]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - fs/btrfs/ioctl.c
projects / mirror_ubuntu-zesty-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
btrfs: tree_search, copy_to_sk: return EOVERFLOW for too small buffer
[mirror_ubuntu-zesty-kernel.git] / fs / btrfs / ioctl.c
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 <linux/uaccess.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 #include "props.h"
60 #include "sysfs.h"
61 #include "qgroup.h"
62
63 #ifdef CONFIG_64BIT
64 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
68 */
69 struct btrfs_ioctl_timespec_32 {
70 __u64 sec;
71 __u32 nsec;
72 } __attribute__ ((__packed__));
73
74 struct btrfs_ioctl_received_subvol_args_32 {
75 char uuid[BTRFS_UUID_SIZE]; /* in */
76 __u64 stransid; /* in */
77 __u64 rtransid; /* out */
78 struct btrfs_ioctl_timespec_32 stime; /* in */
79 struct btrfs_ioctl_timespec_32 rtime; /* out */
80 __u64 flags; /* in */
81 __u64 reserved[16]; /* in */
82 } __attribute__ ((__packed__));
83
84 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
86 #endif
87
88
89 static int btrfs_clone(struct inode *src, struct inode *inode,
90 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
91
92 /* Mask out flags that are inappropriate for the given type of inode. */
93 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
94 {
95 if (S_ISDIR(mode))
96 return flags;
97 else if (S_ISREG(mode))
98 return flags & ~FS_DIRSYNC_FL;
99 else
100 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
101 }
102
103 /*
104 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
105 */
106 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
107 {
108 unsigned int iflags = 0;
109
110 if (flags & BTRFS_INODE_SYNC)
111 iflags |= FS_SYNC_FL;
112 if (flags & BTRFS_INODE_IMMUTABLE)
113 iflags |= FS_IMMUTABLE_FL;
114 if (flags & BTRFS_INODE_APPEND)
115 iflags |= FS_APPEND_FL;
116 if (flags & BTRFS_INODE_NODUMP)
117 iflags |= FS_NODUMP_FL;
118 if (flags & BTRFS_INODE_NOATIME)
119 iflags |= FS_NOATIME_FL;
120 if (flags & BTRFS_INODE_DIRSYNC)
121 iflags |= FS_DIRSYNC_FL;
122 if (flags & BTRFS_INODE_NODATACOW)
123 iflags |= FS_NOCOW_FL;
124
125 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
126 iflags |= FS_COMPR_FL;
127 else if (flags & BTRFS_INODE_NOCOMPRESS)
128 iflags |= FS_NOCOMP_FL;
129
130 return iflags;
131 }
132
133 /*
134 * Update inode->i_flags based on the btrfs internal flags.
135 */
136 void btrfs_update_iflags(struct inode *inode)
137 {
138 struct btrfs_inode *ip = BTRFS_I(inode);
139
140 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
141
142 if (ip->flags & BTRFS_INODE_SYNC)
143 inode->i_flags |= S_SYNC;
144 if (ip->flags & BTRFS_INODE_IMMUTABLE)
145 inode->i_flags |= S_IMMUTABLE;
146 if (ip->flags & BTRFS_INODE_APPEND)
147 inode->i_flags |= S_APPEND;
148 if (ip->flags & BTRFS_INODE_NOATIME)
149 inode->i_flags |= S_NOATIME;
150 if (ip->flags & BTRFS_INODE_DIRSYNC)
151 inode->i_flags |= S_DIRSYNC;
152 }
153
154 /*
155 * Inherit flags from the parent inode.
156 *
157 * Currently only the compression flags and the cow flags are inherited.
158 */
159 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
160 {
161 unsigned int flags;
162
163 if (!dir)
164 return;
165
166 flags = BTRFS_I(dir)->flags;
167
168 if (flags & BTRFS_INODE_NOCOMPRESS) {
169 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
170 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
171 } else if (flags & BTRFS_INODE_COMPRESS) {
172 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
173 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
174 }
175
176 if (flags & BTRFS_INODE_NODATACOW) {
177 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
178 if (S_ISREG(inode->i_mode))
179 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
180 }
181
182 btrfs_update_iflags(inode);
183 }
184
185 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
186 {
187 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
188 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
189
190 if (copy_to_user(arg, &flags, sizeof(flags)))
191 return -EFAULT;
192 return 0;
193 }
194
195 static int check_flags(unsigned int flags)
196 {
197 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
198 FS_NOATIME_FL | FS_NODUMP_FL | \
199 FS_SYNC_FL | FS_DIRSYNC_FL | \
200 FS_NOCOMP_FL | FS_COMPR_FL |
201 FS_NOCOW_FL))
202 return -EOPNOTSUPP;
203
204 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
205 return -EINVAL;
206
207 return 0;
208 }
209
210 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
211 {
212 struct inode *inode = file_inode(file);
213 struct btrfs_inode *ip = BTRFS_I(inode);
214 struct btrfs_root *root = ip->root;
215 struct btrfs_trans_handle *trans;
216 unsigned int flags, oldflags;
217 int ret;
218 u64 ip_oldflags;
219 unsigned int i_oldflags;
220 umode_t mode;
221
222 if (!inode_owner_or_capable(inode))
223 return -EPERM;
224
225 if (btrfs_root_readonly(root))
226 return -EROFS;
227
228 if (copy_from_user(&flags, arg, sizeof(flags)))
229 return -EFAULT;
230
231 ret = check_flags(flags);
232 if (ret)
233 return ret;
234
235 ret = mnt_want_write_file(file);
236 if (ret)
237 return ret;
238
239 mutex_lock(&inode->i_mutex);
240
241 ip_oldflags = ip->flags;
242 i_oldflags = inode->i_flags;
243 mode = inode->i_mode;
244
245 flags = btrfs_mask_flags(inode->i_mode, flags);
246 oldflags = btrfs_flags_to_ioctl(ip->flags);
247 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
248 if (!capable(CAP_LINUX_IMMUTABLE)) {
249 ret = -EPERM;
250 goto out_unlock;
251 }
252 }
253
254 if (flags & FS_SYNC_FL)
255 ip->flags |= BTRFS_INODE_SYNC;
256 else
257 ip->flags &= ~BTRFS_INODE_SYNC;
258 if (flags & FS_IMMUTABLE_FL)
259 ip->flags |= BTRFS_INODE_IMMUTABLE;
260 else
261 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
262 if (flags & FS_APPEND_FL)
263 ip->flags |= BTRFS_INODE_APPEND;
264 else
265 ip->flags &= ~BTRFS_INODE_APPEND;
266 if (flags & FS_NODUMP_FL)
267 ip->flags |= BTRFS_INODE_NODUMP;
268 else
269 ip->flags &= ~BTRFS_INODE_NODUMP;
270 if (flags & FS_NOATIME_FL)
271 ip->flags |= BTRFS_INODE_NOATIME;
272 else
273 ip->flags &= ~BTRFS_INODE_NOATIME;
274 if (flags & FS_DIRSYNC_FL)
275 ip->flags |= BTRFS_INODE_DIRSYNC;
276 else
277 ip->flags &= ~BTRFS_INODE_DIRSYNC;
278 if (flags & FS_NOCOW_FL) {
279 if (S_ISREG(mode)) {
280 /*
281 * It's safe to turn csums off here, no extents exist.
282 * Otherwise we want the flag to reflect the real COW
283 * status of the file and will not set it.
284 */
285 if (inode->i_size == 0)
286 ip->flags |= BTRFS_INODE_NODATACOW
287 | BTRFS_INODE_NODATASUM;
288 } else {
289 ip->flags |= BTRFS_INODE_NODATACOW;
290 }
291 } else {
292 /*
293 * Revert back under same assuptions as above
294 */
295 if (S_ISREG(mode)) {
296 if (inode->i_size == 0)
297 ip->flags &= ~(BTRFS_INODE_NODATACOW
298 | BTRFS_INODE_NODATASUM);
299 } else {
300 ip->flags &= ~BTRFS_INODE_NODATACOW;
301 }
302 }
303
304 /*
305 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
306 * flag may be changed automatically if compression code won't make
307 * things smaller.
308 */
309 if (flags & FS_NOCOMP_FL) {
310 ip->flags &= ~BTRFS_INODE_COMPRESS;
311 ip->flags |= BTRFS_INODE_NOCOMPRESS;
312
313 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
314 if (ret && ret != -ENODATA)
315 goto out_drop;
316 } else if (flags & FS_COMPR_FL) {
317 const char *comp;
318
319 ip->flags |= BTRFS_INODE_COMPRESS;
320 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
321
322 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
323 comp = "lzo";
324 else
325 comp = "zlib";
326 ret = btrfs_set_prop(inode, "btrfs.compression",
327 comp, strlen(comp), 0);
328 if (ret)
329 goto out_drop;
330
331 } else {
332 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
333 }
334
335 trans = btrfs_start_transaction(root, 1);
336 if (IS_ERR(trans)) {
337 ret = PTR_ERR(trans);
338 goto out_drop;
339 }
340
341 btrfs_update_iflags(inode);
342 inode_inc_iversion(inode);
343 inode->i_ctime = CURRENT_TIME;
344 ret = btrfs_update_inode(trans, root, inode);
345
346 btrfs_end_transaction(trans, root);
347 out_drop:
348 if (ret) {
349 ip->flags = ip_oldflags;
350 inode->i_flags = i_oldflags;
351 }
352
353 out_unlock:
354 mutex_unlock(&inode->i_mutex);
355 mnt_drop_write_file(file);
356 return ret;
357 }
358
359 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
360 {
361 struct inode *inode = file_inode(file);
362
363 return put_user(inode->i_generation, arg);
364 }
365
366 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
367 {
368 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
369 struct btrfs_device *device;
370 struct request_queue *q;
371 struct fstrim_range range;
372 u64 minlen = ULLONG_MAX;
373 u64 num_devices = 0;
374 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
375 int ret;
376
377 if (!capable(CAP_SYS_ADMIN))
378 return -EPERM;
379
380 rcu_read_lock();
381 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
382 dev_list) {
383 if (!device->bdev)
384 continue;
385 q = bdev_get_queue(device->bdev);
386 if (blk_queue_discard(q)) {
387 num_devices++;
388 minlen = min((u64)q->limits.discard_granularity,
389 minlen);
390 }
391 }
392 rcu_read_unlock();
393
394 if (!num_devices)
395 return -EOPNOTSUPP;
396 if (copy_from_user(&range, arg, sizeof(range)))
397 return -EFAULT;
398 if (range.start > total_bytes ||
399 range.len < fs_info->sb->s_blocksize)
400 return -EINVAL;
401
402 range.len = min(range.len, total_bytes - range.start);
403 range.minlen = max(range.minlen, minlen);
404 ret = btrfs_trim_fs(fs_info->tree_root, &range);
405 if (ret < 0)
406 return ret;
407
408 if (copy_to_user(arg, &range, sizeof(range)))
409 return -EFAULT;
410
411 return 0;
412 }
413
414 int btrfs_is_empty_uuid(u8 *uuid)
415 {
416 int i;
417
418 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
419 if (uuid[i])
420 return 0;
421 }
422 return 1;
423 }
424
425 static noinline int create_subvol(struct inode *dir,
426 struct dentry *dentry,
427 char *name, int namelen,
428 u64 *async_transid,
429 struct btrfs_qgroup_inherit *inherit)
430 {
431 struct btrfs_trans_handle *trans;
432 struct btrfs_key key;
433 struct btrfs_root_item root_item;
434 struct btrfs_inode_item *inode_item;
435 struct extent_buffer *leaf;
436 struct btrfs_root *root = BTRFS_I(dir)->root;
437 struct btrfs_root *new_root;
438 struct btrfs_block_rsv block_rsv;
439 struct timespec cur_time = CURRENT_TIME;
440 struct inode *inode;
441 int ret;
442 int err;
443 u64 objectid;
444 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
445 u64 index = 0;
446 u64 qgroup_reserved;
447 uuid_le new_uuid;
448
449 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
450 if (ret)
451 return ret;
452
453 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
454 /*
455 * The same as the snapshot creation, please see the comment
456 * of create_snapshot().
457 */
458 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
459 8, &qgroup_reserved, false);
460 if (ret)
461 return ret;
462
463 trans = btrfs_start_transaction(root, 0);
464 if (IS_ERR(trans)) {
465 ret = PTR_ERR(trans);
466 btrfs_subvolume_release_metadata(root, &block_rsv,
467 qgroup_reserved);
468 return ret;
469 }
470 trans->block_rsv = &block_rsv;
471 trans->bytes_reserved = block_rsv.size;
472
473 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
474 if (ret)
475 goto fail;
476
477 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
478 0, objectid, NULL, 0, 0, 0);
479 if (IS_ERR(leaf)) {
480 ret = PTR_ERR(leaf);
481 goto fail;
482 }
483
484 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
485 btrfs_set_header_bytenr(leaf, leaf->start);
486 btrfs_set_header_generation(leaf, trans->transid);
487 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
488 btrfs_set_header_owner(leaf, objectid);
489
490 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
491 BTRFS_FSID_SIZE);
492 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
493 btrfs_header_chunk_tree_uuid(leaf),
494 BTRFS_UUID_SIZE);
495 btrfs_mark_buffer_dirty(leaf);
496
497 memset(&root_item, 0, sizeof(root_item));
498
499 inode_item = &root_item.inode;
500 btrfs_set_stack_inode_generation(inode_item, 1);
501 btrfs_set_stack_inode_size(inode_item, 3);
502 btrfs_set_stack_inode_nlink(inode_item, 1);
503 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
504 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
505
506 btrfs_set_root_flags(&root_item, 0);
507 btrfs_set_root_limit(&root_item, 0);
508 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
509
510 btrfs_set_root_bytenr(&root_item, leaf->start);
511 btrfs_set_root_generation(&root_item, trans->transid);
512 btrfs_set_root_level(&root_item, 0);
513 btrfs_set_root_refs(&root_item, 1);
514 btrfs_set_root_used(&root_item, leaf->len);
515 btrfs_set_root_last_snapshot(&root_item, 0);
516
517 btrfs_set_root_generation_v2(&root_item,
518 btrfs_root_generation(&root_item));
519 uuid_le_gen(&new_uuid);
520 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
521 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
522 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
523 root_item.ctime = root_item.otime;
524 btrfs_set_root_ctransid(&root_item, trans->transid);
525 btrfs_set_root_otransid(&root_item, trans->transid);
526
527 btrfs_tree_unlock(leaf);
528 free_extent_buffer(leaf);
529 leaf = NULL;
530
531 btrfs_set_root_dirid(&root_item, new_dirid);
532
533 key.objectid = objectid;
534 key.offset = 0;
535 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
536 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
537 &root_item);
538 if (ret)
539 goto fail;
540
541 key.offset = (u64)-1;
542 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
543 if (IS_ERR(new_root)) {
544 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
545 ret = PTR_ERR(new_root);
546 goto fail;
547 }
548
549 btrfs_record_root_in_trans(trans, new_root);
550
551 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
552 if (ret) {
553 /* We potentially lose an unused inode item here */
554 btrfs_abort_transaction(trans, root, ret);
555 goto fail;
556 }
557
558 /*
559 * insert the directory item
560 */
561 ret = btrfs_set_inode_index(dir, &index);
562 if (ret) {
563 btrfs_abort_transaction(trans, root, ret);
564 goto fail;
565 }
566
567 ret = btrfs_insert_dir_item(trans, root,
568 name, namelen, dir, &key,
569 BTRFS_FT_DIR, index);
570 if (ret) {
571 btrfs_abort_transaction(trans, root, ret);
572 goto fail;
573 }
574
575 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
576 ret = btrfs_update_inode(trans, root, dir);
577 BUG_ON(ret);
578
579 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
580 objectid, root->root_key.objectid,
581 btrfs_ino(dir), index, name, namelen);
582 BUG_ON(ret);
583
584 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
585 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
586 objectid);
587 if (ret)
588 btrfs_abort_transaction(trans, root, ret);
589
590 fail:
591 trans->block_rsv = NULL;
592 trans->bytes_reserved = 0;
593 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
594
595 if (async_transid) {
596 *async_transid = trans->transid;
597 err = btrfs_commit_transaction_async(trans, root, 1);
598 if (err)
599 err = btrfs_commit_transaction(trans, root);
600 } else {
601 err = btrfs_commit_transaction(trans, root);
602 }
603 if (err && !ret)
604 ret = err;
605
606 if (!ret) {
607 inode = btrfs_lookup_dentry(dir, dentry);
608 if (IS_ERR(inode))
609 return PTR_ERR(inode);
610 d_instantiate(dentry, inode);
611 }
612 return ret;
613 }
614
615 static void btrfs_wait_nocow_write(struct btrfs_root *root)
616 {
617 s64 writers;
618 DEFINE_WAIT(wait);
619
620 do {
621 prepare_to_wait(&root->subv_writers->wait, &wait,
622 TASK_UNINTERRUPTIBLE);
623
624 writers = percpu_counter_sum(&root->subv_writers->counter);
625 if (writers)
626 schedule();
627
628 finish_wait(&root->subv_writers->wait, &wait);
629 } while (writers);
630 }
631
632 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
633 struct dentry *dentry, char *name, int namelen,
634 u64 *async_transid, bool readonly,
635 struct btrfs_qgroup_inherit *inherit)
636 {
637 struct inode *inode;
638 struct btrfs_pending_snapshot *pending_snapshot;
639 struct btrfs_trans_handle *trans;
640 int ret;
641
642 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
643 return -EINVAL;
644
645 atomic_inc(&root->will_be_snapshoted);
646 smp_mb__after_atomic_inc();
647 btrfs_wait_nocow_write(root);
648
649 ret = btrfs_start_delalloc_inodes(root, 0);
650 if (ret)
651 goto out;
652
653 btrfs_wait_ordered_extents(root, -1);
654
655 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
656 if (!pending_snapshot) {
657 ret = -ENOMEM;
658 goto out;
659 }
660
661 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
662 BTRFS_BLOCK_RSV_TEMP);
663 /*
664 * 1 - parent dir inode
665 * 2 - dir entries
666 * 1 - root item
667 * 2 - root ref/backref
668 * 1 - root of snapshot
669 * 1 - UUID item
670 */
671 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
672 &pending_snapshot->block_rsv, 8,
673 &pending_snapshot->qgroup_reserved,
674 false);
675 if (ret)
676 goto free;
677
678 pending_snapshot->dentry = dentry;
679 pending_snapshot->root = root;
680 pending_snapshot->readonly = readonly;
681 pending_snapshot->dir = dir;
682 pending_snapshot->inherit = inherit;
683
684 trans = btrfs_start_transaction(root, 0);
685 if (IS_ERR(trans)) {
686 ret = PTR_ERR(trans);
687 goto fail;
688 }
689
690 spin_lock(&root->fs_info->trans_lock);
691 list_add(&pending_snapshot->list,
692 &trans->transaction->pending_snapshots);
693 spin_unlock(&root->fs_info->trans_lock);
694 if (async_transid) {
695 *async_transid = trans->transid;
696 ret = btrfs_commit_transaction_async(trans,
697 root->fs_info->extent_root, 1);
698 if (ret)
699 ret = btrfs_commit_transaction(trans, root);
700 } else {
701 ret = btrfs_commit_transaction(trans,
702 root->fs_info->extent_root);
703 }
704 if (ret)
705 goto fail;
706
707 ret = pending_snapshot->error;
708 if (ret)
709 goto fail;
710
711 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
712 if (ret)
713 goto fail;
714
715 /*
716 * If orphan cleanup did remove any orphans, it means the tree was
717 * modified and therefore the commit root is not the same as the
718 * current root anymore. This is a problem, because send uses the
719 * commit root and therefore can see inode items that don't exist
720 * in the current root anymore, and for example make calls to
721 * btrfs_iget, which will do tree lookups based on the current root
722 * and not on the commit root. Those lookups will fail, returning a
723 * -ESTALE error, and making send fail with that error. So make sure
724 * a send does not see any orphans we have just removed, and that it
725 * will see the same inodes regardless of whether a transaction
726 * commit happened before it started (meaning that the commit root
727 * will be the same as the current root) or not.
728 */
729 if (readonly && pending_snapshot->snap->node !=
730 pending_snapshot->snap->commit_root) {
731 trans = btrfs_join_transaction(pending_snapshot->snap);
732 if (IS_ERR(trans) && PTR_ERR(trans) != -ENOENT) {
733 ret = PTR_ERR(trans);
734 goto fail;
735 }
736 if (!IS_ERR(trans)) {
737 ret = btrfs_commit_transaction(trans,
738 pending_snapshot->snap);
739 if (ret)
740 goto fail;
741 }
742 }
743
744 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
745 if (IS_ERR(inode)) {
746 ret = PTR_ERR(inode);
747 goto fail;
748 }
749
750 d_instantiate(dentry, inode);
751 ret = 0;
752 fail:
753 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
754 &pending_snapshot->block_rsv,
755 pending_snapshot->qgroup_reserved);
756 free:
757 kfree(pending_snapshot);
758 out:
759 atomic_dec(&root->will_be_snapshoted);
760 return ret;
761 }
762
763 /* copy of check_sticky in fs/namei.c()
764 * It's inline, so penalty for filesystems that don't use sticky bit is
765 * minimal.
766 */
767 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
768 {
769 kuid_t fsuid = current_fsuid();
770
771 if (!(dir->i_mode & S_ISVTX))
772 return 0;
773 if (uid_eq(inode->i_uid, fsuid))
774 return 0;
775 if (uid_eq(dir->i_uid, fsuid))
776 return 0;
777 return !capable(CAP_FOWNER);
778 }
779
780 /* copy of may_delete in fs/namei.c()
781 * Check whether we can remove a link victim from directory dir, check
782 * whether the type of victim is right.
783 * 1. We can't do it if dir is read-only (done in permission())
784 * 2. We should have write and exec permissions on dir
785 * 3. We can't remove anything from append-only dir
786 * 4. We can't do anything with immutable dir (done in permission())
787 * 5. If the sticky bit on dir is set we should either
788 * a. be owner of dir, or
789 * b. be owner of victim, or
790 * c. have CAP_FOWNER capability
791 * 6. If the victim is append-only or immutable we can't do antyhing with
792 * links pointing to it.
793 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
794 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
795 * 9. We can't remove a root or mountpoint.
796 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
797 * nfs_async_unlink().
798 */
799
800 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
801 {
802 int error;
803
804 if (!victim->d_inode)
805 return -ENOENT;
806
807 BUG_ON(victim->d_parent->d_inode != dir);
808 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
809
810 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
811 if (error)
812 return error;
813 if (IS_APPEND(dir))
814 return -EPERM;
815 if (btrfs_check_sticky(dir, victim->d_inode)||
816 IS_APPEND(victim->d_inode)||
817 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
818 return -EPERM;
819 if (isdir) {
820 if (!S_ISDIR(victim->d_inode->i_mode))
821 return -ENOTDIR;
822 if (IS_ROOT(victim))
823 return -EBUSY;
824 } else if (S_ISDIR(victim->d_inode->i_mode))
825 return -EISDIR;
826 if (IS_DEADDIR(dir))
827 return -ENOENT;
828 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
829 return -EBUSY;
830 return 0;
831 }
832
833 /* copy of may_create in fs/namei.c() */
834 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
835 {
836 if (child->d_inode)
837 return -EEXIST;
838 if (IS_DEADDIR(dir))
839 return -ENOENT;
840 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
841 }
842
843 /*
844 * Create a new subvolume below @parent. This is largely modeled after
845 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
846 * inside this filesystem so it's quite a bit simpler.
847 */
848 static noinline int btrfs_mksubvol(struct path *parent,
849 char *name, int namelen,
850 struct btrfs_root *snap_src,
851 u64 *async_transid, bool readonly,
852 struct btrfs_qgroup_inherit *inherit)
853 {
854 struct inode *dir = parent->dentry->d_inode;
855 struct dentry *dentry;
856 int error;
857
858 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
859 if (error == -EINTR)
860 return error;
861
862 dentry = lookup_one_len(name, parent->dentry, namelen);
863 error = PTR_ERR(dentry);
864 if (IS_ERR(dentry))
865 goto out_unlock;
866
867 error = -EEXIST;
868 if (dentry->d_inode)
869 goto out_dput;
870
871 error = btrfs_may_create(dir, dentry);
872 if (error)
873 goto out_dput;
874
875 /*
876 * even if this name doesn't exist, we may get hash collisions.
877 * check for them now when we can safely fail
878 */
879 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
880 dir->i_ino, name,
881 namelen);
882 if (error)
883 goto out_dput;
884
885 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
886
887 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
888 goto out_up_read;
889
890 if (snap_src) {
891 error = create_snapshot(snap_src, dir, dentry, name, namelen,
892 async_transid, readonly, inherit);
893 } else {
894 error = create_subvol(dir, dentry, name, namelen,
895 async_transid, inherit);
896 }
897 if (!error)
898 fsnotify_mkdir(dir, dentry);
899 out_up_read:
900 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
901 out_dput:
902 dput(dentry);
903 out_unlock:
904 mutex_unlock(&dir->i_mutex);
905 return error;
906 }
907
908 /*
909 * When we're defragging a range, we don't want to kick it off again
910 * if it is really just waiting for delalloc to send it down.
911 * If we find a nice big extent or delalloc range for the bytes in the
912 * file you want to defrag, we return 0 to let you know to skip this
913 * part of the file
914 */
915 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
916 {
917 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
918 struct extent_map *em = NULL;
919 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
920 u64 end;
921
922 read_lock(&em_tree->lock);
923 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
924 read_unlock(&em_tree->lock);
925
926 if (em) {
927 end = extent_map_end(em);
928 free_extent_map(em);
929 if (end - offset > thresh)
930 return 0;
931 }
932 /* if we already have a nice delalloc here, just stop */
933 thresh /= 2;
934 end = count_range_bits(io_tree, &offset, offset + thresh,
935 thresh, EXTENT_DELALLOC, 1);
936 if (end >= thresh)
937 return 0;
938 return 1;
939 }
940
941 /*
942 * helper function to walk through a file and find extents
943 * newer than a specific transid, and smaller than thresh.
944 *
945 * This is used by the defragging code to find new and small
946 * extents
947 */
948 static int find_new_extents(struct btrfs_root *root,
949 struct inode *inode, u64 newer_than,
950 u64 *off, int thresh)
951 {
952 struct btrfs_path *path;
953 struct btrfs_key min_key;
954 struct extent_buffer *leaf;
955 struct btrfs_file_extent_item *extent;
956 int type;
957 int ret;
958 u64 ino = btrfs_ino(inode);
959
960 path = btrfs_alloc_path();
961 if (!path)
962 return -ENOMEM;
963
964 min_key.objectid = ino;
965 min_key.type = BTRFS_EXTENT_DATA_KEY;
966 min_key.offset = *off;
967
968 while (1) {
969 path->keep_locks = 1;
970 ret = btrfs_search_forward(root, &min_key, path, newer_than);
971 if (ret != 0)
972 goto none;
973 path->keep_locks = 0;
974 btrfs_unlock_up_safe(path, 1);
975 process_slot:
976 if (min_key.objectid != ino)
977 goto none;
978 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
979 goto none;
980
981 leaf = path->nodes[0];
982 extent = btrfs_item_ptr(leaf, path->slots[0],
983 struct btrfs_file_extent_item);
984
985 type = btrfs_file_extent_type(leaf, extent);
986 if (type == BTRFS_FILE_EXTENT_REG &&
987 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
988 check_defrag_in_cache(inode, min_key.offset, thresh)) {
989 *off = min_key.offset;
990 btrfs_free_path(path);
991 return 0;
992 }
993
994 path->slots[0]++;
995 if (path->slots[0] < btrfs_header_nritems(leaf)) {
996 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
997 goto process_slot;
998 }
999
1000 if (min_key.offset == (u64)-1)
1001 goto none;
1002
1003 min_key.offset++;
1004 btrfs_release_path(path);
1005 }
1006 none:
1007 btrfs_free_path(path);
1008 return -ENOENT;
1009 }
1010
1011 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1012 {
1013 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1014 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1015 struct extent_map *em;
1016 u64 len = PAGE_CACHE_SIZE;
1017
1018 /*
1019 * hopefully we have this extent in the tree already, try without
1020 * the full extent lock
1021 */
1022 read_lock(&em_tree->lock);
1023 em = lookup_extent_mapping(em_tree, start, len);
1024 read_unlock(&em_tree->lock);
1025
1026 if (!em) {
1027 struct extent_state *cached = NULL;
1028 u64 end = start + len - 1;
1029
1030 /* get the big lock and read metadata off disk */
1031 lock_extent_bits(io_tree, start, end, 0, &cached);
1032 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1033 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1034
1035 if (IS_ERR(em))
1036 return NULL;
1037 }
1038
1039 return em;
1040 }
1041
1042 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1043 {
1044 struct extent_map *next;
1045 bool ret = true;
1046
1047 /* this is the last extent */
1048 if (em->start + em->len >= i_size_read(inode))
1049 return false;
1050
1051 next = defrag_lookup_extent(inode, em->start + em->len);
1052 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE ||
1053 (em->block_start + em->block_len == next->block_start))
1054 ret = false;
1055
1056 free_extent_map(next);
1057 return ret;
1058 }
1059
1060 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
1061 u64 *last_len, u64 *skip, u64 *defrag_end,
1062 int compress)
1063 {
1064 struct extent_map *em;
1065 int ret = 1;
1066 bool next_mergeable = true;
1067
1068 /*
1069 * make sure that once we start defragging an extent, we keep on
1070 * defragging it
1071 */
1072 if (start < *defrag_end)
1073 return 1;
1074
1075 *skip = 0;
1076
1077 em = defrag_lookup_extent(inode, start);
1078 if (!em)
1079 return 0;
1080
1081 /* this will cover holes, and inline extents */
1082 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1083 ret = 0;
1084 goto out;
1085 }
1086
1087 next_mergeable = defrag_check_next_extent(inode, em);
1088
1089 /*
1090 * we hit a real extent, if it is big or the next extent is not a
1091 * real extent, don't bother defragging it
1092 */
1093 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1094 (em->len >= thresh || !next_mergeable))
1095 ret = 0;
1096 out:
1097 /*
1098 * last_len ends up being a counter of how many bytes we've defragged.
1099 * every time we choose not to defrag an extent, we reset *last_len
1100 * so that the next tiny extent will force a defrag.
1101 *
1102 * The end result of this is that tiny extents before a single big
1103 * extent will force at least part of that big extent to be defragged.
1104 */
1105 if (ret) {
1106 *defrag_end = extent_map_end(em);
1107 } else {
1108 *last_len = 0;
1109 *skip = extent_map_end(em);
1110 *defrag_end = 0;
1111 }
1112
1113 free_extent_map(em);
1114 return ret;
1115 }
1116
1117 /*
1118 * it doesn't do much good to defrag one or two pages
1119 * at a time. This pulls in a nice chunk of pages
1120 * to COW and defrag.
1121 *
1122 * It also makes sure the delalloc code has enough
1123 * dirty data to avoid making new small extents as part
1124 * of the defrag
1125 *
1126 * It's a good idea to start RA on this range
1127 * before calling this.
1128 */
1129 static int cluster_pages_for_defrag(struct inode *inode,
1130 struct page **pages,
1131 unsigned long start_index,
1132 unsigned long num_pages)
1133 {
1134 unsigned long file_end;
1135 u64 isize = i_size_read(inode);
1136 u64 page_start;
1137 u64 page_end;
1138 u64 page_cnt;
1139 int ret;
1140 int i;
1141 int i_done;
1142 struct btrfs_ordered_extent *ordered;
1143 struct extent_state *cached_state = NULL;
1144 struct extent_io_tree *tree;
1145 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1146
1147 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1148 if (!isize || start_index > file_end)
1149 return 0;
1150
1151 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1152
1153 ret = btrfs_delalloc_reserve_space(inode,
1154 page_cnt << PAGE_CACHE_SHIFT);
1155 if (ret)
1156 return ret;
1157 i_done = 0;
1158 tree = &BTRFS_I(inode)->io_tree;
1159
1160 /* step one, lock all the pages */
1161 for (i = 0; i < page_cnt; i++) {
1162 struct page *page;
1163 again:
1164 page = find_or_create_page(inode->i_mapping,
1165 start_index + i, mask);
1166 if (!page)
1167 break;
1168
1169 page_start = page_offset(page);
1170 page_end = page_start + PAGE_CACHE_SIZE - 1;
1171 while (1) {
1172 lock_extent_bits(tree, page_start, page_end,
1173 0, &cached_state);
1174 ordered = btrfs_lookup_ordered_extent(inode,
1175 page_start);
1176 unlock_extent_cached(tree, page_start, page_end,
1177 &cached_state, GFP_NOFS);
1178 if (!ordered)
1179 break;
1180
1181 unlock_page(page);
1182 btrfs_start_ordered_extent(inode, ordered, 1);
1183 btrfs_put_ordered_extent(ordered);
1184 lock_page(page);
1185 /*
1186 * we unlocked the page above, so we need check if
1187 * it was released or not.
1188 */
1189 if (page->mapping != inode->i_mapping) {
1190 unlock_page(page);
1191 page_cache_release(page);
1192 goto again;
1193 }
1194 }
1195
1196 if (!PageUptodate(page)) {
1197 btrfs_readpage(NULL, page);
1198 lock_page(page);
1199 if (!PageUptodate(page)) {
1200 unlock_page(page);
1201 page_cache_release(page);
1202 ret = -EIO;
1203 break;
1204 }
1205 }
1206
1207 if (page->mapping != inode->i_mapping) {
1208 unlock_page(page);
1209 page_cache_release(page);
1210 goto again;
1211 }
1212
1213 pages[i] = page;
1214 i_done++;
1215 }
1216 if (!i_done || ret)
1217 goto out;
1218
1219 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1220 goto out;
1221
1222 /*
1223 * so now we have a nice long stream of locked
1224 * and up to date pages, lets wait on them
1225 */
1226 for (i = 0; i < i_done; i++)
1227 wait_on_page_writeback(pages[i]);
1228
1229 page_start = page_offset(pages[0]);
1230 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1231
1232 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1233 page_start, page_end - 1, 0, &cached_state);
1234 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1235 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1236 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1237 &cached_state, GFP_NOFS);
1238
1239 if (i_done != page_cnt) {
1240 spin_lock(&BTRFS_I(inode)->lock);
1241 BTRFS_I(inode)->outstanding_extents++;
1242 spin_unlock(&BTRFS_I(inode)->lock);
1243 btrfs_delalloc_release_space(inode,
1244 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1245 }
1246
1247
1248 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1249 &cached_state, GFP_NOFS);
1250
1251 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1252 page_start, page_end - 1, &cached_state,
1253 GFP_NOFS);
1254
1255 for (i = 0; i < i_done; i++) {
1256 clear_page_dirty_for_io(pages[i]);
1257 ClearPageChecked(pages[i]);
1258 set_page_extent_mapped(pages[i]);
1259 set_page_dirty(pages[i]);
1260 unlock_page(pages[i]);
1261 page_cache_release(pages[i]);
1262 }
1263 return i_done;
1264 out:
1265 for (i = 0; i < i_done; i++) {
1266 unlock_page(pages[i]);
1267 page_cache_release(pages[i]);
1268 }
1269 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1270 return ret;
1271
1272 }
1273
1274 int btrfs_defrag_file(struct inode *inode, struct file *file,
1275 struct btrfs_ioctl_defrag_range_args *range,
1276 u64 newer_than, unsigned long max_to_defrag)
1277 {
1278 struct btrfs_root *root = BTRFS_I(inode)->root;
1279 struct file_ra_state *ra = NULL;
1280 unsigned long last_index;
1281 u64 isize = i_size_read(inode);
1282 u64 last_len = 0;
1283 u64 skip = 0;
1284 u64 defrag_end = 0;
1285 u64 newer_off = range->start;
1286 unsigned long i;
1287 unsigned long ra_index = 0;
1288 int ret;
1289 int defrag_count = 0;
1290 int compress_type = BTRFS_COMPRESS_ZLIB;
1291 int extent_thresh = range->extent_thresh;
1292 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1293 unsigned long cluster = max_cluster;
1294 u64 new_align = ~((u64)128 * 1024 - 1);
1295 struct page **pages = NULL;
1296
1297 if (isize == 0)
1298 return 0;
1299
1300 if (range->start >= isize)
1301 return -EINVAL;
1302
1303 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1304 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1305 return -EINVAL;
1306 if (range->compress_type)
1307 compress_type = range->compress_type;
1308 }
1309
1310 if (extent_thresh == 0)
1311 extent_thresh = 256 * 1024;
1312
1313 /*
1314 * if we were not given a file, allocate a readahead
1315 * context
1316 */
1317 if (!file) {
1318 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1319 if (!ra)
1320 return -ENOMEM;
1321 file_ra_state_init(ra, inode->i_mapping);
1322 } else {
1323 ra = &file->f_ra;
1324 }
1325
1326 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1327 GFP_NOFS);
1328 if (!pages) {
1329 ret = -ENOMEM;
1330 goto out_ra;
1331 }
1332
1333 /* find the last page to defrag */
1334 if (range->start + range->len > range->start) {
1335 last_index = min_t(u64, isize - 1,
1336 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1337 } else {
1338 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1339 }
1340
1341 if (newer_than) {
1342 ret = find_new_extents(root, inode, newer_than,
1343 &newer_off, 64 * 1024);
1344 if (!ret) {
1345 range->start = newer_off;
1346 /*
1347 * we always align our defrag to help keep
1348 * the extents in the file evenly spaced
1349 */
1350 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1351 } else
1352 goto out_ra;
1353 } else {
1354 i = range->start >> PAGE_CACHE_SHIFT;
1355 }
1356 if (!max_to_defrag)
1357 max_to_defrag = last_index + 1;
1358
1359 /*
1360 * make writeback starts from i, so the defrag range can be
1361 * written sequentially.
1362 */
1363 if (i < inode->i_mapping->writeback_index)
1364 inode->i_mapping->writeback_index = i;
1365
1366 while (i <= last_index && defrag_count < max_to_defrag &&
1367 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1368 PAGE_CACHE_SHIFT)) {
1369 /*
1370 * make sure we stop running if someone unmounts
1371 * the FS
1372 */
1373 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1374 break;
1375
1376 if (btrfs_defrag_cancelled(root->fs_info)) {
1377 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1378 ret = -EAGAIN;
1379 break;
1380 }
1381
1382 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1383 extent_thresh, &last_len, &skip,
1384 &defrag_end, range->flags &
1385 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1386 unsigned long next;
1387 /*
1388 * the should_defrag function tells us how much to skip
1389 * bump our counter by the suggested amount
1390 */
1391 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1392 i = max(i + 1, next);
1393 continue;
1394 }
1395
1396 if (!newer_than) {
1397 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1398 PAGE_CACHE_SHIFT) - i;
1399 cluster = min(cluster, max_cluster);
1400 } else {
1401 cluster = max_cluster;
1402 }
1403
1404 if (i + cluster > ra_index) {
1405 ra_index = max(i, ra_index);
1406 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1407 cluster);
1408 ra_index += max_cluster;
1409 }
1410
1411 mutex_lock(&inode->i_mutex);
1412 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1413 BTRFS_I(inode)->force_compress = compress_type;
1414 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1415 if (ret < 0) {
1416 mutex_unlock(&inode->i_mutex);
1417 goto out_ra;
1418 }
1419
1420 defrag_count += ret;
1421 balance_dirty_pages_ratelimited(inode->i_mapping);
1422 mutex_unlock(&inode->i_mutex);
1423
1424 if (newer_than) {
1425 if (newer_off == (u64)-1)
1426 break;
1427
1428 if (ret > 0)
1429 i += ret;
1430
1431 newer_off = max(newer_off + 1,
1432 (u64)i << PAGE_CACHE_SHIFT);
1433
1434 ret = find_new_extents(root, inode,
1435 newer_than, &newer_off,
1436 64 * 1024);
1437 if (!ret) {
1438 range->start = newer_off;
1439 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1440 } else {
1441 break;
1442 }
1443 } else {
1444 if (ret > 0) {
1445 i += ret;
1446 last_len += ret << PAGE_CACHE_SHIFT;
1447 } else {
1448 i++;
1449 last_len = 0;
1450 }
1451 }
1452 }
1453
1454 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1455 filemap_flush(inode->i_mapping);
1456 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1457 &BTRFS_I(inode)->runtime_flags))
1458 filemap_flush(inode->i_mapping);
1459 }
1460
1461 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1462 /* the filemap_flush will queue IO into the worker threads, but
1463 * we have to make sure the IO is actually started and that
1464 * ordered extents get created before we return
1465 */
1466 atomic_inc(&root->fs_info->async_submit_draining);
1467 while (atomic_read(&root->fs_info->nr_async_submits) ||
1468 atomic_read(&root->fs_info->async_delalloc_pages)) {
1469 wait_event(root->fs_info->async_submit_wait,
1470 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1471 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1472 }
1473 atomic_dec(&root->fs_info->async_submit_draining);
1474 }
1475
1476 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1477 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1478 }
1479
1480 ret = defrag_count;
1481
1482 out_ra:
1483 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1484 mutex_lock(&inode->i_mutex);
1485 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1486 mutex_unlock(&inode->i_mutex);
1487 }
1488 if (!file)
1489 kfree(ra);
1490 kfree(pages);
1491 return ret;
1492 }
1493
1494 static noinline int btrfs_ioctl_resize(struct file *file,
1495 void __user *arg)
1496 {
1497 u64 new_size;
1498 u64 old_size;
1499 u64 devid = 1;
1500 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1501 struct btrfs_ioctl_vol_args *vol_args;
1502 struct btrfs_trans_handle *trans;
1503 struct btrfs_device *device = NULL;
1504 char *sizestr;
1505 char *retptr;
1506 char *devstr = NULL;
1507 int ret = 0;
1508 int mod = 0;
1509
1510 if (!capable(CAP_SYS_ADMIN))
1511 return -EPERM;
1512
1513 ret = mnt_want_write_file(file);
1514 if (ret)
1515 return ret;
1516
1517 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1518 1)) {
1519 mnt_drop_write_file(file);
1520 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1521 }
1522
1523 mutex_lock(&root->fs_info->volume_mutex);
1524 vol_args = memdup_user(arg, sizeof(*vol_args));
1525 if (IS_ERR(vol_args)) {
1526 ret = PTR_ERR(vol_args);
1527 goto out;
1528 }
1529
1530 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1531
1532 sizestr = vol_args->name;
1533 devstr = strchr(sizestr, ':');
1534 if (devstr) {
1535 sizestr = devstr + 1;
1536 *devstr = '\0';
1537 devstr = vol_args->name;
1538 ret = kstrtoull(devstr, 10, &devid);
1539 if (ret)
1540 goto out_free;
1541 if (!devid) {
1542 ret = -EINVAL;
1543 goto out_free;
1544 }
1545 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1546 }
1547
1548 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1549 if (!device) {
1550 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1551 devid);
1552 ret = -ENODEV;
1553 goto out_free;
1554 }
1555
1556 if (!device->writeable) {
1557 btrfs_info(root->fs_info,
1558 "resizer unable to apply on readonly device %llu",
1559 devid);
1560 ret = -EPERM;
1561 goto out_free;
1562 }
1563
1564 if (!strcmp(sizestr, "max"))
1565 new_size = device->bdev->bd_inode->i_size;
1566 else {
1567 if (sizestr[0] == '-') {
1568 mod = -1;
1569 sizestr++;
1570 } else if (sizestr[0] == '+') {
1571 mod = 1;
1572 sizestr++;
1573 }
1574 new_size = memparse(sizestr, &retptr);
1575 if (*retptr != '\0' || new_size == 0) {
1576 ret = -EINVAL;
1577 goto out_free;
1578 }
1579 }
1580
1581 if (device->is_tgtdev_for_dev_replace) {
1582 ret = -EPERM;
1583 goto out_free;
1584 }
1585
1586 old_size = device->total_bytes;
1587
1588 if (mod < 0) {
1589 if (new_size > old_size) {
1590 ret = -EINVAL;
1591 goto out_free;
1592 }
1593 new_size = old_size - new_size;
1594 } else if (mod > 0) {
1595 if (new_size > ULLONG_MAX - old_size) {
1596 ret = -ERANGE;
1597 goto out_free;
1598 }
1599 new_size = old_size + new_size;
1600 }
1601
1602 if (new_size < 256 * 1024 * 1024) {
1603 ret = -EINVAL;
1604 goto out_free;
1605 }
1606 if (new_size > device->bdev->bd_inode->i_size) {
1607 ret = -EFBIG;
1608 goto out_free;
1609 }
1610
1611 do_div(new_size, root->sectorsize);
1612 new_size *= root->sectorsize;
1613
1614 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1615 rcu_str_deref(device->name), new_size);
1616
1617 if (new_size > old_size) {
1618 trans = btrfs_start_transaction(root, 0);
1619 if (IS_ERR(trans)) {
1620 ret = PTR_ERR(trans);
1621 goto out_free;
1622 }
1623 ret = btrfs_grow_device(trans, device, new_size);
1624 btrfs_commit_transaction(trans, root);
1625 } else if (new_size < old_size) {
1626 ret = btrfs_shrink_device(device, new_size);
1627 } /* equal, nothing need to do */
1628
1629 out_free:
1630 kfree(vol_args);
1631 out:
1632 mutex_unlock(&root->fs_info->volume_mutex);
1633 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1634 mnt_drop_write_file(file);
1635 return ret;
1636 }
1637
1638 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1639 char *name, unsigned long fd, int subvol,
1640 u64 *transid, bool readonly,
1641 struct btrfs_qgroup_inherit *inherit)
1642 {
1643 int namelen;
1644 int ret = 0;
1645
1646 ret = mnt_want_write_file(file);
1647 if (ret)
1648 goto out;
1649
1650 namelen = strlen(name);
1651 if (strchr(name, '/')) {
1652 ret = -EINVAL;
1653 goto out_drop_write;
1654 }
1655
1656 if (name[0] == '.' &&
1657 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1658 ret = -EEXIST;
1659 goto out_drop_write;
1660 }
1661
1662 if (subvol) {
1663 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1664 NULL, transid, readonly, inherit);
1665 } else {
1666 struct fd src = fdget(fd);
1667 struct inode *src_inode;
1668 if (!src.file) {
1669 ret = -EINVAL;
1670 goto out_drop_write;
1671 }
1672
1673 src_inode = file_inode(src.file);
1674 if (src_inode->i_sb != file_inode(file)->i_sb) {
1675 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1676 "Snapshot src from another FS");
1677 ret = -EXDEV;
1678 } else if (!inode_owner_or_capable(src_inode)) {
1679 /*
1680 * Subvolume creation is not restricted, but snapshots
1681 * are limited to own subvolumes only
1682 */
1683 ret = -EPERM;
1684 } else {
1685 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1686 BTRFS_I(src_inode)->root,
1687 transid, readonly, inherit);
1688 }
1689 fdput(src);
1690 }
1691 out_drop_write:
1692 mnt_drop_write_file(file);
1693 out:
1694 return ret;
1695 }
1696
1697 static noinline int btrfs_ioctl_snap_create(struct file *file,
1698 void __user *arg, int subvol)
1699 {
1700 struct btrfs_ioctl_vol_args *vol_args;
1701 int ret;
1702
1703 vol_args = memdup_user(arg, sizeof(*vol_args));
1704 if (IS_ERR(vol_args))
1705 return PTR_ERR(vol_args);
1706 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1707
1708 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1709 vol_args->fd, subvol,
1710 NULL, false, NULL);
1711
1712 kfree(vol_args);
1713 return ret;
1714 }
1715
1716 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1717 void __user *arg, int subvol)
1718 {
1719 struct btrfs_ioctl_vol_args_v2 *vol_args;
1720 int ret;
1721 u64 transid = 0;
1722 u64 *ptr = NULL;
1723 bool readonly = false;
1724 struct btrfs_qgroup_inherit *inherit = NULL;
1725
1726 vol_args = memdup_user(arg, sizeof(*vol_args));
1727 if (IS_ERR(vol_args))
1728 return PTR_ERR(vol_args);
1729 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1730
1731 if (vol_args->flags &
1732 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1733 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1734 ret = -EOPNOTSUPP;
1735 goto out;
1736 }
1737
1738 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1739 ptr = &transid;
1740 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1741 readonly = true;
1742 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1743 if (vol_args->size > PAGE_CACHE_SIZE) {
1744 ret = -EINVAL;
1745 goto out;
1746 }
1747 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1748 if (IS_ERR(inherit)) {
1749 ret = PTR_ERR(inherit);
1750 goto out;
1751 }
1752 }
1753
1754 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1755 vol_args->fd, subvol, ptr,
1756 readonly, inherit);
1757
1758 if (ret == 0 && ptr &&
1759 copy_to_user(arg +
1760 offsetof(struct btrfs_ioctl_vol_args_v2,
1761 transid), ptr, sizeof(*ptr)))
1762 ret = -EFAULT;
1763 out:
1764 kfree(vol_args);
1765 kfree(inherit);
1766 return ret;
1767 }
1768
1769 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1770 void __user *arg)
1771 {
1772 struct inode *inode = file_inode(file);
1773 struct btrfs_root *root = BTRFS_I(inode)->root;
1774 int ret = 0;
1775 u64 flags = 0;
1776
1777 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1778 return -EINVAL;
1779
1780 down_read(&root->fs_info->subvol_sem);
1781 if (btrfs_root_readonly(root))
1782 flags |= BTRFS_SUBVOL_RDONLY;
1783 up_read(&root->fs_info->subvol_sem);
1784
1785 if (copy_to_user(arg, &flags, sizeof(flags)))
1786 ret = -EFAULT;
1787
1788 return ret;
1789 }
1790
1791 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1792 void __user *arg)
1793 {
1794 struct inode *inode = file_inode(file);
1795 struct btrfs_root *root = BTRFS_I(inode)->root;
1796 struct btrfs_trans_handle *trans;
1797 u64 root_flags;
1798 u64 flags;
1799 int ret = 0;
1800
1801 if (!inode_owner_or_capable(inode))
1802 return -EPERM;
1803
1804 ret = mnt_want_write_file(file);
1805 if (ret)
1806 goto out;
1807
1808 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1809 ret = -EINVAL;
1810 goto out_drop_write;
1811 }
1812
1813 if (copy_from_user(&flags, arg, sizeof(flags))) {
1814 ret = -EFAULT;
1815 goto out_drop_write;
1816 }
1817
1818 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1819 ret = -EINVAL;
1820 goto out_drop_write;
1821 }
1822
1823 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1824 ret = -EOPNOTSUPP;
1825 goto out_drop_write;
1826 }
1827
1828 down_write(&root->fs_info->subvol_sem);
1829
1830 /* nothing to do */
1831 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1832 goto out_drop_sem;
1833
1834 root_flags = btrfs_root_flags(&root->root_item);
1835 if (flags & BTRFS_SUBVOL_RDONLY) {
1836 btrfs_set_root_flags(&root->root_item,
1837 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1838 } else {
1839 /*
1840 * Block RO -> RW transition if this subvolume is involved in
1841 * send
1842 */
1843 spin_lock(&root->root_item_lock);
1844 if (root->send_in_progress == 0) {
1845 btrfs_set_root_flags(&root->root_item,
1846 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1847 spin_unlock(&root->root_item_lock);
1848 } else {
1849 spin_unlock(&root->root_item_lock);
1850 btrfs_warn(root->fs_info,
1851 "Attempt to set subvolume %llu read-write during send",
1852 root->root_key.objectid);
1853 ret = -EPERM;
1854 goto out_drop_sem;
1855 }
1856 }
1857
1858 trans = btrfs_start_transaction(root, 1);
1859 if (IS_ERR(trans)) {
1860 ret = PTR_ERR(trans);
1861 goto out_reset;
1862 }
1863
1864 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1865 &root->root_key, &root->root_item);
1866
1867 btrfs_commit_transaction(trans, root);
1868 out_reset:
1869 if (ret)
1870 btrfs_set_root_flags(&root->root_item, root_flags);
1871 out_drop_sem:
1872 up_write(&root->fs_info->subvol_sem);
1873 out_drop_write:
1874 mnt_drop_write_file(file);
1875 out:
1876 return ret;
1877 }
1878
1879 /*
1880 * helper to check if the subvolume references other subvolumes
1881 */
1882 static noinline int may_destroy_subvol(struct btrfs_root *root)
1883 {
1884 struct btrfs_path *path;
1885 struct btrfs_dir_item *di;
1886 struct btrfs_key key;
1887 u64 dir_id;
1888 int ret;
1889
1890 path = btrfs_alloc_path();
1891 if (!path)
1892 return -ENOMEM;
1893
1894 /* Make sure this root isn't set as the default subvol */
1895 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1896 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1897 dir_id, "default", 7, 0);
1898 if (di && !IS_ERR(di)) {
1899 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1900 if (key.objectid == root->root_key.objectid) {
1901 ret = -EPERM;
1902 btrfs_err(root->fs_info, "deleting default subvolume "
1903 "%llu is not allowed", key.objectid);
1904 goto out;
1905 }
1906 btrfs_release_path(path);
1907 }
1908
1909 key.objectid = root->root_key.objectid;
1910 key.type = BTRFS_ROOT_REF_KEY;
1911 key.offset = (u64)-1;
1912
1913 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1914 &key, path, 0, 0);
1915 if (ret < 0)
1916 goto out;
1917 BUG_ON(ret == 0);
1918
1919 ret = 0;
1920 if (path->slots[0] > 0) {
1921 path->slots[0]--;
1922 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1923 if (key.objectid == root->root_key.objectid &&
1924 key.type == BTRFS_ROOT_REF_KEY)
1925 ret = -ENOTEMPTY;
1926 }
1927 out:
1928 btrfs_free_path(path);
1929 return ret;
1930 }
1931
1932 static noinline int key_in_sk(struct btrfs_key *key,
1933 struct btrfs_ioctl_search_key *sk)
1934 {
1935 struct btrfs_key test;
1936 int ret;
1937
1938 test.objectid = sk->min_objectid;
1939 test.type = sk->min_type;
1940 test.offset = sk->min_offset;
1941
1942 ret = btrfs_comp_cpu_keys(key, &test);
1943 if (ret < 0)
1944 return 0;
1945
1946 test.objectid = sk->max_objectid;
1947 test.type = sk->max_type;
1948 test.offset = sk->max_offset;
1949
1950 ret = btrfs_comp_cpu_keys(key, &test);
1951 if (ret > 0)
1952 return 0;
1953 return 1;
1954 }
1955
1956 static noinline int copy_to_sk(struct btrfs_root *root,
1957 struct btrfs_path *path,
1958 struct btrfs_key *key,
1959 struct btrfs_ioctl_search_key *sk,
1960 size_t buf_size,
1961 char *buf,
1962 unsigned long *sk_offset,
1963 int *num_found)
1964 {
1965 u64 found_transid;
1966 struct extent_buffer *leaf;
1967 struct btrfs_ioctl_search_header sh;
1968 unsigned long item_off;
1969 unsigned long item_len;
1970 int nritems;
1971 int i;
1972 int slot;
1973 int ret = 0;
1974
1975 leaf = path->nodes[0];
1976 slot = path->slots[0];
1977 nritems = btrfs_header_nritems(leaf);
1978
1979 if (btrfs_header_generation(leaf) > sk->max_transid) {
1980 i = nritems;
1981 goto advance_key;
1982 }
1983 found_transid = btrfs_header_generation(leaf);
1984
1985 for (i = slot; i < nritems; i++) {
1986 item_off = btrfs_item_ptr_offset(leaf, i);
1987 item_len = btrfs_item_size_nr(leaf, i);
1988
1989 btrfs_item_key_to_cpu(leaf, key, i);
1990 if (!key_in_sk(key, sk))
1991 continue;
1992
1993 if (sizeof(sh) + item_len > buf_size) {
1994 if (*num_found) {
1995 ret = 1;
1996 goto out;
1997 }
1998
1999 /*
2000 * return one empty item back for v1, which does not
2001 * handle -EOVERFLOW
2002 */
2003
2004 item_len = 0;
2005 ret = -EOVERFLOW;
2006 }
2007
2008 if (sizeof(sh) + item_len + *sk_offset > buf_size) {
2009 ret = 1;
2010 goto out;
2011 }
2012
2013 sh.objectid = key->objectid;
2014 sh.offset = key->offset;
2015 sh.type = key->type;
2016 sh.len = item_len;
2017 sh.transid = found_transid;
2018
2019 /* copy search result header */
2020 memcpy(buf + *sk_offset, &sh, sizeof(sh));
2021 *sk_offset += sizeof(sh);
2022
2023 if (item_len) {
2024 char *p = buf + *sk_offset;
2025 /* copy the item */
2026 read_extent_buffer(leaf, p,
2027 item_off, item_len);
2028 *sk_offset += item_len;
2029 }
2030 (*num_found)++;
2031
2032 if (ret) /* -EOVERFLOW from above */
2033 goto out;
2034
2035 if (*num_found >= sk->nr_items) {
2036 ret = 1;
2037 goto out;
2038 }
2039 }
2040 advance_key:
2041 ret = 0;
2042 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
2043 key->offset++;
2044 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2045 key->offset = 0;
2046 key->type++;
2047 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2048 key->offset = 0;
2049 key->type = 0;
2050 key->objectid++;
2051 } else
2052 ret = 1;
2053 out:
2054 return ret;
2055 }
2056
2057 static noinline int search_ioctl(struct inode *inode,
2058 struct btrfs_ioctl_search_key *sk,
2059 size_t buf_size,
2060 char *buf)
2061 {
2062 struct btrfs_root *root;
2063 struct btrfs_key key;
2064 struct btrfs_path *path;
2065 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2066 int ret;
2067 int num_found = 0;
2068 unsigned long sk_offset = 0;
2069
2070 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2071 return -EOVERFLOW;
2072
2073 path = btrfs_alloc_path();
2074 if (!path)
2075 return -ENOMEM;
2076
2077 if (sk->tree_id == 0) {
2078 /* search the root of the inode that was passed */
2079 root = BTRFS_I(inode)->root;
2080 } else {
2081 key.objectid = sk->tree_id;
2082 key.type = BTRFS_ROOT_ITEM_KEY;
2083 key.offset = (u64)-1;
2084 root = btrfs_read_fs_root_no_name(info, &key);
2085 if (IS_ERR(root)) {
2086 printk(KERN_ERR "BTRFS: could not find root %llu\n",
2087 sk->tree_id);
2088 btrfs_free_path(path);
2089 return -ENOENT;
2090 }
2091 }
2092
2093 key.objectid = sk->min_objectid;
2094 key.type = sk->min_type;
2095 key.offset = sk->min_offset;
2096
2097 path->keep_locks = 1;
2098
2099 while (1) {
2100 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2101 if (ret != 0) {
2102 if (ret > 0)
2103 ret = 0;
2104 goto err;
2105 }
2106 ret = copy_to_sk(root, path, &key, sk, buf_size, buf,
2107 &sk_offset, &num_found);
2108 btrfs_release_path(path);
2109 if (ret)
2110 break;
2111
2112 }
2113 if (ret > 0)
2114 ret = 0;
2115 err:
2116 sk->nr_items = num_found;
2117 btrfs_free_path(path);
2118 return ret;
2119 }
2120
2121 static noinline int btrfs_ioctl_tree_search(struct file *file,
2122 void __user *argp)
2123 {
2124 struct btrfs_ioctl_search_args *args;
2125 struct inode *inode;
2126 int ret;
2127
2128 if (!capable(CAP_SYS_ADMIN))
2129 return -EPERM;
2130
2131 args = memdup_user(argp, sizeof(*args));
2132 if (IS_ERR(args))
2133 return PTR_ERR(args);
2134
2135 inode = file_inode(file);
2136 ret = search_ioctl(inode, &args->key, sizeof(args->buf), args->buf);
2137
2138 /*
2139 * In the origin implementation an overflow is handled by returning a
2140 * search header with a len of zero, so reset ret.
2141 */
2142 if (ret == -EOVERFLOW)
2143 ret = 0;
2144
2145 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2146 ret = -EFAULT;
2147 kfree(args);
2148 return ret;
2149 }
2150
2151 /*
2152 * Search INODE_REFs to identify path name of 'dirid' directory
2153 * in a 'tree_id' tree. and sets path name to 'name'.
2154 */
2155 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2156 u64 tree_id, u64 dirid, char *name)
2157 {
2158 struct btrfs_root *root;
2159 struct btrfs_key key;
2160 char *ptr;
2161 int ret = -1;
2162 int slot;
2163 int len;
2164 int total_len = 0;
2165 struct btrfs_inode_ref *iref;
2166 struct extent_buffer *l;
2167 struct btrfs_path *path;
2168
2169 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2170 name[0]='\0';
2171 return 0;
2172 }
2173
2174 path = btrfs_alloc_path();
2175 if (!path)
2176 return -ENOMEM;
2177
2178 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2179
2180 key.objectid = tree_id;
2181 key.type = BTRFS_ROOT_ITEM_KEY;
2182 key.offset = (u64)-1;
2183 root = btrfs_read_fs_root_no_name(info, &key);
2184 if (IS_ERR(root)) {
2185 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2186 ret = -ENOENT;
2187 goto out;
2188 }
2189
2190 key.objectid = dirid;
2191 key.type = BTRFS_INODE_REF_KEY;
2192 key.offset = (u64)-1;
2193
2194 while (1) {
2195 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2196 if (ret < 0)
2197 goto out;
2198 else if (ret > 0) {
2199 ret = btrfs_previous_item(root, path, dirid,
2200 BTRFS_INODE_REF_KEY);
2201 if (ret < 0)
2202 goto out;
2203 else if (ret > 0) {
2204 ret = -ENOENT;
2205 goto out;
2206 }
2207 }
2208
2209 l = path->nodes[0];
2210 slot = path->slots[0];
2211 btrfs_item_key_to_cpu(l, &key, slot);
2212
2213 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2214 len = btrfs_inode_ref_name_len(l, iref);
2215 ptr -= len + 1;
2216 total_len += len + 1;
2217 if (ptr < name) {
2218 ret = -ENAMETOOLONG;
2219 goto out;
2220 }
2221
2222 *(ptr + len) = '/';
2223 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2224
2225 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2226 break;
2227
2228 btrfs_release_path(path);
2229 key.objectid = key.offset;
2230 key.offset = (u64)-1;
2231 dirid = key.objectid;
2232 }
2233 memmove(name, ptr, total_len);
2234 name[total_len] = '\0';
2235 ret = 0;
2236 out:
2237 btrfs_free_path(path);
2238 return ret;
2239 }
2240
2241 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2242 void __user *argp)
2243 {
2244 struct btrfs_ioctl_ino_lookup_args *args;
2245 struct inode *inode;
2246 int ret;
2247
2248 if (!capable(CAP_SYS_ADMIN))
2249 return -EPERM;
2250
2251 args = memdup_user(argp, sizeof(*args));
2252 if (IS_ERR(args))
2253 return PTR_ERR(args);
2254
2255 inode = file_inode(file);
2256
2257 if (args->treeid == 0)
2258 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2259
2260 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2261 args->treeid, args->objectid,
2262 args->name);
2263
2264 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2265 ret = -EFAULT;
2266
2267 kfree(args);
2268 return ret;
2269 }
2270
2271 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2272 void __user *arg)
2273 {
2274 struct dentry *parent = file->f_path.dentry;
2275 struct dentry *dentry;
2276 struct inode *dir = parent->d_inode;
2277 struct inode *inode;
2278 struct btrfs_root *root = BTRFS_I(dir)->root;
2279 struct btrfs_root *dest = NULL;
2280 struct btrfs_ioctl_vol_args *vol_args;
2281 struct btrfs_trans_handle *trans;
2282 struct btrfs_block_rsv block_rsv;
2283 u64 root_flags;
2284 u64 qgroup_reserved;
2285 int namelen;
2286 int ret;
2287 int err = 0;
2288
2289 vol_args = memdup_user(arg, sizeof(*vol_args));
2290 if (IS_ERR(vol_args))
2291 return PTR_ERR(vol_args);
2292
2293 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2294 namelen = strlen(vol_args->name);
2295 if (strchr(vol_args->name, '/') ||
2296 strncmp(vol_args->name, "..", namelen) == 0) {
2297 err = -EINVAL;
2298 goto out;
2299 }
2300
2301 err = mnt_want_write_file(file);
2302 if (err)
2303 goto out;
2304
2305
2306 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2307 if (err == -EINTR)
2308 goto out_drop_write;
2309 dentry = lookup_one_len(vol_args->name, parent, namelen);
2310 if (IS_ERR(dentry)) {
2311 err = PTR_ERR(dentry);
2312 goto out_unlock_dir;
2313 }
2314
2315 if (!dentry->d_inode) {
2316 err = -ENOENT;
2317 goto out_dput;
2318 }
2319
2320 inode = dentry->d_inode;
2321 dest = BTRFS_I(inode)->root;
2322 if (!capable(CAP_SYS_ADMIN)) {
2323 /*
2324 * Regular user. Only allow this with a special mount
2325 * option, when the user has write+exec access to the
2326 * subvol root, and when rmdir(2) would have been
2327 * allowed.
2328 *
2329 * Note that this is _not_ check that the subvol is
2330 * empty or doesn't contain data that we wouldn't
2331 * otherwise be able to delete.
2332 *
2333 * Users who want to delete empty subvols should try
2334 * rmdir(2).
2335 */
2336 err = -EPERM;
2337 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2338 goto out_dput;
2339
2340 /*
2341 * Do not allow deletion if the parent dir is the same
2342 * as the dir to be deleted. That means the ioctl
2343 * must be called on the dentry referencing the root
2344 * of the subvol, not a random directory contained
2345 * within it.
2346 */
2347 err = -EINVAL;
2348 if (root == dest)
2349 goto out_dput;
2350
2351 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2352 if (err)
2353 goto out_dput;
2354 }
2355
2356 /* check if subvolume may be deleted by a user */
2357 err = btrfs_may_delete(dir, dentry, 1);
2358 if (err)
2359 goto out_dput;
2360
2361 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2362 err = -EINVAL;
2363 goto out_dput;
2364 }
2365
2366 mutex_lock(&inode->i_mutex);
2367
2368 /*
2369 * Don't allow to delete a subvolume with send in progress. This is
2370 * inside the i_mutex so the error handling that has to drop the bit
2371 * again is not run concurrently.
2372 */
2373 spin_lock(&dest->root_item_lock);
2374 root_flags = btrfs_root_flags(&dest->root_item);
2375 if (dest->send_in_progress == 0) {
2376 btrfs_set_root_flags(&dest->root_item,
2377 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2378 spin_unlock(&dest->root_item_lock);
2379 } else {
2380 spin_unlock(&dest->root_item_lock);
2381 btrfs_warn(root->fs_info,
2382 "Attempt to delete subvolume %llu during send",
2383 dest->root_key.objectid);
2384 err = -EPERM;
2385 goto out_dput;
2386 }
2387
2388 err = d_invalidate(dentry);
2389 if (err)
2390 goto out_unlock;
2391
2392 down_write(&root->fs_info->subvol_sem);
2393
2394 err = may_destroy_subvol(dest);
2395 if (err)
2396 goto out_up_write;
2397
2398 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2399 /*
2400 * One for dir inode, two for dir entries, two for root
2401 * ref/backref.
2402 */
2403 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2404 5, &qgroup_reserved, true);
2405 if (err)
2406 goto out_up_write;
2407
2408 trans = btrfs_start_transaction(root, 0);
2409 if (IS_ERR(trans)) {
2410 err = PTR_ERR(trans);
2411 goto out_release;
2412 }
2413 trans->block_rsv = &block_rsv;
2414 trans->bytes_reserved = block_rsv.size;
2415
2416 ret = btrfs_unlink_subvol(trans, root, dir,
2417 dest->root_key.objectid,
2418 dentry->d_name.name,
2419 dentry->d_name.len);
2420 if (ret) {
2421 err = ret;
2422 btrfs_abort_transaction(trans, root, ret);
2423 goto out_end_trans;
2424 }
2425
2426 btrfs_record_root_in_trans(trans, dest);
2427
2428 memset(&dest->root_item.drop_progress, 0,
2429 sizeof(dest->root_item.drop_progress));
2430 dest->root_item.drop_level = 0;
2431 btrfs_set_root_refs(&dest->root_item, 0);
2432
2433 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2434 ret = btrfs_insert_orphan_item(trans,
2435 root->fs_info->tree_root,
2436 dest->root_key.objectid);
2437 if (ret) {
2438 btrfs_abort_transaction(trans, root, ret);
2439 err = ret;
2440 goto out_end_trans;
2441 }
2442 }
2443
2444 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2445 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2446 dest->root_key.objectid);
2447 if (ret && ret != -ENOENT) {
2448 btrfs_abort_transaction(trans, root, ret);
2449 err = ret;
2450 goto out_end_trans;
2451 }
2452 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2453 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2454 dest->root_item.received_uuid,
2455 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2456 dest->root_key.objectid);
2457 if (ret && ret != -ENOENT) {
2458 btrfs_abort_transaction(trans, root, ret);
2459 err = ret;
2460 goto out_end_trans;
2461 }
2462 }
2463
2464 out_end_trans:
2465 trans->block_rsv = NULL;
2466 trans->bytes_reserved = 0;
2467 ret = btrfs_end_transaction(trans, root);
2468 if (ret && !err)
2469 err = ret;
2470 inode->i_flags |= S_DEAD;
2471 out_release:
2472 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2473 out_up_write:
2474 up_write(&root->fs_info->subvol_sem);
2475 out_unlock:
2476 if (err) {
2477 spin_lock(&dest->root_item_lock);
2478 root_flags = btrfs_root_flags(&dest->root_item);
2479 btrfs_set_root_flags(&dest->root_item,
2480 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2481 spin_unlock(&dest->root_item_lock);
2482 }
2483 mutex_unlock(&inode->i_mutex);
2484 if (!err) {
2485 shrink_dcache_sb(root->fs_info->sb);
2486 btrfs_invalidate_inodes(dest);
2487 d_delete(dentry);
2488 ASSERT(dest->send_in_progress == 0);
2489
2490 /* the last ref */
2491 if (dest->cache_inode) {
2492 iput(dest->cache_inode);
2493 dest->cache_inode = NULL;
2494 }
2495 }
2496 out_dput:
2497 dput(dentry);
2498 out_unlock_dir:
2499 mutex_unlock(&dir->i_mutex);
2500 out_drop_write:
2501 mnt_drop_write_file(file);
2502 out:
2503 kfree(vol_args);
2504 return err;
2505 }
2506
2507 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2508 {
2509 struct inode *inode = file_inode(file);
2510 struct btrfs_root *root = BTRFS_I(inode)->root;
2511 struct btrfs_ioctl_defrag_range_args *range;
2512 int ret;
2513
2514 ret = mnt_want_write_file(file);
2515 if (ret)
2516 return ret;
2517
2518 if (btrfs_root_readonly(root)) {
2519 ret = -EROFS;
2520 goto out;
2521 }
2522
2523 switch (inode->i_mode & S_IFMT) {
2524 case S_IFDIR:
2525 if (!capable(CAP_SYS_ADMIN)) {
2526 ret = -EPERM;
2527 goto out;
2528 }
2529 ret = btrfs_defrag_root(root);
2530 if (ret)
2531 goto out;
2532 ret = btrfs_defrag_root(root->fs_info->extent_root);
2533 break;
2534 case S_IFREG:
2535 if (!(file->f_mode & FMODE_WRITE)) {
2536 ret = -EINVAL;
2537 goto out;
2538 }
2539
2540 range = kzalloc(sizeof(*range), GFP_KERNEL);
2541 if (!range) {
2542 ret = -ENOMEM;
2543 goto out;
2544 }
2545
2546 if (argp) {
2547 if (copy_from_user(range, argp,
2548 sizeof(*range))) {
2549 ret = -EFAULT;
2550 kfree(range);
2551 goto out;
2552 }
2553 /* compression requires us to start the IO */
2554 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2555 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2556 range->extent_thresh = (u32)-1;
2557 }
2558 } else {
2559 /* the rest are all set to zero by kzalloc */
2560 range->len = (u64)-1;
2561 }
2562 ret = btrfs_defrag_file(file_inode(file), file,
2563 range, 0, 0);
2564 if (ret > 0)
2565 ret = 0;
2566 kfree(range);
2567 break;
2568 default:
2569 ret = -EINVAL;
2570 }
2571 out:
2572 mnt_drop_write_file(file);
2573 return ret;
2574 }
2575
2576 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2577 {
2578 struct btrfs_ioctl_vol_args *vol_args;
2579 int ret;
2580
2581 if (!capable(CAP_SYS_ADMIN))
2582 return -EPERM;
2583
2584 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2585 1)) {
2586 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2587 }
2588
2589 mutex_lock(&root->fs_info->volume_mutex);
2590 vol_args = memdup_user(arg, sizeof(*vol_args));
2591 if (IS_ERR(vol_args)) {
2592 ret = PTR_ERR(vol_args);
2593 goto out;
2594 }
2595
2596 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2597 ret = btrfs_init_new_device(root, vol_args->name);
2598
2599 kfree(vol_args);
2600 out:
2601 mutex_unlock(&root->fs_info->volume_mutex);
2602 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2603 return ret;
2604 }
2605
2606 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2607 {
2608 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2609 struct btrfs_ioctl_vol_args *vol_args;
2610 int ret;
2611
2612 if (!capable(CAP_SYS_ADMIN))
2613 return -EPERM;
2614
2615 ret = mnt_want_write_file(file);
2616 if (ret)
2617 return ret;
2618
2619 vol_args = memdup_user(arg, sizeof(*vol_args));
2620 if (IS_ERR(vol_args)) {
2621 ret = PTR_ERR(vol_args);
2622 goto out;
2623 }
2624
2625 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2626
2627 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2628 1)) {
2629 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2630 goto out;
2631 }
2632
2633 mutex_lock(&root->fs_info->volume_mutex);
2634 ret = btrfs_rm_device(root, vol_args->name);
2635 mutex_unlock(&root->fs_info->volume_mutex);
2636 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2637
2638 out:
2639 kfree(vol_args);
2640 mnt_drop_write_file(file);
2641 return ret;
2642 }
2643
2644 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2645 {
2646 struct btrfs_ioctl_fs_info_args *fi_args;
2647 struct btrfs_device *device;
2648 struct btrfs_device *next;
2649 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2650 int ret = 0;
2651
2652 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2653 if (!fi_args)
2654 return -ENOMEM;
2655
2656 mutex_lock(&fs_devices->device_list_mutex);
2657 fi_args->num_devices = fs_devices->num_devices;
2658 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2659
2660 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2661 if (device->devid > fi_args->max_id)
2662 fi_args->max_id = device->devid;
2663 }
2664 mutex_unlock(&fs_devices->device_list_mutex);
2665
2666 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2667 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2668 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2669
2670 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2671 ret = -EFAULT;
2672
2673 kfree(fi_args);
2674 return ret;
2675 }
2676
2677 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2678 {
2679 struct btrfs_ioctl_dev_info_args *di_args;
2680 struct btrfs_device *dev;
2681 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2682 int ret = 0;
2683 char *s_uuid = NULL;
2684
2685 di_args = memdup_user(arg, sizeof(*di_args));
2686 if (IS_ERR(di_args))
2687 return PTR_ERR(di_args);
2688
2689 if (!btrfs_is_empty_uuid(di_args->uuid))
2690 s_uuid = di_args->uuid;
2691
2692 mutex_lock(&fs_devices->device_list_mutex);
2693 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2694
2695 if (!dev) {
2696 ret = -ENODEV;
2697 goto out;
2698 }
2699
2700 di_args->devid = dev->devid;
2701 di_args->bytes_used = dev->bytes_used;
2702 di_args->total_bytes = dev->total_bytes;
2703 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2704 if (dev->name) {
2705 struct rcu_string *name;
2706
2707 rcu_read_lock();
2708 name = rcu_dereference(dev->name);
2709 strncpy(di_args->path, name->str, sizeof(di_args->path));
2710 rcu_read_unlock();
2711 di_args->path[sizeof(di_args->path) - 1] = 0;
2712 } else {
2713 di_args->path[0] = '\0';
2714 }
2715
2716 out:
2717 mutex_unlock(&fs_devices->device_list_mutex);
2718 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2719 ret = -EFAULT;
2720
2721 kfree(di_args);
2722 return ret;
2723 }
2724
2725 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2726 {
2727 struct page *page;
2728 pgoff_t index;
2729 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2730
2731 index = off >> PAGE_CACHE_SHIFT;
2732
2733 page = grab_cache_page(inode->i_mapping, index);
2734 if (!page)
2735 return NULL;
2736
2737 if (!PageUptodate(page)) {
2738 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2739 0))
2740 return NULL;
2741 lock_page(page);
2742 if (!PageUptodate(page)) {
2743 unlock_page(page);
2744 page_cache_release(page);
2745 return NULL;
2746 }
2747 }
2748 unlock_page(page);
2749
2750 return page;
2751 }
2752
2753 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2754 {
2755 /* do any pending delalloc/csum calc on src, one way or
2756 another, and lock file content */
2757 while (1) {
2758 struct btrfs_ordered_extent *ordered;
2759 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2760 ordered = btrfs_lookup_first_ordered_extent(inode,
2761 off + len - 1);
2762 if ((!ordered ||
2763 ordered->file_offset + ordered->len <= off ||
2764 ordered->file_offset >= off + len) &&
2765 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2766 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2767 if (ordered)
2768 btrfs_put_ordered_extent(ordered);
2769 break;
2770 }
2771 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2772 if (ordered)
2773 btrfs_put_ordered_extent(ordered);
2774 btrfs_wait_ordered_range(inode, off, len);
2775 }
2776 }
2777
2778 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2779 struct inode *inode2, u64 loff2, u64 len)
2780 {
2781 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2782 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2783
2784 mutex_unlock(&inode1->i_mutex);
2785 mutex_unlock(&inode2->i_mutex);
2786 }
2787
2788 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2789 struct inode *inode2, u64 loff2, u64 len)
2790 {
2791 if (inode1 < inode2) {
2792 swap(inode1, inode2);
2793 swap(loff1, loff2);
2794 }
2795
2796 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2797 lock_extent_range(inode1, loff1, len);
2798 if (inode1 != inode2) {
2799 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2800 lock_extent_range(inode2, loff2, len);
2801 }
2802 }
2803
2804 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2805 u64 dst_loff, u64 len)
2806 {
2807 int ret = 0;
2808 struct page *src_page, *dst_page;
2809 unsigned int cmp_len = PAGE_CACHE_SIZE;
2810 void *addr, *dst_addr;
2811
2812 while (len) {
2813 if (len < PAGE_CACHE_SIZE)
2814 cmp_len = len;
2815
2816 src_page = extent_same_get_page(src, loff);
2817 if (!src_page)
2818 return -EINVAL;
2819 dst_page = extent_same_get_page(dst, dst_loff);
2820 if (!dst_page) {
2821 page_cache_release(src_page);
2822 return -EINVAL;
2823 }
2824 addr = kmap_atomic(src_page);
2825 dst_addr = kmap_atomic(dst_page);
2826
2827 flush_dcache_page(src_page);
2828 flush_dcache_page(dst_page);
2829
2830 if (memcmp(addr, dst_addr, cmp_len))
2831 ret = BTRFS_SAME_DATA_DIFFERS;
2832
2833 kunmap_atomic(addr);
2834 kunmap_atomic(dst_addr);
2835 page_cache_release(src_page);
2836 page_cache_release(dst_page);
2837
2838 if (ret)
2839 break;
2840
2841 loff += cmp_len;
2842 dst_loff += cmp_len;
2843 len -= cmp_len;
2844 }
2845
2846 return ret;
2847 }
2848
2849 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2850 {
2851 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2852
2853 if (off + len > inode->i_size || off + len < off)
2854 return -EINVAL;
2855 /* Check that we are block aligned - btrfs_clone() requires this */
2856 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2857 return -EINVAL;
2858
2859 return 0;
2860 }
2861
2862 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2863 struct inode *dst, u64 dst_loff)
2864 {
2865 int ret;
2866
2867 /*
2868 * btrfs_clone() can't handle extents in the same file
2869 * yet. Once that works, we can drop this check and replace it
2870 * with a check for the same inode, but overlapping extents.
2871 */
2872 if (src == dst)
2873 return -EINVAL;
2874
2875 btrfs_double_lock(src, loff, dst, dst_loff, len);
2876
2877 ret = extent_same_check_offsets(src, loff, len);
2878 if (ret)
2879 goto out_unlock;
2880
2881 ret = extent_same_check_offsets(dst, dst_loff, len);
2882 if (ret)
2883 goto out_unlock;
2884
2885 /* don't make the dst file partly checksummed */
2886 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2887 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2888 ret = -EINVAL;
2889 goto out_unlock;
2890 }
2891
2892 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2893 if (ret == 0)
2894 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2895
2896 out_unlock:
2897 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2898
2899 return ret;
2900 }
2901
2902 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2903
2904 static long btrfs_ioctl_file_extent_same(struct file *file,
2905 struct btrfs_ioctl_same_args __user *argp)
2906 {
2907 struct btrfs_ioctl_same_args *same;
2908 struct btrfs_ioctl_same_extent_info *info;
2909 struct inode *src = file_inode(file);
2910 u64 off;
2911 u64 len;
2912 int i;
2913 int ret;
2914 unsigned long size;
2915 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2916 bool is_admin = capable(CAP_SYS_ADMIN);
2917 u16 count;
2918
2919 if (!(file->f_mode & FMODE_READ))
2920 return -EINVAL;
2921
2922 ret = mnt_want_write_file(file);
2923 if (ret)
2924 return ret;
2925
2926 if (get_user(count, &argp->dest_count)) {
2927 ret = -EFAULT;
2928 goto out;
2929 }
2930
2931 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
2932
2933 same = memdup_user(argp, size);
2934
2935 if (IS_ERR(same)) {
2936 ret = PTR_ERR(same);
2937 goto out;
2938 }
2939
2940 off = same->logical_offset;
2941 len = same->length;
2942
2943 /*
2944 * Limit the total length we will dedupe for each operation.
2945 * This is intended to bound the total time spent in this
2946 * ioctl to something sane.
2947 */
2948 if (len > BTRFS_MAX_DEDUPE_LEN)
2949 len = BTRFS_MAX_DEDUPE_LEN;
2950
2951 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2952 /*
2953 * Btrfs does not support blocksize < page_size. As a
2954 * result, btrfs_cmp_data() won't correctly handle
2955 * this situation without an update.
2956 */
2957 ret = -EINVAL;
2958 goto out;
2959 }
2960
2961 ret = -EISDIR;
2962 if (S_ISDIR(src->i_mode))
2963 goto out;
2964
2965 ret = -EACCES;
2966 if (!S_ISREG(src->i_mode))
2967 goto out;
2968
2969 /* pre-format output fields to sane values */
2970 for (i = 0; i < count; i++) {
2971 same->info[i].bytes_deduped = 0ULL;
2972 same->info[i].status = 0;
2973 }
2974
2975 for (i = 0, info = same->info; i < count; i++, info++) {
2976 struct inode *dst;
2977 struct fd dst_file = fdget(info->fd);
2978 if (!dst_file.file) {
2979 info->status = -EBADF;
2980 continue;
2981 }
2982 dst = file_inode(dst_file.file);
2983
2984 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
2985 info->status = -EINVAL;
2986 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
2987 info->status = -EXDEV;
2988 } else if (S_ISDIR(dst->i_mode)) {
2989 info->status = -EISDIR;
2990 } else if (!S_ISREG(dst->i_mode)) {
2991 info->status = -EACCES;
2992 } else {
2993 info->status = btrfs_extent_same(src, off, len, dst,
2994 info->logical_offset);
2995 if (info->status == 0)
2996 info->bytes_deduped += len;
2997 }
2998 fdput(dst_file);
2999 }
3000
3001 ret = copy_to_user(argp, same, size);
3002 if (ret)
3003 ret = -EFAULT;
3004
3005 out:
3006 mnt_drop_write_file(file);
3007 return ret;
3008 }
3009
3010 /* Helper to check and see if this root currently has a ref on the given disk
3011 * bytenr. If it does then we need to update the quota for this root. This
3012 * doesn't do anything if quotas aren't enabled.
3013 */
3014 static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3015 u64 disko)
3016 {
3017 struct seq_list tree_mod_seq_elem = {};
3018 struct ulist *roots;
3019 struct ulist_iterator uiter;
3020 struct ulist_node *root_node = NULL;
3021 int ret;
3022
3023 if (!root->fs_info->quota_enabled)
3024 return 1;
3025
3026 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3027 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3028 tree_mod_seq_elem.seq, &roots);
3029 if (ret < 0)
3030 goto out;
3031 ret = 0;
3032 ULIST_ITER_INIT(&uiter);
3033 while ((root_node = ulist_next(roots, &uiter))) {
3034 if (root_node->val == root->objectid) {
3035 ret = 1;
3036 break;
3037 }
3038 }
3039 ulist_free(roots);
3040 out:
3041 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3042 return ret;
3043 }
3044
3045 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3046 struct inode *inode,
3047 u64 endoff,
3048 const u64 destoff,
3049 const u64 olen)
3050 {
3051 struct btrfs_root *root = BTRFS_I(inode)->root;
3052 int ret;
3053
3054 inode_inc_iversion(inode);
3055 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3056 /*
3057 * We round up to the block size at eof when determining which
3058 * extents to clone above, but shouldn't round up the file size.
3059 */
3060 if (endoff > destoff + olen)
3061 endoff = destoff + olen;
3062 if (endoff > inode->i_size)
3063 btrfs_i_size_write(inode, endoff);
3064
3065 ret = btrfs_update_inode(trans, root, inode);
3066 if (ret) {
3067 btrfs_abort_transaction(trans, root, ret);
3068 btrfs_end_transaction(trans, root);
3069 goto out;
3070 }
3071 ret = btrfs_end_transaction(trans, root);
3072 out:
3073 return ret;
3074 }
3075
3076 static void clone_update_extent_map(struct inode *inode,
3077 const struct btrfs_trans_handle *trans,
3078 const struct btrfs_path *path,
3079 struct btrfs_file_extent_item *fi,
3080 const u64 hole_offset,
3081 const u64 hole_len)
3082 {
3083 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3084 struct extent_map *em;
3085 int ret;
3086
3087 em = alloc_extent_map();
3088 if (!em) {
3089 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3090 &BTRFS_I(inode)->runtime_flags);
3091 return;
3092 }
3093
3094 if (fi) {
3095 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3096 em->generation = -1;
3097 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3098 BTRFS_FILE_EXTENT_INLINE)
3099 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3100 &BTRFS_I(inode)->runtime_flags);
3101 } else {
3102 em->start = hole_offset;
3103 em->len = hole_len;
3104 em->ram_bytes = em->len;
3105 em->orig_start = hole_offset;
3106 em->block_start = EXTENT_MAP_HOLE;
3107 em->block_len = 0;
3108 em->orig_block_len = 0;
3109 em->compress_type = BTRFS_COMPRESS_NONE;
3110 em->generation = trans->transid;
3111 }
3112
3113 while (1) {
3114 write_lock(&em_tree->lock);
3115 ret = add_extent_mapping(em_tree, em, 1);
3116 write_unlock(&em_tree->lock);
3117 if (ret != -EEXIST) {
3118 free_extent_map(em);
3119 break;
3120 }
3121 btrfs_drop_extent_cache(inode, em->start,
3122 em->start + em->len - 1, 0);
3123 }
3124
3125 if (unlikely(ret))
3126 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3127 &BTRFS_I(inode)->runtime_flags);
3128 }
3129
3130 /**
3131 * btrfs_clone() - clone a range from inode file to another
3132 *
3133 * @src: Inode to clone from
3134 * @inode: Inode to clone to
3135 * @off: Offset within source to start clone from
3136 * @olen: Original length, passed by user, of range to clone
3137 * @olen_aligned: Block-aligned value of olen, extent_same uses
3138 * identical values here
3139 * @destoff: Offset within @inode to start clone
3140 */
3141 static int btrfs_clone(struct inode *src, struct inode *inode,
3142 const u64 off, const u64 olen, const u64 olen_aligned,
3143 const u64 destoff)
3144 {
3145 struct btrfs_root *root = BTRFS_I(inode)->root;
3146 struct btrfs_path *path = NULL;
3147 struct extent_buffer *leaf;
3148 struct btrfs_trans_handle *trans;
3149 char *buf = NULL;
3150 struct btrfs_key key;
3151 u32 nritems;
3152 int slot;
3153 int ret;
3154 int no_quota;
3155 const u64 len = olen_aligned;
3156 u64 last_disko = 0;
3157 u64 last_dest_end = destoff;
3158
3159 ret = -ENOMEM;
3160 buf = vmalloc(btrfs_level_size(root, 0));
3161 if (!buf)
3162 return ret;
3163
3164 path = btrfs_alloc_path();
3165 if (!path) {
3166 vfree(buf);
3167 return ret;
3168 }
3169
3170 path->reada = 2;
3171 /* clone data */
3172 key.objectid = btrfs_ino(src);
3173 key.type = BTRFS_EXTENT_DATA_KEY;
3174 key.offset = off;
3175
3176 while (1) {
3177 /*
3178 * note the key will change type as we walk through the
3179 * tree.
3180 */
3181 path->leave_spinning = 1;
3182 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3183 0, 0);
3184 if (ret < 0)
3185 goto out;
3186 /*
3187 * First search, if no extent item that starts at offset off was
3188 * found but the previous item is an extent item, it's possible
3189 * it might overlap our target range, therefore process it.
3190 */
3191 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3192 btrfs_item_key_to_cpu(path->nodes[0], &key,
3193 path->slots[0] - 1);
3194 if (key.type == BTRFS_EXTENT_DATA_KEY)
3195 path->slots[0]--;
3196 }
3197
3198 nritems = btrfs_header_nritems(path->nodes[0]);
3199 process_slot:
3200 no_quota = 1;
3201 if (path->slots[0] >= nritems) {
3202 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3203 if (ret < 0)
3204 goto out;
3205 if (ret > 0)
3206 break;
3207 nritems = btrfs_header_nritems(path->nodes[0]);
3208 }
3209 leaf = path->nodes[0];
3210 slot = path->slots[0];
3211
3212 btrfs_item_key_to_cpu(leaf, &key, slot);
3213 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
3214 key.objectid != btrfs_ino(src))
3215 break;
3216
3217 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
3218 struct btrfs_file_extent_item *extent;
3219 int type;
3220 u32 size;
3221 struct btrfs_key new_key;
3222 u64 disko = 0, diskl = 0;
3223 u64 datao = 0, datal = 0;
3224 u8 comp;
3225 u64 drop_start;
3226
3227 extent = btrfs_item_ptr(leaf, slot,
3228 struct btrfs_file_extent_item);
3229 comp = btrfs_file_extent_compression(leaf, extent);
3230 type = btrfs_file_extent_type(leaf, extent);
3231 if (type == BTRFS_FILE_EXTENT_REG ||
3232 type == BTRFS_FILE_EXTENT_PREALLOC) {
3233 disko = btrfs_file_extent_disk_bytenr(leaf,
3234 extent);
3235 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3236 extent);
3237 datao = btrfs_file_extent_offset(leaf, extent);
3238 datal = btrfs_file_extent_num_bytes(leaf,
3239 extent);
3240 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3241 /* take upper bound, may be compressed */
3242 datal = btrfs_file_extent_ram_bytes(leaf,
3243 extent);
3244 }
3245
3246 /*
3247 * The first search might have left us at an extent
3248 * item that ends before our target range's start, can
3249 * happen if we have holes and NO_HOLES feature enabled.
3250 */
3251 if (key.offset + datal <= off) {
3252 path->slots[0]++;
3253 goto process_slot;
3254 } else if (key.offset >= off + len) {
3255 break;
3256 }
3257
3258 size = btrfs_item_size_nr(leaf, slot);
3259 read_extent_buffer(leaf, buf,
3260 btrfs_item_ptr_offset(leaf, slot),
3261 size);
3262
3263 btrfs_release_path(path);
3264 path->leave_spinning = 0;
3265
3266 memcpy(&new_key, &key, sizeof(new_key));
3267 new_key.objectid = btrfs_ino(inode);
3268 if (off <= key.offset)
3269 new_key.offset = key.offset + destoff - off;
3270 else
3271 new_key.offset = destoff;
3272
3273 /*
3274 * Deal with a hole that doesn't have an extent item
3275 * that represents it (NO_HOLES feature enabled).
3276 * This hole is either in the middle of the cloning
3277 * range or at the beginning (fully overlaps it or
3278 * partially overlaps it).
3279 */
3280 if (new_key.offset != last_dest_end)
3281 drop_start = last_dest_end;
3282 else
3283 drop_start = new_key.offset;
3284
3285 /*
3286 * 1 - adjusting old extent (we may have to split it)
3287 * 1 - add new extent
3288 * 1 - inode update
3289 */
3290 trans = btrfs_start_transaction(root, 3);
3291 if (IS_ERR(trans)) {
3292 ret = PTR_ERR(trans);
3293 goto out;
3294 }
3295
3296 if (type == BTRFS_FILE_EXTENT_REG ||
3297 type == BTRFS_FILE_EXTENT_PREALLOC) {
3298 /*
3299 * a | --- range to clone ---| b
3300 * | ------------- extent ------------- |
3301 */
3302
3303 /* subtract range b */
3304 if (key.offset + datal > off + len)
3305 datal = off + len - key.offset;
3306
3307 /* subtract range a */
3308 if (off > key.offset) {
3309 datao += off - key.offset;
3310 datal -= off - key.offset;
3311 }
3312
3313 ret = btrfs_drop_extents(trans, root, inode,
3314 drop_start,
3315 new_key.offset + datal,
3316 1);
3317 if (ret) {
3318 if (ret != -EOPNOTSUPP)
3319 btrfs_abort_transaction(trans,
3320 root, ret);
3321 btrfs_end_transaction(trans, root);
3322 goto out;
3323 }
3324
3325 ret = btrfs_insert_empty_item(trans, root, path,
3326 &new_key, size);
3327 if (ret) {
3328 btrfs_abort_transaction(trans, root,
3329 ret);
3330 btrfs_end_transaction(trans, root);
3331 goto out;
3332 }
3333
3334 leaf = path->nodes[0];
3335 slot = path->slots[0];
3336 write_extent_buffer(leaf, buf,
3337 btrfs_item_ptr_offset(leaf, slot),
3338 size);
3339
3340 extent = btrfs_item_ptr(leaf, slot,
3341 struct btrfs_file_extent_item);
3342
3343 /* disko == 0 means it's a hole */
3344 if (!disko)
3345 datao = 0;
3346
3347 btrfs_set_file_extent_offset(leaf, extent,
3348 datao);
3349 btrfs_set_file_extent_num_bytes(leaf, extent,
3350 datal);
3351
3352 /*
3353 * We need to look up the roots that point at
3354 * this bytenr and see if the new root does. If
3355 * it does not we need to make sure we update
3356 * quotas appropriately.
3357 */
3358 if (disko && root != BTRFS_I(src)->root &&
3359 disko != last_disko) {
3360 no_quota = check_ref(trans, root,
3361 disko);
3362 if (no_quota < 0) {
3363 btrfs_abort_transaction(trans,
3364 root,
3365 ret);
3366 btrfs_end_transaction(trans,
3367 root);
3368 ret = no_quota;
3369 goto out;
3370 }
3371 }
3372
3373 if (disko) {
3374 inode_add_bytes(inode, datal);
3375 ret = btrfs_inc_extent_ref(trans, root,
3376 disko, diskl, 0,
3377 root->root_key.objectid,
3378 btrfs_ino(inode),
3379 new_key.offset - datao,
3380 no_quota);
3381 if (ret) {
3382 btrfs_abort_transaction(trans,
3383 root,
3384 ret);
3385 btrfs_end_transaction(trans,
3386 root);
3387 goto out;
3388
3389 }
3390 }
3391 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3392 u64 skip = 0;
3393 u64 trim = 0;
3394 u64 aligned_end = 0;
3395
3396 if (off > key.offset) {
3397 skip = off - key.offset;
3398 new_key.offset += skip;
3399 }
3400
3401 if (key.offset + datal > off + len)
3402 trim = key.offset + datal - (off + len);
3403
3404 if (comp && (skip || trim)) {
3405 ret = -EINVAL;
3406 btrfs_end_transaction(trans, root);
3407 goto out;
3408 }
3409 size -= skip + trim;
3410 datal -= skip + trim;
3411
3412 aligned_end = ALIGN(new_key.offset + datal,
3413 root->sectorsize);
3414 ret = btrfs_drop_extents(trans, root, inode,
3415 drop_start,
3416 aligned_end,
3417 1);
3418 if (ret) {
3419 if (ret != -EOPNOTSUPP)
3420 btrfs_abort_transaction(trans,
3421 root, ret);
3422 btrfs_end_transaction(trans, root);
3423 goto out;
3424 }
3425
3426 ret = btrfs_insert_empty_item(trans, root, path,
3427 &new_key, size);
3428 if (ret) {
3429 btrfs_abort_transaction(trans, root,
3430 ret);
3431 btrfs_end_transaction(trans, root);
3432 goto out;
3433 }
3434
3435 if (skip) {
3436 u32 start =
3437 btrfs_file_extent_calc_inline_size(0);
3438 memmove(buf+start, buf+start+skip,
3439 datal);
3440 }
3441
3442 leaf = path->nodes[0];
3443 slot = path->slots[0];
3444 write_extent_buffer(leaf, buf,
3445 btrfs_item_ptr_offset(leaf, slot),
3446 size);
3447 inode_add_bytes(inode, datal);
3448 extent = btrfs_item_ptr(leaf, slot,
3449 struct btrfs_file_extent_item);
3450 }
3451
3452 /* If we have an implicit hole (NO_HOLES feature). */
3453 if (drop_start < new_key.offset)
3454 clone_update_extent_map(inode, trans,
3455 path, NULL, drop_start,
3456 new_key.offset - drop_start);
3457
3458 clone_update_extent_map(inode, trans, path,
3459 extent, 0, 0);
3460
3461 btrfs_mark_buffer_dirty(leaf);
3462 btrfs_release_path(path);
3463
3464 last_dest_end = new_key.offset + datal;
3465 ret = clone_finish_inode_update(trans, inode,
3466 last_dest_end,
3467 destoff, olen);
3468 if (ret)
3469 goto out;
3470 if (new_key.offset + datal >= destoff + len)
3471 break;
3472 }
3473 btrfs_release_path(path);
3474 key.offset++;
3475 }
3476 ret = 0;
3477
3478 if (last_dest_end < destoff + len) {
3479 /*
3480 * We have an implicit hole (NO_HOLES feature is enabled) that
3481 * fully or partially overlaps our cloning range at its end.
3482 */
3483 btrfs_release_path(path);
3484
3485 /*
3486 * 1 - remove extent(s)
3487 * 1 - inode update
3488 */
3489 trans = btrfs_start_transaction(root, 2);
3490 if (IS_ERR(trans)) {
3491 ret = PTR_ERR(trans);
3492 goto out;
3493 }
3494 ret = btrfs_drop_extents(trans, root, inode,
3495 last_dest_end, destoff + len, 1);
3496 if (ret) {
3497 if (ret != -EOPNOTSUPP)
3498 btrfs_abort_transaction(trans, root, ret);
3499 btrfs_end_transaction(trans, root);
3500 goto out;
3501 }
3502 ret = clone_finish_inode_update(trans, inode, destoff + len,
3503 destoff, olen);
3504 if (ret)
3505 goto out;
3506 clone_update_extent_map(inode, trans, path, NULL, last_dest_end,
3507 destoff + len - last_dest_end);
3508 }
3509
3510 out:
3511 btrfs_free_path(path);
3512 vfree(buf);
3513 return ret;
3514 }
3515
3516 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3517 u64 off, u64 olen, u64 destoff)
3518 {
3519 struct inode *inode = file_inode(file);
3520 struct btrfs_root *root = BTRFS_I(inode)->root;
3521 struct fd src_file;
3522 struct inode *src;
3523 int ret;
3524 u64 len = olen;
3525 u64 bs = root->fs_info->sb->s_blocksize;
3526 int same_inode = 0;
3527
3528 /*
3529 * TODO:
3530 * - split compressed inline extents. annoying: we need to
3531 * decompress into destination's address_space (the file offset
3532 * may change, so source mapping won't do), then recompress (or
3533 * otherwise reinsert) a subrange.
3534 *
3535 * - split destination inode's inline extents. The inline extents can
3536 * be either compressed or non-compressed.
3537 */
3538
3539 /* the destination must be opened for writing */
3540 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3541 return -EINVAL;
3542
3543 if (btrfs_root_readonly(root))
3544 return -EROFS;
3545
3546 ret = mnt_want_write_file(file);
3547 if (ret)
3548 return ret;
3549
3550 src_file = fdget(srcfd);
3551 if (!src_file.file) {
3552 ret = -EBADF;
3553 goto out_drop_write;
3554 }
3555
3556 ret = -EXDEV;
3557 if (src_file.file->f_path.mnt != file->f_path.mnt)
3558 goto out_fput;
3559
3560 src = file_inode(src_file.file);
3561
3562 ret = -EINVAL;
3563 if (src == inode)
3564 same_inode = 1;
3565
3566 /* the src must be open for reading */
3567 if (!(src_file.file->f_mode & FMODE_READ))
3568 goto out_fput;
3569
3570 /* don't make the dst file partly checksummed */
3571 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3572 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3573 goto out_fput;
3574
3575 ret = -EISDIR;
3576 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3577 goto out_fput;
3578
3579 ret = -EXDEV;
3580 if (src->i_sb != inode->i_sb)
3581 goto out_fput;
3582
3583 if (!same_inode) {
3584 if (inode < src) {
3585 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3586 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3587 } else {
3588 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3589 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3590 }
3591 } else {
3592 mutex_lock(&src->i_mutex);
3593 }
3594
3595 /* determine range to clone */
3596 ret = -EINVAL;
3597 if (off + len > src->i_size || off + len < off)
3598 goto out_unlock;
3599 if (len == 0)
3600 olen = len = src->i_size - off;
3601 /* if we extend to eof, continue to block boundary */
3602 if (off + len == src->i_size)
3603 len = ALIGN(src->i_size, bs) - off;
3604
3605 /* verify the end result is block aligned */
3606 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3607 !IS_ALIGNED(destoff, bs))
3608 goto out_unlock;
3609
3610 /* verify if ranges are overlapped within the same file */
3611 if (same_inode) {
3612 if (destoff + len > off && destoff < off + len)
3613 goto out_unlock;
3614 }
3615
3616 if (destoff > inode->i_size) {
3617 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3618 if (ret)
3619 goto out_unlock;
3620 }
3621
3622 /*
3623 * Lock the target range too. Right after we replace the file extent
3624 * items in the fs tree (which now point to the cloned data), we might
3625 * have a worker replace them with extent items relative to a write
3626 * operation that was issued before this clone operation (i.e. confront
3627 * with inode.c:btrfs_finish_ordered_io).
3628 */
3629 if (same_inode) {
3630 u64 lock_start = min_t(u64, off, destoff);
3631 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3632
3633 lock_extent_range(src, lock_start, lock_len);
3634 } else {
3635 lock_extent_range(src, off, len);
3636 lock_extent_range(inode, destoff, len);
3637 }
3638
3639 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3640
3641 if (same_inode) {
3642 u64 lock_start = min_t(u64, off, destoff);
3643 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3644
3645 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3646 } else {
3647 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3648 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3649 destoff + len - 1);
3650 }
3651 /*
3652 * Truncate page cache pages so that future reads will see the cloned
3653 * data immediately and not the previous data.
3654 */
3655 truncate_inode_pages_range(&inode->i_data, destoff,
3656 PAGE_CACHE_ALIGN(destoff + len) - 1);
3657 out_unlock:
3658 if (!same_inode) {
3659 if (inode < src) {
3660 mutex_unlock(&src->i_mutex);
3661 mutex_unlock(&inode->i_mutex);
3662 } else {
3663 mutex_unlock(&inode->i_mutex);
3664 mutex_unlock(&src->i_mutex);
3665 }
3666 } else {
3667 mutex_unlock(&src->i_mutex);
3668 }
3669 out_fput:
3670 fdput(src_file);
3671 out_drop_write:
3672 mnt_drop_write_file(file);
3673 return ret;
3674 }
3675
3676 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3677 {
3678 struct btrfs_ioctl_clone_range_args args;
3679
3680 if (copy_from_user(&args, argp, sizeof(args)))
3681 return -EFAULT;
3682 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3683 args.src_length, args.dest_offset);
3684 }
3685
3686 /*
3687 * there are many ways the trans_start and trans_end ioctls can lead
3688 * to deadlocks. They should only be used by applications that
3689 * basically own the machine, and have a very in depth understanding
3690 * of all the possible deadlocks and enospc problems.
3691 */
3692 static long btrfs_ioctl_trans_start(struct file *file)
3693 {
3694 struct inode *inode = file_inode(file);
3695 struct btrfs_root *root = BTRFS_I(inode)->root;
3696 struct btrfs_trans_handle *trans;
3697 int ret;
3698
3699 ret = -EPERM;
3700 if (!capable(CAP_SYS_ADMIN))
3701 goto out;
3702
3703 ret = -EINPROGRESS;
3704 if (file->private_data)
3705 goto out;
3706
3707 ret = -EROFS;
3708 if (btrfs_root_readonly(root))
3709 goto out;
3710
3711 ret = mnt_want_write_file(file);
3712 if (ret)
3713 goto out;
3714
3715 atomic_inc(&root->fs_info->open_ioctl_trans);
3716
3717 ret = -ENOMEM;
3718 trans = btrfs_start_ioctl_transaction(root);
3719 if (IS_ERR(trans))
3720 goto out_drop;
3721
3722 file->private_data = trans;
3723 return 0;
3724
3725 out_drop:
3726 atomic_dec(&root->fs_info->open_ioctl_trans);
3727 mnt_drop_write_file(file);
3728 out:
3729 return ret;
3730 }
3731
3732 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3733 {
3734 struct inode *inode = file_inode(file);
3735 struct btrfs_root *root = BTRFS_I(inode)->root;
3736 struct btrfs_root *new_root;
3737 struct btrfs_dir_item *di;
3738 struct btrfs_trans_handle *trans;
3739 struct btrfs_path *path;
3740 struct btrfs_key location;
3741 struct btrfs_disk_key disk_key;
3742 u64 objectid = 0;
3743 u64 dir_id;
3744 int ret;
3745
3746 if (!capable(CAP_SYS_ADMIN))
3747 return -EPERM;
3748
3749 ret = mnt_want_write_file(file);
3750 if (ret)
3751 return ret;
3752
3753 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3754 ret = -EFAULT;
3755 goto out;
3756 }
3757
3758 if (!objectid)
3759 objectid = BTRFS_FS_TREE_OBJECTID;
3760
3761 location.objectid = objectid;
3762 location.type = BTRFS_ROOT_ITEM_KEY;
3763 location.offset = (u64)-1;
3764
3765 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3766 if (IS_ERR(new_root)) {
3767 ret = PTR_ERR(new_root);
3768 goto out;
3769 }
3770
3771 path = btrfs_alloc_path();
3772 if (!path) {
3773 ret = -ENOMEM;
3774 goto out;
3775 }
3776 path->leave_spinning = 1;
3777
3778 trans = btrfs_start_transaction(root, 1);
3779 if (IS_ERR(trans)) {
3780 btrfs_free_path(path);
3781 ret = PTR_ERR(trans);
3782 goto out;
3783 }
3784
3785 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3786 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3787 dir_id, "default", 7, 1);
3788 if (IS_ERR_OR_NULL(di)) {
3789 btrfs_free_path(path);
3790 btrfs_end_transaction(trans, root);
3791 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3792 "item, this isn't going to work");
3793 ret = -ENOENT;
3794 goto out;
3795 }
3796
3797 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3798 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3799 btrfs_mark_buffer_dirty(path->nodes[0]);
3800 btrfs_free_path(path);
3801
3802 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3803 btrfs_end_transaction(trans, root);
3804 out:
3805 mnt_drop_write_file(file);
3806 return ret;
3807 }
3808
3809 void btrfs_get_block_group_info(struct list_head *groups_list,
3810 struct btrfs_ioctl_space_info *space)
3811 {
3812 struct btrfs_block_group_cache *block_group;
3813
3814 space->total_bytes = 0;
3815 space->used_bytes = 0;
3816 space->flags = 0;
3817 list_for_each_entry(block_group, groups_list, list) {
3818 space->flags = block_group->flags;
3819 space->total_bytes += block_group->key.offset;
3820 space->used_bytes +=
3821 btrfs_block_group_used(&block_group->item);
3822 }
3823 }
3824
3825 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3826 {
3827 struct btrfs_ioctl_space_args space_args;
3828 struct btrfs_ioctl_space_info space;
3829 struct btrfs_ioctl_space_info *dest;
3830 struct btrfs_ioctl_space_info *dest_orig;
3831 struct btrfs_ioctl_space_info __user *user_dest;
3832 struct btrfs_space_info *info;
3833 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3834 BTRFS_BLOCK_GROUP_SYSTEM,
3835 BTRFS_BLOCK_GROUP_METADATA,
3836 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3837 int num_types = 4;
3838 int alloc_size;
3839 int ret = 0;
3840 u64 slot_count = 0;
3841 int i, c;
3842
3843 if (copy_from_user(&space_args,
3844 (struct btrfs_ioctl_space_args __user *)arg,
3845 sizeof(space_args)))
3846 return -EFAULT;
3847
3848 for (i = 0; i < num_types; i++) {
3849 struct btrfs_space_info *tmp;
3850
3851 info = NULL;
3852 rcu_read_lock();
3853 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3854 list) {
3855 if (tmp->flags == types[i]) {
3856 info = tmp;
3857 break;
3858 }
3859 }
3860 rcu_read_unlock();
3861
3862 if (!info)
3863 continue;
3864
3865 down_read(&info->groups_sem);
3866 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3867 if (!list_empty(&info->block_groups[c]))
3868 slot_count++;
3869 }
3870 up_read(&info->groups_sem);
3871 }
3872
3873 /*
3874 * Global block reserve, exported as a space_info
3875 */
3876 slot_count++;
3877
3878 /* space_slots == 0 means they are asking for a count */
3879 if (space_args.space_slots == 0) {
3880 space_args.total_spaces = slot_count;
3881 goto out;
3882 }
3883
3884 slot_count = min_t(u64, space_args.space_slots, slot_count);
3885
3886 alloc_size = sizeof(*dest) * slot_count;
3887
3888 /* we generally have at most 6 or so space infos, one for each raid
3889 * level. So, a whole page should be more than enough for everyone
3890 */
3891 if (alloc_size > PAGE_CACHE_SIZE)
3892 return -ENOMEM;
3893
3894 space_args.total_spaces = 0;
3895 dest = kmalloc(alloc_size, GFP_NOFS);
3896 if (!dest)
3897 return -ENOMEM;
3898 dest_orig = dest;
3899
3900 /* now we have a buffer to copy into */
3901 for (i = 0; i < num_types; i++) {
3902 struct btrfs_space_info *tmp;
3903
3904 if (!slot_count)
3905 break;
3906
3907 info = NULL;
3908 rcu_read_lock();
3909 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3910 list) {
3911 if (tmp->flags == types[i]) {
3912 info = tmp;
3913 break;
3914 }
3915 }
3916 rcu_read_unlock();
3917
3918 if (!info)
3919 continue;
3920 down_read(&info->groups_sem);
3921 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3922 if (!list_empty(&info->block_groups[c])) {
3923 btrfs_get_block_group_info(
3924 &info->block_groups[c], &space);
3925 memcpy(dest, &space, sizeof(space));
3926 dest++;
3927 space_args.total_spaces++;
3928 slot_count--;
3929 }
3930 if (!slot_count)
3931 break;
3932 }
3933 up_read(&info->groups_sem);
3934 }
3935
3936 /*
3937 * Add global block reserve
3938 */
3939 if (slot_count) {
3940 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3941
3942 spin_lock(&block_rsv->lock);
3943 space.total_bytes = block_rsv->size;
3944 space.used_bytes = block_rsv->size - block_rsv->reserved;
3945 spin_unlock(&block_rsv->lock);
3946 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3947 memcpy(dest, &space, sizeof(space));
3948 space_args.total_spaces++;
3949 }
3950
3951 user_dest = (struct btrfs_ioctl_space_info __user *)
3952 (arg + sizeof(struct btrfs_ioctl_space_args));
3953
3954 if (copy_to_user(user_dest, dest_orig, alloc_size))
3955 ret = -EFAULT;
3956
3957 kfree(dest_orig);
3958 out:
3959 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3960 ret = -EFAULT;
3961
3962 return ret;
3963 }
3964
3965 /*
3966 * there are many ways the trans_start and trans_end ioctls can lead
3967 * to deadlocks. They should only be used by applications that
3968 * basically own the machine, and have a very in depth understanding
3969 * of all the possible deadlocks and enospc problems.
3970 */
3971 long btrfs_ioctl_trans_end(struct file *file)
3972 {
3973 struct inode *inode = file_inode(file);
3974 struct btrfs_root *root = BTRFS_I(inode)->root;
3975 struct btrfs_trans_handle *trans;
3976
3977 trans = file->private_data;
3978 if (!trans)
3979 return -EINVAL;
3980 file->private_data = NULL;
3981
3982 btrfs_end_transaction(trans, root);
3983
3984 atomic_dec(&root->fs_info->open_ioctl_trans);
3985
3986 mnt_drop_write_file(file);
3987 return 0;
3988 }
3989
3990 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3991 void __user *argp)
3992 {
3993 struct btrfs_trans_handle *trans;
3994 u64 transid;
3995 int ret;
3996
3997 trans = btrfs_attach_transaction_barrier(root);
3998 if (IS_ERR(trans)) {
3999 if (PTR_ERR(trans) != -ENOENT)
4000 return PTR_ERR(trans);
4001
4002 /* No running transaction, don't bother */
4003 transid = root->fs_info->last_trans_committed;
4004 goto out;
4005 }
4006 transid = trans->transid;
4007 ret = btrfs_commit_transaction_async(trans, root, 0);
4008 if (ret) {
4009 btrfs_end_transaction(trans, root);
4010 return ret;
4011 }
4012 out:
4013 if (argp)
4014 if (copy_to_user(argp, &transid, sizeof(transid)))
4015 return -EFAULT;
4016 return 0;
4017 }
4018
4019 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4020 void __user *argp)
4021 {
4022 u64 transid;
4023
4024 if (argp) {
4025 if (copy_from_user(&transid, argp, sizeof(transid)))
4026 return -EFAULT;
4027 } else {
4028 transid = 0; /* current trans */
4029 }
4030 return btrfs_wait_for_commit(root, transid);
4031 }
4032
4033 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4034 {
4035 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4036 struct btrfs_ioctl_scrub_args *sa;
4037 int ret;
4038
4039 if (!capable(CAP_SYS_ADMIN))
4040 return -EPERM;
4041
4042 sa = memdup_user(arg, sizeof(*sa));
4043 if (IS_ERR(sa))
4044 return PTR_ERR(sa);
4045
4046 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4047 ret = mnt_want_write_file(file);
4048 if (ret)
4049 goto out;
4050 }
4051
4052 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4053 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4054 0);
4055
4056 if (copy_to_user(arg, sa, sizeof(*sa)))
4057 ret = -EFAULT;
4058
4059 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4060 mnt_drop_write_file(file);
4061 out:
4062 kfree(sa);
4063 return ret;
4064 }
4065
4066 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4067 {
4068 if (!capable(CAP_SYS_ADMIN))
4069 return -EPERM;
4070
4071 return btrfs_scrub_cancel(root->fs_info);
4072 }
4073
4074 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4075 void __user *arg)
4076 {
4077 struct btrfs_ioctl_scrub_args *sa;
4078 int ret;
4079
4080 if (!capable(CAP_SYS_ADMIN))
4081 return -EPERM;
4082
4083 sa = memdup_user(arg, sizeof(*sa));
4084 if (IS_ERR(sa))
4085 return PTR_ERR(sa);
4086
4087 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4088
4089 if (copy_to_user(arg, sa, sizeof(*sa)))
4090 ret = -EFAULT;
4091
4092 kfree(sa);
4093 return ret;
4094 }
4095
4096 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4097 void __user *arg)
4098 {
4099 struct btrfs_ioctl_get_dev_stats *sa;
4100 int ret;
4101
4102 sa = memdup_user(arg, sizeof(*sa));
4103 if (IS_ERR(sa))
4104 return PTR_ERR(sa);
4105
4106 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4107 kfree(sa);
4108 return -EPERM;
4109 }
4110
4111 ret = btrfs_get_dev_stats(root, sa);
4112
4113 if (copy_to_user(arg, sa, sizeof(*sa)))
4114 ret = -EFAULT;
4115
4116 kfree(sa);
4117 return ret;
4118 }
4119
4120 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4121 {
4122 struct btrfs_ioctl_dev_replace_args *p;
4123 int ret;
4124
4125 if (!capable(CAP_SYS_ADMIN))
4126 return -EPERM;
4127
4128 p = memdup_user(arg, sizeof(*p));
4129 if (IS_ERR(p))
4130 return PTR_ERR(p);
4131
4132 switch (p->cmd) {
4133 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4134 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4135 ret = -EROFS;
4136 goto out;
4137 }
4138 if (atomic_xchg(
4139 &root->fs_info->mutually_exclusive_operation_running,
4140 1)) {
4141 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4142 } else {
4143 ret = btrfs_dev_replace_start(root, p);
4144 atomic_set(
4145 &root->fs_info->mutually_exclusive_operation_running,
4146 0);
4147 }
4148 break;
4149 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4150 btrfs_dev_replace_status(root->fs_info, p);
4151 ret = 0;
4152 break;
4153 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4154 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4155 break;
4156 default:
4157 ret = -EINVAL;
4158 break;
4159 }
4160
4161 if (copy_to_user(arg, p, sizeof(*p)))
4162 ret = -EFAULT;
4163 out:
4164 kfree(p);
4165 return ret;
4166 }
4167
4168 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4169 {
4170 int ret = 0;
4171 int i;
4172 u64 rel_ptr;
4173 int size;
4174 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4175 struct inode_fs_paths *ipath = NULL;
4176 struct btrfs_path *path;
4177
4178 if (!capable(CAP_DAC_READ_SEARCH))
4179 return -EPERM;
4180
4181 path = btrfs_alloc_path();
4182 if (!path) {
4183 ret = -ENOMEM;
4184 goto out;
4185 }
4186
4187 ipa = memdup_user(arg, sizeof(*ipa));
4188 if (IS_ERR(ipa)) {
4189 ret = PTR_ERR(ipa);
4190 ipa = NULL;
4191 goto out;
4192 }
4193
4194 size = min_t(u32, ipa->size, 4096);
4195 ipath = init_ipath(size, root, path);
4196 if (IS_ERR(ipath)) {
4197 ret = PTR_ERR(ipath);
4198 ipath = NULL;
4199 goto out;
4200 }
4201
4202 ret = paths_from_inode(ipa->inum, ipath);
4203 if (ret < 0)
4204 goto out;
4205
4206 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4207 rel_ptr = ipath->fspath->val[i] -
4208 (u64)(unsigned long)ipath->fspath->val;
4209 ipath->fspath->val[i] = rel_ptr;
4210 }
4211
4212 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4213 (void *)(unsigned long)ipath->fspath, size);
4214 if (ret) {
4215 ret = -EFAULT;
4216 goto out;
4217 }
4218
4219 out:
4220 btrfs_free_path(path);
4221 free_ipath(ipath);
4222 kfree(ipa);
4223
4224 return ret;
4225 }
4226
4227 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4228 {
4229 struct btrfs_data_container *inodes = ctx;
4230 const size_t c = 3 * sizeof(u64);
4231
4232 if (inodes->bytes_left >= c) {
4233 inodes->bytes_left -= c;
4234 inodes->val[inodes->elem_cnt] = inum;
4235 inodes->val[inodes->elem_cnt + 1] = offset;
4236 inodes->val[inodes->elem_cnt + 2] = root;
4237 inodes->elem_cnt += 3;
4238 } else {
4239 inodes->bytes_missing += c - inodes->bytes_left;
4240 inodes->bytes_left = 0;
4241 inodes->elem_missed += 3;
4242 }
4243
4244 return 0;
4245 }
4246
4247 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4248 void __user *arg)
4249 {
4250 int ret = 0;
4251 int size;
4252 struct btrfs_ioctl_logical_ino_args *loi;
4253 struct btrfs_data_container *inodes = NULL;
4254 struct btrfs_path *path = NULL;
4255
4256 if (!capable(CAP_SYS_ADMIN))
4257 return -EPERM;
4258
4259 loi = memdup_user(arg, sizeof(*loi));
4260 if (IS_ERR(loi)) {
4261 ret = PTR_ERR(loi);
4262 loi = NULL;
4263 goto out;
4264 }
4265
4266 path = btrfs_alloc_path();
4267 if (!path) {
4268 ret = -ENOMEM;
4269 goto out;
4270 }
4271
4272 size = min_t(u32, loi->size, 64 * 1024);
4273 inodes = init_data_container(size);
4274 if (IS_ERR(inodes)) {
4275 ret = PTR_ERR(inodes);
4276 inodes = NULL;
4277 goto out;
4278 }
4279
4280 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4281 build_ino_list, inodes);
4282 if (ret == -EINVAL)
4283 ret = -ENOENT;
4284 if (ret < 0)
4285 goto out;
4286
4287 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4288 (void *)(unsigned long)inodes, size);
4289 if (ret)
4290 ret = -EFAULT;
4291
4292 out:
4293 btrfs_free_path(path);
4294 vfree(inodes);
4295 kfree(loi);
4296
4297 return ret;
4298 }
4299
4300 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4301 struct btrfs_ioctl_balance_args *bargs)
4302 {
4303 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4304
4305 bargs->flags = bctl->flags;
4306
4307 if (atomic_read(&fs_info->balance_running))
4308 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4309 if (atomic_read(&fs_info->balance_pause_req))
4310 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4311 if (atomic_read(&fs_info->balance_cancel_req))
4312 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4313
4314 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4315 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4316 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4317
4318 if (lock) {
4319 spin_lock(&fs_info->balance_lock);
4320 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4321 spin_unlock(&fs_info->balance_lock);
4322 } else {
4323 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4324 }
4325 }
4326
4327 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4328 {
4329 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4330 struct btrfs_fs_info *fs_info = root->fs_info;
4331 struct btrfs_ioctl_balance_args *bargs;
4332 struct btrfs_balance_control *bctl;
4333 bool need_unlock; /* for mut. excl. ops lock */
4334 int ret;
4335
4336 if (!capable(CAP_SYS_ADMIN))
4337 return -EPERM;
4338
4339 ret = mnt_want_write_file(file);
4340 if (ret)
4341 return ret;
4342
4343 again:
4344 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4345 mutex_lock(&fs_info->volume_mutex);
4346 mutex_lock(&fs_info->balance_mutex);
4347 need_unlock = true;
4348 goto locked;
4349 }
4350
4351 /*
4352 * mut. excl. ops lock is locked. Three possibilites:
4353 * (1) some other op is running
4354 * (2) balance is running
4355 * (3) balance is paused -- special case (think resume)
4356 */
4357 mutex_lock(&fs_info->balance_mutex);
4358 if (fs_info->balance_ctl) {
4359 /* this is either (2) or (3) */
4360 if (!atomic_read(&fs_info->balance_running)) {
4361 mutex_unlock(&fs_info->balance_mutex);
4362 if (!mutex_trylock(&fs_info->volume_mutex))
4363 goto again;
4364 mutex_lock(&fs_info->balance_mutex);
4365
4366 if (fs_info->balance_ctl &&
4367 !atomic_read(&fs_info->balance_running)) {
4368 /* this is (3) */
4369 need_unlock = false;
4370 goto locked;
4371 }
4372
4373 mutex_unlock(&fs_info->balance_mutex);
4374 mutex_unlock(&fs_info->volume_mutex);
4375 goto again;
4376 } else {
4377 /* this is (2) */
4378 mutex_unlock(&fs_info->balance_mutex);
4379 ret = -EINPROGRESS;
4380 goto out;
4381 }
4382 } else {
4383 /* this is (1) */
4384 mutex_unlock(&fs_info->balance_mutex);
4385 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4386 goto out;
4387 }
4388
4389 locked:
4390 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4391
4392 if (arg) {
4393 bargs = memdup_user(arg, sizeof(*bargs));
4394 if (IS_ERR(bargs)) {
4395 ret = PTR_ERR(bargs);
4396 goto out_unlock;
4397 }
4398
4399 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4400 if (!fs_info->balance_ctl) {
4401 ret = -ENOTCONN;
4402 goto out_bargs;
4403 }
4404
4405 bctl = fs_info->balance_ctl;
4406 spin_lock(&fs_info->balance_lock);
4407 bctl->flags |= BTRFS_BALANCE_RESUME;
4408 spin_unlock(&fs_info->balance_lock);
4409
4410 goto do_balance;
4411 }
4412 } else {
4413 bargs = NULL;
4414 }
4415
4416 if (fs_info->balance_ctl) {
4417 ret = -EINPROGRESS;
4418 goto out_bargs;
4419 }
4420
4421 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4422 if (!bctl) {
4423 ret = -ENOMEM;
4424 goto out_bargs;
4425 }
4426
4427 bctl->fs_info = fs_info;
4428 if (arg) {
4429 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4430 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4431 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4432
4433 bctl->flags = bargs->flags;
4434 } else {
4435 /* balance everything - no filters */
4436 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4437 }
4438
4439 do_balance:
4440 /*
4441 * Ownership of bctl and mutually_exclusive_operation_running
4442 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4443 * or, if restriper was paused all the way until unmount, in
4444 * free_fs_info. mutually_exclusive_operation_running is
4445 * cleared in __cancel_balance.
4446 */
4447 need_unlock = false;
4448
4449 ret = btrfs_balance(bctl, bargs);
4450
4451 if (arg) {
4452 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4453 ret = -EFAULT;
4454 }
4455
4456 out_bargs:
4457 kfree(bargs);
4458 out_unlock:
4459 mutex_unlock(&fs_info->balance_mutex);
4460 mutex_unlock(&fs_info->volume_mutex);
4461 if (need_unlock)
4462 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4463 out:
4464 mnt_drop_write_file(file);
4465 return ret;
4466 }
4467
4468 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4469 {
4470 if (!capable(CAP_SYS_ADMIN))
4471 return -EPERM;
4472
4473 switch (cmd) {
4474 case BTRFS_BALANCE_CTL_PAUSE:
4475 return btrfs_pause_balance(root->fs_info);
4476 case BTRFS_BALANCE_CTL_CANCEL:
4477 return btrfs_cancel_balance(root->fs_info);
4478 }
4479
4480 return -EINVAL;
4481 }
4482
4483 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4484 void __user *arg)
4485 {
4486 struct btrfs_fs_info *fs_info = root->fs_info;
4487 struct btrfs_ioctl_balance_args *bargs;
4488 int ret = 0;
4489
4490 if (!capable(CAP_SYS_ADMIN))
4491 return -EPERM;
4492
4493 mutex_lock(&fs_info->balance_mutex);
4494 if (!fs_info->balance_ctl) {
4495 ret = -ENOTCONN;
4496 goto out;
4497 }
4498
4499 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4500 if (!bargs) {
4501 ret = -ENOMEM;
4502 goto out;
4503 }
4504
4505 update_ioctl_balance_args(fs_info, 1, bargs);
4506
4507 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4508 ret = -EFAULT;
4509
4510 kfree(bargs);
4511 out:
4512 mutex_unlock(&fs_info->balance_mutex);
4513 return ret;
4514 }
4515
4516 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4517 {
4518 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4519 struct btrfs_ioctl_quota_ctl_args *sa;
4520 struct btrfs_trans_handle *trans = NULL;
4521 int ret;
4522 int err;
4523
4524 if (!capable(CAP_SYS_ADMIN))
4525 return -EPERM;
4526
4527 ret = mnt_want_write_file(file);
4528 if (ret)
4529 return ret;
4530
4531 sa = memdup_user(arg, sizeof(*sa));
4532 if (IS_ERR(sa)) {
4533 ret = PTR_ERR(sa);
4534 goto drop_write;
4535 }
4536
4537 down_write(&root->fs_info->subvol_sem);
4538 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4539 if (IS_ERR(trans)) {
4540 ret = PTR_ERR(trans);
4541 goto out;
4542 }
4543
4544 switch (sa->cmd) {
4545 case BTRFS_QUOTA_CTL_ENABLE:
4546 ret = btrfs_quota_enable(trans, root->fs_info);
4547 break;
4548 case BTRFS_QUOTA_CTL_DISABLE:
4549 ret = btrfs_quota_disable(trans, root->fs_info);
4550 break;
4551 default:
4552 ret = -EINVAL;
4553 break;
4554 }
4555
4556 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4557 if (err && !ret)
4558 ret = err;
4559 out:
4560 kfree(sa);
4561 up_write(&root->fs_info->subvol_sem);
4562 drop_write:
4563 mnt_drop_write_file(file);
4564 return ret;
4565 }
4566
4567 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4568 {
4569 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4570 struct btrfs_ioctl_qgroup_assign_args *sa;
4571 struct btrfs_trans_handle *trans;
4572 int ret;
4573 int err;
4574
4575 if (!capable(CAP_SYS_ADMIN))
4576 return -EPERM;
4577
4578 ret = mnt_want_write_file(file);
4579 if (ret)
4580 return ret;
4581
4582 sa = memdup_user(arg, sizeof(*sa));
4583 if (IS_ERR(sa)) {
4584 ret = PTR_ERR(sa);
4585 goto drop_write;
4586 }
4587
4588 trans = btrfs_join_transaction(root);
4589 if (IS_ERR(trans)) {
4590 ret = PTR_ERR(trans);
4591 goto out;
4592 }
4593
4594 /* FIXME: check if the IDs really exist */
4595 if (sa->assign) {
4596 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4597 sa->src, sa->dst);
4598 } else {
4599 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4600 sa->src, sa->dst);
4601 }
4602
4603 err = btrfs_end_transaction(trans, root);
4604 if (err && !ret)
4605 ret = err;
4606
4607 out:
4608 kfree(sa);
4609 drop_write:
4610 mnt_drop_write_file(file);
4611 return ret;
4612 }
4613
4614 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4615 {
4616 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4617 struct btrfs_ioctl_qgroup_create_args *sa;
4618 struct btrfs_trans_handle *trans;
4619 int ret;
4620 int err;
4621
4622 if (!capable(CAP_SYS_ADMIN))
4623 return -EPERM;
4624
4625 ret = mnt_want_write_file(file);
4626 if (ret)
4627 return ret;
4628
4629 sa = memdup_user(arg, sizeof(*sa));
4630 if (IS_ERR(sa)) {
4631 ret = PTR_ERR(sa);
4632 goto drop_write;
4633 }
4634
4635 if (!sa->qgroupid) {
4636 ret = -EINVAL;
4637 goto out;
4638 }
4639
4640 trans = btrfs_join_transaction(root);
4641 if (IS_ERR(trans)) {
4642 ret = PTR_ERR(trans);
4643 goto out;
4644 }
4645
4646 /* FIXME: check if the IDs really exist */
4647 if (sa->create) {
4648 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4649 NULL);
4650 } else {
4651 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4652 }
4653
4654 err = btrfs_end_transaction(trans, root);
4655 if (err && !ret)
4656 ret = err;
4657
4658 out:
4659 kfree(sa);
4660 drop_write:
4661 mnt_drop_write_file(file);
4662 return ret;
4663 }
4664
4665 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4666 {
4667 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4668 struct btrfs_ioctl_qgroup_limit_args *sa;
4669 struct btrfs_trans_handle *trans;
4670 int ret;
4671 int err;
4672 u64 qgroupid;
4673
4674 if (!capable(CAP_SYS_ADMIN))
4675 return -EPERM;
4676
4677 ret = mnt_want_write_file(file);
4678 if (ret)
4679 return ret;
4680
4681 sa = memdup_user(arg, sizeof(*sa));
4682 if (IS_ERR(sa)) {
4683 ret = PTR_ERR(sa);
4684 goto drop_write;
4685 }
4686
4687 trans = btrfs_join_transaction(root);
4688 if (IS_ERR(trans)) {
4689 ret = PTR_ERR(trans);
4690 goto out;
4691 }
4692
4693 qgroupid = sa->qgroupid;
4694 if (!qgroupid) {
4695 /* take the current subvol as qgroup */
4696 qgroupid = root->root_key.objectid;
4697 }
4698
4699 /* FIXME: check if the IDs really exist */
4700 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4701
4702 err = btrfs_end_transaction(trans, root);
4703 if (err && !ret)
4704 ret = err;
4705
4706 out:
4707 kfree(sa);
4708 drop_write:
4709 mnt_drop_write_file(file);
4710 return ret;
4711 }
4712
4713 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4714 {
4715 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4716 struct btrfs_ioctl_quota_rescan_args *qsa;
4717 int ret;
4718
4719 if (!capable(CAP_SYS_ADMIN))
4720 return -EPERM;
4721
4722 ret = mnt_want_write_file(file);
4723 if (ret)
4724 return ret;
4725
4726 qsa = memdup_user(arg, sizeof(*qsa));
4727 if (IS_ERR(qsa)) {
4728 ret = PTR_ERR(qsa);
4729 goto drop_write;
4730 }
4731
4732 if (qsa->flags) {
4733 ret = -EINVAL;
4734 goto out;
4735 }
4736
4737 ret = btrfs_qgroup_rescan(root->fs_info);
4738
4739 out:
4740 kfree(qsa);
4741 drop_write:
4742 mnt_drop_write_file(file);
4743 return ret;
4744 }
4745
4746 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4747 {
4748 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4749 struct btrfs_ioctl_quota_rescan_args *qsa;
4750 int ret = 0;
4751
4752 if (!capable(CAP_SYS_ADMIN))
4753 return -EPERM;
4754
4755 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4756 if (!qsa)
4757 return -ENOMEM;
4758
4759 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4760 qsa->flags = 1;
4761 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4762 }
4763
4764 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4765 ret = -EFAULT;
4766
4767 kfree(qsa);
4768 return ret;
4769 }
4770
4771 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4772 {
4773 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4774
4775 if (!capable(CAP_SYS_ADMIN))
4776 return -EPERM;
4777
4778 return btrfs_qgroup_wait_for_completion(root->fs_info);
4779 }
4780
4781 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4782 struct btrfs_ioctl_received_subvol_args *sa)
4783 {
4784 struct inode *inode = file_inode(file);
4785 struct btrfs_root *root = BTRFS_I(inode)->root;
4786 struct btrfs_root_item *root_item = &root->root_item;
4787 struct btrfs_trans_handle *trans;
4788 struct timespec ct = CURRENT_TIME;
4789 int ret = 0;
4790 int received_uuid_changed;
4791
4792 if (!inode_owner_or_capable(inode))
4793 return -EPERM;
4794
4795 ret = mnt_want_write_file(file);
4796 if (ret < 0)
4797 return ret;
4798
4799 down_write(&root->fs_info->subvol_sem);
4800
4801 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4802 ret = -EINVAL;
4803 goto out;
4804 }
4805
4806 if (btrfs_root_readonly(root)) {
4807 ret = -EROFS;
4808 goto out;
4809 }
4810
4811 /*
4812 * 1 - root item
4813 * 2 - uuid items (received uuid + subvol uuid)
4814 */
4815 trans = btrfs_start_transaction(root, 3);
4816 if (IS_ERR(trans)) {
4817 ret = PTR_ERR(trans);
4818 trans = NULL;
4819 goto out;
4820 }
4821
4822 sa->rtransid = trans->transid;
4823 sa->rtime.sec = ct.tv_sec;
4824 sa->rtime.nsec = ct.tv_nsec;
4825
4826 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4827 BTRFS_UUID_SIZE);
4828 if (received_uuid_changed &&
4829 !btrfs_is_empty_uuid(root_item->received_uuid))
4830 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4831 root_item->received_uuid,
4832 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4833 root->root_key.objectid);
4834 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4835 btrfs_set_root_stransid(root_item, sa->stransid);
4836 btrfs_set_root_rtransid(root_item, sa->rtransid);
4837 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4838 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4839 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4840 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4841
4842 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4843 &root->root_key, &root->root_item);
4844 if (ret < 0) {
4845 btrfs_end_transaction(trans, root);
4846 goto out;
4847 }
4848 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4849 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4850 sa->uuid,
4851 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4852 root->root_key.objectid);
4853 if (ret < 0 && ret != -EEXIST) {
4854 btrfs_abort_transaction(trans, root, ret);
4855 goto out;
4856 }
4857 }
4858 ret = btrfs_commit_transaction(trans, root);
4859 if (ret < 0) {
4860 btrfs_abort_transaction(trans, root, ret);
4861 goto out;
4862 }
4863
4864 out:
4865 up_write(&root->fs_info->subvol_sem);
4866 mnt_drop_write_file(file);
4867 return ret;
4868 }
4869
4870 #ifdef CONFIG_64BIT
4871 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4872 void __user *arg)
4873 {
4874 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4875 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4876 int ret = 0;
4877
4878 args32 = memdup_user(arg, sizeof(*args32));
4879 if (IS_ERR(args32)) {
4880 ret = PTR_ERR(args32);
4881 args32 = NULL;
4882 goto out;
4883 }
4884
4885 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4886 if (!args64) {
4887 ret = -ENOMEM;
4888 goto out;
4889 }
4890
4891 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4892 args64->stransid = args32->stransid;
4893 args64->rtransid = args32->rtransid;
4894 args64->stime.sec = args32->stime.sec;
4895 args64->stime.nsec = args32->stime.nsec;
4896 args64->rtime.sec = args32->rtime.sec;
4897 args64->rtime.nsec = args32->rtime.nsec;
4898 args64->flags = args32->flags;
4899
4900 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4901 if (ret)
4902 goto out;
4903
4904 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4905 args32->stransid = args64->stransid;
4906 args32->rtransid = args64->rtransid;
4907 args32->stime.sec = args64->stime.sec;
4908 args32->stime.nsec = args64->stime.nsec;
4909 args32->rtime.sec = args64->rtime.sec;
4910 args32->rtime.nsec = args64->rtime.nsec;
4911 args32->flags = args64->flags;
4912
4913 ret = copy_to_user(arg, args32, sizeof(*args32));
4914 if (ret)
4915 ret = -EFAULT;
4916
4917 out:
4918 kfree(args32);
4919 kfree(args64);
4920 return ret;
4921 }
4922 #endif
4923
4924 static long btrfs_ioctl_set_received_subvol(struct file *file,
4925 void __user *arg)
4926 {
4927 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4928 int ret = 0;
4929
4930 sa = memdup_user(arg, sizeof(*sa));
4931 if (IS_ERR(sa)) {
4932 ret = PTR_ERR(sa);
4933 sa = NULL;
4934 goto out;
4935 }
4936
4937 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4938
4939 if (ret)
4940 goto out;
4941
4942 ret = copy_to_user(arg, sa, sizeof(*sa));
4943 if (ret)
4944 ret = -EFAULT;
4945
4946 out:
4947 kfree(sa);
4948 return ret;
4949 }
4950
4951 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4952 {
4953 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4954 size_t len;
4955 int ret;
4956 char label[BTRFS_LABEL_SIZE];
4957
4958 spin_lock(&root->fs_info->super_lock);
4959 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4960 spin_unlock(&root->fs_info->super_lock);
4961
4962 len = strnlen(label, BTRFS_LABEL_SIZE);
4963
4964 if (len == BTRFS_LABEL_SIZE) {
4965 btrfs_warn(root->fs_info,
4966 "label is too long, return the first %zu bytes", --len);
4967 }
4968
4969 ret = copy_to_user(arg, label, len);
4970
4971 return ret ? -EFAULT : 0;
4972 }
4973
4974 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4975 {
4976 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4977 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4978 struct btrfs_trans_handle *trans;
4979 char label[BTRFS_LABEL_SIZE];
4980 int ret;
4981
4982 if (!capable(CAP_SYS_ADMIN))
4983 return -EPERM;
4984
4985 if (copy_from_user(label, arg, sizeof(label)))
4986 return -EFAULT;
4987
4988 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4989 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
4990 BTRFS_LABEL_SIZE - 1);
4991 return -EINVAL;
4992 }
4993
4994 ret = mnt_want_write_file(file);
4995 if (ret)
4996 return ret;
4997
4998 trans = btrfs_start_transaction(root, 0);
4999 if (IS_ERR(trans)) {
5000 ret = PTR_ERR(trans);
5001 goto out_unlock;
5002 }
5003
5004 spin_lock(&root->fs_info->super_lock);
5005 strcpy(super_block->label, label);
5006 spin_unlock(&root->fs_info->super_lock);
5007 ret = btrfs_commit_transaction(trans, root);
5008
5009 out_unlock:
5010 mnt_drop_write_file(file);
5011 return ret;
5012 }
5013
5014 #define INIT_FEATURE_FLAGS(suffix) \
5015 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5016 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5017 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5018
5019 static int btrfs_ioctl_get_supported_features(struct file *file,
5020 void __user *arg)
5021 {
5022 static struct btrfs_ioctl_feature_flags features[3] = {
5023 INIT_FEATURE_FLAGS(SUPP),
5024 INIT_FEATURE_FLAGS(SAFE_SET),
5025 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5026 };
5027
5028 if (copy_to_user(arg, &features, sizeof(features)))
5029 return -EFAULT;
5030
5031 return 0;
5032 }
5033
5034 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5035 {
5036 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5037 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5038 struct btrfs_ioctl_feature_flags features;
5039
5040 features.compat_flags = btrfs_super_compat_flags(super_block);
5041 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5042 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5043
5044 if (copy_to_user(arg, &features, sizeof(features)))
5045 return -EFAULT;
5046
5047 return 0;
5048 }
5049
5050 static int check_feature_bits(struct btrfs_root *root,
5051 enum btrfs_feature_set set,
5052 u64 change_mask, u64 flags, u64 supported_flags,
5053 u64 safe_set, u64 safe_clear)
5054 {
5055 const char *type = btrfs_feature_set_names[set];
5056 char *names;
5057 u64 disallowed, unsupported;
5058 u64 set_mask = flags & change_mask;
5059 u64 clear_mask = ~flags & change_mask;
5060
5061 unsupported = set_mask & ~supported_flags;
5062 if (unsupported) {
5063 names = btrfs_printable_features(set, unsupported);
5064 if (names) {
5065 btrfs_warn(root->fs_info,
5066 "this kernel does not support the %s feature bit%s",
5067 names, strchr(names, ',') ? "s" : "");
5068 kfree(names);
5069 } else
5070 btrfs_warn(root->fs_info,
5071 "this kernel does not support %s bits 0x%llx",
5072 type, unsupported);
5073 return -EOPNOTSUPP;
5074 }
5075
5076 disallowed = set_mask & ~safe_set;
5077 if (disallowed) {
5078 names = btrfs_printable_features(set, disallowed);
5079 if (names) {
5080 btrfs_warn(root->fs_info,
5081 "can't set the %s feature bit%s while mounted",
5082 names, strchr(names, ',') ? "s" : "");
5083 kfree(names);
5084 } else
5085 btrfs_warn(root->fs_info,
5086 "can't set %s bits 0x%llx while mounted",
5087 type, disallowed);
5088 return -EPERM;
5089 }
5090
5091 disallowed = clear_mask & ~safe_clear;
5092 if (disallowed) {
5093 names = btrfs_printable_features(set, disallowed);
5094 if (names) {
5095 btrfs_warn(root->fs_info,
5096 "can't clear the %s feature bit%s while mounted",
5097 names, strchr(names, ',') ? "s" : "");
5098 kfree(names);
5099 } else
5100 btrfs_warn(root->fs_info,
5101 "can't clear %s bits 0x%llx while mounted",
5102 type, disallowed);
5103 return -EPERM;
5104 }
5105
5106 return 0;
5107 }
5108
5109 #define check_feature(root, change_mask, flags, mask_base) \
5110 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5111 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5112 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5113 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5114
5115 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5116 {
5117 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5118 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5119 struct btrfs_ioctl_feature_flags flags[2];
5120 struct btrfs_trans_handle *trans;
5121 u64 newflags;
5122 int ret;
5123
5124 if (!capable(CAP_SYS_ADMIN))
5125 return -EPERM;
5126
5127 if (copy_from_user(flags, arg, sizeof(flags)))
5128 return -EFAULT;
5129
5130 /* Nothing to do */
5131 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5132 !flags[0].incompat_flags)
5133 return 0;
5134
5135 ret = check_feature(root, flags[0].compat_flags,
5136 flags[1].compat_flags, COMPAT);
5137 if (ret)
5138 return ret;
5139
5140 ret = check_feature(root, flags[0].compat_ro_flags,
5141 flags[1].compat_ro_flags, COMPAT_RO);
5142 if (ret)
5143 return ret;
5144
5145 ret = check_feature(root, flags[0].incompat_flags,
5146 flags[1].incompat_flags, INCOMPAT);
5147 if (ret)
5148 return ret;
5149
5150 trans = btrfs_start_transaction(root, 0);
5151 if (IS_ERR(trans))
5152 return PTR_ERR(trans);
5153
5154 spin_lock(&root->fs_info->super_lock);
5155 newflags = btrfs_super_compat_flags(super_block);
5156 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5157 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5158 btrfs_set_super_compat_flags(super_block, newflags);
5159
5160 newflags = btrfs_super_compat_ro_flags(super_block);
5161 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5162 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5163 btrfs_set_super_compat_ro_flags(super_block, newflags);
5164
5165 newflags = btrfs_super_incompat_flags(super_block);
5166 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5167 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5168 btrfs_set_super_incompat_flags(super_block, newflags);
5169 spin_unlock(&root->fs_info->super_lock);
5170
5171 return btrfs_commit_transaction(trans, root);
5172 }
5173
5174 long btrfs_ioctl(struct file *file, unsigned int
5175 cmd, unsigned long arg)
5176 {
5177 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5178 void __user *argp = (void __user *)arg;
5179
5180 switch (cmd) {
5181 case FS_IOC_GETFLAGS:
5182 return btrfs_ioctl_getflags(file, argp);
5183 case FS_IOC_SETFLAGS:
5184 return btrfs_ioctl_setflags(file, argp);
5185 case FS_IOC_GETVERSION:
5186 return btrfs_ioctl_getversion(file, argp);
5187 case FITRIM:
5188 return btrfs_ioctl_fitrim(file, argp);
5189 case BTRFS_IOC_SNAP_CREATE:
5190 return btrfs_ioctl_snap_create(file, argp, 0);
5191 case BTRFS_IOC_SNAP_CREATE_V2:
5192 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5193 case BTRFS_IOC_SUBVOL_CREATE:
5194 return btrfs_ioctl_snap_create(file, argp, 1);
5195 case BTRFS_IOC_SUBVOL_CREATE_V2:
5196 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5197 case BTRFS_IOC_SNAP_DESTROY:
5198 return btrfs_ioctl_snap_destroy(file, argp);
5199 case BTRFS_IOC_SUBVOL_GETFLAGS:
5200 return btrfs_ioctl_subvol_getflags(file, argp);
5201 case BTRFS_IOC_SUBVOL_SETFLAGS:
5202 return btrfs_ioctl_subvol_setflags(file, argp);
5203 case BTRFS_IOC_DEFAULT_SUBVOL:
5204 return btrfs_ioctl_default_subvol(file, argp);
5205 case BTRFS_IOC_DEFRAG:
5206 return btrfs_ioctl_defrag(file, NULL);
5207 case BTRFS_IOC_DEFRAG_RANGE:
5208 return btrfs_ioctl_defrag(file, argp);
5209 case BTRFS_IOC_RESIZE:
5210 return btrfs_ioctl_resize(file, argp);
5211 case BTRFS_IOC_ADD_DEV:
5212 return btrfs_ioctl_add_dev(root, argp);
5213 case BTRFS_IOC_RM_DEV:
5214 return btrfs_ioctl_rm_dev(file, argp);
5215 case BTRFS_IOC_FS_INFO:
5216 return btrfs_ioctl_fs_info(root, argp);
5217 case BTRFS_IOC_DEV_INFO:
5218 return btrfs_ioctl_dev_info(root, argp);
5219 case BTRFS_IOC_BALANCE:
5220 return btrfs_ioctl_balance(file, NULL);
5221 case BTRFS_IOC_CLONE:
5222 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5223 case BTRFS_IOC_CLONE_RANGE:
5224 return btrfs_ioctl_clone_range(file, argp);
5225 case BTRFS_IOC_TRANS_START:
5226 return btrfs_ioctl_trans_start(file);
5227 case BTRFS_IOC_TRANS_END:
5228 return btrfs_ioctl_trans_end(file);
5229 case BTRFS_IOC_TREE_SEARCH:
5230 return btrfs_ioctl_tree_search(file, argp);
5231 case BTRFS_IOC_INO_LOOKUP:
5232 return btrfs_ioctl_ino_lookup(file, argp);
5233 case BTRFS_IOC_INO_PATHS:
5234 return btrfs_ioctl_ino_to_path(root, argp);
5235 case BTRFS_IOC_LOGICAL_INO:
5236 return btrfs_ioctl_logical_to_ino(root, argp);
5237 case BTRFS_IOC_SPACE_INFO:
5238 return btrfs_ioctl_space_info(root, argp);
5239 case BTRFS_IOC_SYNC: {
5240 int ret;
5241
5242 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5243 if (ret)
5244 return ret;
5245 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
5246 return ret;
5247 }
5248 case BTRFS_IOC_START_SYNC:
5249 return btrfs_ioctl_start_sync(root, argp);
5250 case BTRFS_IOC_WAIT_SYNC:
5251 return btrfs_ioctl_wait_sync(root, argp);
5252 case BTRFS_IOC_SCRUB:
5253 return btrfs_ioctl_scrub(file, argp);
5254 case BTRFS_IOC_SCRUB_CANCEL:
5255 return btrfs_ioctl_scrub_cancel(root, argp);
5256 case BTRFS_IOC_SCRUB_PROGRESS:
5257 return btrfs_ioctl_scrub_progress(root, argp);
5258 case BTRFS_IOC_BALANCE_V2:
5259 return btrfs_ioctl_balance(file, argp);
5260 case BTRFS_IOC_BALANCE_CTL:
5261 return btrfs_ioctl_balance_ctl(root, arg);
5262 case BTRFS_IOC_BALANCE_PROGRESS:
5263 return btrfs_ioctl_balance_progress(root, argp);
5264 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5265 return btrfs_ioctl_set_received_subvol(file, argp);
5266 #ifdef CONFIG_64BIT
5267 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5268 return btrfs_ioctl_set_received_subvol_32(file, argp);
5269 #endif
5270 case BTRFS_IOC_SEND:
5271 return btrfs_ioctl_send(file, argp);
5272 case BTRFS_IOC_GET_DEV_STATS:
5273 return btrfs_ioctl_get_dev_stats(root, argp);
5274 case BTRFS_IOC_QUOTA_CTL:
5275 return btrfs_ioctl_quota_ctl(file, argp);
5276 case BTRFS_IOC_QGROUP_ASSIGN:
5277 return btrfs_ioctl_qgroup_assign(file, argp);
5278 case BTRFS_IOC_QGROUP_CREATE:
5279 return btrfs_ioctl_qgroup_create(file, argp);
5280 case BTRFS_IOC_QGROUP_LIMIT:
5281 return btrfs_ioctl_qgroup_limit(file, argp);
5282 case BTRFS_IOC_QUOTA_RESCAN:
5283 return btrfs_ioctl_quota_rescan(file, argp);
5284 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5285 return btrfs_ioctl_quota_rescan_status(file, argp);
5286 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5287 return btrfs_ioctl_quota_rescan_wait(file, argp);
5288 case BTRFS_IOC_DEV_REPLACE:
5289 return btrfs_ioctl_dev_replace(root, argp);
5290 case BTRFS_IOC_GET_FSLABEL:
5291 return btrfs_ioctl_get_fslabel(file, argp);
5292 case BTRFS_IOC_SET_FSLABEL:
5293 return btrfs_ioctl_set_fslabel(file, argp);
5294 case BTRFS_IOC_FILE_EXTENT_SAME:
5295 return btrfs_ioctl_file_extent_same(file, argp);
5296 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5297 return btrfs_ioctl_get_supported_features(file, argp);
5298 case BTRFS_IOC_GET_FEATURES:
5299 return btrfs_ioctl_get_features(file, argp);
5300 case BTRFS_IOC_SET_FEATURES:
5301 return btrfs_ioctl_set_features(file, argp);
5302 }
5303
5304 return -ENOTTY;
5305 }
ubuntu-zesty-kernel mirror