file_extent_num_bytes(leaf,
2485 extent);
2486 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2487 /* take upper bound, may be compressed */
2488 datal = btrfs_file_extent_ram_bytes(leaf,
2489 extent);
2490 }
2491 btrfs_release_path(path);
2492
2493 if (key.offset + datal <= off ||
2494 key.offset >= off+len)
2495 goto next;
2496
2497 memcpy(&new_key, &key, sizeof(new_key));
2498 new_key.objectid = btrfs_ino(inode);
2499 if (off <= key.offset)
2500 new_key.offset = key.offset + destoff - off;
2501 else
2502 new_key.offset = destoff;
2503
2504 /*
2505 * 1 - adjusting old extent (we may have to split it)
2506 * 1 - add new extent
2507 * 1 - inode update
2508 */
2509 trans = btrfs_start_transaction(root, 3);
2510 if (IS_ERR(trans)) {
2511 ret = PTR_ERR(trans);
2512 goto out;
2513 }
2514
2515 if (type == BTRFS_FILE_EXTENT_REG ||
2516 type == BTRFS_FILE_EXTENT_PREALLOC) {
2517 /*
2518 * a | --- range to clone ---| b
2519 * | ------------- extent ------------- |
2520 */
2521
2522 /* substract range b */
2523 if (key.offset + datal > off + len)
2524 datal = off + len - key.offset;
2525
2526 /* substract range a */
2527 if (off > key.offset) {
2528 datao += off - key.offset;
2529 datal -= off - key.offset;
2530 }
2531
2532 ret = btrfs_drop_extents(trans, inode,
2533 new_key.offset,
2534 new_key.offset + datal,
2535 &hint_byte, 1);
2536 if (ret) {
2537 btrfs_abort_transaction(trans, root,
2538 ret);
2539 btrfs_end_transaction(trans, root);
2540 goto out;
2541 }
2542
2543 ret = btrfs_insert_empty_item(trans, root, path,
2544 &new_key, size);
2545 if (ret) {
2546 btrfs_abort_transaction(trans, root,
2547 ret);
2548 btrfs_end_transaction(trans, root);
2549 goto out;
2550 }
2551
2552 leaf = path->nodes[0];
2553 slot = path->slots[0];
2554 write_extent_buffer(leaf, buf,
2555 btrfs_item_ptr_offset(leaf, slot),
2556 size);
2557
2558 extent = btrfs_item_ptr(leaf, slot,
2559 struct btrfs_file_extent_item);
2560
2561 /* disko == 0 means it's a hole */
2562 if (!disko)
2563 datao = 0;
2564
2565 btrfs_set_file_extent_offset(leaf, extent,
2566 datao);
2567 btrfs_set_file_extent_num_bytes(leaf, extent,
2568 datal);
2569 if (disko) {
2570 inode_add_bytes(inode, datal);
2571 ret = btrfs_inc_extent_ref(trans, root,
2572 disko, diskl, 0,
2573 root->root_key.objectid,
2574 btrfs_ino(inode),
2575 new_key.offset - datao,
2576 0);
2577 if (ret) {
2578 btrfs_abort_transaction(trans,
2579 root,
2580 ret);
2581 btrfs_end_transaction(trans,
2582 root);
2583 goto out;
2584
2585 }
2586 }
2587 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2588 u64 skip = 0;
2589 u64 trim = 0;
2590 if (off > key.offset) {
2591 skip = off - key.offset;
2592 new_key.offset += skip;
2593 }
2594
2595 if (key.offset + datal > off+len)
2596 trim = key.offset + datal - (off+len);
2597
2598 if (comp && (skip || trim)) {
2599 ret = -EINVAL;
2600 btrfs_end_transaction(trans, root);
2601 goto out;
2602 }
2603 size -= skip + trim;
2604 datal -= skip + trim;
2605
2606 ret = btrfs_drop_extents(trans, inode,
2607 new_key.offset,
2608 new_key.offset + datal,
2609 &hint_byte, 1);
2610 if (ret) {
2611 btrfs_abort_transaction(trans, root,
2612 ret);
2613 btrfs_end_transaction(trans, root);
2614 goto out;
2615 }
2616
2617 ret = btrfs_insert_empty_item(trans, root, path,
2618 &new_key, size);
2619 if (ret) {
2620 btrfs_abort_transaction(trans, root,
2621 ret);
2622 btrfs_end_transaction(trans, root);
2623 goto out;
2624 }
2625
2626 if (skip) {
2627 u32 start =
2628 btrfs_file_extent_calc_inline_size(0);
2629 memmove(buf+start, buf+start+skip,
2630 datal);
2631 }
2632
2633 leaf = path->nodes[0];
2634 slot = path->slots[0];
2635 write_extent_buffer(leaf, buf,
2636 btrfs_item_ptr_offset(leaf, slot),
2637 size);
2638 inode_add_bytes(inode, datal);
2639 }
2640
2641 btrfs_mark_buffer_dirty(leaf);
2642 btrfs_release_path(path);
2643
2644 inode_inc_iversion(inode);
2645 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2646
2647 /*
2648 * we round up to the block size at eof when
2649 * determining which extents to clone above,
2650 * but shouldn't round up the file size
2651 */
2652 endoff = new_key.offset + datal;
2653 if (endoff > destoff+olen)
2654 endoff = destoff+olen;
2655 if (endoff > inode->i_size)
2656 btrfs_i_size_write(inode, endoff);
2657
2658 ret = btrfs_update_inode(trans, root, inode);
2659 if (ret) {
2660 btrfs_abort_transaction(trans, root, ret);
2661 btrfs_end_transaction(trans, root);
2662 goto out;
2663 }
2664 ret = btrfs_end_transaction(trans, root);
2665 }
2666next:
2667 btrfs_release_path(path);
2668 key.offset++;
2669 }
2670 ret = 0;
2671out:
2672 btrfs_release_path(path);
2673 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2674out_unlock:
2675 mutex_unlock(&src->i_mutex);
2676 mutex_unlock(&inode->i_mutex);
2677 vfree(buf);
2678 btrfs_free_path(path);
2679out_fput:
2680 fput(src_file);
2681out_drop_write:
2682 mnt_drop_write_file(file);
2683 return ret;
2684}
2685
2686static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2687{
2688 struct btrfs_ioctl_clone_range_args args;
2689
2690 if (copy_from_user(&args, argp, sizeof(args)))
2691 return -EFAULT;
2692 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2693 args.src_length, args.dest_offset);
2694}
2695
2696/*
2697 * there are many ways the trans_start and trans_end ioctls can lead
2698 * to deadlocks. They should only be used by applications that
2699 * basically own the machine, and have a very in depth understanding
2700 * of all the possible deadlocks and enospc problems.
2701 */
2702static long btrfs_ioctl_trans_start(struct file *file)
2703{
2704 struct inode *inode = fdentry(file)->d_inode;
2705 struct btrfs_root *root = BTRFS_I(inode)->root;
2706 struct btrfs_trans_handle *trans;
2707 int ret;
2708
2709 ret = -EPERM;
2710 if (!capable(CAP_SYS_ADMIN))
2711 goto out;
2712
2713 ret = -EINPROGRESS;
2714 if (file->private_data)
2715 goto out;
2716
2717 ret = -EROFS;
2718 if (btrfs_root_readonly(root))
2719 goto out;
2720
2721 ret = mnt_want_write_file(file);
2722 if (ret)
2723 goto out;
2724
2725 atomic_inc(&root->fs_info->open_ioctl_trans);
2726
2727 ret = -ENOMEM;
2728 trans = btrfs_start_ioctl_transaction(root);
2729 if (IS_ERR(trans))
2730 goto out_drop;
2731
2732 file->private_data = trans;
2733 return 0;
2734
2735out_drop:
2736 atomic_dec(&root->fs_info->open_ioctl_trans);
2737 mnt_drop_write_file(file);
2738out:
2739 return ret;
2740}
2741
2742static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2743{
2744 struct inode *inode = fdentry(file)->d_inode;
2745 struct btrfs_root *root = BTRFS_I(inode)->root;
2746 struct btrfs_root *new_root;
2747 struct btrfs_dir_item *di;
2748 struct btrfs_trans_handle *trans;
2749 struct btrfs_path *path;
2750 struct btrfs_key location;
2751 struct btrfs_disk_key disk_key;
2752 struct btrfs_super_block *disk_super;
2753 u64 features;
2754 u64 objectid = 0;
2755 u64 dir_id;
2756
2757 if (!capable(CAP_SYS_ADMIN))
2758 return -EPERM;
2759
2760 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2761 return -EFAULT;
2762
2763 if (!objectid)
2764 objectid = root->root_key.objectid;
2765
2766 location.objectid = objectid;
2767 location.type = BTRFS_ROOT_ITEM_KEY;
2768 location.offset = (u64)-1;
2769
2770 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2771 if (IS_ERR(new_root))
2772 return PTR_ERR(new_root);
2773
2774 if (btrfs_root_refs(&new_root->root_item) == 0)
2775 return -ENOENT;
2776
2777 path = btrfs_alloc_path();
2778 if (!path)
2779 return -ENOMEM;
2780 path->leave_spinning = 1;
2781
2782 trans = btrfs_start_transaction(root, 1);
2783 if (IS_ERR(trans)) {
2784 btrfs_free_path(path);
2785 return PTR_ERR(trans);
2786 }
2787
2788 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2789 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2790 dir_id, "default", 7, 1);
2791 if (IS_ERR_OR_NULL(di)) {
2792 btrfs_free_path(path);
2793 btrfs_end_transaction(trans, root);
2794 printk(KERN_ERR "Umm, you don't have the default dir item, "
2795 "this isn't going to work\n");
2796 return -ENOENT;
2797 }
2798
2799 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2800 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2801 btrfs_mark_buffer_dirty(path->nodes[0]);
2802 btrfs_free_path(path);
2803
2804 disk_super = root->fs_info->super_copy;
2805 features = btrfs_super_incompat_flags(disk_super);
2806 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2807 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2808 btrfs_set_super_incompat_flags(disk_super, features);
2809 }
2810 btrfs_end_transaction(trans, root);
2811
2812 return 0;
2813}
2814
2815static void get_block_group_info(struct list_head *groups_list,
2816 struct btrfs_ioctl_space_info *space)
2817{
2818 struct btrfs_block_group_cache *block_group;
2819
2820 space->total_bytes = 0;
2821 space->used_bytes = 0;
2822 space->flags = 0;
2823 list_for_each_entry(block_group, groups_list, list) {
2824 space->flags = block_group->flags;
2825 space->total_bytes += block_group->key.offset;
2826 space->used_bytes +=
2827 btrfs_block_group_used(&block_group->item);
2828 }
2829}
2830
2831long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2832{
2833 struct btrfs_ioctl_space_args space_args;
2834 struct btrfs_ioctl_space_info space;
2835 struct btrfs_ioctl_space_info *dest;
2836 struct btrfs_ioctl_space_info *dest_orig;
2837 struct btrfs_ioctl_space_info __user *user_dest;
2838 struct btrfs_space_info *info;
2839 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2840 BTRFS_BLOCK_GROUP_SYSTEM,
2841 BTRFS_BLOCK_GROUP_METADATA,
2842 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2843 int num_types = 4;
2844 int alloc_size;
2845 int ret = 0;
2846 u64 slot_count = 0;
2847 int i, c;
2848
2849 if (copy_from_user(&space_args,
2850 (struct btrfs_ioctl_space_args __user *)arg,
2851 sizeof(space_args)))
2852 return -EFAULT;
2853
2854 for (i = 0; i < num_types; i++) {
2855 struct btrfs_space_info *tmp;
2856
2857 info = NULL;
2858 rcu_read_lock();
2859 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2860 list) {
2861 if (tmp->flags == types[i]) {
2862 info = tmp;
2863 break;
2864 }
2865 }
2866 rcu_read_unlock();
2867
2868 if (!info)
2869 continue;
2870
2871 down_read(&info->groups_sem);
2872 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2873 if (!list_empty(&info->block_groups[c]))
2874 slot_count++;
2875 }
2876 up_read(&info->groups_sem);
2877 }
2878
2879 /* space_slots == 0 means they are asking for a count */
2880 if (space_args.space_slots == 0) {
2881 space_args.total_spaces = slot_count;
2882 goto out;
2883 }
2884
2885 slot_count = min_t(u64, space_args.space_slots, slot_count);
2886
2887 alloc_size = sizeof(*dest) * slot_count;
2888
2889 /* we generally have at most 6 or so space infos, one for each raid
2890 * level. So, a whole page should be more than enough for everyone
2891 */
2892 if (alloc_size > PAGE_CACHE_SIZE)
2893 return -ENOMEM;
2894
2895 space_args.total_spaces = 0;
2896 dest = kmalloc(alloc_size, GFP_NOFS);
2897 if (!dest)
2898 return -ENOMEM;
2899 dest_orig = dest;
2900
2901 /* now we have a buffer to copy into */
2902 for (i = 0; i < num_types; i++) {
2903 struct btrfs_space_info *tmp;
2904
2905 if (!slot_count)
2906 break;
2907
2908 info = NULL;
2909 rcu_read_lock();
2910 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2911 list) {
2912 if (tmp->flags == types[i]) {
2913 info = tmp;
2914 break;
2915 }
2916 }
2917 rcu_read_unlock();
2918
2919 if (!info)
2920 continue;
2921 down_read(&info->groups_sem);
2922 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2923 if (!list_empty(&info->block_groups[c])) {
2924 get_block_group_info(&info->block_groups[c],
2925 &space);
2926 memcpy(dest, &space, sizeof(space));
2927 dest++;
2928 space_args.total_spaces++;
2929 slot_count--;
2930 }
2931 if (!slot_count)
2932 break;
2933 }
2934 up_read(&info->groups_sem);
2935 }
2936
2937 user_dest = (struct btrfs_ioctl_space_info __user *)
2938 (arg + sizeof(struct btrfs_ioctl_space_args));
2939
2940 if (copy_to_user(user_dest, dest_orig, alloc_size))
2941 ret = -EFAULT;
2942
2943 kfree(dest_orig);
2944out:
2945 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2946 ret = -EFAULT;
2947
2948 return ret;
2949}
2950
2951/*
2952 * there are many ways the trans_start and trans_end ioctls can lead
2953 * to deadlocks. They should only be used by applications that
2954 * basically own the machine, and have a very in depth understanding
2955 * of all the possible deadlocks and enospc problems.
2956 */
2957long btrfs_ioctl_trans_end(struct file *file)
2958{
2959 struct inode *inode = fdentry(file)->d_inode;
2960 struct btrfs_root *root = BTRFS_I(inode)->root;
2961 struct btrfs_trans_handle *trans;
2962
2963 trans = file->private_data;
2964 if (!trans)
2965 return -EINVAL;
2966 file->private_data = NULL;
2967
2968 btrfs_end_transaction(trans, root);
2969
2970 atomic_dec(&root->fs_info->open_ioctl_trans);
2971
2972 mnt_drop_write_file(file);
2973 return 0;
2974}
2975
2976static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2977{
2978 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2979 struct btrfs_trans_handle *trans;
2980 u64 transid;
2981 int ret;
2982
2983 trans = btrfs_start_transaction(root, 0);
2984 if (IS_ERR(trans))
2985 return PTR_ERR(trans);
2986 transid = trans->transid;
2987 ret = btrfs_commit_transaction_async(trans, root, 0);
2988 if (ret) {
2989 btrfs_end_transaction(trans, root);
2990 return ret;
2991 }
2992
2993 if (argp)
2994 if (copy_to_user(argp, &transid, sizeof(transid)))
2995 return -EFAULT;
2996 return 0;
2997}
2998
2999static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
3000{
3001 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3002 u64 transid;
3003
3004 if (argp) {
3005 if (copy_from_user(&transid, argp, sizeof(transid)))
3006 return -EFAULT;
3007 } else {
3008 transid = 0; /* current trans */
3009 }
3010 return btrfs_wait_for_commit(root, transid);
3011}
3012
3013static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
3014{
3015 int ret;
3016 struct btrfs_ioctl_scrub_args *sa;
3017
3018 if (!capable(CAP_SYS_ADMIN))
3019 return -EPERM;
3020
3021 sa = memdup_user(arg, sizeof(*sa));
3022 if (IS_ERR(sa))
3023 return PTR_ERR(sa);
3024
3025 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
3026 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3027
3028 if (copy_to_user(arg, sa, sizeof(*sa)))
3029 ret = -EFAULT;
3030
3031 kfree(sa);
3032 return ret;
3033}
3034
3035static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3036{
3037 if (!capable(CAP_SYS_ADMIN))
3038 return -EPERM;
3039
3040 return btrfs_scrub_cancel(root);
3041}
3042
3043static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3044 void __user *arg)
3045{
3046 struct btrfs_ioctl_scrub_args *sa;
3047 int ret;
3048
3049 if (!capable(CAP_SYS_ADMIN))
3050 return -EPERM;
3051
3052 sa = memdup_user(arg, sizeof(*sa));
3053 if (IS_ERR(sa))
3054 return PTR_ERR(sa);
3055
3056 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3057
3058 if (copy_to_user(arg, sa, sizeof(*sa)))
3059 ret = -EFAULT;
3060
3061 kfree(sa);
3062 return ret;
3063}
3064
3065static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3066 void __user *arg, int reset_after_read)
3067{
3068 struct btrfs_ioctl_get_dev_stats *sa;
3069 int ret;
3070
3071 if (reset_after_read && !capable(CAP_SYS_ADMIN))
3072 return -EPERM;
3073
3074 sa = memdup_user(arg, sizeof(*sa));
3075 if (IS_ERR(sa))
3076 return PTR_ERR(sa);
3077
3078 ret = btrfs_get_dev_stats(root, sa, reset_after_read);
3079
3080 if (copy_to_user(arg, sa, sizeof(*sa)))
3081 ret = -EFAULT;
3082
3083 kfree(sa);
3084 return ret;
3085}
3086
3087static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3088{
3089 int ret = 0;
3090 int i;
3091 u64 rel_ptr;
3092 int size;
3093 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3094 struct inode_fs_paths *ipath = NULL;
3095 struct btrfs_path *path;
3096
3097 if (!capable(CAP_SYS_ADMIN))
3098 return -EPERM;
3099
3100 path = btrfs_alloc_path();
3101 if (!path) {
3102 ret = -ENOMEM;
3103 goto out;
3104 }
3105
3106 ipa = memdup_user(arg, sizeof(*ipa));
3107 if (IS_ERR(ipa)) {
3108 ret = PTR_ERR(ipa);
3109 ipa = NULL;
3110 goto out;
3111 }
3112
3113 size = min_t(u32, ipa->size, 4096);
3114 ipath = init_ipath(size, root, path);
3115 if (IS_ERR(ipath)) {
3116 ret = PTR_ERR(ipath);
3117 ipath = NULL;
3118 goto out;
3119 }
3120
3121 ret = paths_from_inode(ipa->inum, ipath);
3122 if (ret < 0)
3123 goto out;
3124
3125 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3126 rel_ptr = ipath->fspath->val[i] -
3127 (u64)(unsigned long)ipath->fspath->val;
3128 ipath->fspath->val[i] = rel_ptr;
3129 }
3130
3131 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3132 (void *)(unsigned long)ipath->fspath, size);
3133 if (ret) {
3134 ret = -EFAULT;
3135 goto out;
3136 }
3137
3138out:
3139 btrfs_free_path(path);
3140 free_ipath(ipath);
3141 kfree(ipa);
3142
3143 return ret;
3144}
3145
3146static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3147{
3148 struct btrfs_data_container *inodes = ctx;
3149 const size_t c = 3 * sizeof(u64);
3150
3151 if (inodes->bytes_left >= c) {
3152 inodes->bytes_left -= c;
3153 inodes->val[inodes->elem_cnt] = inum;
3154 inodes->val[inodes->elem_cnt + 1] = offset;
3155 inodes->val[inodes->elem_cnt + 2] = root;
3156 inodes->elem_cnt += 3;
3157 } else {
3158 inodes->bytes_missing += c - inodes->bytes_left;
3159 inodes->bytes_left = 0;
3160 inodes->elem_missed += 3;
3161 }
3162
3163 return 0;
3164}
3165
3166static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3167 void __user *arg)
3168{
3169 int ret = 0;
3170 int size;
3171 u64 extent_item_pos;
3172 struct btrfs_ioctl_logical_ino_args *loi;
3173 struct btrfs_data_container *inodes = NULL;
3174 struct btrfs_path *path = NULL;
3175 struct btrfs_key key;
3176
3177 if (!capable(CAP_SYS_ADMIN))
3178 return -EPERM;
3179
3180 loi = memdup_user(arg, sizeof(*loi));
3181 if (IS_ERR(loi)) {
3182 ret = PTR_ERR(loi);
3183 loi = NULL;
3184 goto out;
3185 }
3186
3187 path = btrfs_alloc_path();
3188 if (!path) {
3189 ret = -ENOMEM;
3190 goto out;
3191 }
3192
3193 size = min_t(u32, loi->size, 4096);
3194 inodes = init_data_container(size);
3195 if (IS_ERR(inodes)) {
3196 ret = PTR_ERR(inodes);
3197 inodes = NULL;
3198 goto out;
3199 }
3200
3201 ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3202 btrfs_release_path(path);
3203
3204 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3205 ret = -ENOENT;
3206 if (ret < 0)
3207 goto out;
3208
3209 extent_item_pos = loi->logical - key.objectid;
3210 ret = iterate_extent_inodes(root->fs_info, key.objectid,
3211 extent_item_pos, 0, build_ino_list,
3212 inodes);
3213
3214 if (ret < 0)
3215 goto out;
3216
3217 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3218 (void *)(unsigned long)inodes, size);
3219 if (ret)
3220 ret = -EFAULT;
3221
3222out:
3223 btrfs_free_path(path);
3224 kfree(inodes);
3225 kfree(loi);
3226
3227 return ret;
3228}
3229
3230void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3231 struct btrfs_ioctl_balance_args *bargs)
3232{
3233 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3234
3235 bargs->flags = bctl->flags;
3236
3237 if (atomic_read(&fs_info->balance_running))
3238 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3239 if (atomic_read(&fs_info->balance_pause_req))
3240 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3241 if (atomic_read(&fs_info->balance_cancel_req))
3242 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3243
3244 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3245 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3246 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3247
3248 if (lock) {
3249 spin_lock(&fs_info->balance_lock);
3250 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3251 spin_unlock(&fs_info->balance_lock);
3252 } else {
3253 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3254 }
3255}
3256
3257static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3258{
3259 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3260 struct btrfs_fs_info *fs_info = root->fs_info;
3261 struct btrfs_ioctl_balance_args *bargs;
3262 struct btrfs_balance_control *bctl;
3263 int ret;
3264
3265 if (!capable(CAP_SYS_ADMIN))
3266 return -EPERM;
3267
3268 if (fs_info->sb->s_flags & MS_RDONLY)
3269 return -EROFS;
3270
3271 ret = mnt_want_write(file->f_path.mnt);
3272 if (ret)
3273 return ret;
3274
3275 mutex_lock(&fs_info->volume_mutex);
3276 mutex_lock(&fs_info->balance_mutex);
3277
3278 if (arg) {
3279 bargs = memdup_user(arg, sizeof(*bargs));
3280 if (IS_ERR(bargs)) {
3281 ret = PTR_ERR(bargs);
3282 goto out;
3283 }
3284
3285 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3286 if (!fs_info->balance_ctl) {
3287 ret = -ENOTCONN;
3288 goto out_bargs;
3289 }
3290
3291 bctl = fs_info->balance_ctl;
3292 spin_lock(&fs_info->balance_lock);
3293 bctl->flags |= BTRFS_BALANCE_RESUME;
3294 spin_unlock(&fs_info->balance_lock);
3295
3296 goto do_balance;
3297 }
3298 } else {
3299 bargs = NULL;
3300 }
3301
3302 if (fs_info->balance_ctl) {
3303 ret = -EINPROGRESS;
3304 goto out_bargs;
3305 }
3306
3307 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3308 if (!bctl) {
3309 ret = -ENOMEM;
3310 goto out_bargs;
3311 }
3312
3313 bctl->fs_info = fs_info;
3314 if (arg) {
3315 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3316 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3317 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3318
3319 bctl->flags = bargs->flags;
3320 } else {
3321 /* balance everything - no filters */
3322 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3323 }
3324
3325do_balance:
3326 ret = btrfs_balance(bctl, bargs);
3327 /*
3328 * bctl is freed in __cancel_balance or in free_fs_info if
3329 * restriper was paused all the way until unmount
3330 */
3331 if (arg) {
3332 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3333 ret = -EFAULT;
3334 }
3335
3336out_bargs:
3337 kfree(bargs);
3338out:
3339 mutex_unlock(&fs_info->balance_mutex);
3340 mutex_unlock(&fs_info->volume_mutex);
3341 mnt_drop_write(file->f_path.mnt);
3342 return ret;
3343}
3344
3345static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3346{
3347 if (!capable(CAP_SYS_ADMIN))
3348 return -EPERM;
3349
3350 switch (cmd) {
3351 case BTRFS_BALANCE_CTL_PAUSE:
3352 return btrfs_pause_balance(root->fs_info);
3353 case BTRFS_BALANCE_CTL_CANCEL:
3354 return btrfs_cancel_balance(root->fs_info);
3355 }
3356
3357 return -EINVAL;
3358}
3359
3360static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3361 void __user *arg)
3362{
3363 struct btrfs_fs_info *fs_info = root->fs_info;
3364 struct btrfs_ioctl_balance_args *bargs;
3365 int ret = 0;
3366
3367 if (!capable(CAP_SYS_ADMIN))
3368 return -EPERM;
3369
3370 mutex_lock(&fs_info->balance_mutex);
3371 if (!fs_info->balance_ctl) {
3372 ret = -ENOTCONN;
3373 goto out;
3374 }
3375
3376 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3377 if (!bargs) {
3378 ret = -ENOMEM;
3379 goto out;
3380 }
3381
3382 update_ioctl_balance_args(fs_info, 1, bargs);
3383
3384 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3385 ret = -EFAULT;
3386
3387 kfree(bargs);
3388out:
3389 mutex_unlock(&fs_info->balance_mutex);
3390 return ret;
3391}
3392
3393long btrfs_ioctl(struct file *file, unsigned int
3394 cmd, unsigned long arg)
3395{
3396 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3397 void __user *argp = (void __user *)arg;
3398
3399 switch (cmd) {
3400 case FS_IOC_GETFLAGS:
3401 return btrfs_ioctl_getflags(file, argp);
3402 case FS_IOC_SETFLAGS:
3403 return btrfs_ioctl_setflags(file, argp);
3404 case FS_IOC_GETVERSION:
3405 return btrfs_ioctl_getversion(file, argp);
3406 case FITRIM:
3407 return btrfs_ioctl_fitrim(file, argp);
3408 case BTRFS_IOC_SNAP_CREATE:
3409 return btrfs_ioctl_snap_create(file, argp, 0);
3410 case BTRFS_IOC_SNAP_CREATE_V2:
3411 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3412 case BTRFS_IOC_SUBVOL_CREATE:
3413 return btrfs_ioctl_snap_create(file, argp, 1);
3414 case BTRFS_IOC_SNAP_DESTROY:
3415 return btrfs_ioctl_snap_destroy(file, argp);
3416 case BTRFS_IOC_SUBVOL_GETFLAGS:
3417 return btrfs_ioctl_subvol_getflags(file, argp);
3418 case BTRFS_IOC_SUBVOL_SETFLAGS:
3419 return btrfs_ioctl_subvol_setflags(file, argp);
3420 case BTRFS_IOC_DEFAULT_SUBVOL:
3421 return btrfs_ioctl_default_subvol(file, argp);
3422 case BTRFS_IOC_DEFRAG:
3423 return btrfs_ioctl_defrag(file, NULL);
3424 case BTRFS_IOC_DEFRAG_RANGE:
3425 return btrfs_ioctl_defrag(file, argp);
3426 case BTRFS_IOC_RESIZE:
3427 return btrfs_ioctl_resize(root, argp);
3428 case BTRFS_IOC_ADD_DEV:
3429 return btrfs_ioctl_add_dev(root, argp);
3430 case BTRFS_IOC_RM_DEV:
3431 return btrfs_ioctl_rm_dev(root, argp);
3432 case BTRFS_IOC_FS_INFO:
3433 return btrfs_ioctl_fs_info(root, argp);
3434 case BTRFS_IOC_DEV_INFO:
3435 return btrfs_ioctl_dev_info(root, argp);
3436 case BTRFS_IOC_BALANCE:
3437 return btrfs_ioctl_balance(file, NULL);
3438 case BTRFS_IOC_CLONE:
3439 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3440 case BTRFS_IOC_CLONE_RANGE:
3441 return btrfs_ioctl_clone_range(file, argp);
3442 case BTRFS_IOC_TRANS_START:
3443 return btrfs_ioctl_trans_start(file);
3444 case BTRFS_IOC_TRANS_END:
3445 return btrfs_ioctl_trans_end(file);
3446 case BTRFS_IOC_TREE_SEARCH:
3447 return btrfs_ioctl_tree_search(file, argp);
3448 case BTRFS_IOC_INO_LOOKUP:
3449 return btrfs_ioctl_ino_lookup(file, argp);
3450 case BTRFS_IOC_INO_PATHS:
3451 return btrfs_ioctl_ino_to_path(root, argp);
3452 case BTRFS_IOC_LOGICAL_INO:
3453 return btrfs_ioctl_logical_to_ino(root, argp);
3454 case BTRFS_IOC_SPACE_INFO:
3455 return btrfs_ioctl_space_info(root, argp);
3456 case BTRFS_IOC_SYNC:
3457 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3458 return 0;
3459 case BTRFS_IOC_START_SYNC:
3460 return btrfs_ioctl_start_sync(file, argp);
3461 case BTRFS_IOC_WAIT_SYNC:
3462 return btrfs_ioctl_wait_sync(file, argp);
3463 case BTRFS_IOC_SCRUB:
3464 return btrfs_ioctl_scrub(root, argp);
3465 case BTRFS_IOC_SCRUB_CANCEL:
3466 return btrfs_ioctl_scrub_cancel(root, argp);
3467 case BTRFS_IOC_SCRUB_PROGRESS:
3468 return btrfs_ioctl_scrub_progress(root, argp);
3469 case BTRFS_IOC_BALANCE_V2:
3470 return btrfs_ioctl_balance(file, argp);
3471 case BTRFS_IOC_BALANCE_CTL:
3472 return btrfs_ioctl_balance_ctl(root, arg);
3473 case BTRFS_IOC_BALANCE_PROGRESS:
3474 return btrfs_ioctl_balance_progress(root, argp);
3475 case BTRFS_IOC_GET_DEV_STATS:
3476 return btrfs_ioctl_get_dev_stats(root, argp, 0);
3477 case BTRFS_IOC_GET_AND_RESET_DEV_STATS:
3478 return btrfs_ioctl_get_dev_stats(root, argp, 1);
3479 }
3480
3481 return -ENOTTY;
3482}
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (C) 2007 Oracle. All rights reserved.
4 */
5
6#include <linux/kernel.h>
7#include <linux/bio.h>
8#include <linux/file.h>
9#include <linux/fs.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>
22#include <linux/mm.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>
32#include "ctree.h"
33#include "disk-io.h"
34#include "export.h"
35#include "transaction.h"
36#include "btrfs_inode.h"
37#include "print-tree.h"
38#include "volumes.h"
39#include "locking.h"
40#include "backref.h"
41#include "rcu-string.h"
42#include "send.h"
43#include "dev-replace.h"
44#include "props.h"
45#include "sysfs.h"
46#include "qgroup.h"
47#include "tree-log.h"
48#include "compression.h"
49#include "space-info.h"
50#include "delalloc-space.h"
51#include "block-group.h"
52#include "subpage.h"
53#include "fs.h"
54#include "accessors.h"
55#include "extent-tree.h"
56#include "root-tree.h"
57#include "defrag.h"
58#include "dir-item.h"
59#include "uuid-tree.h"
60#include "ioctl.h"
61#include "file.h"
62#include "scrub.h"
63#include "super.h"
64
65#ifdef CONFIG_64BIT
66/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
67 * structures are incorrect, as the timespec structure from userspace
68 * is 4 bytes too small. We define these alternatives here to teach
69 * the kernel about the 32-bit struct packing.
70 */
71struct btrfs_ioctl_timespec_32 {
72 __u64 sec;
73 __u32 nsec;
74} __attribute__ ((__packed__));
75
76struct btrfs_ioctl_received_subvol_args_32 {
77 char uuid[BTRFS_UUID_SIZE]; /* in */
78 __u64 stransid; /* in */
79 __u64 rtransid; /* out */
80 struct btrfs_ioctl_timespec_32 stime; /* in */
81 struct btrfs_ioctl_timespec_32 rtime; /* out */
82 __u64 flags; /* in */
83 __u64 reserved[16]; /* in */
84} __attribute__ ((__packed__));
85
86#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
87 struct btrfs_ioctl_received_subvol_args_32)
88#endif
89
90#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
91struct btrfs_ioctl_send_args_32 {
92 __s64 send_fd; /* in */
93 __u64 clone_sources_count; /* in */
94 compat_uptr_t clone_sources; /* in */
95 __u64 parent_root; /* in */
96 __u64 flags; /* in */
97 __u32 version; /* in */
98 __u8 reserved[28]; /* in */
99} __attribute__ ((__packed__));
100
101#define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
102 struct btrfs_ioctl_send_args_32)
103
104struct btrfs_ioctl_encoded_io_args_32 {
105 compat_uptr_t iov;
106 compat_ulong_t iovcnt;
107 __s64 offset;
108 __u64 flags;
109 __u64 len;
110 __u64 unencoded_len;
111 __u64 unencoded_offset;
112 __u32 compression;
113 __u32 encryption;
114 __u8 reserved[64];
115};
116
117#define BTRFS_IOC_ENCODED_READ_32 _IOR(BTRFS_IOCTL_MAGIC, 64, \
118 struct btrfs_ioctl_encoded_io_args_32)
119#define BTRFS_IOC_ENCODED_WRITE_32 _IOW(BTRFS_IOCTL_MAGIC, 64, \
120 struct btrfs_ioctl_encoded_io_args_32)
121#endif
122
123/* Mask out flags that are inappropriate for the given type of inode. */
124static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
125 unsigned int flags)
126{
127 if (S_ISDIR(inode->i_mode))
128 return flags;
129 else if (S_ISREG(inode->i_mode))
130 return flags & ~FS_DIRSYNC_FL;
131 else
132 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
133}
134
135/*
136 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
137 * ioctl.
138 */
139static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode)
140{
141 unsigned int iflags = 0;
142 u32 flags = binode->flags;
143 u32 ro_flags = binode->ro_flags;
144
145 if (flags & BTRFS_INODE_SYNC)
146 iflags |= FS_SYNC_FL;
147 if (flags & BTRFS_INODE_IMMUTABLE)
148 iflags |= FS_IMMUTABLE_FL;
149 if (flags & BTRFS_INODE_APPEND)
150 iflags |= FS_APPEND_FL;
151 if (flags & BTRFS_INODE_NODUMP)
152 iflags |= FS_NODUMP_FL;
153 if (flags & BTRFS_INODE_NOATIME)
154 iflags |= FS_NOATIME_FL;
155 if (flags & BTRFS_INODE_DIRSYNC)
156 iflags |= FS_DIRSYNC_FL;
157 if (flags & BTRFS_INODE_NODATACOW)
158 iflags |= FS_NOCOW_FL;
159 if (ro_flags & BTRFS_INODE_RO_VERITY)
160 iflags |= FS_VERITY_FL;
161
162 if (flags & BTRFS_INODE_NOCOMPRESS)
163 iflags |= FS_NOCOMP_FL;
164 else if (flags & BTRFS_INODE_COMPRESS)
165 iflags |= FS_COMPR_FL;
166
167 return iflags;
168}
169
170/*
171 * Update inode->i_flags based on the btrfs internal flags.
172 */
173void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
174{
175 struct btrfs_inode *binode = BTRFS_I(inode);
176 unsigned int new_fl = 0;
177
178 if (binode->flags & BTRFS_INODE_SYNC)
179 new_fl |= S_SYNC;
180 if (binode->flags & BTRFS_INODE_IMMUTABLE)
181 new_fl |= S_IMMUTABLE;
182 if (binode->flags & BTRFS_INODE_APPEND)
183 new_fl |= S_APPEND;
184 if (binode->flags & BTRFS_INODE_NOATIME)
185 new_fl |= S_NOATIME;
186 if (binode->flags & BTRFS_INODE_DIRSYNC)
187 new_fl |= S_DIRSYNC;
188 if (binode->ro_flags & BTRFS_INODE_RO_VERITY)
189 new_fl |= S_VERITY;
190
191 set_mask_bits(&inode->i_flags,
192 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC |
193 S_VERITY, new_fl);
194}
195
196/*
197 * Check if @flags are a supported and valid set of FS_*_FL flags and that
198 * the old and new flags are not conflicting
199 */
200static int check_fsflags(unsigned int old_flags, unsigned int flags)
201{
202 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
203 FS_NOATIME_FL | FS_NODUMP_FL | \
204 FS_SYNC_FL | FS_DIRSYNC_FL | \
205 FS_NOCOMP_FL | FS_COMPR_FL |
206 FS_NOCOW_FL))
207 return -EOPNOTSUPP;
208
209 /* COMPR and NOCOMP on new/old are valid */
210 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
211 return -EINVAL;
212
213 if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
214 return -EINVAL;
215
216 /* NOCOW and compression options are mutually exclusive */
217 if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
218 return -EINVAL;
219 if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
220 return -EINVAL;
221
222 return 0;
223}
224
225static int check_fsflags_compatible(struct btrfs_fs_info *fs_info,
226 unsigned int flags)
227{
228 if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL))
229 return -EPERM;
230
231 return 0;
232}
233
234/*
235 * Set flags/xflags from the internal inode flags. The remaining items of
236 * fsxattr are zeroed.
237 */
238int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
239{
240 struct btrfs_inode *binode = BTRFS_I(d_inode(dentry));
241
242 fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode));
243 return 0;
244}
245
246int btrfs_fileattr_set(struct mnt_idmap *idmap,
247 struct dentry *dentry, struct fileattr *fa)
248{
249 struct inode *inode = d_inode(dentry);
250 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
251 struct btrfs_inode *binode = BTRFS_I(inode);
252 struct btrfs_root *root = binode->root;
253 struct btrfs_trans_handle *trans;
254 unsigned int fsflags, old_fsflags;
255 int ret;
256 const char *comp = NULL;
257 u32 binode_flags;
258
259 if (btrfs_root_readonly(root))
260 return -EROFS;
261
262 if (fileattr_has_fsx(fa))
263 return -EOPNOTSUPP;
264
265 fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags);
266 old_fsflags = btrfs_inode_flags_to_fsflags(binode);
267 ret = check_fsflags(old_fsflags, fsflags);
268 if (ret)
269 return ret;
270
271 ret = check_fsflags_compatible(fs_info, fsflags);
272 if (ret)
273 return ret;
274
275 binode_flags = binode->flags;
276 if (fsflags & FS_SYNC_FL)
277 binode_flags |= BTRFS_INODE_SYNC;
278 else
279 binode_flags &= ~BTRFS_INODE_SYNC;
280 if (fsflags & FS_IMMUTABLE_FL)
281 binode_flags |= BTRFS_INODE_IMMUTABLE;
282 else
283 binode_flags &= ~BTRFS_INODE_IMMUTABLE;
284 if (fsflags & FS_APPEND_FL)
285 binode_flags |= BTRFS_INODE_APPEND;
286 else
287 binode_flags &= ~BTRFS_INODE_APPEND;
288 if (fsflags & FS_NODUMP_FL)
289 binode_flags |= BTRFS_INODE_NODUMP;
290 else
291 binode_flags &= ~BTRFS_INODE_NODUMP;
292 if (fsflags & FS_NOATIME_FL)
293 binode_flags |= BTRFS_INODE_NOATIME;
294 else
295 binode_flags &= ~BTRFS_INODE_NOATIME;
296
297 /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */
298 if (!fa->flags_valid) {
299 /* 1 item for the inode */
300 trans = btrfs_start_transaction(root, 1);
301 if (IS_ERR(trans))
302 return PTR_ERR(trans);
303 goto update_flags;
304 }
305
306 if (fsflags & FS_DIRSYNC_FL)
307 binode_flags |= BTRFS_INODE_DIRSYNC;
308 else
309 binode_flags &= ~BTRFS_INODE_DIRSYNC;
310 if (fsflags & FS_NOCOW_FL) {
311 if (S_ISREG(inode->i_mode)) {
312 /*
313 * It's safe to turn csums off here, no extents exist.
314 * Otherwise we want the flag to reflect the real COW
315 * status of the file and will not set it.
316 */
317 if (inode->i_size == 0)
318 binode_flags |= BTRFS_INODE_NODATACOW |
319 BTRFS_INODE_NODATASUM;
320 } else {
321 binode_flags |= BTRFS_INODE_NODATACOW;
322 }
323 } else {
324 /*
325 * Revert back under same assumptions as above
326 */
327 if (S_ISREG(inode->i_mode)) {
328 if (inode->i_size == 0)
329 binode_flags &= ~(BTRFS_INODE_NODATACOW |
330 BTRFS_INODE_NODATASUM);
331 } else {
332 binode_flags &= ~BTRFS_INODE_NODATACOW;
333 }
334 }
335
336 /*
337 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
338 * flag may be changed automatically if compression code won't make
339 * things smaller.
340 */
341 if (fsflags & FS_NOCOMP_FL) {
342 binode_flags &= ~BTRFS_INODE_COMPRESS;
343 binode_flags |= BTRFS_INODE_NOCOMPRESS;
344 } else if (fsflags & FS_COMPR_FL) {
345
346 if (IS_SWAPFILE(inode))
347 return -ETXTBSY;
348
349 binode_flags |= BTRFS_INODE_COMPRESS;
350 binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
351
352 comp = btrfs_compress_type2str(fs_info->compress_type);
353 if (!comp || comp[0] == 0)
354 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
355 } else {
356 binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
357 }
358
359 /*
360 * 1 for inode item
361 * 2 for properties
362 */
363 trans = btrfs_start_transaction(root, 3);
364 if (IS_ERR(trans))
365 return PTR_ERR(trans);
366
367 if (comp) {
368 ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
369 strlen(comp), 0);
370 if (ret) {
371 btrfs_abort_transaction(trans, ret);
372 goto out_end_trans;
373 }
374 } else {
375 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
376 0, 0);
377 if (ret && ret != -ENODATA) {
378 btrfs_abort_transaction(trans, ret);
379 goto out_end_trans;
380 }
381 }
382
383update_flags:
384 binode->flags = binode_flags;
385 btrfs_sync_inode_flags_to_i_flags(inode);
386 inode_inc_iversion(inode);
387 inode_set_ctime_current(inode);
388 ret = btrfs_update_inode(trans, BTRFS_I(inode));
389
390 out_end_trans:
391 btrfs_end_transaction(trans);
392 return ret;
393}
394
395/*
396 * Start exclusive operation @type, return true on success
397 */
398bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
399 enum btrfs_exclusive_operation type)
400{
401 bool ret = false;
402
403 spin_lock(&fs_info->super_lock);
404 if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) {
405 fs_info->exclusive_operation = type;
406 ret = true;
407 }
408 spin_unlock(&fs_info->super_lock);
409
410 return ret;
411}
412
413/*
414 * Conditionally allow to enter the exclusive operation in case it's compatible
415 * with the running one. This must be paired with btrfs_exclop_start_unlock and
416 * btrfs_exclop_finish.
417 *
418 * Compatibility:
419 * - the same type is already running
420 * - when trying to add a device and balance has been paused
421 * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller
422 * must check the condition first that would allow none -> @type
423 */
424bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
425 enum btrfs_exclusive_operation type)
426{
427 spin_lock(&fs_info->super_lock);
428 if (fs_info->exclusive_operation == type ||
429 (fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED &&
430 type == BTRFS_EXCLOP_DEV_ADD))
431 return true;
432
433 spin_unlock(&fs_info->super_lock);
434 return false;
435}
436
437void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info)
438{
439 spin_unlock(&fs_info->super_lock);
440}
441
442void btrfs_exclop_finish(struct btrfs_fs_info *fs_info)
443{
444 spin_lock(&fs_info->super_lock);
445 WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE);
446 spin_unlock(&fs_info->super_lock);
447 sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation");
448}
449
450void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
451 enum btrfs_exclusive_operation op)
452{
453 switch (op) {
454 case BTRFS_EXCLOP_BALANCE_PAUSED:
455 spin_lock(&fs_info->super_lock);
456 ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE ||
457 fs_info->exclusive_operation == BTRFS_EXCLOP_DEV_ADD ||
458 fs_info->exclusive_operation == BTRFS_EXCLOP_NONE ||
459 fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
460 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED;
461 spin_unlock(&fs_info->super_lock);
462 break;
463 case BTRFS_EXCLOP_BALANCE:
464 spin_lock(&fs_info->super_lock);
465 ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
466 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE;
467 spin_unlock(&fs_info->super_lock);
468 break;
469 default:
470 btrfs_warn(fs_info,
471 "invalid exclop balance operation %d requested", op);
472 }
473}
474
475static int btrfs_ioctl_getversion(struct inode *inode, int __user *arg)
476{
477 return put_user(inode->i_generation, arg);
478}
479
480static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
481 void __user *arg)
482{
483 struct btrfs_device *device;
484 struct fstrim_range range;
485 u64 minlen = ULLONG_MAX;
486 u64 num_devices = 0;
487 int ret;
488
489 if (!capable(CAP_SYS_ADMIN))
490 return -EPERM;
491
492 /*
493 * btrfs_trim_block_group() depends on space cache, which is not
494 * available in zoned filesystem. So, disallow fitrim on a zoned
495 * filesystem for now.
496 */
497 if (btrfs_is_zoned(fs_info))
498 return -EOPNOTSUPP;
499
500 /*
501 * If the fs is mounted with nologreplay, which requires it to be
502 * mounted in RO mode as well, we can not allow discard on free space
503 * inside block groups, because log trees refer to extents that are not
504 * pinned in a block group's free space cache (pinning the extents is
505 * precisely the first phase of replaying a log tree).
506 */
507 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
508 return -EROFS;
509
510 rcu_read_lock();
511 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
512 dev_list) {
513 if (!device->bdev || !bdev_max_discard_sectors(device->bdev))
514 continue;
515 num_devices++;
516 minlen = min_t(u64, bdev_discard_granularity(device->bdev),
517 minlen);
518 }
519 rcu_read_unlock();
520
521 if (!num_devices)
522 return -EOPNOTSUPP;
523 if (copy_from_user(&range, arg, sizeof(range)))
524 return -EFAULT;
525
526 /*
527 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
528 * block group is in the logical address space, which can be any
529 * sectorsize aligned bytenr in the range [0, U64_MAX].
530 */
531 if (range.len < fs_info->sb->s_blocksize)
532 return -EINVAL;
533
534 range.minlen = max(range.minlen, minlen);
535 ret = btrfs_trim_fs(fs_info, &range);
536 if (ret < 0)
537 return ret;
538
539 if (copy_to_user(arg, &range, sizeof(range)))
540 return -EFAULT;
541
542 return 0;
543}
544
545int __pure btrfs_is_empty_uuid(u8 *uuid)
546{
547 int i;
548
549 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
550 if (uuid[i])
551 return 0;
552 }
553 return 1;
554}
555
556/*
557 * Calculate the number of transaction items to reserve for creating a subvolume
558 * or snapshot, not including the inode, directory entries, or parent directory.
559 */
560static unsigned int create_subvol_num_items(struct btrfs_qgroup_inherit *inherit)
561{
562 /*
563 * 1 to add root block
564 * 1 to add root item
565 * 1 to add root ref
566 * 1 to add root backref
567 * 1 to add UUID item
568 * 1 to add qgroup info
569 * 1 to add qgroup limit
570 *
571 * Ideally the last two would only be accounted if qgroups are enabled,
572 * but that can change between now and the time we would insert them.
573 */
574 unsigned int num_items = 7;
575
576 if (inherit) {
577 /* 2 to add qgroup relations for each inherited qgroup */
578 num_items += 2 * inherit->num_qgroups;
579 }
580 return num_items;
581}
582
583static noinline int create_subvol(struct mnt_idmap *idmap,
584 struct inode *dir, struct dentry *dentry,
585 struct btrfs_qgroup_inherit *inherit)
586{
587 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
588 struct btrfs_trans_handle *trans;
589 struct btrfs_key key;
590 struct btrfs_root_item *root_item;
591 struct btrfs_inode_item *inode_item;
592 struct extent_buffer *leaf;
593 struct btrfs_root *root = BTRFS_I(dir)->root;
594 struct btrfs_root *new_root;
595 struct btrfs_block_rsv block_rsv;
596 struct timespec64 cur_time = current_time(dir);
597 struct btrfs_new_inode_args new_inode_args = {
598 .dir = dir,
599 .dentry = dentry,
600 .subvol = true,
601 };
602 unsigned int trans_num_items;
603 int ret;
604 dev_t anon_dev;
605 u64 objectid;
606
607 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
608 if (!root_item)
609 return -ENOMEM;
610
611 ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
612 if (ret)
613 goto out_root_item;
614
615 /*
616 * Don't create subvolume whose level is not zero. Or qgroup will be
617 * screwed up since it assumes subvolume qgroup's level to be 0.
618 */
619 if (btrfs_qgroup_level(objectid)) {
620 ret = -ENOSPC;
621 goto out_root_item;
622 }
623
624 ret = get_anon_bdev(&anon_dev);
625 if (ret < 0)
626 goto out_root_item;
627
628 new_inode_args.inode = btrfs_new_subvol_inode(idmap, dir);
629 if (!new_inode_args.inode) {
630 ret = -ENOMEM;
631 goto out_anon_dev;
632 }
633 ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items);
634 if (ret)
635 goto out_inode;
636 trans_num_items += create_subvol_num_items(inherit);
637
638 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
639 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
640 trans_num_items, false);
641 if (ret)
642 goto out_new_inode_args;
643
644 trans = btrfs_start_transaction(root, 0);
645 if (IS_ERR(trans)) {
646 ret = PTR_ERR(trans);
647 btrfs_subvolume_release_metadata(root, &block_rsv);
648 goto out_new_inode_args;
649 }
650 trans->block_rsv = &block_rsv;
651 trans->bytes_reserved = block_rsv.size;
652 /* Tree log can't currently deal with an inode which is a new root. */
653 btrfs_set_log_full_commit(trans);
654
655 ret = btrfs_qgroup_inherit(trans, 0, objectid, root->root_key.objectid, inherit);
656 if (ret)
657 goto out;
658
659 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
660 0, BTRFS_NESTING_NORMAL);
661 if (IS_ERR(leaf)) {
662 ret = PTR_ERR(leaf);
663 goto out;
664 }
665
666 btrfs_mark_buffer_dirty(trans, leaf);
667
668 inode_item = &root_item->inode;
669 btrfs_set_stack_inode_generation(inode_item, 1);
670 btrfs_set_stack_inode_size(inode_item, 3);
671 btrfs_set_stack_inode_nlink(inode_item, 1);
672 btrfs_set_stack_inode_nbytes(inode_item,
673 fs_info->nodesize);
674 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
675
676 btrfs_set_root_flags(root_item, 0);
677 btrfs_set_root_limit(root_item, 0);
678 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
679
680 btrfs_set_root_bytenr(root_item, leaf->start);
681 btrfs_set_root_generation(root_item, trans->transid);
682 btrfs_set_root_level(root_item, 0);
683 btrfs_set_root_refs(root_item, 1);
684 btrfs_set_root_used(root_item, leaf->len);
685 btrfs_set_root_last_snapshot(root_item, 0);
686
687 btrfs_set_root_generation_v2(root_item,
688 btrfs_root_generation(root_item));
689 generate_random_guid(root_item->uuid);
690 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
691 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
692 root_item->ctime = root_item->otime;
693 btrfs_set_root_ctransid(root_item, trans->transid);
694 btrfs_set_root_otransid(root_item, trans->transid);
695
696 btrfs_tree_unlock(leaf);
697
698 btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
699
700 key.objectid = objectid;
701 key.offset = 0;
702 key.type = BTRFS_ROOT_ITEM_KEY;
703 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
704 root_item);
705 if (ret) {
706 /*
707 * Since we don't abort the transaction in this case, free the
708 * tree block so that we don't leak space and leave the
709 * filesystem in an inconsistent state (an extent item in the
710 * extent tree with a backreference for a root that does not
711 * exists).
712 */
713 btrfs_tree_lock(leaf);
714 btrfs_clear_buffer_dirty(trans, leaf);
715 btrfs_tree_unlock(leaf);
716 btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
717 free_extent_buffer(leaf);
718 goto out;
719 }
720
721 free_extent_buffer(leaf);
722 leaf = NULL;
723
724 new_root = btrfs_get_new_fs_root(fs_info, objectid, &anon_dev);
725 if (IS_ERR(new_root)) {
726 ret = PTR_ERR(new_root);
727 btrfs_abort_transaction(trans, ret);
728 goto out;
729 }
730 /* anon_dev is owned by new_root now. */
731 anon_dev = 0;
732 BTRFS_I(new_inode_args.inode)->root = new_root;
733 /* ... and new_root is owned by new_inode_args.inode now. */
734
735 ret = btrfs_record_root_in_trans(trans, new_root);
736 if (ret) {
737 btrfs_abort_transaction(trans, ret);
738 goto out;
739 }
740
741 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
742 BTRFS_UUID_KEY_SUBVOL, objectid);
743 if (ret) {
744 btrfs_abort_transaction(trans, ret);
745 goto out;
746 }
747
748 ret = btrfs_create_new_inode(trans, &new_inode_args);
749 if (ret) {
750 btrfs_abort_transaction(trans, ret);
751 goto out;
752 }
753
754 d_instantiate_new(dentry, new_inode_args.inode);
755 new_inode_args.inode = NULL;
756
757out:
758 trans->block_rsv = NULL;
759 trans->bytes_reserved = 0;
760 btrfs_subvolume_release_metadata(root, &block_rsv);
761
762 btrfs_end_transaction(trans);
763out_new_inode_args:
764 btrfs_new_inode_args_destroy(&new_inode_args);
765out_inode:
766 iput(new_inode_args.inode);
767out_anon_dev:
768 if (anon_dev)
769 free_anon_bdev(anon_dev);
770out_root_item:
771 kfree(root_item);
772 return ret;
773}
774
775static int create_snapshot(struct btrfs_root *root, struct inode *dir,
776 struct dentry *dentry, bool readonly,
777 struct btrfs_qgroup_inherit *inherit)
778{
779 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
780 struct inode *inode;
781 struct btrfs_pending_snapshot *pending_snapshot;
782 unsigned int trans_num_items;
783 struct btrfs_trans_handle *trans;
784 int ret;
785
786 /* We do not support snapshotting right now. */
787 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
788 btrfs_warn(fs_info,
789 "extent tree v2 doesn't support snapshotting yet");
790 return -EOPNOTSUPP;
791 }
792
793 if (btrfs_root_refs(&root->root_item) == 0)
794 return -ENOENT;
795
796 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
797 return -EINVAL;
798
799 if (atomic_read(&root->nr_swapfiles)) {
800 btrfs_warn(fs_info,
801 "cannot snapshot subvolume with active swapfile");
802 return -ETXTBSY;
803 }
804
805 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
806 if (!pending_snapshot)
807 return -ENOMEM;
808
809 ret = get_anon_bdev(&pending_snapshot->anon_dev);
810 if (ret < 0)
811 goto free_pending;
812 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
813 GFP_KERNEL);
814 pending_snapshot->path = btrfs_alloc_path();
815 if (!pending_snapshot->root_item || !pending_snapshot->path) {
816 ret = -ENOMEM;
817 goto free_pending;
818 }
819
820 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
821 BTRFS_BLOCK_RSV_TEMP);
822 /*
823 * 1 to add dir item
824 * 1 to add dir index
825 * 1 to update parent inode item
826 */
827 trans_num_items = create_subvol_num_items(inherit) + 3;
828 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
829 &pending_snapshot->block_rsv,
830 trans_num_items, false);
831 if (ret)
832 goto free_pending;
833
834 pending_snapshot->dentry = dentry;
835 pending_snapshot->root = root;
836 pending_snapshot->readonly = readonly;
837 pending_snapshot->dir = dir;
838 pending_snapshot->inherit = inherit;
839
840 trans = btrfs_start_transaction(root, 0);
841 if (IS_ERR(trans)) {
842 ret = PTR_ERR(trans);
843 goto fail;
844 }
845
846 trans->pending_snapshot = pending_snapshot;
847
848 ret = btrfs_commit_transaction(trans);
849 if (ret)
850 goto fail;
851
852 ret = pending_snapshot->error;
853 if (ret)
854 goto fail;
855
856 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
857 if (ret)
858 goto fail;
859
860 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
861 if (IS_ERR(inode)) {
862 ret = PTR_ERR(inode);
863 goto fail;
864 }
865
866 d_instantiate(dentry, inode);
867 ret = 0;
868 pending_snapshot->anon_dev = 0;
869fail:
870 /* Prevent double freeing of anon_dev */
871 if (ret && pending_snapshot->snap)
872 pending_snapshot->snap->anon_dev = 0;
873 btrfs_put_root(pending_snapshot->snap);
874 btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
875free_pending:
876 if (pending_snapshot->anon_dev)
877 free_anon_bdev(pending_snapshot->anon_dev);
878 kfree(pending_snapshot->root_item);
879 btrfs_free_path(pending_snapshot->path);
880 kfree(pending_snapshot);
881
882 return ret;
883}
884
885/* copy of may_delete in fs/namei.c()
886 * Check whether we can remove a link victim from directory dir, check
887 * whether the type of victim is right.
888 * 1. We can't do it if dir is read-only (done in permission())
889 * 2. We should have write and exec permissions on dir
890 * 3. We can't remove anything from append-only dir
891 * 4. We can't do anything with immutable dir (done in permission())
892 * 5. If the sticky bit on dir is set we should either
893 * a. be owner of dir, or
894 * b. be owner of victim, or
895 * c. have CAP_FOWNER capability
896 * 6. If the victim is append-only or immutable we can't do anything with
897 * links pointing to it.
898 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
899 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
900 * 9. We can't remove a root or mountpoint.
901 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
902 * nfs_async_unlink().
903 */
904
905static int btrfs_may_delete(struct mnt_idmap *idmap,
906 struct inode *dir, struct dentry *victim, int isdir)
907{
908 int error;
909
910 if (d_really_is_negative(victim))
911 return -ENOENT;
912
913 BUG_ON(d_inode(victim->d_parent) != dir);
914 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
915
916 error = inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
917 if (error)
918 return error;
919 if (IS_APPEND(dir))
920 return -EPERM;
921 if (check_sticky(idmap, dir, d_inode(victim)) ||
922 IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
923 IS_SWAPFILE(d_inode(victim)))
924 return -EPERM;
925 if (isdir) {
926 if (!d_is_dir(victim))
927 return -ENOTDIR;
928 if (IS_ROOT(victim))
929 return -EBUSY;
930 } else if (d_is_dir(victim))
931 return -EISDIR;
932 if (IS_DEADDIR(dir))
933 return -ENOENT;
934 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
935 return -EBUSY;
936 return 0;
937}
938
939/* copy of may_create in fs/namei.c() */
940static inline int btrfs_may_create(struct mnt_idmap *idmap,
941 struct inode *dir, struct dentry *child)
942{
943 if (d_really_is_positive(child))
944 return -EEXIST;
945 if (IS_DEADDIR(dir))
946 return -ENOENT;
947 if (!fsuidgid_has_mapping(dir->i_sb, idmap))
948 return -EOVERFLOW;
949 return inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
950}
951
952/*
953 * Create a new subvolume below @parent. This is largely modeled after
954 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
955 * inside this filesystem so it's quite a bit simpler.
956 */
957static noinline int btrfs_mksubvol(const struct path *parent,
958 struct mnt_idmap *idmap,
959 const char *name, int namelen,
960 struct btrfs_root *snap_src,
961 bool readonly,
962 struct btrfs_qgroup_inherit *inherit)
963{
964 struct inode *dir = d_inode(parent->dentry);
965 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
966 struct dentry *dentry;
967 struct fscrypt_str name_str = FSTR_INIT((char *)name, namelen);
968 int error;
969
970 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
971 if (error == -EINTR)
972 return error;
973
974 dentry = lookup_one(idmap, name, parent->dentry, namelen);
975 error = PTR_ERR(dentry);
976 if (IS_ERR(dentry))
977 goto out_unlock;
978
979 error = btrfs_may_create(idmap, dir, dentry);
980 if (error)
981 goto out_dput;
982
983 /*
984 * even if this name doesn't exist, we may get hash collisions.
985 * check for them now when we can safely fail
986 */
987 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
988 dir->i_ino, &name_str);
989 if (error)
990 goto out_dput;
991
992 down_read(&fs_info->subvol_sem);
993
994 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
995 goto out_up_read;
996
997 if (snap_src)
998 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
999 else
1000 error = create_subvol(idmap, dir, dentry, inherit);
1001
1002 if (!error)
1003 fsnotify_mkdir(dir, dentry);
1004out_up_read:
1005 up_read(&fs_info->subvol_sem);
1006out_dput:
1007 dput(dentry);
1008out_unlock:
1009 btrfs_inode_unlock(BTRFS_I(dir), 0);
1010 return error;
1011}
1012
1013static noinline int btrfs_mksnapshot(const struct path *parent,
1014 struct mnt_idmap *idmap,
1015 const char *name, int namelen,
1016 struct btrfs_root *root,
1017 bool readonly,
1018 struct btrfs_qgroup_inherit *inherit)
1019{
1020 int ret;
1021 bool snapshot_force_cow = false;
1022
1023 /*
1024 * Force new buffered writes to reserve space even when NOCOW is
1025 * possible. This is to avoid later writeback (running dealloc) to
1026 * fallback to COW mode and unexpectedly fail with ENOSPC.
1027 */
1028 btrfs_drew_read_lock(&root->snapshot_lock);
1029
1030 ret = btrfs_start_delalloc_snapshot(root, false);
1031 if (ret)
1032 goto out;
1033
1034 /*
1035 * All previous writes have started writeback in NOCOW mode, so now
1036 * we force future writes to fallback to COW mode during snapshot
1037 * creation.
1038 */
1039 atomic_inc(&root->snapshot_force_cow);
1040 snapshot_force_cow = true;
1041
1042 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
1043
1044 ret = btrfs_mksubvol(parent, idmap, name, namelen,
1045 root, readonly, inherit);
1046out:
1047 if (snapshot_force_cow)
1048 atomic_dec(&root->snapshot_force_cow);
1049 btrfs_drew_read_unlock(&root->snapshot_lock);
1050 return ret;
1051}
1052
1053/*
1054 * Try to start exclusive operation @type or cancel it if it's running.
1055 *
1056 * Return:
1057 * 0 - normal mode, newly claimed op started
1058 * >0 - normal mode, something else is running,
1059 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1060 * ECANCELED - cancel mode, successful cancel
1061 * ENOTCONN - cancel mode, operation not running anymore
1062 */
1063static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1064 enum btrfs_exclusive_operation type, bool cancel)
1065{
1066 if (!cancel) {
1067 /* Start normal op */
1068 if (!btrfs_exclop_start(fs_info, type))
1069 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1070 /* Exclusive operation is now claimed */
1071 return 0;
1072 }
1073
1074 /* Cancel running op */
1075 if (btrfs_exclop_start_try_lock(fs_info, type)) {
1076 /*
1077 * This blocks any exclop finish from setting it to NONE, so we
1078 * request cancellation. Either it runs and we will wait for it,
1079 * or it has finished and no waiting will happen.
1080 */
1081 atomic_inc(&fs_info->reloc_cancel_req);
1082 btrfs_exclop_start_unlock(fs_info);
1083
1084 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1085 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1086 TASK_INTERRUPTIBLE);
1087
1088 return -ECANCELED;
1089 }
1090
1091 /* Something else is running or none */
1092 return -ENOTCONN;
1093}
1094
1095static noinline int btrfs_ioctl_resize(struct file *file,
1096 void __user *arg)
1097{
1098 BTRFS_DEV_LOOKUP_ARGS(args);
1099 struct inode *inode = file_inode(file);
1100 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1101 u64 new_size;
1102 u64 old_size;
1103 u64 devid = 1;
1104 struct btrfs_root *root = BTRFS_I(inode)->root;
1105 struct btrfs_ioctl_vol_args *vol_args;
1106 struct btrfs_trans_handle *trans;
1107 struct btrfs_device *device = NULL;
1108 char *sizestr;
1109 char *retptr;
1110 char *devstr = NULL;
1111 int ret = 0;
1112 int mod = 0;
1113 bool cancel;
1114
1115 if (!capable(CAP_SYS_ADMIN))
1116 return -EPERM;
1117
1118 ret = mnt_want_write_file(file);
1119 if (ret)
1120 return ret;
1121
1122 /*
1123 * Read the arguments before checking exclusivity to be able to
1124 * distinguish regular resize and cancel
1125 */
1126 vol_args = memdup_user(arg, sizeof(*vol_args));
1127 if (IS_ERR(vol_args)) {
1128 ret = PTR_ERR(vol_args);
1129 goto out_drop;
1130 }
1131 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1132 sizestr = vol_args->name;
1133 cancel = (strcmp("cancel", sizestr) == 0);
1134 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1135 if (ret)
1136 goto out_free;
1137 /* Exclusive operation is now claimed */
1138
1139 devstr = strchr(sizestr, ':');
1140 if (devstr) {
1141 sizestr = devstr + 1;
1142 *devstr = '\0';
1143 devstr = vol_args->name;
1144 ret = kstrtoull(devstr, 10, &devid);
1145 if (ret)
1146 goto out_finish;
1147 if (!devid) {
1148 ret = -EINVAL;
1149 goto out_finish;
1150 }
1151 btrfs_info(fs_info, "resizing devid %llu", devid);
1152 }
1153
1154 args.devid = devid;
1155 device = btrfs_find_device(fs_info->fs_devices, &args);
1156 if (!device) {
1157 btrfs_info(fs_info, "resizer unable to find device %llu",
1158 devid);
1159 ret = -ENODEV;
1160 goto out_finish;
1161 }
1162
1163 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1164 btrfs_info(fs_info,
1165 "resizer unable to apply on readonly device %llu",
1166 devid);
1167 ret = -EPERM;
1168 goto out_finish;
1169 }
1170
1171 if (!strcmp(sizestr, "max"))
1172 new_size = bdev_nr_bytes(device->bdev);
1173 else {
1174 if (sizestr[0] == '-') {
1175 mod = -1;
1176 sizestr++;
1177 } else if (sizestr[0] == '+') {
1178 mod = 1;
1179 sizestr++;
1180 }
1181 new_size = memparse(sizestr, &retptr);
1182 if (*retptr != '\0' || new_size == 0) {
1183 ret = -EINVAL;
1184 goto out_finish;
1185 }
1186 }
1187
1188 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1189 ret = -EPERM;
1190 goto out_finish;
1191 }
1192
1193 old_size = btrfs_device_get_total_bytes(device);
1194
1195 if (mod < 0) {
1196 if (new_size > old_size) {
1197 ret = -EINVAL;
1198 goto out_finish;
1199 }
1200 new_size = old_size - new_size;
1201 } else if (mod > 0) {
1202 if (new_size > ULLONG_MAX - old_size) {
1203 ret = -ERANGE;
1204 goto out_finish;
1205 }
1206 new_size = old_size + new_size;
1207 }
1208
1209 if (new_size < SZ_256M) {
1210 ret = -EINVAL;
1211 goto out_finish;
1212 }
1213 if (new_size > bdev_nr_bytes(device->bdev)) {
1214 ret = -EFBIG;
1215 goto out_finish;
1216 }
1217
1218 new_size = round_down(new_size, fs_info->sectorsize);
1219
1220 if (new_size > old_size) {
1221 trans = btrfs_start_transaction(root, 0);
1222 if (IS_ERR(trans)) {
1223 ret = PTR_ERR(trans);
1224 goto out_finish;
1225 }
1226 ret = btrfs_grow_device(trans, device, new_size);
1227 btrfs_commit_transaction(trans);
1228 } else if (new_size < old_size) {
1229 ret = btrfs_shrink_device(device, new_size);
1230 } /* equal, nothing need to do */
1231
1232 if (ret == 0 && new_size != old_size)
1233 btrfs_info_in_rcu(fs_info,
1234 "resize device %s (devid %llu) from %llu to %llu",
1235 btrfs_dev_name(device), device->devid,
1236 old_size, new_size);
1237out_finish:
1238 btrfs_exclop_finish(fs_info);
1239out_free:
1240 kfree(vol_args);
1241out_drop:
1242 mnt_drop_write_file(file);
1243 return ret;
1244}
1245
1246static noinline int __btrfs_ioctl_snap_create(struct file *file,
1247 struct mnt_idmap *idmap,
1248 const char *name, unsigned long fd, int subvol,
1249 bool readonly,
1250 struct btrfs_qgroup_inherit *inherit)
1251{
1252 int namelen;
1253 int ret = 0;
1254
1255 if (!S_ISDIR(file_inode(file)->i_mode))
1256 return -ENOTDIR;
1257
1258 ret = mnt_want_write_file(file);
1259 if (ret)
1260 goto out;
1261
1262 namelen = strlen(name);
1263 if (strchr(name, '/')) {
1264 ret = -EINVAL;
1265 goto out_drop_write;
1266 }
1267
1268 if (name[0] == '.' &&
1269 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1270 ret = -EEXIST;
1271 goto out_drop_write;
1272 }
1273
1274 if (subvol) {
1275 ret = btrfs_mksubvol(&file->f_path, idmap, name,
1276 namelen, NULL, readonly, inherit);
1277 } else {
1278 struct fd src = fdget(fd);
1279 struct inode *src_inode;
1280 if (!src.file) {
1281 ret = -EINVAL;
1282 goto out_drop_write;
1283 }
1284
1285 src_inode = file_inode(src.file);
1286 if (src_inode->i_sb != file_inode(file)->i_sb) {
1287 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1288 "Snapshot src from another FS");
1289 ret = -EXDEV;
1290 } else if (!inode_owner_or_capable(idmap, src_inode)) {
1291 /*
1292 * Subvolume creation is not restricted, but snapshots
1293 * are limited to own subvolumes only
1294 */
1295 ret = -EPERM;
1296 } else if (btrfs_ino(BTRFS_I(src_inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1297 /*
1298 * Snapshots must be made with the src_inode referring
1299 * to the subvolume inode, otherwise the permission
1300 * checking above is useless because we may have
1301 * permission on a lower directory but not the subvol
1302 * itself.
1303 */
1304 ret = -EINVAL;
1305 } else {
1306 ret = btrfs_mksnapshot(&file->f_path, idmap,
1307 name, namelen,
1308 BTRFS_I(src_inode)->root,
1309 readonly, inherit);
1310 }
1311 fdput(src);
1312 }
1313out_drop_write:
1314 mnt_drop_write_file(file);
1315out:
1316 return ret;
1317}
1318
1319static noinline int btrfs_ioctl_snap_create(struct file *file,
1320 void __user *arg, int subvol)
1321{
1322 struct btrfs_ioctl_vol_args *vol_args;
1323 int ret;
1324
1325 if (!S_ISDIR(file_inode(file)->i_mode))
1326 return -ENOTDIR;
1327
1328 vol_args = memdup_user(arg, sizeof(*vol_args));
1329 if (IS_ERR(vol_args))
1330 return PTR_ERR(vol_args);
1331 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1332
1333 ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1334 vol_args->name, vol_args->fd, subvol,
1335 false, NULL);
1336
1337 kfree(vol_args);
1338 return ret;
1339}
1340
1341static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1342 void __user *arg, int subvol)
1343{
1344 struct btrfs_ioctl_vol_args_v2 *vol_args;
1345 int ret;
1346 bool readonly = false;
1347 struct btrfs_qgroup_inherit *inherit = NULL;
1348
1349 if (!S_ISDIR(file_inode(file)->i_mode))
1350 return -ENOTDIR;
1351
1352 vol_args = memdup_user(arg, sizeof(*vol_args));
1353 if (IS_ERR(vol_args))
1354 return PTR_ERR(vol_args);
1355 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1356
1357 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1358 ret = -EOPNOTSUPP;
1359 goto free_args;
1360 }
1361
1362 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1363 readonly = true;
1364 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1365 u64 nums;
1366
1367 if (vol_args->size < sizeof(*inherit) ||
1368 vol_args->size > PAGE_SIZE) {
1369 ret = -EINVAL;
1370 goto free_args;
1371 }
1372 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1373 if (IS_ERR(inherit)) {
1374 ret = PTR_ERR(inherit);
1375 goto free_args;
1376 }
1377
1378 if (inherit->num_qgroups > PAGE_SIZE ||
1379 inherit->num_ref_copies > PAGE_SIZE ||
1380 inherit->num_excl_copies > PAGE_SIZE) {
1381 ret = -EINVAL;
1382 goto free_inherit;
1383 }
1384
1385 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
1386 2 * inherit->num_excl_copies;
1387 if (vol_args->size != struct_size(inherit, qgroups, nums)) {
1388 ret = -EINVAL;
1389 goto free_inherit;
1390 }
1391 }
1392
1393 ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1394 vol_args->name, vol_args->fd, subvol,
1395 readonly, inherit);
1396 if (ret)
1397 goto free_inherit;
1398free_inherit:
1399 kfree(inherit);
1400free_args:
1401 kfree(vol_args);
1402 return ret;
1403}
1404
1405static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode,
1406 void __user *arg)
1407{
1408 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1409 struct btrfs_root *root = BTRFS_I(inode)->root;
1410 int ret = 0;
1411 u64 flags = 0;
1412
1413 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1414 return -EINVAL;
1415
1416 down_read(&fs_info->subvol_sem);
1417 if (btrfs_root_readonly(root))
1418 flags |= BTRFS_SUBVOL_RDONLY;
1419 up_read(&fs_info->subvol_sem);
1420
1421 if (copy_to_user(arg, &flags, sizeof(flags)))
1422 ret = -EFAULT;
1423
1424 return ret;
1425}
1426
1427static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1428 void __user *arg)
1429{
1430 struct inode *inode = file_inode(file);
1431 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1432 struct btrfs_root *root = BTRFS_I(inode)->root;
1433 struct btrfs_trans_handle *trans;
1434 u64 root_flags;
1435 u64 flags;
1436 int ret = 0;
1437
1438 if (!inode_owner_or_capable(file_mnt_idmap(file), inode))
1439 return -EPERM;
1440
1441 ret = mnt_want_write_file(file);
1442 if (ret)
1443 goto out;
1444
1445 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1446 ret = -EINVAL;
1447 goto out_drop_write;
1448 }
1449
1450 if (copy_from_user(&flags, arg, sizeof(flags))) {
1451 ret = -EFAULT;
1452 goto out_drop_write;
1453 }
1454
1455 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1456 ret = -EOPNOTSUPP;
1457 goto out_drop_write;
1458 }
1459
1460 down_write(&fs_info->subvol_sem);
1461
1462 /* nothing to do */
1463 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1464 goto out_drop_sem;
1465
1466 root_flags = btrfs_root_flags(&root->root_item);
1467 if (flags & BTRFS_SUBVOL_RDONLY) {
1468 btrfs_set_root_flags(&root->root_item,
1469 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1470 } else {
1471 /*
1472 * Block RO -> RW transition if this subvolume is involved in
1473 * send
1474 */
1475 spin_lock(&root->root_item_lock);
1476 if (root->send_in_progress == 0) {
1477 btrfs_set_root_flags(&root->root_item,
1478 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1479 spin_unlock(&root->root_item_lock);
1480 } else {
1481 spin_unlock(&root->root_item_lock);
1482 btrfs_warn(fs_info,
1483 "Attempt to set subvolume %llu read-write during send",
1484 root->root_key.objectid);
1485 ret = -EPERM;
1486 goto out_drop_sem;
1487 }
1488 }
1489
1490 trans = btrfs_start_transaction(root, 1);
1491 if (IS_ERR(trans)) {
1492 ret = PTR_ERR(trans);
1493 goto out_reset;
1494 }
1495
1496 ret = btrfs_update_root(trans, fs_info->tree_root,
1497 &root->root_key, &root->root_item);
1498 if (ret < 0) {
1499 btrfs_end_transaction(trans);
1500 goto out_reset;
1501 }
1502
1503 ret = btrfs_commit_transaction(trans);
1504
1505out_reset:
1506 if (ret)
1507 btrfs_set_root_flags(&root->root_item, root_flags);
1508out_drop_sem:
1509 up_write(&fs_info->subvol_sem);
1510out_drop_write:
1511 mnt_drop_write_file(file);
1512out:
1513 return ret;
1514}
1515
1516static noinline int key_in_sk(struct btrfs_key *key,
1517 struct btrfs_ioctl_search_key *sk)
1518{
1519 struct btrfs_key test;
1520 int ret;
1521
1522 test.objectid = sk->min_objectid;
1523 test.type = sk->min_type;
1524 test.offset = sk->min_offset;
1525
1526 ret = btrfs_comp_cpu_keys(key, &test);
1527 if (ret < 0)
1528 return 0;
1529
1530 test.objectid = sk->max_objectid;
1531 test.type = sk->max_type;
1532 test.offset = sk->max_offset;
1533
1534 ret = btrfs_comp_cpu_keys(key, &test);
1535 if (ret > 0)
1536 return 0;
1537 return 1;
1538}
1539
1540static noinline int copy_to_sk(struct btrfs_path *path,
1541 struct btrfs_key *key,
1542 struct btrfs_ioctl_search_key *sk,
1543 u64 *buf_size,
1544 char __user *ubuf,
1545 unsigned long *sk_offset,
1546 int *num_found)
1547{
1548 u64 found_transid;
1549 struct extent_buffer *leaf;
1550 struct btrfs_ioctl_search_header sh;
1551 struct btrfs_key test;
1552 unsigned long item_off;
1553 unsigned long item_len;
1554 int nritems;
1555 int i;
1556 int slot;
1557 int ret = 0;
1558
1559 leaf = path->nodes[0];
1560 slot = path->slots[0];
1561 nritems = btrfs_header_nritems(leaf);
1562
1563 if (btrfs_header_generation(leaf) > sk->max_transid) {
1564 i = nritems;
1565 goto advance_key;
1566 }
1567 found_transid = btrfs_header_generation(leaf);
1568
1569 for (i = slot; i < nritems; i++) {
1570 item_off = btrfs_item_ptr_offset(leaf, i);
1571 item_len = btrfs_item_size(leaf, i);
1572
1573 btrfs_item_key_to_cpu(leaf, key, i);
1574 if (!key_in_sk(key, sk))
1575 continue;
1576
1577 if (sizeof(sh) + item_len > *buf_size) {
1578 if (*num_found) {
1579 ret = 1;
1580 goto out;
1581 }
1582
1583 /*
1584 * return one empty item back for v1, which does not
1585 * handle -EOVERFLOW
1586 */
1587
1588 *buf_size = sizeof(sh) + item_len;
1589 item_len = 0;
1590 ret = -EOVERFLOW;
1591 }
1592
1593 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1594 ret = 1;
1595 goto out;
1596 }
1597
1598 sh.objectid = key->objectid;
1599 sh.offset = key->offset;
1600 sh.type = key->type;
1601 sh.len = item_len;
1602 sh.transid = found_transid;
1603
1604 /*
1605 * Copy search result header. If we fault then loop again so we
1606 * can fault in the pages and -EFAULT there if there's a
1607 * problem. Otherwise we'll fault and then copy the buffer in
1608 * properly this next time through
1609 */
1610 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
1611 ret = 0;
1612 goto out;
1613 }
1614
1615 *sk_offset += sizeof(sh);
1616
1617 if (item_len) {
1618 char __user *up = ubuf + *sk_offset;
1619 /*
1620 * Copy the item, same behavior as above, but reset the
1621 * * sk_offset so we copy the full thing again.
1622 */
1623 if (read_extent_buffer_to_user_nofault(leaf, up,
1624 item_off, item_len)) {
1625 ret = 0;
1626 *sk_offset -= sizeof(sh);
1627 goto out;
1628 }
1629
1630 *sk_offset += item_len;
1631 }
1632 (*num_found)++;
1633
1634 if (ret) /* -EOVERFLOW from above */
1635 goto out;
1636
1637 if (*num_found >= sk->nr_items) {
1638 ret = 1;
1639 goto out;
1640 }
1641 }
1642advance_key:
1643 ret = 0;
1644 test.objectid = sk->max_objectid;
1645 test.type = sk->max_type;
1646 test.offset = sk->max_offset;
1647 if (btrfs_comp_cpu_keys(key, &test) >= 0)
1648 ret = 1;
1649 else if (key->offset < (u64)-1)
1650 key->offset++;
1651 else if (key->type < (u8)-1) {
1652 key->offset = 0;
1653 key->type++;
1654 } else if (key->objectid < (u64)-1) {
1655 key->offset = 0;
1656 key->type = 0;
1657 key->objectid++;
1658 } else
1659 ret = 1;
1660out:
1661 /*
1662 * 0: all items from this leaf copied, continue with next
1663 * 1: * more items can be copied, but unused buffer is too small
1664 * * all items were found
1665 * Either way, it will stops the loop which iterates to the next
1666 * leaf
1667 * -EOVERFLOW: item was to large for buffer
1668 * -EFAULT: could not copy extent buffer back to userspace
1669 */
1670 return ret;
1671}
1672
1673static noinline int search_ioctl(struct inode *inode,
1674 struct btrfs_ioctl_search_key *sk,
1675 u64 *buf_size,
1676 char __user *ubuf)
1677{
1678 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
1679 struct btrfs_root *root;
1680 struct btrfs_key key;
1681 struct btrfs_path *path;
1682 int ret;
1683 int num_found = 0;
1684 unsigned long sk_offset = 0;
1685
1686 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
1687 *buf_size = sizeof(struct btrfs_ioctl_search_header);
1688 return -EOVERFLOW;
1689 }
1690
1691 path = btrfs_alloc_path();
1692 if (!path)
1693 return -ENOMEM;
1694
1695 if (sk->tree_id == 0) {
1696 /* search the root of the inode that was passed */
1697 root = btrfs_grab_root(BTRFS_I(inode)->root);
1698 } else {
1699 root = btrfs_get_fs_root(info, sk->tree_id, true);
1700 if (IS_ERR(root)) {
1701 btrfs_free_path(path);
1702 return PTR_ERR(root);
1703 }
1704 }
1705
1706 key.objectid = sk->min_objectid;
1707 key.type = sk->min_type;
1708 key.offset = sk->min_offset;
1709
1710 while (1) {
1711 ret = -EFAULT;
1712 /*
1713 * Ensure that the whole user buffer is faulted in at sub-page
1714 * granularity, otherwise the loop may live-lock.
1715 */
1716 if (fault_in_subpage_writeable(ubuf + sk_offset,
1717 *buf_size - sk_offset))
1718 break;
1719
1720 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
1721 if (ret != 0) {
1722 if (ret > 0)
1723 ret = 0;
1724 goto err;
1725 }
1726 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
1727 &sk_offset, &num_found);
1728 btrfs_release_path(path);
1729 if (ret)
1730 break;
1731
1732 }
1733 if (ret > 0)
1734 ret = 0;
1735err:
1736 sk->nr_items = num_found;
1737 btrfs_put_root(root);
1738 btrfs_free_path(path);
1739 return ret;
1740}
1741
1742static noinline int btrfs_ioctl_tree_search(struct inode *inode,
1743 void __user *argp)
1744{
1745 struct btrfs_ioctl_search_args __user *uargs = argp;
1746 struct btrfs_ioctl_search_key sk;
1747 int ret;
1748 u64 buf_size;
1749
1750 if (!capable(CAP_SYS_ADMIN))
1751 return -EPERM;
1752
1753 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
1754 return -EFAULT;
1755
1756 buf_size = sizeof(uargs->buf);
1757
1758 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
1759
1760 /*
1761 * In the origin implementation an overflow is handled by returning a
1762 * search header with a len of zero, so reset ret.
1763 */
1764 if (ret == -EOVERFLOW)
1765 ret = 0;
1766
1767 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
1768 ret = -EFAULT;
1769 return ret;
1770}
1771
1772static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode,
1773 void __user *argp)
1774{
1775 struct btrfs_ioctl_search_args_v2 __user *uarg = argp;
1776 struct btrfs_ioctl_search_args_v2 args;
1777 int ret;
1778 u64 buf_size;
1779 const u64 buf_limit = SZ_16M;
1780
1781 if (!capable(CAP_SYS_ADMIN))
1782 return -EPERM;
1783
1784 /* copy search header and buffer size */
1785 if (copy_from_user(&args, uarg, sizeof(args)))
1786 return -EFAULT;
1787
1788 buf_size = args.buf_size;
1789
1790 /* limit result size to 16MB */
1791 if (buf_size > buf_limit)
1792 buf_size = buf_limit;
1793
1794 ret = search_ioctl(inode, &args.key, &buf_size,
1795 (char __user *)(&uarg->buf[0]));
1796 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
1797 ret = -EFAULT;
1798 else if (ret == -EOVERFLOW &&
1799 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
1800 ret = -EFAULT;
1801
1802 return ret;
1803}
1804
1805/*
1806 * Search INODE_REFs to identify path name of 'dirid' directory
1807 * in a 'tree_id' tree. and sets path name to 'name'.
1808 */
1809static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1810 u64 tree_id, u64 dirid, char *name)
1811{
1812 struct btrfs_root *root;
1813 struct btrfs_key key;
1814 char *ptr;
1815 int ret = -1;
1816 int slot;
1817 int len;
1818 int total_len = 0;
1819 struct btrfs_inode_ref *iref;
1820 struct extent_buffer *l;
1821 struct btrfs_path *path;
1822
1823 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1824 name[0]='\0';
1825 return 0;
1826 }
1827
1828 path = btrfs_alloc_path();
1829 if (!path)
1830 return -ENOMEM;
1831
1832 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
1833
1834 root = btrfs_get_fs_root(info, tree_id, true);
1835 if (IS_ERR(root)) {
1836 ret = PTR_ERR(root);
1837 root = NULL;
1838 goto out;
1839 }
1840
1841 key.objectid = dirid;
1842 key.type = BTRFS_INODE_REF_KEY;
1843 key.offset = (u64)-1;
1844
1845 while (1) {
1846 ret = btrfs_search_backwards(root, &key, path);
1847 if (ret < 0)
1848 goto out;
1849 else if (ret > 0) {
1850 ret = -ENOENT;
1851 goto out;
1852 }
1853
1854 l = path->nodes[0];
1855 slot = path->slots[0];
1856
1857 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1858 len = btrfs_inode_ref_name_len(l, iref);
1859 ptr -= len + 1;
1860 total_len += len + 1;
1861 if (ptr < name) {
1862 ret = -ENAMETOOLONG;
1863 goto out;
1864 }
1865
1866 *(ptr + len) = '/';
1867 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
1868
1869 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1870 break;
1871
1872 btrfs_release_path(path);
1873 key.objectid = key.offset;
1874 key.offset = (u64)-1;
1875 dirid = key.objectid;
1876 }
1877 memmove(name, ptr, total_len);
1878 name[total_len] = '\0';
1879 ret = 0;
1880out:
1881 btrfs_put_root(root);
1882 btrfs_free_path(path);
1883 return ret;
1884}
1885
1886static int btrfs_search_path_in_tree_user(struct mnt_idmap *idmap,
1887 struct inode *inode,
1888 struct btrfs_ioctl_ino_lookup_user_args *args)
1889{
1890 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
1891 struct super_block *sb = inode->i_sb;
1892 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
1893 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
1894 u64 dirid = args->dirid;
1895 unsigned long item_off;
1896 unsigned long item_len;
1897 struct btrfs_inode_ref *iref;
1898 struct btrfs_root_ref *rref;
1899 struct btrfs_root *root = NULL;
1900 struct btrfs_path *path;
1901 struct btrfs_key key, key2;
1902 struct extent_buffer *leaf;
1903 struct inode *temp_inode;
1904 char *ptr;
1905 int slot;
1906 int len;
1907 int total_len = 0;
1908 int ret;
1909
1910 path = btrfs_alloc_path();
1911 if (!path)
1912 return -ENOMEM;
1913
1914 /*
1915 * If the bottom subvolume does not exist directly under upper_limit,
1916 * construct the path in from the bottom up.
1917 */
1918 if (dirid != upper_limit.objectid) {
1919 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
1920
1921 root = btrfs_get_fs_root(fs_info, treeid, true);
1922 if (IS_ERR(root)) {
1923 ret = PTR_ERR(root);
1924 goto out;
1925 }
1926
1927 key.objectid = dirid;
1928 key.type = BTRFS_INODE_REF_KEY;
1929 key.offset = (u64)-1;
1930 while (1) {
1931 ret = btrfs_search_backwards(root, &key, path);
1932 if (ret < 0)
1933 goto out_put;
1934 else if (ret > 0) {
1935 ret = -ENOENT;
1936 goto out_put;
1937 }
1938
1939 leaf = path->nodes[0];
1940 slot = path->slots[0];
1941
1942 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
1943 len = btrfs_inode_ref_name_len(leaf, iref);
1944 ptr -= len + 1;
1945 total_len += len + 1;
1946 if (ptr < args->path) {
1947 ret = -ENAMETOOLONG;
1948 goto out_put;
1949 }
1950
1951 *(ptr + len) = '/';
1952 read_extent_buffer(leaf, ptr,
1953 (unsigned long)(iref + 1), len);
1954
1955 /* Check the read+exec permission of this directory */
1956 ret = btrfs_previous_item(root, path, dirid,
1957 BTRFS_INODE_ITEM_KEY);
1958 if (ret < 0) {
1959 goto out_put;
1960 } else if (ret > 0) {
1961 ret = -ENOENT;
1962 goto out_put;
1963 }
1964
1965 leaf = path->nodes[0];
1966 slot = path->slots[0];
1967 btrfs_item_key_to_cpu(leaf, &key2, slot);
1968 if (key2.objectid != dirid) {
1969 ret = -ENOENT;
1970 goto out_put;
1971 }
1972
1973 /*
1974 * We don't need the path anymore, so release it and
1975 * avoid deadlocks and lockdep warnings in case
1976 * btrfs_iget() needs to lookup the inode from its root
1977 * btree and lock the same leaf.
1978 */
1979 btrfs_release_path(path);
1980 temp_inode = btrfs_iget(sb, key2.objectid, root);
1981 if (IS_ERR(temp_inode)) {
1982 ret = PTR_ERR(temp_inode);
1983 goto out_put;
1984 }
1985 ret = inode_permission(idmap, temp_inode,
1986 MAY_READ | MAY_EXEC);
1987 iput(temp_inode);
1988 if (ret) {
1989 ret = -EACCES;
1990 goto out_put;
1991 }
1992
1993 if (key.offset == upper_limit.objectid)
1994 break;
1995 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
1996 ret = -EACCES;
1997 goto out_put;
1998 }
1999
2000 key.objectid = key.offset;
2001 key.offset = (u64)-1;
2002 dirid = key.objectid;
2003 }
2004
2005 memmove(args->path, ptr, total_len);
2006 args->path[total_len] = '\0';
2007 btrfs_put_root(root);
2008 root = NULL;
2009 btrfs_release_path(path);
2010 }
2011
2012 /* Get the bottom subvolume's name from ROOT_REF */
2013 key.objectid = treeid;
2014 key.type = BTRFS_ROOT_REF_KEY;
2015 key.offset = args->treeid;
2016 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2017 if (ret < 0) {
2018 goto out;
2019 } else if (ret > 0) {
2020 ret = -ENOENT;
2021 goto out;
2022 }
2023
2024 leaf = path->nodes[0];
2025 slot = path->slots[0];
2026 btrfs_item_key_to_cpu(leaf, &key, slot);
2027
2028 item_off = btrfs_item_ptr_offset(leaf, slot);
2029 item_len = btrfs_item_size(leaf, slot);
2030 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2031 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2032 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2033 ret = -EINVAL;
2034 goto out;
2035 }
2036
2037 /* Copy subvolume's name */
2038 item_off += sizeof(struct btrfs_root_ref);
2039 item_len -= sizeof(struct btrfs_root_ref);
2040 read_extent_buffer(leaf, args->name, item_off, item_len);
2041 args->name[item_len] = 0;
2042
2043out_put:
2044 btrfs_put_root(root);
2045out:
2046 btrfs_free_path(path);
2047 return ret;
2048}
2049
2050static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root,
2051 void __user *argp)
2052{
2053 struct btrfs_ioctl_ino_lookup_args *args;
2054 int ret = 0;
2055
2056 args = memdup_user(argp, sizeof(*args));
2057 if (IS_ERR(args))
2058 return PTR_ERR(args);
2059
2060 /*
2061 * Unprivileged query to obtain the containing subvolume root id. The
2062 * path is reset so it's consistent with btrfs_search_path_in_tree.
2063 */
2064 if (args->treeid == 0)
2065 args->treeid = root->root_key.objectid;
2066
2067 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2068 args->name[0] = 0;
2069 goto out;
2070 }
2071
2072 if (!capable(CAP_SYS_ADMIN)) {
2073 ret = -EPERM;
2074 goto out;
2075 }
2076
2077 ret = btrfs_search_path_in_tree(root->fs_info,
2078 args->treeid, args->objectid,
2079 args->name);
2080
2081out:
2082 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2083 ret = -EFAULT;
2084
2085 kfree(args);
2086 return ret;
2087}
2088
2089/*
2090 * Version of ino_lookup ioctl (unprivileged)
2091 *
2092 * The main differences from ino_lookup ioctl are:
2093 *
2094 * 1. Read + Exec permission will be checked using inode_permission() during
2095 * path construction. -EACCES will be returned in case of failure.
2096 * 2. Path construction will be stopped at the inode number which corresponds
2097 * to the fd with which this ioctl is called. If constructed path does not
2098 * exist under fd's inode, -EACCES will be returned.
2099 * 3. The name of bottom subvolume is also searched and filled.
2100 */
2101static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2102{
2103 struct btrfs_ioctl_ino_lookup_user_args *args;
2104 struct inode *inode;
2105 int ret;
2106
2107 args = memdup_user(argp, sizeof(*args));
2108 if (IS_ERR(args))
2109 return PTR_ERR(args);
2110
2111 inode = file_inode(file);
2112
2113 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2114 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2115 /*
2116 * The subvolume does not exist under fd with which this is
2117 * called
2118 */
2119 kfree(args);
2120 return -EACCES;
2121 }
2122
2123 ret = btrfs_search_path_in_tree_user(file_mnt_idmap(file), inode, args);
2124
2125 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2126 ret = -EFAULT;
2127
2128 kfree(args);
2129 return ret;
2130}
2131
2132/* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2133static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp)
2134{
2135 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2136 struct btrfs_fs_info *fs_info;
2137 struct btrfs_root *root;
2138 struct btrfs_path *path;
2139 struct btrfs_key key;
2140 struct btrfs_root_item *root_item;
2141 struct btrfs_root_ref *rref;
2142 struct extent_buffer *leaf;
2143 unsigned long item_off;
2144 unsigned long item_len;
2145 int slot;
2146 int ret = 0;
2147
2148 path = btrfs_alloc_path();
2149 if (!path)
2150 return -ENOMEM;
2151
2152 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2153 if (!subvol_info) {
2154 btrfs_free_path(path);
2155 return -ENOMEM;
2156 }
2157
2158 fs_info = BTRFS_I(inode)->root->fs_info;
2159
2160 /* Get root_item of inode's subvolume */
2161 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2162 root = btrfs_get_fs_root(fs_info, key.objectid, true);
2163 if (IS_ERR(root)) {
2164 ret = PTR_ERR(root);
2165 goto out_free;
2166 }
2167 root_item = &root->root_item;
2168
2169 subvol_info->treeid = key.objectid;
2170
2171 subvol_info->generation = btrfs_root_generation(root_item);
2172 subvol_info->flags = btrfs_root_flags(root_item);
2173
2174 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2175 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2176 BTRFS_UUID_SIZE);
2177 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2178 BTRFS_UUID_SIZE);
2179
2180 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2181 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2182 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2183
2184 subvol_info->otransid = btrfs_root_otransid(root_item);
2185 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2186 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2187
2188 subvol_info->stransid = btrfs_root_stransid(root_item);
2189 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2190 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2191
2192 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2193 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2194 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2195
2196 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2197 /* Search root tree for ROOT_BACKREF of this subvolume */
2198 key.type = BTRFS_ROOT_BACKREF_KEY;
2199 key.offset = 0;
2200 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2201 if (ret < 0) {
2202 goto out;
2203 } else if (path->slots[0] >=
2204 btrfs_header_nritems(path->nodes[0])) {
2205 ret = btrfs_next_leaf(fs_info->tree_root, path);
2206 if (ret < 0) {
2207 goto out;
2208 } else if (ret > 0) {
2209 ret = -EUCLEAN;
2210 goto out;
2211 }
2212 }
2213
2214 leaf = path->nodes[0];
2215 slot = path->slots[0];
2216 btrfs_item_key_to_cpu(leaf, &key, slot);
2217 if (key.objectid == subvol_info->treeid &&
2218 key.type == BTRFS_ROOT_BACKREF_KEY) {
2219 subvol_info->parent_id = key.offset;
2220
2221 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2222 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2223
2224 item_off = btrfs_item_ptr_offset(leaf, slot)
2225 + sizeof(struct btrfs_root_ref);
2226 item_len = btrfs_item_size(leaf, slot)
2227 - sizeof(struct btrfs_root_ref);
2228 read_extent_buffer(leaf, subvol_info->name,
2229 item_off, item_len);
2230 } else {
2231 ret = -ENOENT;
2232 goto out;
2233 }
2234 }
2235
2236 btrfs_free_path(path);
2237 path = NULL;
2238 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2239 ret = -EFAULT;
2240
2241out:
2242 btrfs_put_root(root);
2243out_free:
2244 btrfs_free_path(path);
2245 kfree(subvol_info);
2246 return ret;
2247}
2248
2249/*
2250 * Return ROOT_REF information of the subvolume containing this inode
2251 * except the subvolume name.
2252 */
2253static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root,
2254 void __user *argp)
2255{
2256 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2257 struct btrfs_root_ref *rref;
2258 struct btrfs_path *path;
2259 struct btrfs_key key;
2260 struct extent_buffer *leaf;
2261 u64 objectid;
2262 int slot;
2263 int ret;
2264 u8 found;
2265
2266 path = btrfs_alloc_path();
2267 if (!path)
2268 return -ENOMEM;
2269
2270 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2271 if (IS_ERR(rootrefs)) {
2272 btrfs_free_path(path);
2273 return PTR_ERR(rootrefs);
2274 }
2275
2276 objectid = root->root_key.objectid;
2277 key.objectid = objectid;
2278 key.type = BTRFS_ROOT_REF_KEY;
2279 key.offset = rootrefs->min_treeid;
2280 found = 0;
2281
2282 root = root->fs_info->tree_root;
2283 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2284 if (ret < 0) {
2285 goto out;
2286 } else if (path->slots[0] >=
2287 btrfs_header_nritems(path->nodes[0])) {
2288 ret = btrfs_next_leaf(root, path);
2289 if (ret < 0) {
2290 goto out;
2291 } else if (ret > 0) {
2292 ret = -EUCLEAN;
2293 goto out;
2294 }
2295 }
2296 while (1) {
2297 leaf = path->nodes[0];
2298 slot = path->slots[0];
2299
2300 btrfs_item_key_to_cpu(leaf, &key, slot);
2301 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2302 ret = 0;
2303 goto out;
2304 }
2305
2306 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2307 ret = -EOVERFLOW;
2308 goto out;
2309 }
2310
2311 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2312 rootrefs->rootref[found].treeid = key.offset;
2313 rootrefs->rootref[found].dirid =
2314 btrfs_root_ref_dirid(leaf, rref);
2315 found++;
2316
2317 ret = btrfs_next_item(root, path);
2318 if (ret < 0) {
2319 goto out;
2320 } else if (ret > 0) {
2321 ret = -EUCLEAN;
2322 goto out;
2323 }
2324 }
2325
2326out:
2327 btrfs_free_path(path);
2328
2329 if (!ret || ret == -EOVERFLOW) {
2330 rootrefs->num_items = found;
2331 /* update min_treeid for next search */
2332 if (found)
2333 rootrefs->min_treeid =
2334 rootrefs->rootref[found - 1].treeid + 1;
2335 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2336 ret = -EFAULT;
2337 }
2338
2339 kfree(rootrefs);
2340
2341 return ret;
2342}
2343
2344static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2345 void __user *arg,
2346 bool destroy_v2)
2347{
2348 struct dentry *parent = file->f_path.dentry;
2349 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2350 struct dentry *dentry;
2351 struct inode *dir = d_inode(parent);
2352 struct inode *inode;
2353 struct btrfs_root *root = BTRFS_I(dir)->root;
2354 struct btrfs_root *dest = NULL;
2355 struct btrfs_ioctl_vol_args *vol_args = NULL;
2356 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2357 struct mnt_idmap *idmap = file_mnt_idmap(file);
2358 char *subvol_name, *subvol_name_ptr = NULL;
2359 int subvol_namelen;
2360 int err = 0;
2361 bool destroy_parent = false;
2362
2363 /* We don't support snapshots with extent tree v2 yet. */
2364 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2365 btrfs_err(fs_info,
2366 "extent tree v2 doesn't support snapshot deletion yet");
2367 return -EOPNOTSUPP;
2368 }
2369
2370 if (destroy_v2) {
2371 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2372 if (IS_ERR(vol_args2))
2373 return PTR_ERR(vol_args2);
2374
2375 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2376 err = -EOPNOTSUPP;
2377 goto out;
2378 }
2379
2380 /*
2381 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2382 * name, same as v1 currently does.
2383 */
2384 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2385 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2386 subvol_name = vol_args2->name;
2387
2388 err = mnt_want_write_file(file);
2389 if (err)
2390 goto out;
2391 } else {
2392 struct inode *old_dir;
2393
2394 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2395 err = -EINVAL;
2396 goto out;
2397 }
2398
2399 err = mnt_want_write_file(file);
2400 if (err)
2401 goto out;
2402
2403 dentry = btrfs_get_dentry(fs_info->sb,
2404 BTRFS_FIRST_FREE_OBJECTID,
2405 vol_args2->subvolid, 0);
2406 if (IS_ERR(dentry)) {
2407 err = PTR_ERR(dentry);
2408 goto out_drop_write;
2409 }
2410
2411 /*
2412 * Change the default parent since the subvolume being
2413 * deleted can be outside of the current mount point.
2414 */
2415 parent = btrfs_get_parent(dentry);
2416
2417 /*
2418 * At this point dentry->d_name can point to '/' if the
2419 * subvolume we want to destroy is outsite of the
2420 * current mount point, so we need to release the
2421 * current dentry and execute the lookup to return a new
2422 * one with ->d_name pointing to the
2423 * <mount point>/subvol_name.
2424 */
2425 dput(dentry);
2426 if (IS_ERR(parent)) {
2427 err = PTR_ERR(parent);
2428 goto out_drop_write;
2429 }
2430 old_dir = dir;
2431 dir = d_inode(parent);
2432
2433 /*
2434 * If v2 was used with SPEC_BY_ID, a new parent was
2435 * allocated since the subvolume can be outside of the
2436 * current mount point. Later on we need to release this
2437 * new parent dentry.
2438 */
2439 destroy_parent = true;
2440
2441 /*
2442 * On idmapped mounts, deletion via subvolid is
2443 * restricted to subvolumes that are immediate
2444 * ancestors of the inode referenced by the file
2445 * descriptor in the ioctl. Otherwise the idmapping
2446 * could potentially be abused to delete subvolumes
2447 * anywhere in the filesystem the user wouldn't be able
2448 * to delete without an idmapped mount.
2449 */
2450 if (old_dir != dir && idmap != &nop_mnt_idmap) {
2451 err = -EOPNOTSUPP;
2452 goto free_parent;
2453 }
2454
2455 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2456 fs_info, vol_args2->subvolid);
2457 if (IS_ERR(subvol_name_ptr)) {
2458 err = PTR_ERR(subvol_name_ptr);
2459 goto free_parent;
2460 }
2461 /* subvol_name_ptr is already nul terminated */
2462 subvol_name = (char *)kbasename(subvol_name_ptr);
2463 }
2464 } else {
2465 vol_args = memdup_user(arg, sizeof(*vol_args));
2466 if (IS_ERR(vol_args))
2467 return PTR_ERR(vol_args);
2468
2469 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
2470 subvol_name = vol_args->name;
2471
2472 err = mnt_want_write_file(file);
2473 if (err)
2474 goto out;
2475 }
2476
2477 subvol_namelen = strlen(subvol_name);
2478
2479 if (strchr(subvol_name, '/') ||
2480 strncmp(subvol_name, "..", subvol_namelen) == 0) {
2481 err = -EINVAL;
2482 goto free_subvol_name;
2483 }
2484
2485 if (!S_ISDIR(dir->i_mode)) {
2486 err = -ENOTDIR;
2487 goto free_subvol_name;
2488 }
2489
2490 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2491 if (err == -EINTR)
2492 goto free_subvol_name;
2493 dentry = lookup_one(idmap, subvol_name, parent, subvol_namelen);
2494 if (IS_ERR(dentry)) {
2495 err = PTR_ERR(dentry);
2496 goto out_unlock_dir;
2497 }
2498
2499 if (d_really_is_negative(dentry)) {
2500 err = -ENOENT;
2501 goto out_dput;
2502 }
2503
2504 inode = d_inode(dentry);
2505 dest = BTRFS_I(inode)->root;
2506 if (!capable(CAP_SYS_ADMIN)) {
2507 /*
2508 * Regular user. Only allow this with a special mount
2509 * option, when the user has write+exec access to the
2510 * subvol root, and when rmdir(2) would have been
2511 * allowed.
2512 *
2513 * Note that this is _not_ check that the subvol is
2514 * empty or doesn't contain data that we wouldn't
2515 * otherwise be able to delete.
2516 *
2517 * Users who want to delete empty subvols should try
2518 * rmdir(2).
2519 */
2520 err = -EPERM;
2521 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2522 goto out_dput;
2523
2524 /*
2525 * Do not allow deletion if the parent dir is the same
2526 * as the dir to be deleted. That means the ioctl
2527 * must be called on the dentry referencing the root
2528 * of the subvol, not a random directory contained
2529 * within it.
2530 */
2531 err = -EINVAL;
2532 if (root == dest)
2533 goto out_dput;
2534
2535 err = inode_permission(idmap, inode, MAY_WRITE | MAY_EXEC);
2536 if (err)
2537 goto out_dput;
2538 }
2539
2540 /* check if subvolume may be deleted by a user */
2541 err = btrfs_may_delete(idmap, dir, dentry, 1);
2542 if (err)
2543 goto out_dput;
2544
2545 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2546 err = -EINVAL;
2547 goto out_dput;
2548 }
2549
2550 btrfs_inode_lock(BTRFS_I(inode), 0);
2551 err = btrfs_delete_subvolume(BTRFS_I(dir), dentry);
2552 btrfs_inode_unlock(BTRFS_I(inode), 0);
2553 if (!err)
2554 d_delete_notify(dir, dentry);
2555
2556out_dput:
2557 dput(dentry);
2558out_unlock_dir:
2559 btrfs_inode_unlock(BTRFS_I(dir), 0);
2560free_subvol_name:
2561 kfree(subvol_name_ptr);
2562free_parent:
2563 if (destroy_parent)
2564 dput(parent);
2565out_drop_write:
2566 mnt_drop_write_file(file);
2567out:
2568 kfree(vol_args2);
2569 kfree(vol_args);
2570 return err;
2571}
2572
2573static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2574{
2575 struct inode *inode = file_inode(file);
2576 struct btrfs_root *root = BTRFS_I(inode)->root;
2577 struct btrfs_ioctl_defrag_range_args range = {0};
2578 int ret;
2579
2580 ret = mnt_want_write_file(file);
2581 if (ret)
2582 return ret;
2583
2584 if (btrfs_root_readonly(root)) {
2585 ret = -EROFS;
2586 goto out;
2587 }
2588
2589 switch (inode->i_mode & S_IFMT) {
2590 case S_IFDIR:
2591 if (!capable(CAP_SYS_ADMIN)) {
2592 ret = -EPERM;
2593 goto out;
2594 }
2595 ret = btrfs_defrag_root(root);
2596 break;
2597 case S_IFREG:
2598 /*
2599 * Note that this does not check the file descriptor for write
2600 * access. This prevents defragmenting executables that are
2601 * running and allows defrag on files open in read-only mode.
2602 */
2603 if (!capable(CAP_SYS_ADMIN) &&
2604 inode_permission(&nop_mnt_idmap, inode, MAY_WRITE)) {
2605 ret = -EPERM;
2606 goto out;
2607 }
2608
2609 if (argp) {
2610 if (copy_from_user(&range, argp, sizeof(range))) {
2611 ret = -EFAULT;
2612 goto out;
2613 }
2614 if (range.flags & ~BTRFS_DEFRAG_RANGE_FLAGS_SUPP) {
2615 ret = -EOPNOTSUPP;
2616 goto out;
2617 }
2618 /* compression requires us to start the IO */
2619 if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2620 range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
2621 range.extent_thresh = (u32)-1;
2622 }
2623 } else {
2624 /* the rest are all set to zero by kzalloc */
2625 range.len = (u64)-1;
2626 }
2627 ret = btrfs_defrag_file(file_inode(file), &file->f_ra,
2628 &range, BTRFS_OLDEST_GENERATION, 0);
2629 if (ret > 0)
2630 ret = 0;
2631 break;
2632 default:
2633 ret = -EINVAL;
2634 }
2635out:
2636 mnt_drop_write_file(file);
2637 return ret;
2638}
2639
2640static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2641{
2642 struct btrfs_ioctl_vol_args *vol_args;
2643 bool restore_op = false;
2644 int ret;
2645
2646 if (!capable(CAP_SYS_ADMIN))
2647 return -EPERM;
2648
2649 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2650 btrfs_err(fs_info, "device add not supported on extent tree v2 yet");
2651 return -EINVAL;
2652 }
2653
2654 if (fs_info->fs_devices->temp_fsid) {
2655 btrfs_err(fs_info,
2656 "device add not supported on cloned temp-fsid mount");
2657 return -EINVAL;
2658 }
2659
2660 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) {
2661 if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD))
2662 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2663
2664 /*
2665 * We can do the device add because we have a paused balanced,
2666 * change the exclusive op type and remember we should bring
2667 * back the paused balance
2668 */
2669 fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD;
2670 btrfs_exclop_start_unlock(fs_info);
2671 restore_op = true;
2672 }
2673
2674 vol_args = memdup_user(arg, sizeof(*vol_args));
2675 if (IS_ERR(vol_args)) {
2676 ret = PTR_ERR(vol_args);
2677 goto out;
2678 }
2679
2680 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2681 ret = btrfs_init_new_device(fs_info, vol_args->name);
2682
2683 if (!ret)
2684 btrfs_info(fs_info, "disk added %s", vol_args->name);
2685
2686 kfree(vol_args);
2687out:
2688 if (restore_op)
2689 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
2690 else
2691 btrfs_exclop_finish(fs_info);
2692 return ret;
2693}
2694
2695static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2696{
2697 BTRFS_DEV_LOOKUP_ARGS(args);
2698 struct inode *inode = file_inode(file);
2699 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2700 struct btrfs_ioctl_vol_args_v2 *vol_args;
2701 struct bdev_handle *bdev_handle = NULL;
2702 int ret;
2703 bool cancel = false;
2704
2705 if (!capable(CAP_SYS_ADMIN))
2706 return -EPERM;
2707
2708 vol_args = memdup_user(arg, sizeof(*vol_args));
2709 if (IS_ERR(vol_args))
2710 return PTR_ERR(vol_args);
2711
2712 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
2713 ret = -EOPNOTSUPP;
2714 goto out;
2715 }
2716
2717 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2718 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2719 args.devid = vol_args->devid;
2720 } else if (!strcmp("cancel", vol_args->name)) {
2721 cancel = true;
2722 } else {
2723 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2724 if (ret)
2725 goto out;
2726 }
2727
2728 ret = mnt_want_write_file(file);
2729 if (ret)
2730 goto out;
2731
2732 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2733 cancel);
2734 if (ret)
2735 goto err_drop;
2736
2737 /* Exclusive operation is now claimed */
2738 ret = btrfs_rm_device(fs_info, &args, &bdev_handle);
2739
2740 btrfs_exclop_finish(fs_info);
2741
2742 if (!ret) {
2743 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2744 btrfs_info(fs_info, "device deleted: id %llu",
2745 vol_args->devid);
2746 else
2747 btrfs_info(fs_info, "device deleted: %s",
2748 vol_args->name);
2749 }
2750err_drop:
2751 mnt_drop_write_file(file);
2752 if (bdev_handle)
2753 bdev_release(bdev_handle);
2754out:
2755 btrfs_put_dev_args_from_path(&args);
2756 kfree(vol_args);
2757 return ret;
2758}
2759
2760static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2761{
2762 BTRFS_DEV_LOOKUP_ARGS(args);
2763 struct inode *inode = file_inode(file);
2764 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2765 struct btrfs_ioctl_vol_args *vol_args;
2766 struct bdev_handle *bdev_handle = NULL;
2767 int ret;
2768 bool cancel = false;
2769
2770 if (!capable(CAP_SYS_ADMIN))
2771 return -EPERM;
2772
2773 vol_args = memdup_user(arg, sizeof(*vol_args));
2774 if (IS_ERR(vol_args))
2775 return PTR_ERR(vol_args);
2776
2777 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2778 if (!strcmp("cancel", vol_args->name)) {
2779 cancel = true;
2780 } else {
2781 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2782 if (ret)
2783 goto out;
2784 }
2785
2786 ret = mnt_want_write_file(file);
2787 if (ret)
2788 goto out;
2789
2790 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2791 cancel);
2792 if (ret == 0) {
2793 ret = btrfs_rm_device(fs_info, &args, &bdev_handle);
2794 if (!ret)
2795 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2796 btrfs_exclop_finish(fs_info);
2797 }
2798
2799 mnt_drop_write_file(file);
2800 if (bdev_handle)
2801 bdev_release(bdev_handle);
2802out:
2803 btrfs_put_dev_args_from_path(&args);
2804 kfree(vol_args);
2805 return ret;
2806}
2807
2808static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2809 void __user *arg)
2810{
2811 struct btrfs_ioctl_fs_info_args *fi_args;
2812 struct btrfs_device *device;
2813 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2814 u64 flags_in;
2815 int ret = 0;
2816
2817 fi_args = memdup_user(arg, sizeof(*fi_args));
2818 if (IS_ERR(fi_args))
2819 return PTR_ERR(fi_args);
2820
2821 flags_in = fi_args->flags;
2822 memset(fi_args, 0, sizeof(*fi_args));
2823
2824 rcu_read_lock();
2825 fi_args->num_devices = fs_devices->num_devices;
2826
2827 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2828 if (device->devid > fi_args->max_id)
2829 fi_args->max_id = device->devid;
2830 }
2831 rcu_read_unlock();
2832
2833 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
2834 fi_args->nodesize = fs_info->nodesize;
2835 fi_args->sectorsize = fs_info->sectorsize;
2836 fi_args->clone_alignment = fs_info->sectorsize;
2837
2838 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
2839 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
2840 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
2841 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
2842 }
2843
2844 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
2845 fi_args->generation = btrfs_get_fs_generation(fs_info);
2846 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
2847 }
2848
2849 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
2850 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
2851 sizeof(fi_args->metadata_uuid));
2852 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
2853 }
2854
2855 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2856 ret = -EFAULT;
2857
2858 kfree(fi_args);
2859 return ret;
2860}
2861
2862static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2863 void __user *arg)
2864{
2865 BTRFS_DEV_LOOKUP_ARGS(args);
2866 struct btrfs_ioctl_dev_info_args *di_args;
2867 struct btrfs_device *dev;
2868 int ret = 0;
2869
2870 di_args = memdup_user(arg, sizeof(*di_args));
2871 if (IS_ERR(di_args))
2872 return PTR_ERR(di_args);
2873
2874 args.devid = di_args->devid;
2875 if (!btrfs_is_empty_uuid(di_args->uuid))
2876 args.uuid = di_args->uuid;
2877
2878 rcu_read_lock();
2879 dev = btrfs_find_device(fs_info->fs_devices, &args);
2880 if (!dev) {
2881 ret = -ENODEV;
2882 goto out;
2883 }
2884
2885 di_args->devid = dev->devid;
2886 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2887 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2888 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2889 memcpy(di_args->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
2890 if (dev->name)
2891 strscpy(di_args->path, btrfs_dev_name(dev), sizeof(di_args->path));
2892 else
2893 di_args->path[0] = '\0';
2894
2895out:
2896 rcu_read_unlock();
2897 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2898 ret = -EFAULT;
2899
2900 kfree(di_args);
2901 return ret;
2902}
2903
2904static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2905{
2906 struct inode *inode = file_inode(file);
2907 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2908 struct btrfs_root *root = BTRFS_I(inode)->root;
2909 struct btrfs_root *new_root;
2910 struct btrfs_dir_item *di;
2911 struct btrfs_trans_handle *trans;
2912 struct btrfs_path *path = NULL;
2913 struct btrfs_disk_key disk_key;
2914 struct fscrypt_str name = FSTR_INIT("default", 7);
2915 u64 objectid = 0;
2916 u64 dir_id;
2917 int ret;
2918
2919 if (!capable(CAP_SYS_ADMIN))
2920 return -EPERM;
2921
2922 ret = mnt_want_write_file(file);
2923 if (ret)
2924 return ret;
2925
2926 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2927 ret = -EFAULT;
2928 goto out;
2929 }
2930
2931 if (!objectid)
2932 objectid = BTRFS_FS_TREE_OBJECTID;
2933
2934 new_root = btrfs_get_fs_root(fs_info, objectid, true);
2935 if (IS_ERR(new_root)) {
2936 ret = PTR_ERR(new_root);
2937 goto out;
2938 }
2939 if (!is_fstree(new_root->root_key.objectid)) {
2940 ret = -ENOENT;
2941 goto out_free;
2942 }
2943
2944 path = btrfs_alloc_path();
2945 if (!path) {
2946 ret = -ENOMEM;
2947 goto out_free;
2948 }
2949
2950 trans = btrfs_start_transaction(root, 1);
2951 if (IS_ERR(trans)) {
2952 ret = PTR_ERR(trans);
2953 goto out_free;
2954 }
2955
2956 dir_id = btrfs_super_root_dir(fs_info->super_copy);
2957 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
2958 dir_id, &name, 1);
2959 if (IS_ERR_OR_NULL(di)) {
2960 btrfs_release_path(path);
2961 btrfs_end_transaction(trans);
2962 btrfs_err(fs_info,
2963 "Umm, you don't have the default diritem, this isn't going to work");
2964 ret = -ENOENT;
2965 goto out_free;
2966 }
2967
2968 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2969 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2970 btrfs_mark_buffer_dirty(trans, path->nodes[0]);
2971 btrfs_release_path(path);
2972
2973 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
2974 btrfs_end_transaction(trans);
2975out_free:
2976 btrfs_put_root(new_root);
2977 btrfs_free_path(path);
2978out:
2979 mnt_drop_write_file(file);
2980 return ret;
2981}
2982
2983static void get_block_group_info(struct list_head *groups_list,
2984 struct btrfs_ioctl_space_info *space)
2985{
2986 struct btrfs_block_group *block_group;
2987
2988 space->total_bytes = 0;
2989 space->used_bytes = 0;
2990 space->flags = 0;
2991 list_for_each_entry(block_group, groups_list, list) {
2992 space->flags = block_group->flags;
2993 space->total_bytes += block_group->length;
2994 space->used_bytes += block_group->used;
2995 }
2996}
2997
2998static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
2999 void __user *arg)
3000{
3001 struct btrfs_ioctl_space_args space_args = { 0 };
3002 struct btrfs_ioctl_space_info space;
3003 struct btrfs_ioctl_space_info *dest;
3004 struct btrfs_ioctl_space_info *dest_orig;
3005 struct btrfs_ioctl_space_info __user *user_dest;
3006 struct btrfs_space_info *info;
3007 static const u64 types[] = {
3008 BTRFS_BLOCK_GROUP_DATA,
3009 BTRFS_BLOCK_GROUP_SYSTEM,
3010 BTRFS_BLOCK_GROUP_METADATA,
3011 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3012 };
3013 int num_types = 4;
3014 int alloc_size;
3015 int ret = 0;
3016 u64 slot_count = 0;
3017 int i, c;
3018
3019 if (copy_from_user(&space_args,
3020 (struct btrfs_ioctl_space_args __user *)arg,
3021 sizeof(space_args)))
3022 return -EFAULT;
3023
3024 for (i = 0; i < num_types; i++) {
3025 struct btrfs_space_info *tmp;
3026
3027 info = NULL;
3028 list_for_each_entry(tmp, &fs_info->space_info, list) {
3029 if (tmp->flags == types[i]) {
3030 info = tmp;
3031 break;
3032 }
3033 }
3034
3035 if (!info)
3036 continue;
3037
3038 down_read(&info->groups_sem);
3039 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3040 if (!list_empty(&info->block_groups[c]))
3041 slot_count++;
3042 }
3043 up_read(&info->groups_sem);
3044 }
3045
3046 /*
3047 * Global block reserve, exported as a space_info
3048 */
3049 slot_count++;
3050
3051 /* space_slots == 0 means they are asking for a count */
3052 if (space_args.space_slots == 0) {
3053 space_args.total_spaces = slot_count;
3054 goto out;
3055 }
3056
3057 slot_count = min_t(u64, space_args.space_slots, slot_count);
3058
3059 alloc_size = sizeof(*dest) * slot_count;
3060
3061 /* we generally have at most 6 or so space infos, one for each raid
3062 * level. So, a whole page should be more than enough for everyone
3063 */
3064 if (alloc_size > PAGE_SIZE)
3065 return -ENOMEM;
3066
3067 space_args.total_spaces = 0;
3068 dest = kmalloc(alloc_size, GFP_KERNEL);
3069 if (!dest)
3070 return -ENOMEM;
3071 dest_orig = dest;
3072
3073 /* now we have a buffer to copy into */
3074 for (i = 0; i < num_types; i++) {
3075 struct btrfs_space_info *tmp;
3076
3077 if (!slot_count)
3078 break;
3079
3080 info = NULL;
3081 list_for_each_entry(tmp, &fs_info->space_info, list) {
3082 if (tmp->flags == types[i]) {
3083 info = tmp;
3084 break;
3085 }
3086 }
3087
3088 if (!info)
3089 continue;
3090 down_read(&info->groups_sem);
3091 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3092 if (!list_empty(&info->block_groups[c])) {
3093 get_block_group_info(&info->block_groups[c],
3094 &space);
3095 memcpy(dest, &space, sizeof(space));
3096 dest++;
3097 space_args.total_spaces++;
3098 slot_count--;
3099 }
3100 if (!slot_count)
3101 break;
3102 }
3103 up_read(&info->groups_sem);
3104 }
3105
3106 /*
3107 * Add global block reserve
3108 */
3109 if (slot_count) {
3110 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3111
3112 spin_lock(&block_rsv->lock);
3113 space.total_bytes = block_rsv->size;
3114 space.used_bytes = block_rsv->size - block_rsv->reserved;
3115 spin_unlock(&block_rsv->lock);
3116 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3117 memcpy(dest, &space, sizeof(space));
3118 space_args.total_spaces++;
3119 }
3120
3121 user_dest = (struct btrfs_ioctl_space_info __user *)
3122 (arg + sizeof(struct btrfs_ioctl_space_args));
3123
3124 if (copy_to_user(user_dest, dest_orig, alloc_size))
3125 ret = -EFAULT;
3126
3127 kfree(dest_orig);
3128out:
3129 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3130 ret = -EFAULT;
3131
3132 return ret;
3133}
3134
3135static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3136 void __user *argp)
3137{
3138 struct btrfs_trans_handle *trans;
3139 u64 transid;
3140
3141 /*
3142 * Start orphan cleanup here for the given root in case it hasn't been
3143 * started already by other means. Errors are handled in the other
3144 * functions during transaction commit.
3145 */
3146 btrfs_orphan_cleanup(root);
3147
3148 trans = btrfs_attach_transaction_barrier(root);
3149 if (IS_ERR(trans)) {
3150 if (PTR_ERR(trans) != -ENOENT)
3151 return PTR_ERR(trans);
3152
3153 /* No running transaction, don't bother */
3154 transid = btrfs_get_last_trans_committed(root->fs_info);
3155 goto out;
3156 }
3157 transid = trans->transid;
3158 btrfs_commit_transaction_async(trans);
3159out:
3160 if (argp)
3161 if (copy_to_user(argp, &transid, sizeof(transid)))
3162 return -EFAULT;
3163 return 0;
3164}
3165
3166static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3167 void __user *argp)
3168{
3169 /* By default wait for the current transaction. */
3170 u64 transid = 0;
3171
3172 if (argp)
3173 if (copy_from_user(&transid, argp, sizeof(transid)))
3174 return -EFAULT;
3175
3176 return btrfs_wait_for_commit(fs_info, transid);
3177}
3178
3179static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3180{
3181 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3182 struct btrfs_ioctl_scrub_args *sa;
3183 int ret;
3184
3185 if (!capable(CAP_SYS_ADMIN))
3186 return -EPERM;
3187
3188 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3189 btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet");
3190 return -EINVAL;
3191 }
3192
3193 sa = memdup_user(arg, sizeof(*sa));
3194 if (IS_ERR(sa))
3195 return PTR_ERR(sa);
3196
3197 if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) {
3198 ret = -EOPNOTSUPP;
3199 goto out;
3200 }
3201
3202 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3203 ret = mnt_want_write_file(file);
3204 if (ret)
3205 goto out;
3206 }
3207
3208 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3209 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3210 0);
3211
3212 /*
3213 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3214 * error. This is important as it allows user space to know how much
3215 * progress scrub has done. For example, if scrub is canceled we get
3216 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3217 * space. Later user space can inspect the progress from the structure
3218 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3219 * previously (btrfs-progs does this).
3220 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3221 * then return -EFAULT to signal the structure was not copied or it may
3222 * be corrupt and unreliable due to a partial copy.
3223 */
3224 if (copy_to_user(arg, sa, sizeof(*sa)))
3225 ret = -EFAULT;
3226
3227 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3228 mnt_drop_write_file(file);
3229out:
3230 kfree(sa);
3231 return ret;
3232}
3233
3234static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3235{
3236 if (!capable(CAP_SYS_ADMIN))
3237 return -EPERM;
3238
3239 return btrfs_scrub_cancel(fs_info);
3240}
3241
3242static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3243 void __user *arg)
3244{
3245 struct btrfs_ioctl_scrub_args *sa;
3246 int ret;
3247
3248 if (!capable(CAP_SYS_ADMIN))
3249 return -EPERM;
3250
3251 sa = memdup_user(arg, sizeof(*sa));
3252 if (IS_ERR(sa))
3253 return PTR_ERR(sa);
3254
3255 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3256
3257 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3258 ret = -EFAULT;
3259
3260 kfree(sa);
3261 return ret;
3262}
3263
3264static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3265 void __user *arg)
3266{
3267 struct btrfs_ioctl_get_dev_stats *sa;
3268 int ret;
3269
3270 sa = memdup_user(arg, sizeof(*sa));
3271 if (IS_ERR(sa))
3272 return PTR_ERR(sa);
3273
3274 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3275 kfree(sa);
3276 return -EPERM;
3277 }
3278
3279 ret = btrfs_get_dev_stats(fs_info, sa);
3280
3281 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3282 ret = -EFAULT;
3283
3284 kfree(sa);
3285 return ret;
3286}
3287
3288static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3289 void __user *arg)
3290{
3291 struct btrfs_ioctl_dev_replace_args *p;
3292 int ret;
3293
3294 if (!capable(CAP_SYS_ADMIN))
3295 return -EPERM;
3296
3297 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3298 btrfs_err(fs_info, "device replace not supported on extent tree v2 yet");
3299 return -EINVAL;
3300 }
3301
3302 p = memdup_user(arg, sizeof(*p));
3303 if (IS_ERR(p))
3304 return PTR_ERR(p);
3305
3306 switch (p->cmd) {
3307 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3308 if (sb_rdonly(fs_info->sb)) {
3309 ret = -EROFS;
3310 goto out;
3311 }
3312 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3313 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3314 } else {
3315 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3316 btrfs_exclop_finish(fs_info);
3317 }
3318 break;
3319 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3320 btrfs_dev_replace_status(fs_info, p);
3321 ret = 0;
3322 break;
3323 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3324 p->result = btrfs_dev_replace_cancel(fs_info);
3325 ret = 0;
3326 break;
3327 default:
3328 ret = -EINVAL;
3329 break;
3330 }
3331
3332 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3333 ret = -EFAULT;
3334out:
3335 kfree(p);
3336 return ret;
3337}
3338
3339static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3340{
3341 int ret = 0;
3342 int i;
3343 u64 rel_ptr;
3344 int size;
3345 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3346 struct inode_fs_paths *ipath = NULL;
3347 struct btrfs_path *path;
3348
3349 if (!capable(CAP_DAC_READ_SEARCH))
3350 return -EPERM;
3351
3352 path = btrfs_alloc_path();
3353 if (!path) {
3354 ret = -ENOMEM;
3355 goto out;
3356 }
3357
3358 ipa = memdup_user(arg, sizeof(*ipa));
3359 if (IS_ERR(ipa)) {
3360 ret = PTR_ERR(ipa);
3361 ipa = NULL;
3362 goto out;
3363 }
3364
3365 size = min_t(u32, ipa->size, 4096);
3366 ipath = init_ipath(size, root, path);
3367 if (IS_ERR(ipath)) {
3368 ret = PTR_ERR(ipath);
3369 ipath = NULL;
3370 goto out;
3371 }
3372
3373 ret = paths_from_inode(ipa->inum, ipath);
3374 if (ret < 0)
3375 goto out;
3376
3377 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3378 rel_ptr = ipath->fspath->val[i] -
3379 (u64)(unsigned long)ipath->fspath->val;
3380 ipath->fspath->val[i] = rel_ptr;
3381 }
3382
3383 btrfs_free_path(path);
3384 path = NULL;
3385 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3386 ipath->fspath, size);
3387 if (ret) {
3388 ret = -EFAULT;
3389 goto out;
3390 }
3391
3392out:
3393 btrfs_free_path(path);
3394 free_ipath(ipath);
3395 kfree(ipa);
3396
3397 return ret;
3398}
3399
3400static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3401 void __user *arg, int version)
3402{
3403 int ret = 0;
3404 int size;
3405 struct btrfs_ioctl_logical_ino_args *loi;
3406 struct btrfs_data_container *inodes = NULL;
3407 struct btrfs_path *path = NULL;
3408 bool ignore_offset;
3409
3410 if (!capable(CAP_SYS_ADMIN))
3411 return -EPERM;
3412
3413 loi = memdup_user(arg, sizeof(*loi));
3414 if (IS_ERR(loi))
3415 return PTR_ERR(loi);
3416
3417 if (version == 1) {
3418 ignore_offset = false;
3419 size = min_t(u32, loi->size, SZ_64K);
3420 } else {
3421 /* All reserved bits must be 0 for now */
3422 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3423 ret = -EINVAL;
3424 goto out_loi;
3425 }
3426 /* Only accept flags we have defined so far */
3427 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3428 ret = -EINVAL;
3429 goto out_loi;
3430 }
3431 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3432 size = min_t(u32, loi->size, SZ_16M);
3433 }
3434
3435 inodes = init_data_container(size);
3436 if (IS_ERR(inodes)) {
3437 ret = PTR_ERR(inodes);
3438 goto out_loi;
3439 }
3440
3441 path = btrfs_alloc_path();
3442 if (!path) {
3443 ret = -ENOMEM;
3444 goto out;
3445 }
3446 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3447 inodes, ignore_offset);
3448 btrfs_free_path(path);
3449 if (ret == -EINVAL)
3450 ret = -ENOENT;
3451 if (ret < 0)
3452 goto out;
3453
3454 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3455 size);
3456 if (ret)
3457 ret = -EFAULT;
3458
3459out:
3460 kvfree(inodes);
3461out_loi:
3462 kfree(loi);
3463
3464 return ret;
3465}
3466
3467void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3468 struct btrfs_ioctl_balance_args *bargs)
3469{
3470 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3471
3472 bargs->flags = bctl->flags;
3473
3474 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3475 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3476 if (atomic_read(&fs_info->balance_pause_req))
3477 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3478 if (atomic_read(&fs_info->balance_cancel_req))
3479 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3480
3481 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3482 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3483 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3484
3485 spin_lock(&fs_info->balance_lock);
3486 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3487 spin_unlock(&fs_info->balance_lock);
3488}
3489
3490/*
3491 * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as
3492 * required.
3493 *
3494 * @fs_info: the filesystem
3495 * @excl_acquired: ptr to boolean value which is set to false in case balance
3496 * is being resumed
3497 *
3498 * Return 0 on success in which case both fs_info::balance is acquired as well
3499 * as exclusive ops are blocked. In case of failure return an error code.
3500 */
3501static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired)
3502{
3503 int ret;
3504
3505 /*
3506 * Exclusive operation is locked. Three possibilities:
3507 * (1) some other op is running
3508 * (2) balance is running
3509 * (3) balance is paused -- special case (think resume)
3510 */
3511 while (1) {
3512 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
3513 *excl_acquired = true;
3514 mutex_lock(&fs_info->balance_mutex);
3515 return 0;
3516 }
3517
3518 mutex_lock(&fs_info->balance_mutex);
3519 if (fs_info->balance_ctl) {
3520 /* This is either (2) or (3) */
3521 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3522 /* This is (2) */
3523 ret = -EINPROGRESS;
3524 goto out_failure;
3525
3526 } else {
3527 mutex_unlock(&fs_info->balance_mutex);
3528 /*
3529 * Lock released to allow other waiters to
3530 * continue, we'll reexamine the status again.
3531 */
3532 mutex_lock(&fs_info->balance_mutex);
3533
3534 if (fs_info->balance_ctl &&
3535 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3536 /* This is (3) */
3537 *excl_acquired = false;
3538 return 0;
3539 }
3540 }
3541 } else {
3542 /* This is (1) */
3543 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3544 goto out_failure;
3545 }
3546
3547 mutex_unlock(&fs_info->balance_mutex);
3548 }
3549
3550out_failure:
3551 mutex_unlock(&fs_info->balance_mutex);
3552 *excl_acquired = false;
3553 return ret;
3554}
3555
3556static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3557{
3558 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3559 struct btrfs_fs_info *fs_info = root->fs_info;
3560 struct btrfs_ioctl_balance_args *bargs;
3561 struct btrfs_balance_control *bctl;
3562 bool need_unlock = true;
3563 int ret;
3564
3565 if (!capable(CAP_SYS_ADMIN))
3566 return -EPERM;
3567
3568 ret = mnt_want_write_file(file);
3569 if (ret)
3570 return ret;
3571
3572 bargs = memdup_user(arg, sizeof(*bargs));
3573 if (IS_ERR(bargs)) {
3574 ret = PTR_ERR(bargs);
3575 bargs = NULL;
3576 goto out;
3577 }
3578
3579 ret = btrfs_try_lock_balance(fs_info, &need_unlock);
3580 if (ret)
3581 goto out;
3582
3583 lockdep_assert_held(&fs_info->balance_mutex);
3584
3585 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3586 if (!fs_info->balance_ctl) {
3587 ret = -ENOTCONN;
3588 goto out_unlock;
3589 }
3590
3591 bctl = fs_info->balance_ctl;
3592 spin_lock(&fs_info->balance_lock);
3593 bctl->flags |= BTRFS_BALANCE_RESUME;
3594 spin_unlock(&fs_info->balance_lock);
3595 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE);
3596
3597 goto do_balance;
3598 }
3599
3600 if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3601 ret = -EINVAL;
3602 goto out_unlock;
3603 }
3604
3605 if (fs_info->balance_ctl) {
3606 ret = -EINPROGRESS;
3607 goto out_unlock;
3608 }
3609
3610 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
3611 if (!bctl) {
3612 ret = -ENOMEM;
3613 goto out_unlock;
3614 }
3615
3616 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3617 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3618 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3619
3620 bctl->flags = bargs->flags;
3621do_balance:
3622 /*
3623 * Ownership of bctl and exclusive operation goes to btrfs_balance.
3624 * bctl is freed in reset_balance_state, or, if restriper was paused
3625 * all the way until unmount, in free_fs_info. The flag should be
3626 * cleared after reset_balance_state.
3627 */
3628 need_unlock = false;
3629
3630 ret = btrfs_balance(fs_info, bctl, bargs);
3631 bctl = NULL;
3632
3633 if (ret == 0 || ret == -ECANCELED) {
3634 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3635 ret = -EFAULT;
3636 }
3637
3638 kfree(bctl);
3639out_unlock:
3640 mutex_unlock(&fs_info->balance_mutex);
3641 if (need_unlock)
3642 btrfs_exclop_finish(fs_info);
3643out:
3644 mnt_drop_write_file(file);
3645 kfree(bargs);
3646 return ret;
3647}
3648
3649static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
3650{
3651 if (!capable(CAP_SYS_ADMIN))
3652 return -EPERM;
3653
3654 switch (cmd) {
3655 case BTRFS_BALANCE_CTL_PAUSE:
3656 return btrfs_pause_balance(fs_info);
3657 case BTRFS_BALANCE_CTL_CANCEL:
3658 return btrfs_cancel_balance(fs_info);
3659 }
3660
3661 return -EINVAL;
3662}
3663
3664static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
3665 void __user *arg)
3666{
3667 struct btrfs_ioctl_balance_args *bargs;
3668 int ret = 0;
3669
3670 if (!capable(CAP_SYS_ADMIN))
3671 return -EPERM;
3672
3673 mutex_lock(&fs_info->balance_mutex);
3674 if (!fs_info->balance_ctl) {
3675 ret = -ENOTCONN;
3676 goto out;
3677 }
3678
3679 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
3680 if (!bargs) {
3681 ret = -ENOMEM;
3682 goto out;
3683 }
3684
3685 btrfs_update_ioctl_balance_args(fs_info, bargs);
3686
3687 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3688 ret = -EFAULT;
3689
3690 kfree(bargs);
3691out:
3692 mutex_unlock(&fs_info->balance_mutex);
3693 return ret;
3694}
3695
3696static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3697{
3698 struct inode *inode = file_inode(file);
3699 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3700 struct btrfs_ioctl_quota_ctl_args *sa;
3701 int ret;
3702
3703 if (!capable(CAP_SYS_ADMIN))
3704 return -EPERM;
3705
3706 ret = mnt_want_write_file(file);
3707 if (ret)
3708 return ret;
3709
3710 sa = memdup_user(arg, sizeof(*sa));
3711 if (IS_ERR(sa)) {
3712 ret = PTR_ERR(sa);
3713 goto drop_write;
3714 }
3715
3716 down_write(&fs_info->subvol_sem);
3717
3718 switch (sa->cmd) {
3719 case BTRFS_QUOTA_CTL_ENABLE:
3720 case BTRFS_QUOTA_CTL_ENABLE_SIMPLE_QUOTA:
3721 ret = btrfs_quota_enable(fs_info, sa);
3722 break;
3723 case BTRFS_QUOTA_CTL_DISABLE:
3724 ret = btrfs_quota_disable(fs_info);
3725 break;
3726 default:
3727 ret = -EINVAL;
3728 break;
3729 }
3730
3731 kfree(sa);
3732 up_write(&fs_info->subvol_sem);
3733drop_write:
3734 mnt_drop_write_file(file);
3735 return ret;
3736}
3737
3738static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3739{
3740 struct inode *inode = file_inode(file);
3741 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3742 struct btrfs_root *root = BTRFS_I(inode)->root;
3743 struct btrfs_ioctl_qgroup_assign_args *sa;
3744 struct btrfs_trans_handle *trans;
3745 int ret;
3746 int err;
3747
3748 if (!capable(CAP_SYS_ADMIN))
3749 return -EPERM;
3750
3751 ret = mnt_want_write_file(file);
3752 if (ret)
3753 return ret;
3754
3755 sa = memdup_user(arg, sizeof(*sa));
3756 if (IS_ERR(sa)) {
3757 ret = PTR_ERR(sa);
3758 goto drop_write;
3759 }
3760
3761 trans = btrfs_join_transaction(root);
3762 if (IS_ERR(trans)) {
3763 ret = PTR_ERR(trans);
3764 goto out;
3765 }
3766
3767 if (sa->assign) {
3768 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
3769 } else {
3770 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
3771 }
3772
3773 /* update qgroup status and info */
3774 mutex_lock(&fs_info->qgroup_ioctl_lock);
3775 err = btrfs_run_qgroups(trans);
3776 mutex_unlock(&fs_info->qgroup_ioctl_lock);
3777 if (err < 0)
3778 btrfs_handle_fs_error(fs_info, err,
3779 "failed to update qgroup status and info");
3780 err = btrfs_end_transaction(trans);
3781 if (err && !ret)
3782 ret = err;
3783
3784out:
3785 kfree(sa);
3786drop_write:
3787 mnt_drop_write_file(file);
3788 return ret;
3789}
3790
3791static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3792{
3793 struct inode *inode = file_inode(file);
3794 struct btrfs_root *root = BTRFS_I(inode)->root;
3795 struct btrfs_ioctl_qgroup_create_args *sa;
3796 struct btrfs_trans_handle *trans;
3797 int ret;
3798 int err;
3799
3800 if (!capable(CAP_SYS_ADMIN))
3801 return -EPERM;
3802
3803 ret = mnt_want_write_file(file);
3804 if (ret)
3805 return ret;
3806
3807 sa = memdup_user(arg, sizeof(*sa));
3808 if (IS_ERR(sa)) {
3809 ret = PTR_ERR(sa);
3810 goto drop_write;
3811 }
3812
3813 if (!sa->qgroupid) {
3814 ret = -EINVAL;
3815 goto out;
3816 }
3817
3818 if (sa->create && is_fstree(sa->qgroupid)) {
3819 ret = -EINVAL;
3820 goto out;
3821 }
3822
3823 trans = btrfs_join_transaction(root);
3824 if (IS_ERR(trans)) {
3825 ret = PTR_ERR(trans);
3826 goto out;
3827 }
3828
3829 if (sa->create) {
3830 ret = btrfs_create_qgroup(trans, sa->qgroupid);
3831 } else {
3832 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
3833 }
3834
3835 err = btrfs_end_transaction(trans);
3836 if (err && !ret)
3837 ret = err;
3838
3839out:
3840 kfree(sa);
3841drop_write:
3842 mnt_drop_write_file(file);
3843 return ret;
3844}
3845
3846static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3847{
3848 struct inode *inode = file_inode(file);
3849 struct btrfs_root *root = BTRFS_I(inode)->root;
3850 struct btrfs_ioctl_qgroup_limit_args *sa;
3851 struct btrfs_trans_handle *trans;
3852 int ret;
3853 int err;
3854 u64 qgroupid;
3855
3856 if (!capable(CAP_SYS_ADMIN))
3857 return -EPERM;
3858
3859 ret = mnt_want_write_file(file);
3860 if (ret)
3861 return ret;
3862
3863 sa = memdup_user(arg, sizeof(*sa));
3864 if (IS_ERR(sa)) {
3865 ret = PTR_ERR(sa);
3866 goto drop_write;
3867 }
3868
3869 trans = btrfs_join_transaction(root);
3870 if (IS_ERR(trans)) {
3871 ret = PTR_ERR(trans);
3872 goto out;
3873 }
3874
3875 qgroupid = sa->qgroupid;
3876 if (!qgroupid) {
3877 /* take the current subvol as qgroup */
3878 qgroupid = root->root_key.objectid;
3879 }
3880
3881 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
3882
3883 err = btrfs_end_transaction(trans);
3884 if (err && !ret)
3885 ret = err;
3886
3887out:
3888 kfree(sa);
3889drop_write:
3890 mnt_drop_write_file(file);
3891 return ret;
3892}
3893
3894static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3895{
3896 struct inode *inode = file_inode(file);
3897 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3898 struct btrfs_ioctl_quota_rescan_args *qsa;
3899 int ret;
3900
3901 if (!capable(CAP_SYS_ADMIN))
3902 return -EPERM;
3903
3904 ret = mnt_want_write_file(file);
3905 if (ret)
3906 return ret;
3907
3908 qsa = memdup_user(arg, sizeof(*qsa));
3909 if (IS_ERR(qsa)) {
3910 ret = PTR_ERR(qsa);
3911 goto drop_write;
3912 }
3913
3914 if (qsa->flags) {
3915 ret = -EINVAL;
3916 goto out;
3917 }
3918
3919 ret = btrfs_qgroup_rescan(fs_info);
3920
3921out:
3922 kfree(qsa);
3923drop_write:
3924 mnt_drop_write_file(file);
3925 return ret;
3926}
3927
3928static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
3929 void __user *arg)
3930{
3931 struct btrfs_ioctl_quota_rescan_args qsa = {0};
3932
3933 if (!capable(CAP_SYS_ADMIN))
3934 return -EPERM;
3935
3936 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3937 qsa.flags = 1;
3938 qsa.progress = fs_info->qgroup_rescan_progress.objectid;
3939 }
3940
3941 if (copy_to_user(arg, &qsa, sizeof(qsa)))
3942 return -EFAULT;
3943
3944 return 0;
3945}
3946
3947static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
3948 void __user *arg)
3949{
3950 if (!capable(CAP_SYS_ADMIN))
3951 return -EPERM;
3952
3953 return btrfs_qgroup_wait_for_completion(fs_info, true);
3954}
3955
3956static long _btrfs_ioctl_set_received_subvol(struct file *file,
3957 struct mnt_idmap *idmap,
3958 struct btrfs_ioctl_received_subvol_args *sa)
3959{
3960 struct inode *inode = file_inode(file);
3961 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3962 struct btrfs_root *root = BTRFS_I(inode)->root;
3963 struct btrfs_root_item *root_item = &root->root_item;
3964 struct btrfs_trans_handle *trans;
3965 struct timespec64 ct = current_time(inode);
3966 int ret = 0;
3967 int received_uuid_changed;
3968
3969 if (!inode_owner_or_capable(idmap, inode))
3970 return -EPERM;
3971
3972 ret = mnt_want_write_file(file);
3973 if (ret < 0)
3974 return ret;
3975
3976 down_write(&fs_info->subvol_sem);
3977
3978 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3979 ret = -EINVAL;
3980 goto out;
3981 }
3982
3983 if (btrfs_root_readonly(root)) {
3984 ret = -EROFS;
3985 goto out;
3986 }
3987
3988 /*
3989 * 1 - root item
3990 * 2 - uuid items (received uuid + subvol uuid)
3991 */
3992 trans = btrfs_start_transaction(root, 3);
3993 if (IS_ERR(trans)) {
3994 ret = PTR_ERR(trans);
3995 trans = NULL;
3996 goto out;
3997 }
3998
3999 sa->rtransid = trans->transid;
4000 sa->rtime.sec = ct.tv_sec;
4001 sa->rtime.nsec = ct.tv_nsec;
4002
4003 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4004 BTRFS_UUID_SIZE);
4005 if (received_uuid_changed &&
4006 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4007 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4008 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4009 root->root_key.objectid);
4010 if (ret && ret != -ENOENT) {
4011 btrfs_abort_transaction(trans, ret);
4012 btrfs_end_transaction(trans);
4013 goto out;
4014 }
4015 }
4016 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4017 btrfs_set_root_stransid(root_item, sa->stransid);
4018 btrfs_set_root_rtransid(root_item, sa->rtransid);
4019 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4020 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4021 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4022 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4023
4024 ret = btrfs_update_root(trans, fs_info->tree_root,
4025 &root->root_key, &root->root_item);
4026 if (ret < 0) {
4027 btrfs_end_transaction(trans);
4028 goto out;
4029 }
4030 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4031 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4032 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4033 root->root_key.objectid);
4034 if (ret < 0 && ret != -EEXIST) {
4035 btrfs_abort_transaction(trans, ret);
4036 btrfs_end_transaction(trans);
4037 goto out;
4038 }
4039 }
4040 ret = btrfs_commit_transaction(trans);
4041out:
4042 up_write(&fs_info->subvol_sem);
4043 mnt_drop_write_file(file);
4044 return ret;
4045}
4046
4047#ifdef CONFIG_64BIT
4048static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4049 void __user *arg)
4050{
4051 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4052 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4053 int ret = 0;
4054
4055 args32 = memdup_user(arg, sizeof(*args32));
4056 if (IS_ERR(args32))
4057 return PTR_ERR(args32);
4058
4059 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4060 if (!args64) {
4061 ret = -ENOMEM;
4062 goto out;
4063 }
4064
4065 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4066 args64->stransid = args32->stransid;
4067 args64->rtransid = args32->rtransid;
4068 args64->stime.sec = args32->stime.sec;
4069 args64->stime.nsec = args32->stime.nsec;
4070 args64->rtime.sec = args32->rtime.sec;
4071 args64->rtime.nsec = args32->rtime.nsec;
4072 args64->flags = args32->flags;
4073
4074 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), args64);
4075 if (ret)
4076 goto out;
4077
4078 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4079 args32->stransid = args64->stransid;
4080 args32->rtransid = args64->rtransid;
4081 args32->stime.sec = args64->stime.sec;
4082 args32->stime.nsec = args64->stime.nsec;
4083 args32->rtime.sec = args64->rtime.sec;
4084 args32->rtime.nsec = args64->rtime.nsec;
4085 args32->flags = args64->flags;
4086
4087 ret = copy_to_user(arg, args32, sizeof(*args32));
4088 if (ret)
4089 ret = -EFAULT;
4090
4091out:
4092 kfree(args32);
4093 kfree(args64);
4094 return ret;
4095}
4096#endif
4097
4098static long btrfs_ioctl_set_received_subvol(struct file *file,
4099 void __user *arg)
4100{
4101 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4102 int ret = 0;
4103
4104 sa = memdup_user(arg, sizeof(*sa));
4105 if (IS_ERR(sa))
4106 return PTR_ERR(sa);
4107
4108 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), sa);
4109
4110 if (ret)
4111 goto out;
4112
4113 ret = copy_to_user(arg, sa, sizeof(*sa));
4114 if (ret)
4115 ret = -EFAULT;
4116
4117out:
4118 kfree(sa);
4119 return ret;
4120}
4121
4122static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4123 void __user *arg)
4124{
4125 size_t len;
4126 int ret;
4127 char label[BTRFS_LABEL_SIZE];
4128
4129 spin_lock(&fs_info->super_lock);
4130 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4131 spin_unlock(&fs_info->super_lock);
4132
4133 len = strnlen(label, BTRFS_LABEL_SIZE);
4134
4135 if (len == BTRFS_LABEL_SIZE) {
4136 btrfs_warn(fs_info,
4137 "label is too long, return the first %zu bytes",
4138 --len);
4139 }
4140
4141 ret = copy_to_user(arg, label, len);
4142
4143 return ret ? -EFAULT : 0;
4144}
4145
4146static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4147{
4148 struct inode *inode = file_inode(file);
4149 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4150 struct btrfs_root *root = BTRFS_I(inode)->root;
4151 struct btrfs_super_block *super_block = fs_info->super_copy;
4152 struct btrfs_trans_handle *trans;
4153 char label[BTRFS_LABEL_SIZE];
4154 int ret;
4155
4156 if (!capable(CAP_SYS_ADMIN))
4157 return -EPERM;
4158
4159 if (copy_from_user(label, arg, sizeof(label)))
4160 return -EFAULT;
4161
4162 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4163 btrfs_err(fs_info,
4164 "unable to set label with more than %d bytes",
4165 BTRFS_LABEL_SIZE - 1);
4166 return -EINVAL;
4167 }
4168
4169 ret = mnt_want_write_file(file);
4170 if (ret)
4171 return ret;
4172
4173 trans = btrfs_start_transaction(root, 0);
4174 if (IS_ERR(trans)) {
4175 ret = PTR_ERR(trans);
4176 goto out_unlock;
4177 }
4178
4179 spin_lock(&fs_info->super_lock);
4180 strcpy(super_block->label, label);
4181 spin_unlock(&fs_info->super_lock);
4182 ret = btrfs_commit_transaction(trans);
4183
4184out_unlock:
4185 mnt_drop_write_file(file);
4186 return ret;
4187}
4188
4189#define INIT_FEATURE_FLAGS(suffix) \
4190 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4191 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4192 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4193
4194int btrfs_ioctl_get_supported_features(void __user *arg)
4195{
4196 static const struct btrfs_ioctl_feature_flags features[3] = {
4197 INIT_FEATURE_FLAGS(SUPP),
4198 INIT_FEATURE_FLAGS(SAFE_SET),
4199 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4200 };
4201
4202 if (copy_to_user(arg, &features, sizeof(features)))
4203 return -EFAULT;
4204
4205 return 0;
4206}
4207
4208static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4209 void __user *arg)
4210{
4211 struct btrfs_super_block *super_block = fs_info->super_copy;
4212 struct btrfs_ioctl_feature_flags features;
4213
4214 features.compat_flags = btrfs_super_compat_flags(super_block);
4215 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4216 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4217
4218 if (copy_to_user(arg, &features, sizeof(features)))
4219 return -EFAULT;
4220
4221 return 0;
4222}
4223
4224static int check_feature_bits(struct btrfs_fs_info *fs_info,
4225 enum btrfs_feature_set set,
4226 u64 change_mask, u64 flags, u64 supported_flags,
4227 u64 safe_set, u64 safe_clear)
4228{
4229 const char *type = btrfs_feature_set_name(set);
4230 char *names;
4231 u64 disallowed, unsupported;
4232 u64 set_mask = flags & change_mask;
4233 u64 clear_mask = ~flags & change_mask;
4234
4235 unsupported = set_mask & ~supported_flags;
4236 if (unsupported) {
4237 names = btrfs_printable_features(set, unsupported);
4238 if (names) {
4239 btrfs_warn(fs_info,
4240 "this kernel does not support the %s feature bit%s",
4241 names, strchr(names, ',') ? "s" : "");
4242 kfree(names);
4243 } else
4244 btrfs_warn(fs_info,
4245 "this kernel does not support %s bits 0x%llx",
4246 type, unsupported);
4247 return -EOPNOTSUPP;
4248 }
4249
4250 disallowed = set_mask & ~safe_set;
4251 if (disallowed) {
4252 names = btrfs_printable_features(set, disallowed);
4253 if (names) {
4254 btrfs_warn(fs_info,
4255 "can't set the %s feature bit%s while mounted",
4256 names, strchr(names, ',') ? "s" : "");
4257 kfree(names);
4258 } else
4259 btrfs_warn(fs_info,
4260 "can't set %s bits 0x%llx while mounted",
4261 type, disallowed);
4262 return -EPERM;
4263 }
4264
4265 disallowed = clear_mask & ~safe_clear;
4266 if (disallowed) {
4267 names = btrfs_printable_features(set, disallowed);
4268 if (names) {
4269 btrfs_warn(fs_info,
4270 "can't clear the %s feature bit%s while mounted",
4271 names, strchr(names, ',') ? "s" : "");
4272 kfree(names);
4273 } else
4274 btrfs_warn(fs_info,
4275 "can't clear %s bits 0x%llx while mounted",
4276 type, disallowed);
4277 return -EPERM;
4278 }
4279
4280 return 0;
4281}
4282
4283#define check_feature(fs_info, change_mask, flags, mask_base) \
4284check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
4285 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4286 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4287 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4288
4289static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4290{
4291 struct inode *inode = file_inode(file);
4292 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4293 struct btrfs_root *root = BTRFS_I(inode)->root;
4294 struct btrfs_super_block *super_block = fs_info->super_copy;
4295 struct btrfs_ioctl_feature_flags flags[2];
4296 struct btrfs_trans_handle *trans;
4297 u64 newflags;
4298 int ret;
4299
4300 if (!capable(CAP_SYS_ADMIN))
4301 return -EPERM;
4302
4303 if (copy_from_user(flags, arg, sizeof(flags)))
4304 return -EFAULT;
4305
4306 /* Nothing to do */
4307 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4308 !flags[0].incompat_flags)
4309 return 0;
4310
4311 ret = check_feature(fs_info, flags[0].compat_flags,
4312 flags[1].compat_flags, COMPAT);
4313 if (ret)
4314 return ret;
4315
4316 ret = check_feature(fs_info, flags[0].compat_ro_flags,
4317 flags[1].compat_ro_flags, COMPAT_RO);
4318 if (ret)
4319 return ret;
4320
4321 ret = check_feature(fs_info, flags[0].incompat_flags,
4322 flags[1].incompat_flags, INCOMPAT);
4323 if (ret)
4324 return ret;
4325
4326 ret = mnt_want_write_file(file);
4327 if (ret)
4328 return ret;
4329
4330 trans = btrfs_start_transaction(root, 0);
4331 if (IS_ERR(trans)) {
4332 ret = PTR_ERR(trans);
4333 goto out_drop_write;
4334 }
4335
4336 spin_lock(&fs_info->super_lock);
4337 newflags = btrfs_super_compat_flags(super_block);
4338 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4339 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4340 btrfs_set_super_compat_flags(super_block, newflags);
4341
4342 newflags = btrfs_super_compat_ro_flags(super_block);
4343 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4344 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4345 btrfs_set_super_compat_ro_flags(super_block, newflags);
4346
4347 newflags = btrfs_super_incompat_flags(super_block);
4348 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4349 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4350 btrfs_set_super_incompat_flags(super_block, newflags);
4351 spin_unlock(&fs_info->super_lock);
4352
4353 ret = btrfs_commit_transaction(trans);
4354out_drop_write:
4355 mnt_drop_write_file(file);
4356
4357 return ret;
4358}
4359
4360static int _btrfs_ioctl_send(struct inode *inode, void __user *argp, bool compat)
4361{
4362 struct btrfs_ioctl_send_args *arg;
4363 int ret;
4364
4365 if (compat) {
4366#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4367 struct btrfs_ioctl_send_args_32 args32 = { 0 };
4368
4369 ret = copy_from_user(&args32, argp, sizeof(args32));
4370 if (ret)
4371 return -EFAULT;
4372 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4373 if (!arg)
4374 return -ENOMEM;
4375 arg->send_fd = args32.send_fd;
4376 arg->clone_sources_count = args32.clone_sources_count;
4377 arg->clone_sources = compat_ptr(args32.clone_sources);
4378 arg->parent_root = args32.parent_root;
4379 arg->flags = args32.flags;
4380 arg->version = args32.version;
4381 memcpy(arg->reserved, args32.reserved,
4382 sizeof(args32.reserved));
4383#else
4384 return -ENOTTY;
4385#endif
4386 } else {
4387 arg = memdup_user(argp, sizeof(*arg));
4388 if (IS_ERR(arg))
4389 return PTR_ERR(arg);
4390 }
4391 ret = btrfs_ioctl_send(inode, arg);
4392 kfree(arg);
4393 return ret;
4394}
4395
4396static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,
4397 bool compat)
4398{
4399 struct btrfs_ioctl_encoded_io_args args = { 0 };
4400 size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args,
4401 flags);
4402 size_t copy_end;
4403 struct iovec iovstack[UIO_FASTIOV];
4404 struct iovec *iov = iovstack;
4405 struct iov_iter iter;
4406 loff_t pos;
4407 struct kiocb kiocb;
4408 ssize_t ret;
4409
4410 if (!capable(CAP_SYS_ADMIN)) {
4411 ret = -EPERM;
4412 goto out_acct;
4413 }
4414
4415 if (compat) {
4416#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4417 struct btrfs_ioctl_encoded_io_args_32 args32;
4418
4419 copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32,
4420 flags);
4421 if (copy_from_user(&args32, argp, copy_end)) {
4422 ret = -EFAULT;
4423 goto out_acct;
4424 }
4425 args.iov = compat_ptr(args32.iov);
4426 args.iovcnt = args32.iovcnt;
4427 args.offset = args32.offset;
4428 args.flags = args32.flags;
4429#else
4430 return -ENOTTY;
4431#endif
4432 } else {
4433 copy_end = copy_end_kernel;
4434 if (copy_from_user(&args, argp, copy_end)) {
4435 ret = -EFAULT;
4436 goto out_acct;
4437 }
4438 }
4439 if (args.flags != 0) {
4440 ret = -EINVAL;
4441 goto out_acct;
4442 }
4443
4444 ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4445 &iov, &iter);
4446 if (ret < 0)
4447 goto out_acct;
4448
4449 if (iov_iter_count(&iter) == 0) {
4450 ret = 0;
4451 goto out_iov;
4452 }
4453 pos = args.offset;
4454 ret = rw_verify_area(READ, file, &pos, args.len);
4455 if (ret < 0)
4456 goto out_iov;
4457
4458 init_sync_kiocb(&kiocb, file);
4459 kiocb.ki_pos = pos;
4460
4461 ret = btrfs_encoded_read(&kiocb, &iter, &args);
4462 if (ret >= 0) {
4463 fsnotify_access(file);
4464 if (copy_to_user(argp + copy_end,
4465 (char *)&args + copy_end_kernel,
4466 sizeof(args) - copy_end_kernel))
4467 ret = -EFAULT;
4468 }
4469
4470out_iov:
4471 kfree(iov);
4472out_acct:
4473 if (ret > 0)
4474 add_rchar(current, ret);
4475 inc_syscr(current);
4476 return ret;
4477}
4478
4479static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat)
4480{
4481 struct btrfs_ioctl_encoded_io_args args;
4482 struct iovec iovstack[UIO_FASTIOV];
4483 struct iovec *iov = iovstack;
4484 struct iov_iter iter;
4485 loff_t pos;
4486 struct kiocb kiocb;
4487 ssize_t ret;
4488
4489 if (!capable(CAP_SYS_ADMIN)) {
4490 ret = -EPERM;
4491 goto out_acct;
4492 }
4493
4494 if (!(file->f_mode & FMODE_WRITE)) {
4495 ret = -EBADF;
4496 goto out_acct;
4497 }
4498
4499 if (compat) {
4500#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4501 struct btrfs_ioctl_encoded_io_args_32 args32;
4502
4503 if (copy_from_user(&args32, argp, sizeof(args32))) {
4504 ret = -EFAULT;
4505 goto out_acct;
4506 }
4507 args.iov = compat_ptr(args32.iov);
4508 args.iovcnt = args32.iovcnt;
4509 args.offset = args32.offset;
4510 args.flags = args32.flags;
4511 args.len = args32.len;
4512 args.unencoded_len = args32.unencoded_len;
4513 args.unencoded_offset = args32.unencoded_offset;
4514 args.compression = args32.compression;
4515 args.encryption = args32.encryption;
4516 memcpy(args.reserved, args32.reserved, sizeof(args.reserved));
4517#else
4518 return -ENOTTY;
4519#endif
4520 } else {
4521 if (copy_from_user(&args, argp, sizeof(args))) {
4522 ret = -EFAULT;
4523 goto out_acct;
4524 }
4525 }
4526
4527 ret = -EINVAL;
4528 if (args.flags != 0)
4529 goto out_acct;
4530 if (memchr_inv(args.reserved, 0, sizeof(args.reserved)))
4531 goto out_acct;
4532 if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
4533 args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
4534 goto out_acct;
4535 if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
4536 args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
4537 goto out_acct;
4538 if (args.unencoded_offset > args.unencoded_len)
4539 goto out_acct;
4540 if (args.len > args.unencoded_len - args.unencoded_offset)
4541 goto out_acct;
4542
4543 ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4544 &iov, &iter);
4545 if (ret < 0)
4546 goto out_acct;
4547
4548 if (iov_iter_count(&iter) == 0) {
4549 ret = 0;
4550 goto out_iov;
4551 }
4552 pos = args.offset;
4553 ret = rw_verify_area(WRITE, file, &pos, args.len);
4554 if (ret < 0)
4555 goto out_iov;
4556
4557 init_sync_kiocb(&kiocb, file);
4558 ret = kiocb_set_rw_flags(&kiocb, 0);
4559 if (ret)
4560 goto out_iov;
4561 kiocb.ki_pos = pos;
4562
4563 file_start_write(file);
4564
4565 ret = btrfs_do_write_iter(&kiocb, &iter, &args);
4566 if (ret > 0)
4567 fsnotify_modify(file);
4568
4569 file_end_write(file);
4570out_iov:
4571 kfree(iov);
4572out_acct:
4573 if (ret > 0)
4574 add_wchar(current, ret);
4575 inc_syscw(current);
4576 return ret;
4577}
4578
4579long btrfs_ioctl(struct file *file, unsigned int
4580 cmd, unsigned long arg)
4581{
4582 struct inode *inode = file_inode(file);
4583 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4584 struct btrfs_root *root = BTRFS_I(inode)->root;
4585 void __user *argp = (void __user *)arg;
4586
4587 switch (cmd) {
4588 case FS_IOC_GETVERSION:
4589 return btrfs_ioctl_getversion(inode, argp);
4590 case FS_IOC_GETFSLABEL:
4591 return btrfs_ioctl_get_fslabel(fs_info, argp);
4592 case FS_IOC_SETFSLABEL:
4593 return btrfs_ioctl_set_fslabel(file, argp);
4594 case FITRIM:
4595 return btrfs_ioctl_fitrim(fs_info, argp);
4596 case BTRFS_IOC_SNAP_CREATE:
4597 return btrfs_ioctl_snap_create(file, argp, 0);
4598 case BTRFS_IOC_SNAP_CREATE_V2:
4599 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4600 case BTRFS_IOC_SUBVOL_CREATE:
4601 return btrfs_ioctl_snap_create(file, argp, 1);
4602 case BTRFS_IOC_SUBVOL_CREATE_V2:
4603 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4604 case BTRFS_IOC_SNAP_DESTROY:
4605 return btrfs_ioctl_snap_destroy(file, argp, false);
4606 case BTRFS_IOC_SNAP_DESTROY_V2:
4607 return btrfs_ioctl_snap_destroy(file, argp, true);
4608 case BTRFS_IOC_SUBVOL_GETFLAGS:
4609 return btrfs_ioctl_subvol_getflags(inode, argp);
4610 case BTRFS_IOC_SUBVOL_SETFLAGS:
4611 return btrfs_ioctl_subvol_setflags(file, argp);
4612 case BTRFS_IOC_DEFAULT_SUBVOL:
4613 return btrfs_ioctl_default_subvol(file, argp);
4614 case BTRFS_IOC_DEFRAG:
4615 return btrfs_ioctl_defrag(file, NULL);
4616 case BTRFS_IOC_DEFRAG_RANGE:
4617 return btrfs_ioctl_defrag(file, argp);
4618 case BTRFS_IOC_RESIZE:
4619 return btrfs_ioctl_resize(file, argp);
4620 case BTRFS_IOC_ADD_DEV:
4621 return btrfs_ioctl_add_dev(fs_info, argp);
4622 case BTRFS_IOC_RM_DEV:
4623 return btrfs_ioctl_rm_dev(file, argp);
4624 case BTRFS_IOC_RM_DEV_V2:
4625 return btrfs_ioctl_rm_dev_v2(file, argp);
4626 case BTRFS_IOC_FS_INFO:
4627 return btrfs_ioctl_fs_info(fs_info, argp);
4628 case BTRFS_IOC_DEV_INFO:
4629 return btrfs_ioctl_dev_info(fs_info, argp);
4630 case BTRFS_IOC_TREE_SEARCH:
4631 return btrfs_ioctl_tree_search(inode, argp);
4632 case BTRFS_IOC_TREE_SEARCH_V2:
4633 return btrfs_ioctl_tree_search_v2(inode, argp);
4634 case BTRFS_IOC_INO_LOOKUP:
4635 return btrfs_ioctl_ino_lookup(root, argp);
4636 case BTRFS_IOC_INO_PATHS:
4637 return btrfs_ioctl_ino_to_path(root, argp);
4638 case BTRFS_IOC_LOGICAL_INO:
4639 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4640 case BTRFS_IOC_LOGICAL_INO_V2:
4641 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4642 case BTRFS_IOC_SPACE_INFO:
4643 return btrfs_ioctl_space_info(fs_info, argp);
4644 case BTRFS_IOC_SYNC: {
4645 int ret;
4646
4647 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4648 if (ret)
4649 return ret;
4650 ret = btrfs_sync_fs(inode->i_sb, 1);
4651 /*
4652 * The transaction thread may want to do more work,
4653 * namely it pokes the cleaner kthread that will start
4654 * processing uncleaned subvols.
4655 */
4656 wake_up_process(fs_info->transaction_kthread);
4657 return ret;
4658 }
4659 case BTRFS_IOC_START_SYNC:
4660 return btrfs_ioctl_start_sync(root, argp);
4661 case BTRFS_IOC_WAIT_SYNC:
4662 return btrfs_ioctl_wait_sync(fs_info, argp);
4663 case BTRFS_IOC_SCRUB:
4664 return btrfs_ioctl_scrub(file, argp);
4665 case BTRFS_IOC_SCRUB_CANCEL:
4666 return btrfs_ioctl_scrub_cancel(fs_info);
4667 case BTRFS_IOC_SCRUB_PROGRESS:
4668 return btrfs_ioctl_scrub_progress(fs_info, argp);
4669 case BTRFS_IOC_BALANCE_V2:
4670 return btrfs_ioctl_balance(file, argp);
4671 case BTRFS_IOC_BALANCE_CTL:
4672 return btrfs_ioctl_balance_ctl(fs_info, arg);
4673 case BTRFS_IOC_BALANCE_PROGRESS:
4674 return btrfs_ioctl_balance_progress(fs_info, argp);
4675 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4676 return btrfs_ioctl_set_received_subvol(file, argp);
4677#ifdef CONFIG_64BIT
4678 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4679 return btrfs_ioctl_set_received_subvol_32(file, argp);
4680#endif
4681 case BTRFS_IOC_SEND:
4682 return _btrfs_ioctl_send(inode, argp, false);
4683#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4684 case BTRFS_IOC_SEND_32:
4685 return _btrfs_ioctl_send(inode, argp, true);
4686#endif
4687 case BTRFS_IOC_GET_DEV_STATS:
4688 return btrfs_ioctl_get_dev_stats(fs_info, argp);
4689 case BTRFS_IOC_QUOTA_CTL:
4690 return btrfs_ioctl_quota_ctl(file, argp);
4691 case BTRFS_IOC_QGROUP_ASSIGN:
4692 return btrfs_ioctl_qgroup_assign(file, argp);
4693 case BTRFS_IOC_QGROUP_CREATE:
4694 return btrfs_ioctl_qgroup_create(file, argp);
4695 case BTRFS_IOC_QGROUP_LIMIT:
4696 return btrfs_ioctl_qgroup_limit(file, argp);
4697 case BTRFS_IOC_QUOTA_RESCAN:
4698 return btrfs_ioctl_quota_rescan(file, argp);
4699 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4700 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
4701 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4702 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
4703 case BTRFS_IOC_DEV_REPLACE:
4704 return btrfs_ioctl_dev_replace(fs_info, argp);
4705 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4706 return btrfs_ioctl_get_supported_features(argp);
4707 case BTRFS_IOC_GET_FEATURES:
4708 return btrfs_ioctl_get_features(fs_info, argp);
4709 case BTRFS_IOC_SET_FEATURES:
4710 return btrfs_ioctl_set_features(file, argp);
4711 case BTRFS_IOC_GET_SUBVOL_INFO:
4712 return btrfs_ioctl_get_subvol_info(inode, argp);
4713 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
4714 return btrfs_ioctl_get_subvol_rootref(root, argp);
4715 case BTRFS_IOC_INO_LOOKUP_USER:
4716 return btrfs_ioctl_ino_lookup_user(file, argp);
4717 case FS_IOC_ENABLE_VERITY:
4718 return fsverity_ioctl_enable(file, (const void __user *)argp);
4719 case FS_IOC_MEASURE_VERITY:
4720 return fsverity_ioctl_measure(file, argp);
4721 case BTRFS_IOC_ENCODED_READ:
4722 return btrfs_ioctl_encoded_read(file, argp, false);
4723 case BTRFS_IOC_ENCODED_WRITE:
4724 return btrfs_ioctl_encoded_write(file, argp, false);
4725#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4726 case BTRFS_IOC_ENCODED_READ_32:
4727 return btrfs_ioctl_encoded_read(file, argp, true);
4728 case BTRFS_IOC_ENCODED_WRITE_32:
4729 return btrfs_ioctl_encoded_write(file, argp, true);
4730#endif
4731 }
4732
4733 return -ENOTTY;
4734}
4735
4736#ifdef CONFIG_COMPAT
4737long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4738{
4739 /*
4740 * These all access 32-bit values anyway so no further
4741 * handling is necessary.
4742 */
4743 switch (cmd) {
4744 case FS_IOC32_GETVERSION:
4745 cmd = FS_IOC_GETVERSION;
4746 break;
4747 }
4748
4749 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4750}
4751#endif