?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'nfs-for-3.9-4' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
[linux/fpc-iii.git] / fs / btrfs / ioctl.c
blob2c02310ff2d96320ebf2d348ed339dd0ea047e0f
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 <linux/uuid.h>
45 #include <linux/btrfs.h>
46 #include "compat.h"
47 #include "ctree.h"
48 #include "disk-io.h"
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
52 #include "volumes.h"
53 #include "locking.h"
54 #include "inode-map.h"
55 #include "backref.h"
56 #include "rcu-string.h"
57 #include "send.h"
58 #include "dev-replace.h"
59
60 /* Mask out flags that are inappropriate for the given type of inode. */
61 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
62 {
63 if (S_ISDIR(mode))
64 return flags;
65 else if (S_ISREG(mode))
66 return flags & ~FS_DIRSYNC_FL;
67 else
68 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
69 }
70
71 /*
72 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
73 */
74 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
75 {
76 unsigned int iflags = 0;
77
78 if (flags & BTRFS_INODE_SYNC)
79 iflags |= FS_SYNC_FL;
80 if (flags & BTRFS_INODE_IMMUTABLE)
81 iflags |= FS_IMMUTABLE_FL;
82 if (flags & BTRFS_INODE_APPEND)
83 iflags |= FS_APPEND_FL;
84 if (flags & BTRFS_INODE_NODUMP)
85 iflags |= FS_NODUMP_FL;
86 if (flags & BTRFS_INODE_NOATIME)
87 iflags |= FS_NOATIME_FL;
88 if (flags & BTRFS_INODE_DIRSYNC)
89 iflags |= FS_DIRSYNC_FL;
90 if (flags & BTRFS_INODE_NODATACOW)
91 iflags |= FS_NOCOW_FL;
92
93 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
94 iflags |= FS_COMPR_FL;
95 else if (flags & BTRFS_INODE_NOCOMPRESS)
96 iflags |= FS_NOCOMP_FL;
97
98 return iflags;
99 }
100
101 /*
102 * Update inode->i_flags based on the btrfs internal flags.
103 */
104 void btrfs_update_iflags(struct inode *inode)
105 {
106 struct btrfs_inode *ip = BTRFS_I(inode);
107
108 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
109
110 if (ip->flags & BTRFS_INODE_SYNC)
111 inode->i_flags |= S_SYNC;
112 if (ip->flags & BTRFS_INODE_IMMUTABLE)
113 inode->i_flags |= S_IMMUTABLE;
114 if (ip->flags & BTRFS_INODE_APPEND)
115 inode->i_flags |= S_APPEND;
116 if (ip->flags & BTRFS_INODE_NOATIME)
117 inode->i_flags |= S_NOATIME;
118 if (ip->flags & BTRFS_INODE_DIRSYNC)
119 inode->i_flags |= S_DIRSYNC;
120 }
121
122 /*
123 * Inherit flags from the parent inode.
124 *
125 * Currently only the compression flags and the cow flags are inherited.
126 */
127 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
128 {
129 unsigned int flags;
130
131 if (!dir)
132 return;
133
134 flags = BTRFS_I(dir)->flags;
135
136 if (flags & BTRFS_INODE_NOCOMPRESS) {
137 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
138 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
139 } else if (flags & BTRFS_INODE_COMPRESS) {
140 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
141 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
142 }
143
144 if (flags & BTRFS_INODE_NODATACOW) {
145 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
146 if (S_ISREG(inode->i_mode))
147 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
148 }
149
150 btrfs_update_iflags(inode);
151 }
152
153 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
154 {
155 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
156 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
157
158 if (copy_to_user(arg, &flags, sizeof(flags)))
159 return -EFAULT;
160 return 0;
161 }
162
163 static int check_flags(unsigned int flags)
164 {
165 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
166 FS_NOATIME_FL | FS_NODUMP_FL | \
167 FS_SYNC_FL | FS_DIRSYNC_FL | \
168 FS_NOCOMP_FL | FS_COMPR_FL |
169 FS_NOCOW_FL))
170 return -EOPNOTSUPP;
171
172 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
173 return -EINVAL;
174
175 return 0;
176 }
177
178 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
179 {
180 struct inode *inode = file_inode(file);
181 struct btrfs_inode *ip = BTRFS_I(inode);
182 struct btrfs_root *root = ip->root;
183 struct btrfs_trans_handle *trans;
184 unsigned int flags, oldflags;
185 int ret;
186 u64 ip_oldflags;
187 unsigned int i_oldflags;
188 umode_t mode;
189
190 if (btrfs_root_readonly(root))
191 return -EROFS;
192
193 if (copy_from_user(&flags, arg, sizeof(flags)))
194 return -EFAULT;
195
196 ret = check_flags(flags);
197 if (ret)
198 return ret;
199
200 if (!inode_owner_or_capable(inode))
201 return -EACCES;
202
203 ret = mnt_want_write_file(file);
204 if (ret)
205 return ret;
206
207 mutex_lock(&inode->i_mutex);
208
209 ip_oldflags = ip->flags;
210 i_oldflags = inode->i_flags;
211 mode = inode->i_mode;
212
213 flags = btrfs_mask_flags(inode->i_mode, flags);
214 oldflags = btrfs_flags_to_ioctl(ip->flags);
215 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
216 if (!capable(CAP_LINUX_IMMUTABLE)) {
217 ret = -EPERM;
218 goto out_unlock;
219 }
220 }
221
222 if (flags & FS_SYNC_FL)
223 ip->flags |= BTRFS_INODE_SYNC;
224 else
225 ip->flags &= ~BTRFS_INODE_SYNC;
226 if (flags & FS_IMMUTABLE_FL)
227 ip->flags |= BTRFS_INODE_IMMUTABLE;
228 else
229 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
230 if (flags & FS_APPEND_FL)
231 ip->flags |= BTRFS_INODE_APPEND;
232 else
233 ip->flags &= ~BTRFS_INODE_APPEND;
234 if (flags & FS_NODUMP_FL)
235 ip->flags |= BTRFS_INODE_NODUMP;
236 else
237 ip->flags &= ~BTRFS_INODE_NODUMP;
238 if (flags & FS_NOATIME_FL)
239 ip->flags |= BTRFS_INODE_NOATIME;
240 else
241 ip->flags &= ~BTRFS_INODE_NOATIME;
242 if (flags & FS_DIRSYNC_FL)
243 ip->flags |= BTRFS_INODE_DIRSYNC;
244 else
245 ip->flags &= ~BTRFS_INODE_DIRSYNC;
246 if (flags & FS_NOCOW_FL) {
247 if (S_ISREG(mode)) {
248 /*
249 * It's safe to turn csums off here, no extents exist.
250 * Otherwise we want the flag to reflect the real COW
251 * status of the file and will not set it.
252 */
253 if (inode->i_size == 0)
254 ip->flags |= BTRFS_INODE_NODATACOW
255 | BTRFS_INODE_NODATASUM;
256 } else {
257 ip->flags |= BTRFS_INODE_NODATACOW;
258 }
259 } else {
260 /*
261 * Revert back under same assuptions as above
262 */
263 if (S_ISREG(mode)) {
264 if (inode->i_size == 0)
265 ip->flags &= ~(BTRFS_INODE_NODATACOW
266 | BTRFS_INODE_NODATASUM);
267 } else {
268 ip->flags &= ~BTRFS_INODE_NODATACOW;
269 }
270 }
271
272 /*
273 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
274 * flag may be changed automatically if compression code won't make
275 * things smaller.
276 */
277 if (flags & FS_NOCOMP_FL) {
278 ip->flags &= ~BTRFS_INODE_COMPRESS;
279 ip->flags |= BTRFS_INODE_NOCOMPRESS;
280 } else if (flags & FS_COMPR_FL) {
281 ip->flags |= BTRFS_INODE_COMPRESS;
282 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
283 } else {
284 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
285 }
286
287 trans = btrfs_start_transaction(root, 1);
288 if (IS_ERR(trans)) {
289 ret = PTR_ERR(trans);
290 goto out_drop;
291 }
292
293 btrfs_update_iflags(inode);
294 inode_inc_iversion(inode);
295 inode->i_ctime = CURRENT_TIME;
296 ret = btrfs_update_inode(trans, root, inode);
297
298 btrfs_end_transaction(trans, root);
299 out_drop:
300 if (ret) {
301 ip->flags = ip_oldflags;
302 inode->i_flags = i_oldflags;
303 }
304
305 out_unlock:
306 mutex_unlock(&inode->i_mutex);
307 mnt_drop_write_file(file);
308 return ret;
309 }
310
311 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
312 {
313 struct inode *inode = file_inode(file);
314
315 return put_user(inode->i_generation, arg);
316 }
317
318 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
319 {
320 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
321 struct btrfs_device *device;
322 struct request_queue *q;
323 struct fstrim_range range;
324 u64 minlen = ULLONG_MAX;
325 u64 num_devices = 0;
326 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
327 int ret;
328
329 if (!capable(CAP_SYS_ADMIN))
330 return -EPERM;
331
332 rcu_read_lock();
333 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
334 dev_list) {
335 if (!device->bdev)
336 continue;
337 q = bdev_get_queue(device->bdev);
338 if (blk_queue_discard(q)) {
339 num_devices++;
340 minlen = min((u64)q->limits.discard_granularity,
341 minlen);
342 }
343 }
344 rcu_read_unlock();
345
346 if (!num_devices)
347 return -EOPNOTSUPP;
348 if (copy_from_user(&range, arg, sizeof(range)))
349 return -EFAULT;
350 if (range.start > total_bytes ||
351 range.len < fs_info->sb->s_blocksize)
352 return -EINVAL;
353
354 range.len = min(range.len, total_bytes - range.start);
355 range.minlen = max(range.minlen, minlen);
356 ret = btrfs_trim_fs(fs_info->tree_root, &range);
357 if (ret < 0)
358 return ret;
359
360 if (copy_to_user(arg, &range, sizeof(range)))
361 return -EFAULT;
362
363 return 0;
364 }
365
366 static noinline int create_subvol(struct inode *dir,
367 struct dentry *dentry,
368 char *name, int namelen,
369 u64 *async_transid,
370 struct btrfs_qgroup_inherit *inherit)
371 {
372 struct btrfs_trans_handle *trans;
373 struct btrfs_key key;
374 struct btrfs_root_item root_item;
375 struct btrfs_inode_item *inode_item;
376 struct extent_buffer *leaf;
377 struct btrfs_root *root = BTRFS_I(dir)->root;
378 struct btrfs_root *new_root;
379 struct btrfs_block_rsv block_rsv;
380 struct timespec cur_time = CURRENT_TIME;
381 int ret;
382 int err;
383 u64 objectid;
384 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
385 u64 index = 0;
386 u64 qgroup_reserved;
387 uuid_le new_uuid;
388
389 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
390 if (ret)
391 return ret;
392
393 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
394 /*
395 * The same as the snapshot creation, please see the comment
396 * of create_snapshot().
397 */
398 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
399 7, &qgroup_reserved);
400 if (ret)
401 return ret;
402
403 trans = btrfs_start_transaction(root, 0);
404 if (IS_ERR(trans)) {
405 ret = PTR_ERR(trans);
406 goto out;
407 }
408 trans->block_rsv = &block_rsv;
409 trans->bytes_reserved = block_rsv.size;
410
411 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
412 if (ret)
413 goto fail;
414
415 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
416 0, objectid, NULL, 0, 0, 0);
417 if (IS_ERR(leaf)) {
418 ret = PTR_ERR(leaf);
419 goto fail;
420 }
421
422 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
423 btrfs_set_header_bytenr(leaf, leaf->start);
424 btrfs_set_header_generation(leaf, trans->transid);
425 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
426 btrfs_set_header_owner(leaf, objectid);
427
428 write_extent_buffer(leaf, root->fs_info->fsid,
429 (unsigned long)btrfs_header_fsid(leaf),
430 BTRFS_FSID_SIZE);
431 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
432 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
433 BTRFS_UUID_SIZE);
434 btrfs_mark_buffer_dirty(leaf);
435
436 memset(&root_item, 0, sizeof(root_item));
437
438 inode_item = &root_item.inode;
439 inode_item->generation = cpu_to_le64(1);
440 inode_item->size = cpu_to_le64(3);
441 inode_item->nlink = cpu_to_le32(1);
442 inode_item->nbytes = cpu_to_le64(root->leafsize);
443 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
444
445 root_item.flags = 0;
446 root_item.byte_limit = 0;
447 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
448
449 btrfs_set_root_bytenr(&root_item, leaf->start);
450 btrfs_set_root_generation(&root_item, trans->transid);
451 btrfs_set_root_level(&root_item, 0);
452 btrfs_set_root_refs(&root_item, 1);
453 btrfs_set_root_used(&root_item, leaf->len);
454 btrfs_set_root_last_snapshot(&root_item, 0);
455
456 btrfs_set_root_generation_v2(&root_item,
457 btrfs_root_generation(&root_item));
458 uuid_le_gen(&new_uuid);
459 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
460 root_item.otime.sec = cpu_to_le64(cur_time.tv_sec);
461 root_item.otime.nsec = cpu_to_le32(cur_time.tv_nsec);
462 root_item.ctime = root_item.otime;
463 btrfs_set_root_ctransid(&root_item, trans->transid);
464 btrfs_set_root_otransid(&root_item, trans->transid);
465
466 btrfs_tree_unlock(leaf);
467 free_extent_buffer(leaf);
468 leaf = NULL;
469
470 btrfs_set_root_dirid(&root_item, new_dirid);
471
472 key.objectid = objectid;
473 key.offset = 0;
474 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
475 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
476 &root_item);
477 if (ret)
478 goto fail;
479
480 key.offset = (u64)-1;
481 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
482 if (IS_ERR(new_root)) {
483 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
484 ret = PTR_ERR(new_root);
485 goto fail;
486 }
487
488 btrfs_record_root_in_trans(trans, new_root);
489
490 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
491 if (ret) {
492 /* We potentially lose an unused inode item here */
493 btrfs_abort_transaction(trans, root, ret);
494 goto fail;
495 }
496
497 /*
498 * insert the directory item
499 */
500 ret = btrfs_set_inode_index(dir, &index);
501 if (ret) {
502 btrfs_abort_transaction(trans, root, ret);
503 goto fail;
504 }
505
506 ret = btrfs_insert_dir_item(trans, root,
507 name, namelen, dir, &key,
508 BTRFS_FT_DIR, index);
509 if (ret) {
510 btrfs_abort_transaction(trans, root, ret);
511 goto fail;
512 }
513
514 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
515 ret = btrfs_update_inode(trans, root, dir);
516 BUG_ON(ret);
517
518 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
519 objectid, root->root_key.objectid,
520 btrfs_ino(dir), index, name, namelen);
521
522 BUG_ON(ret);
523
524 fail:
525 trans->block_rsv = NULL;
526 trans->bytes_reserved = 0;
527 if (async_transid) {
528 *async_transid = trans->transid;
529 err = btrfs_commit_transaction_async(trans, root, 1);
530 if (err)
531 err = btrfs_commit_transaction(trans, root);
532 } else {
533 err = btrfs_commit_transaction(trans, root);
534 }
535 if (err && !ret)
536 ret = err;
537
538 if (!ret)
539 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
540 out:
541 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
542 return ret;
543 }
544
545 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
546 struct dentry *dentry, char *name, int namelen,
547 u64 *async_transid, bool readonly,
548 struct btrfs_qgroup_inherit *inherit)
549 {
550 struct inode *inode;
551 struct btrfs_pending_snapshot *pending_snapshot;
552 struct btrfs_trans_handle *trans;
553 int ret;
554
555 if (!root->ref_cows)
556 return -EINVAL;
557
558 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
559 if (!pending_snapshot)
560 return -ENOMEM;
561
562 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
563 BTRFS_BLOCK_RSV_TEMP);
564 /*
565 * 1 - parent dir inode
566 * 2 - dir entries
567 * 1 - root item
568 * 2 - root ref/backref
569 * 1 - root of snapshot
570 */
571 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
572 &pending_snapshot->block_rsv, 7,
573 &pending_snapshot->qgroup_reserved);
574 if (ret)
575 goto out;
576
577 pending_snapshot->dentry = dentry;
578 pending_snapshot->root = root;
579 pending_snapshot->readonly = readonly;
580 pending_snapshot->dir = dir;
581 pending_snapshot->inherit = inherit;
582
583 trans = btrfs_start_transaction(root, 0);
584 if (IS_ERR(trans)) {
585 ret = PTR_ERR(trans);
586 goto fail;
587 }
588
589 spin_lock(&root->fs_info->trans_lock);
590 list_add(&pending_snapshot->list,
591 &trans->transaction->pending_snapshots);
592 spin_unlock(&root->fs_info->trans_lock);
593 if (async_transid) {
594 *async_transid = trans->transid;
595 ret = btrfs_commit_transaction_async(trans,
596 root->fs_info->extent_root, 1);
597 if (ret)
598 ret = btrfs_commit_transaction(trans, root);
599 } else {
600 ret = btrfs_commit_transaction(trans,
601 root->fs_info->extent_root);
602 }
603 if (ret)
604 goto fail;
605
606 ret = pending_snapshot->error;
607 if (ret)
608 goto fail;
609
610 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
611 if (ret)
612 goto fail;
613
614 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
615 if (IS_ERR(inode)) {
616 ret = PTR_ERR(inode);
617 goto fail;
618 }
619 BUG_ON(!inode);
620 d_instantiate(dentry, inode);
621 ret = 0;
622 fail:
623 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
624 &pending_snapshot->block_rsv,
625 pending_snapshot->qgroup_reserved);
626 out:
627 kfree(pending_snapshot);
628 return ret;
629 }
630
631 /* copy of check_sticky in fs/namei.c()
632 * It's inline, so penalty for filesystems that don't use sticky bit is
633 * minimal.
634 */
635 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
636 {
637 kuid_t fsuid = current_fsuid();
638
639 if (!(dir->i_mode & S_ISVTX))
640 return 0;
641 if (uid_eq(inode->i_uid, fsuid))
642 return 0;
643 if (uid_eq(dir->i_uid, fsuid))
644 return 0;
645 return !capable(CAP_FOWNER);
646 }
647
648 /* copy of may_delete in fs/namei.c()
649 * Check whether we can remove a link victim from directory dir, check
650 * whether the type of victim is right.
651 * 1. We can't do it if dir is read-only (done in permission())
652 * 2. We should have write and exec permissions on dir
653 * 3. We can't remove anything from append-only dir
654 * 4. We can't do anything with immutable dir (done in permission())
655 * 5. If the sticky bit on dir is set we should either
656 * a. be owner of dir, or
657 * b. be owner of victim, or
658 * c. have CAP_FOWNER capability
659 * 6. If the victim is append-only or immutable we can't do antyhing with
660 * links pointing to it.
661 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
662 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
663 * 9. We can't remove a root or mountpoint.
664 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
665 * nfs_async_unlink().
666 */
667
668 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
669 {
670 int error;
671
672 if (!victim->d_inode)
673 return -ENOENT;
674
675 BUG_ON(victim->d_parent->d_inode != dir);
676 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
677
678 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
679 if (error)
680 return error;
681 if (IS_APPEND(dir))
682 return -EPERM;
683 if (btrfs_check_sticky(dir, victim->d_inode)||
684 IS_APPEND(victim->d_inode)||
685 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
686 return -EPERM;
687 if (isdir) {
688 if (!S_ISDIR(victim->d_inode->i_mode))
689 return -ENOTDIR;
690 if (IS_ROOT(victim))
691 return -EBUSY;
692 } else if (S_ISDIR(victim->d_inode->i_mode))
693 return -EISDIR;
694 if (IS_DEADDIR(dir))
695 return -ENOENT;
696 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
697 return -EBUSY;
698 return 0;
699 }
700
701 /* copy of may_create in fs/namei.c() */
702 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
703 {
704 if (child->d_inode)
705 return -EEXIST;
706 if (IS_DEADDIR(dir))
707 return -ENOENT;
708 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
709 }
710
711 /*
712 * Create a new subvolume below @parent. This is largely modeled after
713 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
714 * inside this filesystem so it's quite a bit simpler.
715 */
716 static noinline int btrfs_mksubvol(struct path *parent,
717 char *name, int namelen,
718 struct btrfs_root *snap_src,
719 u64 *async_transid, bool readonly,
720 struct btrfs_qgroup_inherit *inherit)
721 {
722 struct inode *dir = parent->dentry->d_inode;
723 struct dentry *dentry;
724 int error;
725
726 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
727
728 dentry = lookup_one_len(name, parent->dentry, namelen);
729 error = PTR_ERR(dentry);
730 if (IS_ERR(dentry))
731 goto out_unlock;
732
733 error = -EEXIST;
734 if (dentry->d_inode)
735 goto out_dput;
736
737 error = btrfs_may_create(dir, dentry);
738 if (error)
739 goto out_dput;
740
741 /*
742 * even if this name doesn't exist, we may get hash collisions.
743 * check for them now when we can safely fail
744 */
745 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
746 dir->i_ino, name,
747 namelen);
748 if (error)
749 goto out_dput;
750
751 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
752
753 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
754 goto out_up_read;
755
756 if (snap_src) {
757 error = create_snapshot(snap_src, dir, dentry, name, namelen,
758 async_transid, readonly, inherit);
759 } else {
760 error = create_subvol(dir, dentry, name, namelen,
761 async_transid, inherit);
762 }
763 if (!error)
764 fsnotify_mkdir(dir, dentry);
765 out_up_read:
766 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
767 out_dput:
768 dput(dentry);
769 out_unlock:
770 mutex_unlock(&dir->i_mutex);
771 return error;
772 }
773
774 /*
775 * When we're defragging a range, we don't want to kick it off again
776 * if it is really just waiting for delalloc to send it down.
777 * If we find a nice big extent or delalloc range for the bytes in the
778 * file you want to defrag, we return 0 to let you know to skip this
779 * part of the file
780 */
781 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
782 {
783 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
784 struct extent_map *em = NULL;
785 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
786 u64 end;
787
788 read_lock(&em_tree->lock);
789 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
790 read_unlock(&em_tree->lock);
791
792 if (em) {
793 end = extent_map_end(em);
794 free_extent_map(em);
795 if (end - offset > thresh)
796 return 0;
797 }
798 /* if we already have a nice delalloc here, just stop */
799 thresh /= 2;
800 end = count_range_bits(io_tree, &offset, offset + thresh,
801 thresh, EXTENT_DELALLOC, 1);
802 if (end >= thresh)
803 return 0;
804 return 1;
805 }
806
807 /*
808 * helper function to walk through a file and find extents
809 * newer than a specific transid, and smaller than thresh.
810 *
811 * This is used by the defragging code to find new and small
812 * extents
813 */
814 static int find_new_extents(struct btrfs_root *root,
815 struct inode *inode, u64 newer_than,
816 u64 *off, int thresh)
817 {
818 struct btrfs_path *path;
819 struct btrfs_key min_key;
820 struct btrfs_key max_key;
821 struct extent_buffer *leaf;
822 struct btrfs_file_extent_item *extent;
823 int type;
824 int ret;
825 u64 ino = btrfs_ino(inode);
826
827 path = btrfs_alloc_path();
828 if (!path)
829 return -ENOMEM;
830
831 min_key.objectid = ino;
832 min_key.type = BTRFS_EXTENT_DATA_KEY;
833 min_key.offset = *off;
834
835 max_key.objectid = ino;
836 max_key.type = (u8)-1;
837 max_key.offset = (u64)-1;
838
839 path->keep_locks = 1;
840
841 while(1) {
842 ret = btrfs_search_forward(root, &min_key, &max_key,
843 path, newer_than);
844 if (ret != 0)
845 goto none;
846 if (min_key.objectid != ino)
847 goto none;
848 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
849 goto none;
850
851 leaf = path->nodes[0];
852 extent = btrfs_item_ptr(leaf, path->slots[0],
853 struct btrfs_file_extent_item);
854
855 type = btrfs_file_extent_type(leaf, extent);
856 if (type == BTRFS_FILE_EXTENT_REG &&
857 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
858 check_defrag_in_cache(inode, min_key.offset, thresh)) {
859 *off = min_key.offset;
860 btrfs_free_path(path);
861 return 0;
862 }
863
864 if (min_key.offset == (u64)-1)
865 goto none;
866
867 min_key.offset++;
868 btrfs_release_path(path);
869 }
870 none:
871 btrfs_free_path(path);
872 return -ENOENT;
873 }
874
875 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
876 {
877 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
878 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
879 struct extent_map *em;
880 u64 len = PAGE_CACHE_SIZE;
881
882 /*
883 * hopefully we have this extent in the tree already, try without
884 * the full extent lock
885 */
886 read_lock(&em_tree->lock);
887 em = lookup_extent_mapping(em_tree, start, len);
888 read_unlock(&em_tree->lock);
889
890 if (!em) {
891 /* get the big lock and read metadata off disk */
892 lock_extent(io_tree, start, start + len - 1);
893 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
894 unlock_extent(io_tree, start, start + len - 1);
895
896 if (IS_ERR(em))
897 return NULL;
898 }
899
900 return em;
901 }
902
903 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
904 {
905 struct extent_map *next;
906 bool ret = true;
907
908 /* this is the last extent */
909 if (em->start + em->len >= i_size_read(inode))
910 return false;
911
912 next = defrag_lookup_extent(inode, em->start + em->len);
913 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
914 ret = false;
915
916 free_extent_map(next);
917 return ret;
918 }
919
920 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
921 u64 *last_len, u64 *skip, u64 *defrag_end,
922 int compress)
923 {
924 struct extent_map *em;
925 int ret = 1;
926 bool next_mergeable = true;
927
928 /*
929 * make sure that once we start defragging an extent, we keep on
930 * defragging it
931 */
932 if (start < *defrag_end)
933 return 1;
934
935 *skip = 0;
936
937 em = defrag_lookup_extent(inode, start);
938 if (!em)
939 return 0;
940
941 /* this will cover holes, and inline extents */
942 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
943 ret = 0;
944 goto out;
945 }
946
947 next_mergeable = defrag_check_next_extent(inode, em);
948
949 /*
950 * we hit a real extent, if it is big or the next extent is not a
951 * real extent, don't bother defragging it
952 */
953 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
954 (em->len >= thresh || !next_mergeable))
955 ret = 0;
956 out:
957 /*
958 * last_len ends up being a counter of how many bytes we've defragged.
959 * every time we choose not to defrag an extent, we reset *last_len
960 * so that the next tiny extent will force a defrag.
961 *
962 * The end result of this is that tiny extents before a single big
963 * extent will force at least part of that big extent to be defragged.
964 */
965 if (ret) {
966 *defrag_end = extent_map_end(em);
967 } else {
968 *last_len = 0;
969 *skip = extent_map_end(em);
970 *defrag_end = 0;
971 }
972
973 free_extent_map(em);
974 return ret;
975 }
976
977 /*
978 * it doesn't do much good to defrag one or two pages
979 * at a time. This pulls in a nice chunk of pages
980 * to COW and defrag.
981 *
982 * It also makes sure the delalloc code has enough
983 * dirty data to avoid making new small extents as part
984 * of the defrag
985 *
986 * It's a good idea to start RA on this range
987 * before calling this.
988 */
989 static int cluster_pages_for_defrag(struct inode *inode,
990 struct page **pages,
991 unsigned long start_index,
992 int num_pages)
993 {
994 unsigned long file_end;
995 u64 isize = i_size_read(inode);
996 u64 page_start;
997 u64 page_end;
998 u64 page_cnt;
999 int ret;
1000 int i;
1001 int i_done;
1002 struct btrfs_ordered_extent *ordered;
1003 struct extent_state *cached_state = NULL;
1004 struct extent_io_tree *tree;
1005 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1006
1007 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1008 if (!isize || start_index > file_end)
1009 return 0;
1010
1011 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1012
1013 ret = btrfs_delalloc_reserve_space(inode,
1014 page_cnt << PAGE_CACHE_SHIFT);
1015 if (ret)
1016 return ret;
1017 i_done = 0;
1018 tree = &BTRFS_I(inode)->io_tree;
1019
1020 /* step one, lock all the pages */
1021 for (i = 0; i < page_cnt; i++) {
1022 struct page *page;
1023 again:
1024 page = find_or_create_page(inode->i_mapping,
1025 start_index + i, mask);
1026 if (!page)
1027 break;
1028
1029 page_start = page_offset(page);
1030 page_end = page_start + PAGE_CACHE_SIZE - 1;
1031 while (1) {
1032 lock_extent(tree, page_start, page_end);
1033 ordered = btrfs_lookup_ordered_extent(inode,
1034 page_start);
1035 unlock_extent(tree, page_start, page_end);
1036 if (!ordered)
1037 break;
1038
1039 unlock_page(page);
1040 btrfs_start_ordered_extent(inode, ordered, 1);
1041 btrfs_put_ordered_extent(ordered);
1042 lock_page(page);
1043 /*
1044 * we unlocked the page above, so we need check if
1045 * it was released or not.
1046 */
1047 if (page->mapping != inode->i_mapping) {
1048 unlock_page(page);
1049 page_cache_release(page);
1050 goto again;
1051 }
1052 }
1053
1054 if (!PageUptodate(page)) {
1055 btrfs_readpage(NULL, page);
1056 lock_page(page);
1057 if (!PageUptodate(page)) {
1058 unlock_page(page);
1059 page_cache_release(page);
1060 ret = -EIO;
1061 break;
1062 }
1063 }
1064
1065 if (page->mapping != inode->i_mapping) {
1066 unlock_page(page);
1067 page_cache_release(page);
1068 goto again;
1069 }
1070
1071 pages[i] = page;
1072 i_done++;
1073 }
1074 if (!i_done || ret)
1075 goto out;
1076
1077 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1078 goto out;
1079
1080 /*
1081 * so now we have a nice long stream of locked
1082 * and up to date pages, lets wait on them
1083 */
1084 for (i = 0; i < i_done; i++)
1085 wait_on_page_writeback(pages[i]);
1086
1087 page_start = page_offset(pages[0]);
1088 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1089
1090 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1091 page_start, page_end - 1, 0, &cached_state);
1092 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1093 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1094 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1095 &cached_state, GFP_NOFS);
1096
1097 if (i_done != page_cnt) {
1098 spin_lock(&BTRFS_I(inode)->lock);
1099 BTRFS_I(inode)->outstanding_extents++;
1100 spin_unlock(&BTRFS_I(inode)->lock);
1101 btrfs_delalloc_release_space(inode,
1102 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1103 }
1104
1105
1106 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1107 &cached_state, GFP_NOFS);
1108
1109 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1110 page_start, page_end - 1, &cached_state,
1111 GFP_NOFS);
1112
1113 for (i = 0; i < i_done; i++) {
1114 clear_page_dirty_for_io(pages[i]);
1115 ClearPageChecked(pages[i]);
1116 set_page_extent_mapped(pages[i]);
1117 set_page_dirty(pages[i]);
1118 unlock_page(pages[i]);
1119 page_cache_release(pages[i]);
1120 }
1121 return i_done;
1122 out:
1123 for (i = 0; i < i_done; i++) {
1124 unlock_page(pages[i]);
1125 page_cache_release(pages[i]);
1126 }
1127 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1128 return ret;
1129
1130 }
1131
1132 int btrfs_defrag_file(struct inode *inode, struct file *file,
1133 struct btrfs_ioctl_defrag_range_args *range,
1134 u64 newer_than, unsigned long max_to_defrag)
1135 {
1136 struct btrfs_root *root = BTRFS_I(inode)->root;
1137 struct file_ra_state *ra = NULL;
1138 unsigned long last_index;
1139 u64 isize = i_size_read(inode);
1140 u64 last_len = 0;
1141 u64 skip = 0;
1142 u64 defrag_end = 0;
1143 u64 newer_off = range->start;
1144 unsigned long i;
1145 unsigned long ra_index = 0;
1146 int ret;
1147 int defrag_count = 0;
1148 int compress_type = BTRFS_COMPRESS_ZLIB;
1149 int extent_thresh = range->extent_thresh;
1150 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1151 int cluster = max_cluster;
1152 u64 new_align = ~((u64)128 * 1024 - 1);
1153 struct page **pages = NULL;
1154
1155 if (extent_thresh == 0)
1156 extent_thresh = 256 * 1024;
1157
1158 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1159 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1160 return -EINVAL;
1161 if (range->compress_type)
1162 compress_type = range->compress_type;
1163 }
1164
1165 if (isize == 0)
1166 return 0;
1167
1168 /*
1169 * if we were not given a file, allocate a readahead
1170 * context
1171 */
1172 if (!file) {
1173 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1174 if (!ra)
1175 return -ENOMEM;
1176 file_ra_state_init(ra, inode->i_mapping);
1177 } else {
1178 ra = &file->f_ra;
1179 }
1180
1181 pages = kmalloc(sizeof(struct page *) * max_cluster,
1182 GFP_NOFS);
1183 if (!pages) {
1184 ret = -ENOMEM;
1185 goto out_ra;
1186 }
1187
1188 /* find the last page to defrag */
1189 if (range->start + range->len > range->start) {
1190 last_index = min_t(u64, isize - 1,
1191 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1192 } else {
1193 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1194 }
1195
1196 if (newer_than) {
1197 ret = find_new_extents(root, inode, newer_than,
1198 &newer_off, 64 * 1024);
1199 if (!ret) {
1200 range->start = newer_off;
1201 /*
1202 * we always align our defrag to help keep
1203 * the extents in the file evenly spaced
1204 */
1205 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1206 } else
1207 goto out_ra;
1208 } else {
1209 i = range->start >> PAGE_CACHE_SHIFT;
1210 }
1211 if (!max_to_defrag)
1212 max_to_defrag = last_index + 1;
1213
1214 /*
1215 * make writeback starts from i, so the defrag range can be
1216 * written sequentially.
1217 */
1218 if (i < inode->i_mapping->writeback_index)
1219 inode->i_mapping->writeback_index = i;
1220
1221 while (i <= last_index && defrag_count < max_to_defrag &&
1222 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1223 PAGE_CACHE_SHIFT)) {
1224 /*
1225 * make sure we stop running if someone unmounts
1226 * the FS
1227 */
1228 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1229 break;
1230
1231 if (btrfs_defrag_cancelled(root->fs_info)) {
1232 printk(KERN_DEBUG "btrfs: defrag_file cancelled\n");
1233 ret = -EAGAIN;
1234 break;
1235 }
1236
1237 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1238 extent_thresh, &last_len, &skip,
1239 &defrag_end, range->flags &
1240 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1241 unsigned long next;
1242 /*
1243 * the should_defrag function tells us how much to skip
1244 * bump our counter by the suggested amount
1245 */
1246 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1247 i = max(i + 1, next);
1248 continue;
1249 }
1250
1251 if (!newer_than) {
1252 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1253 PAGE_CACHE_SHIFT) - i;
1254 cluster = min(cluster, max_cluster);
1255 } else {
1256 cluster = max_cluster;
1257 }
1258
1259 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1260 BTRFS_I(inode)->force_compress = compress_type;
1261
1262 if (i + cluster > ra_index) {
1263 ra_index = max(i, ra_index);
1264 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1265 cluster);
1266 ra_index += max_cluster;
1267 }
1268
1269 mutex_lock(&inode->i_mutex);
1270 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1271 if (ret < 0) {
1272 mutex_unlock(&inode->i_mutex);
1273 goto out_ra;
1274 }
1275
1276 defrag_count += ret;
1277 balance_dirty_pages_ratelimited(inode->i_mapping);
1278 mutex_unlock(&inode->i_mutex);
1279
1280 if (newer_than) {
1281 if (newer_off == (u64)-1)
1282 break;
1283
1284 if (ret > 0)
1285 i += ret;
1286
1287 newer_off = max(newer_off + 1,
1288 (u64)i << PAGE_CACHE_SHIFT);
1289
1290 ret = find_new_extents(root, inode,
1291 newer_than, &newer_off,
1292 64 * 1024);
1293 if (!ret) {
1294 range->start = newer_off;
1295 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1296 } else {
1297 break;
1298 }
1299 } else {
1300 if (ret > 0) {
1301 i += ret;
1302 last_len += ret << PAGE_CACHE_SHIFT;
1303 } else {
1304 i++;
1305 last_len = 0;
1306 }
1307 }
1308 }
1309
1310 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1311 filemap_flush(inode->i_mapping);
1312
1313 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1314 /* the filemap_flush will queue IO into the worker threads, but
1315 * we have to make sure the IO is actually started and that
1316 * ordered extents get created before we return
1317 */
1318 atomic_inc(&root->fs_info->async_submit_draining);
1319 while (atomic_read(&root->fs_info->nr_async_submits) ||
1320 atomic_read(&root->fs_info->async_delalloc_pages)) {
1321 wait_event(root->fs_info->async_submit_wait,
1322 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1323 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1324 }
1325 atomic_dec(&root->fs_info->async_submit_draining);
1326
1327 mutex_lock(&inode->i_mutex);
1328 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1329 mutex_unlock(&inode->i_mutex);
1330 }
1331
1332 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1333 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1334 }
1335
1336 ret = defrag_count;
1337
1338 out_ra:
1339 if (!file)
1340 kfree(ra);
1341 kfree(pages);
1342 return ret;
1343 }
1344
1345 static noinline int btrfs_ioctl_resize(struct file *file,
1346 void __user *arg)
1347 {
1348 u64 new_size;
1349 u64 old_size;
1350 u64 devid = 1;
1351 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1352 struct btrfs_ioctl_vol_args *vol_args;
1353 struct btrfs_trans_handle *trans;
1354 struct btrfs_device *device = NULL;
1355 char *sizestr;
1356 char *devstr = NULL;
1357 int ret = 0;
1358 int mod = 0;
1359
1360 if (!capable(CAP_SYS_ADMIN))
1361 return -EPERM;
1362
1363 ret = mnt_want_write_file(file);
1364 if (ret)
1365 return ret;
1366
1367 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1368 1)) {
1369 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1370 mnt_drop_write_file(file);
1371 return -EINVAL;
1372 }
1373
1374 mutex_lock(&root->fs_info->volume_mutex);
1375 vol_args = memdup_user(arg, sizeof(*vol_args));
1376 if (IS_ERR(vol_args)) {
1377 ret = PTR_ERR(vol_args);
1378 goto out;
1379 }
1380
1381 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1382
1383 sizestr = vol_args->name;
1384 devstr = strchr(sizestr, ':');
1385 if (devstr) {
1386 char *end;
1387 sizestr = devstr + 1;
1388 *devstr = '\0';
1389 devstr = vol_args->name;
1390 devid = simple_strtoull(devstr, &end, 10);
1391 if (!devid) {
1392 ret = -EINVAL;
1393 goto out_free;
1394 }
1395 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1396 (unsigned long long)devid);
1397 }
1398
1399 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1400 if (!device) {
1401 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1402 (unsigned long long)devid);
1403 ret = -ENODEV;
1404 goto out_free;
1405 }
1406
1407 if (!device->writeable) {
1408 printk(KERN_INFO "btrfs: resizer unable to apply on "
1409 "readonly device %llu\n",
1410 (unsigned long long)devid);
1411 ret = -EPERM;
1412 goto out_free;
1413 }
1414
1415 if (!strcmp(sizestr, "max"))
1416 new_size = device->bdev->bd_inode->i_size;
1417 else {
1418 if (sizestr[0] == '-') {
1419 mod = -1;
1420 sizestr++;
1421 } else if (sizestr[0] == '+') {
1422 mod = 1;
1423 sizestr++;
1424 }
1425 new_size = memparse(sizestr, NULL);
1426 if (new_size == 0) {
1427 ret = -EINVAL;
1428 goto out_free;
1429 }
1430 }
1431
1432 if (device->is_tgtdev_for_dev_replace) {
1433 ret = -EPERM;
1434 goto out_free;
1435 }
1436
1437 old_size = device->total_bytes;
1438
1439 if (mod < 0) {
1440 if (new_size > old_size) {
1441 ret = -EINVAL;
1442 goto out_free;
1443 }
1444 new_size = old_size - new_size;
1445 } else if (mod > 0) {
1446 new_size = old_size + new_size;
1447 }
1448
1449 if (new_size < 256 * 1024 * 1024) {
1450 ret = -EINVAL;
1451 goto out_free;
1452 }
1453 if (new_size > device->bdev->bd_inode->i_size) {
1454 ret = -EFBIG;
1455 goto out_free;
1456 }
1457
1458 do_div(new_size, root->sectorsize);
1459 new_size *= root->sectorsize;
1460
1461 printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1462 rcu_str_deref(device->name),
1463 (unsigned long long)new_size);
1464
1465 if (new_size > old_size) {
1466 trans = btrfs_start_transaction(root, 0);
1467 if (IS_ERR(trans)) {
1468 ret = PTR_ERR(trans);
1469 goto out_free;
1470 }
1471 ret = btrfs_grow_device(trans, device, new_size);
1472 btrfs_commit_transaction(trans, root);
1473 } else if (new_size < old_size) {
1474 ret = btrfs_shrink_device(device, new_size);
1475 } /* equal, nothing need to do */
1476
1477 out_free:
1478 kfree(vol_args);
1479 out:
1480 mutex_unlock(&root->fs_info->volume_mutex);
1481 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1482 mnt_drop_write_file(file);
1483 return ret;
1484 }
1485
1486 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1487 char *name, unsigned long fd, int subvol,
1488 u64 *transid, bool readonly,
1489 struct btrfs_qgroup_inherit *inherit)
1490 {
1491 int namelen;
1492 int ret = 0;
1493
1494 ret = mnt_want_write_file(file);
1495 if (ret)
1496 goto out;
1497
1498 namelen = strlen(name);
1499 if (strchr(name, '/')) {
1500 ret = -EINVAL;
1501 goto out_drop_write;
1502 }
1503
1504 if (name[0] == '.' &&
1505 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1506 ret = -EEXIST;
1507 goto out_drop_write;
1508 }
1509
1510 if (subvol) {
1511 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1512 NULL, transid, readonly, inherit);
1513 } else {
1514 struct fd src = fdget(fd);
1515 struct inode *src_inode;
1516 if (!src.file) {
1517 ret = -EINVAL;
1518 goto out_drop_write;
1519 }
1520
1521 src_inode = file_inode(src.file);
1522 if (src_inode->i_sb != file_inode(file)->i_sb) {
1523 printk(KERN_INFO "btrfs: Snapshot src from "
1524 "another FS\n");
1525 ret = -EINVAL;
1526 } else {
1527 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1528 BTRFS_I(src_inode)->root,
1529 transid, readonly, inherit);
1530 }
1531 fdput(src);
1532 }
1533 out_drop_write:
1534 mnt_drop_write_file(file);
1535 out:
1536 return ret;
1537 }
1538
1539 static noinline int btrfs_ioctl_snap_create(struct file *file,
1540 void __user *arg, int subvol)
1541 {
1542 struct btrfs_ioctl_vol_args *vol_args;
1543 int ret;
1544
1545 vol_args = memdup_user(arg, sizeof(*vol_args));
1546 if (IS_ERR(vol_args))
1547 return PTR_ERR(vol_args);
1548 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1549
1550 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1551 vol_args->fd, subvol,
1552 NULL, false, NULL);
1553
1554 kfree(vol_args);
1555 return ret;
1556 }
1557
1558 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1559 void __user *arg, int subvol)
1560 {
1561 struct btrfs_ioctl_vol_args_v2 *vol_args;
1562 int ret;
1563 u64 transid = 0;
1564 u64 *ptr = NULL;
1565 bool readonly = false;
1566 struct btrfs_qgroup_inherit *inherit = NULL;
1567
1568 vol_args = memdup_user(arg, sizeof(*vol_args));
1569 if (IS_ERR(vol_args))
1570 return PTR_ERR(vol_args);
1571 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1572
1573 if (vol_args->flags &
1574 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1575 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1576 ret = -EOPNOTSUPP;
1577 goto out;
1578 }
1579
1580 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1581 ptr = &transid;
1582 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1583 readonly = true;
1584 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1585 if (vol_args->size > PAGE_CACHE_SIZE) {
1586 ret = -EINVAL;
1587 goto out;
1588 }
1589 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1590 if (IS_ERR(inherit)) {
1591 ret = PTR_ERR(inherit);
1592 goto out;
1593 }
1594 }
1595
1596 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1597 vol_args->fd, subvol, ptr,
1598 readonly, inherit);
1599
1600 if (ret == 0 && ptr &&
1601 copy_to_user(arg +
1602 offsetof(struct btrfs_ioctl_vol_args_v2,
1603 transid), ptr, sizeof(*ptr)))
1604 ret = -EFAULT;
1605 out:
1606 kfree(vol_args);
1607 kfree(inherit);
1608 return ret;
1609 }
1610
1611 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1612 void __user *arg)
1613 {
1614 struct inode *inode = file_inode(file);
1615 struct btrfs_root *root = BTRFS_I(inode)->root;
1616 int ret = 0;
1617 u64 flags = 0;
1618
1619 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1620 return -EINVAL;
1621
1622 down_read(&root->fs_info->subvol_sem);
1623 if (btrfs_root_readonly(root))
1624 flags |= BTRFS_SUBVOL_RDONLY;
1625 up_read(&root->fs_info->subvol_sem);
1626
1627 if (copy_to_user(arg, &flags, sizeof(flags)))
1628 ret = -EFAULT;
1629
1630 return ret;
1631 }
1632
1633 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1634 void __user *arg)
1635 {
1636 struct inode *inode = file_inode(file);
1637 struct btrfs_root *root = BTRFS_I(inode)->root;
1638 struct btrfs_trans_handle *trans;
1639 u64 root_flags;
1640 u64 flags;
1641 int ret = 0;
1642
1643 ret = mnt_want_write_file(file);
1644 if (ret)
1645 goto out;
1646
1647 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1648 ret = -EINVAL;
1649 goto out_drop_write;
1650 }
1651
1652 if (copy_from_user(&flags, arg, sizeof(flags))) {
1653 ret = -EFAULT;
1654 goto out_drop_write;
1655 }
1656
1657 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1658 ret = -EINVAL;
1659 goto out_drop_write;
1660 }
1661
1662 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1663 ret = -EOPNOTSUPP;
1664 goto out_drop_write;
1665 }
1666
1667 if (!inode_owner_or_capable(inode)) {
1668 ret = -EACCES;
1669 goto out_drop_write;
1670 }
1671
1672 down_write(&root->fs_info->subvol_sem);
1673
1674 /* nothing to do */
1675 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1676 goto out_drop_sem;
1677
1678 root_flags = btrfs_root_flags(&root->root_item);
1679 if (flags & BTRFS_SUBVOL_RDONLY)
1680 btrfs_set_root_flags(&root->root_item,
1681 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1682 else
1683 btrfs_set_root_flags(&root->root_item,
1684 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1685
1686 trans = btrfs_start_transaction(root, 1);
1687 if (IS_ERR(trans)) {
1688 ret = PTR_ERR(trans);
1689 goto out_reset;
1690 }
1691
1692 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1693 &root->root_key, &root->root_item);
1694
1695 btrfs_commit_transaction(trans, root);
1696 out_reset:
1697 if (ret)
1698 btrfs_set_root_flags(&root->root_item, root_flags);
1699 out_drop_sem:
1700 up_write(&root->fs_info->subvol_sem);
1701 out_drop_write:
1702 mnt_drop_write_file(file);
1703 out:
1704 return ret;
1705 }
1706
1707 /*
1708 * helper to check if the subvolume references other subvolumes
1709 */
1710 static noinline int may_destroy_subvol(struct btrfs_root *root)
1711 {
1712 struct btrfs_path *path;
1713 struct btrfs_key key;
1714 int ret;
1715
1716 path = btrfs_alloc_path();
1717 if (!path)
1718 return -ENOMEM;
1719
1720 key.objectid = root->root_key.objectid;
1721 key.type = BTRFS_ROOT_REF_KEY;
1722 key.offset = (u64)-1;
1723
1724 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1725 &key, path, 0, 0);
1726 if (ret < 0)
1727 goto out;
1728 BUG_ON(ret == 0);
1729
1730 ret = 0;
1731 if (path->slots[0] > 0) {
1732 path->slots[0]--;
1733 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1734 if (key.objectid == root->root_key.objectid &&
1735 key.type == BTRFS_ROOT_REF_KEY)
1736 ret = -ENOTEMPTY;
1737 }
1738 out:
1739 btrfs_free_path(path);
1740 return ret;
1741 }
1742
1743 static noinline int key_in_sk(struct btrfs_key *key,
1744 struct btrfs_ioctl_search_key *sk)
1745 {
1746 struct btrfs_key test;
1747 int ret;
1748
1749 test.objectid = sk->min_objectid;
1750 test.type = sk->min_type;
1751 test.offset = sk->min_offset;
1752
1753 ret = btrfs_comp_cpu_keys(key, &test);
1754 if (ret < 0)
1755 return 0;
1756
1757 test.objectid = sk->max_objectid;
1758 test.type = sk->max_type;
1759 test.offset = sk->max_offset;
1760
1761 ret = btrfs_comp_cpu_keys(key, &test);
1762 if (ret > 0)
1763 return 0;
1764 return 1;
1765 }
1766
1767 static noinline int copy_to_sk(struct btrfs_root *root,
1768 struct btrfs_path *path,
1769 struct btrfs_key *key,
1770 struct btrfs_ioctl_search_key *sk,
1771 char *buf,
1772 unsigned long *sk_offset,
1773 int *num_found)
1774 {
1775 u64 found_transid;
1776 struct extent_buffer *leaf;
1777 struct btrfs_ioctl_search_header sh;
1778 unsigned long item_off;
1779 unsigned long item_len;
1780 int nritems;
1781 int i;
1782 int slot;
1783 int ret = 0;
1784
1785 leaf = path->nodes[0];
1786 slot = path->slots[0];
1787 nritems = btrfs_header_nritems(leaf);
1788
1789 if (btrfs_header_generation(leaf) > sk->max_transid) {
1790 i = nritems;
1791 goto advance_key;
1792 }
1793 found_transid = btrfs_header_generation(leaf);
1794
1795 for (i = slot; i < nritems; i++) {
1796 item_off = btrfs_item_ptr_offset(leaf, i);
1797 item_len = btrfs_item_size_nr(leaf, i);
1798
1799 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1800 item_len = 0;
1801
1802 if (sizeof(sh) + item_len + *sk_offset >
1803 BTRFS_SEARCH_ARGS_BUFSIZE) {
1804 ret = 1;
1805 goto overflow;
1806 }
1807
1808 btrfs_item_key_to_cpu(leaf, key, i);
1809 if (!key_in_sk(key, sk))
1810 continue;
1811
1812 sh.objectid = key->objectid;
1813 sh.offset = key->offset;
1814 sh.type = key->type;
1815 sh.len = item_len;
1816 sh.transid = found_transid;
1817
1818 /* copy search result header */
1819 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1820 *sk_offset += sizeof(sh);
1821
1822 if (item_len) {
1823 char *p = buf + *sk_offset;
1824 /* copy the item */
1825 read_extent_buffer(leaf, p,
1826 item_off, item_len);
1827 *sk_offset += item_len;
1828 }
1829 (*num_found)++;
1830
1831 if (*num_found >= sk->nr_items)
1832 break;
1833 }
1834 advance_key:
1835 ret = 0;
1836 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1837 key->offset++;
1838 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1839 key->offset = 0;
1840 key->type++;
1841 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1842 key->offset = 0;
1843 key->type = 0;
1844 key->objectid++;
1845 } else
1846 ret = 1;
1847 overflow:
1848 return ret;
1849 }
1850
1851 static noinline int search_ioctl(struct inode *inode,
1852 struct btrfs_ioctl_search_args *args)
1853 {
1854 struct btrfs_root *root;
1855 struct btrfs_key key;
1856 struct btrfs_key max_key;
1857 struct btrfs_path *path;
1858 struct btrfs_ioctl_search_key *sk = &args->key;
1859 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1860 int ret;
1861 int num_found = 0;
1862 unsigned long sk_offset = 0;
1863
1864 path = btrfs_alloc_path();
1865 if (!path)
1866 return -ENOMEM;
1867
1868 if (sk->tree_id == 0) {
1869 /* search the root of the inode that was passed */
1870 root = BTRFS_I(inode)->root;
1871 } else {
1872 key.objectid = sk->tree_id;
1873 key.type = BTRFS_ROOT_ITEM_KEY;
1874 key.offset = (u64)-1;
1875 root = btrfs_read_fs_root_no_name(info, &key);
1876 if (IS_ERR(root)) {
1877 printk(KERN_ERR "could not find root %llu\n",
1878 sk->tree_id);
1879 btrfs_free_path(path);
1880 return -ENOENT;
1881 }
1882 }
1883
1884 key.objectid = sk->min_objectid;
1885 key.type = sk->min_type;
1886 key.offset = sk->min_offset;
1887
1888 max_key.objectid = sk->max_objectid;
1889 max_key.type = sk->max_type;
1890 max_key.offset = sk->max_offset;
1891
1892 path->keep_locks = 1;
1893
1894 while(1) {
1895 ret = btrfs_search_forward(root, &key, &max_key, path,
1896 sk->min_transid);
1897 if (ret != 0) {
1898 if (ret > 0)
1899 ret = 0;
1900 goto err;
1901 }
1902 ret = copy_to_sk(root, path, &key, sk, args->buf,
1903 &sk_offset, &num_found);
1904 btrfs_release_path(path);
1905 if (ret || num_found >= sk->nr_items)
1906 break;
1907
1908 }
1909 ret = 0;
1910 err:
1911 sk->nr_items = num_found;
1912 btrfs_free_path(path);
1913 return ret;
1914 }
1915
1916 static noinline int btrfs_ioctl_tree_search(struct file *file,
1917 void __user *argp)
1918 {
1919 struct btrfs_ioctl_search_args *args;
1920 struct inode *inode;
1921 int ret;
1922
1923 if (!capable(CAP_SYS_ADMIN))
1924 return -EPERM;
1925
1926 args = memdup_user(argp, sizeof(*args));
1927 if (IS_ERR(args))
1928 return PTR_ERR(args);
1929
1930 inode = file_inode(file);
1931 ret = search_ioctl(inode, args);
1932 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1933 ret = -EFAULT;
1934 kfree(args);
1935 return ret;
1936 }
1937
1938 /*
1939 * Search INODE_REFs to identify path name of 'dirid' directory
1940 * in a 'tree_id' tree. and sets path name to 'name'.
1941 */
1942 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1943 u64 tree_id, u64 dirid, char *name)
1944 {
1945 struct btrfs_root *root;
1946 struct btrfs_key key;
1947 char *ptr;
1948 int ret = -1;
1949 int slot;
1950 int len;
1951 int total_len = 0;
1952 struct btrfs_inode_ref *iref;
1953 struct extent_buffer *l;
1954 struct btrfs_path *path;
1955
1956 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1957 name[0]='\0';
1958 return 0;
1959 }
1960
1961 path = btrfs_alloc_path();
1962 if (!path)
1963 return -ENOMEM;
1964
1965 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1966
1967 key.objectid = tree_id;
1968 key.type = BTRFS_ROOT_ITEM_KEY;
1969 key.offset = (u64)-1;
1970 root = btrfs_read_fs_root_no_name(info, &key);
1971 if (IS_ERR(root)) {
1972 printk(KERN_ERR "could not find root %llu\n", tree_id);
1973 ret = -ENOENT;
1974 goto out;
1975 }
1976
1977 key.objectid = dirid;
1978 key.type = BTRFS_INODE_REF_KEY;
1979 key.offset = (u64)-1;
1980
1981 while(1) {
1982 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1983 if (ret < 0)
1984 goto out;
1985
1986 l = path->nodes[0];
1987 slot = path->slots[0];
1988 if (ret > 0 && slot > 0)
1989 slot--;
1990 btrfs_item_key_to_cpu(l, &key, slot);
1991
1992 if (ret > 0 && (key.objectid != dirid ||
1993 key.type != BTRFS_INODE_REF_KEY)) {
1994 ret = -ENOENT;
1995 goto out;
1996 }
1997
1998 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1999 len = btrfs_inode_ref_name_len(l, iref);
2000 ptr -= len + 1;
2001 total_len += len + 1;
2002 if (ptr < name)
2003 goto out;
2004
2005 *(ptr + len) = '/';
2006 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
2007
2008 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2009 break;
2010
2011 btrfs_release_path(path);
2012 key.objectid = key.offset;
2013 key.offset = (u64)-1;
2014 dirid = key.objectid;
2015 }
2016 if (ptr < name)
2017 goto out;
2018 memmove(name, ptr, total_len);
2019 name[total_len]='\0';
2020 ret = 0;
2021 out:
2022 btrfs_free_path(path);
2023 return ret;
2024 }
2025
2026 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2027 void __user *argp)
2028 {
2029 struct btrfs_ioctl_ino_lookup_args *args;
2030 struct inode *inode;
2031 int ret;
2032
2033 if (!capable(CAP_SYS_ADMIN))
2034 return -EPERM;
2035
2036 args = memdup_user(argp, sizeof(*args));
2037 if (IS_ERR(args))
2038 return PTR_ERR(args);
2039
2040 inode = file_inode(file);
2041
2042 if (args->treeid == 0)
2043 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2044
2045 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2046 args->treeid, args->objectid,
2047 args->name);
2048
2049 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2050 ret = -EFAULT;
2051
2052 kfree(args);
2053 return ret;
2054 }
2055
2056 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2057 void __user *arg)
2058 {
2059 struct dentry *parent = fdentry(file);
2060 struct dentry *dentry;
2061 struct inode *dir = parent->d_inode;
2062 struct inode *inode;
2063 struct btrfs_root *root = BTRFS_I(dir)->root;
2064 struct btrfs_root *dest = NULL;
2065 struct btrfs_ioctl_vol_args *vol_args;
2066 struct btrfs_trans_handle *trans;
2067 struct btrfs_block_rsv block_rsv;
2068 u64 qgroup_reserved;
2069 int namelen;
2070 int ret;
2071 int err = 0;
2072
2073 vol_args = memdup_user(arg, sizeof(*vol_args));
2074 if (IS_ERR(vol_args))
2075 return PTR_ERR(vol_args);
2076
2077 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2078 namelen = strlen(vol_args->name);
2079 if (strchr(vol_args->name, '/') ||
2080 strncmp(vol_args->name, "..", namelen) == 0) {
2081 err = -EINVAL;
2082 goto out;
2083 }
2084
2085 err = mnt_want_write_file(file);
2086 if (err)
2087 goto out;
2088
2089 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
2090 dentry = lookup_one_len(vol_args->name, parent, namelen);
2091 if (IS_ERR(dentry)) {
2092 err = PTR_ERR(dentry);
2093 goto out_unlock_dir;
2094 }
2095
2096 if (!dentry->d_inode) {
2097 err = -ENOENT;
2098 goto out_dput;
2099 }
2100
2101 inode = dentry->d_inode;
2102 dest = BTRFS_I(inode)->root;
2103 if (!capable(CAP_SYS_ADMIN)){
2104 /*
2105 * Regular user. Only allow this with a special mount
2106 * option, when the user has write+exec access to the
2107 * subvol root, and when rmdir(2) would have been
2108 * allowed.
2109 *
2110 * Note that this is _not_ check that the subvol is
2111 * empty or doesn't contain data that we wouldn't
2112 * otherwise be able to delete.
2113 *
2114 * Users who want to delete empty subvols should try
2115 * rmdir(2).
2116 */
2117 err = -EPERM;
2118 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2119 goto out_dput;
2120
2121 /*
2122 * Do not allow deletion if the parent dir is the same
2123 * as the dir to be deleted. That means the ioctl
2124 * must be called on the dentry referencing the root
2125 * of the subvol, not a random directory contained
2126 * within it.
2127 */
2128 err = -EINVAL;
2129 if (root == dest)
2130 goto out_dput;
2131
2132 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2133 if (err)
2134 goto out_dput;
2135 }
2136
2137 /* check if subvolume may be deleted by a user */
2138 err = btrfs_may_delete(dir, dentry, 1);
2139 if (err)
2140 goto out_dput;
2141
2142 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2143 err = -EINVAL;
2144 goto out_dput;
2145 }
2146
2147 mutex_lock(&inode->i_mutex);
2148 err = d_invalidate(dentry);
2149 if (err)
2150 goto out_unlock;
2151
2152 down_write(&root->fs_info->subvol_sem);
2153
2154 err = may_destroy_subvol(dest);
2155 if (err)
2156 goto out_up_write;
2157
2158 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2159 /*
2160 * One for dir inode, two for dir entries, two for root
2161 * ref/backref.
2162 */
2163 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2164 5, &qgroup_reserved);
2165 if (err)
2166 goto out_up_write;
2167
2168 trans = btrfs_start_transaction(root, 0);
2169 if (IS_ERR(trans)) {
2170 err = PTR_ERR(trans);
2171 goto out_release;
2172 }
2173 trans->block_rsv = &block_rsv;
2174 trans->bytes_reserved = block_rsv.size;
2175
2176 ret = btrfs_unlink_subvol(trans, root, dir,
2177 dest->root_key.objectid,
2178 dentry->d_name.name,
2179 dentry->d_name.len);
2180 if (ret) {
2181 err = ret;
2182 btrfs_abort_transaction(trans, root, ret);
2183 goto out_end_trans;
2184 }
2185
2186 btrfs_record_root_in_trans(trans, dest);
2187
2188 memset(&dest->root_item.drop_progress, 0,
2189 sizeof(dest->root_item.drop_progress));
2190 dest->root_item.drop_level = 0;
2191 btrfs_set_root_refs(&dest->root_item, 0);
2192
2193 if (!xchg(&dest->orphan_item_inserted, 1)) {
2194 ret = btrfs_insert_orphan_item(trans,
2195 root->fs_info->tree_root,
2196 dest->root_key.objectid);
2197 if (ret) {
2198 btrfs_abort_transaction(trans, root, ret);
2199 err = ret;
2200 goto out_end_trans;
2201 }
2202 }
2203 out_end_trans:
2204 trans->block_rsv = NULL;
2205 trans->bytes_reserved = 0;
2206 ret = btrfs_end_transaction(trans, root);
2207 if (ret && !err)
2208 err = ret;
2209 inode->i_flags |= S_DEAD;
2210 out_release:
2211 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2212 out_up_write:
2213 up_write(&root->fs_info->subvol_sem);
2214 out_unlock:
2215 mutex_unlock(&inode->i_mutex);
2216 if (!err) {
2217 shrink_dcache_sb(root->fs_info->sb);
2218 btrfs_invalidate_inodes(dest);
2219 d_delete(dentry);
2220
2221 /* the last ref */
2222 if (dest->cache_inode) {
2223 iput(dest->cache_inode);
2224 dest->cache_inode = NULL;
2225 }
2226 }
2227 out_dput:
2228 dput(dentry);
2229 out_unlock_dir:
2230 mutex_unlock(&dir->i_mutex);
2231 mnt_drop_write_file(file);
2232 out:
2233 kfree(vol_args);
2234 return err;
2235 }
2236
2237 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2238 {
2239 struct inode *inode = file_inode(file);
2240 struct btrfs_root *root = BTRFS_I(inode)->root;
2241 struct btrfs_ioctl_defrag_range_args *range;
2242 int ret;
2243
2244 ret = mnt_want_write_file(file);
2245 if (ret)
2246 return ret;
2247
2248 if (btrfs_root_readonly(root)) {
2249 ret = -EROFS;
2250 goto out;
2251 }
2252
2253 switch (inode->i_mode & S_IFMT) {
2254 case S_IFDIR:
2255 if (!capable(CAP_SYS_ADMIN)) {
2256 ret = -EPERM;
2257 goto out;
2258 }
2259 ret = btrfs_defrag_root(root);
2260 if (ret)
2261 goto out;
2262 ret = btrfs_defrag_root(root->fs_info->extent_root);
2263 break;
2264 case S_IFREG:
2265 if (!(file->f_mode & FMODE_WRITE)) {
2266 ret = -EINVAL;
2267 goto out;
2268 }
2269
2270 range = kzalloc(sizeof(*range), GFP_KERNEL);
2271 if (!range) {
2272 ret = -ENOMEM;
2273 goto out;
2274 }
2275
2276 if (argp) {
2277 if (copy_from_user(range, argp,
2278 sizeof(*range))) {
2279 ret = -EFAULT;
2280 kfree(range);
2281 goto out;
2282 }
2283 /* compression requires us to start the IO */
2284 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2285 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2286 range->extent_thresh = (u32)-1;
2287 }
2288 } else {
2289 /* the rest are all set to zero by kzalloc */
2290 range->len = (u64)-1;
2291 }
2292 ret = btrfs_defrag_file(file_inode(file), file,
2293 range, 0, 0);
2294 if (ret > 0)
2295 ret = 0;
2296 kfree(range);
2297 break;
2298 default:
2299 ret = -EINVAL;
2300 }
2301 out:
2302 mnt_drop_write_file(file);
2303 return ret;
2304 }
2305
2306 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2307 {
2308 struct btrfs_ioctl_vol_args *vol_args;
2309 int ret;
2310
2311 if (!capable(CAP_SYS_ADMIN))
2312 return -EPERM;
2313
2314 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2315 1)) {
2316 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2317 return -EINVAL;
2318 }
2319
2320 mutex_lock(&root->fs_info->volume_mutex);
2321 vol_args = memdup_user(arg, sizeof(*vol_args));
2322 if (IS_ERR(vol_args)) {
2323 ret = PTR_ERR(vol_args);
2324 goto out;
2325 }
2326
2327 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2328 ret = btrfs_init_new_device(root, vol_args->name);
2329
2330 kfree(vol_args);
2331 out:
2332 mutex_unlock(&root->fs_info->volume_mutex);
2333 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2334 return ret;
2335 }
2336
2337 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2338 {
2339 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2340 struct btrfs_ioctl_vol_args *vol_args;
2341 int ret;
2342
2343 if (!capable(CAP_SYS_ADMIN))
2344 return -EPERM;
2345
2346 ret = mnt_want_write_file(file);
2347 if (ret)
2348 return ret;
2349
2350 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2351 1)) {
2352 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2353 mnt_drop_write_file(file);
2354 return -EINVAL;
2355 }
2356
2357 mutex_lock(&root->fs_info->volume_mutex);
2358 vol_args = memdup_user(arg, sizeof(*vol_args));
2359 if (IS_ERR(vol_args)) {
2360 ret = PTR_ERR(vol_args);
2361 goto out;
2362 }
2363
2364 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2365 ret = btrfs_rm_device(root, vol_args->name);
2366
2367 kfree(vol_args);
2368 out:
2369 mutex_unlock(&root->fs_info->volume_mutex);
2370 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2371 mnt_drop_write_file(file);
2372 return ret;
2373 }
2374
2375 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2376 {
2377 struct btrfs_ioctl_fs_info_args *fi_args;
2378 struct btrfs_device *device;
2379 struct btrfs_device *next;
2380 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2381 int ret = 0;
2382
2383 if (!capable(CAP_SYS_ADMIN))
2384 return -EPERM;
2385
2386 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2387 if (!fi_args)
2388 return -ENOMEM;
2389
2390 fi_args->num_devices = fs_devices->num_devices;
2391 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2392
2393 mutex_lock(&fs_devices->device_list_mutex);
2394 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2395 if (device->devid > fi_args->max_id)
2396 fi_args->max_id = device->devid;
2397 }
2398 mutex_unlock(&fs_devices->device_list_mutex);
2399
2400 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2401 ret = -EFAULT;
2402
2403 kfree(fi_args);
2404 return ret;
2405 }
2406
2407 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2408 {
2409 struct btrfs_ioctl_dev_info_args *di_args;
2410 struct btrfs_device *dev;
2411 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2412 int ret = 0;
2413 char *s_uuid = NULL;
2414 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2415
2416 if (!capable(CAP_SYS_ADMIN))
2417 return -EPERM;
2418
2419 di_args = memdup_user(arg, sizeof(*di_args));
2420 if (IS_ERR(di_args))
2421 return PTR_ERR(di_args);
2422
2423 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2424 s_uuid = di_args->uuid;
2425
2426 mutex_lock(&fs_devices->device_list_mutex);
2427 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2428 mutex_unlock(&fs_devices->device_list_mutex);
2429
2430 if (!dev) {
2431 ret = -ENODEV;
2432 goto out;
2433 }
2434
2435 di_args->devid = dev->devid;
2436 di_args->bytes_used = dev->bytes_used;
2437 di_args->total_bytes = dev->total_bytes;
2438 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2439 if (dev->name) {
2440 struct rcu_string *name;
2441
2442 rcu_read_lock();
2443 name = rcu_dereference(dev->name);
2444 strncpy(di_args->path, name->str, sizeof(di_args->path));
2445 rcu_read_unlock();
2446 di_args->path[sizeof(di_args->path) - 1] = 0;
2447 } else {
2448 di_args->path[0] = '\0';
2449 }
2450
2451 out:
2452 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2453 ret = -EFAULT;
2454
2455 kfree(di_args);
2456 return ret;
2457 }
2458
2459 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2460 u64 off, u64 olen, u64 destoff)
2461 {
2462 struct inode *inode = file_inode(file);
2463 struct btrfs_root *root = BTRFS_I(inode)->root;
2464 struct fd src_file;
2465 struct inode *src;
2466 struct btrfs_trans_handle *trans;
2467 struct btrfs_path *path;
2468 struct extent_buffer *leaf;
2469 char *buf;
2470 struct btrfs_key key;
2471 u32 nritems;
2472 int slot;
2473 int ret;
2474 u64 len = olen;
2475 u64 bs = root->fs_info->sb->s_blocksize;
2476
2477 /*
2478 * TODO:
2479 * - split compressed inline extents. annoying: we need to
2480 * decompress into destination's address_space (the file offset
2481 * may change, so source mapping won't do), then recompress (or
2482 * otherwise reinsert) a subrange.
2483 * - allow ranges within the same file to be cloned (provided
2484 * they don't overlap)?
2485 */
2486
2487 /* the destination must be opened for writing */
2488 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2489 return -EINVAL;
2490
2491 if (btrfs_root_readonly(root))
2492 return -EROFS;
2493
2494 ret = mnt_want_write_file(file);
2495 if (ret)
2496 return ret;
2497
2498 src_file = fdget(srcfd);
2499 if (!src_file.file) {
2500 ret = -EBADF;
2501 goto out_drop_write;
2502 }
2503
2504 ret = -EXDEV;
2505 if (src_file.file->f_path.mnt != file->f_path.mnt)
2506 goto out_fput;
2507
2508 src = file_inode(src_file.file);
2509
2510 ret = -EINVAL;
2511 if (src == inode)
2512 goto out_fput;
2513
2514 /* the src must be open for reading */
2515 if (!(src_file.file->f_mode & FMODE_READ))
2516 goto out_fput;
2517
2518 /* don't make the dst file partly checksummed */
2519 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2520 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2521 goto out_fput;
2522
2523 ret = -EISDIR;
2524 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2525 goto out_fput;
2526
2527 ret = -EXDEV;
2528 if (src->i_sb != inode->i_sb)
2529 goto out_fput;
2530
2531 ret = -ENOMEM;
2532 buf = vmalloc(btrfs_level_size(root, 0));
2533 if (!buf)
2534 goto out_fput;
2535
2536 path = btrfs_alloc_path();
2537 if (!path) {
2538 vfree(buf);
2539 goto out_fput;
2540 }
2541 path->reada = 2;
2542
2543 if (inode < src) {
2544 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2545 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2546 } else {
2547 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2548 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2549 }
2550
2551 /* determine range to clone */
2552 ret = -EINVAL;
2553 if (off + len > src->i_size || off + len < off)
2554 goto out_unlock;
2555 if (len == 0)
2556 olen = len = src->i_size - off;
2557 /* if we extend to eof, continue to block boundary */
2558 if (off + len == src->i_size)
2559 len = ALIGN(src->i_size, bs) - off;
2560
2561 /* verify the end result is block aligned */
2562 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2563 !IS_ALIGNED(destoff, bs))
2564 goto out_unlock;
2565
2566 if (destoff > inode->i_size) {
2567 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2568 if (ret)
2569 goto out_unlock;
2570 }
2571
2572 /* truncate page cache pages from target inode range */
2573 truncate_inode_pages_range(&inode->i_data, destoff,
2574 PAGE_CACHE_ALIGN(destoff + len) - 1);
2575
2576 /* do any pending delalloc/csum calc on src, one way or
2577 another, and lock file content */
2578 while (1) {
2579 struct btrfs_ordered_extent *ordered;
2580 lock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2581 ordered = btrfs_lookup_first_ordered_extent(src, off + len - 1);
2582 if (!ordered &&
2583 !test_range_bit(&BTRFS_I(src)->io_tree, off, off + len - 1,
2584 EXTENT_DELALLOC, 0, NULL))
2585 break;
2586 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2587 if (ordered)
2588 btrfs_put_ordered_extent(ordered);
2589 btrfs_wait_ordered_range(src, off, len);
2590 }
2591
2592 /* clone data */
2593 key.objectid = btrfs_ino(src);
2594 key.type = BTRFS_EXTENT_DATA_KEY;
2595 key.offset = 0;
2596
2597 while (1) {
2598 /*
2599 * note the key will change type as we walk through the
2600 * tree.
2601 */
2602 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2603 0, 0);
2604 if (ret < 0)
2605 goto out;
2606
2607 nritems = btrfs_header_nritems(path->nodes[0]);
2608 if (path->slots[0] >= nritems) {
2609 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2610 if (ret < 0)
2611 goto out;
2612 if (ret > 0)
2613 break;
2614 nritems = btrfs_header_nritems(path->nodes[0]);
2615 }
2616 leaf = path->nodes[0];
2617 slot = path->slots[0];
2618
2619 btrfs_item_key_to_cpu(leaf, &key, slot);
2620 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2621 key.objectid != btrfs_ino(src))
2622 break;
2623
2624 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2625 struct btrfs_file_extent_item *extent;
2626 int type;
2627 u32 size;
2628 struct btrfs_key new_key;
2629 u64 disko = 0, diskl = 0;
2630 u64 datao = 0, datal = 0;
2631 u8 comp;
2632 u64 endoff;
2633
2634 size = btrfs_item_size_nr(leaf, slot);
2635 read_extent_buffer(leaf, buf,
2636 btrfs_item_ptr_offset(leaf, slot),
2637 size);
2638
2639 extent = btrfs_item_ptr(leaf, slot,
2640 struct btrfs_file_extent_item);
2641 comp = btrfs_file_extent_compression(leaf, extent);
2642 type = btrfs_file_extent_type(leaf, extent);
2643 if (type == BTRFS_FILE_EXTENT_REG ||
2644 type == BTRFS_FILE_EXTENT_PREALLOC) {
2645 disko = btrfs_file_extent_disk_bytenr(leaf,
2646 extent);
2647 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2648 extent);
2649 datao = btrfs_file_extent_offset(leaf, extent);
2650 datal = btrfs_file_extent_num_bytes(leaf,
2651 extent);
2652 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2653 /* take upper bound, may be compressed */
2654 datal = btrfs_file_extent_ram_bytes(leaf,
2655 extent);
2656 }
2657 btrfs_release_path(path);
2658
2659 if (key.offset + datal <= off ||
2660 key.offset >= off + len - 1)
2661 goto next;
2662
2663 memcpy(&new_key, &key, sizeof(new_key));
2664 new_key.objectid = btrfs_ino(inode);
2665 if (off <= key.offset)
2666 new_key.offset = key.offset + destoff - off;
2667 else
2668 new_key.offset = destoff;
2669
2670 /*
2671 * 1 - adjusting old extent (we may have to split it)
2672 * 1 - add new extent
2673 * 1 - inode update
2674 */
2675 trans = btrfs_start_transaction(root, 3);
2676 if (IS_ERR(trans)) {
2677 ret = PTR_ERR(trans);
2678 goto out;
2679 }
2680
2681 if (type == BTRFS_FILE_EXTENT_REG ||
2682 type == BTRFS_FILE_EXTENT_PREALLOC) {
2683 /*
2684 * a | --- range to clone ---| b
2685 * | ------------- extent ------------- |
2686 */
2687
2688 /* substract range b */
2689 if (key.offset + datal > off + len)
2690 datal = off + len - key.offset;
2691
2692 /* substract range a */
2693 if (off > key.offset) {
2694 datao += off - key.offset;
2695 datal -= off - key.offset;
2696 }
2697
2698 ret = btrfs_drop_extents(trans, root, inode,
2699 new_key.offset,
2700 new_key.offset + datal,
2701 1);
2702 if (ret) {
2703 btrfs_abort_transaction(trans, root,
2704 ret);
2705 btrfs_end_transaction(trans, root);
2706 goto out;
2707 }
2708
2709 ret = btrfs_insert_empty_item(trans, root, path,
2710 &new_key, size);
2711 if (ret) {
2712 btrfs_abort_transaction(trans, root,
2713 ret);
2714 btrfs_end_transaction(trans, root);
2715 goto out;
2716 }
2717
2718 leaf = path->nodes[0];
2719 slot = path->slots[0];
2720 write_extent_buffer(leaf, buf,
2721 btrfs_item_ptr_offset(leaf, slot),
2722 size);
2723
2724 extent = btrfs_item_ptr(leaf, slot,
2725 struct btrfs_file_extent_item);
2726
2727 /* disko == 0 means it's a hole */
2728 if (!disko)
2729 datao = 0;
2730
2731 btrfs_set_file_extent_offset(leaf, extent,
2732 datao);
2733 btrfs_set_file_extent_num_bytes(leaf, extent,
2734 datal);
2735 if (disko) {
2736 inode_add_bytes(inode, datal);
2737 ret = btrfs_inc_extent_ref(trans, root,
2738 disko, diskl, 0,
2739 root->root_key.objectid,
2740 btrfs_ino(inode),
2741 new_key.offset - datao,
2742 0);
2743 if (ret) {
2744 btrfs_abort_transaction(trans,
2745 root,
2746 ret);
2747 btrfs_end_transaction(trans,
2748 root);
2749 goto out;
2750
2751 }
2752 }
2753 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2754 u64 skip = 0;
2755 u64 trim = 0;
2756 if (off > key.offset) {
2757 skip = off - key.offset;
2758 new_key.offset += skip;
2759 }
2760
2761 if (key.offset + datal > off + len)
2762 trim = key.offset + datal - (off + len);
2763
2764 if (comp && (skip || trim)) {
2765 ret = -EINVAL;
2766 btrfs_end_transaction(trans, root);
2767 goto out;
2768 }
2769 size -= skip + trim;
2770 datal -= skip + trim;
2771
2772 ret = btrfs_drop_extents(trans, root, inode,
2773 new_key.offset,
2774 new_key.offset + datal,
2775 1);
2776 if (ret) {
2777 btrfs_abort_transaction(trans, root,
2778 ret);
2779 btrfs_end_transaction(trans, root);
2780 goto out;
2781 }
2782
2783 ret = btrfs_insert_empty_item(trans, root, path,
2784 &new_key, size);
2785 if (ret) {
2786 btrfs_abort_transaction(trans, root,
2787 ret);
2788 btrfs_end_transaction(trans, root);
2789 goto out;
2790 }
2791
2792 if (skip) {
2793 u32 start =
2794 btrfs_file_extent_calc_inline_size(0);
2795 memmove(buf+start, buf+start+skip,
2796 datal);
2797 }
2798
2799 leaf = path->nodes[0];
2800 slot = path->slots[0];
2801 write_extent_buffer(leaf, buf,
2802 btrfs_item_ptr_offset(leaf, slot),
2803 size);
2804 inode_add_bytes(inode, datal);
2805 }
2806
2807 btrfs_mark_buffer_dirty(leaf);
2808 btrfs_release_path(path);
2809
2810 inode_inc_iversion(inode);
2811 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2812
2813 /*
2814 * we round up to the block size at eof when
2815 * determining which extents to clone above,
2816 * but shouldn't round up the file size
2817 */
2818 endoff = new_key.offset + datal;
2819 if (endoff > destoff+olen)
2820 endoff = destoff+olen;
2821 if (endoff > inode->i_size)
2822 btrfs_i_size_write(inode, endoff);
2823
2824 ret = btrfs_update_inode(trans, root, inode);
2825 if (ret) {
2826 btrfs_abort_transaction(trans, root, ret);
2827 btrfs_end_transaction(trans, root);
2828 goto out;
2829 }
2830 ret = btrfs_end_transaction(trans, root);
2831 }
2832 next:
2833 btrfs_release_path(path);
2834 key.offset++;
2835 }
2836 ret = 0;
2837 out:
2838 btrfs_release_path(path);
2839 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2840 out_unlock:
2841 mutex_unlock(&src->i_mutex);
2842 mutex_unlock(&inode->i_mutex);
2843 vfree(buf);
2844 btrfs_free_path(path);
2845 out_fput:
2846 fdput(src_file);
2847 out_drop_write:
2848 mnt_drop_write_file(file);
2849 return ret;
2850 }
2851
2852 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2853 {
2854 struct btrfs_ioctl_clone_range_args args;
2855
2856 if (copy_from_user(&args, argp, sizeof(args)))
2857 return -EFAULT;
2858 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2859 args.src_length, args.dest_offset);
2860 }
2861
2862 /*
2863 * there are many ways the trans_start and trans_end ioctls can lead
2864 * to deadlocks. They should only be used by applications that
2865 * basically own the machine, and have a very in depth understanding
2866 * of all the possible deadlocks and enospc problems.
2867 */
2868 static long btrfs_ioctl_trans_start(struct file *file)
2869 {
2870 struct inode *inode = file_inode(file);
2871 struct btrfs_root *root = BTRFS_I(inode)->root;
2872 struct btrfs_trans_handle *trans;
2873 int ret;
2874
2875 ret = -EPERM;
2876 if (!capable(CAP_SYS_ADMIN))
2877 goto out;
2878
2879 ret = -EINPROGRESS;
2880 if (file->private_data)
2881 goto out;
2882
2883 ret = -EROFS;
2884 if (btrfs_root_readonly(root))
2885 goto out;
2886
2887 ret = mnt_want_write_file(file);
2888 if (ret)
2889 goto out;
2890
2891 atomic_inc(&root->fs_info->open_ioctl_trans);
2892
2893 ret = -ENOMEM;
2894 trans = btrfs_start_ioctl_transaction(root);
2895 if (IS_ERR(trans))
2896 goto out_drop;
2897
2898 file->private_data = trans;
2899 return 0;
2900
2901 out_drop:
2902 atomic_dec(&root->fs_info->open_ioctl_trans);
2903 mnt_drop_write_file(file);
2904 out:
2905 return ret;
2906 }
2907
2908 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2909 {
2910 struct inode *inode = file_inode(file);
2911 struct btrfs_root *root = BTRFS_I(inode)->root;
2912 struct btrfs_root *new_root;
2913 struct btrfs_dir_item *di;
2914 struct btrfs_trans_handle *trans;
2915 struct btrfs_path *path;
2916 struct btrfs_key location;
2917 struct btrfs_disk_key disk_key;
2918 u64 objectid = 0;
2919 u64 dir_id;
2920 int ret;
2921
2922 if (!capable(CAP_SYS_ADMIN))
2923 return -EPERM;
2924
2925 ret = mnt_want_write_file(file);
2926 if (ret)
2927 return ret;
2928
2929 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2930 ret = -EFAULT;
2931 goto out;
2932 }
2933
2934 if (!objectid)
2935 objectid = root->root_key.objectid;
2936
2937 location.objectid = objectid;
2938 location.type = BTRFS_ROOT_ITEM_KEY;
2939 location.offset = (u64)-1;
2940
2941 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2942 if (IS_ERR(new_root)) {
2943 ret = PTR_ERR(new_root);
2944 goto out;
2945 }
2946
2947 if (btrfs_root_refs(&new_root->root_item) == 0) {
2948 ret = -ENOENT;
2949 goto out;
2950 }
2951
2952 path = btrfs_alloc_path();
2953 if (!path) {
2954 ret = -ENOMEM;
2955 goto out;
2956 }
2957 path->leave_spinning = 1;
2958
2959 trans = btrfs_start_transaction(root, 1);
2960 if (IS_ERR(trans)) {
2961 btrfs_free_path(path);
2962 ret = PTR_ERR(trans);
2963 goto out;
2964 }
2965
2966 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2967 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2968 dir_id, "default", 7, 1);
2969 if (IS_ERR_OR_NULL(di)) {
2970 btrfs_free_path(path);
2971 btrfs_end_transaction(trans, root);
2972 printk(KERN_ERR "Umm, you don't have the default dir item, "
2973 "this isn't going to work\n");
2974 ret = -ENOENT;
2975 goto out;
2976 }
2977
2978 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2979 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2980 btrfs_mark_buffer_dirty(path->nodes[0]);
2981 btrfs_free_path(path);
2982
2983 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
2984 btrfs_end_transaction(trans, root);
2985 out:
2986 mnt_drop_write_file(file);
2987 return ret;
2988 }
2989
2990 void btrfs_get_block_group_info(struct list_head *groups_list,
2991 struct btrfs_ioctl_space_info *space)
2992 {
2993 struct btrfs_block_group_cache *block_group;
2994
2995 space->total_bytes = 0;
2996 space->used_bytes = 0;
2997 space->flags = 0;
2998 list_for_each_entry(block_group, groups_list, list) {
2999 space->flags = block_group->flags;
3000 space->total_bytes += block_group->key.offset;
3001 space->used_bytes +=
3002 btrfs_block_group_used(&block_group->item);
3003 }
3004 }
3005
3006 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3007 {
3008 struct btrfs_ioctl_space_args space_args;
3009 struct btrfs_ioctl_space_info space;
3010 struct btrfs_ioctl_space_info *dest;
3011 struct btrfs_ioctl_space_info *dest_orig;
3012 struct btrfs_ioctl_space_info __user *user_dest;
3013 struct btrfs_space_info *info;
3014 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3015 BTRFS_BLOCK_GROUP_SYSTEM,
3016 BTRFS_BLOCK_GROUP_METADATA,
3017 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3018 int num_types = 4;
3019 int alloc_size;
3020 int ret = 0;
3021 u64 slot_count = 0;
3022 int i, c;
3023
3024 if (copy_from_user(&space_args,
3025 (struct btrfs_ioctl_space_args __user *)arg,
3026 sizeof(space_args)))
3027 return -EFAULT;
3028
3029 for (i = 0; i < num_types; i++) {
3030 struct btrfs_space_info *tmp;
3031
3032 info = NULL;
3033 rcu_read_lock();
3034 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3035 list) {
3036 if (tmp->flags == types[i]) {
3037 info = tmp;
3038 break;
3039 }
3040 }
3041 rcu_read_unlock();
3042
3043 if (!info)
3044 continue;
3045
3046 down_read(&info->groups_sem);
3047 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3048 if (!list_empty(&info->block_groups[c]))
3049 slot_count++;
3050 }
3051 up_read(&info->groups_sem);
3052 }
3053
3054 /* space_slots == 0 means they are asking for a count */
3055 if (space_args.space_slots == 0) {
3056 space_args.total_spaces = slot_count;
3057 goto out;
3058 }
3059
3060 slot_count = min_t(u64, space_args.space_slots, slot_count);
3061
3062 alloc_size = sizeof(*dest) * slot_count;
3063
3064 /* we generally have at most 6 or so space infos, one for each raid
3065 * level. So, a whole page should be more than enough for everyone
3066 */
3067 if (alloc_size > PAGE_CACHE_SIZE)
3068 return -ENOMEM;
3069
3070 space_args.total_spaces = 0;
3071 dest = kmalloc(alloc_size, GFP_NOFS);
3072 if (!dest)
3073 return -ENOMEM;
3074 dest_orig = dest;
3075
3076 /* now we have a buffer to copy into */
3077 for (i = 0; i < num_types; i++) {
3078 struct btrfs_space_info *tmp;
3079
3080 if (!slot_count)
3081 break;
3082
3083 info = NULL;
3084 rcu_read_lock();
3085 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3086 list) {
3087 if (tmp->flags == types[i]) {
3088 info = tmp;
3089 break;
3090 }
3091 }
3092 rcu_read_unlock();
3093
3094 if (!info)
3095 continue;
3096 down_read(&info->groups_sem);
3097 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3098 if (!list_empty(&info->block_groups[c])) {
3099 btrfs_get_block_group_info(
3100 &info->block_groups[c], &space);
3101 memcpy(dest, &space, sizeof(space));
3102 dest++;
3103 space_args.total_spaces++;
3104 slot_count--;
3105 }
3106 if (!slot_count)
3107 break;
3108 }
3109 up_read(&info->groups_sem);
3110 }
3111
3112 user_dest = (struct btrfs_ioctl_space_info __user *)
3113 (arg + sizeof(struct btrfs_ioctl_space_args));
3114
3115 if (copy_to_user(user_dest, dest_orig, alloc_size))
3116 ret = -EFAULT;
3117
3118 kfree(dest_orig);
3119 out:
3120 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3121 ret = -EFAULT;
3122
3123 return ret;
3124 }
3125
3126 /*
3127 * there are many ways the trans_start and trans_end ioctls can lead
3128 * to deadlocks. They should only be used by applications that
3129 * basically own the machine, and have a very in depth understanding
3130 * of all the possible deadlocks and enospc problems.
3131 */
3132 long btrfs_ioctl_trans_end(struct file *file)
3133 {
3134 struct inode *inode = file_inode(file);
3135 struct btrfs_root *root = BTRFS_I(inode)->root;
3136 struct btrfs_trans_handle *trans;
3137
3138 trans = file->private_data;
3139 if (!trans)
3140 return -EINVAL;
3141 file->private_data = NULL;
3142
3143 btrfs_end_transaction(trans, root);
3144
3145 atomic_dec(&root->fs_info->open_ioctl_trans);
3146
3147 mnt_drop_write_file(file);
3148 return 0;
3149 }
3150
3151 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3152 void __user *argp)
3153 {
3154 struct btrfs_trans_handle *trans;
3155 u64 transid;
3156 int ret;
3157
3158 trans = btrfs_attach_transaction_barrier(root);
3159 if (IS_ERR(trans)) {
3160 if (PTR_ERR(trans) != -ENOENT)
3161 return PTR_ERR(trans);
3162
3163 /* No running transaction, don't bother */
3164 transid = root->fs_info->last_trans_committed;
3165 goto out;
3166 }
3167 transid = trans->transid;
3168 ret = btrfs_commit_transaction_async(trans, root, 0);
3169 if (ret) {
3170 btrfs_end_transaction(trans, root);
3171 return ret;
3172 }
3173 out:
3174 if (argp)
3175 if (copy_to_user(argp, &transid, sizeof(transid)))
3176 return -EFAULT;
3177 return 0;
3178 }
3179
3180 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3181 void __user *argp)
3182 {
3183 u64 transid;
3184
3185 if (argp) {
3186 if (copy_from_user(&transid, argp, sizeof(transid)))
3187 return -EFAULT;
3188 } else {
3189 transid = 0; /* current trans */
3190 }
3191 return btrfs_wait_for_commit(root, transid);
3192 }
3193
3194 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3195 {
3196 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3197 struct btrfs_ioctl_scrub_args *sa;
3198 int ret;
3199
3200 if (!capable(CAP_SYS_ADMIN))
3201 return -EPERM;
3202
3203 sa = memdup_user(arg, sizeof(*sa));
3204 if (IS_ERR(sa))
3205 return PTR_ERR(sa);
3206
3207 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3208 ret = mnt_want_write_file(file);
3209 if (ret)
3210 goto out;
3211 }
3212
3213 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
3214 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3215 0);
3216
3217 if (copy_to_user(arg, sa, sizeof(*sa)))
3218 ret = -EFAULT;
3219
3220 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3221 mnt_drop_write_file(file);
3222 out:
3223 kfree(sa);
3224 return ret;
3225 }
3226
3227 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3228 {
3229 if (!capable(CAP_SYS_ADMIN))
3230 return -EPERM;
3231
3232 return btrfs_scrub_cancel(root->fs_info);
3233 }
3234
3235 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3236 void __user *arg)
3237 {
3238 struct btrfs_ioctl_scrub_args *sa;
3239 int ret;
3240
3241 if (!capable(CAP_SYS_ADMIN))
3242 return -EPERM;
3243
3244 sa = memdup_user(arg, sizeof(*sa));
3245 if (IS_ERR(sa))
3246 return PTR_ERR(sa);
3247
3248 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3249
3250 if (copy_to_user(arg, sa, sizeof(*sa)))
3251 ret = -EFAULT;
3252
3253 kfree(sa);
3254 return ret;
3255 }
3256
3257 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3258 void __user *arg)
3259 {
3260 struct btrfs_ioctl_get_dev_stats *sa;
3261 int ret;
3262
3263 sa = memdup_user(arg, sizeof(*sa));
3264 if (IS_ERR(sa))
3265 return PTR_ERR(sa);
3266
3267 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3268 kfree(sa);
3269 return -EPERM;
3270 }
3271
3272 ret = btrfs_get_dev_stats(root, sa);
3273
3274 if (copy_to_user(arg, sa, sizeof(*sa)))
3275 ret = -EFAULT;
3276
3277 kfree(sa);
3278 return ret;
3279 }
3280
3281 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3282 {
3283 struct btrfs_ioctl_dev_replace_args *p;
3284 int ret;
3285
3286 if (!capable(CAP_SYS_ADMIN))
3287 return -EPERM;
3288
3289 p = memdup_user(arg, sizeof(*p));
3290 if (IS_ERR(p))
3291 return PTR_ERR(p);
3292
3293 switch (p->cmd) {
3294 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3295 if (atomic_xchg(
3296 &root->fs_info->mutually_exclusive_operation_running,
3297 1)) {
3298 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3299 ret = -EINPROGRESS;
3300 } else {
3301 ret = btrfs_dev_replace_start(root, p);
3302 atomic_set(
3303 &root->fs_info->mutually_exclusive_operation_running,
3304 0);
3305 }
3306 break;
3307 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3308 btrfs_dev_replace_status(root->fs_info, p);
3309 ret = 0;
3310 break;
3311 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3312 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3313 break;
3314 default:
3315 ret = -EINVAL;
3316 break;
3317 }
3318
3319 if (copy_to_user(arg, p, sizeof(*p)))
3320 ret = -EFAULT;
3321
3322 kfree(p);
3323 return ret;
3324 }
3325
3326 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3327 {
3328 int ret = 0;
3329 int i;
3330 u64 rel_ptr;
3331 int size;
3332 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3333 struct inode_fs_paths *ipath = NULL;
3334 struct btrfs_path *path;
3335
3336 if (!capable(CAP_DAC_READ_SEARCH))
3337 return -EPERM;
3338
3339 path = btrfs_alloc_path();
3340 if (!path) {
3341 ret = -ENOMEM;
3342 goto out;
3343 }
3344
3345 ipa = memdup_user(arg, sizeof(*ipa));
3346 if (IS_ERR(ipa)) {
3347 ret = PTR_ERR(ipa);
3348 ipa = NULL;
3349 goto out;
3350 }
3351
3352 size = min_t(u32, ipa->size, 4096);
3353 ipath = init_ipath(size, root, path);
3354 if (IS_ERR(ipath)) {
3355 ret = PTR_ERR(ipath);
3356 ipath = NULL;
3357 goto out;
3358 }
3359
3360 ret = paths_from_inode(ipa->inum, ipath);
3361 if (ret < 0)
3362 goto out;
3363
3364 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3365 rel_ptr = ipath->fspath->val[i] -
3366 (u64)(unsigned long)ipath->fspath->val;
3367 ipath->fspath->val[i] = rel_ptr;
3368 }
3369
3370 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3371 (void *)(unsigned long)ipath->fspath, size);
3372 if (ret) {
3373 ret = -EFAULT;
3374 goto out;
3375 }
3376
3377 out:
3378 btrfs_free_path(path);
3379 free_ipath(ipath);
3380 kfree(ipa);
3381
3382 return ret;
3383 }
3384
3385 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3386 {
3387 struct btrfs_data_container *inodes = ctx;
3388 const size_t c = 3 * sizeof(u64);
3389
3390 if (inodes->bytes_left >= c) {
3391 inodes->bytes_left -= c;
3392 inodes->val[inodes->elem_cnt] = inum;
3393 inodes->val[inodes->elem_cnt + 1] = offset;
3394 inodes->val[inodes->elem_cnt + 2] = root;
3395 inodes->elem_cnt += 3;
3396 } else {
3397 inodes->bytes_missing += c - inodes->bytes_left;
3398 inodes->bytes_left = 0;
3399 inodes->elem_missed += 3;
3400 }
3401
3402 return 0;
3403 }
3404
3405 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3406 void __user *arg)
3407 {
3408 int ret = 0;
3409 int size;
3410 struct btrfs_ioctl_logical_ino_args *loi;
3411 struct btrfs_data_container *inodes = NULL;
3412 struct btrfs_path *path = NULL;
3413
3414 if (!capable(CAP_SYS_ADMIN))
3415 return -EPERM;
3416
3417 loi = memdup_user(arg, sizeof(*loi));
3418 if (IS_ERR(loi)) {
3419 ret = PTR_ERR(loi);
3420 loi = NULL;
3421 goto out;
3422 }
3423
3424 path = btrfs_alloc_path();
3425 if (!path) {
3426 ret = -ENOMEM;
3427 goto out;
3428 }
3429
3430 size = min_t(u32, loi->size, 64 * 1024);
3431 inodes = init_data_container(size);
3432 if (IS_ERR(inodes)) {
3433 ret = PTR_ERR(inodes);
3434 inodes = NULL;
3435 goto out;
3436 }
3437
3438 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3439 build_ino_list, inodes);
3440 if (ret == -EINVAL)
3441 ret = -ENOENT;
3442 if (ret < 0)
3443 goto out;
3444
3445 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3446 (void *)(unsigned long)inodes, size);
3447 if (ret)
3448 ret = -EFAULT;
3449
3450 out:
3451 btrfs_free_path(path);
3452 vfree(inodes);
3453 kfree(loi);
3454
3455 return ret;
3456 }
3457
3458 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3459 struct btrfs_ioctl_balance_args *bargs)
3460 {
3461 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3462
3463 bargs->flags = bctl->flags;
3464
3465 if (atomic_read(&fs_info->balance_running))
3466 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3467 if (atomic_read(&fs_info->balance_pause_req))
3468 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3469 if (atomic_read(&fs_info->balance_cancel_req))
3470 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3471
3472 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3473 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3474 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3475
3476 if (lock) {
3477 spin_lock(&fs_info->balance_lock);
3478 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3479 spin_unlock(&fs_info->balance_lock);
3480 } else {
3481 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3482 }
3483 }
3484
3485 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3486 {
3487 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3488 struct btrfs_fs_info *fs_info = root->fs_info;
3489 struct btrfs_ioctl_balance_args *bargs;
3490 struct btrfs_balance_control *bctl;
3491 bool need_unlock; /* for mut. excl. ops lock */
3492 int ret;
3493
3494 if (!capable(CAP_SYS_ADMIN))
3495 return -EPERM;
3496
3497 ret = mnt_want_write_file(file);
3498 if (ret)
3499 return ret;
3500
3501 again:
3502 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3503 mutex_lock(&fs_info->volume_mutex);
3504 mutex_lock(&fs_info->balance_mutex);
3505 need_unlock = true;
3506 goto locked;
3507 }
3508
3509 /*
3510 * mut. excl. ops lock is locked. Three possibilites:
3511 * (1) some other op is running
3512 * (2) balance is running
3513 * (3) balance is paused -- special case (think resume)
3514 */
3515 mutex_lock(&fs_info->balance_mutex);
3516 if (fs_info->balance_ctl) {
3517 /* this is either (2) or (3) */
3518 if (!atomic_read(&fs_info->balance_running)) {
3519 mutex_unlock(&fs_info->balance_mutex);
3520 if (!mutex_trylock(&fs_info->volume_mutex))
3521 goto again;
3522 mutex_lock(&fs_info->balance_mutex);
3523
3524 if (fs_info->balance_ctl &&
3525 !atomic_read(&fs_info->balance_running)) {
3526 /* this is (3) */
3527 need_unlock = false;
3528 goto locked;
3529 }
3530
3531 mutex_unlock(&fs_info->balance_mutex);
3532 mutex_unlock(&fs_info->volume_mutex);
3533 goto again;
3534 } else {
3535 /* this is (2) */
3536 mutex_unlock(&fs_info->balance_mutex);
3537 ret = -EINPROGRESS;
3538 goto out;
3539 }
3540 } else {
3541 /* this is (1) */
3542 mutex_unlock(&fs_info->balance_mutex);
3543 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3544 ret = -EINVAL;
3545 goto out;
3546 }
3547
3548 locked:
3549 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
3550
3551 if (arg) {
3552 bargs = memdup_user(arg, sizeof(*bargs));
3553 if (IS_ERR(bargs)) {
3554 ret = PTR_ERR(bargs);
3555 goto out_unlock;
3556 }
3557
3558 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3559 if (!fs_info->balance_ctl) {
3560 ret = -ENOTCONN;
3561 goto out_bargs;
3562 }
3563
3564 bctl = fs_info->balance_ctl;
3565 spin_lock(&fs_info->balance_lock);
3566 bctl->flags |= BTRFS_BALANCE_RESUME;
3567 spin_unlock(&fs_info->balance_lock);
3568
3569 goto do_balance;
3570 }
3571 } else {
3572 bargs = NULL;
3573 }
3574
3575 if (fs_info->balance_ctl) {
3576 ret = -EINPROGRESS;
3577 goto out_bargs;
3578 }
3579
3580 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3581 if (!bctl) {
3582 ret = -ENOMEM;
3583 goto out_bargs;
3584 }
3585
3586 bctl->fs_info = fs_info;
3587 if (arg) {
3588 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3589 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3590 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3591
3592 bctl->flags = bargs->flags;
3593 } else {
3594 /* balance everything - no filters */
3595 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3596 }
3597
3598 do_balance:
3599 /*
3600 * Ownership of bctl and mutually_exclusive_operation_running
3601 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
3602 * or, if restriper was paused all the way until unmount, in
3603 * free_fs_info. mutually_exclusive_operation_running is
3604 * cleared in __cancel_balance.
3605 */
3606 need_unlock = false;
3607
3608 ret = btrfs_balance(bctl, bargs);
3609
3610 if (arg) {
3611 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3612 ret = -EFAULT;
3613 }
3614
3615 out_bargs:
3616 kfree(bargs);
3617 out_unlock:
3618 mutex_unlock(&fs_info->balance_mutex);
3619 mutex_unlock(&fs_info->volume_mutex);
3620 if (need_unlock)
3621 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
3622 out:
3623 mnt_drop_write_file(file);
3624 return ret;
3625 }
3626
3627 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3628 {
3629 if (!capable(CAP_SYS_ADMIN))
3630 return -EPERM;
3631
3632 switch (cmd) {
3633 case BTRFS_BALANCE_CTL_PAUSE:
3634 return btrfs_pause_balance(root->fs_info);
3635 case BTRFS_BALANCE_CTL_CANCEL:
3636 return btrfs_cancel_balance(root->fs_info);
3637 }
3638
3639 return -EINVAL;
3640 }
3641
3642 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3643 void __user *arg)
3644 {
3645 struct btrfs_fs_info *fs_info = root->fs_info;
3646 struct btrfs_ioctl_balance_args *bargs;
3647 int ret = 0;
3648
3649 if (!capable(CAP_SYS_ADMIN))
3650 return -EPERM;
3651
3652 mutex_lock(&fs_info->balance_mutex);
3653 if (!fs_info->balance_ctl) {
3654 ret = -ENOTCONN;
3655 goto out;
3656 }
3657
3658 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3659 if (!bargs) {
3660 ret = -ENOMEM;
3661 goto out;
3662 }
3663
3664 update_ioctl_balance_args(fs_info, 1, bargs);
3665
3666 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3667 ret = -EFAULT;
3668
3669 kfree(bargs);
3670 out:
3671 mutex_unlock(&fs_info->balance_mutex);
3672 return ret;
3673 }
3674
3675 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3676 {
3677 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3678 struct btrfs_ioctl_quota_ctl_args *sa;
3679 struct btrfs_trans_handle *trans = NULL;
3680 int ret;
3681 int err;
3682
3683 if (!capable(CAP_SYS_ADMIN))
3684 return -EPERM;
3685
3686 ret = mnt_want_write_file(file);
3687 if (ret)
3688 return ret;
3689
3690 sa = memdup_user(arg, sizeof(*sa));
3691 if (IS_ERR(sa)) {
3692 ret = PTR_ERR(sa);
3693 goto drop_write;
3694 }
3695
3696 if (sa->cmd != BTRFS_QUOTA_CTL_RESCAN) {
3697 trans = btrfs_start_transaction(root, 2);
3698 if (IS_ERR(trans)) {
3699 ret = PTR_ERR(trans);
3700 goto out;
3701 }
3702 }
3703
3704 switch (sa->cmd) {
3705 case BTRFS_QUOTA_CTL_ENABLE:
3706 ret = btrfs_quota_enable(trans, root->fs_info);
3707 break;
3708 case BTRFS_QUOTA_CTL_DISABLE:
3709 ret = btrfs_quota_disable(trans, root->fs_info);
3710 break;
3711 case BTRFS_QUOTA_CTL_RESCAN:
3712 ret = btrfs_quota_rescan(root->fs_info);
3713 break;
3714 default:
3715 ret = -EINVAL;
3716 break;
3717 }
3718
3719 if (copy_to_user(arg, sa, sizeof(*sa)))
3720 ret = -EFAULT;
3721
3722 if (trans) {
3723 err = btrfs_commit_transaction(trans, root);
3724 if (err && !ret)
3725 ret = err;
3726 }
3727 out:
3728 kfree(sa);
3729 drop_write:
3730 mnt_drop_write_file(file);
3731 return ret;
3732 }
3733
3734 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3735 {
3736 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3737 struct btrfs_ioctl_qgroup_assign_args *sa;
3738 struct btrfs_trans_handle *trans;
3739 int ret;
3740 int err;
3741
3742 if (!capable(CAP_SYS_ADMIN))
3743 return -EPERM;
3744
3745 ret = mnt_want_write_file(file);
3746 if (ret)
3747 return ret;
3748
3749 sa = memdup_user(arg, sizeof(*sa));
3750 if (IS_ERR(sa)) {
3751 ret = PTR_ERR(sa);
3752 goto drop_write;
3753 }
3754
3755 trans = btrfs_join_transaction(root);
3756 if (IS_ERR(trans)) {
3757 ret = PTR_ERR(trans);
3758 goto out;
3759 }
3760
3761 /* FIXME: check if the IDs really exist */
3762 if (sa->assign) {
3763 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3764 sa->src, sa->dst);
3765 } else {
3766 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3767 sa->src, sa->dst);
3768 }
3769
3770 err = btrfs_end_transaction(trans, root);
3771 if (err && !ret)
3772 ret = err;
3773
3774 out:
3775 kfree(sa);
3776 drop_write:
3777 mnt_drop_write_file(file);
3778 return ret;
3779 }
3780
3781 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3782 {
3783 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3784 struct btrfs_ioctl_qgroup_create_args *sa;
3785 struct btrfs_trans_handle *trans;
3786 int ret;
3787 int err;
3788
3789 if (!capable(CAP_SYS_ADMIN))
3790 return -EPERM;
3791
3792 ret = mnt_want_write_file(file);
3793 if (ret)
3794 return ret;
3795
3796 sa = memdup_user(arg, sizeof(*sa));
3797 if (IS_ERR(sa)) {
3798 ret = PTR_ERR(sa);
3799 goto drop_write;
3800 }
3801
3802 if (!sa->qgroupid) {
3803 ret = -EINVAL;
3804 goto out;
3805 }
3806
3807 trans = btrfs_join_transaction(root);
3808 if (IS_ERR(trans)) {
3809 ret = PTR_ERR(trans);
3810 goto out;
3811 }
3812
3813 /* FIXME: check if the IDs really exist */
3814 if (sa->create) {
3815 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3816 NULL);
3817 } else {
3818 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3819 }
3820
3821 err = btrfs_end_transaction(trans, root);
3822 if (err && !ret)
3823 ret = err;
3824
3825 out:
3826 kfree(sa);
3827 drop_write:
3828 mnt_drop_write_file(file);
3829 return ret;
3830 }
3831
3832 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3833 {
3834 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3835 struct btrfs_ioctl_qgroup_limit_args *sa;
3836 struct btrfs_trans_handle *trans;
3837 int ret;
3838 int err;
3839 u64 qgroupid;
3840
3841 if (!capable(CAP_SYS_ADMIN))
3842 return -EPERM;
3843
3844 ret = mnt_want_write_file(file);
3845 if (ret)
3846 return ret;
3847
3848 sa = memdup_user(arg, sizeof(*sa));
3849 if (IS_ERR(sa)) {
3850 ret = PTR_ERR(sa);
3851 goto drop_write;
3852 }
3853
3854 trans = btrfs_join_transaction(root);
3855 if (IS_ERR(trans)) {
3856 ret = PTR_ERR(trans);
3857 goto out;
3858 }
3859
3860 qgroupid = sa->qgroupid;
3861 if (!qgroupid) {
3862 /* take the current subvol as qgroup */
3863 qgroupid = root->root_key.objectid;
3864 }
3865
3866 /* FIXME: check if the IDs really exist */
3867 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3868
3869 err = btrfs_end_transaction(trans, root);
3870 if (err && !ret)
3871 ret = err;
3872
3873 out:
3874 kfree(sa);
3875 drop_write:
3876 mnt_drop_write_file(file);
3877 return ret;
3878 }
3879
3880 static long btrfs_ioctl_set_received_subvol(struct file *file,
3881 void __user *arg)
3882 {
3883 struct btrfs_ioctl_received_subvol_args *sa = NULL;
3884 struct inode *inode = file_inode(file);
3885 struct btrfs_root *root = BTRFS_I(inode)->root;
3886 struct btrfs_root_item *root_item = &root->root_item;
3887 struct btrfs_trans_handle *trans;
3888 struct timespec ct = CURRENT_TIME;
3889 int ret = 0;
3890
3891 ret = mnt_want_write_file(file);
3892 if (ret < 0)
3893 return ret;
3894
3895 down_write(&root->fs_info->subvol_sem);
3896
3897 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3898 ret = -EINVAL;
3899 goto out;
3900 }
3901
3902 if (btrfs_root_readonly(root)) {
3903 ret = -EROFS;
3904 goto out;
3905 }
3906
3907 if (!inode_owner_or_capable(inode)) {
3908 ret = -EACCES;
3909 goto out;
3910 }
3911
3912 sa = memdup_user(arg, sizeof(*sa));
3913 if (IS_ERR(sa)) {
3914 ret = PTR_ERR(sa);
3915 sa = NULL;
3916 goto out;
3917 }
3918
3919 trans = btrfs_start_transaction(root, 1);
3920 if (IS_ERR(trans)) {
3921 ret = PTR_ERR(trans);
3922 trans = NULL;
3923 goto out;
3924 }
3925
3926 sa->rtransid = trans->transid;
3927 sa->rtime.sec = ct.tv_sec;
3928 sa->rtime.nsec = ct.tv_nsec;
3929
3930 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3931 btrfs_set_root_stransid(root_item, sa->stransid);
3932 btrfs_set_root_rtransid(root_item, sa->rtransid);
3933 root_item->stime.sec = cpu_to_le64(sa->stime.sec);
3934 root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
3935 root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
3936 root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
3937
3938 ret = btrfs_update_root(trans, root->fs_info->tree_root,
3939 &root->root_key, &root->root_item);
3940 if (ret < 0) {
3941 btrfs_end_transaction(trans, root);
3942 trans = NULL;
3943 goto out;
3944 } else {
3945 ret = btrfs_commit_transaction(trans, root);
3946 if (ret < 0)
3947 goto out;
3948 }
3949
3950 ret = copy_to_user(arg, sa, sizeof(*sa));
3951 if (ret)
3952 ret = -EFAULT;
3953
3954 out:
3955 kfree(sa);
3956 up_write(&root->fs_info->subvol_sem);
3957 mnt_drop_write_file(file);
3958 return ret;
3959 }
3960
3961 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
3962 {
3963 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3964 const char *label = root->fs_info->super_copy->label;
3965 size_t len = strnlen(label, BTRFS_LABEL_SIZE);
3966 int ret;
3967
3968 if (len == BTRFS_LABEL_SIZE) {
3969 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
3970 --len);
3971 }
3972
3973 mutex_lock(&root->fs_info->volume_mutex);
3974 ret = copy_to_user(arg, label, len);
3975 mutex_unlock(&root->fs_info->volume_mutex);
3976
3977 return ret ? -EFAULT : 0;
3978 }
3979
3980 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
3981 {
3982 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3983 struct btrfs_super_block *super_block = root->fs_info->super_copy;
3984 struct btrfs_trans_handle *trans;
3985 char label[BTRFS_LABEL_SIZE];
3986 int ret;
3987
3988 if (!capable(CAP_SYS_ADMIN))
3989 return -EPERM;
3990
3991 if (copy_from_user(label, arg, sizeof(label)))
3992 return -EFAULT;
3993
3994 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
3995 pr_err("btrfs: unable to set label with more than %d bytes\n",
3996 BTRFS_LABEL_SIZE - 1);
3997 return -EINVAL;
3998 }
3999
4000 ret = mnt_want_write_file(file);
4001 if (ret)
4002 return ret;
4003
4004 mutex_lock(&root->fs_info->volume_mutex);
4005 trans = btrfs_start_transaction(root, 0);
4006 if (IS_ERR(trans)) {
4007 ret = PTR_ERR(trans);
4008 goto out_unlock;
4009 }
4010
4011 strcpy(super_block->label, label);
4012 ret = btrfs_end_transaction(trans, root);
4013
4014 out_unlock:
4015 mutex_unlock(&root->fs_info->volume_mutex);
4016 mnt_drop_write_file(file);
4017 return ret;
4018 }
4019
4020 long btrfs_ioctl(struct file *file, unsigned int
4021 cmd, unsigned long arg)
4022 {
4023 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4024 void __user *argp = (void __user *)arg;
4025
4026 switch (cmd) {
4027 case FS_IOC_GETFLAGS:
4028 return btrfs_ioctl_getflags(file, argp);
4029 case FS_IOC_SETFLAGS:
4030 return btrfs_ioctl_setflags(file, argp);
4031 case FS_IOC_GETVERSION:
4032 return btrfs_ioctl_getversion(file, argp);
4033 case FITRIM:
4034 return btrfs_ioctl_fitrim(file, argp);
4035 case BTRFS_IOC_SNAP_CREATE:
4036 return btrfs_ioctl_snap_create(file, argp, 0);
4037 case BTRFS_IOC_SNAP_CREATE_V2:
4038 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4039 case BTRFS_IOC_SUBVOL_CREATE:
4040 return btrfs_ioctl_snap_create(file, argp, 1);
4041 case BTRFS_IOC_SUBVOL_CREATE_V2:
4042 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4043 case BTRFS_IOC_SNAP_DESTROY:
4044 return btrfs_ioctl_snap_destroy(file, argp);
4045 case BTRFS_IOC_SUBVOL_GETFLAGS:
4046 return btrfs_ioctl_subvol_getflags(file, argp);
4047 case BTRFS_IOC_SUBVOL_SETFLAGS:
4048 return btrfs_ioctl_subvol_setflags(file, argp);
4049 case BTRFS_IOC_DEFAULT_SUBVOL:
4050 return btrfs_ioctl_default_subvol(file, argp);
4051 case BTRFS_IOC_DEFRAG:
4052 return btrfs_ioctl_defrag(file, NULL);
4053 case BTRFS_IOC_DEFRAG_RANGE:
4054 return btrfs_ioctl_defrag(file, argp);
4055 case BTRFS_IOC_RESIZE:
4056 return btrfs_ioctl_resize(file, argp);
4057 case BTRFS_IOC_ADD_DEV:
4058 return btrfs_ioctl_add_dev(root, argp);
4059 case BTRFS_IOC_RM_DEV:
4060 return btrfs_ioctl_rm_dev(file, argp);
4061 case BTRFS_IOC_FS_INFO:
4062 return btrfs_ioctl_fs_info(root, argp);
4063 case BTRFS_IOC_DEV_INFO:
4064 return btrfs_ioctl_dev_info(root, argp);
4065 case BTRFS_IOC_BALANCE:
4066 return btrfs_ioctl_balance(file, NULL);
4067 case BTRFS_IOC_CLONE:
4068 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4069 case BTRFS_IOC_CLONE_RANGE:
4070 return btrfs_ioctl_clone_range(file, argp);
4071 case BTRFS_IOC_TRANS_START:
4072 return btrfs_ioctl_trans_start(file);
4073 case BTRFS_IOC_TRANS_END:
4074 return btrfs_ioctl_trans_end(file);
4075 case BTRFS_IOC_TREE_SEARCH:
4076 return btrfs_ioctl_tree_search(file, argp);
4077 case BTRFS_IOC_INO_LOOKUP:
4078 return btrfs_ioctl_ino_lookup(file, argp);
4079 case BTRFS_IOC_INO_PATHS:
4080 return btrfs_ioctl_ino_to_path(root, argp);
4081 case BTRFS_IOC_LOGICAL_INO:
4082 return btrfs_ioctl_logical_to_ino(root, argp);
4083 case BTRFS_IOC_SPACE_INFO:
4084 return btrfs_ioctl_space_info(root, argp);
4085 case BTRFS_IOC_SYNC:
4086 btrfs_sync_fs(file->f_dentry->d_sb, 1);
4087 return 0;
4088 case BTRFS_IOC_START_SYNC:
4089 return btrfs_ioctl_start_sync(root, argp);
4090 case BTRFS_IOC_WAIT_SYNC:
4091 return btrfs_ioctl_wait_sync(root, argp);
4092 case BTRFS_IOC_SCRUB:
4093 return btrfs_ioctl_scrub(file, argp);
4094 case BTRFS_IOC_SCRUB_CANCEL:
4095 return btrfs_ioctl_scrub_cancel(root, argp);
4096 case BTRFS_IOC_SCRUB_PROGRESS:
4097 return btrfs_ioctl_scrub_progress(root, argp);
4098 case BTRFS_IOC_BALANCE_V2:
4099 return btrfs_ioctl_balance(file, argp);
4100 case BTRFS_IOC_BALANCE_CTL:
4101 return btrfs_ioctl_balance_ctl(root, arg);
4102 case BTRFS_IOC_BALANCE_PROGRESS:
4103 return btrfs_ioctl_balance_progress(root, argp);
4104 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4105 return btrfs_ioctl_set_received_subvol(file, argp);
4106 case BTRFS_IOC_SEND:
4107 return btrfs_ioctl_send(file, argp);
4108 case BTRFS_IOC_GET_DEV_STATS:
4109 return btrfs_ioctl_get_dev_stats(root, argp);
4110 case BTRFS_IOC_QUOTA_CTL:
4111 return btrfs_ioctl_quota_ctl(file, argp);
4112 case BTRFS_IOC_QGROUP_ASSIGN:
4113 return btrfs_ioctl_qgroup_assign(file, argp);
4114 case BTRFS_IOC_QGROUP_CREATE:
4115 return btrfs_ioctl_qgroup_create(file, argp);
4116 case BTRFS_IOC_QGROUP_LIMIT:
4117 return btrfs_ioctl_qgroup_limit(file, argp);
4118 case BTRFS_IOC_DEV_REPLACE:
4119 return btrfs_ioctl_dev_replace(root, argp);
4120 case BTRFS_IOC_GET_FSLABEL:
4121 return btrfs_ioctl_get_fslabel(file, argp);
4122 case BTRFS_IOC_SET_FSLABEL:
4123 return btrfs_ioctl_set_fslabel(file, argp);
4124 }
4125
4126 return -ENOTTY;
4127 }
Linux with added FPC-III support