]> git.ipfire.org Git - people/ms/linux.git/blob - fs/btrfs/ioctl.c
projects / people / ms / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge tag 'soc-fixes-6.0-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc
[people/ms/linux.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 int no_time_update);
92
93 /* Mask out flags that are inappropriate for the given type of inode. */
94 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
95 {
96 if (S_ISDIR(mode))
97 return flags;
98 else if (S_ISREG(mode))
99 return flags & ~FS_DIRSYNC_FL;
100 else
101 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
102 }
103
104 /*
105 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
106 */
107 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
108 {
109 unsigned int iflags = 0;
110
111 if (flags & BTRFS_INODE_SYNC)
112 iflags |= FS_SYNC_FL;
113 if (flags & BTRFS_INODE_IMMUTABLE)
114 iflags |= FS_IMMUTABLE_FL;
115 if (flags & BTRFS_INODE_APPEND)
116 iflags |= FS_APPEND_FL;
117 if (flags & BTRFS_INODE_NODUMP)
118 iflags |= FS_NODUMP_FL;
119 if (flags & BTRFS_INODE_NOATIME)
120 iflags |= FS_NOATIME_FL;
121 if (flags & BTRFS_INODE_DIRSYNC)
122 iflags |= FS_DIRSYNC_FL;
123 if (flags & BTRFS_INODE_NODATACOW)
124 iflags |= FS_NOCOW_FL;
125
126 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
127 iflags |= FS_COMPR_FL;
128 else if (flags & BTRFS_INODE_NOCOMPRESS)
129 iflags |= FS_NOCOMP_FL;
130
131 return iflags;
132 }
133
134 /*
135 * Update inode->i_flags based on the btrfs internal flags.
136 */
137 void btrfs_update_iflags(struct inode *inode)
138 {
139 struct btrfs_inode *ip = BTRFS_I(inode);
140 unsigned int new_fl = 0;
141
142 if (ip->flags & BTRFS_INODE_SYNC)
143 new_fl |= S_SYNC;
144 if (ip->flags & BTRFS_INODE_IMMUTABLE)
145 new_fl |= S_IMMUTABLE;
146 if (ip->flags & BTRFS_INODE_APPEND)
147 new_fl |= S_APPEND;
148 if (ip->flags & BTRFS_INODE_NOATIME)
149 new_fl |= S_NOATIME;
150 if (ip->flags & BTRFS_INODE_DIRSYNC)
151 new_fl |= S_DIRSYNC;
152
153 set_mask_bits(&inode->i_flags,
154 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
155 new_fl);
156 }
157
158 /*
159 * Inherit flags from the parent inode.
160 *
161 * Currently only the compression flags and the cow flags are inherited.
162 */
163 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
164 {
165 unsigned int flags;
166
167 if (!dir)
168 return;
169
170 flags = BTRFS_I(dir)->flags;
171
172 if (flags & BTRFS_INODE_NOCOMPRESS) {
173 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
174 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
175 } else if (flags & BTRFS_INODE_COMPRESS) {
176 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
177 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
178 }
179
180 if (flags & BTRFS_INODE_NODATACOW) {
181 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
182 if (S_ISREG(inode->i_mode))
183 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
184 }
185
186 btrfs_update_iflags(inode);
187 }
188
189 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
190 {
191 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
192 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
193
194 if (copy_to_user(arg, &flags, sizeof(flags)))
195 return -EFAULT;
196 return 0;
197 }
198
199 static int check_flags(unsigned int flags)
200 {
201 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
202 FS_NOATIME_FL | FS_NODUMP_FL | \
203 FS_SYNC_FL | FS_DIRSYNC_FL | \
204 FS_NOCOMP_FL | FS_COMPR_FL |
205 FS_NOCOW_FL))
206 return -EOPNOTSUPP;
207
208 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
209 return -EINVAL;
210
211 return 0;
212 }
213
214 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
215 {
216 struct inode *inode = file_inode(file);
217 struct btrfs_inode *ip = BTRFS_I(inode);
218 struct btrfs_root *root = ip->root;
219 struct btrfs_trans_handle *trans;
220 unsigned int flags, oldflags;
221 int ret;
222 u64 ip_oldflags;
223 unsigned int i_oldflags;
224 umode_t mode;
225
226 if (!inode_owner_or_capable(inode))
227 return -EPERM;
228
229 if (btrfs_root_readonly(root))
230 return -EROFS;
231
232 if (copy_from_user(&flags, arg, sizeof(flags)))
233 return -EFAULT;
234
235 ret = check_flags(flags);
236 if (ret)
237 return ret;
238
239 ret = mnt_want_write_file(file);
240 if (ret)
241 return ret;
242
243 mutex_lock(&inode->i_mutex);
244
245 ip_oldflags = ip->flags;
246 i_oldflags = inode->i_flags;
247 mode = inode->i_mode;
248
249 flags = btrfs_mask_flags(inode->i_mode, flags);
250 oldflags = btrfs_flags_to_ioctl(ip->flags);
251 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
252 if (!capable(CAP_LINUX_IMMUTABLE)) {
253 ret = -EPERM;
254 goto out_unlock;
255 }
256 }
257
258 if (flags & FS_SYNC_FL)
259 ip->flags |= BTRFS_INODE_SYNC;
260 else
261 ip->flags &= ~BTRFS_INODE_SYNC;
262 if (flags & FS_IMMUTABLE_FL)
263 ip->flags |= BTRFS_INODE_IMMUTABLE;
264 else
265 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
266 if (flags & FS_APPEND_FL)
267 ip->flags |= BTRFS_INODE_APPEND;
268 else
269 ip->flags &= ~BTRFS_INODE_APPEND;
270 if (flags & FS_NODUMP_FL)
271 ip->flags |= BTRFS_INODE_NODUMP;
272 else
273 ip->flags &= ~BTRFS_INODE_NODUMP;
274 if (flags & FS_NOATIME_FL)
275 ip->flags |= BTRFS_INODE_NOATIME;
276 else
277 ip->flags &= ~BTRFS_INODE_NOATIME;
278 if (flags & FS_DIRSYNC_FL)
279 ip->flags |= BTRFS_INODE_DIRSYNC;
280 else
281 ip->flags &= ~BTRFS_INODE_DIRSYNC;
282 if (flags & FS_NOCOW_FL) {
283 if (S_ISREG(mode)) {
284 /*
285 * It's safe to turn csums off here, no extents exist.
286 * Otherwise we want the flag to reflect the real COW
287 * status of the file and will not set it.
288 */
289 if (inode->i_size == 0)
290 ip->flags |= BTRFS_INODE_NODATACOW
291 | BTRFS_INODE_NODATASUM;
292 } else {
293 ip->flags |= BTRFS_INODE_NODATACOW;
294 }
295 } else {
296 /*
297 * Revert back under same assuptions as above
298 */
299 if (S_ISREG(mode)) {
300 if (inode->i_size == 0)
301 ip->flags &= ~(BTRFS_INODE_NODATACOW
302 | BTRFS_INODE_NODATASUM);
303 } else {
304 ip->flags &= ~BTRFS_INODE_NODATACOW;
305 }
306 }
307
308 /*
309 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
310 * flag may be changed automatically if compression code won't make
311 * things smaller.
312 */
313 if (flags & FS_NOCOMP_FL) {
314 ip->flags &= ~BTRFS_INODE_COMPRESS;
315 ip->flags |= BTRFS_INODE_NOCOMPRESS;
316
317 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
318 if (ret && ret != -ENODATA)
319 goto out_drop;
320 } else if (flags & FS_COMPR_FL) {
321 const char *comp;
322
323 ip->flags |= BTRFS_INODE_COMPRESS;
324 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
325
326 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
327 comp = "lzo";
328 else
329 comp = "zlib";
330 ret = btrfs_set_prop(inode, "btrfs.compression",
331 comp, strlen(comp), 0);
332 if (ret)
333 goto out_drop;
334
335 } else {
336 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
337 if (ret && ret != -ENODATA)
338 goto out_drop;
339 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
340 }
341
342 trans = btrfs_start_transaction(root, 1);
343 if (IS_ERR(trans)) {
344 ret = PTR_ERR(trans);
345 goto out_drop;
346 }
347
348 btrfs_update_iflags(inode);
349 inode_inc_iversion(inode);
350 inode->i_ctime = CURRENT_TIME;
351 ret = btrfs_update_inode(trans, root, inode);
352
353 btrfs_end_transaction(trans, root);
354 out_drop:
355 if (ret) {
356 ip->flags = ip_oldflags;
357 inode->i_flags = i_oldflags;
358 }
359
360 out_unlock:
361 mutex_unlock(&inode->i_mutex);
362 mnt_drop_write_file(file);
363 return ret;
364 }
365
366 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
367 {
368 struct inode *inode = file_inode(file);
369
370 return put_user(inode->i_generation, arg);
371 }
372
373 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
374 {
375 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
376 struct btrfs_device *device;
377 struct request_queue *q;
378 struct fstrim_range range;
379 u64 minlen = ULLONG_MAX;
380 u64 num_devices = 0;
381 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
382 int ret;
383
384 if (!capable(CAP_SYS_ADMIN))
385 return -EPERM;
386
387 rcu_read_lock();
388 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
389 dev_list) {
390 if (!device->bdev)
391 continue;
392 q = bdev_get_queue(device->bdev);
393 if (blk_queue_discard(q)) {
394 num_devices++;
395 minlen = min((u64)q->limits.discard_granularity,
396 minlen);
397 }
398 }
399 rcu_read_unlock();
400
401 if (!num_devices)
402 return -EOPNOTSUPP;
403 if (copy_from_user(&range, arg, sizeof(range)))
404 return -EFAULT;
405 if (range.start > total_bytes ||
406 range.len < fs_info->sb->s_blocksize)
407 return -EINVAL;
408
409 range.len = min(range.len, total_bytes - range.start);
410 range.minlen = max(range.minlen, minlen);
411 ret = btrfs_trim_fs(fs_info->tree_root, &range);
412 if (ret < 0)
413 return ret;
414
415 if (copy_to_user(arg, &range, sizeof(range)))
416 return -EFAULT;
417
418 return 0;
419 }
420
421 int btrfs_is_empty_uuid(u8 *uuid)
422 {
423 int i;
424
425 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
426 if (uuid[i])
427 return 0;
428 }
429 return 1;
430 }
431
432 static noinline int create_subvol(struct inode *dir,
433 struct dentry *dentry,
434 char *name, int namelen,
435 u64 *async_transid,
436 struct btrfs_qgroup_inherit *inherit)
437 {
438 struct btrfs_trans_handle *trans;
439 struct btrfs_key key;
440 struct btrfs_root_item root_item;
441 struct btrfs_inode_item *inode_item;
442 struct extent_buffer *leaf;
443 struct btrfs_root *root = BTRFS_I(dir)->root;
444 struct btrfs_root *new_root;
445 struct btrfs_block_rsv block_rsv;
446 struct timespec cur_time = CURRENT_TIME;
447 struct inode *inode;
448 int ret;
449 int err;
450 u64 objectid;
451 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
452 u64 index = 0;
453 u64 qgroup_reserved;
454 uuid_le new_uuid;
455
456 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
457 if (ret)
458 return ret;
459
460 /*
461 * Don't create subvolume whose level is not zero. Or qgroup will be
462 * screwed up since it assume subvolme qgroup's level to be 0.
463 */
464 if (btrfs_qgroup_level(objectid))
465 return -ENOSPC;
466
467 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
468 /*
469 * The same as the snapshot creation, please see the comment
470 * of create_snapshot().
471 */
472 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
473 8, &qgroup_reserved, false);
474 if (ret)
475 return ret;
476
477 trans = btrfs_start_transaction(root, 0);
478 if (IS_ERR(trans)) {
479 ret = PTR_ERR(trans);
480 btrfs_subvolume_release_metadata(root, &block_rsv,
481 qgroup_reserved);
482 return ret;
483 }
484 trans->block_rsv = &block_rsv;
485 trans->bytes_reserved = block_rsv.size;
486
487 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
488 if (ret)
489 goto fail;
490
491 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
492 if (IS_ERR(leaf)) {
493 ret = PTR_ERR(leaf);
494 goto fail;
495 }
496
497 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
498 btrfs_set_header_bytenr(leaf, leaf->start);
499 btrfs_set_header_generation(leaf, trans->transid);
500 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
501 btrfs_set_header_owner(leaf, objectid);
502
503 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
504 BTRFS_FSID_SIZE);
505 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
506 btrfs_header_chunk_tree_uuid(leaf),
507 BTRFS_UUID_SIZE);
508 btrfs_mark_buffer_dirty(leaf);
509
510 memset(&root_item, 0, sizeof(root_item));
511
512 inode_item = &root_item.inode;
513 btrfs_set_stack_inode_generation(inode_item, 1);
514 btrfs_set_stack_inode_size(inode_item, 3);
515 btrfs_set_stack_inode_nlink(inode_item, 1);
516 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
517 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
518
519 btrfs_set_root_flags(&root_item, 0);
520 btrfs_set_root_limit(&root_item, 0);
521 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
522
523 btrfs_set_root_bytenr(&root_item, leaf->start);
524 btrfs_set_root_generation(&root_item, trans->transid);
525 btrfs_set_root_level(&root_item, 0);
526 btrfs_set_root_refs(&root_item, 1);
527 btrfs_set_root_used(&root_item, leaf->len);
528 btrfs_set_root_last_snapshot(&root_item, 0);
529
530 btrfs_set_root_generation_v2(&root_item,
531 btrfs_root_generation(&root_item));
532 uuid_le_gen(&new_uuid);
533 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
534 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
535 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
536 root_item.ctime = root_item.otime;
537 btrfs_set_root_ctransid(&root_item, trans->transid);
538 btrfs_set_root_otransid(&root_item, trans->transid);
539
540 btrfs_tree_unlock(leaf);
541 free_extent_buffer(leaf);
542 leaf = NULL;
543
544 btrfs_set_root_dirid(&root_item, new_dirid);
545
546 key.objectid = objectid;
547 key.offset = 0;
548 key.type = BTRFS_ROOT_ITEM_KEY;
549 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
550 &root_item);
551 if (ret)
552 goto fail;
553
554 key.offset = (u64)-1;
555 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
556 if (IS_ERR(new_root)) {
557 ret = PTR_ERR(new_root);
558 btrfs_abort_transaction(trans, root, ret);
559 goto fail;
560 }
561
562 btrfs_record_root_in_trans(trans, new_root);
563
564 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
565 if (ret) {
566 /* We potentially lose an unused inode item here */
567 btrfs_abort_transaction(trans, root, ret);
568 goto fail;
569 }
570
571 /*
572 * insert the directory item
573 */
574 ret = btrfs_set_inode_index(dir, &index);
575 if (ret) {
576 btrfs_abort_transaction(trans, root, ret);
577 goto fail;
578 }
579
580 ret = btrfs_insert_dir_item(trans, root,
581 name, namelen, dir, &key,
582 BTRFS_FT_DIR, index);
583 if (ret) {
584 btrfs_abort_transaction(trans, root, ret);
585 goto fail;
586 }
587
588 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
589 ret = btrfs_update_inode(trans, root, dir);
590 BUG_ON(ret);
591
592 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
593 objectid, root->root_key.objectid,
594 btrfs_ino(dir), index, name, namelen);
595 BUG_ON(ret);
596
597 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
598 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
599 objectid);
600 if (ret)
601 btrfs_abort_transaction(trans, root, ret);
602
603 fail:
604 trans->block_rsv = NULL;
605 trans->bytes_reserved = 0;
606 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
607
608 if (async_transid) {
609 *async_transid = trans->transid;
610 err = btrfs_commit_transaction_async(trans, root, 1);
611 if (err)
612 err = btrfs_commit_transaction(trans, root);
613 } else {
614 err = btrfs_commit_transaction(trans, root);
615 }
616 if (err && !ret)
617 ret = err;
618
619 if (!ret) {
620 inode = btrfs_lookup_dentry(dir, dentry);
621 if (IS_ERR(inode))
622 return PTR_ERR(inode);
623 d_instantiate(dentry, inode);
624 }
625 return ret;
626 }
627
628 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
629 {
630 s64 writers;
631 DEFINE_WAIT(wait);
632
633 do {
634 prepare_to_wait(&root->subv_writers->wait, &wait,
635 TASK_UNINTERRUPTIBLE);
636
637 writers = percpu_counter_sum(&root->subv_writers->counter);
638 if (writers)
639 schedule();
640
641 finish_wait(&root->subv_writers->wait, &wait);
642 } while (writers);
643 }
644
645 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
646 struct dentry *dentry, char *name, int namelen,
647 u64 *async_transid, bool readonly,
648 struct btrfs_qgroup_inherit *inherit)
649 {
650 struct inode *inode;
651 struct btrfs_pending_snapshot *pending_snapshot;
652 struct btrfs_trans_handle *trans;
653 int ret;
654
655 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
656 return -EINVAL;
657
658 atomic_inc(&root->will_be_snapshoted);
659 smp_mb__after_atomic();
660 btrfs_wait_for_no_snapshoting_writes(root);
661
662 ret = btrfs_start_delalloc_inodes(root, 0);
663 if (ret)
664 goto out;
665
666 btrfs_wait_ordered_extents(root, -1);
667
668 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
669 if (!pending_snapshot) {
670 ret = -ENOMEM;
671 goto out;
672 }
673
674 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
675 BTRFS_BLOCK_RSV_TEMP);
676 /*
677 * 1 - parent dir inode
678 * 2 - dir entries
679 * 1 - root item
680 * 2 - root ref/backref
681 * 1 - root of snapshot
682 * 1 - UUID item
683 */
684 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
685 &pending_snapshot->block_rsv, 8,
686 &pending_snapshot->qgroup_reserved,
687 false);
688 if (ret)
689 goto free;
690
691 pending_snapshot->dentry = dentry;
692 pending_snapshot->root = root;
693 pending_snapshot->readonly = readonly;
694 pending_snapshot->dir = dir;
695 pending_snapshot->inherit = inherit;
696
697 trans = btrfs_start_transaction(root, 0);
698 if (IS_ERR(trans)) {
699 ret = PTR_ERR(trans);
700 goto fail;
701 }
702
703 spin_lock(&root->fs_info->trans_lock);
704 list_add(&pending_snapshot->list,
705 &trans->transaction->pending_snapshots);
706 spin_unlock(&root->fs_info->trans_lock);
707 if (async_transid) {
708 *async_transid = trans->transid;
709 ret = btrfs_commit_transaction_async(trans,
710 root->fs_info->extent_root, 1);
711 if (ret)
712 ret = btrfs_commit_transaction(trans, root);
713 } else {
714 ret = btrfs_commit_transaction(trans,
715 root->fs_info->extent_root);
716 }
717 if (ret)
718 goto fail;
719
720 ret = pending_snapshot->error;
721 if (ret)
722 goto fail;
723
724 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
725 if (ret)
726 goto fail;
727
728 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
729 if (IS_ERR(inode)) {
730 ret = PTR_ERR(inode);
731 goto fail;
732 }
733
734 d_instantiate(dentry, inode);
735 ret = 0;
736 fail:
737 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
738 &pending_snapshot->block_rsv,
739 pending_snapshot->qgroup_reserved);
740 free:
741 kfree(pending_snapshot);
742 out:
743 if (atomic_dec_and_test(&root->will_be_snapshoted))
744 wake_up_atomic_t(&root->will_be_snapshoted);
745 return ret;
746 }
747
748 /* copy of may_delete in fs/namei.c()
749 * Check whether we can remove a link victim from directory dir, check
750 * whether the type of victim is right.
751 * 1. We can't do it if dir is read-only (done in permission())
752 * 2. We should have write and exec permissions on dir
753 * 3. We can't remove anything from append-only dir
754 * 4. We can't do anything with immutable dir (done in permission())
755 * 5. If the sticky bit on dir is set we should either
756 * a. be owner of dir, or
757 * b. be owner of victim, or
758 * c. have CAP_FOWNER capability
759 * 6. If the victim is append-only or immutable we can't do antyhing with
760 * links pointing to it.
761 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
762 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
763 * 9. We can't remove a root or mountpoint.
764 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
765 * nfs_async_unlink().
766 */
767
768 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
769 {
770 int error;
771
772 if (d_really_is_negative(victim))
773 return -ENOENT;
774
775 BUG_ON(d_inode(victim->d_parent) != dir);
776 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
777
778 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
779 if (error)
780 return error;
781 if (IS_APPEND(dir))
782 return -EPERM;
783 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
784 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
785 return -EPERM;
786 if (isdir) {
787 if (!d_is_dir(victim))
788 return -ENOTDIR;
789 if (IS_ROOT(victim))
790 return -EBUSY;
791 } else if (d_is_dir(victim))
792 return -EISDIR;
793 if (IS_DEADDIR(dir))
794 return -ENOENT;
795 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
796 return -EBUSY;
797 return 0;
798 }
799
800 /* copy of may_create in fs/namei.c() */
801 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
802 {
803 if (d_really_is_positive(child))
804 return -EEXIST;
805 if (IS_DEADDIR(dir))
806 return -ENOENT;
807 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
808 }
809
810 /*
811 * Create a new subvolume below @parent. This is largely modeled after
812 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
813 * inside this filesystem so it's quite a bit simpler.
814 */
815 static noinline int btrfs_mksubvol(struct path *parent,
816 char *name, int namelen,
817 struct btrfs_root *snap_src,
818 u64 *async_transid, bool readonly,
819 struct btrfs_qgroup_inherit *inherit)
820 {
821 struct inode *dir = d_inode(parent->dentry);
822 struct dentry *dentry;
823 int error;
824
825 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
826 if (error == -EINTR)
827 return error;
828
829 dentry = lookup_one_len(name, parent->dentry, namelen);
830 error = PTR_ERR(dentry);
831 if (IS_ERR(dentry))
832 goto out_unlock;
833
834 error = -EEXIST;
835 if (d_really_is_positive(dentry))
836 goto out_dput;
837
838 error = btrfs_may_create(dir, dentry);
839 if (error)
840 goto out_dput;
841
842 /*
843 * even if this name doesn't exist, we may get hash collisions.
844 * check for them now when we can safely fail
845 */
846 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
847 dir->i_ino, name,
848 namelen);
849 if (error)
850 goto out_dput;
851
852 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
853
854 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
855 goto out_up_read;
856
857 if (snap_src) {
858 error = create_snapshot(snap_src, dir, dentry, name, namelen,
859 async_transid, readonly, inherit);
860 } else {
861 error = create_subvol(dir, dentry, name, namelen,
862 async_transid, inherit);
863 }
864 if (!error)
865 fsnotify_mkdir(dir, dentry);
866 out_up_read:
867 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
868 out_dput:
869 dput(dentry);
870 out_unlock:
871 mutex_unlock(&dir->i_mutex);
872 return error;
873 }
874
875 /*
876 * When we're defragging a range, we don't want to kick it off again
877 * if it is really just waiting for delalloc to send it down.
878 * If we find a nice big extent or delalloc range for the bytes in the
879 * file you want to defrag, we return 0 to let you know to skip this
880 * part of the file
881 */
882 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
883 {
884 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
885 struct extent_map *em = NULL;
886 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
887 u64 end;
888
889 read_lock(&em_tree->lock);
890 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
891 read_unlock(&em_tree->lock);
892
893 if (em) {
894 end = extent_map_end(em);
895 free_extent_map(em);
896 if (end - offset > thresh)
897 return 0;
898 }
899 /* if we already have a nice delalloc here, just stop */
900 thresh /= 2;
901 end = count_range_bits(io_tree, &offset, offset + thresh,
902 thresh, EXTENT_DELALLOC, 1);
903 if (end >= thresh)
904 return 0;
905 return 1;
906 }
907
908 /*
909 * helper function to walk through a file and find extents
910 * newer than a specific transid, and smaller than thresh.
911 *
912 * This is used by the defragging code to find new and small
913 * extents
914 */
915 static int find_new_extents(struct btrfs_root *root,
916 struct inode *inode, u64 newer_than,
917 u64 *off, u32 thresh)
918 {
919 struct btrfs_path *path;
920 struct btrfs_key min_key;
921 struct extent_buffer *leaf;
922 struct btrfs_file_extent_item *extent;
923 int type;
924 int ret;
925 u64 ino = btrfs_ino(inode);
926
927 path = btrfs_alloc_path();
928 if (!path)
929 return -ENOMEM;
930
931 min_key.objectid = ino;
932 min_key.type = BTRFS_EXTENT_DATA_KEY;
933 min_key.offset = *off;
934
935 while (1) {
936 ret = btrfs_search_forward(root, &min_key, path, newer_than);
937 if (ret != 0)
938 goto none;
939 process_slot:
940 if (min_key.objectid != ino)
941 goto none;
942 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
943 goto none;
944
945 leaf = path->nodes[0];
946 extent = btrfs_item_ptr(leaf, path->slots[0],
947 struct btrfs_file_extent_item);
948
949 type = btrfs_file_extent_type(leaf, extent);
950 if (type == BTRFS_FILE_EXTENT_REG &&
951 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
952 check_defrag_in_cache(inode, min_key.offset, thresh)) {
953 *off = min_key.offset;
954 btrfs_free_path(path);
955 return 0;
956 }
957
958 path->slots[0]++;
959 if (path->slots[0] < btrfs_header_nritems(leaf)) {
960 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
961 goto process_slot;
962 }
963
964 if (min_key.offset == (u64)-1)
965 goto none;
966
967 min_key.offset++;
968 btrfs_release_path(path);
969 }
970 none:
971 btrfs_free_path(path);
972 return -ENOENT;
973 }
974
975 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
976 {
977 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
978 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
979 struct extent_map *em;
980 u64 len = PAGE_CACHE_SIZE;
981
982 /*
983 * hopefully we have this extent in the tree already, try without
984 * the full extent lock
985 */
986 read_lock(&em_tree->lock);
987 em = lookup_extent_mapping(em_tree, start, len);
988 read_unlock(&em_tree->lock);
989
990 if (!em) {
991 struct extent_state *cached = NULL;
992 u64 end = start + len - 1;
993
994 /* get the big lock and read metadata off disk */
995 lock_extent_bits(io_tree, start, end, 0, &cached);
996 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
997 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
998
999 if (IS_ERR(em))
1000 return NULL;
1001 }
1002
1003 return em;
1004 }
1005
1006 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1007 {
1008 struct extent_map *next;
1009 bool ret = true;
1010
1011 /* this is the last extent */
1012 if (em->start + em->len >= i_size_read(inode))
1013 return false;
1014
1015 next = defrag_lookup_extent(inode, em->start + em->len);
1016 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1017 ret = false;
1018 else if ((em->block_start + em->block_len == next->block_start) &&
1019 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
1020 ret = false;
1021
1022 free_extent_map(next);
1023 return ret;
1024 }
1025
1026 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1027 u64 *last_len, u64 *skip, u64 *defrag_end,
1028 int compress)
1029 {
1030 struct extent_map *em;
1031 int ret = 1;
1032 bool next_mergeable = true;
1033 bool prev_mergeable = true;
1034
1035 /*
1036 * make sure that once we start defragging an extent, we keep on
1037 * defragging it
1038 */
1039 if (start < *defrag_end)
1040 return 1;
1041
1042 *skip = 0;
1043
1044 em = defrag_lookup_extent(inode, start);
1045 if (!em)
1046 return 0;
1047
1048 /* this will cover holes, and inline extents */
1049 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1050 ret = 0;
1051 goto out;
1052 }
1053
1054 if (!*defrag_end)
1055 prev_mergeable = false;
1056
1057 next_mergeable = defrag_check_next_extent(inode, em);
1058 /*
1059 * we hit a real extent, if it is big or the next extent is not a
1060 * real extent, don't bother defragging it
1061 */
1062 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1063 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1064 ret = 0;
1065 out:
1066 /*
1067 * last_len ends up being a counter of how many bytes we've defragged.
1068 * every time we choose not to defrag an extent, we reset *last_len
1069 * so that the next tiny extent will force a defrag.
1070 *
1071 * The end result of this is that tiny extents before a single big
1072 * extent will force at least part of that big extent to be defragged.
1073 */
1074 if (ret) {
1075 *defrag_end = extent_map_end(em);
1076 } else {
1077 *last_len = 0;
1078 *skip = extent_map_end(em);
1079 *defrag_end = 0;
1080 }
1081
1082 free_extent_map(em);
1083 return ret;
1084 }
1085
1086 /*
1087 * it doesn't do much good to defrag one or two pages
1088 * at a time. This pulls in a nice chunk of pages
1089 * to COW and defrag.
1090 *
1091 * It also makes sure the delalloc code has enough
1092 * dirty data to avoid making new small extents as part
1093 * of the defrag
1094 *
1095 * It's a good idea to start RA on this range
1096 * before calling this.
1097 */
1098 static int cluster_pages_for_defrag(struct inode *inode,
1099 struct page **pages,
1100 unsigned long start_index,
1101 unsigned long num_pages)
1102 {
1103 unsigned long file_end;
1104 u64 isize = i_size_read(inode);
1105 u64 page_start;
1106 u64 page_end;
1107 u64 page_cnt;
1108 int ret;
1109 int i;
1110 int i_done;
1111 struct btrfs_ordered_extent *ordered;
1112 struct extent_state *cached_state = NULL;
1113 struct extent_io_tree *tree;
1114 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1115
1116 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1117 if (!isize || start_index > file_end)
1118 return 0;
1119
1120 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1121
1122 ret = btrfs_delalloc_reserve_space(inode,
1123 start_index << PAGE_CACHE_SHIFT,
1124 page_cnt << PAGE_CACHE_SHIFT);
1125 if (ret)
1126 return ret;
1127 i_done = 0;
1128 tree = &BTRFS_I(inode)->io_tree;
1129
1130 /* step one, lock all the pages */
1131 for (i = 0; i < page_cnt; i++) {
1132 struct page *page;
1133 again:
1134 page = find_or_create_page(inode->i_mapping,
1135 start_index + i, mask);
1136 if (!page)
1137 break;
1138
1139 page_start = page_offset(page);
1140 page_end = page_start + PAGE_CACHE_SIZE - 1;
1141 while (1) {
1142 lock_extent_bits(tree, page_start, page_end,
1143 0, &cached_state);
1144 ordered = btrfs_lookup_ordered_extent(inode,
1145 page_start);
1146 unlock_extent_cached(tree, page_start, page_end,
1147 &cached_state, GFP_NOFS);
1148 if (!ordered)
1149 break;
1150
1151 unlock_page(page);
1152 btrfs_start_ordered_extent(inode, ordered, 1);
1153 btrfs_put_ordered_extent(ordered);
1154 lock_page(page);
1155 /*
1156 * we unlocked the page above, so we need check if
1157 * it was released or not.
1158 */
1159 if (page->mapping != inode->i_mapping) {
1160 unlock_page(page);
1161 page_cache_release(page);
1162 goto again;
1163 }
1164 }
1165
1166 if (!PageUptodate(page)) {
1167 btrfs_readpage(NULL, page);
1168 lock_page(page);
1169 if (!PageUptodate(page)) {
1170 unlock_page(page);
1171 page_cache_release(page);
1172 ret = -EIO;
1173 break;
1174 }
1175 }
1176
1177 if (page->mapping != inode->i_mapping) {
1178 unlock_page(page);
1179 page_cache_release(page);
1180 goto again;
1181 }
1182
1183 pages[i] = page;
1184 i_done++;
1185 }
1186 if (!i_done || ret)
1187 goto out;
1188
1189 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1190 goto out;
1191
1192 /*
1193 * so now we have a nice long stream of locked
1194 * and up to date pages, lets wait on them
1195 */
1196 for (i = 0; i < i_done; i++)
1197 wait_on_page_writeback(pages[i]);
1198
1199 page_start = page_offset(pages[0]);
1200 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1201
1202 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1203 page_start, page_end - 1, 0, &cached_state);
1204 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1205 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1206 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1207 &cached_state, GFP_NOFS);
1208
1209 if (i_done != page_cnt) {
1210 spin_lock(&BTRFS_I(inode)->lock);
1211 BTRFS_I(inode)->outstanding_extents++;
1212 spin_unlock(&BTRFS_I(inode)->lock);
1213 btrfs_delalloc_release_space(inode,
1214 start_index << PAGE_CACHE_SHIFT,
1215 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1216 }
1217
1218
1219 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1220 &cached_state, GFP_NOFS);
1221
1222 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1223 page_start, page_end - 1, &cached_state,
1224 GFP_NOFS);
1225
1226 for (i = 0; i < i_done; i++) {
1227 clear_page_dirty_for_io(pages[i]);
1228 ClearPageChecked(pages[i]);
1229 set_page_extent_mapped(pages[i]);
1230 set_page_dirty(pages[i]);
1231 unlock_page(pages[i]);
1232 page_cache_release(pages[i]);
1233 }
1234 return i_done;
1235 out:
1236 for (i = 0; i < i_done; i++) {
1237 unlock_page(pages[i]);
1238 page_cache_release(pages[i]);
1239 }
1240 btrfs_delalloc_release_space(inode,
1241 start_index << PAGE_CACHE_SHIFT,
1242 page_cnt << PAGE_CACHE_SHIFT);
1243 return ret;
1244
1245 }
1246
1247 int btrfs_defrag_file(struct inode *inode, struct file *file,
1248 struct btrfs_ioctl_defrag_range_args *range,
1249 u64 newer_than, unsigned long max_to_defrag)
1250 {
1251 struct btrfs_root *root = BTRFS_I(inode)->root;
1252 struct file_ra_state *ra = NULL;
1253 unsigned long last_index;
1254 u64 isize = i_size_read(inode);
1255 u64 last_len = 0;
1256 u64 skip = 0;
1257 u64 defrag_end = 0;
1258 u64 newer_off = range->start;
1259 unsigned long i;
1260 unsigned long ra_index = 0;
1261 int ret;
1262 int defrag_count = 0;
1263 int compress_type = BTRFS_COMPRESS_ZLIB;
1264 u32 extent_thresh = range->extent_thresh;
1265 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1266 unsigned long cluster = max_cluster;
1267 u64 new_align = ~((u64)128 * 1024 - 1);
1268 struct page **pages = NULL;
1269
1270 if (isize == 0)
1271 return 0;
1272
1273 if (range->start >= isize)
1274 return -EINVAL;
1275
1276 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1277 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1278 return -EINVAL;
1279 if (range->compress_type)
1280 compress_type = range->compress_type;
1281 }
1282
1283 if (extent_thresh == 0)
1284 extent_thresh = 256 * 1024;
1285
1286 /*
1287 * if we were not given a file, allocate a readahead
1288 * context
1289 */
1290 if (!file) {
1291 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1292 if (!ra)
1293 return -ENOMEM;
1294 file_ra_state_init(ra, inode->i_mapping);
1295 } else {
1296 ra = &file->f_ra;
1297 }
1298
1299 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1300 GFP_NOFS);
1301 if (!pages) {
1302 ret = -ENOMEM;
1303 goto out_ra;
1304 }
1305
1306 /* find the last page to defrag */
1307 if (range->start + range->len > range->start) {
1308 last_index = min_t(u64, isize - 1,
1309 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1310 } else {
1311 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1312 }
1313
1314 if (newer_than) {
1315 ret = find_new_extents(root, inode, newer_than,
1316 &newer_off, 64 * 1024);
1317 if (!ret) {
1318 range->start = newer_off;
1319 /*
1320 * we always align our defrag to help keep
1321 * the extents in the file evenly spaced
1322 */
1323 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1324 } else
1325 goto out_ra;
1326 } else {
1327 i = range->start >> PAGE_CACHE_SHIFT;
1328 }
1329 if (!max_to_defrag)
1330 max_to_defrag = last_index - i + 1;
1331
1332 /*
1333 * make writeback starts from i, so the defrag range can be
1334 * written sequentially.
1335 */
1336 if (i < inode->i_mapping->writeback_index)
1337 inode->i_mapping->writeback_index = i;
1338
1339 while (i <= last_index && defrag_count < max_to_defrag &&
1340 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
1341 /*
1342 * make sure we stop running if someone unmounts
1343 * the FS
1344 */
1345 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1346 break;
1347
1348 if (btrfs_defrag_cancelled(root->fs_info)) {
1349 btrfs_debug(root->fs_info, "defrag_file cancelled");
1350 ret = -EAGAIN;
1351 break;
1352 }
1353
1354 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1355 extent_thresh, &last_len, &skip,
1356 &defrag_end, range->flags &
1357 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1358 unsigned long next;
1359 /*
1360 * the should_defrag function tells us how much to skip
1361 * bump our counter by the suggested amount
1362 */
1363 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
1364 i = max(i + 1, next);
1365 continue;
1366 }
1367
1368 if (!newer_than) {
1369 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1370 PAGE_CACHE_SHIFT) - i;
1371 cluster = min(cluster, max_cluster);
1372 } else {
1373 cluster = max_cluster;
1374 }
1375
1376 if (i + cluster > ra_index) {
1377 ra_index = max(i, ra_index);
1378 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1379 cluster);
1380 ra_index += cluster;
1381 }
1382
1383 mutex_lock(&inode->i_mutex);
1384 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1385 BTRFS_I(inode)->force_compress = compress_type;
1386 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1387 if (ret < 0) {
1388 mutex_unlock(&inode->i_mutex);
1389 goto out_ra;
1390 }
1391
1392 defrag_count += ret;
1393 balance_dirty_pages_ratelimited(inode->i_mapping);
1394 mutex_unlock(&inode->i_mutex);
1395
1396 if (newer_than) {
1397 if (newer_off == (u64)-1)
1398 break;
1399
1400 if (ret > 0)
1401 i += ret;
1402
1403 newer_off = max(newer_off + 1,
1404 (u64)i << PAGE_CACHE_SHIFT);
1405
1406 ret = find_new_extents(root, inode,
1407 newer_than, &newer_off,
1408 64 * 1024);
1409 if (!ret) {
1410 range->start = newer_off;
1411 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1412 } else {
1413 break;
1414 }
1415 } else {
1416 if (ret > 0) {
1417 i += ret;
1418 last_len += ret << PAGE_CACHE_SHIFT;
1419 } else {
1420 i++;
1421 last_len = 0;
1422 }
1423 }
1424 }
1425
1426 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1427 filemap_flush(inode->i_mapping);
1428 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1429 &BTRFS_I(inode)->runtime_flags))
1430 filemap_flush(inode->i_mapping);
1431 }
1432
1433 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1434 /* the filemap_flush will queue IO into the worker threads, but
1435 * we have to make sure the IO is actually started and that
1436 * ordered extents get created before we return
1437 */
1438 atomic_inc(&root->fs_info->async_submit_draining);
1439 while (atomic_read(&root->fs_info->nr_async_submits) ||
1440 atomic_read(&root->fs_info->async_delalloc_pages)) {
1441 wait_event(root->fs_info->async_submit_wait,
1442 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1443 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1444 }
1445 atomic_dec(&root->fs_info->async_submit_draining);
1446 }
1447
1448 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1449 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1450 }
1451
1452 ret = defrag_count;
1453
1454 out_ra:
1455 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1456 mutex_lock(&inode->i_mutex);
1457 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1458 mutex_unlock(&inode->i_mutex);
1459 }
1460 if (!file)
1461 kfree(ra);
1462 kfree(pages);
1463 return ret;
1464 }
1465
1466 static noinline int btrfs_ioctl_resize(struct file *file,
1467 void __user *arg)
1468 {
1469 u64 new_size;
1470 u64 old_size;
1471 u64 devid = 1;
1472 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1473 struct btrfs_ioctl_vol_args *vol_args;
1474 struct btrfs_trans_handle *trans;
1475 struct btrfs_device *device = NULL;
1476 char *sizestr;
1477 char *retptr;
1478 char *devstr = NULL;
1479 int ret = 0;
1480 int mod = 0;
1481
1482 if (!capable(CAP_SYS_ADMIN))
1483 return -EPERM;
1484
1485 ret = mnt_want_write_file(file);
1486 if (ret)
1487 return ret;
1488
1489 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1490 1)) {
1491 mnt_drop_write_file(file);
1492 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1493 }
1494
1495 mutex_lock(&root->fs_info->volume_mutex);
1496 vol_args = memdup_user(arg, sizeof(*vol_args));
1497 if (IS_ERR(vol_args)) {
1498 ret = PTR_ERR(vol_args);
1499 goto out;
1500 }
1501
1502 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1503
1504 sizestr = vol_args->name;
1505 devstr = strchr(sizestr, ':');
1506 if (devstr) {
1507 sizestr = devstr + 1;
1508 *devstr = '\0';
1509 devstr = vol_args->name;
1510 ret = kstrtoull(devstr, 10, &devid);
1511 if (ret)
1512 goto out_free;
1513 if (!devid) {
1514 ret = -EINVAL;
1515 goto out_free;
1516 }
1517 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1518 }
1519
1520 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1521 if (!device) {
1522 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1523 devid);
1524 ret = -ENODEV;
1525 goto out_free;
1526 }
1527
1528 if (!device->writeable) {
1529 btrfs_info(root->fs_info,
1530 "resizer unable to apply on readonly device %llu",
1531 devid);
1532 ret = -EPERM;
1533 goto out_free;
1534 }
1535
1536 if (!strcmp(sizestr, "max"))
1537 new_size = device->bdev->bd_inode->i_size;
1538 else {
1539 if (sizestr[0] == '-') {
1540 mod = -1;
1541 sizestr++;
1542 } else if (sizestr[0] == '+') {
1543 mod = 1;
1544 sizestr++;
1545 }
1546 new_size = memparse(sizestr, &retptr);
1547 if (*retptr != '\0' || new_size == 0) {
1548 ret = -EINVAL;
1549 goto out_free;
1550 }
1551 }
1552
1553 if (device->is_tgtdev_for_dev_replace) {
1554 ret = -EPERM;
1555 goto out_free;
1556 }
1557
1558 old_size = btrfs_device_get_total_bytes(device);
1559
1560 if (mod < 0) {
1561 if (new_size > old_size) {
1562 ret = -EINVAL;
1563 goto out_free;
1564 }
1565 new_size = old_size - new_size;
1566 } else if (mod > 0) {
1567 if (new_size > ULLONG_MAX - old_size) {
1568 ret = -ERANGE;
1569 goto out_free;
1570 }
1571 new_size = old_size + new_size;
1572 }
1573
1574 if (new_size < 256 * 1024 * 1024) {
1575 ret = -EINVAL;
1576 goto out_free;
1577 }
1578 if (new_size > device->bdev->bd_inode->i_size) {
1579 ret = -EFBIG;
1580 goto out_free;
1581 }
1582
1583 new_size = div_u64(new_size, root->sectorsize);
1584 new_size *= root->sectorsize;
1585
1586 btrfs_info_in_rcu(root->fs_info, "new size for %s is %llu",
1587 rcu_str_deref(device->name), new_size);
1588
1589 if (new_size > old_size) {
1590 trans = btrfs_start_transaction(root, 0);
1591 if (IS_ERR(trans)) {
1592 ret = PTR_ERR(trans);
1593 goto out_free;
1594 }
1595 ret = btrfs_grow_device(trans, device, new_size);
1596 btrfs_commit_transaction(trans, root);
1597 } else if (new_size < old_size) {
1598 ret = btrfs_shrink_device(device, new_size);
1599 } /* equal, nothing need to do */
1600
1601 out_free:
1602 kfree(vol_args);
1603 out:
1604 mutex_unlock(&root->fs_info->volume_mutex);
1605 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1606 mnt_drop_write_file(file);
1607 return ret;
1608 }
1609
1610 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1611 char *name, unsigned long fd, int subvol,
1612 u64 *transid, bool readonly,
1613 struct btrfs_qgroup_inherit *inherit)
1614 {
1615 int namelen;
1616 int ret = 0;
1617
1618 ret = mnt_want_write_file(file);
1619 if (ret)
1620 goto out;
1621
1622 namelen = strlen(name);
1623 if (strchr(name, '/')) {
1624 ret = -EINVAL;
1625 goto out_drop_write;
1626 }
1627
1628 if (name[0] == '.' &&
1629 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1630 ret = -EEXIST;
1631 goto out_drop_write;
1632 }
1633
1634 if (subvol) {
1635 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1636 NULL, transid, readonly, inherit);
1637 } else {
1638 struct fd src = fdget(fd);
1639 struct inode *src_inode;
1640 if (!src.file) {
1641 ret = -EINVAL;
1642 goto out_drop_write;
1643 }
1644
1645 src_inode = file_inode(src.file);
1646 if (src_inode->i_sb != file_inode(file)->i_sb) {
1647 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1648 "Snapshot src from another FS");
1649 ret = -EXDEV;
1650 } else if (!inode_owner_or_capable(src_inode)) {
1651 /*
1652 * Subvolume creation is not restricted, but snapshots
1653 * are limited to own subvolumes only
1654 */
1655 ret = -EPERM;
1656 } else {
1657 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1658 BTRFS_I(src_inode)->root,
1659 transid, readonly, inherit);
1660 }
1661 fdput(src);
1662 }
1663 out_drop_write:
1664 mnt_drop_write_file(file);
1665 out:
1666 return ret;
1667 }
1668
1669 static noinline int btrfs_ioctl_snap_create(struct file *file,
1670 void __user *arg, int subvol)
1671 {
1672 struct btrfs_ioctl_vol_args *vol_args;
1673 int ret;
1674
1675 vol_args = memdup_user(arg, sizeof(*vol_args));
1676 if (IS_ERR(vol_args))
1677 return PTR_ERR(vol_args);
1678 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1679
1680 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1681 vol_args->fd, subvol,
1682 NULL, false, NULL);
1683
1684 kfree(vol_args);
1685 return ret;
1686 }
1687
1688 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1689 void __user *arg, int subvol)
1690 {
1691 struct btrfs_ioctl_vol_args_v2 *vol_args;
1692 int ret;
1693 u64 transid = 0;
1694 u64 *ptr = NULL;
1695 bool readonly = false;
1696 struct btrfs_qgroup_inherit *inherit = NULL;
1697
1698 vol_args = memdup_user(arg, sizeof(*vol_args));
1699 if (IS_ERR(vol_args))
1700 return PTR_ERR(vol_args);
1701 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1702
1703 if (vol_args->flags &
1704 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1705 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1706 ret = -EOPNOTSUPP;
1707 goto free_args;
1708 }
1709
1710 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1711 ptr = &transid;
1712 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1713 readonly = true;
1714 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1715 if (vol_args->size > PAGE_CACHE_SIZE) {
1716 ret = -EINVAL;
1717 goto free_args;
1718 }
1719 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1720 if (IS_ERR(inherit)) {
1721 ret = PTR_ERR(inherit);
1722 goto free_args;
1723 }
1724 }
1725
1726 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1727 vol_args->fd, subvol, ptr,
1728 readonly, inherit);
1729 if (ret)
1730 goto free_inherit;
1731
1732 if (ptr && copy_to_user(arg +
1733 offsetof(struct btrfs_ioctl_vol_args_v2,
1734 transid),
1735 ptr, sizeof(*ptr)))
1736 ret = -EFAULT;
1737
1738 free_inherit:
1739 kfree(inherit);
1740 free_args:
1741 kfree(vol_args);
1742 return ret;
1743 }
1744
1745 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1746 void __user *arg)
1747 {
1748 struct inode *inode = file_inode(file);
1749 struct btrfs_root *root = BTRFS_I(inode)->root;
1750 int ret = 0;
1751 u64 flags = 0;
1752
1753 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1754 return -EINVAL;
1755
1756 down_read(&root->fs_info->subvol_sem);
1757 if (btrfs_root_readonly(root))
1758 flags |= BTRFS_SUBVOL_RDONLY;
1759 up_read(&root->fs_info->subvol_sem);
1760
1761 if (copy_to_user(arg, &flags, sizeof(flags)))
1762 ret = -EFAULT;
1763
1764 return ret;
1765 }
1766
1767 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1768 void __user *arg)
1769 {
1770 struct inode *inode = file_inode(file);
1771 struct btrfs_root *root = BTRFS_I(inode)->root;
1772 struct btrfs_trans_handle *trans;
1773 u64 root_flags;
1774 u64 flags;
1775 int ret = 0;
1776
1777 if (!inode_owner_or_capable(inode))
1778 return -EPERM;
1779
1780 ret = mnt_want_write_file(file);
1781 if (ret)
1782 goto out;
1783
1784 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1785 ret = -EINVAL;
1786 goto out_drop_write;
1787 }
1788
1789 if (copy_from_user(&flags, arg, sizeof(flags))) {
1790 ret = -EFAULT;
1791 goto out_drop_write;
1792 }
1793
1794 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1795 ret = -EINVAL;
1796 goto out_drop_write;
1797 }
1798
1799 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1800 ret = -EOPNOTSUPP;
1801 goto out_drop_write;
1802 }
1803
1804 down_write(&root->fs_info->subvol_sem);
1805
1806 /* nothing to do */
1807 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1808 goto out_drop_sem;
1809
1810 root_flags = btrfs_root_flags(&root->root_item);
1811 if (flags & BTRFS_SUBVOL_RDONLY) {
1812 btrfs_set_root_flags(&root->root_item,
1813 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1814 } else {
1815 /*
1816 * Block RO -> RW transition if this subvolume is involved in
1817 * send
1818 */
1819 spin_lock(&root->root_item_lock);
1820 if (root->send_in_progress == 0) {
1821 btrfs_set_root_flags(&root->root_item,
1822 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1823 spin_unlock(&root->root_item_lock);
1824 } else {
1825 spin_unlock(&root->root_item_lock);
1826 btrfs_warn(root->fs_info,
1827 "Attempt to set subvolume %llu read-write during send",
1828 root->root_key.objectid);
1829 ret = -EPERM;
1830 goto out_drop_sem;
1831 }
1832 }
1833
1834 trans = btrfs_start_transaction(root, 1);
1835 if (IS_ERR(trans)) {
1836 ret = PTR_ERR(trans);
1837 goto out_reset;
1838 }
1839
1840 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1841 &root->root_key, &root->root_item);
1842
1843 btrfs_commit_transaction(trans, root);
1844 out_reset:
1845 if (ret)
1846 btrfs_set_root_flags(&root->root_item, root_flags);
1847 out_drop_sem:
1848 up_write(&root->fs_info->subvol_sem);
1849 out_drop_write:
1850 mnt_drop_write_file(file);
1851 out:
1852 return ret;
1853 }
1854
1855 /*
1856 * helper to check if the subvolume references other subvolumes
1857 */
1858 static noinline int may_destroy_subvol(struct btrfs_root *root)
1859 {
1860 struct btrfs_path *path;
1861 struct btrfs_dir_item *di;
1862 struct btrfs_key key;
1863 u64 dir_id;
1864 int ret;
1865
1866 path = btrfs_alloc_path();
1867 if (!path)
1868 return -ENOMEM;
1869
1870 /* Make sure this root isn't set as the default subvol */
1871 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1872 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1873 dir_id, "default", 7, 0);
1874 if (di && !IS_ERR(di)) {
1875 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1876 if (key.objectid == root->root_key.objectid) {
1877 ret = -EPERM;
1878 btrfs_err(root->fs_info, "deleting default subvolume "
1879 "%llu is not allowed", key.objectid);
1880 goto out;
1881 }
1882 btrfs_release_path(path);
1883 }
1884
1885 key.objectid = root->root_key.objectid;
1886 key.type = BTRFS_ROOT_REF_KEY;
1887 key.offset = (u64)-1;
1888
1889 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1890 &key, path, 0, 0);
1891 if (ret < 0)
1892 goto out;
1893 BUG_ON(ret == 0);
1894
1895 ret = 0;
1896 if (path->slots[0] > 0) {
1897 path->slots[0]--;
1898 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1899 if (key.objectid == root->root_key.objectid &&
1900 key.type == BTRFS_ROOT_REF_KEY)
1901 ret = -ENOTEMPTY;
1902 }
1903 out:
1904 btrfs_free_path(path);
1905 return ret;
1906 }
1907
1908 static noinline int key_in_sk(struct btrfs_key *key,
1909 struct btrfs_ioctl_search_key *sk)
1910 {
1911 struct btrfs_key test;
1912 int ret;
1913
1914 test.objectid = sk->min_objectid;
1915 test.type = sk->min_type;
1916 test.offset = sk->min_offset;
1917
1918 ret = btrfs_comp_cpu_keys(key, &test);
1919 if (ret < 0)
1920 return 0;
1921
1922 test.objectid = sk->max_objectid;
1923 test.type = sk->max_type;
1924 test.offset = sk->max_offset;
1925
1926 ret = btrfs_comp_cpu_keys(key, &test);
1927 if (ret > 0)
1928 return 0;
1929 return 1;
1930 }
1931
1932 static noinline int copy_to_sk(struct btrfs_root *root,
1933 struct btrfs_path *path,
1934 struct btrfs_key *key,
1935 struct btrfs_ioctl_search_key *sk,
1936 size_t *buf_size,
1937 char __user *ubuf,
1938 unsigned long *sk_offset,
1939 int *num_found)
1940 {
1941 u64 found_transid;
1942 struct extent_buffer *leaf;
1943 struct btrfs_ioctl_search_header sh;
1944 struct btrfs_key test;
1945 unsigned long item_off;
1946 unsigned long item_len;
1947 int nritems;
1948 int i;
1949 int slot;
1950 int ret = 0;
1951
1952 leaf = path->nodes[0];
1953 slot = path->slots[0];
1954 nritems = btrfs_header_nritems(leaf);
1955
1956 if (btrfs_header_generation(leaf) > sk->max_transid) {
1957 i = nritems;
1958 goto advance_key;
1959 }
1960 found_transid = btrfs_header_generation(leaf);
1961
1962 for (i = slot; i < nritems; i++) {
1963 item_off = btrfs_item_ptr_offset(leaf, i);
1964 item_len = btrfs_item_size_nr(leaf, i);
1965
1966 btrfs_item_key_to_cpu(leaf, key, i);
1967 if (!key_in_sk(key, sk))
1968 continue;
1969
1970 if (sizeof(sh) + item_len > *buf_size) {
1971 if (*num_found) {
1972 ret = 1;
1973 goto out;
1974 }
1975
1976 /*
1977 * return one empty item back for v1, which does not
1978 * handle -EOVERFLOW
1979 */
1980
1981 *buf_size = sizeof(sh) + item_len;
1982 item_len = 0;
1983 ret = -EOVERFLOW;
1984 }
1985
1986 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1987 ret = 1;
1988 goto out;
1989 }
1990
1991 sh.objectid = key->objectid;
1992 sh.offset = key->offset;
1993 sh.type = key->type;
1994 sh.len = item_len;
1995 sh.transid = found_transid;
1996
1997 /* copy search result header */
1998 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1999 ret = -EFAULT;
2000 goto out;
2001 }
2002
2003 *sk_offset += sizeof(sh);
2004
2005 if (item_len) {
2006 char __user *up = ubuf + *sk_offset;
2007 /* copy the item */
2008 if (read_extent_buffer_to_user(leaf, up,
2009 item_off, item_len)) {
2010 ret = -EFAULT;
2011 goto out;
2012 }
2013
2014 *sk_offset += item_len;
2015 }
2016 (*num_found)++;
2017
2018 if (ret) /* -EOVERFLOW from above */
2019 goto out;
2020
2021 if (*num_found >= sk->nr_items) {
2022 ret = 1;
2023 goto out;
2024 }
2025 }
2026 advance_key:
2027 ret = 0;
2028 test.objectid = sk->max_objectid;
2029 test.type = sk->max_type;
2030 test.offset = sk->max_offset;
2031 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2032 ret = 1;
2033 else if (key->offset < (u64)-1)
2034 key->offset++;
2035 else if (key->type < (u8)-1) {
2036 key->offset = 0;
2037 key->type++;
2038 } else if (key->objectid < (u64)-1) {
2039 key->offset = 0;
2040 key->type = 0;
2041 key->objectid++;
2042 } else
2043 ret = 1;
2044 out:
2045 /*
2046 * 0: all items from this leaf copied, continue with next
2047 * 1: * more items can be copied, but unused buffer is too small
2048 * * all items were found
2049 * Either way, it will stops the loop which iterates to the next
2050 * leaf
2051 * -EOVERFLOW: item was to large for buffer
2052 * -EFAULT: could not copy extent buffer back to userspace
2053 */
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 __user *ubuf)
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 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2072 return -EOVERFLOW;
2073 }
2074
2075 path = btrfs_alloc_path();
2076 if (!path)
2077 return -ENOMEM;
2078
2079 if (sk->tree_id == 0) {
2080 /* search the root of the inode that was passed */
2081 root = BTRFS_I(inode)->root;
2082 } else {
2083 key.objectid = sk->tree_id;
2084 key.type = BTRFS_ROOT_ITEM_KEY;
2085 key.offset = (u64)-1;
2086 root = btrfs_read_fs_root_no_name(info, &key);
2087 if (IS_ERR(root)) {
2088 btrfs_err(info, "could not find root %llu",
2089 sk->tree_id);
2090 btrfs_free_path(path);
2091 return -ENOENT;
2092 }
2093 }
2094
2095 key.objectid = sk->min_objectid;
2096 key.type = sk->min_type;
2097 key.offset = sk->min_offset;
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, ubuf,
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 __user *uargs;
2125 struct btrfs_ioctl_search_key sk;
2126 struct inode *inode;
2127 int ret;
2128 size_t buf_size;
2129
2130 if (!capable(CAP_SYS_ADMIN))
2131 return -EPERM;
2132
2133 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2134
2135 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2136 return -EFAULT;
2137
2138 buf_size = sizeof(uargs->buf);
2139
2140 inode = file_inode(file);
2141 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2142
2143 /*
2144 * In the origin implementation an overflow is handled by returning a
2145 * search header with a len of zero, so reset ret.
2146 */
2147 if (ret == -EOVERFLOW)
2148 ret = 0;
2149
2150 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2151 ret = -EFAULT;
2152 return ret;
2153 }
2154
2155 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2156 void __user *argp)
2157 {
2158 struct btrfs_ioctl_search_args_v2 __user *uarg;
2159 struct btrfs_ioctl_search_args_v2 args;
2160 struct inode *inode;
2161 int ret;
2162 size_t buf_size;
2163 const size_t buf_limit = 16 * 1024 * 1024;
2164
2165 if (!capable(CAP_SYS_ADMIN))
2166 return -EPERM;
2167
2168 /* copy search header and buffer size */
2169 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2170 if (copy_from_user(&args, uarg, sizeof(args)))
2171 return -EFAULT;
2172
2173 buf_size = args.buf_size;
2174
2175 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2176 return -EOVERFLOW;
2177
2178 /* limit result size to 16MB */
2179 if (buf_size > buf_limit)
2180 buf_size = buf_limit;
2181
2182 inode = file_inode(file);
2183 ret = search_ioctl(inode, &args.key, &buf_size,
2184 (char *)(&uarg->buf[0]));
2185 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2186 ret = -EFAULT;
2187 else if (ret == -EOVERFLOW &&
2188 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2189 ret = -EFAULT;
2190
2191 return ret;
2192 }
2193
2194 /*
2195 * Search INODE_REFs to identify path name of 'dirid' directory
2196 * in a 'tree_id' tree. and sets path name to 'name'.
2197 */
2198 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2199 u64 tree_id, u64 dirid, char *name)
2200 {
2201 struct btrfs_root *root;
2202 struct btrfs_key key;
2203 char *ptr;
2204 int ret = -1;
2205 int slot;
2206 int len;
2207 int total_len = 0;
2208 struct btrfs_inode_ref *iref;
2209 struct extent_buffer *l;
2210 struct btrfs_path *path;
2211
2212 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2213 name[0]='\0';
2214 return 0;
2215 }
2216
2217 path = btrfs_alloc_path();
2218 if (!path)
2219 return -ENOMEM;
2220
2221 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2222
2223 key.objectid = tree_id;
2224 key.type = BTRFS_ROOT_ITEM_KEY;
2225 key.offset = (u64)-1;
2226 root = btrfs_read_fs_root_no_name(info, &key);
2227 if (IS_ERR(root)) {
2228 btrfs_err(info, "could not find root %llu", tree_id);
2229 ret = -ENOENT;
2230 goto out;
2231 }
2232
2233 key.objectid = dirid;
2234 key.type = BTRFS_INODE_REF_KEY;
2235 key.offset = (u64)-1;
2236
2237 while (1) {
2238 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2239 if (ret < 0)
2240 goto out;
2241 else if (ret > 0) {
2242 ret = btrfs_previous_item(root, path, dirid,
2243 BTRFS_INODE_REF_KEY);
2244 if (ret < 0)
2245 goto out;
2246 else if (ret > 0) {
2247 ret = -ENOENT;
2248 goto out;
2249 }
2250 }
2251
2252 l = path->nodes[0];
2253 slot = path->slots[0];
2254 btrfs_item_key_to_cpu(l, &key, slot);
2255
2256 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2257 len = btrfs_inode_ref_name_len(l, iref);
2258 ptr -= len + 1;
2259 total_len += len + 1;
2260 if (ptr < name) {
2261 ret = -ENAMETOOLONG;
2262 goto out;
2263 }
2264
2265 *(ptr + len) = '/';
2266 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2267
2268 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2269 break;
2270
2271 btrfs_release_path(path);
2272 key.objectid = key.offset;
2273 key.offset = (u64)-1;
2274 dirid = key.objectid;
2275 }
2276 memmove(name, ptr, total_len);
2277 name[total_len] = '\0';
2278 ret = 0;
2279 out:
2280 btrfs_free_path(path);
2281 return ret;
2282 }
2283
2284 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2285 void __user *argp)
2286 {
2287 struct btrfs_ioctl_ino_lookup_args *args;
2288 struct inode *inode;
2289 int ret = 0;
2290
2291 args = memdup_user(argp, sizeof(*args));
2292 if (IS_ERR(args))
2293 return PTR_ERR(args);
2294
2295 inode = file_inode(file);
2296
2297 /*
2298 * Unprivileged query to obtain the containing subvolume root id. The
2299 * path is reset so it's consistent with btrfs_search_path_in_tree.
2300 */
2301 if (args->treeid == 0)
2302 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2303
2304 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2305 args->name[0] = 0;
2306 goto out;
2307 }
2308
2309 if (!capable(CAP_SYS_ADMIN)) {
2310 ret = -EPERM;
2311 goto out;
2312 }
2313
2314 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2315 args->treeid, args->objectid,
2316 args->name);
2317
2318 out:
2319 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2320 ret = -EFAULT;
2321
2322 kfree(args);
2323 return ret;
2324 }
2325
2326 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2327 void __user *arg)
2328 {
2329 struct dentry *parent = file->f_path.dentry;
2330 struct dentry *dentry;
2331 struct inode *dir = d_inode(parent);
2332 struct inode *inode;
2333 struct btrfs_root *root = BTRFS_I(dir)->root;
2334 struct btrfs_root *dest = NULL;
2335 struct btrfs_ioctl_vol_args *vol_args;
2336 struct btrfs_trans_handle *trans;
2337 struct btrfs_block_rsv block_rsv;
2338 u64 root_flags;
2339 u64 qgroup_reserved;
2340 int namelen;
2341 int ret;
2342 int err = 0;
2343
2344 vol_args = memdup_user(arg, sizeof(*vol_args));
2345 if (IS_ERR(vol_args))
2346 return PTR_ERR(vol_args);
2347
2348 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2349 namelen = strlen(vol_args->name);
2350 if (strchr(vol_args->name, '/') ||
2351 strncmp(vol_args->name, "..", namelen) == 0) {
2352 err = -EINVAL;
2353 goto out;
2354 }
2355
2356 err = mnt_want_write_file(file);
2357 if (err)
2358 goto out;
2359
2360
2361 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2362 if (err == -EINTR)
2363 goto out_drop_write;
2364 dentry = lookup_one_len(vol_args->name, parent, namelen);
2365 if (IS_ERR(dentry)) {
2366 err = PTR_ERR(dentry);
2367 goto out_unlock_dir;
2368 }
2369
2370 if (d_really_is_negative(dentry)) {
2371 err = -ENOENT;
2372 goto out_dput;
2373 }
2374
2375 inode = d_inode(dentry);
2376 dest = BTRFS_I(inode)->root;
2377 if (!capable(CAP_SYS_ADMIN)) {
2378 /*
2379 * Regular user. Only allow this with a special mount
2380 * option, when the user has write+exec access to the
2381 * subvol root, and when rmdir(2) would have been
2382 * allowed.
2383 *
2384 * Note that this is _not_ check that the subvol is
2385 * empty or doesn't contain data that we wouldn't
2386 * otherwise be able to delete.
2387 *
2388 * Users who want to delete empty subvols should try
2389 * rmdir(2).
2390 */
2391 err = -EPERM;
2392 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2393 goto out_dput;
2394
2395 /*
2396 * Do not allow deletion if the parent dir is the same
2397 * as the dir to be deleted. That means the ioctl
2398 * must be called on the dentry referencing the root
2399 * of the subvol, not a random directory contained
2400 * within it.
2401 */
2402 err = -EINVAL;
2403 if (root == dest)
2404 goto out_dput;
2405
2406 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2407 if (err)
2408 goto out_dput;
2409 }
2410
2411 /* check if subvolume may be deleted by a user */
2412 err = btrfs_may_delete(dir, dentry, 1);
2413 if (err)
2414 goto out_dput;
2415
2416 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2417 err = -EINVAL;
2418 goto out_dput;
2419 }
2420
2421 mutex_lock(&inode->i_mutex);
2422
2423 /*
2424 * Don't allow to delete a subvolume with send in progress. This is
2425 * inside the i_mutex so the error handling that has to drop the bit
2426 * again is not run concurrently.
2427 */
2428 spin_lock(&dest->root_item_lock);
2429 root_flags = btrfs_root_flags(&dest->root_item);
2430 if (dest->send_in_progress == 0) {
2431 btrfs_set_root_flags(&dest->root_item,
2432 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2433 spin_unlock(&dest->root_item_lock);
2434 } else {
2435 spin_unlock(&dest->root_item_lock);
2436 btrfs_warn(root->fs_info,
2437 "Attempt to delete subvolume %llu during send",
2438 dest->root_key.objectid);
2439 err = -EPERM;
2440 goto out_unlock_inode;
2441 }
2442
2443 down_write(&root->fs_info->subvol_sem);
2444
2445 err = may_destroy_subvol(dest);
2446 if (err)
2447 goto out_up_write;
2448
2449 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2450 /*
2451 * One for dir inode, two for dir entries, two for root
2452 * ref/backref.
2453 */
2454 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2455 5, &qgroup_reserved, true);
2456 if (err)
2457 goto out_up_write;
2458
2459 trans = btrfs_start_transaction(root, 0);
2460 if (IS_ERR(trans)) {
2461 err = PTR_ERR(trans);
2462 goto out_release;
2463 }
2464 trans->block_rsv = &block_rsv;
2465 trans->bytes_reserved = block_rsv.size;
2466
2467 ret = btrfs_unlink_subvol(trans, root, dir,
2468 dest->root_key.objectid,
2469 dentry->d_name.name,
2470 dentry->d_name.len);
2471 if (ret) {
2472 err = ret;
2473 btrfs_abort_transaction(trans, root, ret);
2474 goto out_end_trans;
2475 }
2476
2477 btrfs_record_root_in_trans(trans, dest);
2478
2479 memset(&dest->root_item.drop_progress, 0,
2480 sizeof(dest->root_item.drop_progress));
2481 dest->root_item.drop_level = 0;
2482 btrfs_set_root_refs(&dest->root_item, 0);
2483
2484 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2485 ret = btrfs_insert_orphan_item(trans,
2486 root->fs_info->tree_root,
2487 dest->root_key.objectid);
2488 if (ret) {
2489 btrfs_abort_transaction(trans, root, ret);
2490 err = ret;
2491 goto out_end_trans;
2492 }
2493 }
2494
2495 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2496 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2497 dest->root_key.objectid);
2498 if (ret && ret != -ENOENT) {
2499 btrfs_abort_transaction(trans, root, ret);
2500 err = ret;
2501 goto out_end_trans;
2502 }
2503 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2504 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2505 dest->root_item.received_uuid,
2506 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2507 dest->root_key.objectid);
2508 if (ret && ret != -ENOENT) {
2509 btrfs_abort_transaction(trans, root, ret);
2510 err = ret;
2511 goto out_end_trans;
2512 }
2513 }
2514
2515 out_end_trans:
2516 trans->block_rsv = NULL;
2517 trans->bytes_reserved = 0;
2518 ret = btrfs_end_transaction(trans, root);
2519 if (ret && !err)
2520 err = ret;
2521 inode->i_flags |= S_DEAD;
2522 out_release:
2523 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2524 out_up_write:
2525 up_write(&root->fs_info->subvol_sem);
2526 if (err) {
2527 spin_lock(&dest->root_item_lock);
2528 root_flags = btrfs_root_flags(&dest->root_item);
2529 btrfs_set_root_flags(&dest->root_item,
2530 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2531 spin_unlock(&dest->root_item_lock);
2532 }
2533 out_unlock_inode:
2534 mutex_unlock(&inode->i_mutex);
2535 if (!err) {
2536 d_invalidate(dentry);
2537 btrfs_invalidate_inodes(dest);
2538 d_delete(dentry);
2539 ASSERT(dest->send_in_progress == 0);
2540
2541 /* the last ref */
2542 if (dest->ino_cache_inode) {
2543 iput(dest->ino_cache_inode);
2544 dest->ino_cache_inode = NULL;
2545 }
2546 }
2547 out_dput:
2548 dput(dentry);
2549 out_unlock_dir:
2550 mutex_unlock(&dir->i_mutex);
2551 out_drop_write:
2552 mnt_drop_write_file(file);
2553 out:
2554 kfree(vol_args);
2555 return err;
2556 }
2557
2558 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2559 {
2560 struct inode *inode = file_inode(file);
2561 struct btrfs_root *root = BTRFS_I(inode)->root;
2562 struct btrfs_ioctl_defrag_range_args *range;
2563 int ret;
2564
2565 ret = mnt_want_write_file(file);
2566 if (ret)
2567 return ret;
2568
2569 if (btrfs_root_readonly(root)) {
2570 ret = -EROFS;
2571 goto out;
2572 }
2573
2574 switch (inode->i_mode & S_IFMT) {
2575 case S_IFDIR:
2576 if (!capable(CAP_SYS_ADMIN)) {
2577 ret = -EPERM;
2578 goto out;
2579 }
2580 ret = btrfs_defrag_root(root);
2581 if (ret)
2582 goto out;
2583 ret = btrfs_defrag_root(root->fs_info->extent_root);
2584 break;
2585 case S_IFREG:
2586 if (!(file->f_mode & FMODE_WRITE)) {
2587 ret = -EINVAL;
2588 goto out;
2589 }
2590
2591 range = kzalloc(sizeof(*range), GFP_KERNEL);
2592 if (!range) {
2593 ret = -ENOMEM;
2594 goto out;
2595 }
2596
2597 if (argp) {
2598 if (copy_from_user(range, argp,
2599 sizeof(*range))) {
2600 ret = -EFAULT;
2601 kfree(range);
2602 goto out;
2603 }
2604 /* compression requires us to start the IO */
2605 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2606 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2607 range->extent_thresh = (u32)-1;
2608 }
2609 } else {
2610 /* the rest are all set to zero by kzalloc */
2611 range->len = (u64)-1;
2612 }
2613 ret = btrfs_defrag_file(file_inode(file), file,
2614 range, 0, 0);
2615 if (ret > 0)
2616 ret = 0;
2617 kfree(range);
2618 break;
2619 default:
2620 ret = -EINVAL;
2621 }
2622 out:
2623 mnt_drop_write_file(file);
2624 return ret;
2625 }
2626
2627 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2628 {
2629 struct btrfs_ioctl_vol_args *vol_args;
2630 int ret;
2631
2632 if (!capable(CAP_SYS_ADMIN))
2633 return -EPERM;
2634
2635 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2636 1)) {
2637 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2638 }
2639
2640 mutex_lock(&root->fs_info->volume_mutex);
2641 vol_args = memdup_user(arg, sizeof(*vol_args));
2642 if (IS_ERR(vol_args)) {
2643 ret = PTR_ERR(vol_args);
2644 goto out;
2645 }
2646
2647 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2648 ret = btrfs_init_new_device(root, vol_args->name);
2649
2650 if (!ret)
2651 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2652
2653 kfree(vol_args);
2654 out:
2655 mutex_unlock(&root->fs_info->volume_mutex);
2656 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2657 return ret;
2658 }
2659
2660 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2661 {
2662 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2663 struct btrfs_ioctl_vol_args *vol_args;
2664 int ret;
2665
2666 if (!capable(CAP_SYS_ADMIN))
2667 return -EPERM;
2668
2669 ret = mnt_want_write_file(file);
2670 if (ret)
2671 return ret;
2672
2673 vol_args = memdup_user(arg, sizeof(*vol_args));
2674 if (IS_ERR(vol_args)) {
2675 ret = PTR_ERR(vol_args);
2676 goto err_drop;
2677 }
2678
2679 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2680
2681 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2682 1)) {
2683 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2684 goto out;
2685 }
2686
2687 mutex_lock(&root->fs_info->volume_mutex);
2688 ret = btrfs_rm_device(root, vol_args->name);
2689 mutex_unlock(&root->fs_info->volume_mutex);
2690 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2691
2692 if (!ret)
2693 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2694
2695 out:
2696 kfree(vol_args);
2697 err_drop:
2698 mnt_drop_write_file(file);
2699 return ret;
2700 }
2701
2702 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2703 {
2704 struct btrfs_ioctl_fs_info_args *fi_args;
2705 struct btrfs_device *device;
2706 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2707 int ret = 0;
2708
2709 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2710 if (!fi_args)
2711 return -ENOMEM;
2712
2713 mutex_lock(&fs_devices->device_list_mutex);
2714 fi_args->num_devices = fs_devices->num_devices;
2715 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2716
2717 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2718 if (device->devid > fi_args->max_id)
2719 fi_args->max_id = device->devid;
2720 }
2721 mutex_unlock(&fs_devices->device_list_mutex);
2722
2723 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2724 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2725 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2726
2727 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2728 ret = -EFAULT;
2729
2730 kfree(fi_args);
2731 return ret;
2732 }
2733
2734 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2735 {
2736 struct btrfs_ioctl_dev_info_args *di_args;
2737 struct btrfs_device *dev;
2738 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2739 int ret = 0;
2740 char *s_uuid = NULL;
2741
2742 di_args = memdup_user(arg, sizeof(*di_args));
2743 if (IS_ERR(di_args))
2744 return PTR_ERR(di_args);
2745
2746 if (!btrfs_is_empty_uuid(di_args->uuid))
2747 s_uuid = di_args->uuid;
2748
2749 mutex_lock(&fs_devices->device_list_mutex);
2750 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2751
2752 if (!dev) {
2753 ret = -ENODEV;
2754 goto out;
2755 }
2756
2757 di_args->devid = dev->devid;
2758 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2759 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2760 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2761 if (dev->name) {
2762 struct rcu_string *name;
2763
2764 rcu_read_lock();
2765 name = rcu_dereference(dev->name);
2766 strncpy(di_args->path, name->str, sizeof(di_args->path));
2767 rcu_read_unlock();
2768 di_args->path[sizeof(di_args->path) - 1] = 0;
2769 } else {
2770 di_args->path[0] = '\0';
2771 }
2772
2773 out:
2774 mutex_unlock(&fs_devices->device_list_mutex);
2775 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2776 ret = -EFAULT;
2777
2778 kfree(di_args);
2779 return ret;
2780 }
2781
2782 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2783 {
2784 struct page *page;
2785 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2786
2787 page = grab_cache_page(inode->i_mapping, index);
2788 if (!page)
2789 return NULL;
2790
2791 if (!PageUptodate(page)) {
2792 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2793 0))
2794 return NULL;
2795 lock_page(page);
2796 if (!PageUptodate(page)) {
2797 unlock_page(page);
2798 page_cache_release(page);
2799 return NULL;
2800 }
2801 }
2802 unlock_page(page);
2803
2804 return page;
2805 }
2806
2807 static int gather_extent_pages(struct inode *inode, struct page **pages,
2808 int num_pages, u64 off)
2809 {
2810 int i;
2811 pgoff_t index = off >> PAGE_CACHE_SHIFT;
2812
2813 for (i = 0; i < num_pages; i++) {
2814 pages[i] = extent_same_get_page(inode, index + i);
2815 if (!pages[i])
2816 return -ENOMEM;
2817 }
2818 return 0;
2819 }
2820
2821 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2822 {
2823 /* do any pending delalloc/csum calc on src, one way or
2824 another, and lock file content */
2825 while (1) {
2826 struct btrfs_ordered_extent *ordered;
2827 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2828 ordered = btrfs_lookup_first_ordered_extent(inode,
2829 off + len - 1);
2830 if ((!ordered ||
2831 ordered->file_offset + ordered->len <= off ||
2832 ordered->file_offset >= off + len) &&
2833 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2834 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2835 if (ordered)
2836 btrfs_put_ordered_extent(ordered);
2837 break;
2838 }
2839 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2840 if (ordered)
2841 btrfs_put_ordered_extent(ordered);
2842 btrfs_wait_ordered_range(inode, off, len);
2843 }
2844 }
2845
2846 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2847 {
2848 mutex_unlock(&inode1->i_mutex);
2849 mutex_unlock(&inode2->i_mutex);
2850 }
2851
2852 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2853 {
2854 if (inode1 < inode2)
2855 swap(inode1, inode2);
2856
2857 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2858 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2859 }
2860
2861 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2862 struct inode *inode2, u64 loff2, u64 len)
2863 {
2864 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2865 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2866 }
2867
2868 static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2869 struct inode *inode2, u64 loff2, u64 len)
2870 {
2871 if (inode1 < inode2) {
2872 swap(inode1, inode2);
2873 swap(loff1, loff2);
2874 }
2875 lock_extent_range(inode1, loff1, len);
2876 lock_extent_range(inode2, loff2, len);
2877 }
2878
2879 struct cmp_pages {
2880 int num_pages;
2881 struct page **src_pages;
2882 struct page **dst_pages;
2883 };
2884
2885 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2886 {
2887 int i;
2888 struct page *pg;
2889
2890 for (i = 0; i < cmp->num_pages; i++) {
2891 pg = cmp->src_pages[i];
2892 if (pg)
2893 page_cache_release(pg);
2894 pg = cmp->dst_pages[i];
2895 if (pg)
2896 page_cache_release(pg);
2897 }
2898 kfree(cmp->src_pages);
2899 kfree(cmp->dst_pages);
2900 }
2901
2902 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2903 struct inode *dst, u64 dst_loff,
2904 u64 len, struct cmp_pages *cmp)
2905 {
2906 int ret;
2907 int num_pages = PAGE_CACHE_ALIGN(len) >> PAGE_CACHE_SHIFT;
2908 struct page **src_pgarr, **dst_pgarr;
2909
2910 /*
2911 * We must gather up all the pages before we initiate our
2912 * extent locking. We use an array for the page pointers. Size
2913 * of the array is bounded by len, which is in turn bounded by
2914 * BTRFS_MAX_DEDUPE_LEN.
2915 */
2916 src_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2917 dst_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2918 if (!src_pgarr || !dst_pgarr) {
2919 kfree(src_pgarr);
2920 kfree(dst_pgarr);
2921 return -ENOMEM;
2922 }
2923 cmp->num_pages = num_pages;
2924 cmp->src_pages = src_pgarr;
2925 cmp->dst_pages = dst_pgarr;
2926
2927 ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff);
2928 if (ret)
2929 goto out;
2930
2931 ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff);
2932
2933 out:
2934 if (ret)
2935 btrfs_cmp_data_free(cmp);
2936 return 0;
2937 }
2938
2939 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2940 u64 dst_loff, u64 len, struct cmp_pages *cmp)
2941 {
2942 int ret = 0;
2943 int i;
2944 struct page *src_page, *dst_page;
2945 unsigned int cmp_len = PAGE_CACHE_SIZE;
2946 void *addr, *dst_addr;
2947
2948 i = 0;
2949 while (len) {
2950 if (len < PAGE_CACHE_SIZE)
2951 cmp_len = len;
2952
2953 BUG_ON(i >= cmp->num_pages);
2954
2955 src_page = cmp->src_pages[i];
2956 dst_page = cmp->dst_pages[i];
2957
2958 addr = kmap_atomic(src_page);
2959 dst_addr = kmap_atomic(dst_page);
2960
2961 flush_dcache_page(src_page);
2962 flush_dcache_page(dst_page);
2963
2964 if (memcmp(addr, dst_addr, cmp_len))
2965 ret = BTRFS_SAME_DATA_DIFFERS;
2966
2967 kunmap_atomic(addr);
2968 kunmap_atomic(dst_addr);
2969
2970 if (ret)
2971 break;
2972
2973 len -= cmp_len;
2974 i++;
2975 }
2976
2977 return ret;
2978 }
2979
2980 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
2981 u64 olen)
2982 {
2983 u64 len = *plen;
2984 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2985
2986 if (off + olen > inode->i_size || off + olen < off)
2987 return -EINVAL;
2988
2989 /* if we extend to eof, continue to block boundary */
2990 if (off + len == inode->i_size)
2991 *plen = len = ALIGN(inode->i_size, bs) - off;
2992
2993 /* Check that we are block aligned - btrfs_clone() requires this */
2994 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2995 return -EINVAL;
2996
2997 return 0;
2998 }
2999
3000 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3001 struct inode *dst, u64 dst_loff)
3002 {
3003 int ret;
3004 u64 len = olen;
3005 struct cmp_pages cmp;
3006 int same_inode = 0;
3007 u64 same_lock_start = 0;
3008 u64 same_lock_len = 0;
3009
3010 if (src == dst)
3011 same_inode = 1;
3012
3013 if (len == 0)
3014 return 0;
3015
3016 if (same_inode) {
3017 mutex_lock(&src->i_mutex);
3018
3019 ret = extent_same_check_offsets(src, loff, &len, olen);
3020 if (ret)
3021 goto out_unlock;
3022
3023 /*
3024 * Single inode case wants the same checks, except we
3025 * don't want our length pushed out past i_size as
3026 * comparing that data range makes no sense.
3027 *
3028 * extent_same_check_offsets() will do this for an
3029 * unaligned length at i_size, so catch it here and
3030 * reject the request.
3031 *
3032 * This effectively means we require aligned extents
3033 * for the single-inode case, whereas the other cases
3034 * allow an unaligned length so long as it ends at
3035 * i_size.
3036 */
3037 if (len != olen) {
3038 ret = -EINVAL;
3039 goto out_unlock;
3040 }
3041
3042 /* Check for overlapping ranges */
3043 if (dst_loff + len > loff && dst_loff < loff + len) {
3044 ret = -EINVAL;
3045 goto out_unlock;
3046 }
3047
3048 same_lock_start = min_t(u64, loff, dst_loff);
3049 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3050 } else {
3051 btrfs_double_inode_lock(src, dst);
3052
3053 ret = extent_same_check_offsets(src, loff, &len, olen);
3054 if (ret)
3055 goto out_unlock;
3056
3057 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3058 if (ret)
3059 goto out_unlock;
3060 }
3061
3062 /* don't make the dst file partly checksummed */
3063 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3064 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3065 ret = -EINVAL;
3066 goto out_unlock;
3067 }
3068
3069 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3070 if (ret)
3071 goto out_unlock;
3072
3073 if (same_inode)
3074 lock_extent_range(src, same_lock_start, same_lock_len);
3075 else
3076 btrfs_double_extent_lock(src, loff, dst, dst_loff, len);
3077
3078 /* pass original length for comparison so we stay within i_size */
3079 ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp);
3080 if (ret == 0)
3081 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3082
3083 if (same_inode)
3084 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3085 same_lock_start + same_lock_len - 1);
3086 else
3087 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3088
3089 btrfs_cmp_data_free(&cmp);
3090 out_unlock:
3091 if (same_inode)
3092 mutex_unlock(&src->i_mutex);
3093 else
3094 btrfs_double_inode_unlock(src, dst);
3095
3096 return ret;
3097 }
3098
3099 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
3100
3101 static long btrfs_ioctl_file_extent_same(struct file *file,
3102 struct btrfs_ioctl_same_args __user *argp)
3103 {
3104 struct btrfs_ioctl_same_args *same = NULL;
3105 struct btrfs_ioctl_same_extent_info *info;
3106 struct inode *src = file_inode(file);
3107 u64 off;
3108 u64 len;
3109 int i;
3110 int ret;
3111 unsigned long size;
3112 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3113 bool is_admin = capable(CAP_SYS_ADMIN);
3114 u16 count;
3115
3116 if (!(file->f_mode & FMODE_READ))
3117 return -EINVAL;
3118
3119 ret = mnt_want_write_file(file);
3120 if (ret)
3121 return ret;
3122
3123 if (get_user(count, &argp->dest_count)) {
3124 ret = -EFAULT;
3125 goto out;
3126 }
3127
3128 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
3129
3130 same = memdup_user(argp, size);
3131
3132 if (IS_ERR(same)) {
3133 ret = PTR_ERR(same);
3134 same = NULL;
3135 goto out;
3136 }
3137
3138 off = same->logical_offset;
3139 len = same->length;
3140
3141 /*
3142 * Limit the total length we will dedupe for each operation.
3143 * This is intended to bound the total time spent in this
3144 * ioctl to something sane.
3145 */
3146 if (len > BTRFS_MAX_DEDUPE_LEN)
3147 len = BTRFS_MAX_DEDUPE_LEN;
3148
3149 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
3150 /*
3151 * Btrfs does not support blocksize < page_size. As a
3152 * result, btrfs_cmp_data() won't correctly handle
3153 * this situation without an update.
3154 */
3155 ret = -EINVAL;
3156 goto out;
3157 }
3158
3159 ret = -EISDIR;
3160 if (S_ISDIR(src->i_mode))
3161 goto out;
3162
3163 ret = -EACCES;
3164 if (!S_ISREG(src->i_mode))
3165 goto out;
3166
3167 /* pre-format output fields to sane values */
3168 for (i = 0; i < count; i++) {
3169 same->info[i].bytes_deduped = 0ULL;
3170 same->info[i].status = 0;
3171 }
3172
3173 for (i = 0, info = same->info; i < count; i++, info++) {
3174 struct inode *dst;
3175 struct fd dst_file = fdget(info->fd);
3176 if (!dst_file.file) {
3177 info->status = -EBADF;
3178 continue;
3179 }
3180 dst = file_inode(dst_file.file);
3181
3182 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
3183 info->status = -EINVAL;
3184 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3185 info->status = -EXDEV;
3186 } else if (S_ISDIR(dst->i_mode)) {
3187 info->status = -EISDIR;
3188 } else if (!S_ISREG(dst->i_mode)) {
3189 info->status = -EACCES;
3190 } else {
3191 info->status = btrfs_extent_same(src, off, len, dst,
3192 info->logical_offset);
3193 if (info->status == 0)
3194 info->bytes_deduped += len;
3195 }
3196 fdput(dst_file);
3197 }
3198
3199 ret = copy_to_user(argp, same, size);
3200 if (ret)
3201 ret = -EFAULT;
3202
3203 out:
3204 mnt_drop_write_file(file);
3205 kfree(same);
3206 return ret;
3207 }
3208
3209 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3210 struct inode *inode,
3211 u64 endoff,
3212 const u64 destoff,
3213 const u64 olen,
3214 int no_time_update)
3215 {
3216 struct btrfs_root *root = BTRFS_I(inode)->root;
3217 int ret;
3218
3219 inode_inc_iversion(inode);
3220 if (!no_time_update)
3221 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3222 /*
3223 * We round up to the block size at eof when determining which
3224 * extents to clone above, but shouldn't round up the file size.
3225 */
3226 if (endoff > destoff + olen)
3227 endoff = destoff + olen;
3228 if (endoff > inode->i_size)
3229 btrfs_i_size_write(inode, endoff);
3230
3231 ret = btrfs_update_inode(trans, root, inode);
3232 if (ret) {
3233 btrfs_abort_transaction(trans, root, ret);
3234 btrfs_end_transaction(trans, root);
3235 goto out;
3236 }
3237 ret = btrfs_end_transaction(trans, root);
3238 out:
3239 return ret;
3240 }
3241
3242 static void clone_update_extent_map(struct inode *inode,
3243 const struct btrfs_trans_handle *trans,
3244 const struct btrfs_path *path,
3245 const u64 hole_offset,
3246 const u64 hole_len)
3247 {
3248 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3249 struct extent_map *em;
3250 int ret;
3251
3252 em = alloc_extent_map();
3253 if (!em) {
3254 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3255 &BTRFS_I(inode)->runtime_flags);
3256 return;
3257 }
3258
3259 if (path) {
3260 struct btrfs_file_extent_item *fi;
3261
3262 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3263 struct btrfs_file_extent_item);
3264 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3265 em->generation = -1;
3266 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3267 BTRFS_FILE_EXTENT_INLINE)
3268 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3269 &BTRFS_I(inode)->runtime_flags);
3270 } else {
3271 em->start = hole_offset;
3272 em->len = hole_len;
3273 em->ram_bytes = em->len;
3274 em->orig_start = hole_offset;
3275 em->block_start = EXTENT_MAP_HOLE;
3276 em->block_len = 0;
3277 em->orig_block_len = 0;
3278 em->compress_type = BTRFS_COMPRESS_NONE;
3279 em->generation = trans->transid;
3280 }
3281
3282 while (1) {
3283 write_lock(&em_tree->lock);
3284 ret = add_extent_mapping(em_tree, em, 1);
3285 write_unlock(&em_tree->lock);
3286 if (ret != -EEXIST) {
3287 free_extent_map(em);
3288 break;
3289 }
3290 btrfs_drop_extent_cache(inode, em->start,
3291 em->start + em->len - 1, 0);
3292 }
3293
3294 if (ret)
3295 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3296 &BTRFS_I(inode)->runtime_flags);
3297 }
3298
3299 /*
3300 * Make sure we do not end up inserting an inline extent into a file that has
3301 * already other (non-inline) extents. If a file has an inline extent it can
3302 * not have any other extents and the (single) inline extent must start at the
3303 * file offset 0. Failing to respect these rules will lead to file corruption,
3304 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3305 *
3306 * We can have extents that have been already written to disk or we can have
3307 * dirty ranges still in delalloc, in which case the extent maps and items are
3308 * created only when we run delalloc, and the delalloc ranges might fall outside
3309 * the range we are currently locking in the inode's io tree. So we check the
3310 * inode's i_size because of that (i_size updates are done while holding the
3311 * i_mutex, which we are holding here).
3312 * We also check to see if the inode has a size not greater than "datal" but has
3313 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3314 * protected against such concurrent fallocate calls by the i_mutex).
3315 *
3316 * If the file has no extents but a size greater than datal, do not allow the
3317 * copy because we would need turn the inline extent into a non-inline one (even
3318 * with NO_HOLES enabled). If we find our destination inode only has one inline
3319 * extent, just overwrite it with the source inline extent if its size is less
3320 * than the source extent's size, or we could copy the source inline extent's
3321 * data into the destination inode's inline extent if the later is greater then
3322 * the former.
3323 */
3324 static int clone_copy_inline_extent(struct inode *src,
3325 struct inode *dst,
3326 struct btrfs_trans_handle *trans,
3327 struct btrfs_path *path,
3328 struct btrfs_key *new_key,
3329 const u64 drop_start,
3330 const u64 datal,
3331 const u64 skip,
3332 const u64 size,
3333 char *inline_data)
3334 {
3335 struct btrfs_root *root = BTRFS_I(dst)->root;
3336 const u64 aligned_end = ALIGN(new_key->offset + datal,
3337 root->sectorsize);
3338 int ret;
3339 struct btrfs_key key;
3340
3341 if (new_key->offset > 0)
3342 return -EOPNOTSUPP;
3343
3344 key.objectid = btrfs_ino(dst);
3345 key.type = BTRFS_EXTENT_DATA_KEY;
3346 key.offset = 0;
3347 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3348 if (ret < 0) {
3349 return ret;
3350 } else if (ret > 0) {
3351 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3352 ret = btrfs_next_leaf(root, path);
3353 if (ret < 0)
3354 return ret;
3355 else if (ret > 0)
3356 goto copy_inline_extent;
3357 }
3358 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3359 if (key.objectid == btrfs_ino(dst) &&
3360 key.type == BTRFS_EXTENT_DATA_KEY) {
3361 ASSERT(key.offset > 0);
3362 return -EOPNOTSUPP;
3363 }
3364 } else if (i_size_read(dst) <= datal) {
3365 struct btrfs_file_extent_item *ei;
3366 u64 ext_len;
3367
3368 /*
3369 * If the file size is <= datal, make sure there are no other
3370 * extents following (can happen do to an fallocate call with
3371 * the flag FALLOC_FL_KEEP_SIZE).
3372 */
3373 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3374 struct btrfs_file_extent_item);
3375 /*
3376 * If it's an inline extent, it can not have other extents
3377 * following it.
3378 */
3379 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3380 BTRFS_FILE_EXTENT_INLINE)
3381 goto copy_inline_extent;
3382
3383 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3384 if (ext_len > aligned_end)
3385 return -EOPNOTSUPP;
3386
3387 ret = btrfs_next_item(root, path);
3388 if (ret < 0) {
3389 return ret;
3390 } else if (ret == 0) {
3391 btrfs_item_key_to_cpu(path->nodes[0], &key,
3392 path->slots[0]);
3393 if (key.objectid == btrfs_ino(dst) &&
3394 key.type == BTRFS_EXTENT_DATA_KEY)
3395 return -EOPNOTSUPP;
3396 }
3397 }
3398
3399 copy_inline_extent:
3400 /*
3401 * We have no extent items, or we have an extent at offset 0 which may
3402 * or may not be inlined. All these cases are dealt the same way.
3403 */
3404 if (i_size_read(dst) > datal) {
3405 /*
3406 * If the destination inode has an inline extent...
3407 * This would require copying the data from the source inline
3408 * extent into the beginning of the destination's inline extent.
3409 * But this is really complex, both extents can be compressed
3410 * or just one of them, which would require decompressing and
3411 * re-compressing data (which could increase the new compressed
3412 * size, not allowing the compressed data to fit anymore in an
3413 * inline extent).
3414 * So just don't support this case for now (it should be rare,
3415 * we are not really saving space when cloning inline extents).
3416 */
3417 return -EOPNOTSUPP;
3418 }
3419
3420 btrfs_release_path(path);
3421 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3422 if (ret)
3423 return ret;
3424 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3425 if (ret)
3426 return ret;
3427
3428 if (skip) {
3429 const u32 start = btrfs_file_extent_calc_inline_size(0);
3430
3431 memmove(inline_data + start, inline_data + start + skip, datal);
3432 }
3433
3434 write_extent_buffer(path->nodes[0], inline_data,
3435 btrfs_item_ptr_offset(path->nodes[0],
3436 path->slots[0]),
3437 size);
3438 inode_add_bytes(dst, datal);
3439
3440 return 0;
3441 }
3442
3443 /**
3444 * btrfs_clone() - clone a range from inode file to another
3445 *
3446 * @src: Inode to clone from
3447 * @inode: Inode to clone to
3448 * @off: Offset within source to start clone from
3449 * @olen: Original length, passed by user, of range to clone
3450 * @olen_aligned: Block-aligned value of olen
3451 * @destoff: Offset within @inode to start clone
3452 * @no_time_update: Whether to update mtime/ctime on the target inode
3453 */
3454 static int btrfs_clone(struct inode *src, struct inode *inode,
3455 const u64 off, const u64 olen, const u64 olen_aligned,
3456 const u64 destoff, int no_time_update)
3457 {
3458 struct btrfs_root *root = BTRFS_I(inode)->root;
3459 struct btrfs_path *path = NULL;
3460 struct extent_buffer *leaf;
3461 struct btrfs_trans_handle *trans;
3462 char *buf = NULL;
3463 struct btrfs_key key;
3464 u32 nritems;
3465 int slot;
3466 int ret;
3467 const u64 len = olen_aligned;
3468 u64 last_dest_end = destoff;
3469
3470 ret = -ENOMEM;
3471 buf = vmalloc(root->nodesize);
3472 if (!buf)
3473 return ret;
3474
3475 path = btrfs_alloc_path();
3476 if (!path) {
3477 vfree(buf);
3478 return ret;
3479 }
3480
3481 path->reada = 2;
3482 /* clone data */
3483 key.objectid = btrfs_ino(src);
3484 key.type = BTRFS_EXTENT_DATA_KEY;
3485 key.offset = off;
3486
3487 while (1) {
3488 u64 next_key_min_offset = key.offset + 1;
3489
3490 /*
3491 * note the key will change type as we walk through the
3492 * tree.
3493 */
3494 path->leave_spinning = 1;
3495 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3496 0, 0);
3497 if (ret < 0)
3498 goto out;
3499 /*
3500 * First search, if no extent item that starts at offset off was
3501 * found but the previous item is an extent item, it's possible
3502 * it might overlap our target range, therefore process it.
3503 */
3504 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3505 btrfs_item_key_to_cpu(path->nodes[0], &key,
3506 path->slots[0] - 1);
3507 if (key.type == BTRFS_EXTENT_DATA_KEY)
3508 path->slots[0]--;
3509 }
3510
3511 nritems = btrfs_header_nritems(path->nodes[0]);
3512 process_slot:
3513 if (path->slots[0] >= nritems) {
3514 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3515 if (ret < 0)
3516 goto out;
3517 if (ret > 0)
3518 break;
3519 nritems = btrfs_header_nritems(path->nodes[0]);
3520 }
3521 leaf = path->nodes[0];
3522 slot = path->slots[0];
3523
3524 btrfs_item_key_to_cpu(leaf, &key, slot);
3525 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3526 key.objectid != btrfs_ino(src))
3527 break;
3528
3529 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3530 struct btrfs_file_extent_item *extent;
3531 int type;
3532 u32 size;
3533 struct btrfs_key new_key;
3534 u64 disko = 0, diskl = 0;
3535 u64 datao = 0, datal = 0;
3536 u8 comp;
3537 u64 drop_start;
3538
3539 extent = btrfs_item_ptr(leaf, slot,
3540 struct btrfs_file_extent_item);
3541 comp = btrfs_file_extent_compression(leaf, extent);
3542 type = btrfs_file_extent_type(leaf, extent);
3543 if (type == BTRFS_FILE_EXTENT_REG ||
3544 type == BTRFS_FILE_EXTENT_PREALLOC) {
3545 disko = btrfs_file_extent_disk_bytenr(leaf,
3546 extent);
3547 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3548 extent);
3549 datao = btrfs_file_extent_offset(leaf, extent);
3550 datal = btrfs_file_extent_num_bytes(leaf,
3551 extent);
3552 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3553 /* take upper bound, may be compressed */
3554 datal = btrfs_file_extent_ram_bytes(leaf,
3555 extent);
3556 }
3557
3558 /*
3559 * The first search might have left us at an extent
3560 * item that ends before our target range's start, can
3561 * happen if we have holes and NO_HOLES feature enabled.
3562 */
3563 if (key.offset + datal <= off) {
3564 path->slots[0]++;
3565 goto process_slot;
3566 } else if (key.offset >= off + len) {
3567 break;
3568 }
3569 next_key_min_offset = key.offset + datal;
3570 size = btrfs_item_size_nr(leaf, slot);
3571 read_extent_buffer(leaf, buf,
3572 btrfs_item_ptr_offset(leaf, slot),
3573 size);
3574
3575 btrfs_release_path(path);
3576 path->leave_spinning = 0;
3577
3578 memcpy(&new_key, &key, sizeof(new_key));
3579 new_key.objectid = btrfs_ino(inode);
3580 if (off <= key.offset)
3581 new_key.offset = key.offset + destoff - off;
3582 else
3583 new_key.offset = destoff;
3584
3585 /*
3586 * Deal with a hole that doesn't have an extent item
3587 * that represents it (NO_HOLES feature enabled).
3588 * This hole is either in the middle of the cloning
3589 * range or at the beginning (fully overlaps it or
3590 * partially overlaps it).
3591 */
3592 if (new_key.offset != last_dest_end)
3593 drop_start = last_dest_end;
3594 else
3595 drop_start = new_key.offset;
3596
3597 /*
3598 * 1 - adjusting old extent (we may have to split it)
3599 * 1 - add new extent
3600 * 1 - inode update
3601 */
3602 trans = btrfs_start_transaction(root, 3);
3603 if (IS_ERR(trans)) {
3604 ret = PTR_ERR(trans);
3605 goto out;
3606 }
3607
3608 if (type == BTRFS_FILE_EXTENT_REG ||
3609 type == BTRFS_FILE_EXTENT_PREALLOC) {
3610 /*
3611 * a | --- range to clone ---| b
3612 * | ------------- extent ------------- |
3613 */
3614
3615 /* subtract range b */
3616 if (key.offset + datal > off + len)
3617 datal = off + len - key.offset;
3618
3619 /* subtract range a */
3620 if (off > key.offset) {
3621 datao += off - key.offset;
3622 datal -= off - key.offset;
3623 }
3624
3625 ret = btrfs_drop_extents(trans, root, inode,
3626 drop_start,
3627 new_key.offset + datal,
3628 1);
3629 if (ret) {
3630 if (ret != -EOPNOTSUPP)
3631 btrfs_abort_transaction(trans,
3632 root, ret);
3633 btrfs_end_transaction(trans, root);
3634 goto out;
3635 }
3636
3637 ret = btrfs_insert_empty_item(trans, root, path,
3638 &new_key, size);
3639 if (ret) {
3640 btrfs_abort_transaction(trans, root,
3641 ret);
3642 btrfs_end_transaction(trans, root);
3643 goto out;
3644 }
3645
3646 leaf = path->nodes[0];
3647 slot = path->slots[0];
3648 write_extent_buffer(leaf, buf,
3649 btrfs_item_ptr_offset(leaf, slot),
3650 size);
3651
3652 extent = btrfs_item_ptr(leaf, slot,
3653 struct btrfs_file_extent_item);
3654
3655 /* disko == 0 means it's a hole */
3656 if (!disko)
3657 datao = 0;
3658
3659 btrfs_set_file_extent_offset(leaf, extent,
3660 datao);
3661 btrfs_set_file_extent_num_bytes(leaf, extent,
3662 datal);
3663
3664 if (disko) {
3665 inode_add_bytes(inode, datal);
3666 ret = btrfs_inc_extent_ref(trans, root,
3667 disko, diskl, 0,
3668 root->root_key.objectid,
3669 btrfs_ino(inode),
3670 new_key.offset - datao);
3671 if (ret) {
3672 btrfs_abort_transaction(trans,
3673 root,
3674 ret);
3675 btrfs_end_transaction(trans,
3676 root);
3677 goto out;
3678
3679 }
3680 }
3681 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3682 u64 skip = 0;
3683 u64 trim = 0;
3684
3685 if (off > key.offset) {
3686 skip = off - key.offset;
3687 new_key.offset += skip;
3688 }
3689
3690 if (key.offset + datal > off + len)
3691 trim = key.offset + datal - (off + len);
3692
3693 if (comp && (skip || trim)) {
3694 ret = -EINVAL;
3695 btrfs_end_transaction(trans, root);
3696 goto out;
3697 }
3698 size -= skip + trim;
3699 datal -= skip + trim;
3700
3701 ret = clone_copy_inline_extent(src, inode,
3702 trans, path,
3703 &new_key,
3704 drop_start,
3705 datal,
3706 skip, size, buf);
3707 if (ret) {
3708 if (ret != -EOPNOTSUPP)
3709 btrfs_abort_transaction(trans,
3710 root,
3711 ret);
3712 btrfs_end_transaction(trans, root);
3713 goto out;
3714 }
3715 leaf = path->nodes[0];
3716 slot = path->slots[0];
3717 }
3718
3719 /* If we have an implicit hole (NO_HOLES feature). */
3720 if (drop_start < new_key.offset)
3721 clone_update_extent_map(inode, trans,
3722 NULL, drop_start,
3723 new_key.offset - drop_start);
3724
3725 clone_update_extent_map(inode, trans, path, 0, 0);
3726
3727 btrfs_mark_buffer_dirty(leaf);
3728 btrfs_release_path(path);
3729
3730 last_dest_end = ALIGN(new_key.offset + datal,
3731 root->sectorsize);
3732 ret = clone_finish_inode_update(trans, inode,
3733 last_dest_end,
3734 destoff, olen,
3735 no_time_update);
3736 if (ret)
3737 goto out;
3738 if (new_key.offset + datal >= destoff + len)
3739 break;
3740 }
3741 btrfs_release_path(path);
3742 key.offset = next_key_min_offset;
3743 }
3744 ret = 0;
3745
3746 if (last_dest_end < destoff + len) {
3747 /*
3748 * We have an implicit hole (NO_HOLES feature is enabled) that
3749 * fully or partially overlaps our cloning range at its end.
3750 */
3751 btrfs_release_path(path);
3752
3753 /*
3754 * 1 - remove extent(s)
3755 * 1 - inode update
3756 */
3757 trans = btrfs_start_transaction(root, 2);
3758 if (IS_ERR(trans)) {
3759 ret = PTR_ERR(trans);
3760 goto out;
3761 }
3762 ret = btrfs_drop_extents(trans, root, inode,
3763 last_dest_end, destoff + len, 1);
3764 if (ret) {
3765 if (ret != -EOPNOTSUPP)
3766 btrfs_abort_transaction(trans, root, ret);
3767 btrfs_end_transaction(trans, root);
3768 goto out;
3769 }
3770 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3771 destoff + len - last_dest_end);
3772 ret = clone_finish_inode_update(trans, inode, destoff + len,
3773 destoff, olen, no_time_update);
3774 }
3775
3776 out:
3777 btrfs_free_path(path);
3778 vfree(buf);
3779 return ret;
3780 }
3781
3782 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3783 u64 off, u64 olen, u64 destoff)
3784 {
3785 struct inode *inode = file_inode(file);
3786 struct btrfs_root *root = BTRFS_I(inode)->root;
3787 struct fd src_file;
3788 struct inode *src;
3789 int ret;
3790 u64 len = olen;
3791 u64 bs = root->fs_info->sb->s_blocksize;
3792 int same_inode = 0;
3793
3794 /*
3795 * TODO:
3796 * - split compressed inline extents. annoying: we need to
3797 * decompress into destination's address_space (the file offset
3798 * may change, so source mapping won't do), then recompress (or
3799 * otherwise reinsert) a subrange.
3800 *
3801 * - split destination inode's inline extents. The inline extents can
3802 * be either compressed or non-compressed.
3803 */
3804
3805 /* the destination must be opened for writing */
3806 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3807 return -EINVAL;
3808
3809 if (btrfs_root_readonly(root))
3810 return -EROFS;
3811
3812 ret = mnt_want_write_file(file);
3813 if (ret)
3814 return ret;
3815
3816 src_file = fdget(srcfd);
3817 if (!src_file.file) {
3818 ret = -EBADF;
3819 goto out_drop_write;
3820 }
3821
3822 ret = -EXDEV;
3823 if (src_file.file->f_path.mnt != file->f_path.mnt)
3824 goto out_fput;
3825
3826 src = file_inode(src_file.file);
3827
3828 ret = -EINVAL;
3829 if (src == inode)
3830 same_inode = 1;
3831
3832 /* the src must be open for reading */
3833 if (!(src_file.file->f_mode & FMODE_READ))
3834 goto out_fput;
3835
3836 /* don't make the dst file partly checksummed */
3837 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3838 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3839 goto out_fput;
3840
3841 ret = -EISDIR;
3842 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3843 goto out_fput;
3844
3845 ret = -EXDEV;
3846 if (src->i_sb != inode->i_sb)
3847 goto out_fput;
3848
3849 if (!same_inode) {
3850 btrfs_double_inode_lock(src, inode);
3851 } else {
3852 mutex_lock(&src->i_mutex);
3853 }
3854
3855 /* determine range to clone */
3856 ret = -EINVAL;
3857 if (off + len > src->i_size || off + len < off)
3858 goto out_unlock;
3859 if (len == 0)
3860 olen = len = src->i_size - off;
3861 /* if we extend to eof, continue to block boundary */
3862 if (off + len == src->i_size)
3863 len = ALIGN(src->i_size, bs) - off;
3864
3865 if (len == 0) {
3866 ret = 0;
3867 goto out_unlock;
3868 }
3869
3870 /* verify the end result is block aligned */
3871 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3872 !IS_ALIGNED(destoff, bs))
3873 goto out_unlock;
3874
3875 /* verify if ranges are overlapped within the same file */
3876 if (same_inode) {
3877 if (destoff + len > off && destoff < off + len)
3878 goto out_unlock;
3879 }
3880
3881 if (destoff > inode->i_size) {
3882 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3883 if (ret)
3884 goto out_unlock;
3885 }
3886
3887 /*
3888 * Lock the target range too. Right after we replace the file extent
3889 * items in the fs tree (which now point to the cloned data), we might
3890 * have a worker replace them with extent items relative to a write
3891 * operation that was issued before this clone operation (i.e. confront
3892 * with inode.c:btrfs_finish_ordered_io).
3893 */
3894 if (same_inode) {
3895 u64 lock_start = min_t(u64, off, destoff);
3896 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3897
3898 lock_extent_range(src, lock_start, lock_len);
3899 } else {
3900 btrfs_double_extent_lock(src, off, inode, destoff, len);
3901 }
3902
3903 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3904
3905 if (same_inode) {
3906 u64 lock_start = min_t(u64, off, destoff);
3907 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3908
3909 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3910 } else {
3911 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3912 }
3913 /*
3914 * Truncate page cache pages so that future reads will see the cloned
3915 * data immediately and not the previous data.
3916 */
3917 truncate_inode_pages_range(&inode->i_data, destoff,
3918 PAGE_CACHE_ALIGN(destoff + len) - 1);
3919 out_unlock:
3920 if (!same_inode)
3921 btrfs_double_inode_unlock(src, inode);
3922 else
3923 mutex_unlock(&src->i_mutex);
3924 out_fput:
3925 fdput(src_file);
3926 out_drop_write:
3927 mnt_drop_write_file(file);
3928 return ret;
3929 }
3930
3931 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3932 {
3933 struct btrfs_ioctl_clone_range_args args;
3934
3935 if (copy_from_user(&args, argp, sizeof(args)))
3936 return -EFAULT;
3937 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3938 args.src_length, args.dest_offset);
3939 }
3940
3941 /*
3942 * there are many ways the trans_start and trans_end ioctls can lead
3943 * to deadlocks. They should only be used by applications that
3944 * basically own the machine, and have a very in depth understanding
3945 * of all the possible deadlocks and enospc problems.
3946 */
3947 static long btrfs_ioctl_trans_start(struct file *file)
3948 {
3949 struct inode *inode = file_inode(file);
3950 struct btrfs_root *root = BTRFS_I(inode)->root;
3951 struct btrfs_trans_handle *trans;
3952 int ret;
3953
3954 ret = -EPERM;
3955 if (!capable(CAP_SYS_ADMIN))
3956 goto out;
3957
3958 ret = -EINPROGRESS;
3959 if (file->private_data)
3960 goto out;
3961
3962 ret = -EROFS;
3963 if (btrfs_root_readonly(root))
3964 goto out;
3965
3966 ret = mnt_want_write_file(file);
3967 if (ret)
3968 goto out;
3969
3970 atomic_inc(&root->fs_info->open_ioctl_trans);
3971
3972 ret = -ENOMEM;
3973 trans = btrfs_start_ioctl_transaction(root);
3974 if (IS_ERR(trans))
3975 goto out_drop;
3976
3977 file->private_data = trans;
3978 return 0;
3979
3980 out_drop:
3981 atomic_dec(&root->fs_info->open_ioctl_trans);
3982 mnt_drop_write_file(file);
3983 out:
3984 return ret;
3985 }
3986
3987 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3988 {
3989 struct inode *inode = file_inode(file);
3990 struct btrfs_root *root = BTRFS_I(inode)->root;
3991 struct btrfs_root *new_root;
3992 struct btrfs_dir_item *di;
3993 struct btrfs_trans_handle *trans;
3994 struct btrfs_path *path;
3995 struct btrfs_key location;
3996 struct btrfs_disk_key disk_key;
3997 u64 objectid = 0;
3998 u64 dir_id;
3999 int ret;
4000
4001 if (!capable(CAP_SYS_ADMIN))
4002 return -EPERM;
4003
4004 ret = mnt_want_write_file(file);
4005 if (ret)
4006 return ret;
4007
4008 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4009 ret = -EFAULT;
4010 goto out;
4011 }
4012
4013 if (!objectid)
4014 objectid = BTRFS_FS_TREE_OBJECTID;
4015
4016 location.objectid = objectid;
4017 location.type = BTRFS_ROOT_ITEM_KEY;
4018 location.offset = (u64)-1;
4019
4020 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
4021 if (IS_ERR(new_root)) {
4022 ret = PTR_ERR(new_root);
4023 goto out;
4024 }
4025
4026 path = btrfs_alloc_path();
4027 if (!path) {
4028 ret = -ENOMEM;
4029 goto out;
4030 }
4031 path->leave_spinning = 1;
4032
4033 trans = btrfs_start_transaction(root, 1);
4034 if (IS_ERR(trans)) {
4035 btrfs_free_path(path);
4036 ret = PTR_ERR(trans);
4037 goto out;
4038 }
4039
4040 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
4041 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
4042 dir_id, "default", 7, 1);
4043 if (IS_ERR_OR_NULL(di)) {
4044 btrfs_free_path(path);
4045 btrfs_end_transaction(trans, root);
4046 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
4047 "item, this isn't going to work");
4048 ret = -ENOENT;
4049 goto out;
4050 }
4051
4052 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4053 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4054 btrfs_mark_buffer_dirty(path->nodes[0]);
4055 btrfs_free_path(path);
4056
4057 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
4058 btrfs_end_transaction(trans, root);
4059 out:
4060 mnt_drop_write_file(file);
4061 return ret;
4062 }
4063
4064 void btrfs_get_block_group_info(struct list_head *groups_list,
4065 struct btrfs_ioctl_space_info *space)
4066 {
4067 struct btrfs_block_group_cache *block_group;
4068
4069 space->total_bytes = 0;
4070 space->used_bytes = 0;
4071 space->flags = 0;
4072 list_for_each_entry(block_group, groups_list, list) {
4073 space->flags = block_group->flags;
4074 space->total_bytes += block_group->key.offset;
4075 space->used_bytes +=
4076 btrfs_block_group_used(&block_group->item);
4077 }
4078 }
4079
4080 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
4081 {
4082 struct btrfs_ioctl_space_args space_args;
4083 struct btrfs_ioctl_space_info space;
4084 struct btrfs_ioctl_space_info *dest;
4085 struct btrfs_ioctl_space_info *dest_orig;
4086 struct btrfs_ioctl_space_info __user *user_dest;
4087 struct btrfs_space_info *info;
4088 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4089 BTRFS_BLOCK_GROUP_SYSTEM,
4090 BTRFS_BLOCK_GROUP_METADATA,
4091 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4092 int num_types = 4;
4093 int alloc_size;
4094 int ret = 0;
4095 u64 slot_count = 0;
4096 int i, c;
4097
4098 if (copy_from_user(&space_args,
4099 (struct btrfs_ioctl_space_args __user *)arg,
4100 sizeof(space_args)))
4101 return -EFAULT;
4102
4103 for (i = 0; i < num_types; i++) {
4104 struct btrfs_space_info *tmp;
4105
4106 info = NULL;
4107 rcu_read_lock();
4108 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4109 list) {
4110 if (tmp->flags == types[i]) {
4111 info = tmp;
4112 break;
4113 }
4114 }
4115 rcu_read_unlock();
4116
4117 if (!info)
4118 continue;
4119
4120 down_read(&info->groups_sem);
4121 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4122 if (!list_empty(&info->block_groups[c]))
4123 slot_count++;
4124 }
4125 up_read(&info->groups_sem);
4126 }
4127
4128 /*
4129 * Global block reserve, exported as a space_info
4130 */
4131 slot_count++;
4132
4133 /* space_slots == 0 means they are asking for a count */
4134 if (space_args.space_slots == 0) {
4135 space_args.total_spaces = slot_count;
4136 goto out;
4137 }
4138
4139 slot_count = min_t(u64, space_args.space_slots, slot_count);
4140
4141 alloc_size = sizeof(*dest) * slot_count;
4142
4143 /* we generally have at most 6 or so space infos, one for each raid
4144 * level. So, a whole page should be more than enough for everyone
4145 */
4146 if (alloc_size > PAGE_CACHE_SIZE)
4147 return -ENOMEM;
4148
4149 space_args.total_spaces = 0;
4150 dest = kmalloc(alloc_size, GFP_NOFS);
4151 if (!dest)
4152 return -ENOMEM;
4153 dest_orig = dest;
4154
4155 /* now we have a buffer to copy into */
4156 for (i = 0; i < num_types; i++) {
4157 struct btrfs_space_info *tmp;
4158
4159 if (!slot_count)
4160 break;
4161
4162 info = NULL;
4163 rcu_read_lock();
4164 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4165 list) {
4166 if (tmp->flags == types[i]) {
4167 info = tmp;
4168 break;
4169 }
4170 }
4171 rcu_read_unlock();
4172
4173 if (!info)
4174 continue;
4175 down_read(&info->groups_sem);
4176 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4177 if (!list_empty(&info->block_groups[c])) {
4178 btrfs_get_block_group_info(
4179 &info->block_groups[c], &space);
4180 memcpy(dest, &space, sizeof(space));
4181 dest++;
4182 space_args.total_spaces++;
4183 slot_count--;
4184 }
4185 if (!slot_count)
4186 break;
4187 }
4188 up_read(&info->groups_sem);
4189 }
4190
4191 /*
4192 * Add global block reserve
4193 */
4194 if (slot_count) {
4195 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4196
4197 spin_lock(&block_rsv->lock);
4198 space.total_bytes = block_rsv->size;
4199 space.used_bytes = block_rsv->size - block_rsv->reserved;
4200 spin_unlock(&block_rsv->lock);
4201 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4202 memcpy(dest, &space, sizeof(space));
4203 space_args.total_spaces++;
4204 }
4205
4206 user_dest = (struct btrfs_ioctl_space_info __user *)
4207 (arg + sizeof(struct btrfs_ioctl_space_args));
4208
4209 if (copy_to_user(user_dest, dest_orig, alloc_size))
4210 ret = -EFAULT;
4211
4212 kfree(dest_orig);
4213 out:
4214 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4215 ret = -EFAULT;
4216
4217 return ret;
4218 }
4219
4220 /*
4221 * there are many ways the trans_start and trans_end ioctls can lead
4222 * to deadlocks. They should only be used by applications that
4223 * basically own the machine, and have a very in depth understanding
4224 * of all the possible deadlocks and enospc problems.
4225 */
4226 long btrfs_ioctl_trans_end(struct file *file)
4227 {
4228 struct inode *inode = file_inode(file);
4229 struct btrfs_root *root = BTRFS_I(inode)->root;
4230 struct btrfs_trans_handle *trans;
4231
4232 trans = file->private_data;
4233 if (!trans)
4234 return -EINVAL;
4235 file->private_data = NULL;
4236
4237 btrfs_end_transaction(trans, root);
4238
4239 atomic_dec(&root->fs_info->open_ioctl_trans);
4240
4241 mnt_drop_write_file(file);
4242 return 0;
4243 }
4244
4245 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4246 void __user *argp)
4247 {
4248 struct btrfs_trans_handle *trans;
4249 u64 transid;
4250 int ret;
4251
4252 trans = btrfs_attach_transaction_barrier(root);
4253 if (IS_ERR(trans)) {
4254 if (PTR_ERR(trans) != -ENOENT)
4255 return PTR_ERR(trans);
4256
4257 /* No running transaction, don't bother */
4258 transid = root->fs_info->last_trans_committed;
4259 goto out;
4260 }
4261 transid = trans->transid;
4262 ret = btrfs_commit_transaction_async(trans, root, 0);
4263 if (ret) {
4264 btrfs_end_transaction(trans, root);
4265 return ret;
4266 }
4267 out:
4268 if (argp)
4269 if (copy_to_user(argp, &transid, sizeof(transid)))
4270 return -EFAULT;
4271 return 0;
4272 }
4273
4274 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4275 void __user *argp)
4276 {
4277 u64 transid;
4278
4279 if (argp) {
4280 if (copy_from_user(&transid, argp, sizeof(transid)))
4281 return -EFAULT;
4282 } else {
4283 transid = 0; /* current trans */
4284 }
4285 return btrfs_wait_for_commit(root, transid);
4286 }
4287
4288 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4289 {
4290 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4291 struct btrfs_ioctl_scrub_args *sa;
4292 int ret;
4293
4294 if (!capable(CAP_SYS_ADMIN))
4295 return -EPERM;
4296
4297 sa = memdup_user(arg, sizeof(*sa));
4298 if (IS_ERR(sa))
4299 return PTR_ERR(sa);
4300
4301 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4302 ret = mnt_want_write_file(file);
4303 if (ret)
4304 goto out;
4305 }
4306
4307 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4308 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4309 0);
4310
4311 if (copy_to_user(arg, sa, sizeof(*sa)))
4312 ret = -EFAULT;
4313
4314 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4315 mnt_drop_write_file(file);
4316 out:
4317 kfree(sa);
4318 return ret;
4319 }
4320
4321 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4322 {
4323 if (!capable(CAP_SYS_ADMIN))
4324 return -EPERM;
4325
4326 return btrfs_scrub_cancel(root->fs_info);
4327 }
4328
4329 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4330 void __user *arg)
4331 {
4332 struct btrfs_ioctl_scrub_args *sa;
4333 int ret;
4334
4335 if (!capable(CAP_SYS_ADMIN))
4336 return -EPERM;
4337
4338 sa = memdup_user(arg, sizeof(*sa));
4339 if (IS_ERR(sa))
4340 return PTR_ERR(sa);
4341
4342 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4343
4344 if (copy_to_user(arg, sa, sizeof(*sa)))
4345 ret = -EFAULT;
4346
4347 kfree(sa);
4348 return ret;
4349 }
4350
4351 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4352 void __user *arg)
4353 {
4354 struct btrfs_ioctl_get_dev_stats *sa;
4355 int ret;
4356
4357 sa = memdup_user(arg, sizeof(*sa));
4358 if (IS_ERR(sa))
4359 return PTR_ERR(sa);
4360
4361 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4362 kfree(sa);
4363 return -EPERM;
4364 }
4365
4366 ret = btrfs_get_dev_stats(root, sa);
4367
4368 if (copy_to_user(arg, sa, sizeof(*sa)))
4369 ret = -EFAULT;
4370
4371 kfree(sa);
4372 return ret;
4373 }
4374
4375 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4376 {
4377 struct btrfs_ioctl_dev_replace_args *p;
4378 int ret;
4379
4380 if (!capable(CAP_SYS_ADMIN))
4381 return -EPERM;
4382
4383 p = memdup_user(arg, sizeof(*p));
4384 if (IS_ERR(p))
4385 return PTR_ERR(p);
4386
4387 switch (p->cmd) {
4388 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4389 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4390 ret = -EROFS;
4391 goto out;
4392 }
4393 if (atomic_xchg(
4394 &root->fs_info->mutually_exclusive_operation_running,
4395 1)) {
4396 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4397 } else {
4398 ret = btrfs_dev_replace_start(root, p);
4399 atomic_set(
4400 &root->fs_info->mutually_exclusive_operation_running,
4401 0);
4402 }
4403 break;
4404 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4405 btrfs_dev_replace_status(root->fs_info, p);
4406 ret = 0;
4407 break;
4408 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4409 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4410 break;
4411 default:
4412 ret = -EINVAL;
4413 break;
4414 }
4415
4416 if (copy_to_user(arg, p, sizeof(*p)))
4417 ret = -EFAULT;
4418 out:
4419 kfree(p);
4420 return ret;
4421 }
4422
4423 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4424 {
4425 int ret = 0;
4426 int i;
4427 u64 rel_ptr;
4428 int size;
4429 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4430 struct inode_fs_paths *ipath = NULL;
4431 struct btrfs_path *path;
4432
4433 if (!capable(CAP_DAC_READ_SEARCH))
4434 return -EPERM;
4435
4436 path = btrfs_alloc_path();
4437 if (!path) {
4438 ret = -ENOMEM;
4439 goto out;
4440 }
4441
4442 ipa = memdup_user(arg, sizeof(*ipa));
4443 if (IS_ERR(ipa)) {
4444 ret = PTR_ERR(ipa);
4445 ipa = NULL;
4446 goto out;
4447 }
4448
4449 size = min_t(u32, ipa->size, 4096);
4450 ipath = init_ipath(size, root, path);
4451 if (IS_ERR(ipath)) {
4452 ret = PTR_ERR(ipath);
4453 ipath = NULL;
4454 goto out;
4455 }
4456
4457 ret = paths_from_inode(ipa->inum, ipath);
4458 if (ret < 0)
4459 goto out;
4460
4461 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4462 rel_ptr = ipath->fspath->val[i] -
4463 (u64)(unsigned long)ipath->fspath->val;
4464 ipath->fspath->val[i] = rel_ptr;
4465 }
4466
4467 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4468 (void *)(unsigned long)ipath->fspath, size);
4469 if (ret) {
4470 ret = -EFAULT;
4471 goto out;
4472 }
4473
4474 out:
4475 btrfs_free_path(path);
4476 free_ipath(ipath);
4477 kfree(ipa);
4478
4479 return ret;
4480 }
4481
4482 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4483 {
4484 struct btrfs_data_container *inodes = ctx;
4485 const size_t c = 3 * sizeof(u64);
4486
4487 if (inodes->bytes_left >= c) {
4488 inodes->bytes_left -= c;
4489 inodes->val[inodes->elem_cnt] = inum;
4490 inodes->val[inodes->elem_cnt + 1] = offset;
4491 inodes->val[inodes->elem_cnt + 2] = root;
4492 inodes->elem_cnt += 3;
4493 } else {
4494 inodes->bytes_missing += c - inodes->bytes_left;
4495 inodes->bytes_left = 0;
4496 inodes->elem_missed += 3;
4497 }
4498
4499 return 0;
4500 }
4501
4502 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4503 void __user *arg)
4504 {
4505 int ret = 0;
4506 int size;
4507 struct btrfs_ioctl_logical_ino_args *loi;
4508 struct btrfs_data_container *inodes = NULL;
4509 struct btrfs_path *path = NULL;
4510
4511 if (!capable(CAP_SYS_ADMIN))
4512 return -EPERM;
4513
4514 loi = memdup_user(arg, sizeof(*loi));
4515 if (IS_ERR(loi)) {
4516 ret = PTR_ERR(loi);
4517 loi = NULL;
4518 goto out;
4519 }
4520
4521 path = btrfs_alloc_path();
4522 if (!path) {
4523 ret = -ENOMEM;
4524 goto out;
4525 }
4526
4527 size = min_t(u32, loi->size, 64 * 1024);
4528 inodes = init_data_container(size);
4529 if (IS_ERR(inodes)) {
4530 ret = PTR_ERR(inodes);
4531 inodes = NULL;
4532 goto out;
4533 }
4534
4535 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4536 build_ino_list, inodes);
4537 if (ret == -EINVAL)
4538 ret = -ENOENT;
4539 if (ret < 0)
4540 goto out;
4541
4542 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4543 (void *)(unsigned long)inodes, size);
4544 if (ret)
4545 ret = -EFAULT;
4546
4547 out:
4548 btrfs_free_path(path);
4549 vfree(inodes);
4550 kfree(loi);
4551
4552 return ret;
4553 }
4554
4555 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4556 struct btrfs_ioctl_balance_args *bargs)
4557 {
4558 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4559
4560 bargs->flags = bctl->flags;
4561
4562 if (atomic_read(&fs_info->balance_running))
4563 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4564 if (atomic_read(&fs_info->balance_pause_req))
4565 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4566 if (atomic_read(&fs_info->balance_cancel_req))
4567 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4568
4569 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4570 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4571 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4572
4573 if (lock) {
4574 spin_lock(&fs_info->balance_lock);
4575 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4576 spin_unlock(&fs_info->balance_lock);
4577 } else {
4578 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4579 }
4580 }
4581
4582 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4583 {
4584 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4585 struct btrfs_fs_info *fs_info = root->fs_info;
4586 struct btrfs_ioctl_balance_args *bargs;
4587 struct btrfs_balance_control *bctl;
4588 bool need_unlock; /* for mut. excl. ops lock */
4589 int ret;
4590
4591 if (!capable(CAP_SYS_ADMIN))
4592 return -EPERM;
4593
4594 ret = mnt_want_write_file(file);
4595 if (ret)
4596 return ret;
4597
4598 again:
4599 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4600 mutex_lock(&fs_info->volume_mutex);
4601 mutex_lock(&fs_info->balance_mutex);
4602 need_unlock = true;
4603 goto locked;
4604 }
4605
4606 /*
4607 * mut. excl. ops lock is locked. Three possibilites:
4608 * (1) some other op is running
4609 * (2) balance is running
4610 * (3) balance is paused -- special case (think resume)
4611 */
4612 mutex_lock(&fs_info->balance_mutex);
4613 if (fs_info->balance_ctl) {
4614 /* this is either (2) or (3) */
4615 if (!atomic_read(&fs_info->balance_running)) {
4616 mutex_unlock(&fs_info->balance_mutex);
4617 if (!mutex_trylock(&fs_info->volume_mutex))
4618 goto again;
4619 mutex_lock(&fs_info->balance_mutex);
4620
4621 if (fs_info->balance_ctl &&
4622 !atomic_read(&fs_info->balance_running)) {
4623 /* this is (3) */
4624 need_unlock = false;
4625 goto locked;
4626 }
4627
4628 mutex_unlock(&fs_info->balance_mutex);
4629 mutex_unlock(&fs_info->volume_mutex);
4630 goto again;
4631 } else {
4632 /* this is (2) */
4633 mutex_unlock(&fs_info->balance_mutex);
4634 ret = -EINPROGRESS;
4635 goto out;
4636 }
4637 } else {
4638 /* this is (1) */
4639 mutex_unlock(&fs_info->balance_mutex);
4640 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4641 goto out;
4642 }
4643
4644 locked:
4645 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4646
4647 if (arg) {
4648 bargs = memdup_user(arg, sizeof(*bargs));
4649 if (IS_ERR(bargs)) {
4650 ret = PTR_ERR(bargs);
4651 goto out_unlock;
4652 }
4653
4654 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4655 if (!fs_info->balance_ctl) {
4656 ret = -ENOTCONN;
4657 goto out_bargs;
4658 }
4659
4660 bctl = fs_info->balance_ctl;
4661 spin_lock(&fs_info->balance_lock);
4662 bctl->flags |= BTRFS_BALANCE_RESUME;
4663 spin_unlock(&fs_info->balance_lock);
4664
4665 goto do_balance;
4666 }
4667 } else {
4668 bargs = NULL;
4669 }
4670
4671 if (fs_info->balance_ctl) {
4672 ret = -EINPROGRESS;
4673 goto out_bargs;
4674 }
4675
4676 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4677 if (!bctl) {
4678 ret = -ENOMEM;
4679 goto out_bargs;
4680 }
4681
4682 bctl->fs_info = fs_info;
4683 if (arg) {
4684 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4685 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4686 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4687
4688 bctl->flags = bargs->flags;
4689 } else {
4690 /* balance everything - no filters */
4691 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4692 }
4693
4694 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4695 ret = -EINVAL;
4696 goto out_bctl;
4697 }
4698
4699 do_balance:
4700 /*
4701 * Ownership of bctl and mutually_exclusive_operation_running
4702 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4703 * or, if restriper was paused all the way until unmount, in
4704 * free_fs_info. mutually_exclusive_operation_running is
4705 * cleared in __cancel_balance.
4706 */
4707 need_unlock = false;
4708
4709 ret = btrfs_balance(bctl, bargs);
4710 bctl = NULL;
4711
4712 if (arg) {
4713 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4714 ret = -EFAULT;
4715 }
4716
4717 out_bctl:
4718 kfree(bctl);
4719 out_bargs:
4720 kfree(bargs);
4721 out_unlock:
4722 mutex_unlock(&fs_info->balance_mutex);
4723 mutex_unlock(&fs_info->volume_mutex);
4724 if (need_unlock)
4725 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4726 out:
4727 mnt_drop_write_file(file);
4728 return ret;
4729 }
4730
4731 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4732 {
4733 if (!capable(CAP_SYS_ADMIN))
4734 return -EPERM;
4735
4736 switch (cmd) {
4737 case BTRFS_BALANCE_CTL_PAUSE:
4738 return btrfs_pause_balance(root->fs_info);
4739 case BTRFS_BALANCE_CTL_CANCEL:
4740 return btrfs_cancel_balance(root->fs_info);
4741 }
4742
4743 return -EINVAL;
4744 }
4745
4746 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4747 void __user *arg)
4748 {
4749 struct btrfs_fs_info *fs_info = root->fs_info;
4750 struct btrfs_ioctl_balance_args *bargs;
4751 int ret = 0;
4752
4753 if (!capable(CAP_SYS_ADMIN))
4754 return -EPERM;
4755
4756 mutex_lock(&fs_info->balance_mutex);
4757 if (!fs_info->balance_ctl) {
4758 ret = -ENOTCONN;
4759 goto out;
4760 }
4761
4762 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4763 if (!bargs) {
4764 ret = -ENOMEM;
4765 goto out;
4766 }
4767
4768 update_ioctl_balance_args(fs_info, 1, bargs);
4769
4770 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4771 ret = -EFAULT;
4772
4773 kfree(bargs);
4774 out:
4775 mutex_unlock(&fs_info->balance_mutex);
4776 return ret;
4777 }
4778
4779 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4780 {
4781 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4782 struct btrfs_ioctl_quota_ctl_args *sa;
4783 struct btrfs_trans_handle *trans = NULL;
4784 int ret;
4785 int err;
4786
4787 if (!capable(CAP_SYS_ADMIN))
4788 return -EPERM;
4789
4790 ret = mnt_want_write_file(file);
4791 if (ret)
4792 return ret;
4793
4794 sa = memdup_user(arg, sizeof(*sa));
4795 if (IS_ERR(sa)) {
4796 ret = PTR_ERR(sa);
4797 goto drop_write;
4798 }
4799
4800 down_write(&root->fs_info->subvol_sem);
4801 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4802 if (IS_ERR(trans)) {
4803 ret = PTR_ERR(trans);
4804 goto out;
4805 }
4806
4807 switch (sa->cmd) {
4808 case BTRFS_QUOTA_CTL_ENABLE:
4809 ret = btrfs_quota_enable(trans, root->fs_info);
4810 break;
4811 case BTRFS_QUOTA_CTL_DISABLE:
4812 ret = btrfs_quota_disable(trans, root->fs_info);
4813 break;
4814 default:
4815 ret = -EINVAL;
4816 break;
4817 }
4818
4819 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4820 if (err && !ret)
4821 ret = err;
4822 out:
4823 kfree(sa);
4824 up_write(&root->fs_info->subvol_sem);
4825 drop_write:
4826 mnt_drop_write_file(file);
4827 return ret;
4828 }
4829
4830 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4831 {
4832 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4833 struct btrfs_ioctl_qgroup_assign_args *sa;
4834 struct btrfs_trans_handle *trans;
4835 int ret;
4836 int err;
4837
4838 if (!capable(CAP_SYS_ADMIN))
4839 return -EPERM;
4840
4841 ret = mnt_want_write_file(file);
4842 if (ret)
4843 return ret;
4844
4845 sa = memdup_user(arg, sizeof(*sa));
4846 if (IS_ERR(sa)) {
4847 ret = PTR_ERR(sa);
4848 goto drop_write;
4849 }
4850
4851 trans = btrfs_join_transaction(root);
4852 if (IS_ERR(trans)) {
4853 ret = PTR_ERR(trans);
4854 goto out;
4855 }
4856
4857 /* FIXME: check if the IDs really exist */
4858 if (sa->assign) {
4859 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4860 sa->src, sa->dst);
4861 } else {
4862 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4863 sa->src, sa->dst);
4864 }
4865
4866 /* update qgroup status and info */
4867 err = btrfs_run_qgroups(trans, root->fs_info);
4868 if (err < 0)
4869 btrfs_std_error(root->fs_info, ret,
4870 "failed to update qgroup status and info\n");
4871 err = btrfs_end_transaction(trans, root);
4872 if (err && !ret)
4873 ret = err;
4874
4875 out:
4876 kfree(sa);
4877 drop_write:
4878 mnt_drop_write_file(file);
4879 return ret;
4880 }
4881
4882 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4883 {
4884 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4885 struct btrfs_ioctl_qgroup_create_args *sa;
4886 struct btrfs_trans_handle *trans;
4887 int ret;
4888 int err;
4889
4890 if (!capable(CAP_SYS_ADMIN))
4891 return -EPERM;
4892
4893 ret = mnt_want_write_file(file);
4894 if (ret)
4895 return ret;
4896
4897 sa = memdup_user(arg, sizeof(*sa));
4898 if (IS_ERR(sa)) {
4899 ret = PTR_ERR(sa);
4900 goto drop_write;
4901 }
4902
4903 if (!sa->qgroupid) {
4904 ret = -EINVAL;
4905 goto out;
4906 }
4907
4908 trans = btrfs_join_transaction(root);
4909 if (IS_ERR(trans)) {
4910 ret = PTR_ERR(trans);
4911 goto out;
4912 }
4913
4914 /* FIXME: check if the IDs really exist */
4915 if (sa->create) {
4916 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
4917 } else {
4918 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4919 }
4920
4921 err = btrfs_end_transaction(trans, root);
4922 if (err && !ret)
4923 ret = err;
4924
4925 out:
4926 kfree(sa);
4927 drop_write:
4928 mnt_drop_write_file(file);
4929 return ret;
4930 }
4931
4932 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4933 {
4934 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4935 struct btrfs_ioctl_qgroup_limit_args *sa;
4936 struct btrfs_trans_handle *trans;
4937 int ret;
4938 int err;
4939 u64 qgroupid;
4940
4941 if (!capable(CAP_SYS_ADMIN))
4942 return -EPERM;
4943
4944 ret = mnt_want_write_file(file);
4945 if (ret)
4946 return ret;
4947
4948 sa = memdup_user(arg, sizeof(*sa));
4949 if (IS_ERR(sa)) {
4950 ret = PTR_ERR(sa);
4951 goto drop_write;
4952 }
4953
4954 trans = btrfs_join_transaction(root);
4955 if (IS_ERR(trans)) {
4956 ret = PTR_ERR(trans);
4957 goto out;
4958 }
4959
4960 qgroupid = sa->qgroupid;
4961 if (!qgroupid) {
4962 /* take the current subvol as qgroup */
4963 qgroupid = root->root_key.objectid;
4964 }
4965
4966 /* FIXME: check if the IDs really exist */
4967 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4968
4969 err = btrfs_end_transaction(trans, root);
4970 if (err && !ret)
4971 ret = err;
4972
4973 out:
4974 kfree(sa);
4975 drop_write:
4976 mnt_drop_write_file(file);
4977 return ret;
4978 }
4979
4980 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4981 {
4982 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4983 struct btrfs_ioctl_quota_rescan_args *qsa;
4984 int ret;
4985
4986 if (!capable(CAP_SYS_ADMIN))
4987 return -EPERM;
4988
4989 ret = mnt_want_write_file(file);
4990 if (ret)
4991 return ret;
4992
4993 qsa = memdup_user(arg, sizeof(*qsa));
4994 if (IS_ERR(qsa)) {
4995 ret = PTR_ERR(qsa);
4996 goto drop_write;
4997 }
4998
4999 if (qsa->flags) {
5000 ret = -EINVAL;
5001 goto out;
5002 }
5003
5004 ret = btrfs_qgroup_rescan(root->fs_info);
5005
5006 out:
5007 kfree(qsa);
5008 drop_write:
5009 mnt_drop_write_file(file);
5010 return ret;
5011 }
5012
5013 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5014 {
5015 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5016 struct btrfs_ioctl_quota_rescan_args *qsa;
5017 int ret = 0;
5018
5019 if (!capable(CAP_SYS_ADMIN))
5020 return -EPERM;
5021
5022 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
5023 if (!qsa)
5024 return -ENOMEM;
5025
5026 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5027 qsa->flags = 1;
5028 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
5029 }
5030
5031 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5032 ret = -EFAULT;
5033
5034 kfree(qsa);
5035 return ret;
5036 }
5037
5038 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5039 {
5040 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5041
5042 if (!capable(CAP_SYS_ADMIN))
5043 return -EPERM;
5044
5045 return btrfs_qgroup_wait_for_completion(root->fs_info);
5046 }
5047
5048 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5049 struct btrfs_ioctl_received_subvol_args *sa)
5050 {
5051 struct inode *inode = file_inode(file);
5052 struct btrfs_root *root = BTRFS_I(inode)->root;
5053 struct btrfs_root_item *root_item = &root->root_item;
5054 struct btrfs_trans_handle *trans;
5055 struct timespec ct = CURRENT_TIME;
5056 int ret = 0;
5057 int received_uuid_changed;
5058
5059 if (!inode_owner_or_capable(inode))
5060 return -EPERM;
5061
5062 ret = mnt_want_write_file(file);
5063 if (ret < 0)
5064 return ret;
5065
5066 down_write(&root->fs_info->subvol_sem);
5067
5068 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
5069 ret = -EINVAL;
5070 goto out;
5071 }
5072
5073 if (btrfs_root_readonly(root)) {
5074 ret = -EROFS;
5075 goto out;
5076 }
5077
5078 /*
5079 * 1 - root item
5080 * 2 - uuid items (received uuid + subvol uuid)
5081 */
5082 trans = btrfs_start_transaction(root, 3);
5083 if (IS_ERR(trans)) {
5084 ret = PTR_ERR(trans);
5085 trans = NULL;
5086 goto out;
5087 }
5088
5089 sa->rtransid = trans->transid;
5090 sa->rtime.sec = ct.tv_sec;
5091 sa->rtime.nsec = ct.tv_nsec;
5092
5093 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5094 BTRFS_UUID_SIZE);
5095 if (received_uuid_changed &&
5096 !btrfs_is_empty_uuid(root_item->received_uuid))
5097 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
5098 root_item->received_uuid,
5099 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5100 root->root_key.objectid);
5101 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5102 btrfs_set_root_stransid(root_item, sa->stransid);
5103 btrfs_set_root_rtransid(root_item, sa->rtransid);
5104 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5105 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5106 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5107 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5108
5109 ret = btrfs_update_root(trans, root->fs_info->tree_root,
5110 &root->root_key, &root->root_item);
5111 if (ret < 0) {
5112 btrfs_end_transaction(trans, root);
5113 goto out;
5114 }
5115 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5116 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
5117 sa->uuid,
5118 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5119 root->root_key.objectid);
5120 if (ret < 0 && ret != -EEXIST) {
5121 btrfs_abort_transaction(trans, root, ret);
5122 goto out;
5123 }
5124 }
5125 ret = btrfs_commit_transaction(trans, root);
5126 if (ret < 0) {
5127 btrfs_abort_transaction(trans, root, ret);
5128 goto out;
5129 }
5130
5131 out:
5132 up_write(&root->fs_info->subvol_sem);
5133 mnt_drop_write_file(file);
5134 return ret;
5135 }
5136
5137 #ifdef CONFIG_64BIT
5138 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5139 void __user *arg)
5140 {
5141 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5142 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5143 int ret = 0;
5144
5145 args32 = memdup_user(arg, sizeof(*args32));
5146 if (IS_ERR(args32)) {
5147 ret = PTR_ERR(args32);
5148 args32 = NULL;
5149 goto out;
5150 }
5151
5152 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
5153 if (!args64) {
5154 ret = -ENOMEM;
5155 goto out;
5156 }
5157
5158 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5159 args64->stransid = args32->stransid;
5160 args64->rtransid = args32->rtransid;
5161 args64->stime.sec = args32->stime.sec;
5162 args64->stime.nsec = args32->stime.nsec;
5163 args64->rtime.sec = args32->rtime.sec;
5164 args64->rtime.nsec = args32->rtime.nsec;
5165 args64->flags = args32->flags;
5166
5167 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5168 if (ret)
5169 goto out;
5170
5171 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5172 args32->stransid = args64->stransid;
5173 args32->rtransid = args64->rtransid;
5174 args32->stime.sec = args64->stime.sec;
5175 args32->stime.nsec = args64->stime.nsec;
5176 args32->rtime.sec = args64->rtime.sec;
5177 args32->rtime.nsec = args64->rtime.nsec;
5178 args32->flags = args64->flags;
5179
5180 ret = copy_to_user(arg, args32, sizeof(*args32));
5181 if (ret)
5182 ret = -EFAULT;
5183
5184 out:
5185 kfree(args32);
5186 kfree(args64);
5187 return ret;
5188 }
5189 #endif
5190
5191 static long btrfs_ioctl_set_received_subvol(struct file *file,
5192 void __user *arg)
5193 {
5194 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5195 int ret = 0;
5196
5197 sa = memdup_user(arg, sizeof(*sa));
5198 if (IS_ERR(sa)) {
5199 ret = PTR_ERR(sa);
5200 sa = NULL;
5201 goto out;
5202 }
5203
5204 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5205
5206 if (ret)
5207 goto out;
5208
5209 ret = copy_to_user(arg, sa, sizeof(*sa));
5210 if (ret)
5211 ret = -EFAULT;
5212
5213 out:
5214 kfree(sa);
5215 return ret;
5216 }
5217
5218 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5219 {
5220 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5221 size_t len;
5222 int ret;
5223 char label[BTRFS_LABEL_SIZE];
5224
5225 spin_lock(&root->fs_info->super_lock);
5226 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5227 spin_unlock(&root->fs_info->super_lock);
5228
5229 len = strnlen(label, BTRFS_LABEL_SIZE);
5230
5231 if (len == BTRFS_LABEL_SIZE) {
5232 btrfs_warn(root->fs_info,
5233 "label is too long, return the first %zu bytes", --len);
5234 }
5235
5236 ret = copy_to_user(arg, label, len);
5237
5238 return ret ? -EFAULT : 0;
5239 }
5240
5241 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5242 {
5243 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5244 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5245 struct btrfs_trans_handle *trans;
5246 char label[BTRFS_LABEL_SIZE];
5247 int ret;
5248
5249 if (!capable(CAP_SYS_ADMIN))
5250 return -EPERM;
5251
5252 if (copy_from_user(label, arg, sizeof(label)))
5253 return -EFAULT;
5254
5255 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5256 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5257 BTRFS_LABEL_SIZE - 1);
5258 return -EINVAL;
5259 }
5260
5261 ret = mnt_want_write_file(file);
5262 if (ret)
5263 return ret;
5264
5265 trans = btrfs_start_transaction(root, 0);
5266 if (IS_ERR(trans)) {
5267 ret = PTR_ERR(trans);
5268 goto out_unlock;
5269 }
5270
5271 spin_lock(&root->fs_info->super_lock);
5272 strcpy(super_block->label, label);
5273 spin_unlock(&root->fs_info->super_lock);
5274 ret = btrfs_commit_transaction(trans, root);
5275
5276 out_unlock:
5277 mnt_drop_write_file(file);
5278 return ret;
5279 }
5280
5281 #define INIT_FEATURE_FLAGS(suffix) \
5282 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5283 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5284 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5285
5286 static int btrfs_ioctl_get_supported_features(struct file *file,
5287 void __user *arg)
5288 {
5289 static struct btrfs_ioctl_feature_flags features[3] = {
5290 INIT_FEATURE_FLAGS(SUPP),
5291 INIT_FEATURE_FLAGS(SAFE_SET),
5292 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5293 };
5294
5295 if (copy_to_user(arg, &features, sizeof(features)))
5296 return -EFAULT;
5297
5298 return 0;
5299 }
5300
5301 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5302 {
5303 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5304 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5305 struct btrfs_ioctl_feature_flags features;
5306
5307 features.compat_flags = btrfs_super_compat_flags(super_block);
5308 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5309 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5310
5311 if (copy_to_user(arg, &features, sizeof(features)))
5312 return -EFAULT;
5313
5314 return 0;
5315 }
5316
5317 static int check_feature_bits(struct btrfs_root *root,
5318 enum btrfs_feature_set set,
5319 u64 change_mask, u64 flags, u64 supported_flags,
5320 u64 safe_set, u64 safe_clear)
5321 {
5322 const char *type = btrfs_feature_set_names[set];
5323 char *names;
5324 u64 disallowed, unsupported;
5325 u64 set_mask = flags & change_mask;
5326 u64 clear_mask = ~flags & change_mask;
5327
5328 unsupported = set_mask & ~supported_flags;
5329 if (unsupported) {
5330 names = btrfs_printable_features(set, unsupported);
5331 if (names) {
5332 btrfs_warn(root->fs_info,
5333 "this kernel does not support the %s feature bit%s",
5334 names, strchr(names, ',') ? "s" : "");
5335 kfree(names);
5336 } else
5337 btrfs_warn(root->fs_info,
5338 "this kernel does not support %s bits 0x%llx",
5339 type, unsupported);
5340 return -EOPNOTSUPP;
5341 }
5342
5343 disallowed = set_mask & ~safe_set;
5344 if (disallowed) {
5345 names = btrfs_printable_features(set, disallowed);
5346 if (names) {
5347 btrfs_warn(root->fs_info,
5348 "can't set the %s feature bit%s while mounted",
5349 names, strchr(names, ',') ? "s" : "");
5350 kfree(names);
5351 } else
5352 btrfs_warn(root->fs_info,
5353 "can't set %s bits 0x%llx while mounted",
5354 type, disallowed);
5355 return -EPERM;
5356 }
5357
5358 disallowed = clear_mask & ~safe_clear;
5359 if (disallowed) {
5360 names = btrfs_printable_features(set, disallowed);
5361 if (names) {
5362 btrfs_warn(root->fs_info,
5363 "can't clear the %s feature bit%s while mounted",
5364 names, strchr(names, ',') ? "s" : "");
5365 kfree(names);
5366 } else
5367 btrfs_warn(root->fs_info,
5368 "can't clear %s bits 0x%llx while mounted",
5369 type, disallowed);
5370 return -EPERM;
5371 }
5372
5373 return 0;
5374 }
5375
5376 #define check_feature(root, change_mask, flags, mask_base) \
5377 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5378 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5379 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5380 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5381
5382 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5383 {
5384 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5385 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5386 struct btrfs_ioctl_feature_flags flags[2];
5387 struct btrfs_trans_handle *trans;
5388 u64 newflags;
5389 int ret;
5390
5391 if (!capable(CAP_SYS_ADMIN))
5392 return -EPERM;
5393
5394 if (copy_from_user(flags, arg, sizeof(flags)))
5395 return -EFAULT;
5396
5397 /* Nothing to do */
5398 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5399 !flags[0].incompat_flags)
5400 return 0;
5401
5402 ret = check_feature(root, flags[0].compat_flags,
5403 flags[1].compat_flags, COMPAT);
5404 if (ret)
5405 return ret;
5406
5407 ret = check_feature(root, flags[0].compat_ro_flags,
5408 flags[1].compat_ro_flags, COMPAT_RO);
5409 if (ret)
5410 return ret;
5411
5412 ret = check_feature(root, flags[0].incompat_flags,
5413 flags[1].incompat_flags, INCOMPAT);
5414 if (ret)
5415 return ret;
5416
5417 trans = btrfs_start_transaction(root, 0);
5418 if (IS_ERR(trans))
5419 return PTR_ERR(trans);
5420
5421 spin_lock(&root->fs_info->super_lock);
5422 newflags = btrfs_super_compat_flags(super_block);
5423 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5424 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5425 btrfs_set_super_compat_flags(super_block, newflags);
5426
5427 newflags = btrfs_super_compat_ro_flags(super_block);
5428 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5429 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5430 btrfs_set_super_compat_ro_flags(super_block, newflags);
5431
5432 newflags = btrfs_super_incompat_flags(super_block);
5433 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5434 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5435 btrfs_set_super_incompat_flags(super_block, newflags);
5436 spin_unlock(&root->fs_info->super_lock);
5437
5438 return btrfs_commit_transaction(trans, root);
5439 }
5440
5441 long btrfs_ioctl(struct file *file, unsigned int
5442 cmd, unsigned long arg)
5443 {
5444 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5445 void __user *argp = (void __user *)arg;
5446
5447 switch (cmd) {
5448 case FS_IOC_GETFLAGS:
5449 return btrfs_ioctl_getflags(file, argp);
5450 case FS_IOC_SETFLAGS:
5451 return btrfs_ioctl_setflags(file, argp);
5452 case FS_IOC_GETVERSION:
5453 return btrfs_ioctl_getversion(file, argp);
5454 case FITRIM:
5455 return btrfs_ioctl_fitrim(file, argp);
5456 case BTRFS_IOC_SNAP_CREATE:
5457 return btrfs_ioctl_snap_create(file, argp, 0);
5458 case BTRFS_IOC_SNAP_CREATE_V2:
5459 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5460 case BTRFS_IOC_SUBVOL_CREATE:
5461 return btrfs_ioctl_snap_create(file, argp, 1);
5462 case BTRFS_IOC_SUBVOL_CREATE_V2:
5463 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5464 case BTRFS_IOC_SNAP_DESTROY:
5465 return btrfs_ioctl_snap_destroy(file, argp);
5466 case BTRFS_IOC_SUBVOL_GETFLAGS:
5467 return btrfs_ioctl_subvol_getflags(file, argp);
5468 case BTRFS_IOC_SUBVOL_SETFLAGS:
5469 return btrfs_ioctl_subvol_setflags(file, argp);
5470 case BTRFS_IOC_DEFAULT_SUBVOL:
5471 return btrfs_ioctl_default_subvol(file, argp);
5472 case BTRFS_IOC_DEFRAG:
5473 return btrfs_ioctl_defrag(file, NULL);
5474 case BTRFS_IOC_DEFRAG_RANGE:
5475 return btrfs_ioctl_defrag(file, argp);
5476 case BTRFS_IOC_RESIZE:
5477 return btrfs_ioctl_resize(file, argp);
5478 case BTRFS_IOC_ADD_DEV:
5479 return btrfs_ioctl_add_dev(root, argp);
5480 case BTRFS_IOC_RM_DEV:
5481 return btrfs_ioctl_rm_dev(file, argp);
5482 case BTRFS_IOC_FS_INFO:
5483 return btrfs_ioctl_fs_info(root, argp);
5484 case BTRFS_IOC_DEV_INFO:
5485 return btrfs_ioctl_dev_info(root, argp);
5486 case BTRFS_IOC_BALANCE:
5487 return btrfs_ioctl_balance(file, NULL);
5488 case BTRFS_IOC_CLONE:
5489 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5490 case BTRFS_IOC_CLONE_RANGE:
5491 return btrfs_ioctl_clone_range(file, argp);
5492 case BTRFS_IOC_TRANS_START:
5493 return btrfs_ioctl_trans_start(file);
5494 case BTRFS_IOC_TRANS_END:
5495 return btrfs_ioctl_trans_end(file);
5496 case BTRFS_IOC_TREE_SEARCH:
5497 return btrfs_ioctl_tree_search(file, argp);
5498 case BTRFS_IOC_TREE_SEARCH_V2:
5499 return btrfs_ioctl_tree_search_v2(file, argp);
5500 case BTRFS_IOC_INO_LOOKUP:
5501 return btrfs_ioctl_ino_lookup(file, argp);
5502 case BTRFS_IOC_INO_PATHS:
5503 return btrfs_ioctl_ino_to_path(root, argp);
5504 case BTRFS_IOC_LOGICAL_INO:
5505 return btrfs_ioctl_logical_to_ino(root, argp);
5506 case BTRFS_IOC_SPACE_INFO:
5507 return btrfs_ioctl_space_info(root, argp);
5508 case BTRFS_IOC_SYNC: {
5509 int ret;
5510
5511 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5512 if (ret)
5513 return ret;
5514 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
5515 /*
5516 * The transaction thread may want to do more work,
5517 * namely it pokes the cleaner ktread that will start
5518 * processing uncleaned subvols.
5519 */
5520 wake_up_process(root->fs_info->transaction_kthread);
5521 return ret;
5522 }
5523 case BTRFS_IOC_START_SYNC:
5524 return btrfs_ioctl_start_sync(root, argp);
5525 case BTRFS_IOC_WAIT_SYNC:
5526 return btrfs_ioctl_wait_sync(root, argp);
5527 case BTRFS_IOC_SCRUB:
5528 return btrfs_ioctl_scrub(file, argp);
5529 case BTRFS_IOC_SCRUB_CANCEL:
5530 return btrfs_ioctl_scrub_cancel(root, argp);
5531 case BTRFS_IOC_SCRUB_PROGRESS:
5532 return btrfs_ioctl_scrub_progress(root, argp);
5533 case BTRFS_IOC_BALANCE_V2:
5534 return btrfs_ioctl_balance(file, argp);
5535 case BTRFS_IOC_BALANCE_CTL:
5536 return btrfs_ioctl_balance_ctl(root, arg);
5537 case BTRFS_IOC_BALANCE_PROGRESS:
5538 return btrfs_ioctl_balance_progress(root, argp);
5539 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5540 return btrfs_ioctl_set_received_subvol(file, argp);
5541 #ifdef CONFIG_64BIT
5542 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5543 return btrfs_ioctl_set_received_subvol_32(file, argp);
5544 #endif
5545 case BTRFS_IOC_SEND:
5546 return btrfs_ioctl_send(file, argp);
5547 case BTRFS_IOC_GET_DEV_STATS:
5548 return btrfs_ioctl_get_dev_stats(root, argp);
5549 case BTRFS_IOC_QUOTA_CTL:
5550 return btrfs_ioctl_quota_ctl(file, argp);
5551 case BTRFS_IOC_QGROUP_ASSIGN:
5552 return btrfs_ioctl_qgroup_assign(file, argp);
5553 case BTRFS_IOC_QGROUP_CREATE:
5554 return btrfs_ioctl_qgroup_create(file, argp);
5555 case BTRFS_IOC_QGROUP_LIMIT:
5556 return btrfs_ioctl_qgroup_limit(file, argp);
5557 case BTRFS_IOC_QUOTA_RESCAN:
5558 return btrfs_ioctl_quota_rescan(file, argp);
5559 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5560 return btrfs_ioctl_quota_rescan_status(file, argp);
5561 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5562 return btrfs_ioctl_quota_rescan_wait(file, argp);
5563 case BTRFS_IOC_DEV_REPLACE:
5564 return btrfs_ioctl_dev_replace(root, argp);
5565 case BTRFS_IOC_GET_FSLABEL:
5566 return btrfs_ioctl_get_fslabel(file, argp);
5567 case BTRFS_IOC_SET_FSLABEL:
5568 return btrfs_ioctl_set_fslabel(file, argp);
5569 case BTRFS_IOC_FILE_EXTENT_SAME:
5570 return btrfs_ioctl_file_extent_same(file, argp);
5571 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5572 return btrfs_ioctl_get_supported_features(file, argp);
5573 case BTRFS_IOC_GET_FEATURES:
5574 return btrfs_ioctl_get_features(file, argp);
5575 case BTRFS_IOC_SET_FEATURES:
5576 return btrfs_ioctl_set_features(file, argp);
5577 }
5578
5579 return -ENOTTY;
5580 }
Michael's Linux tree.