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