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 "ctree.h"
31#include "disk-io.h"
32#include "export.h"
33#include "transaction.h"
34#include "btrfs_inode.h"
35#include "print-tree.h"
36#include "volumes.h"
37#include "locking.h"
38#include "backref.h"
39#include "rcu-string.h"
40#include "send.h"
41#include "dev-replace.h"
42#include "props.h"
43#include "sysfs.h"
44#include "qgroup.h"
45#include "tree-log.h"
46#include "compression.h"
47#include "space-info.h"
48#include "delalloc-space.h"
49#include "block-group.h"
50
51#ifdef CONFIG_64BIT
52/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
53 * structures are incorrect, as the timespec structure from userspace
54 * is 4 bytes too small. We define these alternatives here to teach
55 * the kernel about the 32-bit struct packing.
56 */
57struct btrfs_ioctl_timespec_32 {
58 __u64 sec;
59 __u32 nsec;
60} __attribute__ ((__packed__));
61
62struct btrfs_ioctl_received_subvol_args_32 {
63 char uuid[BTRFS_UUID_SIZE]; /* in */
64 __u64 stransid; /* in */
65 __u64 rtransid; /* out */
66 struct btrfs_ioctl_timespec_32 stime; /* in */
67 struct btrfs_ioctl_timespec_32 rtime; /* out */
68 __u64 flags; /* in */
69 __u64 reserved[16]; /* in */
70} __attribute__ ((__packed__));
71
72#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
73 struct btrfs_ioctl_received_subvol_args_32)
74#endif
75
76#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
77struct btrfs_ioctl_send_args_32 {
78 __s64 send_fd; /* in */
79 __u64 clone_sources_count; /* in */
80 compat_uptr_t clone_sources; /* in */
81 __u64 parent_root; /* in */
82 __u64 flags; /* in */
83 __u64 reserved[4]; /* in */
84} __attribute__ ((__packed__));
85
86#define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
87 struct btrfs_ioctl_send_args_32)
88#endif
89
90/* Mask out flags that are inappropriate for the given type of inode. */
91static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
92 unsigned int flags)
93{
94 if (S_ISDIR(inode->i_mode))
95 return flags;
96 else if (S_ISREG(inode->i_mode))
97 return flags & ~FS_DIRSYNC_FL;
98 else
99 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
100}
101
102/*
103 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
104 * ioctl.
105 */
106static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags)
107{
108 unsigned int iflags = 0;
109
110 if (flags & BTRFS_INODE_SYNC)
111 iflags |= FS_SYNC_FL;
112 if (flags & BTRFS_INODE_IMMUTABLE)
113 iflags |= FS_IMMUTABLE_FL;
114 if (flags & BTRFS_INODE_APPEND)
115 iflags |= FS_APPEND_FL;
116 if (flags & BTRFS_INODE_NODUMP)
117 iflags |= FS_NODUMP_FL;
118 if (flags & BTRFS_INODE_NOATIME)
119 iflags |= FS_NOATIME_FL;
120 if (flags & BTRFS_INODE_DIRSYNC)
121 iflags |= FS_DIRSYNC_FL;
122 if (flags & BTRFS_INODE_NODATACOW)
123 iflags |= FS_NOCOW_FL;
124
125 if (flags & BTRFS_INODE_NOCOMPRESS)
126 iflags |= FS_NOCOMP_FL;
127 else if (flags & BTRFS_INODE_COMPRESS)
128 iflags |= FS_COMPR_FL;
129
130 return iflags;
131}
132
133/*
134 * Update inode->i_flags based on the btrfs internal flags.
135 */
136void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
137{
138 struct btrfs_inode *binode = BTRFS_I(inode);
139 unsigned int new_fl = 0;
140
141 if (binode->flags & BTRFS_INODE_SYNC)
142 new_fl |= S_SYNC;
143 if (binode->flags & BTRFS_INODE_IMMUTABLE)
144 new_fl |= S_IMMUTABLE;
145 if (binode->flags & BTRFS_INODE_APPEND)
146 new_fl |= S_APPEND;
147 if (binode->flags & BTRFS_INODE_NOATIME)
148 new_fl |= S_NOATIME;
149 if (binode->flags & BTRFS_INODE_DIRSYNC)
150 new_fl |= S_DIRSYNC;
151
152 set_mask_bits(&inode->i_flags,
153 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
154 new_fl);
155}
156
157/*
158 * Check if @flags are a supported and valid set of FS_*_FL flags and that
159 * the old and new flags are not conflicting
160 */
161static int check_fsflags(unsigned int old_flags, unsigned int flags)
162{
163 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
164 FS_NOATIME_FL | FS_NODUMP_FL | \
165 FS_SYNC_FL | FS_DIRSYNC_FL | \
166 FS_NOCOMP_FL | FS_COMPR_FL |
167 FS_NOCOW_FL))
168 return -EOPNOTSUPP;
169
170 /* COMPR and NOCOMP on new/old are valid */
171 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
172 return -EINVAL;
173
174 if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
175 return -EINVAL;
176
177 /* NOCOW and compression options are mutually exclusive */
178 if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
179 return -EINVAL;
180 if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
181 return -EINVAL;
182
183 return 0;
184}
185
186static int check_fsflags_compatible(struct btrfs_fs_info *fs_info,
187 unsigned int flags)
188{
189 if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL))
190 return -EPERM;
191
192 return 0;
193}
194
195/*
196 * Set flags/xflags from the internal inode flags. The remaining items of
197 * fsxattr are zeroed.
198 */
199int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
200{
201 struct btrfs_inode *binode = BTRFS_I(d_inode(dentry));
202
203 fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode->flags));
204 return 0;
205}
206
207int btrfs_fileattr_set(struct user_namespace *mnt_userns,
208 struct dentry *dentry, struct fileattr *fa)
209{
210 struct inode *inode = d_inode(dentry);
211 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
212 struct btrfs_inode *binode = BTRFS_I(inode);
213 struct btrfs_root *root = binode->root;
214 struct btrfs_trans_handle *trans;
215 unsigned int fsflags, old_fsflags;
216 int ret;
217 const char *comp = NULL;
218 u32 binode_flags;
219
220 if (btrfs_root_readonly(root))
221 return -EROFS;
222
223 if (fileattr_has_fsx(fa))
224 return -EOPNOTSUPP;
225
226 fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags);
227 old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags);
228 ret = check_fsflags(old_fsflags, fsflags);
229 if (ret)
230 return ret;
231
232 ret = check_fsflags_compatible(fs_info, fsflags);
233 if (ret)
234 return ret;
235
236 binode_flags = binode->flags;
237 if (fsflags & FS_SYNC_FL)
238 binode_flags |= BTRFS_INODE_SYNC;
239 else
240 binode_flags &= ~BTRFS_INODE_SYNC;
241 if (fsflags & FS_IMMUTABLE_FL)
242 binode_flags |= BTRFS_INODE_IMMUTABLE;
243 else
244 binode_flags &= ~BTRFS_INODE_IMMUTABLE;
245 if (fsflags & FS_APPEND_FL)
246 binode_flags |= BTRFS_INODE_APPEND;
247 else
248 binode_flags &= ~BTRFS_INODE_APPEND;
249 if (fsflags & FS_NODUMP_FL)
250 binode_flags |= BTRFS_INODE_NODUMP;
251 else
252 binode_flags &= ~BTRFS_INODE_NODUMP;
253 if (fsflags & FS_NOATIME_FL)
254 binode_flags |= BTRFS_INODE_NOATIME;
255 else
256 binode_flags &= ~BTRFS_INODE_NOATIME;
257
258 /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */
259 if (!fa->flags_valid) {
260 /* 1 item for the inode */
261 trans = btrfs_start_transaction(root, 1);
262 if (IS_ERR(trans))
263 return PTR_ERR(trans);
264 goto update_flags;
265 }
266
267 if (fsflags & FS_DIRSYNC_FL)
268 binode_flags |= BTRFS_INODE_DIRSYNC;
269 else
270 binode_flags &= ~BTRFS_INODE_DIRSYNC;
271 if (fsflags & FS_NOCOW_FL) {
272 if (S_ISREG(inode->i_mode)) {
273 /*
274 * It's safe to turn csums off here, no extents exist.
275 * Otherwise we want the flag to reflect the real COW
276 * status of the file and will not set it.
277 */
278 if (inode->i_size == 0)
279 binode_flags |= BTRFS_INODE_NODATACOW |
280 BTRFS_INODE_NODATASUM;
281 } else {
282 binode_flags |= BTRFS_INODE_NODATACOW;
283 }
284 } else {
285 /*
286 * Revert back under same assumptions as above
287 */
288 if (S_ISREG(inode->i_mode)) {
289 if (inode->i_size == 0)
290 binode_flags &= ~(BTRFS_INODE_NODATACOW |
291 BTRFS_INODE_NODATASUM);
292 } else {
293 binode_flags &= ~BTRFS_INODE_NODATACOW;
294 }
295 }
296
297 /*
298 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
299 * flag may be changed automatically if compression code won't make
300 * things smaller.
301 */
302 if (fsflags & FS_NOCOMP_FL) {
303 binode_flags &= ~BTRFS_INODE_COMPRESS;
304 binode_flags |= BTRFS_INODE_NOCOMPRESS;
305 } else if (fsflags & FS_COMPR_FL) {
306
307 if (IS_SWAPFILE(inode))
308 return -ETXTBSY;
309
310 binode_flags |= BTRFS_INODE_COMPRESS;
311 binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
312
313 comp = btrfs_compress_type2str(fs_info->compress_type);
314 if (!comp || comp[0] == 0)
315 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
316 } else {
317 binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
318 }
319
320 /*
321 * 1 for inode item
322 * 2 for properties
323 */
324 trans = btrfs_start_transaction(root, 3);
325 if (IS_ERR(trans))
326 return PTR_ERR(trans);
327
328 if (comp) {
329 ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
330 strlen(comp), 0);
331 if (ret) {
332 btrfs_abort_transaction(trans, ret);
333 goto out_end_trans;
334 }
335 } else {
336 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
337 0, 0);
338 if (ret && ret != -ENODATA) {
339 btrfs_abort_transaction(trans, ret);
340 goto out_end_trans;
341 }
342 }
343
344update_flags:
345 binode->flags = binode_flags;
346 btrfs_sync_inode_flags_to_i_flags(inode);
347 inode_inc_iversion(inode);
348 inode->i_ctime = current_time(inode);
349 ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
350
351 out_end_trans:
352 btrfs_end_transaction(trans);
353 return ret;
354}
355
356/*
357 * Start exclusive operation @type, return true on success
358 */
359bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
360 enum btrfs_exclusive_operation type)
361{
362 bool ret = false;
363
364 spin_lock(&fs_info->super_lock);
365 if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) {
366 fs_info->exclusive_operation = type;
367 ret = true;
368 }
369 spin_unlock(&fs_info->super_lock);
370
371 return ret;
372}
373
374/*
375 * Conditionally allow to enter the exclusive operation in case it's compatible
376 * with the running one. This must be paired with btrfs_exclop_start_unlock and
377 * btrfs_exclop_finish.
378 *
379 * Compatibility:
380 * - the same type is already running
381 * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller
382 * must check the condition first that would allow none -> @type
383 */
384bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
385 enum btrfs_exclusive_operation type)
386{
387 spin_lock(&fs_info->super_lock);
388 if (fs_info->exclusive_operation == type)
389 return true;
390
391 spin_unlock(&fs_info->super_lock);
392 return false;
393}
394
395void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info)
396{
397 spin_unlock(&fs_info->super_lock);
398}
399
400void btrfs_exclop_finish(struct btrfs_fs_info *fs_info)
401{
402 spin_lock(&fs_info->super_lock);
403 WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE);
404 spin_unlock(&fs_info->super_lock);
405 sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation");
406}
407
408static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
409{
410 struct inode *inode = file_inode(file);
411
412 return put_user(inode->i_generation, arg);
413}
414
415static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
416 void __user *arg)
417{
418 struct btrfs_device *device;
419 struct request_queue *q;
420 struct fstrim_range range;
421 u64 minlen = ULLONG_MAX;
422 u64 num_devices = 0;
423 int ret;
424
425 if (!capable(CAP_SYS_ADMIN))
426 return -EPERM;
427
428 /*
429 * btrfs_trim_block_group() depends on space cache, which is not
430 * available in zoned filesystem. So, disallow fitrim on a zoned
431 * filesystem for now.
432 */
433 if (btrfs_is_zoned(fs_info))
434 return -EOPNOTSUPP;
435
436 /*
437 * If the fs is mounted with nologreplay, which requires it to be
438 * mounted in RO mode as well, we can not allow discard on free space
439 * inside block groups, because log trees refer to extents that are not
440 * pinned in a block group's free space cache (pinning the extents is
441 * precisely the first phase of replaying a log tree).
442 */
443 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
444 return -EROFS;
445
446 rcu_read_lock();
447 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
448 dev_list) {
449 if (!device->bdev)
450 continue;
451 q = bdev_get_queue(device->bdev);
452 if (blk_queue_discard(q)) {
453 num_devices++;
454 minlen = min_t(u64, q->limits.discard_granularity,
455 minlen);
456 }
457 }
458 rcu_read_unlock();
459
460 if (!num_devices)
461 return -EOPNOTSUPP;
462 if (copy_from_user(&range, arg, sizeof(range)))
463 return -EFAULT;
464
465 /*
466 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
467 * block group is in the logical address space, which can be any
468 * sectorsize aligned bytenr in the range [0, U64_MAX].
469 */
470 if (range.len < fs_info->sb->s_blocksize)
471 return -EINVAL;
472
473 range.minlen = max(range.minlen, minlen);
474 ret = btrfs_trim_fs(fs_info, &range);
475 if (ret < 0)
476 return ret;
477
478 if (copy_to_user(arg, &range, sizeof(range)))
479 return -EFAULT;
480
481 return 0;
482}
483
484int __pure btrfs_is_empty_uuid(u8 *uuid)
485{
486 int i;
487
488 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
489 if (uuid[i])
490 return 0;
491 }
492 return 1;
493}
494
495static noinline int create_subvol(struct inode *dir,
496 struct dentry *dentry,
497 const char *name, int namelen,
498 struct btrfs_qgroup_inherit *inherit)
499{
500 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
501 struct btrfs_trans_handle *trans;
502 struct btrfs_key key;
503 struct btrfs_root_item *root_item;
504 struct btrfs_inode_item *inode_item;
505 struct extent_buffer *leaf;
506 struct btrfs_root *root = BTRFS_I(dir)->root;
507 struct btrfs_root *new_root;
508 struct btrfs_block_rsv block_rsv;
509 struct timespec64 cur_time = current_time(dir);
510 struct inode *inode;
511 int ret;
512 int err;
513 dev_t anon_dev = 0;
514 u64 objectid;
515 u64 index = 0;
516
517 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
518 if (!root_item)
519 return -ENOMEM;
520
521 ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
522 if (ret)
523 goto fail_free;
524
525 ret = get_anon_bdev(&anon_dev);
526 if (ret < 0)
527 goto fail_free;
528
529 /*
530 * Don't create subvolume whose level is not zero. Or qgroup will be
531 * screwed up since it assumes subvolume qgroup's level to be 0.
532 */
533 if (btrfs_qgroup_level(objectid)) {
534 ret = -ENOSPC;
535 goto fail_free;
536 }
537
538 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
539 /*
540 * The same as the snapshot creation, please see the comment
541 * of create_snapshot().
542 */
543 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false);
544 if (ret)
545 goto fail_free;
546
547 trans = btrfs_start_transaction(root, 0);
548 if (IS_ERR(trans)) {
549 ret = PTR_ERR(trans);
550 btrfs_subvolume_release_metadata(root, &block_rsv);
551 goto fail_free;
552 }
553 trans->block_rsv = &block_rsv;
554 trans->bytes_reserved = block_rsv.size;
555
556 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
557 if (ret)
558 goto fail;
559
560 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
561 BTRFS_NESTING_NORMAL);
562 if (IS_ERR(leaf)) {
563 ret = PTR_ERR(leaf);
564 goto fail;
565 }
566
567 btrfs_mark_buffer_dirty(leaf);
568
569 inode_item = &root_item->inode;
570 btrfs_set_stack_inode_generation(inode_item, 1);
571 btrfs_set_stack_inode_size(inode_item, 3);
572 btrfs_set_stack_inode_nlink(inode_item, 1);
573 btrfs_set_stack_inode_nbytes(inode_item,
574 fs_info->nodesize);
575 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
576
577 btrfs_set_root_flags(root_item, 0);
578 btrfs_set_root_limit(root_item, 0);
579 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
580
581 btrfs_set_root_bytenr(root_item, leaf->start);
582 btrfs_set_root_generation(root_item, trans->transid);
583 btrfs_set_root_level(root_item, 0);
584 btrfs_set_root_refs(root_item, 1);
585 btrfs_set_root_used(root_item, leaf->len);
586 btrfs_set_root_last_snapshot(root_item, 0);
587
588 btrfs_set_root_generation_v2(root_item,
589 btrfs_root_generation(root_item));
590 generate_random_guid(root_item->uuid);
591 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
592 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
593 root_item->ctime = root_item->otime;
594 btrfs_set_root_ctransid(root_item, trans->transid);
595 btrfs_set_root_otransid(root_item, trans->transid);
596
597 btrfs_tree_unlock(leaf);
598
599 btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
600
601 key.objectid = objectid;
602 key.offset = 0;
603 key.type = BTRFS_ROOT_ITEM_KEY;
604 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
605 root_item);
606 if (ret) {
607 /*
608 * Since we don't abort the transaction in this case, free the
609 * tree block so that we don't leak space and leave the
610 * filesystem in an inconsistent state (an extent item in the
611 * extent tree without backreferences). Also no need to have
612 * the tree block locked since it is not in any tree at this
613 * point, so no other task can find it and use it.
614 */
615 btrfs_free_tree_block(trans, root, leaf, 0, 1);
616 free_extent_buffer(leaf);
617 goto fail;
618 }
619
620 free_extent_buffer(leaf);
621 leaf = NULL;
622
623 key.offset = (u64)-1;
624 new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
625 if (IS_ERR(new_root)) {
626 free_anon_bdev(anon_dev);
627 ret = PTR_ERR(new_root);
628 btrfs_abort_transaction(trans, ret);
629 goto fail;
630 }
631 /* Freeing will be done in btrfs_put_root() of new_root */
632 anon_dev = 0;
633
634 ret = btrfs_record_root_in_trans(trans, new_root);
635 if (ret) {
636 btrfs_put_root(new_root);
637 btrfs_abort_transaction(trans, ret);
638 goto fail;
639 }
640
641 ret = btrfs_create_subvol_root(trans, new_root, root);
642 btrfs_put_root(new_root);
643 if (ret) {
644 /* We potentially lose an unused inode item here */
645 btrfs_abort_transaction(trans, ret);
646 goto fail;
647 }
648
649 /*
650 * insert the directory item
651 */
652 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
653 if (ret) {
654 btrfs_abort_transaction(trans, ret);
655 goto fail;
656 }
657
658 ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
659 BTRFS_FT_DIR, index);
660 if (ret) {
661 btrfs_abort_transaction(trans, ret);
662 goto fail;
663 }
664
665 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
666 ret = btrfs_update_inode(trans, root, BTRFS_I(dir));
667 if (ret) {
668 btrfs_abort_transaction(trans, ret);
669 goto fail;
670 }
671
672 ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
673 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
674 if (ret) {
675 btrfs_abort_transaction(trans, ret);
676 goto fail;
677 }
678
679 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
680 BTRFS_UUID_KEY_SUBVOL, objectid);
681 if (ret)
682 btrfs_abort_transaction(trans, ret);
683
684fail:
685 kfree(root_item);
686 trans->block_rsv = NULL;
687 trans->bytes_reserved = 0;
688 btrfs_subvolume_release_metadata(root, &block_rsv);
689
690 err = btrfs_commit_transaction(trans);
691 if (err && !ret)
692 ret = err;
693
694 if (!ret) {
695 inode = btrfs_lookup_dentry(dir, dentry);
696 if (IS_ERR(inode))
697 return PTR_ERR(inode);
698 d_instantiate(dentry, inode);
699 }
700 return ret;
701
702fail_free:
703 if (anon_dev)
704 free_anon_bdev(anon_dev);
705 kfree(root_item);
706 return ret;
707}
708
709static int create_snapshot(struct btrfs_root *root, struct inode *dir,
710 struct dentry *dentry, bool readonly,
711 struct btrfs_qgroup_inherit *inherit)
712{
713 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
714 struct inode *inode;
715 struct btrfs_pending_snapshot *pending_snapshot;
716 struct btrfs_trans_handle *trans;
717 int ret;
718
719 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
720 return -EINVAL;
721
722 if (atomic_read(&root->nr_swapfiles)) {
723 btrfs_warn(fs_info,
724 "cannot snapshot subvolume with active swapfile");
725 return -ETXTBSY;
726 }
727
728 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
729 if (!pending_snapshot)
730 return -ENOMEM;
731
732 ret = get_anon_bdev(&pending_snapshot->anon_dev);
733 if (ret < 0)
734 goto free_pending;
735 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
736 GFP_KERNEL);
737 pending_snapshot->path = btrfs_alloc_path();
738 if (!pending_snapshot->root_item || !pending_snapshot->path) {
739 ret = -ENOMEM;
740 goto free_pending;
741 }
742
743 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
744 BTRFS_BLOCK_RSV_TEMP);
745 /*
746 * 1 - parent dir inode
747 * 2 - dir entries
748 * 1 - root item
749 * 2 - root ref/backref
750 * 1 - root of snapshot
751 * 1 - UUID item
752 */
753 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
754 &pending_snapshot->block_rsv, 8,
755 false);
756 if (ret)
757 goto free_pending;
758
759 pending_snapshot->dentry = dentry;
760 pending_snapshot->root = root;
761 pending_snapshot->readonly = readonly;
762 pending_snapshot->dir = dir;
763 pending_snapshot->inherit = inherit;
764
765 trans = btrfs_start_transaction(root, 0);
766 if (IS_ERR(trans)) {
767 ret = PTR_ERR(trans);
768 goto fail;
769 }
770
771 spin_lock(&fs_info->trans_lock);
772 list_add(&pending_snapshot->list,
773 &trans->transaction->pending_snapshots);
774 spin_unlock(&fs_info->trans_lock);
775
776 ret = btrfs_commit_transaction(trans);
777 if (ret)
778 goto fail;
779
780 ret = pending_snapshot->error;
781 if (ret)
782 goto fail;
783
784 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
785 if (ret)
786 goto fail;
787
788 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
789 if (IS_ERR(inode)) {
790 ret = PTR_ERR(inode);
791 goto fail;
792 }
793
794 d_instantiate(dentry, inode);
795 ret = 0;
796 pending_snapshot->anon_dev = 0;
797fail:
798 /* Prevent double freeing of anon_dev */
799 if (ret && pending_snapshot->snap)
800 pending_snapshot->snap->anon_dev = 0;
801 btrfs_put_root(pending_snapshot->snap);
802 btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
803free_pending:
804 if (pending_snapshot->anon_dev)
805 free_anon_bdev(pending_snapshot->anon_dev);
806 kfree(pending_snapshot->root_item);
807 btrfs_free_path(pending_snapshot->path);
808 kfree(pending_snapshot);
809
810 return ret;
811}
812
813/* copy of may_delete in fs/namei.c()
814 * Check whether we can remove a link victim from directory dir, check
815 * whether the type of victim is right.
816 * 1. We can't do it if dir is read-only (done in permission())
817 * 2. We should have write and exec permissions on dir
818 * 3. We can't remove anything from append-only dir
819 * 4. We can't do anything with immutable dir (done in permission())
820 * 5. If the sticky bit on dir is set we should either
821 * a. be owner of dir, or
822 * b. be owner of victim, or
823 * c. have CAP_FOWNER capability
824 * 6. If the victim is append-only or immutable we can't do anything with
825 * links pointing to it.
826 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
827 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
828 * 9. We can't remove a root or mountpoint.
829 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
830 * nfs_async_unlink().
831 */
832
833static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
834{
835 int error;
836
837 if (d_really_is_negative(victim))
838 return -ENOENT;
839
840 BUG_ON(d_inode(victim->d_parent) != dir);
841 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
842
843 error = inode_permission(&init_user_ns, dir, MAY_WRITE | MAY_EXEC);
844 if (error)
845 return error;
846 if (IS_APPEND(dir))
847 return -EPERM;
848 if (check_sticky(&init_user_ns, dir, d_inode(victim)) ||
849 IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
850 IS_SWAPFILE(d_inode(victim)))
851 return -EPERM;
852 if (isdir) {
853 if (!d_is_dir(victim))
854 return -ENOTDIR;
855 if (IS_ROOT(victim))
856 return -EBUSY;
857 } else if (d_is_dir(victim))
858 return -EISDIR;
859 if (IS_DEADDIR(dir))
860 return -ENOENT;
861 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
862 return -EBUSY;
863 return 0;
864}
865
866/* copy of may_create in fs/namei.c() */
867static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
868{
869 if (d_really_is_positive(child))
870 return -EEXIST;
871 if (IS_DEADDIR(dir))
872 return -ENOENT;
873 return inode_permission(&init_user_ns, dir, MAY_WRITE | MAY_EXEC);
874}
875
876/*
877 * Create a new subvolume below @parent. This is largely modeled after
878 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
879 * inside this filesystem so it's quite a bit simpler.
880 */
881static noinline int btrfs_mksubvol(const struct path *parent,
882 const char *name, int namelen,
883 struct btrfs_root *snap_src,
884 bool readonly,
885 struct btrfs_qgroup_inherit *inherit)
886{
887 struct inode *dir = d_inode(parent->dentry);
888 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
889 struct dentry *dentry;
890 int error;
891
892 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
893 if (error == -EINTR)
894 return error;
895
896 dentry = lookup_one_len(name, parent->dentry, namelen);
897 error = PTR_ERR(dentry);
898 if (IS_ERR(dentry))
899 goto out_unlock;
900
901 error = btrfs_may_create(dir, dentry);
902 if (error)
903 goto out_dput;
904
905 /*
906 * even if this name doesn't exist, we may get hash collisions.
907 * check for them now when we can safely fail
908 */
909 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
910 dir->i_ino, name,
911 namelen);
912 if (error)
913 goto out_dput;
914
915 down_read(&fs_info->subvol_sem);
916
917 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
918 goto out_up_read;
919
920 if (snap_src)
921 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
922 else
923 error = create_subvol(dir, dentry, name, namelen, inherit);
924
925 if (!error)
926 fsnotify_mkdir(dir, dentry);
927out_up_read:
928 up_read(&fs_info->subvol_sem);
929out_dput:
930 dput(dentry);
931out_unlock:
932 btrfs_inode_unlock(dir, 0);
933 return error;
934}
935
936static noinline int btrfs_mksnapshot(const struct path *parent,
937 const char *name, int namelen,
938 struct btrfs_root *root,
939 bool readonly,
940 struct btrfs_qgroup_inherit *inherit)
941{
942 int ret;
943 bool snapshot_force_cow = false;
944
945 /*
946 * Force new buffered writes to reserve space even when NOCOW is
947 * possible. This is to avoid later writeback (running dealloc) to
948 * fallback to COW mode and unexpectedly fail with ENOSPC.
949 */
950 btrfs_drew_read_lock(&root->snapshot_lock);
951
952 ret = btrfs_start_delalloc_snapshot(root, false);
953 if (ret)
954 goto out;
955
956 /*
957 * All previous writes have started writeback in NOCOW mode, so now
958 * we force future writes to fallback to COW mode during snapshot
959 * creation.
960 */
961 atomic_inc(&root->snapshot_force_cow);
962 snapshot_force_cow = true;
963
964 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
965
966 ret = btrfs_mksubvol(parent, name, namelen,
967 root, readonly, inherit);
968out:
969 if (snapshot_force_cow)
970 atomic_dec(&root->snapshot_force_cow);
971 btrfs_drew_read_unlock(&root->snapshot_lock);
972 return ret;
973}
974
975/*
976 * When we're defragging a range, we don't want to kick it off again
977 * if it is really just waiting for delalloc to send it down.
978 * If we find a nice big extent or delalloc range for the bytes in the
979 * file you want to defrag, we return 0 to let you know to skip this
980 * part of the file
981 */
982static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
983{
984 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
985 struct extent_map *em = NULL;
986 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
987 u64 end;
988
989 read_lock(&em_tree->lock);
990 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
991 read_unlock(&em_tree->lock);
992
993 if (em) {
994 end = extent_map_end(em);
995 free_extent_map(em);
996 if (end - offset > thresh)
997 return 0;
998 }
999 /* if we already have a nice delalloc here, just stop */
1000 thresh /= 2;
1001 end = count_range_bits(io_tree, &offset, offset + thresh,
1002 thresh, EXTENT_DELALLOC, 1);
1003 if (end >= thresh)
1004 return 0;
1005 return 1;
1006}
1007
1008/*
1009 * helper function to walk through a file and find extents
1010 * newer than a specific transid, and smaller than thresh.
1011 *
1012 * This is used by the defragging code to find new and small
1013 * extents
1014 */
1015static int find_new_extents(struct btrfs_root *root,
1016 struct inode *inode, u64 newer_than,
1017 u64 *off, u32 thresh)
1018{
1019 struct btrfs_path *path;
1020 struct btrfs_key min_key;
1021 struct extent_buffer *leaf;
1022 struct btrfs_file_extent_item *extent;
1023 int type;
1024 int ret;
1025 u64 ino = btrfs_ino(BTRFS_I(inode));
1026
1027 path = btrfs_alloc_path();
1028 if (!path)
1029 return -ENOMEM;
1030
1031 min_key.objectid = ino;
1032 min_key.type = BTRFS_EXTENT_DATA_KEY;
1033 min_key.offset = *off;
1034
1035 while (1) {
1036 ret = btrfs_search_forward(root, &min_key, path, newer_than);
1037 if (ret != 0)
1038 goto none;
1039process_slot:
1040 if (min_key.objectid != ino)
1041 goto none;
1042 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1043 goto none;
1044
1045 leaf = path->nodes[0];
1046 extent = btrfs_item_ptr(leaf, path->slots[0],
1047 struct btrfs_file_extent_item);
1048
1049 type = btrfs_file_extent_type(leaf, extent);
1050 if (type == BTRFS_FILE_EXTENT_REG &&
1051 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1052 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1053 *off = min_key.offset;
1054 btrfs_free_path(path);
1055 return 0;
1056 }
1057
1058 path->slots[0]++;
1059 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1060 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1061 goto process_slot;
1062 }
1063
1064 if (min_key.offset == (u64)-1)
1065 goto none;
1066
1067 min_key.offset++;
1068 btrfs_release_path(path);
1069 }
1070none:
1071 btrfs_free_path(path);
1072 return -ENOENT;
1073}
1074
1075static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1076{
1077 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1078 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1079 struct extent_map *em;
1080 u64 len = PAGE_SIZE;
1081
1082 /*
1083 * hopefully we have this extent in the tree already, try without
1084 * the full extent lock
1085 */
1086 read_lock(&em_tree->lock);
1087 em = lookup_extent_mapping(em_tree, start, len);
1088 read_unlock(&em_tree->lock);
1089
1090 if (!em) {
1091 struct extent_state *cached = NULL;
1092 u64 end = start + len - 1;
1093
1094 /* get the big lock and read metadata off disk */
1095 lock_extent_bits(io_tree, start, end, &cached);
1096 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len);
1097 unlock_extent_cached(io_tree, start, end, &cached);
1098
1099 if (IS_ERR(em))
1100 return NULL;
1101 }
1102
1103 return em;
1104}
1105
1106static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1107{
1108 struct extent_map *next;
1109 bool ret = true;
1110
1111 /* this is the last extent */
1112 if (em->start + em->len >= i_size_read(inode))
1113 return false;
1114
1115 next = defrag_lookup_extent(inode, em->start + em->len);
1116 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1117 ret = false;
1118 else if ((em->block_start + em->block_len == next->block_start) &&
1119 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1120 ret = false;
1121
1122 free_extent_map(next);
1123 return ret;
1124}
1125
1126static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1127 u64 *last_len, u64 *skip, u64 *defrag_end,
1128 int compress)
1129{
1130 struct extent_map *em;
1131 int ret = 1;
1132 bool next_mergeable = true;
1133 bool prev_mergeable = true;
1134
1135 /*
1136 * make sure that once we start defragging an extent, we keep on
1137 * defragging it
1138 */
1139 if (start < *defrag_end)
1140 return 1;
1141
1142 *skip = 0;
1143
1144 em = defrag_lookup_extent(inode, start);
1145 if (!em)
1146 return 0;
1147
1148 /* this will cover holes, and inline extents */
1149 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1150 ret = 0;
1151 goto out;
1152 }
1153
1154 if (!*defrag_end)
1155 prev_mergeable = false;
1156
1157 next_mergeable = defrag_check_next_extent(inode, em);
1158 /*
1159 * we hit a real extent, if it is big or the next extent is not a
1160 * real extent, don't bother defragging it
1161 */
1162 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1163 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1164 ret = 0;
1165out:
1166 /*
1167 * last_len ends up being a counter of how many bytes we've defragged.
1168 * every time we choose not to defrag an extent, we reset *last_len
1169 * so that the next tiny extent will force a defrag.
1170 *
1171 * The end result of this is that tiny extents before a single big
1172 * extent will force at least part of that big extent to be defragged.
1173 */
1174 if (ret) {
1175 *defrag_end = extent_map_end(em);
1176 } else {
1177 *last_len = 0;
1178 *skip = extent_map_end(em);
1179 *defrag_end = 0;
1180 }
1181
1182 free_extent_map(em);
1183 return ret;
1184}
1185
1186/*
1187 * it doesn't do much good to defrag one or two pages
1188 * at a time. This pulls in a nice chunk of pages
1189 * to COW and defrag.
1190 *
1191 * It also makes sure the delalloc code has enough
1192 * dirty data to avoid making new small extents as part
1193 * of the defrag
1194 *
1195 * It's a good idea to start RA on this range
1196 * before calling this.
1197 */
1198static int cluster_pages_for_defrag(struct inode *inode,
1199 struct page **pages,
1200 unsigned long start_index,
1201 unsigned long num_pages)
1202{
1203 unsigned long file_end;
1204 u64 isize = i_size_read(inode);
1205 u64 page_start;
1206 u64 page_end;
1207 u64 page_cnt;
1208 u64 start = (u64)start_index << PAGE_SHIFT;
1209 u64 search_start;
1210 int ret;
1211 int i;
1212 int i_done;
1213 struct btrfs_ordered_extent *ordered;
1214 struct extent_state *cached_state = NULL;
1215 struct extent_io_tree *tree;
1216 struct extent_changeset *data_reserved = NULL;
1217 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1218
1219 file_end = (isize - 1) >> PAGE_SHIFT;
1220 if (!isize || start_index > file_end)
1221 return 0;
1222
1223 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1224
1225 ret = btrfs_delalloc_reserve_space(BTRFS_I(inode), &data_reserved,
1226 start, page_cnt << PAGE_SHIFT);
1227 if (ret)
1228 return ret;
1229 i_done = 0;
1230 tree = &BTRFS_I(inode)->io_tree;
1231
1232 /* step one, lock all the pages */
1233 for (i = 0; i < page_cnt; i++) {
1234 struct page *page;
1235again:
1236 page = find_or_create_page(inode->i_mapping,
1237 start_index + i, mask);
1238 if (!page)
1239 break;
1240
1241 ret = set_page_extent_mapped(page);
1242 if (ret < 0) {
1243 unlock_page(page);
1244 put_page(page);
1245 break;
1246 }
1247
1248 page_start = page_offset(page);
1249 page_end = page_start + PAGE_SIZE - 1;
1250 while (1) {
1251 lock_extent_bits(tree, page_start, page_end,
1252 &cached_state);
1253 ordered = btrfs_lookup_ordered_extent(BTRFS_I(inode),
1254 page_start);
1255 unlock_extent_cached(tree, page_start, page_end,
1256 &cached_state);
1257 if (!ordered)
1258 break;
1259
1260 unlock_page(page);
1261 btrfs_start_ordered_extent(ordered, 1);
1262 btrfs_put_ordered_extent(ordered);
1263 lock_page(page);
1264 /*
1265 * we unlocked the page above, so we need check if
1266 * it was released or not.
1267 */
1268 if (page->mapping != inode->i_mapping) {
1269 unlock_page(page);
1270 put_page(page);
1271 goto again;
1272 }
1273 }
1274
1275 if (!PageUptodate(page)) {
1276 btrfs_readpage(NULL, page);
1277 lock_page(page);
1278 if (!PageUptodate(page)) {
1279 unlock_page(page);
1280 put_page(page);
1281 ret = -EIO;
1282 break;
1283 }
1284 }
1285
1286 if (page->mapping != inode->i_mapping) {
1287 unlock_page(page);
1288 put_page(page);
1289 goto again;
1290 }
1291
1292 pages[i] = page;
1293 i_done++;
1294 }
1295 if (!i_done || ret)
1296 goto out;
1297
1298 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1299 goto out;
1300
1301 /*
1302 * so now we have a nice long stream of locked
1303 * and up to date pages, lets wait on them
1304 */
1305 for (i = 0; i < i_done; i++)
1306 wait_on_page_writeback(pages[i]);
1307
1308 page_start = page_offset(pages[0]);
1309 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1310
1311 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1312 page_start, page_end - 1, &cached_state);
1313
1314 /*
1315 * When defragmenting we skip ranges that have holes or inline extents,
1316 * (check should_defrag_range()), to avoid unnecessary IO and wasting
1317 * space. At btrfs_defrag_file(), we check if a range should be defragged
1318 * before locking the inode and then, if it should, we trigger a sync
1319 * page cache readahead - we lock the inode only after that to avoid
1320 * blocking for too long other tasks that possibly want to operate on
1321 * other file ranges. But before we were able to get the inode lock,
1322 * some other task may have punched a hole in the range, or we may have
1323 * now an inline extent, in which case we should not defrag. So check
1324 * for that here, where we have the inode and the range locked, and bail
1325 * out if that happened.
1326 */
1327 search_start = page_start;
1328 while (search_start < page_end) {
1329 struct extent_map *em;
1330
1331 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, search_start,
1332 page_end - search_start);
1333 if (IS_ERR(em)) {
1334 ret = PTR_ERR(em);
1335 goto out_unlock_range;
1336 }
1337 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1338 free_extent_map(em);
1339 /* Ok, 0 means we did not defrag anything */
1340 ret = 0;
1341 goto out_unlock_range;
1342 }
1343 search_start = extent_map_end(em);
1344 free_extent_map(em);
1345 }
1346
1347 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1348 page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
1349 EXTENT_DEFRAG, 0, 0, &cached_state);
1350
1351 if (i_done != page_cnt) {
1352 spin_lock(&BTRFS_I(inode)->lock);
1353 btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
1354 spin_unlock(&BTRFS_I(inode)->lock);
1355 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1356 start, (page_cnt - i_done) << PAGE_SHIFT, true);
1357 }
1358
1359
1360 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1361 &cached_state);
1362
1363 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1364 page_start, page_end - 1, &cached_state);
1365
1366 for (i = 0; i < i_done; i++) {
1367 clear_page_dirty_for_io(pages[i]);
1368 ClearPageChecked(pages[i]);
1369 set_page_dirty(pages[i]);
1370 unlock_page(pages[i]);
1371 put_page(pages[i]);
1372 }
1373 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1374 extent_changeset_free(data_reserved);
1375 return i_done;
1376
1377out_unlock_range:
1378 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1379 page_start, page_end - 1, &cached_state);
1380out:
1381 for (i = 0; i < i_done; i++) {
1382 unlock_page(pages[i]);
1383 put_page(pages[i]);
1384 }
1385 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1386 start, page_cnt << PAGE_SHIFT, true);
1387 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1388 extent_changeset_free(data_reserved);
1389 return ret;
1390
1391}
1392
1393int btrfs_defrag_file(struct inode *inode, struct file *file,
1394 struct btrfs_ioctl_defrag_range_args *range,
1395 u64 newer_than, unsigned long max_to_defrag)
1396{
1397 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1398 struct btrfs_root *root = BTRFS_I(inode)->root;
1399 struct file_ra_state *ra = NULL;
1400 unsigned long last_index;
1401 u64 isize = i_size_read(inode);
1402 u64 last_len = 0;
1403 u64 skip = 0;
1404 u64 defrag_end = 0;
1405 u64 newer_off = range->start;
1406 unsigned long i;
1407 unsigned long ra_index = 0;
1408 int ret;
1409 int defrag_count = 0;
1410 int compress_type = BTRFS_COMPRESS_ZLIB;
1411 u32 extent_thresh = range->extent_thresh;
1412 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1413 unsigned long cluster = max_cluster;
1414 u64 new_align = ~((u64)SZ_128K - 1);
1415 struct page **pages = NULL;
1416 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1417
1418 if (isize == 0)
1419 return 0;
1420
1421 if (range->start >= isize)
1422 return -EINVAL;
1423
1424 if (do_compress) {
1425 if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES)
1426 return -EINVAL;
1427 if (range->compress_type)
1428 compress_type = range->compress_type;
1429 }
1430
1431 if (extent_thresh == 0)
1432 extent_thresh = SZ_256K;
1433
1434 /*
1435 * If we were not given a file, allocate a readahead context. As
1436 * readahead is just an optimization, defrag will work without it so
1437 * we don't error out.
1438 */
1439 if (!file) {
1440 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1441 if (ra)
1442 file_ra_state_init(ra, inode->i_mapping);
1443 } else {
1444 ra = &file->f_ra;
1445 }
1446
1447 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1448 if (!pages) {
1449 ret = -ENOMEM;
1450 goto out_ra;
1451 }
1452
1453 /* find the last page to defrag */
1454 if (range->start + range->len > range->start) {
1455 last_index = min_t(u64, isize - 1,
1456 range->start + range->len - 1) >> PAGE_SHIFT;
1457 } else {
1458 last_index = (isize - 1) >> PAGE_SHIFT;
1459 }
1460
1461 if (newer_than) {
1462 ret = find_new_extents(root, inode, newer_than,
1463 &newer_off, SZ_64K);
1464 if (!ret) {
1465 range->start = newer_off;
1466 /*
1467 * we always align our defrag to help keep
1468 * the extents in the file evenly spaced
1469 */
1470 i = (newer_off & new_align) >> PAGE_SHIFT;
1471 } else
1472 goto out_ra;
1473 } else {
1474 i = range->start >> PAGE_SHIFT;
1475 }
1476 if (!max_to_defrag)
1477 max_to_defrag = last_index - i + 1;
1478
1479 /*
1480 * make writeback starts from i, so the defrag range can be
1481 * written sequentially.
1482 */
1483 if (i < inode->i_mapping->writeback_index)
1484 inode->i_mapping->writeback_index = i;
1485
1486 while (i <= last_index && defrag_count < max_to_defrag &&
1487 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1488 /*
1489 * make sure we stop running if someone unmounts
1490 * the FS
1491 */
1492 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1493 break;
1494
1495 if (btrfs_defrag_cancelled(fs_info)) {
1496 btrfs_debug(fs_info, "defrag_file cancelled");
1497 ret = -EAGAIN;
1498 goto error;
1499 }
1500
1501 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1502 extent_thresh, &last_len, &skip,
1503 &defrag_end, do_compress)){
1504 unsigned long next;
1505 /*
1506 * the should_defrag function tells us how much to skip
1507 * bump our counter by the suggested amount
1508 */
1509 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1510 i = max(i + 1, next);
1511 continue;
1512 }
1513
1514 if (!newer_than) {
1515 cluster = (PAGE_ALIGN(defrag_end) >>
1516 PAGE_SHIFT) - i;
1517 cluster = min(cluster, max_cluster);
1518 } else {
1519 cluster = max_cluster;
1520 }
1521
1522 if (i + cluster > ra_index) {
1523 ra_index = max(i, ra_index);
1524 if (ra)
1525 page_cache_sync_readahead(inode->i_mapping, ra,
1526 file, ra_index, cluster);
1527 ra_index += cluster;
1528 }
1529
1530 btrfs_inode_lock(inode, 0);
1531 if (IS_SWAPFILE(inode)) {
1532 ret = -ETXTBSY;
1533 } else {
1534 if (do_compress)
1535 BTRFS_I(inode)->defrag_compress = compress_type;
1536 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1537 }
1538 if (ret < 0) {
1539 btrfs_inode_unlock(inode, 0);
1540 goto out_ra;
1541 }
1542
1543 defrag_count += ret;
1544 balance_dirty_pages_ratelimited(inode->i_mapping);
1545 btrfs_inode_unlock(inode, 0);
1546
1547 if (newer_than) {
1548 if (newer_off == (u64)-1)
1549 break;
1550
1551 if (ret > 0)
1552 i += ret;
1553
1554 newer_off = max(newer_off + 1,
1555 (u64)i << PAGE_SHIFT);
1556
1557 ret = find_new_extents(root, inode, newer_than,
1558 &newer_off, SZ_64K);
1559 if (!ret) {
1560 range->start = newer_off;
1561 i = (newer_off & new_align) >> PAGE_SHIFT;
1562 } else {
1563 break;
1564 }
1565 } else {
1566 if (ret > 0) {
1567 i += ret;
1568 last_len += ret << PAGE_SHIFT;
1569 } else {
1570 i++;
1571 last_len = 0;
1572 }
1573 }
1574 }
1575
1576 ret = defrag_count;
1577error:
1578 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1579 filemap_flush(inode->i_mapping);
1580 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1581 &BTRFS_I(inode)->runtime_flags))
1582 filemap_flush(inode->i_mapping);
1583 }
1584
1585 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1586 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1587 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1588 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1589 }
1590
1591out_ra:
1592 if (do_compress) {
1593 btrfs_inode_lock(inode, 0);
1594 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1595 btrfs_inode_unlock(inode, 0);
1596 }
1597 if (!file)
1598 kfree(ra);
1599 kfree(pages);
1600 return ret;
1601}
1602
1603/*
1604 * Try to start exclusive operation @type or cancel it if it's running.
1605 *
1606 * Return:
1607 * 0 - normal mode, newly claimed op started
1608 * >0 - normal mode, something else is running,
1609 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1610 * ECANCELED - cancel mode, successful cancel
1611 * ENOTCONN - cancel mode, operation not running anymore
1612 */
1613static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1614 enum btrfs_exclusive_operation type, bool cancel)
1615{
1616 if (!cancel) {
1617 /* Start normal op */
1618 if (!btrfs_exclop_start(fs_info, type))
1619 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1620 /* Exclusive operation is now claimed */
1621 return 0;
1622 }
1623
1624 /* Cancel running op */
1625 if (btrfs_exclop_start_try_lock(fs_info, type)) {
1626 /*
1627 * This blocks any exclop finish from setting it to NONE, so we
1628 * request cancellation. Either it runs and we will wait for it,
1629 * or it has finished and no waiting will happen.
1630 */
1631 atomic_inc(&fs_info->reloc_cancel_req);
1632 btrfs_exclop_start_unlock(fs_info);
1633
1634 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1635 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1636 TASK_INTERRUPTIBLE);
1637
1638 return -ECANCELED;
1639 }
1640
1641 /* Something else is running or none */
1642 return -ENOTCONN;
1643}
1644
1645static noinline int btrfs_ioctl_resize(struct file *file,
1646 void __user *arg)
1647{
1648 struct inode *inode = file_inode(file);
1649 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1650 u64 new_size;
1651 u64 old_size;
1652 u64 devid = 1;
1653 struct btrfs_root *root = BTRFS_I(inode)->root;
1654 struct btrfs_ioctl_vol_args *vol_args;
1655 struct btrfs_trans_handle *trans;
1656 struct btrfs_device *device = NULL;
1657 char *sizestr;
1658 char *retptr;
1659 char *devstr = NULL;
1660 int ret = 0;
1661 int mod = 0;
1662 bool cancel;
1663
1664 if (!capable(CAP_SYS_ADMIN))
1665 return -EPERM;
1666
1667 ret = mnt_want_write_file(file);
1668 if (ret)
1669 return ret;
1670
1671 /*
1672 * Read the arguments before checking exclusivity to be able to
1673 * distinguish regular resize and cancel
1674 */
1675 vol_args = memdup_user(arg, sizeof(*vol_args));
1676 if (IS_ERR(vol_args)) {
1677 ret = PTR_ERR(vol_args);
1678 goto out_drop;
1679 }
1680 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1681 sizestr = vol_args->name;
1682 cancel = (strcmp("cancel", sizestr) == 0);
1683 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1684 if (ret)
1685 goto out_free;
1686 /* Exclusive operation is now claimed */
1687
1688 devstr = strchr(sizestr, ':');
1689 if (devstr) {
1690 sizestr = devstr + 1;
1691 *devstr = '\0';
1692 devstr = vol_args->name;
1693 ret = kstrtoull(devstr, 10, &devid);
1694 if (ret)
1695 goto out_finish;
1696 if (!devid) {
1697 ret = -EINVAL;
1698 goto out_finish;
1699 }
1700 btrfs_info(fs_info, "resizing devid %llu", devid);
1701 }
1702
1703 device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL);
1704 if (!device) {
1705 btrfs_info(fs_info, "resizer unable to find device %llu",
1706 devid);
1707 ret = -ENODEV;
1708 goto out_finish;
1709 }
1710
1711 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1712 btrfs_info(fs_info,
1713 "resizer unable to apply on readonly device %llu",
1714 devid);
1715 ret = -EPERM;
1716 goto out_finish;
1717 }
1718
1719 if (!strcmp(sizestr, "max"))
1720 new_size = device->bdev->bd_inode->i_size;
1721 else {
1722 if (sizestr[0] == '-') {
1723 mod = -1;
1724 sizestr++;
1725 } else if (sizestr[0] == '+') {
1726 mod = 1;
1727 sizestr++;
1728 }
1729 new_size = memparse(sizestr, &retptr);
1730 if (*retptr != '\0' || new_size == 0) {
1731 ret = -EINVAL;
1732 goto out_finish;
1733 }
1734 }
1735
1736 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1737 ret = -EPERM;
1738 goto out_finish;
1739 }
1740
1741 old_size = btrfs_device_get_total_bytes(device);
1742
1743 if (mod < 0) {
1744 if (new_size > old_size) {
1745 ret = -EINVAL;
1746 goto out_finish;
1747 }
1748 new_size = old_size - new_size;
1749 } else if (mod > 0) {
1750 if (new_size > ULLONG_MAX - old_size) {
1751 ret = -ERANGE;
1752 goto out_finish;
1753 }
1754 new_size = old_size + new_size;
1755 }
1756
1757 if (new_size < SZ_256M) {
1758 ret = -EINVAL;
1759 goto out_finish;
1760 }
1761 if (new_size > device->bdev->bd_inode->i_size) {
1762 ret = -EFBIG;
1763 goto out_finish;
1764 }
1765
1766 new_size = round_down(new_size, fs_info->sectorsize);
1767
1768 if (new_size > old_size) {
1769 trans = btrfs_start_transaction(root, 0);
1770 if (IS_ERR(trans)) {
1771 ret = PTR_ERR(trans);
1772 goto out_finish;
1773 }
1774 ret = btrfs_grow_device(trans, device, new_size);
1775 btrfs_commit_transaction(trans);
1776 } else if (new_size < old_size) {
1777 ret = btrfs_shrink_device(device, new_size);
1778 } /* equal, nothing need to do */
1779
1780 if (ret == 0 && new_size != old_size)
1781 btrfs_info_in_rcu(fs_info,
1782 "resize device %s (devid %llu) from %llu to %llu",
1783 rcu_str_deref(device->name), device->devid,
1784 old_size, new_size);
1785out_finish:
1786 btrfs_exclop_finish(fs_info);
1787out_free:
1788 kfree(vol_args);
1789out_drop:
1790 mnt_drop_write_file(file);
1791 return ret;
1792}
1793
1794static noinline int __btrfs_ioctl_snap_create(struct file *file,
1795 const char *name, unsigned long fd, int subvol,
1796 bool readonly,
1797 struct btrfs_qgroup_inherit *inherit)
1798{
1799 int namelen;
1800 int ret = 0;
1801
1802 if (!S_ISDIR(file_inode(file)->i_mode))
1803 return -ENOTDIR;
1804
1805 ret = mnt_want_write_file(file);
1806 if (ret)
1807 goto out;
1808
1809 namelen = strlen(name);
1810 if (strchr(name, '/')) {
1811 ret = -EINVAL;
1812 goto out_drop_write;
1813 }
1814
1815 if (name[0] == '.' &&
1816 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1817 ret = -EEXIST;
1818 goto out_drop_write;
1819 }
1820
1821 if (subvol) {
1822 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1823 NULL, readonly, inherit);
1824 } else {
1825 struct fd src = fdget(fd);
1826 struct inode *src_inode;
1827 if (!src.file) {
1828 ret = -EINVAL;
1829 goto out_drop_write;
1830 }
1831
1832 src_inode = file_inode(src.file);
1833 if (src_inode->i_sb != file_inode(file)->i_sb) {
1834 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1835 "Snapshot src from another FS");
1836 ret = -EXDEV;
1837 } else if (!inode_owner_or_capable(&init_user_ns, src_inode)) {
1838 /*
1839 * Subvolume creation is not restricted, but snapshots
1840 * are limited to own subvolumes only
1841 */
1842 ret = -EPERM;
1843 } else {
1844 ret = btrfs_mksnapshot(&file->f_path, name, namelen,
1845 BTRFS_I(src_inode)->root,
1846 readonly, inherit);
1847 }
1848 fdput(src);
1849 }
1850out_drop_write:
1851 mnt_drop_write_file(file);
1852out:
1853 return ret;
1854}
1855
1856static noinline int btrfs_ioctl_snap_create(struct file *file,
1857 void __user *arg, int subvol)
1858{
1859 struct btrfs_ioctl_vol_args *vol_args;
1860 int ret;
1861
1862 if (!S_ISDIR(file_inode(file)->i_mode))
1863 return -ENOTDIR;
1864
1865 vol_args = memdup_user(arg, sizeof(*vol_args));
1866 if (IS_ERR(vol_args))
1867 return PTR_ERR(vol_args);
1868 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1869
1870 ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1871 subvol, false, NULL);
1872
1873 kfree(vol_args);
1874 return ret;
1875}
1876
1877static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1878 void __user *arg, int subvol)
1879{
1880 struct btrfs_ioctl_vol_args_v2 *vol_args;
1881 int ret;
1882 bool readonly = false;
1883 struct btrfs_qgroup_inherit *inherit = NULL;
1884
1885 if (!S_ISDIR(file_inode(file)->i_mode))
1886 return -ENOTDIR;
1887
1888 vol_args = memdup_user(arg, sizeof(*vol_args));
1889 if (IS_ERR(vol_args))
1890 return PTR_ERR(vol_args);
1891 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1892
1893 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1894 ret = -EOPNOTSUPP;
1895 goto free_args;
1896 }
1897
1898 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1899 readonly = true;
1900 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1901 u64 nums;
1902
1903 if (vol_args->size < sizeof(*inherit) ||
1904 vol_args->size > PAGE_SIZE) {
1905 ret = -EINVAL;
1906 goto free_args;
1907 }
1908 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1909 if (IS_ERR(inherit)) {
1910 ret = PTR_ERR(inherit);
1911 goto free_args;
1912 }
1913
1914 if (inherit->num_qgroups > PAGE_SIZE ||
1915 inherit->num_ref_copies > PAGE_SIZE ||
1916 inherit->num_excl_copies > PAGE_SIZE) {
1917 ret = -EINVAL;
1918 goto free_inherit;
1919 }
1920
1921 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
1922 2 * inherit->num_excl_copies;
1923 if (vol_args->size != struct_size(inherit, qgroups, nums)) {
1924 ret = -EINVAL;
1925 goto free_inherit;
1926 }
1927 }
1928
1929 ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1930 subvol, readonly, inherit);
1931 if (ret)
1932 goto free_inherit;
1933free_inherit:
1934 kfree(inherit);
1935free_args:
1936 kfree(vol_args);
1937 return ret;
1938}
1939
1940static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1941 void __user *arg)
1942{
1943 struct inode *inode = file_inode(file);
1944 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1945 struct btrfs_root *root = BTRFS_I(inode)->root;
1946 int ret = 0;
1947 u64 flags = 0;
1948
1949 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1950 return -EINVAL;
1951
1952 down_read(&fs_info->subvol_sem);
1953 if (btrfs_root_readonly(root))
1954 flags |= BTRFS_SUBVOL_RDONLY;
1955 up_read(&fs_info->subvol_sem);
1956
1957 if (copy_to_user(arg, &flags, sizeof(flags)))
1958 ret = -EFAULT;
1959
1960 return ret;
1961}
1962
1963static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1964 void __user *arg)
1965{
1966 struct inode *inode = file_inode(file);
1967 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1968 struct btrfs_root *root = BTRFS_I(inode)->root;
1969 struct btrfs_trans_handle *trans;
1970 u64 root_flags;
1971 u64 flags;
1972 int ret = 0;
1973
1974 if (!inode_owner_or_capable(&init_user_ns, inode))
1975 return -EPERM;
1976
1977 ret = mnt_want_write_file(file);
1978 if (ret)
1979 goto out;
1980
1981 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1982 ret = -EINVAL;
1983 goto out_drop_write;
1984 }
1985
1986 if (copy_from_user(&flags, arg, sizeof(flags))) {
1987 ret = -EFAULT;
1988 goto out_drop_write;
1989 }
1990
1991 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1992 ret = -EOPNOTSUPP;
1993 goto out_drop_write;
1994 }
1995
1996 down_write(&fs_info->subvol_sem);
1997
1998 /* nothing to do */
1999 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
2000 goto out_drop_sem;
2001
2002 root_flags = btrfs_root_flags(&root->root_item);
2003 if (flags & BTRFS_SUBVOL_RDONLY) {
2004 btrfs_set_root_flags(&root->root_item,
2005 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2006 } else {
2007 /*
2008 * Block RO -> RW transition if this subvolume is involved in
2009 * send
2010 */
2011 spin_lock(&root->root_item_lock);
2012 if (root->send_in_progress == 0) {
2013 btrfs_set_root_flags(&root->root_item,
2014 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2015 spin_unlock(&root->root_item_lock);
2016 } else {
2017 spin_unlock(&root->root_item_lock);
2018 btrfs_warn(fs_info,
2019 "Attempt to set subvolume %llu read-write during send",
2020 root->root_key.objectid);
2021 ret = -EPERM;
2022 goto out_drop_sem;
2023 }
2024 }
2025
2026 trans = btrfs_start_transaction(root, 1);
2027 if (IS_ERR(trans)) {
2028 ret = PTR_ERR(trans);
2029 goto out_reset;
2030 }
2031
2032 ret = btrfs_update_root(trans, fs_info->tree_root,
2033 &root->root_key, &root->root_item);
2034 if (ret < 0) {
2035 btrfs_end_transaction(trans);
2036 goto out_reset;
2037 }
2038
2039 ret = btrfs_commit_transaction(trans);
2040
2041out_reset:
2042 if (ret)
2043 btrfs_set_root_flags(&root->root_item, root_flags);
2044out_drop_sem:
2045 up_write(&fs_info->subvol_sem);
2046out_drop_write:
2047 mnt_drop_write_file(file);
2048out:
2049 return ret;
2050}
2051
2052static noinline int key_in_sk(struct btrfs_key *key,
2053 struct btrfs_ioctl_search_key *sk)
2054{
2055 struct btrfs_key test;
2056 int ret;
2057
2058 test.objectid = sk->min_objectid;
2059 test.type = sk->min_type;
2060 test.offset = sk->min_offset;
2061
2062 ret = btrfs_comp_cpu_keys(key, &test);
2063 if (ret < 0)
2064 return 0;
2065
2066 test.objectid = sk->max_objectid;
2067 test.type = sk->max_type;
2068 test.offset = sk->max_offset;
2069
2070 ret = btrfs_comp_cpu_keys(key, &test);
2071 if (ret > 0)
2072 return 0;
2073 return 1;
2074}
2075
2076static noinline int copy_to_sk(struct btrfs_path *path,
2077 struct btrfs_key *key,
2078 struct btrfs_ioctl_search_key *sk,
2079 size_t *buf_size,
2080 char __user *ubuf,
2081 unsigned long *sk_offset,
2082 int *num_found)
2083{
2084 u64 found_transid;
2085 struct extent_buffer *leaf;
2086 struct btrfs_ioctl_search_header sh;
2087 struct btrfs_key test;
2088 unsigned long item_off;
2089 unsigned long item_len;
2090 int nritems;
2091 int i;
2092 int slot;
2093 int ret = 0;
2094
2095 leaf = path->nodes[0];
2096 slot = path->slots[0];
2097 nritems = btrfs_header_nritems(leaf);
2098
2099 if (btrfs_header_generation(leaf) > sk->max_transid) {
2100 i = nritems;
2101 goto advance_key;
2102 }
2103 found_transid = btrfs_header_generation(leaf);
2104
2105 for (i = slot; i < nritems; i++) {
2106 item_off = btrfs_item_ptr_offset(leaf, i);
2107 item_len = btrfs_item_size_nr(leaf, i);
2108
2109 btrfs_item_key_to_cpu(leaf, key, i);
2110 if (!key_in_sk(key, sk))
2111 continue;
2112
2113 if (sizeof(sh) + item_len > *buf_size) {
2114 if (*num_found) {
2115 ret = 1;
2116 goto out;
2117 }
2118
2119 /*
2120 * return one empty item back for v1, which does not
2121 * handle -EOVERFLOW
2122 */
2123
2124 *buf_size = sizeof(sh) + item_len;
2125 item_len = 0;
2126 ret = -EOVERFLOW;
2127 }
2128
2129 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2130 ret = 1;
2131 goto out;
2132 }
2133
2134 sh.objectid = key->objectid;
2135 sh.offset = key->offset;
2136 sh.type = key->type;
2137 sh.len = item_len;
2138 sh.transid = found_transid;
2139
2140 /*
2141 * Copy search result header. If we fault then loop again so we
2142 * can fault in the pages and -EFAULT there if there's a
2143 * problem. Otherwise we'll fault and then copy the buffer in
2144 * properly this next time through
2145 */
2146 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
2147 ret = 0;
2148 goto out;
2149 }
2150
2151 *sk_offset += sizeof(sh);
2152
2153 if (item_len) {
2154 char __user *up = ubuf + *sk_offset;
2155 /*
2156 * Copy the item, same behavior as above, but reset the
2157 * * sk_offset so we copy the full thing again.
2158 */
2159 if (read_extent_buffer_to_user_nofault(leaf, up,
2160 item_off, item_len)) {
2161 ret = 0;
2162 *sk_offset -= sizeof(sh);
2163 goto out;
2164 }
2165
2166 *sk_offset += item_len;
2167 }
2168 (*num_found)++;
2169
2170 if (ret) /* -EOVERFLOW from above */
2171 goto out;
2172
2173 if (*num_found >= sk->nr_items) {
2174 ret = 1;
2175 goto out;
2176 }
2177 }
2178advance_key:
2179 ret = 0;
2180 test.objectid = sk->max_objectid;
2181 test.type = sk->max_type;
2182 test.offset = sk->max_offset;
2183 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2184 ret = 1;
2185 else if (key->offset < (u64)-1)
2186 key->offset++;
2187 else if (key->type < (u8)-1) {
2188 key->offset = 0;
2189 key->type++;
2190 } else if (key->objectid < (u64)-1) {
2191 key->offset = 0;
2192 key->type = 0;
2193 key->objectid++;
2194 } else
2195 ret = 1;
2196out:
2197 /*
2198 * 0: all items from this leaf copied, continue with next
2199 * 1: * more items can be copied, but unused buffer is too small
2200 * * all items were found
2201 * Either way, it will stops the loop which iterates to the next
2202 * leaf
2203 * -EOVERFLOW: item was to large for buffer
2204 * -EFAULT: could not copy extent buffer back to userspace
2205 */
2206 return ret;
2207}
2208
2209static noinline int search_ioctl(struct inode *inode,
2210 struct btrfs_ioctl_search_key *sk,
2211 size_t *buf_size,
2212 char __user *ubuf)
2213{
2214 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2215 struct btrfs_root *root;
2216 struct btrfs_key key;
2217 struct btrfs_path *path;
2218 int ret;
2219 int num_found = 0;
2220 unsigned long sk_offset = 0;
2221
2222 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2223 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2224 return -EOVERFLOW;
2225 }
2226
2227 path = btrfs_alloc_path();
2228 if (!path)
2229 return -ENOMEM;
2230
2231 if (sk->tree_id == 0) {
2232 /* search the root of the inode that was passed */
2233 root = btrfs_grab_root(BTRFS_I(inode)->root);
2234 } else {
2235 root = btrfs_get_fs_root(info, sk->tree_id, true);
2236 if (IS_ERR(root)) {
2237 btrfs_free_path(path);
2238 return PTR_ERR(root);
2239 }
2240 }
2241
2242 key.objectid = sk->min_objectid;
2243 key.type = sk->min_type;
2244 key.offset = sk->min_offset;
2245
2246 while (1) {
2247 ret = fault_in_pages_writeable(ubuf + sk_offset,
2248 *buf_size - sk_offset);
2249 if (ret)
2250 break;
2251
2252 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2253 if (ret != 0) {
2254 if (ret > 0)
2255 ret = 0;
2256 goto err;
2257 }
2258 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2259 &sk_offset, &num_found);
2260 btrfs_release_path(path);
2261 if (ret)
2262 break;
2263
2264 }
2265 if (ret > 0)
2266 ret = 0;
2267err:
2268 sk->nr_items = num_found;
2269 btrfs_put_root(root);
2270 btrfs_free_path(path);
2271 return ret;
2272}
2273
2274static noinline int btrfs_ioctl_tree_search(struct file *file,
2275 void __user *argp)
2276{
2277 struct btrfs_ioctl_search_args __user *uargs;
2278 struct btrfs_ioctl_search_key sk;
2279 struct inode *inode;
2280 int ret;
2281 size_t buf_size;
2282
2283 if (!capable(CAP_SYS_ADMIN))
2284 return -EPERM;
2285
2286 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2287
2288 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2289 return -EFAULT;
2290
2291 buf_size = sizeof(uargs->buf);
2292
2293 inode = file_inode(file);
2294 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2295
2296 /*
2297 * In the origin implementation an overflow is handled by returning a
2298 * search header with a len of zero, so reset ret.
2299 */
2300 if (ret == -EOVERFLOW)
2301 ret = 0;
2302
2303 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2304 ret = -EFAULT;
2305 return ret;
2306}
2307
2308static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2309 void __user *argp)
2310{
2311 struct btrfs_ioctl_search_args_v2 __user *uarg;
2312 struct btrfs_ioctl_search_args_v2 args;
2313 struct inode *inode;
2314 int ret;
2315 size_t buf_size;
2316 const size_t buf_limit = SZ_16M;
2317
2318 if (!capable(CAP_SYS_ADMIN))
2319 return -EPERM;
2320
2321 /* copy search header and buffer size */
2322 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2323 if (copy_from_user(&args, uarg, sizeof(args)))
2324 return -EFAULT;
2325
2326 buf_size = args.buf_size;
2327
2328 /* limit result size to 16MB */
2329 if (buf_size > buf_limit)
2330 buf_size = buf_limit;
2331
2332 inode = file_inode(file);
2333 ret = search_ioctl(inode, &args.key, &buf_size,
2334 (char __user *)(&uarg->buf[0]));
2335 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2336 ret = -EFAULT;
2337 else if (ret == -EOVERFLOW &&
2338 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2339 ret = -EFAULT;
2340
2341 return ret;
2342}
2343
2344/*
2345 * Search INODE_REFs to identify path name of 'dirid' directory
2346 * in a 'tree_id' tree. and sets path name to 'name'.
2347 */
2348static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2349 u64 tree_id, u64 dirid, char *name)
2350{
2351 struct btrfs_root *root;
2352 struct btrfs_key key;
2353 char *ptr;
2354 int ret = -1;
2355 int slot;
2356 int len;
2357 int total_len = 0;
2358 struct btrfs_inode_ref *iref;
2359 struct extent_buffer *l;
2360 struct btrfs_path *path;
2361
2362 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2363 name[0]='\0';
2364 return 0;
2365 }
2366
2367 path = btrfs_alloc_path();
2368 if (!path)
2369 return -ENOMEM;
2370
2371 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2372
2373 root = btrfs_get_fs_root(info, tree_id, true);
2374 if (IS_ERR(root)) {
2375 ret = PTR_ERR(root);
2376 root = NULL;
2377 goto out;
2378 }
2379
2380 key.objectid = dirid;
2381 key.type = BTRFS_INODE_REF_KEY;
2382 key.offset = (u64)-1;
2383
2384 while (1) {
2385 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2386 if (ret < 0)
2387 goto out;
2388 else if (ret > 0) {
2389 ret = btrfs_previous_item(root, path, dirid,
2390 BTRFS_INODE_REF_KEY);
2391 if (ret < 0)
2392 goto out;
2393 else if (ret > 0) {
2394 ret = -ENOENT;
2395 goto out;
2396 }
2397 }
2398
2399 l = path->nodes[0];
2400 slot = path->slots[0];
2401 btrfs_item_key_to_cpu(l, &key, slot);
2402
2403 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2404 len = btrfs_inode_ref_name_len(l, iref);
2405 ptr -= len + 1;
2406 total_len += len + 1;
2407 if (ptr < name) {
2408 ret = -ENAMETOOLONG;
2409 goto out;
2410 }
2411
2412 *(ptr + len) = '/';
2413 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2414
2415 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2416 break;
2417
2418 btrfs_release_path(path);
2419 key.objectid = key.offset;
2420 key.offset = (u64)-1;
2421 dirid = key.objectid;
2422 }
2423 memmove(name, ptr, total_len);
2424 name[total_len] = '\0';
2425 ret = 0;
2426out:
2427 btrfs_put_root(root);
2428 btrfs_free_path(path);
2429 return ret;
2430}
2431
2432static int btrfs_search_path_in_tree_user(struct inode *inode,
2433 struct btrfs_ioctl_ino_lookup_user_args *args)
2434{
2435 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2436 struct super_block *sb = inode->i_sb;
2437 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2438 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2439 u64 dirid = args->dirid;
2440 unsigned long item_off;
2441 unsigned long item_len;
2442 struct btrfs_inode_ref *iref;
2443 struct btrfs_root_ref *rref;
2444 struct btrfs_root *root = NULL;
2445 struct btrfs_path *path;
2446 struct btrfs_key key, key2;
2447 struct extent_buffer *leaf;
2448 struct inode *temp_inode;
2449 char *ptr;
2450 int slot;
2451 int len;
2452 int total_len = 0;
2453 int ret;
2454
2455 path = btrfs_alloc_path();
2456 if (!path)
2457 return -ENOMEM;
2458
2459 /*
2460 * If the bottom subvolume does not exist directly under upper_limit,
2461 * construct the path in from the bottom up.
2462 */
2463 if (dirid != upper_limit.objectid) {
2464 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2465
2466 root = btrfs_get_fs_root(fs_info, treeid, true);
2467 if (IS_ERR(root)) {
2468 ret = PTR_ERR(root);
2469 goto out;
2470 }
2471
2472 key.objectid = dirid;
2473 key.type = BTRFS_INODE_REF_KEY;
2474 key.offset = (u64)-1;
2475 while (1) {
2476 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2477 if (ret < 0) {
2478 goto out_put;
2479 } else if (ret > 0) {
2480 ret = btrfs_previous_item(root, path, dirid,
2481 BTRFS_INODE_REF_KEY);
2482 if (ret < 0) {
2483 goto out_put;
2484 } else if (ret > 0) {
2485 ret = -ENOENT;
2486 goto out_put;
2487 }
2488 }
2489
2490 leaf = path->nodes[0];
2491 slot = path->slots[0];
2492 btrfs_item_key_to_cpu(leaf, &key, slot);
2493
2494 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2495 len = btrfs_inode_ref_name_len(leaf, iref);
2496 ptr -= len + 1;
2497 total_len += len + 1;
2498 if (ptr < args->path) {
2499 ret = -ENAMETOOLONG;
2500 goto out_put;
2501 }
2502
2503 *(ptr + len) = '/';
2504 read_extent_buffer(leaf, ptr,
2505 (unsigned long)(iref + 1), len);
2506
2507 /* Check the read+exec permission of this directory */
2508 ret = btrfs_previous_item(root, path, dirid,
2509 BTRFS_INODE_ITEM_KEY);
2510 if (ret < 0) {
2511 goto out_put;
2512 } else if (ret > 0) {
2513 ret = -ENOENT;
2514 goto out_put;
2515 }
2516
2517 leaf = path->nodes[0];
2518 slot = path->slots[0];
2519 btrfs_item_key_to_cpu(leaf, &key2, slot);
2520 if (key2.objectid != dirid) {
2521 ret = -ENOENT;
2522 goto out_put;
2523 }
2524
2525 temp_inode = btrfs_iget(sb, key2.objectid, root);
2526 if (IS_ERR(temp_inode)) {
2527 ret = PTR_ERR(temp_inode);
2528 goto out_put;
2529 }
2530 ret = inode_permission(&init_user_ns, temp_inode,
2531 MAY_READ | MAY_EXEC);
2532 iput(temp_inode);
2533 if (ret) {
2534 ret = -EACCES;
2535 goto out_put;
2536 }
2537
2538 if (key.offset == upper_limit.objectid)
2539 break;
2540 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2541 ret = -EACCES;
2542 goto out_put;
2543 }
2544
2545 btrfs_release_path(path);
2546 key.objectid = key.offset;
2547 key.offset = (u64)-1;
2548 dirid = key.objectid;
2549 }
2550
2551 memmove(args->path, ptr, total_len);
2552 args->path[total_len] = '\0';
2553 btrfs_put_root(root);
2554 root = NULL;
2555 btrfs_release_path(path);
2556 }
2557
2558 /* Get the bottom subvolume's name from ROOT_REF */
2559 key.objectid = treeid;
2560 key.type = BTRFS_ROOT_REF_KEY;
2561 key.offset = args->treeid;
2562 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2563 if (ret < 0) {
2564 goto out;
2565 } else if (ret > 0) {
2566 ret = -ENOENT;
2567 goto out;
2568 }
2569
2570 leaf = path->nodes[0];
2571 slot = path->slots[0];
2572 btrfs_item_key_to_cpu(leaf, &key, slot);
2573
2574 item_off = btrfs_item_ptr_offset(leaf, slot);
2575 item_len = btrfs_item_size_nr(leaf, slot);
2576 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2577 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2578 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2579 ret = -EINVAL;
2580 goto out;
2581 }
2582
2583 /* Copy subvolume's name */
2584 item_off += sizeof(struct btrfs_root_ref);
2585 item_len -= sizeof(struct btrfs_root_ref);
2586 read_extent_buffer(leaf, args->name, item_off, item_len);
2587 args->name[item_len] = 0;
2588
2589out_put:
2590 btrfs_put_root(root);
2591out:
2592 btrfs_free_path(path);
2593 return ret;
2594}
2595
2596static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2597 void __user *argp)
2598{
2599 struct btrfs_ioctl_ino_lookup_args *args;
2600 struct inode *inode;
2601 int ret = 0;
2602
2603 args = memdup_user(argp, sizeof(*args));
2604 if (IS_ERR(args))
2605 return PTR_ERR(args);
2606
2607 inode = file_inode(file);
2608
2609 /*
2610 * Unprivileged query to obtain the containing subvolume root id. The
2611 * path is reset so it's consistent with btrfs_search_path_in_tree.
2612 */
2613 if (args->treeid == 0)
2614 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2615
2616 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2617 args->name[0] = 0;
2618 goto out;
2619 }
2620
2621 if (!capable(CAP_SYS_ADMIN)) {
2622 ret = -EPERM;
2623 goto out;
2624 }
2625
2626 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2627 args->treeid, args->objectid,
2628 args->name);
2629
2630out:
2631 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2632 ret = -EFAULT;
2633
2634 kfree(args);
2635 return ret;
2636}
2637
2638/*
2639 * Version of ino_lookup ioctl (unprivileged)
2640 *
2641 * The main differences from ino_lookup ioctl are:
2642 *
2643 * 1. Read + Exec permission will be checked using inode_permission() during
2644 * path construction. -EACCES will be returned in case of failure.
2645 * 2. Path construction will be stopped at the inode number which corresponds
2646 * to the fd with which this ioctl is called. If constructed path does not
2647 * exist under fd's inode, -EACCES will be returned.
2648 * 3. The name of bottom subvolume is also searched and filled.
2649 */
2650static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2651{
2652 struct btrfs_ioctl_ino_lookup_user_args *args;
2653 struct inode *inode;
2654 int ret;
2655
2656 args = memdup_user(argp, sizeof(*args));
2657 if (IS_ERR(args))
2658 return PTR_ERR(args);
2659
2660 inode = file_inode(file);
2661
2662 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2663 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2664 /*
2665 * The subvolume does not exist under fd with which this is
2666 * called
2667 */
2668 kfree(args);
2669 return -EACCES;
2670 }
2671
2672 ret = btrfs_search_path_in_tree_user(inode, args);
2673
2674 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2675 ret = -EFAULT;
2676
2677 kfree(args);
2678 return ret;
2679}
2680
2681/* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2682static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2683{
2684 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2685 struct btrfs_fs_info *fs_info;
2686 struct btrfs_root *root;
2687 struct btrfs_path *path;
2688 struct btrfs_key key;
2689 struct btrfs_root_item *root_item;
2690 struct btrfs_root_ref *rref;
2691 struct extent_buffer *leaf;
2692 unsigned long item_off;
2693 unsigned long item_len;
2694 struct inode *inode;
2695 int slot;
2696 int ret = 0;
2697
2698 path = btrfs_alloc_path();
2699 if (!path)
2700 return -ENOMEM;
2701
2702 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2703 if (!subvol_info) {
2704 btrfs_free_path(path);
2705 return -ENOMEM;
2706 }
2707
2708 inode = file_inode(file);
2709 fs_info = BTRFS_I(inode)->root->fs_info;
2710
2711 /* Get root_item of inode's subvolume */
2712 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2713 root = btrfs_get_fs_root(fs_info, key.objectid, true);
2714 if (IS_ERR(root)) {
2715 ret = PTR_ERR(root);
2716 goto out_free;
2717 }
2718 root_item = &root->root_item;
2719
2720 subvol_info->treeid = key.objectid;
2721
2722 subvol_info->generation = btrfs_root_generation(root_item);
2723 subvol_info->flags = btrfs_root_flags(root_item);
2724
2725 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2726 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2727 BTRFS_UUID_SIZE);
2728 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2729 BTRFS_UUID_SIZE);
2730
2731 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2732 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2733 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2734
2735 subvol_info->otransid = btrfs_root_otransid(root_item);
2736 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2737 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2738
2739 subvol_info->stransid = btrfs_root_stransid(root_item);
2740 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2741 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2742
2743 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2744 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2745 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2746
2747 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2748 /* Search root tree for ROOT_BACKREF of this subvolume */
2749 key.type = BTRFS_ROOT_BACKREF_KEY;
2750 key.offset = 0;
2751 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2752 if (ret < 0) {
2753 goto out;
2754 } else if (path->slots[0] >=
2755 btrfs_header_nritems(path->nodes[0])) {
2756 ret = btrfs_next_leaf(fs_info->tree_root, path);
2757 if (ret < 0) {
2758 goto out;
2759 } else if (ret > 0) {
2760 ret = -EUCLEAN;
2761 goto out;
2762 }
2763 }
2764
2765 leaf = path->nodes[0];
2766 slot = path->slots[0];
2767 btrfs_item_key_to_cpu(leaf, &key, slot);
2768 if (key.objectid == subvol_info->treeid &&
2769 key.type == BTRFS_ROOT_BACKREF_KEY) {
2770 subvol_info->parent_id = key.offset;
2771
2772 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2773 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2774
2775 item_off = btrfs_item_ptr_offset(leaf, slot)
2776 + sizeof(struct btrfs_root_ref);
2777 item_len = btrfs_item_size_nr(leaf, slot)
2778 - sizeof(struct btrfs_root_ref);
2779 read_extent_buffer(leaf, subvol_info->name,
2780 item_off, item_len);
2781 } else {
2782 ret = -ENOENT;
2783 goto out;
2784 }
2785 }
2786
2787 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2788 ret = -EFAULT;
2789
2790out:
2791 btrfs_put_root(root);
2792out_free:
2793 btrfs_free_path(path);
2794 kfree(subvol_info);
2795 return ret;
2796}
2797
2798/*
2799 * Return ROOT_REF information of the subvolume containing this inode
2800 * except the subvolume name.
2801 */
2802static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2803{
2804 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2805 struct btrfs_root_ref *rref;
2806 struct btrfs_root *root;
2807 struct btrfs_path *path;
2808 struct btrfs_key key;
2809 struct extent_buffer *leaf;
2810 struct inode *inode;
2811 u64 objectid;
2812 int slot;
2813 int ret;
2814 u8 found;
2815
2816 path = btrfs_alloc_path();
2817 if (!path)
2818 return -ENOMEM;
2819
2820 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2821 if (IS_ERR(rootrefs)) {
2822 btrfs_free_path(path);
2823 return PTR_ERR(rootrefs);
2824 }
2825
2826 inode = file_inode(file);
2827 root = BTRFS_I(inode)->root->fs_info->tree_root;
2828 objectid = BTRFS_I(inode)->root->root_key.objectid;
2829
2830 key.objectid = objectid;
2831 key.type = BTRFS_ROOT_REF_KEY;
2832 key.offset = rootrefs->min_treeid;
2833 found = 0;
2834
2835 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2836 if (ret < 0) {
2837 goto out;
2838 } else if (path->slots[0] >=
2839 btrfs_header_nritems(path->nodes[0])) {
2840 ret = btrfs_next_leaf(root, path);
2841 if (ret < 0) {
2842 goto out;
2843 } else if (ret > 0) {
2844 ret = -EUCLEAN;
2845 goto out;
2846 }
2847 }
2848 while (1) {
2849 leaf = path->nodes[0];
2850 slot = path->slots[0];
2851
2852 btrfs_item_key_to_cpu(leaf, &key, slot);
2853 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2854 ret = 0;
2855 goto out;
2856 }
2857
2858 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2859 ret = -EOVERFLOW;
2860 goto out;
2861 }
2862
2863 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2864 rootrefs->rootref[found].treeid = key.offset;
2865 rootrefs->rootref[found].dirid =
2866 btrfs_root_ref_dirid(leaf, rref);
2867 found++;
2868
2869 ret = btrfs_next_item(root, path);
2870 if (ret < 0) {
2871 goto out;
2872 } else if (ret > 0) {
2873 ret = -EUCLEAN;
2874 goto out;
2875 }
2876 }
2877
2878out:
2879 if (!ret || ret == -EOVERFLOW) {
2880 rootrefs->num_items = found;
2881 /* update min_treeid for next search */
2882 if (found)
2883 rootrefs->min_treeid =
2884 rootrefs->rootref[found - 1].treeid + 1;
2885 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2886 ret = -EFAULT;
2887 }
2888
2889 kfree(rootrefs);
2890 btrfs_free_path(path);
2891
2892 return ret;
2893}
2894
2895static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2896 void __user *arg,
2897 bool destroy_v2)
2898{
2899 struct dentry *parent = file->f_path.dentry;
2900 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2901 struct dentry *dentry;
2902 struct inode *dir = d_inode(parent);
2903 struct inode *inode;
2904 struct btrfs_root *root = BTRFS_I(dir)->root;
2905 struct btrfs_root *dest = NULL;
2906 struct btrfs_ioctl_vol_args *vol_args = NULL;
2907 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2908 char *subvol_name, *subvol_name_ptr = NULL;
2909 int subvol_namelen;
2910 int err = 0;
2911 bool destroy_parent = false;
2912
2913 if (destroy_v2) {
2914 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2915 if (IS_ERR(vol_args2))
2916 return PTR_ERR(vol_args2);
2917
2918 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2919 err = -EOPNOTSUPP;
2920 goto out;
2921 }
2922
2923 /*
2924 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2925 * name, same as v1 currently does.
2926 */
2927 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2928 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2929 subvol_name = vol_args2->name;
2930
2931 err = mnt_want_write_file(file);
2932 if (err)
2933 goto out;
2934 } else {
2935 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2936 err = -EINVAL;
2937 goto out;
2938 }
2939
2940 err = mnt_want_write_file(file);
2941 if (err)
2942 goto out;
2943
2944 dentry = btrfs_get_dentry(fs_info->sb,
2945 BTRFS_FIRST_FREE_OBJECTID,
2946 vol_args2->subvolid, 0, 0);
2947 if (IS_ERR(dentry)) {
2948 err = PTR_ERR(dentry);
2949 goto out_drop_write;
2950 }
2951
2952 /*
2953 * Change the default parent since the subvolume being
2954 * deleted can be outside of the current mount point.
2955 */
2956 parent = btrfs_get_parent(dentry);
2957
2958 /*
2959 * At this point dentry->d_name can point to '/' if the
2960 * subvolume we want to destroy is outsite of the
2961 * current mount point, so we need to release the
2962 * current dentry and execute the lookup to return a new
2963 * one with ->d_name pointing to the
2964 * <mount point>/subvol_name.
2965 */
2966 dput(dentry);
2967 if (IS_ERR(parent)) {
2968 err = PTR_ERR(parent);
2969 goto out_drop_write;
2970 }
2971 dir = d_inode(parent);
2972
2973 /*
2974 * If v2 was used with SPEC_BY_ID, a new parent was
2975 * allocated since the subvolume can be outside of the
2976 * current mount point. Later on we need to release this
2977 * new parent dentry.
2978 */
2979 destroy_parent = true;
2980
2981 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2982 fs_info, vol_args2->subvolid);
2983 if (IS_ERR(subvol_name_ptr)) {
2984 err = PTR_ERR(subvol_name_ptr);
2985 goto free_parent;
2986 }
2987 /* subvol_name_ptr is already nul terminated */
2988 subvol_name = (char *)kbasename(subvol_name_ptr);
2989 }
2990 } else {
2991 vol_args = memdup_user(arg, sizeof(*vol_args));
2992 if (IS_ERR(vol_args))
2993 return PTR_ERR(vol_args);
2994
2995 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
2996 subvol_name = vol_args->name;
2997
2998 err = mnt_want_write_file(file);
2999 if (err)
3000 goto out;
3001 }
3002
3003 subvol_namelen = strlen(subvol_name);
3004
3005 if (strchr(subvol_name, '/') ||
3006 strncmp(subvol_name, "..", subvol_namelen) == 0) {
3007 err = -EINVAL;
3008 goto free_subvol_name;
3009 }
3010
3011 if (!S_ISDIR(dir->i_mode)) {
3012 err = -ENOTDIR;
3013 goto free_subvol_name;
3014 }
3015
3016 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
3017 if (err == -EINTR)
3018 goto free_subvol_name;
3019 dentry = lookup_one_len(subvol_name, parent, subvol_namelen);
3020 if (IS_ERR(dentry)) {
3021 err = PTR_ERR(dentry);
3022 goto out_unlock_dir;
3023 }
3024
3025 if (d_really_is_negative(dentry)) {
3026 err = -ENOENT;
3027 goto out_dput;
3028 }
3029
3030 inode = d_inode(dentry);
3031 dest = BTRFS_I(inode)->root;
3032 if (!capable(CAP_SYS_ADMIN)) {
3033 /*
3034 * Regular user. Only allow this with a special mount
3035 * option, when the user has write+exec access to the
3036 * subvol root, and when rmdir(2) would have been
3037 * allowed.
3038 *
3039 * Note that this is _not_ check that the subvol is
3040 * empty or doesn't contain data that we wouldn't
3041 * otherwise be able to delete.
3042 *
3043 * Users who want to delete empty subvols should try
3044 * rmdir(2).
3045 */
3046 err = -EPERM;
3047 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
3048 goto out_dput;
3049
3050 /*
3051 * Do not allow deletion if the parent dir is the same
3052 * as the dir to be deleted. That means the ioctl
3053 * must be called on the dentry referencing the root
3054 * of the subvol, not a random directory contained
3055 * within it.
3056 */
3057 err = -EINVAL;
3058 if (root == dest)
3059 goto out_dput;
3060
3061 err = inode_permission(&init_user_ns, inode,
3062 MAY_WRITE | MAY_EXEC);
3063 if (err)
3064 goto out_dput;
3065 }
3066
3067 /* check if subvolume may be deleted by a user */
3068 err = btrfs_may_delete(dir, dentry, 1);
3069 if (err)
3070 goto out_dput;
3071
3072 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3073 err = -EINVAL;
3074 goto out_dput;
3075 }
3076
3077 btrfs_inode_lock(inode, 0);
3078 err = btrfs_delete_subvolume(dir, dentry);
3079 btrfs_inode_unlock(inode, 0);
3080 if (!err) {
3081 fsnotify_rmdir(dir, dentry);
3082 d_delete(dentry);
3083 }
3084
3085out_dput:
3086 dput(dentry);
3087out_unlock_dir:
3088 btrfs_inode_unlock(dir, 0);
3089free_subvol_name:
3090 kfree(subvol_name_ptr);
3091free_parent:
3092 if (destroy_parent)
3093 dput(parent);
3094out_drop_write:
3095 mnt_drop_write_file(file);
3096out:
3097 kfree(vol_args2);
3098 kfree(vol_args);
3099 return err;
3100}
3101
3102static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
3103{
3104 struct inode *inode = file_inode(file);
3105 struct btrfs_root *root = BTRFS_I(inode)->root;
3106 struct btrfs_ioctl_defrag_range_args *range;
3107 int ret;
3108
3109 ret = mnt_want_write_file(file);
3110 if (ret)
3111 return ret;
3112
3113 if (btrfs_root_readonly(root)) {
3114 ret = -EROFS;
3115 goto out;
3116 }
3117
3118 switch (inode->i_mode & S_IFMT) {
3119 case S_IFDIR:
3120 if (!capable(CAP_SYS_ADMIN)) {
3121 ret = -EPERM;
3122 goto out;
3123 }
3124 ret = btrfs_defrag_root(root);
3125 break;
3126 case S_IFREG:
3127 /*
3128 * Note that this does not check the file descriptor for write
3129 * access. This prevents defragmenting executables that are
3130 * running and allows defrag on files open in read-only mode.
3131 */
3132 if (!capable(CAP_SYS_ADMIN) &&
3133 inode_permission(&init_user_ns, inode, MAY_WRITE)) {
3134 ret = -EPERM;
3135 goto out;
3136 }
3137
3138 range = kzalloc(sizeof(*range), GFP_KERNEL);
3139 if (!range) {
3140 ret = -ENOMEM;
3141 goto out;
3142 }
3143
3144 if (argp) {
3145 if (copy_from_user(range, argp,
3146 sizeof(*range))) {
3147 ret = -EFAULT;
3148 kfree(range);
3149 goto out;
3150 }
3151 /* compression requires us to start the IO */
3152 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3153 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
3154 range->extent_thresh = (u32)-1;
3155 }
3156 } else {
3157 /* the rest are all set to zero by kzalloc */
3158 range->len = (u64)-1;
3159 }
3160 ret = btrfs_defrag_file(file_inode(file), file,
3161 range, BTRFS_OLDEST_GENERATION, 0);
3162 if (ret > 0)
3163 ret = 0;
3164 kfree(range);
3165 break;
3166 default:
3167 ret = -EINVAL;
3168 }
3169out:
3170 mnt_drop_write_file(file);
3171 return ret;
3172}
3173
3174static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
3175{
3176 struct btrfs_ioctl_vol_args *vol_args;
3177 int ret;
3178
3179 if (!capable(CAP_SYS_ADMIN))
3180 return -EPERM;
3181
3182 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD))
3183 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3184
3185 vol_args = memdup_user(arg, sizeof(*vol_args));
3186 if (IS_ERR(vol_args)) {
3187 ret = PTR_ERR(vol_args);
3188 goto out;
3189 }
3190
3191 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3192 ret = btrfs_init_new_device(fs_info, vol_args->name);
3193
3194 if (!ret)
3195 btrfs_info(fs_info, "disk added %s", vol_args->name);
3196
3197 kfree(vol_args);
3198out:
3199 btrfs_exclop_finish(fs_info);
3200 return ret;
3201}
3202
3203static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
3204{
3205 struct inode *inode = file_inode(file);
3206 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3207 struct btrfs_ioctl_vol_args_v2 *vol_args;
3208 struct block_device *bdev = NULL;
3209 fmode_t mode;
3210 int ret;
3211 bool cancel = false;
3212
3213 if (!capable(CAP_SYS_ADMIN))
3214 return -EPERM;
3215
3216 ret = mnt_want_write_file(file);
3217 if (ret)
3218 return ret;
3219
3220 vol_args = memdup_user(arg, sizeof(*vol_args));
3221 if (IS_ERR(vol_args)) {
3222 ret = PTR_ERR(vol_args);
3223 goto err_drop;
3224 }
3225
3226 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
3227 ret = -EOPNOTSUPP;
3228 goto out;
3229 }
3230 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3231 if (!(vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) &&
3232 strcmp("cancel", vol_args->name) == 0)
3233 cancel = true;
3234
3235 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3236 cancel);
3237 if (ret)
3238 goto out;
3239 /* Exclusive operation is now claimed */
3240
3241 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3242 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid, &bdev, &mode);
3243 else
3244 ret = btrfs_rm_device(fs_info, vol_args->name, 0, &bdev, &mode);
3245
3246 btrfs_exclop_finish(fs_info);
3247
3248 if (!ret) {
3249 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3250 btrfs_info(fs_info, "device deleted: id %llu",
3251 vol_args->devid);
3252 else
3253 btrfs_info(fs_info, "device deleted: %s",
3254 vol_args->name);
3255 }
3256out:
3257 kfree(vol_args);
3258err_drop:
3259 mnt_drop_write_file(file);
3260 if (bdev)
3261 blkdev_put(bdev, mode);
3262 return ret;
3263}
3264
3265static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3266{
3267 struct inode *inode = file_inode(file);
3268 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3269 struct btrfs_ioctl_vol_args *vol_args;
3270 struct block_device *bdev = NULL;
3271 fmode_t mode;
3272 int ret;
3273 bool cancel;
3274
3275 if (!capable(CAP_SYS_ADMIN))
3276 return -EPERM;
3277
3278 ret = mnt_want_write_file(file);
3279 if (ret)
3280 return ret;
3281
3282 vol_args = memdup_user(arg, sizeof(*vol_args));
3283 if (IS_ERR(vol_args)) {
3284 ret = PTR_ERR(vol_args);
3285 goto out_drop_write;
3286 }
3287 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3288 cancel = (strcmp("cancel", vol_args->name) == 0);
3289
3290 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3291 cancel);
3292 if (ret == 0) {
3293 ret = btrfs_rm_device(fs_info, vol_args->name, 0, &bdev, &mode);
3294 if (!ret)
3295 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3296 btrfs_exclop_finish(fs_info);
3297 }
3298
3299 kfree(vol_args);
3300out_drop_write:
3301 mnt_drop_write_file(file);
3302 if (bdev)
3303 blkdev_put(bdev, mode);
3304 return ret;
3305}
3306
3307static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3308 void __user *arg)
3309{
3310 struct btrfs_ioctl_fs_info_args *fi_args;
3311 struct btrfs_device *device;
3312 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3313 u64 flags_in;
3314 int ret = 0;
3315
3316 fi_args = memdup_user(arg, sizeof(*fi_args));
3317 if (IS_ERR(fi_args))
3318 return PTR_ERR(fi_args);
3319
3320 flags_in = fi_args->flags;
3321 memset(fi_args, 0, sizeof(*fi_args));
3322
3323 rcu_read_lock();
3324 fi_args->num_devices = fs_devices->num_devices;
3325
3326 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3327 if (device->devid > fi_args->max_id)
3328 fi_args->max_id = device->devid;
3329 }
3330 rcu_read_unlock();
3331
3332 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
3333 fi_args->nodesize = fs_info->nodesize;
3334 fi_args->sectorsize = fs_info->sectorsize;
3335 fi_args->clone_alignment = fs_info->sectorsize;
3336
3337 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
3338 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
3339 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
3340 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
3341 }
3342
3343 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
3344 fi_args->generation = fs_info->generation;
3345 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
3346 }
3347
3348 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
3349 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
3350 sizeof(fi_args->metadata_uuid));
3351 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
3352 }
3353
3354 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3355 ret = -EFAULT;
3356
3357 kfree(fi_args);
3358 return ret;
3359}
3360
3361static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3362 void __user *arg)
3363{
3364 struct btrfs_ioctl_dev_info_args *di_args;
3365 struct btrfs_device *dev;
3366 int ret = 0;
3367 char *s_uuid = NULL;
3368
3369 di_args = memdup_user(arg, sizeof(*di_args));
3370 if (IS_ERR(di_args))
3371 return PTR_ERR(di_args);
3372
3373 if (!btrfs_is_empty_uuid(di_args->uuid))
3374 s_uuid = di_args->uuid;
3375
3376 rcu_read_lock();
3377 dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid,
3378 NULL);
3379
3380 if (!dev) {
3381 ret = -ENODEV;
3382 goto out;
3383 }
3384
3385 di_args->devid = dev->devid;
3386 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3387 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3388 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3389 if (dev->name) {
3390 strncpy(di_args->path, rcu_str_deref(dev->name),
3391 sizeof(di_args->path) - 1);
3392 di_args->path[sizeof(di_args->path) - 1] = 0;
3393 } else {
3394 di_args->path[0] = '\0';
3395 }
3396
3397out:
3398 rcu_read_unlock();
3399 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3400 ret = -EFAULT;
3401
3402 kfree(di_args);
3403 return ret;
3404}
3405
3406static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3407{
3408 struct inode *inode = file_inode(file);
3409 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3410 struct btrfs_root *root = BTRFS_I(inode)->root;
3411 struct btrfs_root *new_root;
3412 struct btrfs_dir_item *di;
3413 struct btrfs_trans_handle *trans;
3414 struct btrfs_path *path = NULL;
3415 struct btrfs_disk_key disk_key;
3416 u64 objectid = 0;
3417 u64 dir_id;
3418 int ret;
3419
3420 if (!capable(CAP_SYS_ADMIN))
3421 return -EPERM;
3422
3423 ret = mnt_want_write_file(file);
3424 if (ret)
3425 return ret;
3426
3427 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3428 ret = -EFAULT;
3429 goto out;
3430 }
3431
3432 if (!objectid)
3433 objectid = BTRFS_FS_TREE_OBJECTID;
3434
3435 new_root = btrfs_get_fs_root(fs_info, objectid, true);
3436 if (IS_ERR(new_root)) {
3437 ret = PTR_ERR(new_root);
3438 goto out;
3439 }
3440 if (!is_fstree(new_root->root_key.objectid)) {
3441 ret = -ENOENT;
3442 goto out_free;
3443 }
3444
3445 path = btrfs_alloc_path();
3446 if (!path) {
3447 ret = -ENOMEM;
3448 goto out_free;
3449 }
3450
3451 trans = btrfs_start_transaction(root, 1);
3452 if (IS_ERR(trans)) {
3453 ret = PTR_ERR(trans);
3454 goto out_free;
3455 }
3456
3457 dir_id = btrfs_super_root_dir(fs_info->super_copy);
3458 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3459 dir_id, "default", 7, 1);
3460 if (IS_ERR_OR_NULL(di)) {
3461 btrfs_release_path(path);
3462 btrfs_end_transaction(trans);
3463 btrfs_err(fs_info,
3464 "Umm, you don't have the default diritem, this isn't going to work");
3465 ret = -ENOENT;
3466 goto out_free;
3467 }
3468
3469 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3470 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3471 btrfs_mark_buffer_dirty(path->nodes[0]);
3472 btrfs_release_path(path);
3473
3474 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3475 btrfs_end_transaction(trans);
3476out_free:
3477 btrfs_put_root(new_root);
3478 btrfs_free_path(path);
3479out:
3480 mnt_drop_write_file(file);
3481 return ret;
3482}
3483
3484static void get_block_group_info(struct list_head *groups_list,
3485 struct btrfs_ioctl_space_info *space)
3486{
3487 struct btrfs_block_group *block_group;
3488
3489 space->total_bytes = 0;
3490 space->used_bytes = 0;
3491 space->flags = 0;
3492 list_for_each_entry(block_group, groups_list, list) {
3493 space->flags = block_group->flags;
3494 space->total_bytes += block_group->length;
3495 space->used_bytes += block_group->used;
3496 }
3497}
3498
3499static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3500 void __user *arg)
3501{
3502 struct btrfs_ioctl_space_args space_args;
3503 struct btrfs_ioctl_space_info space;
3504 struct btrfs_ioctl_space_info *dest;
3505 struct btrfs_ioctl_space_info *dest_orig;
3506 struct btrfs_ioctl_space_info __user *user_dest;
3507 struct btrfs_space_info *info;
3508 static const u64 types[] = {
3509 BTRFS_BLOCK_GROUP_DATA,
3510 BTRFS_BLOCK_GROUP_SYSTEM,
3511 BTRFS_BLOCK_GROUP_METADATA,
3512 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3513 };
3514 int num_types = 4;
3515 int alloc_size;
3516 int ret = 0;
3517 u64 slot_count = 0;
3518 int i, c;
3519
3520 if (copy_from_user(&space_args,
3521 (struct btrfs_ioctl_space_args __user *)arg,
3522 sizeof(space_args)))
3523 return -EFAULT;
3524
3525 for (i = 0; i < num_types; i++) {
3526 struct btrfs_space_info *tmp;
3527
3528 info = NULL;
3529 list_for_each_entry(tmp, &fs_info->space_info, list) {
3530 if (tmp->flags == types[i]) {
3531 info = tmp;
3532 break;
3533 }
3534 }
3535
3536 if (!info)
3537 continue;
3538
3539 down_read(&info->groups_sem);
3540 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3541 if (!list_empty(&info->block_groups[c]))
3542 slot_count++;
3543 }
3544 up_read(&info->groups_sem);
3545 }
3546
3547 /*
3548 * Global block reserve, exported as a space_info
3549 */
3550 slot_count++;
3551
3552 /* space_slots == 0 means they are asking for a count */
3553 if (space_args.space_slots == 0) {
3554 space_args.total_spaces = slot_count;
3555 goto out;
3556 }
3557
3558 slot_count = min_t(u64, space_args.space_slots, slot_count);
3559
3560 alloc_size = sizeof(*dest) * slot_count;
3561
3562 /* we generally have at most 6 or so space infos, one for each raid
3563 * level. So, a whole page should be more than enough for everyone
3564 */
3565 if (alloc_size > PAGE_SIZE)
3566 return -ENOMEM;
3567
3568 space_args.total_spaces = 0;
3569 dest = kmalloc(alloc_size, GFP_KERNEL);
3570 if (!dest)
3571 return -ENOMEM;
3572 dest_orig = dest;
3573
3574 /* now we have a buffer to copy into */
3575 for (i = 0; i < num_types; i++) {
3576 struct btrfs_space_info *tmp;
3577
3578 if (!slot_count)
3579 break;
3580
3581 info = NULL;
3582 list_for_each_entry(tmp, &fs_info->space_info, list) {
3583 if (tmp->flags == types[i]) {
3584 info = tmp;
3585 break;
3586 }
3587 }
3588
3589 if (!info)
3590 continue;
3591 down_read(&info->groups_sem);
3592 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3593 if (!list_empty(&info->block_groups[c])) {
3594 get_block_group_info(&info->block_groups[c],
3595 &space);
3596 memcpy(dest, &space, sizeof(space));
3597 dest++;
3598 space_args.total_spaces++;
3599 slot_count--;
3600 }
3601 if (!slot_count)
3602 break;
3603 }
3604 up_read(&info->groups_sem);
3605 }
3606
3607 /*
3608 * Add global block reserve
3609 */
3610 if (slot_count) {
3611 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3612
3613 spin_lock(&block_rsv->lock);
3614 space.total_bytes = block_rsv->size;
3615 space.used_bytes = block_rsv->size - block_rsv->reserved;
3616 spin_unlock(&block_rsv->lock);
3617 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3618 memcpy(dest, &space, sizeof(space));
3619 space_args.total_spaces++;
3620 }
3621
3622 user_dest = (struct btrfs_ioctl_space_info __user *)
3623 (arg + sizeof(struct btrfs_ioctl_space_args));
3624
3625 if (copy_to_user(user_dest, dest_orig, alloc_size))
3626 ret = -EFAULT;
3627
3628 kfree(dest_orig);
3629out:
3630 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3631 ret = -EFAULT;
3632
3633 return ret;
3634}
3635
3636static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3637 void __user *argp)
3638{
3639 struct btrfs_trans_handle *trans;
3640 u64 transid;
3641 int ret;
3642
3643 trans = btrfs_attach_transaction_barrier(root);
3644 if (IS_ERR(trans)) {
3645 if (PTR_ERR(trans) != -ENOENT)
3646 return PTR_ERR(trans);
3647
3648 /* No running transaction, don't bother */
3649 transid = root->fs_info->last_trans_committed;
3650 goto out;
3651 }
3652 transid = trans->transid;
3653 ret = btrfs_commit_transaction_async(trans);
3654 if (ret) {
3655 btrfs_end_transaction(trans);
3656 return ret;
3657 }
3658out:
3659 if (argp)
3660 if (copy_to_user(argp, &transid, sizeof(transid)))
3661 return -EFAULT;
3662 return 0;
3663}
3664
3665static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3666 void __user *argp)
3667{
3668 u64 transid;
3669
3670 if (argp) {
3671 if (copy_from_user(&transid, argp, sizeof(transid)))
3672 return -EFAULT;
3673 } else {
3674 transid = 0; /* current trans */
3675 }
3676 return btrfs_wait_for_commit(fs_info, transid);
3677}
3678
3679static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3680{
3681 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3682 struct btrfs_ioctl_scrub_args *sa;
3683 int ret;
3684
3685 if (!capable(CAP_SYS_ADMIN))
3686 return -EPERM;
3687
3688 sa = memdup_user(arg, sizeof(*sa));
3689 if (IS_ERR(sa))
3690 return PTR_ERR(sa);
3691
3692 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3693 ret = mnt_want_write_file(file);
3694 if (ret)
3695 goto out;
3696 }
3697
3698 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3699 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3700 0);
3701
3702 /*
3703 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3704 * error. This is important as it allows user space to know how much
3705 * progress scrub has done. For example, if scrub is canceled we get
3706 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3707 * space. Later user space can inspect the progress from the structure
3708 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3709 * previously (btrfs-progs does this).
3710 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3711 * then return -EFAULT to signal the structure was not copied or it may
3712 * be corrupt and unreliable due to a partial copy.
3713 */
3714 if (copy_to_user(arg, sa, sizeof(*sa)))
3715 ret = -EFAULT;
3716
3717 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3718 mnt_drop_write_file(file);
3719out:
3720 kfree(sa);
3721 return ret;
3722}
3723
3724static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3725{
3726 if (!capable(CAP_SYS_ADMIN))
3727 return -EPERM;
3728
3729 return btrfs_scrub_cancel(fs_info);
3730}
3731
3732static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3733 void __user *arg)
3734{
3735 struct btrfs_ioctl_scrub_args *sa;
3736 int ret;
3737
3738 if (!capable(CAP_SYS_ADMIN))
3739 return -EPERM;
3740
3741 sa = memdup_user(arg, sizeof(*sa));
3742 if (IS_ERR(sa))
3743 return PTR_ERR(sa);
3744
3745 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3746
3747 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3748 ret = -EFAULT;
3749
3750 kfree(sa);
3751 return ret;
3752}
3753
3754static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3755 void __user *arg)
3756{
3757 struct btrfs_ioctl_get_dev_stats *sa;
3758 int ret;
3759
3760 sa = memdup_user(arg, sizeof(*sa));
3761 if (IS_ERR(sa))
3762 return PTR_ERR(sa);
3763
3764 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3765 kfree(sa);
3766 return -EPERM;
3767 }
3768
3769 ret = btrfs_get_dev_stats(fs_info, sa);
3770
3771 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3772 ret = -EFAULT;
3773
3774 kfree(sa);
3775 return ret;
3776}
3777
3778static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3779 void __user *arg)
3780{
3781 struct btrfs_ioctl_dev_replace_args *p;
3782 int ret;
3783
3784 if (!capable(CAP_SYS_ADMIN))
3785 return -EPERM;
3786
3787 p = memdup_user(arg, sizeof(*p));
3788 if (IS_ERR(p))
3789 return PTR_ERR(p);
3790
3791 switch (p->cmd) {
3792 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3793 if (sb_rdonly(fs_info->sb)) {
3794 ret = -EROFS;
3795 goto out;
3796 }
3797 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3798 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3799 } else {
3800 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3801 btrfs_exclop_finish(fs_info);
3802 }
3803 break;
3804 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3805 btrfs_dev_replace_status(fs_info, p);
3806 ret = 0;
3807 break;
3808 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3809 p->result = btrfs_dev_replace_cancel(fs_info);
3810 ret = 0;
3811 break;
3812 default:
3813 ret = -EINVAL;
3814 break;
3815 }
3816
3817 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3818 ret = -EFAULT;
3819out:
3820 kfree(p);
3821 return ret;
3822}
3823
3824static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3825{
3826 int ret = 0;
3827 int i;
3828 u64 rel_ptr;
3829 int size;
3830 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3831 struct inode_fs_paths *ipath = NULL;
3832 struct btrfs_path *path;
3833
3834 if (!capable(CAP_DAC_READ_SEARCH))
3835 return -EPERM;
3836
3837 path = btrfs_alloc_path();
3838 if (!path) {
3839 ret = -ENOMEM;
3840 goto out;
3841 }
3842
3843 ipa = memdup_user(arg, sizeof(*ipa));
3844 if (IS_ERR(ipa)) {
3845 ret = PTR_ERR(ipa);
3846 ipa = NULL;
3847 goto out;
3848 }
3849
3850 size = min_t(u32, ipa->size, 4096);
3851 ipath = init_ipath(size, root, path);
3852 if (IS_ERR(ipath)) {
3853 ret = PTR_ERR(ipath);
3854 ipath = NULL;
3855 goto out;
3856 }
3857
3858 ret = paths_from_inode(ipa->inum, ipath);
3859 if (ret < 0)
3860 goto out;
3861
3862 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3863 rel_ptr = ipath->fspath->val[i] -
3864 (u64)(unsigned long)ipath->fspath->val;
3865 ipath->fspath->val[i] = rel_ptr;
3866 }
3867
3868 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3869 ipath->fspath, size);
3870 if (ret) {
3871 ret = -EFAULT;
3872 goto out;
3873 }
3874
3875out:
3876 btrfs_free_path(path);
3877 free_ipath(ipath);
3878 kfree(ipa);
3879
3880 return ret;
3881}
3882
3883static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3884{
3885 struct btrfs_data_container *inodes = ctx;
3886 const size_t c = 3 * sizeof(u64);
3887
3888 if (inodes->bytes_left >= c) {
3889 inodes->bytes_left -= c;
3890 inodes->val[inodes->elem_cnt] = inum;
3891 inodes->val[inodes->elem_cnt + 1] = offset;
3892 inodes->val[inodes->elem_cnt + 2] = root;
3893 inodes->elem_cnt += 3;
3894 } else {
3895 inodes->bytes_missing += c - inodes->bytes_left;
3896 inodes->bytes_left = 0;
3897 inodes->elem_missed += 3;
3898 }
3899
3900 return 0;
3901}
3902
3903static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3904 void __user *arg, int version)
3905{
3906 int ret = 0;
3907 int size;
3908 struct btrfs_ioctl_logical_ino_args *loi;
3909 struct btrfs_data_container *inodes = NULL;
3910 struct btrfs_path *path = NULL;
3911 bool ignore_offset;
3912
3913 if (!capable(CAP_SYS_ADMIN))
3914 return -EPERM;
3915
3916 loi = memdup_user(arg, sizeof(*loi));
3917 if (IS_ERR(loi))
3918 return PTR_ERR(loi);
3919
3920 if (version == 1) {
3921 ignore_offset = false;
3922 size = min_t(u32, loi->size, SZ_64K);
3923 } else {
3924 /* All reserved bits must be 0 for now */
3925 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3926 ret = -EINVAL;
3927 goto out_loi;
3928 }
3929 /* Only accept flags we have defined so far */
3930 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3931 ret = -EINVAL;
3932 goto out_loi;
3933 }
3934 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3935 size = min_t(u32, loi->size, SZ_16M);
3936 }
3937
3938 path = btrfs_alloc_path();
3939 if (!path) {
3940 ret = -ENOMEM;
3941 goto out;
3942 }
3943
3944 inodes = init_data_container(size);
3945 if (IS_ERR(inodes)) {
3946 ret = PTR_ERR(inodes);
3947 inodes = NULL;
3948 goto out;
3949 }
3950
3951 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3952 build_ino_list, inodes, ignore_offset);
3953 if (ret == -EINVAL)
3954 ret = -ENOENT;
3955 if (ret < 0)
3956 goto out;
3957
3958 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3959 size);
3960 if (ret)
3961 ret = -EFAULT;
3962
3963out:
3964 btrfs_free_path(path);
3965 kvfree(inodes);
3966out_loi:
3967 kfree(loi);
3968
3969 return ret;
3970}
3971
3972void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3973 struct btrfs_ioctl_balance_args *bargs)
3974{
3975 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3976
3977 bargs->flags = bctl->flags;
3978
3979 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3980 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3981 if (atomic_read(&fs_info->balance_pause_req))
3982 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3983 if (atomic_read(&fs_info->balance_cancel_req))
3984 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3985
3986 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3987 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3988 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3989
3990 spin_lock(&fs_info->balance_lock);
3991 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3992 spin_unlock(&fs_info->balance_lock);
3993}
3994
3995static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3996{
3997 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3998 struct btrfs_fs_info *fs_info = root->fs_info;
3999 struct btrfs_ioctl_balance_args *bargs;
4000 struct btrfs_balance_control *bctl;
4001 bool need_unlock; /* for mut. excl. ops lock */
4002 int ret;
4003
4004 if (!capable(CAP_SYS_ADMIN))
4005 return -EPERM;
4006
4007 ret = mnt_want_write_file(file);
4008 if (ret)
4009 return ret;
4010
4011again:
4012 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
4013 mutex_lock(&fs_info->balance_mutex);
4014 need_unlock = true;
4015 goto locked;
4016 }
4017
4018 /*
4019 * mut. excl. ops lock is locked. Three possibilities:
4020 * (1) some other op is running
4021 * (2) balance is running
4022 * (3) balance is paused -- special case (think resume)
4023 */
4024 mutex_lock(&fs_info->balance_mutex);
4025 if (fs_info->balance_ctl) {
4026 /* this is either (2) or (3) */
4027 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4028 mutex_unlock(&fs_info->balance_mutex);
4029 /*
4030 * Lock released to allow other waiters to continue,
4031 * we'll reexamine the status again.
4032 */
4033 mutex_lock(&fs_info->balance_mutex);
4034
4035 if (fs_info->balance_ctl &&
4036 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4037 /* this is (3) */
4038 need_unlock = false;
4039 goto locked;
4040 }
4041
4042 mutex_unlock(&fs_info->balance_mutex);
4043 goto again;
4044 } else {
4045 /* this is (2) */
4046 mutex_unlock(&fs_info->balance_mutex);
4047 ret = -EINPROGRESS;
4048 goto out;
4049 }
4050 } else {
4051 /* this is (1) */
4052 mutex_unlock(&fs_info->balance_mutex);
4053 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4054 goto out;
4055 }
4056
4057locked:
4058
4059 if (arg) {
4060 bargs = memdup_user(arg, sizeof(*bargs));
4061 if (IS_ERR(bargs)) {
4062 ret = PTR_ERR(bargs);
4063 goto out_unlock;
4064 }
4065
4066 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4067 if (!fs_info->balance_ctl) {
4068 ret = -ENOTCONN;
4069 goto out_bargs;
4070 }
4071
4072 bctl = fs_info->balance_ctl;
4073 spin_lock(&fs_info->balance_lock);
4074 bctl->flags |= BTRFS_BALANCE_RESUME;
4075 spin_unlock(&fs_info->balance_lock);
4076
4077 goto do_balance;
4078 }
4079 } else {
4080 bargs = NULL;
4081 }
4082
4083 if (fs_info->balance_ctl) {
4084 ret = -EINPROGRESS;
4085 goto out_bargs;
4086 }
4087
4088 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4089 if (!bctl) {
4090 ret = -ENOMEM;
4091 goto out_bargs;
4092 }
4093
4094 if (arg) {
4095 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4096 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4097 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4098
4099 bctl->flags = bargs->flags;
4100 } else {
4101 /* balance everything - no filters */
4102 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4103 }
4104
4105 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4106 ret = -EINVAL;
4107 goto out_bctl;
4108 }
4109
4110do_balance:
4111 /*
4112 * Ownership of bctl and exclusive operation goes to btrfs_balance.
4113 * bctl is freed in reset_balance_state, or, if restriper was paused
4114 * all the way until unmount, in free_fs_info. The flag should be
4115 * cleared after reset_balance_state.
4116 */
4117 need_unlock = false;
4118
4119 ret = btrfs_balance(fs_info, bctl, bargs);
4120 bctl = NULL;
4121
4122 if ((ret == 0 || ret == -ECANCELED) && arg) {
4123 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4124 ret = -EFAULT;
4125 }
4126
4127out_bctl:
4128 kfree(bctl);
4129out_bargs:
4130 kfree(bargs);
4131out_unlock:
4132 mutex_unlock(&fs_info->balance_mutex);
4133 if (need_unlock)
4134 btrfs_exclop_finish(fs_info);
4135out:
4136 mnt_drop_write_file(file);
4137 return ret;
4138}
4139
4140static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4141{
4142 if (!capable(CAP_SYS_ADMIN))
4143 return -EPERM;
4144
4145 switch (cmd) {
4146 case BTRFS_BALANCE_CTL_PAUSE:
4147 return btrfs_pause_balance(fs_info);
4148 case BTRFS_BALANCE_CTL_CANCEL:
4149 return btrfs_cancel_balance(fs_info);
4150 }
4151
4152 return -EINVAL;
4153}
4154
4155static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4156 void __user *arg)
4157{
4158 struct btrfs_ioctl_balance_args *bargs;
4159 int ret = 0;
4160
4161 if (!capable(CAP_SYS_ADMIN))
4162 return -EPERM;
4163
4164 mutex_lock(&fs_info->balance_mutex);
4165 if (!fs_info->balance_ctl) {
4166 ret = -ENOTCONN;
4167 goto out;
4168 }
4169
4170 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4171 if (!bargs) {
4172 ret = -ENOMEM;
4173 goto out;
4174 }
4175
4176 btrfs_update_ioctl_balance_args(fs_info, bargs);
4177
4178 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4179 ret = -EFAULT;
4180
4181 kfree(bargs);
4182out:
4183 mutex_unlock(&fs_info->balance_mutex);
4184 return ret;
4185}
4186
4187static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4188{
4189 struct inode *inode = file_inode(file);
4190 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4191 struct btrfs_ioctl_quota_ctl_args *sa;
4192 int ret;
4193
4194 if (!capable(CAP_SYS_ADMIN))
4195 return -EPERM;
4196
4197 ret = mnt_want_write_file(file);
4198 if (ret)
4199 return ret;
4200
4201 sa = memdup_user(arg, sizeof(*sa));
4202 if (IS_ERR(sa)) {
4203 ret = PTR_ERR(sa);
4204 goto drop_write;
4205 }
4206
4207 down_write(&fs_info->subvol_sem);
4208
4209 switch (sa->cmd) {
4210 case BTRFS_QUOTA_CTL_ENABLE:
4211 ret = btrfs_quota_enable(fs_info);
4212 break;
4213 case BTRFS_QUOTA_CTL_DISABLE:
4214 ret = btrfs_quota_disable(fs_info);
4215 break;
4216 default:
4217 ret = -EINVAL;
4218 break;
4219 }
4220
4221 kfree(sa);
4222 up_write(&fs_info->subvol_sem);
4223drop_write:
4224 mnt_drop_write_file(file);
4225 return ret;
4226}
4227
4228static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4229{
4230 struct inode *inode = file_inode(file);
4231 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4232 struct btrfs_root *root = BTRFS_I(inode)->root;
4233 struct btrfs_ioctl_qgroup_assign_args *sa;
4234 struct btrfs_trans_handle *trans;
4235 int ret;
4236 int err;
4237
4238 if (!capable(CAP_SYS_ADMIN))
4239 return -EPERM;
4240
4241 ret = mnt_want_write_file(file);
4242 if (ret)
4243 return ret;
4244
4245 sa = memdup_user(arg, sizeof(*sa));
4246 if (IS_ERR(sa)) {
4247 ret = PTR_ERR(sa);
4248 goto drop_write;
4249 }
4250
4251 trans = btrfs_join_transaction(root);
4252 if (IS_ERR(trans)) {
4253 ret = PTR_ERR(trans);
4254 goto out;
4255 }
4256
4257 if (sa->assign) {
4258 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
4259 } else {
4260 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
4261 }
4262
4263 /* update qgroup status and info */
4264 err = btrfs_run_qgroups(trans);
4265 if (err < 0)
4266 btrfs_handle_fs_error(fs_info, err,
4267 "failed to update qgroup status and info");
4268 err = btrfs_end_transaction(trans);
4269 if (err && !ret)
4270 ret = err;
4271
4272out:
4273 kfree(sa);
4274drop_write:
4275 mnt_drop_write_file(file);
4276 return ret;
4277}
4278
4279static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4280{
4281 struct inode *inode = file_inode(file);
4282 struct btrfs_root *root = BTRFS_I(inode)->root;
4283 struct btrfs_ioctl_qgroup_create_args *sa;
4284 struct btrfs_trans_handle *trans;
4285 int ret;
4286 int err;
4287
4288 if (!capable(CAP_SYS_ADMIN))
4289 return -EPERM;
4290
4291 ret = mnt_want_write_file(file);
4292 if (ret)
4293 return ret;
4294
4295 sa = memdup_user(arg, sizeof(*sa));
4296 if (IS_ERR(sa)) {
4297 ret = PTR_ERR(sa);
4298 goto drop_write;
4299 }
4300
4301 if (!sa->qgroupid) {
4302 ret = -EINVAL;
4303 goto out;
4304 }
4305
4306 trans = btrfs_join_transaction(root);
4307 if (IS_ERR(trans)) {
4308 ret = PTR_ERR(trans);
4309 goto out;
4310 }
4311
4312 if (sa->create) {
4313 ret = btrfs_create_qgroup(trans, sa->qgroupid);
4314 } else {
4315 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
4316 }
4317
4318 err = btrfs_end_transaction(trans);
4319 if (err && !ret)
4320 ret = err;
4321
4322out:
4323 kfree(sa);
4324drop_write:
4325 mnt_drop_write_file(file);
4326 return ret;
4327}
4328
4329static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4330{
4331 struct inode *inode = file_inode(file);
4332 struct btrfs_root *root = BTRFS_I(inode)->root;
4333 struct btrfs_ioctl_qgroup_limit_args *sa;
4334 struct btrfs_trans_handle *trans;
4335 int ret;
4336 int err;
4337 u64 qgroupid;
4338
4339 if (!capable(CAP_SYS_ADMIN))
4340 return -EPERM;
4341
4342 ret = mnt_want_write_file(file);
4343 if (ret)
4344 return ret;
4345
4346 sa = memdup_user(arg, sizeof(*sa));
4347 if (IS_ERR(sa)) {
4348 ret = PTR_ERR(sa);
4349 goto drop_write;
4350 }
4351
4352 trans = btrfs_join_transaction(root);
4353 if (IS_ERR(trans)) {
4354 ret = PTR_ERR(trans);
4355 goto out;
4356 }
4357
4358 qgroupid = sa->qgroupid;
4359 if (!qgroupid) {
4360 /* take the current subvol as qgroup */
4361 qgroupid = root->root_key.objectid;
4362 }
4363
4364 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
4365
4366 err = btrfs_end_transaction(trans);
4367 if (err && !ret)
4368 ret = err;
4369
4370out:
4371 kfree(sa);
4372drop_write:
4373 mnt_drop_write_file(file);
4374 return ret;
4375}
4376
4377static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4378{
4379 struct inode *inode = file_inode(file);
4380 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4381 struct btrfs_ioctl_quota_rescan_args *qsa;
4382 int ret;
4383
4384 if (!capable(CAP_SYS_ADMIN))
4385 return -EPERM;
4386
4387 ret = mnt_want_write_file(file);
4388 if (ret)
4389 return ret;
4390
4391 qsa = memdup_user(arg, sizeof(*qsa));
4392 if (IS_ERR(qsa)) {
4393 ret = PTR_ERR(qsa);
4394 goto drop_write;
4395 }
4396
4397 if (qsa->flags) {
4398 ret = -EINVAL;
4399 goto out;
4400 }
4401
4402 ret = btrfs_qgroup_rescan(fs_info);
4403
4404out:
4405 kfree(qsa);
4406drop_write:
4407 mnt_drop_write_file(file);
4408 return ret;
4409}
4410
4411static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4412 void __user *arg)
4413{
4414 struct btrfs_ioctl_quota_rescan_args *qsa;
4415 int ret = 0;
4416
4417 if (!capable(CAP_SYS_ADMIN))
4418 return -EPERM;
4419
4420 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
4421 if (!qsa)
4422 return -ENOMEM;
4423
4424 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4425 qsa->flags = 1;
4426 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
4427 }
4428
4429 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4430 ret = -EFAULT;
4431
4432 kfree(qsa);
4433 return ret;
4434}
4435
4436static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4437 void __user *arg)
4438{
4439 if (!capable(CAP_SYS_ADMIN))
4440 return -EPERM;
4441
4442 return btrfs_qgroup_wait_for_completion(fs_info, true);
4443}
4444
4445static long _btrfs_ioctl_set_received_subvol(struct file *file,
4446 struct btrfs_ioctl_received_subvol_args *sa)
4447{
4448 struct inode *inode = file_inode(file);
4449 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4450 struct btrfs_root *root = BTRFS_I(inode)->root;
4451 struct btrfs_root_item *root_item = &root->root_item;
4452 struct btrfs_trans_handle *trans;
4453 struct timespec64 ct = current_time(inode);
4454 int ret = 0;
4455 int received_uuid_changed;
4456
4457 if (!inode_owner_or_capable(&init_user_ns, inode))
4458 return -EPERM;
4459
4460 ret = mnt_want_write_file(file);
4461 if (ret < 0)
4462 return ret;
4463
4464 down_write(&fs_info->subvol_sem);
4465
4466 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4467 ret = -EINVAL;
4468 goto out;
4469 }
4470
4471 if (btrfs_root_readonly(root)) {
4472 ret = -EROFS;
4473 goto out;
4474 }
4475
4476 /*
4477 * 1 - root item
4478 * 2 - uuid items (received uuid + subvol uuid)
4479 */
4480 trans = btrfs_start_transaction(root, 3);
4481 if (IS_ERR(trans)) {
4482 ret = PTR_ERR(trans);
4483 trans = NULL;
4484 goto out;
4485 }
4486
4487 sa->rtransid = trans->transid;
4488 sa->rtime.sec = ct.tv_sec;
4489 sa->rtime.nsec = ct.tv_nsec;
4490
4491 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4492 BTRFS_UUID_SIZE);
4493 if (received_uuid_changed &&
4494 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4495 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4496 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4497 root->root_key.objectid);
4498 if (ret && ret != -ENOENT) {
4499 btrfs_abort_transaction(trans, ret);
4500 btrfs_end_transaction(trans);
4501 goto out;
4502 }
4503 }
4504 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4505 btrfs_set_root_stransid(root_item, sa->stransid);
4506 btrfs_set_root_rtransid(root_item, sa->rtransid);
4507 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4508 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4509 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4510 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4511
4512 ret = btrfs_update_root(trans, fs_info->tree_root,
4513 &root->root_key, &root->root_item);
4514 if (ret < 0) {
4515 btrfs_end_transaction(trans);
4516 goto out;
4517 }
4518 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4519 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4520 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4521 root->root_key.objectid);
4522 if (ret < 0 && ret != -EEXIST) {
4523 btrfs_abort_transaction(trans, ret);
4524 btrfs_end_transaction(trans);
4525 goto out;
4526 }
4527 }
4528 ret = btrfs_commit_transaction(trans);
4529out:
4530 up_write(&fs_info->subvol_sem);
4531 mnt_drop_write_file(file);
4532 return ret;
4533}
4534
4535#ifdef CONFIG_64BIT
4536static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4537 void __user *arg)
4538{
4539 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4540 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4541 int ret = 0;
4542
4543 args32 = memdup_user(arg, sizeof(*args32));
4544 if (IS_ERR(args32))
4545 return PTR_ERR(args32);
4546
4547 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4548 if (!args64) {
4549 ret = -ENOMEM;
4550 goto out;
4551 }
4552
4553 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4554 args64->stransid = args32->stransid;
4555 args64->rtransid = args32->rtransid;
4556 args64->stime.sec = args32->stime.sec;
4557 args64->stime.nsec = args32->stime.nsec;
4558 args64->rtime.sec = args32->rtime.sec;
4559 args64->rtime.nsec = args32->rtime.nsec;
4560 args64->flags = args32->flags;
4561
4562 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4563 if (ret)
4564 goto out;
4565
4566 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4567 args32->stransid = args64->stransid;
4568 args32->rtransid = args64->rtransid;
4569 args32->stime.sec = args64->stime.sec;
4570 args32->stime.nsec = args64->stime.nsec;
4571 args32->rtime.sec = args64->rtime.sec;
4572 args32->rtime.nsec = args64->rtime.nsec;
4573 args32->flags = args64->flags;
4574
4575 ret = copy_to_user(arg, args32, sizeof(*args32));
4576 if (ret)
4577 ret = -EFAULT;
4578
4579out:
4580 kfree(args32);
4581 kfree(args64);
4582 return ret;
4583}
4584#endif
4585
4586static long btrfs_ioctl_set_received_subvol(struct file *file,
4587 void __user *arg)
4588{
4589 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4590 int ret = 0;
4591
4592 sa = memdup_user(arg, sizeof(*sa));
4593 if (IS_ERR(sa))
4594 return PTR_ERR(sa);
4595
4596 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4597
4598 if (ret)
4599 goto out;
4600
4601 ret = copy_to_user(arg, sa, sizeof(*sa));
4602 if (ret)
4603 ret = -EFAULT;
4604
4605out:
4606 kfree(sa);
4607 return ret;
4608}
4609
4610static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4611 void __user *arg)
4612{
4613 size_t len;
4614 int ret;
4615 char label[BTRFS_LABEL_SIZE];
4616
4617 spin_lock(&fs_info->super_lock);
4618 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4619 spin_unlock(&fs_info->super_lock);
4620
4621 len = strnlen(label, BTRFS_LABEL_SIZE);
4622
4623 if (len == BTRFS_LABEL_SIZE) {
4624 btrfs_warn(fs_info,
4625 "label is too long, return the first %zu bytes",
4626 --len);
4627 }
4628
4629 ret = copy_to_user(arg, label, len);
4630
4631 return ret ? -EFAULT : 0;
4632}
4633
4634static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4635{
4636 struct inode *inode = file_inode(file);
4637 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4638 struct btrfs_root *root = BTRFS_I(inode)->root;
4639 struct btrfs_super_block *super_block = fs_info->super_copy;
4640 struct btrfs_trans_handle *trans;
4641 char label[BTRFS_LABEL_SIZE];
4642 int ret;
4643
4644 if (!capable(CAP_SYS_ADMIN))
4645 return -EPERM;
4646
4647 if (copy_from_user(label, arg, sizeof(label)))
4648 return -EFAULT;
4649
4650 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4651 btrfs_err(fs_info,
4652 "unable to set label with more than %d bytes",
4653 BTRFS_LABEL_SIZE - 1);
4654 return -EINVAL;
4655 }
4656
4657 ret = mnt_want_write_file(file);
4658 if (ret)
4659 return ret;
4660
4661 trans = btrfs_start_transaction(root, 0);
4662 if (IS_ERR(trans)) {
4663 ret = PTR_ERR(trans);
4664 goto out_unlock;
4665 }
4666
4667 spin_lock(&fs_info->super_lock);
4668 strcpy(super_block->label, label);
4669 spin_unlock(&fs_info->super_lock);
4670 ret = btrfs_commit_transaction(trans);
4671
4672out_unlock:
4673 mnt_drop_write_file(file);
4674 return ret;
4675}
4676
4677#define INIT_FEATURE_FLAGS(suffix) \
4678 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4679 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4680 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4681
4682int btrfs_ioctl_get_supported_features(void __user *arg)
4683{
4684 static const struct btrfs_ioctl_feature_flags features[3] = {
4685 INIT_FEATURE_FLAGS(SUPP),
4686 INIT_FEATURE_FLAGS(SAFE_SET),
4687 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4688 };
4689
4690 if (copy_to_user(arg, &features, sizeof(features)))
4691 return -EFAULT;
4692
4693 return 0;
4694}
4695
4696static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4697 void __user *arg)
4698{
4699 struct btrfs_super_block *super_block = fs_info->super_copy;
4700 struct btrfs_ioctl_feature_flags features;
4701
4702 features.compat_flags = btrfs_super_compat_flags(super_block);
4703 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4704 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4705
4706 if (copy_to_user(arg, &features, sizeof(features)))
4707 return -EFAULT;
4708
4709 return 0;
4710}
4711
4712static int check_feature_bits(struct btrfs_fs_info *fs_info,
4713 enum btrfs_feature_set set,
4714 u64 change_mask, u64 flags, u64 supported_flags,
4715 u64 safe_set, u64 safe_clear)
4716{
4717 const char *type = btrfs_feature_set_name(set);
4718 char *names;
4719 u64 disallowed, unsupported;
4720 u64 set_mask = flags & change_mask;
4721 u64 clear_mask = ~flags & change_mask;
4722
4723 unsupported = set_mask & ~supported_flags;
4724 if (unsupported) {
4725 names = btrfs_printable_features(set, unsupported);
4726 if (names) {
4727 btrfs_warn(fs_info,
4728 "this kernel does not support the %s feature bit%s",
4729 names, strchr(names, ',') ? "s" : "");
4730 kfree(names);
4731 } else
4732 btrfs_warn(fs_info,
4733 "this kernel does not support %s bits 0x%llx",
4734 type, unsupported);
4735 return -EOPNOTSUPP;
4736 }
4737
4738 disallowed = set_mask & ~safe_set;
4739 if (disallowed) {
4740 names = btrfs_printable_features(set, disallowed);
4741 if (names) {
4742 btrfs_warn(fs_info,
4743 "can't set the %s feature bit%s while mounted",
4744 names, strchr(names, ',') ? "s" : "");
4745 kfree(names);
4746 } else
4747 btrfs_warn(fs_info,
4748 "can't set %s bits 0x%llx while mounted",
4749 type, disallowed);
4750 return -EPERM;
4751 }
4752
4753 disallowed = clear_mask & ~safe_clear;
4754 if (disallowed) {
4755 names = btrfs_printable_features(set, disallowed);
4756 if (names) {
4757 btrfs_warn(fs_info,
4758 "can't clear the %s feature bit%s while mounted",
4759 names, strchr(names, ',') ? "s" : "");
4760 kfree(names);
4761 } else
4762 btrfs_warn(fs_info,
4763 "can't clear %s bits 0x%llx while mounted",
4764 type, disallowed);
4765 return -EPERM;
4766 }
4767
4768 return 0;
4769}
4770
4771#define check_feature(fs_info, change_mask, flags, mask_base) \
4772check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
4773 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4774 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4775 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4776
4777static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4778{
4779 struct inode *inode = file_inode(file);
4780 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4781 struct btrfs_root *root = BTRFS_I(inode)->root;
4782 struct btrfs_super_block *super_block = fs_info->super_copy;
4783 struct btrfs_ioctl_feature_flags flags[2];
4784 struct btrfs_trans_handle *trans;
4785 u64 newflags;
4786 int ret;
4787
4788 if (!capable(CAP_SYS_ADMIN))
4789 return -EPERM;
4790
4791 if (copy_from_user(flags, arg, sizeof(flags)))
4792 return -EFAULT;
4793
4794 /* Nothing to do */
4795 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4796 !flags[0].incompat_flags)
4797 return 0;
4798
4799 ret = check_feature(fs_info, flags[0].compat_flags,
4800 flags[1].compat_flags, COMPAT);
4801 if (ret)
4802 return ret;
4803
4804 ret = check_feature(fs_info, flags[0].compat_ro_flags,
4805 flags[1].compat_ro_flags, COMPAT_RO);
4806 if (ret)
4807 return ret;
4808
4809 ret = check_feature(fs_info, flags[0].incompat_flags,
4810 flags[1].incompat_flags, INCOMPAT);
4811 if (ret)
4812 return ret;
4813
4814 ret = mnt_want_write_file(file);
4815 if (ret)
4816 return ret;
4817
4818 trans = btrfs_start_transaction(root, 0);
4819 if (IS_ERR(trans)) {
4820 ret = PTR_ERR(trans);
4821 goto out_drop_write;
4822 }
4823
4824 spin_lock(&fs_info->super_lock);
4825 newflags = btrfs_super_compat_flags(super_block);
4826 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4827 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4828 btrfs_set_super_compat_flags(super_block, newflags);
4829
4830 newflags = btrfs_super_compat_ro_flags(super_block);
4831 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4832 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4833 btrfs_set_super_compat_ro_flags(super_block, newflags);
4834
4835 newflags = btrfs_super_incompat_flags(super_block);
4836 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4837 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4838 btrfs_set_super_incompat_flags(super_block, newflags);
4839 spin_unlock(&fs_info->super_lock);
4840
4841 ret = btrfs_commit_transaction(trans);
4842out_drop_write:
4843 mnt_drop_write_file(file);
4844
4845 return ret;
4846}
4847
4848static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
4849{
4850 struct btrfs_ioctl_send_args *arg;
4851 int ret;
4852
4853 if (compat) {
4854#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4855 struct btrfs_ioctl_send_args_32 args32;
4856
4857 ret = copy_from_user(&args32, argp, sizeof(args32));
4858 if (ret)
4859 return -EFAULT;
4860 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4861 if (!arg)
4862 return -ENOMEM;
4863 arg->send_fd = args32.send_fd;
4864 arg->clone_sources_count = args32.clone_sources_count;
4865 arg->clone_sources = compat_ptr(args32.clone_sources);
4866 arg->parent_root = args32.parent_root;
4867 arg->flags = args32.flags;
4868 memcpy(arg->reserved, args32.reserved,
4869 sizeof(args32.reserved));
4870#else
4871 return -ENOTTY;
4872#endif
4873 } else {
4874 arg = memdup_user(argp, sizeof(*arg));
4875 if (IS_ERR(arg))
4876 return PTR_ERR(arg);
4877 }
4878 ret = btrfs_ioctl_send(file, arg);
4879 kfree(arg);
4880 return ret;
4881}
4882
4883long btrfs_ioctl(struct file *file, unsigned int
4884 cmd, unsigned long arg)
4885{
4886 struct inode *inode = file_inode(file);
4887 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4888 struct btrfs_root *root = BTRFS_I(inode)->root;
4889 void __user *argp = (void __user *)arg;
4890
4891 switch (cmd) {
4892 case FS_IOC_GETVERSION:
4893 return btrfs_ioctl_getversion(file, argp);
4894 case FS_IOC_GETFSLABEL:
4895 return btrfs_ioctl_get_fslabel(fs_info, argp);
4896 case FS_IOC_SETFSLABEL:
4897 return btrfs_ioctl_set_fslabel(file, argp);
4898 case FITRIM:
4899 return btrfs_ioctl_fitrim(fs_info, argp);
4900 case BTRFS_IOC_SNAP_CREATE:
4901 return btrfs_ioctl_snap_create(file, argp, 0);
4902 case BTRFS_IOC_SNAP_CREATE_V2:
4903 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4904 case BTRFS_IOC_SUBVOL_CREATE:
4905 return btrfs_ioctl_snap_create(file, argp, 1);
4906 case BTRFS_IOC_SUBVOL_CREATE_V2:
4907 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4908 case BTRFS_IOC_SNAP_DESTROY:
4909 return btrfs_ioctl_snap_destroy(file, argp, false);
4910 case BTRFS_IOC_SNAP_DESTROY_V2:
4911 return btrfs_ioctl_snap_destroy(file, argp, true);
4912 case BTRFS_IOC_SUBVOL_GETFLAGS:
4913 return btrfs_ioctl_subvol_getflags(file, argp);
4914 case BTRFS_IOC_SUBVOL_SETFLAGS:
4915 return btrfs_ioctl_subvol_setflags(file, argp);
4916 case BTRFS_IOC_DEFAULT_SUBVOL:
4917 return btrfs_ioctl_default_subvol(file, argp);
4918 case BTRFS_IOC_DEFRAG:
4919 return btrfs_ioctl_defrag(file, NULL);
4920 case BTRFS_IOC_DEFRAG_RANGE:
4921 return btrfs_ioctl_defrag(file, argp);
4922 case BTRFS_IOC_RESIZE:
4923 return btrfs_ioctl_resize(file, argp);
4924 case BTRFS_IOC_ADD_DEV:
4925 return btrfs_ioctl_add_dev(fs_info, argp);
4926 case BTRFS_IOC_RM_DEV:
4927 return btrfs_ioctl_rm_dev(file, argp);
4928 case BTRFS_IOC_RM_DEV_V2:
4929 return btrfs_ioctl_rm_dev_v2(file, argp);
4930 case BTRFS_IOC_FS_INFO:
4931 return btrfs_ioctl_fs_info(fs_info, argp);
4932 case BTRFS_IOC_DEV_INFO:
4933 return btrfs_ioctl_dev_info(fs_info, argp);
4934 case BTRFS_IOC_BALANCE:
4935 return btrfs_ioctl_balance(file, NULL);
4936 case BTRFS_IOC_TREE_SEARCH:
4937 return btrfs_ioctl_tree_search(file, argp);
4938 case BTRFS_IOC_TREE_SEARCH_V2:
4939 return btrfs_ioctl_tree_search_v2(file, argp);
4940 case BTRFS_IOC_INO_LOOKUP:
4941 return btrfs_ioctl_ino_lookup(file, argp);
4942 case BTRFS_IOC_INO_PATHS:
4943 return btrfs_ioctl_ino_to_path(root, argp);
4944 case BTRFS_IOC_LOGICAL_INO:
4945 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4946 case BTRFS_IOC_LOGICAL_INO_V2:
4947 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4948 case BTRFS_IOC_SPACE_INFO:
4949 return btrfs_ioctl_space_info(fs_info, argp);
4950 case BTRFS_IOC_SYNC: {
4951 int ret;
4952
4953 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4954 if (ret)
4955 return ret;
4956 ret = btrfs_sync_fs(inode->i_sb, 1);
4957 /*
4958 * The transaction thread may want to do more work,
4959 * namely it pokes the cleaner kthread that will start
4960 * processing uncleaned subvols.
4961 */
4962 wake_up_process(fs_info->transaction_kthread);
4963 return ret;
4964 }
4965 case BTRFS_IOC_START_SYNC:
4966 return btrfs_ioctl_start_sync(root, argp);
4967 case BTRFS_IOC_WAIT_SYNC:
4968 return btrfs_ioctl_wait_sync(fs_info, argp);
4969 case BTRFS_IOC_SCRUB:
4970 return btrfs_ioctl_scrub(file, argp);
4971 case BTRFS_IOC_SCRUB_CANCEL:
4972 return btrfs_ioctl_scrub_cancel(fs_info);
4973 case BTRFS_IOC_SCRUB_PROGRESS:
4974 return btrfs_ioctl_scrub_progress(fs_info, argp);
4975 case BTRFS_IOC_BALANCE_V2:
4976 return btrfs_ioctl_balance(file, argp);
4977 case BTRFS_IOC_BALANCE_CTL:
4978 return btrfs_ioctl_balance_ctl(fs_info, arg);
4979 case BTRFS_IOC_BALANCE_PROGRESS:
4980 return btrfs_ioctl_balance_progress(fs_info, argp);
4981 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4982 return btrfs_ioctl_set_received_subvol(file, argp);
4983#ifdef CONFIG_64BIT
4984 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4985 return btrfs_ioctl_set_received_subvol_32(file, argp);
4986#endif
4987 case BTRFS_IOC_SEND:
4988 return _btrfs_ioctl_send(file, argp, false);
4989#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4990 case BTRFS_IOC_SEND_32:
4991 return _btrfs_ioctl_send(file, argp, true);
4992#endif
4993 case BTRFS_IOC_GET_DEV_STATS:
4994 return btrfs_ioctl_get_dev_stats(fs_info, argp);
4995 case BTRFS_IOC_QUOTA_CTL:
4996 return btrfs_ioctl_quota_ctl(file, argp);
4997 case BTRFS_IOC_QGROUP_ASSIGN:
4998 return btrfs_ioctl_qgroup_assign(file, argp);
4999 case BTRFS_IOC_QGROUP_CREATE:
5000 return btrfs_ioctl_qgroup_create(file, argp);
5001 case BTRFS_IOC_QGROUP_LIMIT:
5002 return btrfs_ioctl_qgroup_limit(file, argp);
5003 case BTRFS_IOC_QUOTA_RESCAN:
5004 return btrfs_ioctl_quota_rescan(file, argp);
5005 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5006 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
5007 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5008 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
5009 case BTRFS_IOC_DEV_REPLACE:
5010 return btrfs_ioctl_dev_replace(fs_info, argp);
5011 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5012 return btrfs_ioctl_get_supported_features(argp);
5013 case BTRFS_IOC_GET_FEATURES:
5014 return btrfs_ioctl_get_features(fs_info, argp);
5015 case BTRFS_IOC_SET_FEATURES:
5016 return btrfs_ioctl_set_features(file, argp);
5017 case BTRFS_IOC_GET_SUBVOL_INFO:
5018 return btrfs_ioctl_get_subvol_info(file, argp);
5019 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5020 return btrfs_ioctl_get_subvol_rootref(file, argp);
5021 case BTRFS_IOC_INO_LOOKUP_USER:
5022 return btrfs_ioctl_ino_lookup_user(file, argp);
5023 }
5024
5025 return -ENOTTY;
5026}
5027
5028#ifdef CONFIG_COMPAT
5029long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5030{
5031 /*
5032 * These all access 32-bit values anyway so no further
5033 * handling is necessary.
5034 */
5035 switch (cmd) {
5036 case FS_IOC32_GETVERSION:
5037 cmd = FS_IOC_GETVERSION;
5038 break;
5039 }
5040
5041 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5042}
5043#endif