projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
orangefs: sanitize ->llseek()
[deliverable/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 inode_lock(inode);
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 inode_unlock(inode);
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 mutex_lock(&new_root->objectid_mutex);
572 new_root->highest_objectid = new_dirid;
573 mutex_unlock(&new_root->objectid_mutex);
574
575 /*
576 * insert the directory item
577 */
578 ret = btrfs_set_inode_index(dir, &index);
579 if (ret) {
580 btrfs_abort_transaction(trans, root, ret);
581 goto fail;
582 }
583
584 ret = btrfs_insert_dir_item(trans, root,
585 name, namelen, dir, &key,
586 BTRFS_FT_DIR, index);
587 if (ret) {
588 btrfs_abort_transaction(trans, root, ret);
589 goto fail;
590 }
591
592 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
593 ret = btrfs_update_inode(trans, root, dir);
594 BUG_ON(ret);
595
596 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
597 objectid, root->root_key.objectid,
598 btrfs_ino(dir), index, name, namelen);
599 BUG_ON(ret);
600
601 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
602 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
603 objectid);
604 if (ret)
605 btrfs_abort_transaction(trans, root, ret);
606
607 fail:
608 trans->block_rsv = NULL;
609 trans->bytes_reserved = 0;
610 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
611
612 if (async_transid) {
613 *async_transid = trans->transid;
614 err = btrfs_commit_transaction_async(trans, root, 1);
615 if (err)
616 err = btrfs_commit_transaction(trans, root);
617 } else {
618 err = btrfs_commit_transaction(trans, root);
619 }
620 if (err && !ret)
621 ret = err;
622
623 if (!ret) {
624 inode = btrfs_lookup_dentry(dir, dentry);
625 if (IS_ERR(inode))
626 return PTR_ERR(inode);
627 d_instantiate(dentry, inode);
628 }
629 return ret;
630 }
631
632 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
633 {
634 s64 writers;
635 DEFINE_WAIT(wait);
636
637 do {
638 prepare_to_wait(&root->subv_writers->wait, &wait,
639 TASK_UNINTERRUPTIBLE);
640
641 writers = percpu_counter_sum(&root->subv_writers->counter);
642 if (writers)
643 schedule();
644
645 finish_wait(&root->subv_writers->wait, &wait);
646 } while (writers);
647 }
648
649 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
650 struct dentry *dentry, char *name, int namelen,
651 u64 *async_transid, bool readonly,
652 struct btrfs_qgroup_inherit *inherit)
653 {
654 struct inode *inode;
655 struct btrfs_pending_snapshot *pending_snapshot;
656 struct btrfs_trans_handle *trans;
657 int ret;
658
659 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
660 return -EINVAL;
661
662 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
663 if (!pending_snapshot)
664 return -ENOMEM;
665
666 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
667 GFP_NOFS);
668 pending_snapshot->path = btrfs_alloc_path();
669 if (!pending_snapshot->root_item || !pending_snapshot->path) {
670 ret = -ENOMEM;
671 goto free_pending;
672 }
673
674 atomic_inc(&root->will_be_snapshoted);
675 smp_mb__after_atomic();
676 btrfs_wait_for_no_snapshoting_writes(root);
677
678 ret = btrfs_start_delalloc_inodes(root, 0);
679 if (ret)
680 goto dec_and_free;
681
682 btrfs_wait_ordered_extents(root, -1);
683
684 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
685 BTRFS_BLOCK_RSV_TEMP);
686 /*
687 * 1 - parent dir inode
688 * 2 - dir entries
689 * 1 - root item
690 * 2 - root ref/backref
691 * 1 - root of snapshot
692 * 1 - UUID item
693 */
694 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
695 &pending_snapshot->block_rsv, 8,
696 &pending_snapshot->qgroup_reserved,
697 false);
698 if (ret)
699 goto dec_and_free;
700
701 pending_snapshot->dentry = dentry;
702 pending_snapshot->root = root;
703 pending_snapshot->readonly = readonly;
704 pending_snapshot->dir = dir;
705 pending_snapshot->inherit = inherit;
706
707 trans = btrfs_start_transaction(root, 0);
708 if (IS_ERR(trans)) {
709 ret = PTR_ERR(trans);
710 goto fail;
711 }
712
713 spin_lock(&root->fs_info->trans_lock);
714 list_add(&pending_snapshot->list,
715 &trans->transaction->pending_snapshots);
716 spin_unlock(&root->fs_info->trans_lock);
717 if (async_transid) {
718 *async_transid = trans->transid;
719 ret = btrfs_commit_transaction_async(trans,
720 root->fs_info->extent_root, 1);
721 if (ret)
722 ret = btrfs_commit_transaction(trans, root);
723 } else {
724 ret = btrfs_commit_transaction(trans,
725 root->fs_info->extent_root);
726 }
727 if (ret)
728 goto fail;
729
730 ret = pending_snapshot->error;
731 if (ret)
732 goto fail;
733
734 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
735 if (ret)
736 goto fail;
737
738 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
739 if (IS_ERR(inode)) {
740 ret = PTR_ERR(inode);
741 goto fail;
742 }
743
744 d_instantiate(dentry, inode);
745 ret = 0;
746 fail:
747 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
748 &pending_snapshot->block_rsv,
749 pending_snapshot->qgroup_reserved);
750 dec_and_free:
751 if (atomic_dec_and_test(&root->will_be_snapshoted))
752 wake_up_atomic_t(&root->will_be_snapshoted);
753 free_pending:
754 kfree(pending_snapshot->root_item);
755 btrfs_free_path(pending_snapshot->path);
756 kfree(pending_snapshot);
757
758 return ret;
759 }
760
761 /* copy of may_delete in fs/namei.c()
762 * Check whether we can remove a link victim from directory dir, check
763 * whether the type of victim is right.
764 * 1. We can't do it if dir is read-only (done in permission())
765 * 2. We should have write and exec permissions on dir
766 * 3. We can't remove anything from append-only dir
767 * 4. We can't do anything with immutable dir (done in permission())
768 * 5. If the sticky bit on dir is set we should either
769 * a. be owner of dir, or
770 * b. be owner of victim, or
771 * c. have CAP_FOWNER capability
772 * 6. If the victim is append-only or immutable we can't do antyhing with
773 * links pointing to it.
774 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
775 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
776 * 9. We can't remove a root or mountpoint.
777 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
778 * nfs_async_unlink().
779 */
780
781 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
782 {
783 int error;
784
785 if (d_really_is_negative(victim))
786 return -ENOENT;
787
788 BUG_ON(d_inode(victim->d_parent) != dir);
789 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
790
791 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
792 if (error)
793 return error;
794 if (IS_APPEND(dir))
795 return -EPERM;
796 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
797 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
798 return -EPERM;
799 if (isdir) {
800 if (!d_is_dir(victim))
801 return -ENOTDIR;
802 if (IS_ROOT(victim))
803 return -EBUSY;
804 } else if (d_is_dir(victim))
805 return -EISDIR;
806 if (IS_DEADDIR(dir))
807 return -ENOENT;
808 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
809 return -EBUSY;
810 return 0;
811 }
812
813 /* copy of may_create in fs/namei.c() */
814 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
815 {
816 if (d_really_is_positive(child))
817 return -EEXIST;
818 if (IS_DEADDIR(dir))
819 return -ENOENT;
820 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
821 }
822
823 /*
824 * Create a new subvolume below @parent. This is largely modeled after
825 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
826 * inside this filesystem so it's quite a bit simpler.
827 */
828 static noinline int btrfs_mksubvol(struct path *parent,
829 char *name, int namelen,
830 struct btrfs_root *snap_src,
831 u64 *async_transid, bool readonly,
832 struct btrfs_qgroup_inherit *inherit)
833 {
834 struct inode *dir = d_inode(parent->dentry);
835 struct dentry *dentry;
836 int error;
837
838 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
839 if (error == -EINTR)
840 return error;
841
842 dentry = lookup_one_len(name, parent->dentry, namelen);
843 error = PTR_ERR(dentry);
844 if (IS_ERR(dentry))
845 goto out_unlock;
846
847 error = -EEXIST;
848 if (d_really_is_positive(dentry))
849 goto out_dput;
850
851 error = btrfs_may_create(dir, dentry);
852 if (error)
853 goto out_dput;
854
855 /*
856 * even if this name doesn't exist, we may get hash collisions.
857 * check for them now when we can safely fail
858 */
859 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
860 dir->i_ino, name,
861 namelen);
862 if (error)
863 goto out_dput;
864
865 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
866
867 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
868 goto out_up_read;
869
870 if (snap_src) {
871 error = create_snapshot(snap_src, dir, dentry, name, namelen,
872 async_transid, readonly, inherit);
873 } else {
874 error = create_subvol(dir, dentry, name, namelen,
875 async_transid, inherit);
876 }
877 if (!error)
878 fsnotify_mkdir(dir, dentry);
879 out_up_read:
880 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
881 out_dput:
882 dput(dentry);
883 out_unlock:
884 inode_unlock(dir);
885 return error;
886 }
887
888 /*
889 * When we're defragging a range, we don't want to kick it off again
890 * if it is really just waiting for delalloc to send it down.
891 * If we find a nice big extent or delalloc range for the bytes in the
892 * file you want to defrag, we return 0 to let you know to skip this
893 * part of the file
894 */
895 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
896 {
897 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
898 struct extent_map *em = NULL;
899 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
900 u64 end;
901
902 read_lock(&em_tree->lock);
903 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
904 read_unlock(&em_tree->lock);
905
906 if (em) {
907 end = extent_map_end(em);
908 free_extent_map(em);
909 if (end - offset > thresh)
910 return 0;
911 }
912 /* if we already have a nice delalloc here, just stop */
913 thresh /= 2;
914 end = count_range_bits(io_tree, &offset, offset + thresh,
915 thresh, EXTENT_DELALLOC, 1);
916 if (end >= thresh)
917 return 0;
918 return 1;
919 }
920
921 /*
922 * helper function to walk through a file and find extents
923 * newer than a specific transid, and smaller than thresh.
924 *
925 * This is used by the defragging code to find new and small
926 * extents
927 */
928 static int find_new_extents(struct btrfs_root *root,
929 struct inode *inode, u64 newer_than,
930 u64 *off, u32 thresh)
931 {
932 struct btrfs_path *path;
933 struct btrfs_key min_key;
934 struct extent_buffer *leaf;
935 struct btrfs_file_extent_item *extent;
936 int type;
937 int ret;
938 u64 ino = btrfs_ino(inode);
939
940 path = btrfs_alloc_path();
941 if (!path)
942 return -ENOMEM;
943
944 min_key.objectid = ino;
945 min_key.type = BTRFS_EXTENT_DATA_KEY;
946 min_key.offset = *off;
947
948 while (1) {
949 ret = btrfs_search_forward(root, &min_key, path, newer_than);
950 if (ret != 0)
951 goto none;
952 process_slot:
953 if (min_key.objectid != ino)
954 goto none;
955 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
956 goto none;
957
958 leaf = path->nodes[0];
959 extent = btrfs_item_ptr(leaf, path->slots[0],
960 struct btrfs_file_extent_item);
961
962 type = btrfs_file_extent_type(leaf, extent);
963 if (type == BTRFS_FILE_EXTENT_REG &&
964 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
965 check_defrag_in_cache(inode, min_key.offset, thresh)) {
966 *off = min_key.offset;
967 btrfs_free_path(path);
968 return 0;
969 }
970
971 path->slots[0]++;
972 if (path->slots[0] < btrfs_header_nritems(leaf)) {
973 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
974 goto process_slot;
975 }
976
977 if (min_key.offset == (u64)-1)
978 goto none;
979
980 min_key.offset++;
981 btrfs_release_path(path);
982 }
983 none:
984 btrfs_free_path(path);
985 return -ENOENT;
986 }
987
988 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
989 {
990 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
991 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
992 struct extent_map *em;
993 u64 len = PAGE_CACHE_SIZE;
994
995 /*
996 * hopefully we have this extent in the tree already, try without
997 * the full extent lock
998 */
999 read_lock(&em_tree->lock);
1000 em = lookup_extent_mapping(em_tree, start, len);
1001 read_unlock(&em_tree->lock);
1002
1003 if (!em) {
1004 struct extent_state *cached = NULL;
1005 u64 end = start + len - 1;
1006
1007 /* get the big lock and read metadata off disk */
1008 lock_extent_bits(io_tree, start, end, &cached);
1009 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1010 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1011
1012 if (IS_ERR(em))
1013 return NULL;
1014 }
1015
1016 return em;
1017 }
1018
1019 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1020 {
1021 struct extent_map *next;
1022 bool ret = true;
1023
1024 /* this is the last extent */
1025 if (em->start + em->len >= i_size_read(inode))
1026 return false;
1027
1028 next = defrag_lookup_extent(inode, em->start + em->len);
1029 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1030 ret = false;
1031 else if ((em->block_start + em->block_len == next->block_start) &&
1032 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1033 ret = false;
1034
1035 free_extent_map(next);
1036 return ret;
1037 }
1038
1039 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1040 u64 *last_len, u64 *skip, u64 *defrag_end,
1041 int compress)
1042 {
1043 struct extent_map *em;
1044 int ret = 1;
1045 bool next_mergeable = true;
1046 bool prev_mergeable = true;
1047
1048 /*
1049 * make sure that once we start defragging an extent, we keep on
1050 * defragging it
1051 */
1052 if (start < *defrag_end)
1053 return 1;
1054
1055 *skip = 0;
1056
1057 em = defrag_lookup_extent(inode, start);
1058 if (!em)
1059 return 0;
1060
1061 /* this will cover holes, and inline extents */
1062 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1063 ret = 0;
1064 goto out;
1065 }
1066
1067 if (!*defrag_end)
1068 prev_mergeable = false;
1069
1070 next_mergeable = defrag_check_next_extent(inode, em);
1071 /*
1072 * we hit a real extent, if it is big or the next extent is not a
1073 * real extent, don't bother defragging it
1074 */
1075 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1076 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1077 ret = 0;
1078 out:
1079 /*
1080 * last_len ends up being a counter of how many bytes we've defragged.
1081 * every time we choose not to defrag an extent, we reset *last_len
1082 * so that the next tiny extent will force a defrag.
1083 *
1084 * The end result of this is that tiny extents before a single big
1085 * extent will force at least part of that big extent to be defragged.
1086 */
1087 if (ret) {
1088 *defrag_end = extent_map_end(em);
1089 } else {
1090 *last_len = 0;
1091 *skip = extent_map_end(em);
1092 *defrag_end = 0;
1093 }
1094
1095 free_extent_map(em);
1096 return ret;
1097 }
1098
1099 /*
1100 * it doesn't do much good to defrag one or two pages
1101 * at a time. This pulls in a nice chunk of pages
1102 * to COW and defrag.
1103 *
1104 * It also makes sure the delalloc code has enough
1105 * dirty data to avoid making new small extents as part
1106 * of the defrag
1107 *
1108 * It's a good idea to start RA on this range
1109 * before calling this.
1110 */
1111 static int cluster_pages_for_defrag(struct inode *inode,
1112 struct page **pages,
1113 unsigned long start_index,
1114 unsigned long num_pages)
1115 {
1116 unsigned long file_end;
1117 u64 isize = i_size_read(inode);
1118 u64 page_start;
1119 u64 page_end;
1120 u64 page_cnt;
1121 int ret;
1122 int i;
1123 int i_done;
1124 struct btrfs_ordered_extent *ordered;
1125 struct extent_state *cached_state = NULL;
1126 struct extent_io_tree *tree;
1127 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1128
1129 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1130 if (!isize || start_index > file_end)
1131 return 0;
1132
1133 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1134
1135 ret = btrfs_delalloc_reserve_space(inode,
1136 start_index << PAGE_CACHE_SHIFT,
1137 page_cnt << PAGE_CACHE_SHIFT);
1138 if (ret)
1139 return ret;
1140 i_done = 0;
1141 tree = &BTRFS_I(inode)->io_tree;
1142
1143 /* step one, lock all the pages */
1144 for (i = 0; i < page_cnt; i++) {
1145 struct page *page;
1146 again:
1147 page = find_or_create_page(inode->i_mapping,
1148 start_index + i, mask);
1149 if (!page)
1150 break;
1151
1152 page_start = page_offset(page);
1153 page_end = page_start + PAGE_CACHE_SIZE - 1;
1154 while (1) {
1155 lock_extent_bits(tree, page_start, page_end,
1156 &cached_state);
1157 ordered = btrfs_lookup_ordered_extent(inode,
1158 page_start);
1159 unlock_extent_cached(tree, page_start, page_end,
1160 &cached_state, GFP_NOFS);
1161 if (!ordered)
1162 break;
1163
1164 unlock_page(page);
1165 btrfs_start_ordered_extent(inode, ordered, 1);
1166 btrfs_put_ordered_extent(ordered);
1167 lock_page(page);
1168 /*
1169 * we unlocked the page above, so we need check if
1170 * it was released or not.
1171 */
1172 if (page->mapping != inode->i_mapping) {
1173 unlock_page(page);
1174 page_cache_release(page);
1175 goto again;
1176 }
1177 }
1178
1179 if (!PageUptodate(page)) {
1180 btrfs_readpage(NULL, page);
1181 lock_page(page);
1182 if (!PageUptodate(page)) {
1183 unlock_page(page);
1184 page_cache_release(page);
1185 ret = -EIO;
1186 break;
1187 }
1188 }
1189
1190 if (page->mapping != inode->i_mapping) {
1191 unlock_page(page);
1192 page_cache_release(page);
1193 goto again;
1194 }
1195
1196 pages[i] = page;
1197 i_done++;
1198 }
1199 if (!i_done || ret)
1200 goto out;
1201
1202 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1203 goto out;
1204
1205 /*
1206 * so now we have a nice long stream of locked
1207 * and up to date pages, lets wait on them
1208 */
1209 for (i = 0; i < i_done; i++)
1210 wait_on_page_writeback(pages[i]);
1211
1212 page_start = page_offset(pages[0]);
1213 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1214
1215 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1216 page_start, page_end - 1, &cached_state);
1217 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1218 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1219 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1220 &cached_state, GFP_NOFS);
1221
1222 if (i_done != page_cnt) {
1223 spin_lock(&BTRFS_I(inode)->lock);
1224 BTRFS_I(inode)->outstanding_extents++;
1225 spin_unlock(&BTRFS_I(inode)->lock);
1226 btrfs_delalloc_release_space(inode,
1227 start_index << PAGE_CACHE_SHIFT,
1228 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1229 }
1230
1231
1232 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1233 &cached_state, GFP_NOFS);
1234
1235 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1236 page_start, page_end - 1, &cached_state,
1237 GFP_NOFS);
1238
1239 for (i = 0; i < i_done; i++) {
1240 clear_page_dirty_for_io(pages[i]);
1241 ClearPageChecked(pages[i]);
1242 set_page_extent_mapped(pages[i]);
1243 set_page_dirty(pages[i]);
1244 unlock_page(pages[i]);
1245 page_cache_release(pages[i]);
1246 }
1247 return i_done;
1248 out:
1249 for (i = 0; i < i_done; i++) {
1250 unlock_page(pages[i]);
1251 page_cache_release(pages[i]);
1252 }
1253 btrfs_delalloc_release_space(inode,
1254 start_index << PAGE_CACHE_SHIFT,
1255 page_cnt << PAGE_CACHE_SHIFT);
1256 return ret;
1257
1258 }
1259
1260 int btrfs_defrag_file(struct inode *inode, struct file *file,
1261 struct btrfs_ioctl_defrag_range_args *range,
1262 u64 newer_than, unsigned long max_to_defrag)
1263 {
1264 struct btrfs_root *root = BTRFS_I(inode)->root;
1265 struct file_ra_state *ra = NULL;
1266 unsigned long last_index;
1267 u64 isize = i_size_read(inode);
1268 u64 last_len = 0;
1269 u64 skip = 0;
1270 u64 defrag_end = 0;
1271 u64 newer_off = range->start;
1272 unsigned long i;
1273 unsigned long ra_index = 0;
1274 int ret;
1275 int defrag_count = 0;
1276 int compress_type = BTRFS_COMPRESS_ZLIB;
1277 u32 extent_thresh = range->extent_thresh;
1278 unsigned long max_cluster = SZ_256K >> PAGE_CACHE_SHIFT;
1279 unsigned long cluster = max_cluster;
1280 u64 new_align = ~((u64)SZ_128K - 1);
1281 struct page **pages = NULL;
1282
1283 if (isize == 0)
1284 return 0;
1285
1286 if (range->start >= isize)
1287 return -EINVAL;
1288
1289 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1290 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1291 return -EINVAL;
1292 if (range->compress_type)
1293 compress_type = range->compress_type;
1294 }
1295
1296 if (extent_thresh == 0)
1297 extent_thresh = SZ_256K;
1298
1299 /*
1300 * if we were not given a file, allocate a readahead
1301 * context
1302 */
1303 if (!file) {
1304 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1305 if (!ra)
1306 return -ENOMEM;
1307 file_ra_state_init(ra, inode->i_mapping);
1308 } else {
1309 ra = &file->f_ra;
1310 }
1311
1312 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1313 GFP_NOFS);
1314 if (!pages) {
1315 ret = -ENOMEM;
1316 goto out_ra;
1317 }
1318
1319 /* find the last page to defrag */
1320 if (range->start + range->len > range->start) {
1321 last_index = min_t(u64, isize - 1,
1322 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1323 } else {
1324 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1325 }
1326
1327 if (newer_than) {
1328 ret = find_new_extents(root, inode, newer_than,
1329 &newer_off, SZ_64K);
1330 if (!ret) {
1331 range->start = newer_off;
1332 /*
1333 * we always align our defrag to help keep
1334 * the extents in the file evenly spaced
1335 */
1336 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1337 } else
1338 goto out_ra;
1339 } else {
1340 i = range->start >> PAGE_CACHE_SHIFT;
1341 }
1342 if (!max_to_defrag)
1343 max_to_defrag = last_index - i + 1;
1344
1345 /*
1346 * make writeback starts from i, so the defrag range can be
1347 * written sequentially.
1348 */
1349 if (i < inode->i_mapping->writeback_index)
1350 inode->i_mapping->writeback_index = i;
1351
1352 while (i <= last_index && defrag_count < max_to_defrag &&
1353 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
1354 /*
1355 * make sure we stop running if someone unmounts
1356 * the FS
1357 */
1358 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1359 break;
1360
1361 if (btrfs_defrag_cancelled(root->fs_info)) {
1362 btrfs_debug(root->fs_info, "defrag_file cancelled");
1363 ret = -EAGAIN;
1364 break;
1365 }
1366
1367 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1368 extent_thresh, &last_len, &skip,
1369 &defrag_end, range->flags &
1370 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1371 unsigned long next;
1372 /*
1373 * the should_defrag function tells us how much to skip
1374 * bump our counter by the suggested amount
1375 */
1376 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
1377 i = max(i + 1, next);
1378 continue;
1379 }
1380
1381 if (!newer_than) {
1382 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1383 PAGE_CACHE_SHIFT) - i;
1384 cluster = min(cluster, max_cluster);
1385 } else {
1386 cluster = max_cluster;
1387 }
1388
1389 if (i + cluster > ra_index) {
1390 ra_index = max(i, ra_index);
1391 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1392 cluster);
1393 ra_index += cluster;
1394 }
1395
1396 inode_lock(inode);
1397 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1398 BTRFS_I(inode)->force_compress = compress_type;
1399 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1400 if (ret < 0) {
1401 inode_unlock(inode);
1402 goto out_ra;
1403 }
1404
1405 defrag_count += ret;
1406 balance_dirty_pages_ratelimited(inode->i_mapping);
1407 inode_unlock(inode);
1408
1409 if (newer_than) {
1410 if (newer_off == (u64)-1)
1411 break;
1412
1413 if (ret > 0)
1414 i += ret;
1415
1416 newer_off = max(newer_off + 1,
1417 (u64)i << PAGE_CACHE_SHIFT);
1418
1419 ret = find_new_extents(root, inode, newer_than,
1420 &newer_off, SZ_64K);
1421 if (!ret) {
1422 range->start = newer_off;
1423 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1424 } else {
1425 break;
1426 }
1427 } else {
1428 if (ret > 0) {
1429 i += ret;
1430 last_len += ret << PAGE_CACHE_SHIFT;
1431 } else {
1432 i++;
1433 last_len = 0;
1434 }
1435 }
1436 }
1437
1438 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1439 filemap_flush(inode->i_mapping);
1440 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1441 &BTRFS_I(inode)->runtime_flags))
1442 filemap_flush(inode->i_mapping);
1443 }
1444
1445 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1446 /* the filemap_flush will queue IO into the worker threads, but
1447 * we have to make sure the IO is actually started and that
1448 * ordered extents get created before we return
1449 */
1450 atomic_inc(&root->fs_info->async_submit_draining);
1451 while (atomic_read(&root->fs_info->nr_async_submits) ||
1452 atomic_read(&root->fs_info->async_delalloc_pages)) {
1453 wait_event(root->fs_info->async_submit_wait,
1454 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1455 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1456 }
1457 atomic_dec(&root->fs_info->async_submit_draining);
1458 }
1459
1460 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1461 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1462 }
1463
1464 ret = defrag_count;
1465
1466 out_ra:
1467 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1468 inode_lock(inode);
1469 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1470 inode_unlock(inode);
1471 }
1472 if (!file)
1473 kfree(ra);
1474 kfree(pages);
1475 return ret;
1476 }
1477
1478 static noinline int btrfs_ioctl_resize(struct file *file,
1479 void __user *arg)
1480 {
1481 u64 new_size;
1482 u64 old_size;
1483 u64 devid = 1;
1484 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1485 struct btrfs_ioctl_vol_args *vol_args;
1486 struct btrfs_trans_handle *trans;
1487 struct btrfs_device *device = NULL;
1488 char *sizestr;
1489 char *retptr;
1490 char *devstr = NULL;
1491 int ret = 0;
1492 int mod = 0;
1493
1494 if (!capable(CAP_SYS_ADMIN))
1495 return -EPERM;
1496
1497 ret = mnt_want_write_file(file);
1498 if (ret)
1499 return ret;
1500
1501 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1502 1)) {
1503 mnt_drop_write_file(file);
1504 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1505 }
1506
1507 mutex_lock(&root->fs_info->volume_mutex);
1508 vol_args = memdup_user(arg, sizeof(*vol_args));
1509 if (IS_ERR(vol_args)) {
1510 ret = PTR_ERR(vol_args);
1511 goto out;
1512 }
1513
1514 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1515
1516 sizestr = vol_args->name;
1517 devstr = strchr(sizestr, ':');
1518 if (devstr) {
1519 sizestr = devstr + 1;
1520 *devstr = '\0';
1521 devstr = vol_args->name;
1522 ret = kstrtoull(devstr, 10, &devid);
1523 if (ret)
1524 goto out_free;
1525 if (!devid) {
1526 ret = -EINVAL;
1527 goto out_free;
1528 }
1529 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1530 }
1531
1532 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1533 if (!device) {
1534 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1535 devid);
1536 ret = -ENODEV;
1537 goto out_free;
1538 }
1539
1540 if (!device->writeable) {
1541 btrfs_info(root->fs_info,
1542 "resizer unable to apply on readonly device %llu",
1543 devid);
1544 ret = -EPERM;
1545 goto out_free;
1546 }
1547
1548 if (!strcmp(sizestr, "max"))
1549 new_size = device->bdev->bd_inode->i_size;
1550 else {
1551 if (sizestr[0] == '-') {
1552 mod = -1;
1553 sizestr++;
1554 } else if (sizestr[0] == '+') {
1555 mod = 1;
1556 sizestr++;
1557 }
1558 new_size = memparse(sizestr, &retptr);
1559 if (*retptr != '\0' || new_size == 0) {
1560 ret = -EINVAL;
1561 goto out_free;
1562 }
1563 }
1564
1565 if (device->is_tgtdev_for_dev_replace) {
1566 ret = -EPERM;
1567 goto out_free;
1568 }
1569
1570 old_size = btrfs_device_get_total_bytes(device);
1571
1572 if (mod < 0) {
1573 if (new_size > old_size) {
1574 ret = -EINVAL;
1575 goto out_free;
1576 }
1577 new_size = old_size - new_size;
1578 } else if (mod > 0) {
1579 if (new_size > ULLONG_MAX - old_size) {
1580 ret = -ERANGE;
1581 goto out_free;
1582 }
1583 new_size = old_size + new_size;
1584 }
1585
1586 if (new_size < SZ_256M) {
1587 ret = -EINVAL;
1588 goto out_free;
1589 }
1590 if (new_size > device->bdev->bd_inode->i_size) {
1591 ret = -EFBIG;
1592 goto out_free;
1593 }
1594
1595 new_size = div_u64(new_size, root->sectorsize);
1596 new_size *= root->sectorsize;
1597
1598 btrfs_info_in_rcu(root->fs_info, "new size for %s is %llu",
1599 rcu_str_deref(device->name), new_size);
1600
1601 if (new_size > old_size) {
1602 trans = btrfs_start_transaction(root, 0);
1603 if (IS_ERR(trans)) {
1604 ret = PTR_ERR(trans);
1605 goto out_free;
1606 }
1607 ret = btrfs_grow_device(trans, device, new_size);
1608 btrfs_commit_transaction(trans, root);
1609 } else if (new_size < old_size) {
1610 ret = btrfs_shrink_device(device, new_size);
1611 } /* equal, nothing need to do */
1612
1613 out_free:
1614 kfree(vol_args);
1615 out:
1616 mutex_unlock(&root->fs_info->volume_mutex);
1617 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1618 mnt_drop_write_file(file);
1619 return ret;
1620 }
1621
1622 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1623 char *name, unsigned long fd, int subvol,
1624 u64 *transid, bool readonly,
1625 struct btrfs_qgroup_inherit *inherit)
1626 {
1627 int namelen;
1628 int ret = 0;
1629
1630 ret = mnt_want_write_file(file);
1631 if (ret)
1632 goto out;
1633
1634 namelen = strlen(name);
1635 if (strchr(name, '/')) {
1636 ret = -EINVAL;
1637 goto out_drop_write;
1638 }
1639
1640 if (name[0] == '.' &&
1641 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1642 ret = -EEXIST;
1643 goto out_drop_write;
1644 }
1645
1646 if (subvol) {
1647 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1648 NULL, transid, readonly, inherit);
1649 } else {
1650 struct fd src = fdget(fd);
1651 struct inode *src_inode;
1652 if (!src.file) {
1653 ret = -EINVAL;
1654 goto out_drop_write;
1655 }
1656
1657 src_inode = file_inode(src.file);
1658 if (src_inode->i_sb != file_inode(file)->i_sb) {
1659 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1660 "Snapshot src from another FS");
1661 ret = -EXDEV;
1662 } else if (!inode_owner_or_capable(src_inode)) {
1663 /*
1664 * Subvolume creation is not restricted, but snapshots
1665 * are limited to own subvolumes only
1666 */
1667 ret = -EPERM;
1668 } else {
1669 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1670 BTRFS_I(src_inode)->root,
1671 transid, readonly, inherit);
1672 }
1673 fdput(src);
1674 }
1675 out_drop_write:
1676 mnt_drop_write_file(file);
1677 out:
1678 return ret;
1679 }
1680
1681 static noinline int btrfs_ioctl_snap_create(struct file *file,
1682 void __user *arg, int subvol)
1683 {
1684 struct btrfs_ioctl_vol_args *vol_args;
1685 int ret;
1686
1687 vol_args = memdup_user(arg, sizeof(*vol_args));
1688 if (IS_ERR(vol_args))
1689 return PTR_ERR(vol_args);
1690 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1691
1692 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1693 vol_args->fd, subvol,
1694 NULL, false, NULL);
1695
1696 kfree(vol_args);
1697 return ret;
1698 }
1699
1700 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1701 void __user *arg, int subvol)
1702 {
1703 struct btrfs_ioctl_vol_args_v2 *vol_args;
1704 int ret;
1705 u64 transid = 0;
1706 u64 *ptr = NULL;
1707 bool readonly = false;
1708 struct btrfs_qgroup_inherit *inherit = NULL;
1709
1710 vol_args = memdup_user(arg, sizeof(*vol_args));
1711 if (IS_ERR(vol_args))
1712 return PTR_ERR(vol_args);
1713 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1714
1715 if (vol_args->flags &
1716 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1717 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1718 ret = -EOPNOTSUPP;
1719 goto free_args;
1720 }
1721
1722 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1723 ptr = &transid;
1724 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1725 readonly = true;
1726 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1727 if (vol_args->size > PAGE_CACHE_SIZE) {
1728 ret = -EINVAL;
1729 goto free_args;
1730 }
1731 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1732 if (IS_ERR(inherit)) {
1733 ret = PTR_ERR(inherit);
1734 goto free_args;
1735 }
1736 }
1737
1738 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1739 vol_args->fd, subvol, ptr,
1740 readonly, inherit);
1741 if (ret)
1742 goto free_inherit;
1743
1744 if (ptr && copy_to_user(arg +
1745 offsetof(struct btrfs_ioctl_vol_args_v2,
1746 transid),
1747 ptr, sizeof(*ptr)))
1748 ret = -EFAULT;
1749
1750 free_inherit:
1751 kfree(inherit);
1752 free_args:
1753 kfree(vol_args);
1754 return ret;
1755 }
1756
1757 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1758 void __user *arg)
1759 {
1760 struct inode *inode = file_inode(file);
1761 struct btrfs_root *root = BTRFS_I(inode)->root;
1762 int ret = 0;
1763 u64 flags = 0;
1764
1765 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1766 return -EINVAL;
1767
1768 down_read(&root->fs_info->subvol_sem);
1769 if (btrfs_root_readonly(root))
1770 flags |= BTRFS_SUBVOL_RDONLY;
1771 up_read(&root->fs_info->subvol_sem);
1772
1773 if (copy_to_user(arg, &flags, sizeof(flags)))
1774 ret = -EFAULT;
1775
1776 return ret;
1777 }
1778
1779 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1780 void __user *arg)
1781 {
1782 struct inode *inode = file_inode(file);
1783 struct btrfs_root *root = BTRFS_I(inode)->root;
1784 struct btrfs_trans_handle *trans;
1785 u64 root_flags;
1786 u64 flags;
1787 int ret = 0;
1788
1789 if (!inode_owner_or_capable(inode))
1790 return -EPERM;
1791
1792 ret = mnt_want_write_file(file);
1793 if (ret)
1794 goto out;
1795
1796 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1797 ret = -EINVAL;
1798 goto out_drop_write;
1799 }
1800
1801 if (copy_from_user(&flags, arg, sizeof(flags))) {
1802 ret = -EFAULT;
1803 goto out_drop_write;
1804 }
1805
1806 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1807 ret = -EINVAL;
1808 goto out_drop_write;
1809 }
1810
1811 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1812 ret = -EOPNOTSUPP;
1813 goto out_drop_write;
1814 }
1815
1816 down_write(&root->fs_info->subvol_sem);
1817
1818 /* nothing to do */
1819 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1820 goto out_drop_sem;
1821
1822 root_flags = btrfs_root_flags(&root->root_item);
1823 if (flags & BTRFS_SUBVOL_RDONLY) {
1824 btrfs_set_root_flags(&root->root_item,
1825 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1826 } else {
1827 /*
1828 * Block RO -> RW transition if this subvolume is involved in
1829 * send
1830 */
1831 spin_lock(&root->root_item_lock);
1832 if (root->send_in_progress == 0) {
1833 btrfs_set_root_flags(&root->root_item,
1834 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1835 spin_unlock(&root->root_item_lock);
1836 } else {
1837 spin_unlock(&root->root_item_lock);
1838 btrfs_warn(root->fs_info,
1839 "Attempt to set subvolume %llu read-write during send",
1840 root->root_key.objectid);
1841 ret = -EPERM;
1842 goto out_drop_sem;
1843 }
1844 }
1845
1846 trans = btrfs_start_transaction(root, 1);
1847 if (IS_ERR(trans)) {
1848 ret = PTR_ERR(trans);
1849 goto out_reset;
1850 }
1851
1852 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1853 &root->root_key, &root->root_item);
1854
1855 btrfs_commit_transaction(trans, root);
1856 out_reset:
1857 if (ret)
1858 btrfs_set_root_flags(&root->root_item, root_flags);
1859 out_drop_sem:
1860 up_write(&root->fs_info->subvol_sem);
1861 out_drop_write:
1862 mnt_drop_write_file(file);
1863 out:
1864 return ret;
1865 }
1866
1867 /*
1868 * helper to check if the subvolume references other subvolumes
1869 */
1870 static noinline int may_destroy_subvol(struct btrfs_root *root)
1871 {
1872 struct btrfs_path *path;
1873 struct btrfs_dir_item *di;
1874 struct btrfs_key key;
1875 u64 dir_id;
1876 int ret;
1877
1878 path = btrfs_alloc_path();
1879 if (!path)
1880 return -ENOMEM;
1881
1882 /* Make sure this root isn't set as the default subvol */
1883 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1884 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1885 dir_id, "default", 7, 0);
1886 if (di && !IS_ERR(di)) {
1887 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1888 if (key.objectid == root->root_key.objectid) {
1889 ret = -EPERM;
1890 btrfs_err(root->fs_info, "deleting default subvolume "
1891 "%llu is not allowed", key.objectid);
1892 goto out;
1893 }
1894 btrfs_release_path(path);
1895 }
1896
1897 key.objectid = root->root_key.objectid;
1898 key.type = BTRFS_ROOT_REF_KEY;
1899 key.offset = (u64)-1;
1900
1901 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1902 &key, path, 0, 0);
1903 if (ret < 0)
1904 goto out;
1905 BUG_ON(ret == 0);
1906
1907 ret = 0;
1908 if (path->slots[0] > 0) {
1909 path->slots[0]--;
1910 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1911 if (key.objectid == root->root_key.objectid &&
1912 key.type == BTRFS_ROOT_REF_KEY)
1913 ret = -ENOTEMPTY;
1914 }
1915 out:
1916 btrfs_free_path(path);
1917 return ret;
1918 }
1919
1920 static noinline int key_in_sk(struct btrfs_key *key,
1921 struct btrfs_ioctl_search_key *sk)
1922 {
1923 struct btrfs_key test;
1924 int ret;
1925
1926 test.objectid = sk->min_objectid;
1927 test.type = sk->min_type;
1928 test.offset = sk->min_offset;
1929
1930 ret = btrfs_comp_cpu_keys(key, &test);
1931 if (ret < 0)
1932 return 0;
1933
1934 test.objectid = sk->max_objectid;
1935 test.type = sk->max_type;
1936 test.offset = sk->max_offset;
1937
1938 ret = btrfs_comp_cpu_keys(key, &test);
1939 if (ret > 0)
1940 return 0;
1941 return 1;
1942 }
1943
1944 static noinline int copy_to_sk(struct btrfs_root *root,
1945 struct btrfs_path *path,
1946 struct btrfs_key *key,
1947 struct btrfs_ioctl_search_key *sk,
1948 size_t *buf_size,
1949 char __user *ubuf,
1950 unsigned long *sk_offset,
1951 int *num_found)
1952 {
1953 u64 found_transid;
1954 struct extent_buffer *leaf;
1955 struct btrfs_ioctl_search_header sh;
1956 struct btrfs_key test;
1957 unsigned long item_off;
1958 unsigned long item_len;
1959 int nritems;
1960 int i;
1961 int slot;
1962 int ret = 0;
1963
1964 leaf = path->nodes[0];
1965 slot = path->slots[0];
1966 nritems = btrfs_header_nritems(leaf);
1967
1968 if (btrfs_header_generation(leaf) > sk->max_transid) {
1969 i = nritems;
1970 goto advance_key;
1971 }
1972 found_transid = btrfs_header_generation(leaf);
1973
1974 for (i = slot; i < nritems; i++) {
1975 item_off = btrfs_item_ptr_offset(leaf, i);
1976 item_len = btrfs_item_size_nr(leaf, i);
1977
1978 btrfs_item_key_to_cpu(leaf, key, i);
1979 if (!key_in_sk(key, sk))
1980 continue;
1981
1982 if (sizeof(sh) + item_len > *buf_size) {
1983 if (*num_found) {
1984 ret = 1;
1985 goto out;
1986 }
1987
1988 /*
1989 * return one empty item back for v1, which does not
1990 * handle -EOVERFLOW
1991 */
1992
1993 *buf_size = sizeof(sh) + item_len;
1994 item_len = 0;
1995 ret = -EOVERFLOW;
1996 }
1997
1998 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1999 ret = 1;
2000 goto out;
2001 }
2002
2003 sh.objectid = key->objectid;
2004 sh.offset = key->offset;
2005 sh.type = key->type;
2006 sh.len = item_len;
2007 sh.transid = found_transid;
2008
2009 /* copy search result header */
2010 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2011 ret = -EFAULT;
2012 goto out;
2013 }
2014
2015 *sk_offset += sizeof(sh);
2016
2017 if (item_len) {
2018 char __user *up = ubuf + *sk_offset;
2019 /* copy the item */
2020 if (read_extent_buffer_to_user(leaf, up,
2021 item_off, item_len)) {
2022 ret = -EFAULT;
2023 goto out;
2024 }
2025
2026 *sk_offset += item_len;
2027 }
2028 (*num_found)++;
2029
2030 if (ret) /* -EOVERFLOW from above */
2031 goto out;
2032
2033 if (*num_found >= sk->nr_items) {
2034 ret = 1;
2035 goto out;
2036 }
2037 }
2038 advance_key:
2039 ret = 0;
2040 test.objectid = sk->max_objectid;
2041 test.type = sk->max_type;
2042 test.offset = sk->max_offset;
2043 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2044 ret = 1;
2045 else if (key->offset < (u64)-1)
2046 key->offset++;
2047 else if (key->type < (u8)-1) {
2048 key->offset = 0;
2049 key->type++;
2050 } else if (key->objectid < (u64)-1) {
2051 key->offset = 0;
2052 key->type = 0;
2053 key->objectid++;
2054 } else
2055 ret = 1;
2056 out:
2057 /*
2058 * 0: all items from this leaf copied, continue with next
2059 * 1: * more items can be copied, but unused buffer is too small
2060 * * all items were found
2061 * Either way, it will stops the loop which iterates to the next
2062 * leaf
2063 * -EOVERFLOW: item was to large for buffer
2064 * -EFAULT: could not copy extent buffer back to userspace
2065 */
2066 return ret;
2067 }
2068
2069 static noinline int search_ioctl(struct inode *inode,
2070 struct btrfs_ioctl_search_key *sk,
2071 size_t *buf_size,
2072 char __user *ubuf)
2073 {
2074 struct btrfs_root *root;
2075 struct btrfs_key key;
2076 struct btrfs_path *path;
2077 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2078 int ret;
2079 int num_found = 0;
2080 unsigned long sk_offset = 0;
2081
2082 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2083 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2084 return -EOVERFLOW;
2085 }
2086
2087 path = btrfs_alloc_path();
2088 if (!path)
2089 return -ENOMEM;
2090
2091 if (sk->tree_id == 0) {
2092 /* search the root of the inode that was passed */
2093 root = BTRFS_I(inode)->root;
2094 } else {
2095 key.objectid = sk->tree_id;
2096 key.type = BTRFS_ROOT_ITEM_KEY;
2097 key.offset = (u64)-1;
2098 root = btrfs_read_fs_root_no_name(info, &key);
2099 if (IS_ERR(root)) {
2100 btrfs_err(info, "could not find root %llu",
2101 sk->tree_id);
2102 btrfs_free_path(path);
2103 return -ENOENT;
2104 }
2105 }
2106
2107 key.objectid = sk->min_objectid;
2108 key.type = sk->min_type;
2109 key.offset = sk->min_offset;
2110
2111 while (1) {
2112 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2113 if (ret != 0) {
2114 if (ret > 0)
2115 ret = 0;
2116 goto err;
2117 }
2118 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2119 &sk_offset, &num_found);
2120 btrfs_release_path(path);
2121 if (ret)
2122 break;
2123
2124 }
2125 if (ret > 0)
2126 ret = 0;
2127 err:
2128 sk->nr_items = num_found;
2129 btrfs_free_path(path);
2130 return ret;
2131 }
2132
2133 static noinline int btrfs_ioctl_tree_search(struct file *file,
2134 void __user *argp)
2135 {
2136 struct btrfs_ioctl_search_args __user *uargs;
2137 struct btrfs_ioctl_search_key sk;
2138 struct inode *inode;
2139 int ret;
2140 size_t buf_size;
2141
2142 if (!capable(CAP_SYS_ADMIN))
2143 return -EPERM;
2144
2145 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2146
2147 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2148 return -EFAULT;
2149
2150 buf_size = sizeof(uargs->buf);
2151
2152 inode = file_inode(file);
2153 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2154
2155 /*
2156 * In the origin implementation an overflow is handled by returning a
2157 * search header with a len of zero, so reset ret.
2158 */
2159 if (ret == -EOVERFLOW)
2160 ret = 0;
2161
2162 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2163 ret = -EFAULT;
2164 return ret;
2165 }
2166
2167 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2168 void __user *argp)
2169 {
2170 struct btrfs_ioctl_search_args_v2 __user *uarg;
2171 struct btrfs_ioctl_search_args_v2 args;
2172 struct inode *inode;
2173 int ret;
2174 size_t buf_size;
2175 const size_t buf_limit = SZ_16M;
2176
2177 if (!capable(CAP_SYS_ADMIN))
2178 return -EPERM;
2179
2180 /* copy search header and buffer size */
2181 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2182 if (copy_from_user(&args, uarg, sizeof(args)))
2183 return -EFAULT;
2184
2185 buf_size = args.buf_size;
2186
2187 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2188 return -EOVERFLOW;
2189
2190 /* limit result size to 16MB */
2191 if (buf_size > buf_limit)
2192 buf_size = buf_limit;
2193
2194 inode = file_inode(file);
2195 ret = search_ioctl(inode, &args.key, &buf_size,
2196 (char *)(&uarg->buf[0]));
2197 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2198 ret = -EFAULT;
2199 else if (ret == -EOVERFLOW &&
2200 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2201 ret = -EFAULT;
2202
2203 return ret;
2204 }
2205
2206 /*
2207 * Search INODE_REFs to identify path name of 'dirid' directory
2208 * in a 'tree_id' tree. and sets path name to 'name'.
2209 */
2210 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2211 u64 tree_id, u64 dirid, char *name)
2212 {
2213 struct btrfs_root *root;
2214 struct btrfs_key key;
2215 char *ptr;
2216 int ret = -1;
2217 int slot;
2218 int len;
2219 int total_len = 0;
2220 struct btrfs_inode_ref *iref;
2221 struct extent_buffer *l;
2222 struct btrfs_path *path;
2223
2224 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2225 name[0]='\0';
2226 return 0;
2227 }
2228
2229 path = btrfs_alloc_path();
2230 if (!path)
2231 return -ENOMEM;
2232
2233 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2234
2235 key.objectid = tree_id;
2236 key.type = BTRFS_ROOT_ITEM_KEY;
2237 key.offset = (u64)-1;
2238 root = btrfs_read_fs_root_no_name(info, &key);
2239 if (IS_ERR(root)) {
2240 btrfs_err(info, "could not find root %llu", tree_id);
2241 ret = -ENOENT;
2242 goto out;
2243 }
2244
2245 key.objectid = dirid;
2246 key.type = BTRFS_INODE_REF_KEY;
2247 key.offset = (u64)-1;
2248
2249 while (1) {
2250 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2251 if (ret < 0)
2252 goto out;
2253 else if (ret > 0) {
2254 ret = btrfs_previous_item(root, path, dirid,
2255 BTRFS_INODE_REF_KEY);
2256 if (ret < 0)
2257 goto out;
2258 else if (ret > 0) {
2259 ret = -ENOENT;
2260 goto out;
2261 }
2262 }
2263
2264 l = path->nodes[0];
2265 slot = path->slots[0];
2266 btrfs_item_key_to_cpu(l, &key, slot);
2267
2268 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2269 len = btrfs_inode_ref_name_len(l, iref);
2270 ptr -= len + 1;
2271 total_len += len + 1;
2272 if (ptr < name) {
2273 ret = -ENAMETOOLONG;
2274 goto out;
2275 }
2276
2277 *(ptr + len) = '/';
2278 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2279
2280 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2281 break;
2282
2283 btrfs_release_path(path);
2284 key.objectid = key.offset;
2285 key.offset = (u64)-1;
2286 dirid = key.objectid;
2287 }
2288 memmove(name, ptr, total_len);
2289 name[total_len] = '\0';
2290 ret = 0;
2291 out:
2292 btrfs_free_path(path);
2293 return ret;
2294 }
2295
2296 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2297 void __user *argp)
2298 {
2299 struct btrfs_ioctl_ino_lookup_args *args;
2300 struct inode *inode;
2301 int ret = 0;
2302
2303 args = memdup_user(argp, sizeof(*args));
2304 if (IS_ERR(args))
2305 return PTR_ERR(args);
2306
2307 inode = file_inode(file);
2308
2309 /*
2310 * Unprivileged query to obtain the containing subvolume root id. The
2311 * path is reset so it's consistent with btrfs_search_path_in_tree.
2312 */
2313 if (args->treeid == 0)
2314 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2315
2316 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2317 args->name[0] = 0;
2318 goto out;
2319 }
2320
2321 if (!capable(CAP_SYS_ADMIN)) {
2322 ret = -EPERM;
2323 goto out;
2324 }
2325
2326 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2327 args->treeid, args->objectid,
2328 args->name);
2329
2330 out:
2331 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2332 ret = -EFAULT;
2333
2334 kfree(args);
2335 return ret;
2336 }
2337
2338 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2339 void __user *arg)
2340 {
2341 struct dentry *parent = file->f_path.dentry;
2342 struct dentry *dentry;
2343 struct inode *dir = d_inode(parent);
2344 struct inode *inode;
2345 struct btrfs_root *root = BTRFS_I(dir)->root;
2346 struct btrfs_root *dest = NULL;
2347 struct btrfs_ioctl_vol_args *vol_args;
2348 struct btrfs_trans_handle *trans;
2349 struct btrfs_block_rsv block_rsv;
2350 u64 root_flags;
2351 u64 qgroup_reserved;
2352 int namelen;
2353 int ret;
2354 int err = 0;
2355
2356 vol_args = memdup_user(arg, sizeof(*vol_args));
2357 if (IS_ERR(vol_args))
2358 return PTR_ERR(vol_args);
2359
2360 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2361 namelen = strlen(vol_args->name);
2362 if (strchr(vol_args->name, '/') ||
2363 strncmp(vol_args->name, "..", namelen) == 0) {
2364 err = -EINVAL;
2365 goto out;
2366 }
2367
2368 err = mnt_want_write_file(file);
2369 if (err)
2370 goto out;
2371
2372
2373 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2374 if (err == -EINTR)
2375 goto out_drop_write;
2376 dentry = lookup_one_len(vol_args->name, parent, namelen);
2377 if (IS_ERR(dentry)) {
2378 err = PTR_ERR(dentry);
2379 goto out_unlock_dir;
2380 }
2381
2382 if (d_really_is_negative(dentry)) {
2383 err = -ENOENT;
2384 goto out_dput;
2385 }
2386
2387 inode = d_inode(dentry);
2388 dest = BTRFS_I(inode)->root;
2389 if (!capable(CAP_SYS_ADMIN)) {
2390 /*
2391 * Regular user. Only allow this with a special mount
2392 * option, when the user has write+exec access to the
2393 * subvol root, and when rmdir(2) would have been
2394 * allowed.
2395 *
2396 * Note that this is _not_ check that the subvol is
2397 * empty or doesn't contain data that we wouldn't
2398 * otherwise be able to delete.
2399 *
2400 * Users who want to delete empty subvols should try
2401 * rmdir(2).
2402 */
2403 err = -EPERM;
2404 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2405 goto out_dput;
2406
2407 /*
2408 * Do not allow deletion if the parent dir is the same
2409 * as the dir to be deleted. That means the ioctl
2410 * must be called on the dentry referencing the root
2411 * of the subvol, not a random directory contained
2412 * within it.
2413 */
2414 err = -EINVAL;
2415 if (root == dest)
2416 goto out_dput;
2417
2418 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2419 if (err)
2420 goto out_dput;
2421 }
2422
2423 /* check if subvolume may be deleted by a user */
2424 err = btrfs_may_delete(dir, dentry, 1);
2425 if (err)
2426 goto out_dput;
2427
2428 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2429 err = -EINVAL;
2430 goto out_dput;
2431 }
2432
2433 inode_lock(inode);
2434
2435 /*
2436 * Don't allow to delete a subvolume with send in progress. This is
2437 * inside the i_mutex so the error handling that has to drop the bit
2438 * again is not run concurrently.
2439 */
2440 spin_lock(&dest->root_item_lock);
2441 root_flags = btrfs_root_flags(&dest->root_item);
2442 if (dest->send_in_progress == 0) {
2443 btrfs_set_root_flags(&dest->root_item,
2444 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2445 spin_unlock(&dest->root_item_lock);
2446 } else {
2447 spin_unlock(&dest->root_item_lock);
2448 btrfs_warn(root->fs_info,
2449 "Attempt to delete subvolume %llu during send",
2450 dest->root_key.objectid);
2451 err = -EPERM;
2452 goto out_unlock_inode;
2453 }
2454
2455 down_write(&root->fs_info->subvol_sem);
2456
2457 err = may_destroy_subvol(dest);
2458 if (err)
2459 goto out_up_write;
2460
2461 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2462 /*
2463 * One for dir inode, two for dir entries, two for root
2464 * ref/backref.
2465 */
2466 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2467 5, &qgroup_reserved, true);
2468 if (err)
2469 goto out_up_write;
2470
2471 trans = btrfs_start_transaction(root, 0);
2472 if (IS_ERR(trans)) {
2473 err = PTR_ERR(trans);
2474 goto out_release;
2475 }
2476 trans->block_rsv = &block_rsv;
2477 trans->bytes_reserved = block_rsv.size;
2478
2479 ret = btrfs_unlink_subvol(trans, root, dir,
2480 dest->root_key.objectid,
2481 dentry->d_name.name,
2482 dentry->d_name.len);
2483 if (ret) {
2484 err = ret;
2485 btrfs_abort_transaction(trans, root, ret);
2486 goto out_end_trans;
2487 }
2488
2489 btrfs_record_root_in_trans(trans, dest);
2490
2491 memset(&dest->root_item.drop_progress, 0,
2492 sizeof(dest->root_item.drop_progress));
2493 dest->root_item.drop_level = 0;
2494 btrfs_set_root_refs(&dest->root_item, 0);
2495
2496 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2497 ret = btrfs_insert_orphan_item(trans,
2498 root->fs_info->tree_root,
2499 dest->root_key.objectid);
2500 if (ret) {
2501 btrfs_abort_transaction(trans, root, ret);
2502 err = ret;
2503 goto out_end_trans;
2504 }
2505 }
2506
2507 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2508 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2509 dest->root_key.objectid);
2510 if (ret && ret != -ENOENT) {
2511 btrfs_abort_transaction(trans, root, ret);
2512 err = ret;
2513 goto out_end_trans;
2514 }
2515 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2516 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2517 dest->root_item.received_uuid,
2518 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2519 dest->root_key.objectid);
2520 if (ret && ret != -ENOENT) {
2521 btrfs_abort_transaction(trans, root, ret);
2522 err = ret;
2523 goto out_end_trans;
2524 }
2525 }
2526
2527 out_end_trans:
2528 trans->block_rsv = NULL;
2529 trans->bytes_reserved = 0;
2530 ret = btrfs_end_transaction(trans, root);
2531 if (ret && !err)
2532 err = ret;
2533 inode->i_flags |= S_DEAD;
2534 out_release:
2535 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2536 out_up_write:
2537 up_write(&root->fs_info->subvol_sem);
2538 if (err) {
2539 spin_lock(&dest->root_item_lock);
2540 root_flags = btrfs_root_flags(&dest->root_item);
2541 btrfs_set_root_flags(&dest->root_item,
2542 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2543 spin_unlock(&dest->root_item_lock);
2544 }
2545 out_unlock_inode:
2546 inode_unlock(inode);
2547 if (!err) {
2548 d_invalidate(dentry);
2549 btrfs_invalidate_inodes(dest);
2550 d_delete(dentry);
2551 ASSERT(dest->send_in_progress == 0);
2552
2553 /* the last ref */
2554 if (dest->ino_cache_inode) {
2555 iput(dest->ino_cache_inode);
2556 dest->ino_cache_inode = NULL;
2557 }
2558 }
2559 out_dput:
2560 dput(dentry);
2561 out_unlock_dir:
2562 inode_unlock(dir);
2563 out_drop_write:
2564 mnt_drop_write_file(file);
2565 out:
2566 kfree(vol_args);
2567 return err;
2568 }
2569
2570 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2571 {
2572 struct inode *inode = file_inode(file);
2573 struct btrfs_root *root = BTRFS_I(inode)->root;
2574 struct btrfs_ioctl_defrag_range_args *range;
2575 int ret;
2576
2577 ret = mnt_want_write_file(file);
2578 if (ret)
2579 return ret;
2580
2581 if (btrfs_root_readonly(root)) {
2582 ret = -EROFS;
2583 goto out;
2584 }
2585
2586 switch (inode->i_mode & S_IFMT) {
2587 case S_IFDIR:
2588 if (!capable(CAP_SYS_ADMIN)) {
2589 ret = -EPERM;
2590 goto out;
2591 }
2592 ret = btrfs_defrag_root(root);
2593 if (ret)
2594 goto out;
2595 ret = btrfs_defrag_root(root->fs_info->extent_root);
2596 break;
2597 case S_IFREG:
2598 if (!(file->f_mode & FMODE_WRITE)) {
2599 ret = -EINVAL;
2600 goto out;
2601 }
2602
2603 range = kzalloc(sizeof(*range), GFP_KERNEL);
2604 if (!range) {
2605 ret = -ENOMEM;
2606 goto out;
2607 }
2608
2609 if (argp) {
2610 if (copy_from_user(range, argp,
2611 sizeof(*range))) {
2612 ret = -EFAULT;
2613 kfree(range);
2614 goto out;
2615 }
2616 /* compression requires us to start the IO */
2617 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2618 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2619 range->extent_thresh = (u32)-1;
2620 }
2621 } else {
2622 /* the rest are all set to zero by kzalloc */
2623 range->len = (u64)-1;
2624 }
2625 ret = btrfs_defrag_file(file_inode(file), file,
2626 range, 0, 0);
2627 if (ret > 0)
2628 ret = 0;
2629 kfree(range);
2630 break;
2631 default:
2632 ret = -EINVAL;
2633 }
2634 out:
2635 mnt_drop_write_file(file);
2636 return ret;
2637 }
2638
2639 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2640 {
2641 struct btrfs_ioctl_vol_args *vol_args;
2642 int ret;
2643
2644 if (!capable(CAP_SYS_ADMIN))
2645 return -EPERM;
2646
2647 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2648 1)) {
2649 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2650 }
2651
2652 mutex_lock(&root->fs_info->volume_mutex);
2653 vol_args = memdup_user(arg, sizeof(*vol_args));
2654 if (IS_ERR(vol_args)) {
2655 ret = PTR_ERR(vol_args);
2656 goto out;
2657 }
2658
2659 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2660 ret = btrfs_init_new_device(root, vol_args->name);
2661
2662 if (!ret)
2663 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2664
2665 kfree(vol_args);
2666 out:
2667 mutex_unlock(&root->fs_info->volume_mutex);
2668 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2669 return ret;
2670 }
2671
2672 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2673 {
2674 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2675 struct btrfs_ioctl_vol_args *vol_args;
2676 int ret;
2677
2678 if (!capable(CAP_SYS_ADMIN))
2679 return -EPERM;
2680
2681 ret = mnt_want_write_file(file);
2682 if (ret)
2683 return ret;
2684
2685 vol_args = memdup_user(arg, sizeof(*vol_args));
2686 if (IS_ERR(vol_args)) {
2687 ret = PTR_ERR(vol_args);
2688 goto err_drop;
2689 }
2690
2691 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2692
2693 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2694 1)) {
2695 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2696 goto out;
2697 }
2698
2699 mutex_lock(&root->fs_info->volume_mutex);
2700 ret = btrfs_rm_device(root, vol_args->name);
2701 mutex_unlock(&root->fs_info->volume_mutex);
2702 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2703
2704 if (!ret)
2705 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2706
2707 out:
2708 kfree(vol_args);
2709 err_drop:
2710 mnt_drop_write_file(file);
2711 return ret;
2712 }
2713
2714 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2715 {
2716 struct btrfs_ioctl_fs_info_args *fi_args;
2717 struct btrfs_device *device;
2718 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2719 int ret = 0;
2720
2721 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2722 if (!fi_args)
2723 return -ENOMEM;
2724
2725 mutex_lock(&fs_devices->device_list_mutex);
2726 fi_args->num_devices = fs_devices->num_devices;
2727 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2728
2729 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2730 if (device->devid > fi_args->max_id)
2731 fi_args->max_id = device->devid;
2732 }
2733 mutex_unlock(&fs_devices->device_list_mutex);
2734
2735 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2736 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2737 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2738
2739 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2740 ret = -EFAULT;
2741
2742 kfree(fi_args);
2743 return ret;
2744 }
2745
2746 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2747 {
2748 struct btrfs_ioctl_dev_info_args *di_args;
2749 struct btrfs_device *dev;
2750 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2751 int ret = 0;
2752 char *s_uuid = NULL;
2753
2754 di_args = memdup_user(arg, sizeof(*di_args));
2755 if (IS_ERR(di_args))
2756 return PTR_ERR(di_args);
2757
2758 if (!btrfs_is_empty_uuid(di_args->uuid))
2759 s_uuid = di_args->uuid;
2760
2761 mutex_lock(&fs_devices->device_list_mutex);
2762 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2763
2764 if (!dev) {
2765 ret = -ENODEV;
2766 goto out;
2767 }
2768
2769 di_args->devid = dev->devid;
2770 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2771 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2772 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2773 if (dev->name) {
2774 struct rcu_string *name;
2775
2776 rcu_read_lock();
2777 name = rcu_dereference(dev->name);
2778 strncpy(di_args->path, name->str, sizeof(di_args->path));
2779 rcu_read_unlock();
2780 di_args->path[sizeof(di_args->path) - 1] = 0;
2781 } else {
2782 di_args->path[0] = '\0';
2783 }
2784
2785 out:
2786 mutex_unlock(&fs_devices->device_list_mutex);
2787 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2788 ret = -EFAULT;
2789
2790 kfree(di_args);
2791 return ret;
2792 }
2793
2794 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2795 {
2796 struct page *page;
2797
2798 page = grab_cache_page(inode->i_mapping, index);
2799 if (!page)
2800 return ERR_PTR(-ENOMEM);
2801
2802 if (!PageUptodate(page)) {
2803 int ret;
2804
2805 ret = btrfs_readpage(NULL, page);
2806 if (ret)
2807 return ERR_PTR(ret);
2808 lock_page(page);
2809 if (!PageUptodate(page)) {
2810 unlock_page(page);
2811 page_cache_release(page);
2812 return ERR_PTR(-EIO);
2813 }
2814 if (page->mapping != inode->i_mapping) {
2815 unlock_page(page);
2816 page_cache_release(page);
2817 return ERR_PTR(-EAGAIN);
2818 }
2819 }
2820
2821 return page;
2822 }
2823
2824 static int gather_extent_pages(struct inode *inode, struct page **pages,
2825 int num_pages, u64 off)
2826 {
2827 int i;
2828 pgoff_t index = off >> PAGE_CACHE_SHIFT;
2829
2830 for (i = 0; i < num_pages; i++) {
2831 again:
2832 pages[i] = extent_same_get_page(inode, index + i);
2833 if (IS_ERR(pages[i])) {
2834 int err = PTR_ERR(pages[i]);
2835
2836 if (err == -EAGAIN)
2837 goto again;
2838 pages[i] = NULL;
2839 return err;
2840 }
2841 }
2842 return 0;
2843 }
2844
2845 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2846 bool retry_range_locking)
2847 {
2848 /*
2849 * Do any pending delalloc/csum calculations on inode, one way or
2850 * another, and lock file content.
2851 * The locking order is:
2852 *
2853 * 1) pages
2854 * 2) range in the inode's io tree
2855 */
2856 while (1) {
2857 struct btrfs_ordered_extent *ordered;
2858 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2859 ordered = btrfs_lookup_first_ordered_extent(inode,
2860 off + len - 1);
2861 if ((!ordered ||
2862 ordered->file_offset + ordered->len <= off ||
2863 ordered->file_offset >= off + len) &&
2864 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2865 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2866 if (ordered)
2867 btrfs_put_ordered_extent(ordered);
2868 break;
2869 }
2870 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2871 if (ordered)
2872 btrfs_put_ordered_extent(ordered);
2873 if (!retry_range_locking)
2874 return -EAGAIN;
2875 btrfs_wait_ordered_range(inode, off, len);
2876 }
2877 return 0;
2878 }
2879
2880 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2881 {
2882 inode_unlock(inode1);
2883 inode_unlock(inode2);
2884 }
2885
2886 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2887 {
2888 if (inode1 < inode2)
2889 swap(inode1, inode2);
2890
2891 inode_lock_nested(inode1, I_MUTEX_PARENT);
2892 inode_lock_nested(inode2, I_MUTEX_CHILD);
2893 }
2894
2895 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2896 struct inode *inode2, u64 loff2, u64 len)
2897 {
2898 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2899 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2900 }
2901
2902 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2903 struct inode *inode2, u64 loff2, u64 len,
2904 bool retry_range_locking)
2905 {
2906 int ret;
2907
2908 if (inode1 < inode2) {
2909 swap(inode1, inode2);
2910 swap(loff1, loff2);
2911 }
2912 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2913 if (ret)
2914 return ret;
2915 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2916 if (ret)
2917 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2918 loff1 + len - 1);
2919 return ret;
2920 }
2921
2922 struct cmp_pages {
2923 int num_pages;
2924 struct page **src_pages;
2925 struct page **dst_pages;
2926 };
2927
2928 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2929 {
2930 int i;
2931 struct page *pg;
2932
2933 for (i = 0; i < cmp->num_pages; i++) {
2934 pg = cmp->src_pages[i];
2935 if (pg) {
2936 unlock_page(pg);
2937 page_cache_release(pg);
2938 }
2939 pg = cmp->dst_pages[i];
2940 if (pg) {
2941 unlock_page(pg);
2942 page_cache_release(pg);
2943 }
2944 }
2945 kfree(cmp->src_pages);
2946 kfree(cmp->dst_pages);
2947 }
2948
2949 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2950 struct inode *dst, u64 dst_loff,
2951 u64 len, struct cmp_pages *cmp)
2952 {
2953 int ret;
2954 int num_pages = PAGE_CACHE_ALIGN(len) >> PAGE_CACHE_SHIFT;
2955 struct page **src_pgarr, **dst_pgarr;
2956
2957 /*
2958 * We must gather up all the pages before we initiate our
2959 * extent locking. We use an array for the page pointers. Size
2960 * of the array is bounded by len, which is in turn bounded by
2961 * BTRFS_MAX_DEDUPE_LEN.
2962 */
2963 src_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2964 dst_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2965 if (!src_pgarr || !dst_pgarr) {
2966 kfree(src_pgarr);
2967 kfree(dst_pgarr);
2968 return -ENOMEM;
2969 }
2970 cmp->num_pages = num_pages;
2971 cmp->src_pages = src_pgarr;
2972 cmp->dst_pages = dst_pgarr;
2973
2974 ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff);
2975 if (ret)
2976 goto out;
2977
2978 ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff);
2979
2980 out:
2981 if (ret)
2982 btrfs_cmp_data_free(cmp);
2983 return 0;
2984 }
2985
2986 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2987 u64 dst_loff, u64 len, struct cmp_pages *cmp)
2988 {
2989 int ret = 0;
2990 int i;
2991 struct page *src_page, *dst_page;
2992 unsigned int cmp_len = PAGE_CACHE_SIZE;
2993 void *addr, *dst_addr;
2994
2995 i = 0;
2996 while (len) {
2997 if (len < PAGE_CACHE_SIZE)
2998 cmp_len = len;
2999
3000 BUG_ON(i >= cmp->num_pages);
3001
3002 src_page = cmp->src_pages[i];
3003 dst_page = cmp->dst_pages[i];
3004 ASSERT(PageLocked(src_page));
3005 ASSERT(PageLocked(dst_page));
3006
3007 addr = kmap_atomic(src_page);
3008 dst_addr = kmap_atomic(dst_page);
3009
3010 flush_dcache_page(src_page);
3011 flush_dcache_page(dst_page);
3012
3013 if (memcmp(addr, dst_addr, cmp_len))
3014 ret = -EBADE;
3015
3016 kunmap_atomic(addr);
3017 kunmap_atomic(dst_addr);
3018
3019 if (ret)
3020 break;
3021
3022 len -= cmp_len;
3023 i++;
3024 }
3025
3026 return ret;
3027 }
3028
3029 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3030 u64 olen)
3031 {
3032 u64 len = *plen;
3033 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3034
3035 if (off + olen > inode->i_size || off + olen < off)
3036 return -EINVAL;
3037
3038 /* if we extend to eof, continue to block boundary */
3039 if (off + len == inode->i_size)
3040 *plen = len = ALIGN(inode->i_size, bs) - off;
3041
3042 /* Check that we are block aligned - btrfs_clone() requires this */
3043 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3044 return -EINVAL;
3045
3046 return 0;
3047 }
3048
3049 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3050 struct inode *dst, u64 dst_loff)
3051 {
3052 int ret;
3053 u64 len = olen;
3054 struct cmp_pages cmp;
3055 int same_inode = 0;
3056 u64 same_lock_start = 0;
3057 u64 same_lock_len = 0;
3058
3059 if (src == dst)
3060 same_inode = 1;
3061
3062 if (len == 0)
3063 return 0;
3064
3065 if (same_inode) {
3066 inode_lock(src);
3067
3068 ret = extent_same_check_offsets(src, loff, &len, olen);
3069 if (ret)
3070 goto out_unlock;
3071
3072 /*
3073 * Single inode case wants the same checks, except we
3074 * don't want our length pushed out past i_size as
3075 * comparing that data range makes no sense.
3076 *
3077 * extent_same_check_offsets() will do this for an
3078 * unaligned length at i_size, so catch it here and
3079 * reject the request.
3080 *
3081 * This effectively means we require aligned extents
3082 * for the single-inode case, whereas the other cases
3083 * allow an unaligned length so long as it ends at
3084 * i_size.
3085 */
3086 if (len != olen) {
3087 ret = -EINVAL;
3088 goto out_unlock;
3089 }
3090
3091 /* Check for overlapping ranges */
3092 if (dst_loff + len > loff && dst_loff < loff + len) {
3093 ret = -EINVAL;
3094 goto out_unlock;
3095 }
3096
3097 same_lock_start = min_t(u64, loff, dst_loff);
3098 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3099 } else {
3100 btrfs_double_inode_lock(src, dst);
3101
3102 ret = extent_same_check_offsets(src, loff, &len, olen);
3103 if (ret)
3104 goto out_unlock;
3105
3106 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3107 if (ret)
3108 goto out_unlock;
3109 }
3110
3111 /* don't make the dst file partly checksummed */
3112 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3113 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3114 ret = -EINVAL;
3115 goto out_unlock;
3116 }
3117
3118 again:
3119 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3120 if (ret)
3121 goto out_unlock;
3122
3123 if (same_inode)
3124 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3125 false);
3126 else
3127 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3128 false);
3129 /*
3130 * If one of the inodes has dirty pages in the respective range or
3131 * ordered extents, we need to flush dellaloc and wait for all ordered
3132 * extents in the range. We must unlock the pages and the ranges in the
3133 * io trees to avoid deadlocks when flushing delalloc (requires locking
3134 * pages) and when waiting for ordered extents to complete (they require
3135 * range locking).
3136 */
3137 if (ret == -EAGAIN) {
3138 /*
3139 * Ranges in the io trees already unlocked. Now unlock all
3140 * pages before waiting for all IO to complete.
3141 */
3142 btrfs_cmp_data_free(&cmp);
3143 if (same_inode) {
3144 btrfs_wait_ordered_range(src, same_lock_start,
3145 same_lock_len);
3146 } else {
3147 btrfs_wait_ordered_range(src, loff, len);
3148 btrfs_wait_ordered_range(dst, dst_loff, len);
3149 }
3150 goto again;
3151 }
3152 ASSERT(ret == 0);
3153 if (WARN_ON(ret)) {
3154 /* ranges in the io trees already unlocked */
3155 btrfs_cmp_data_free(&cmp);
3156 return ret;
3157 }
3158
3159 /* pass original length for comparison so we stay within i_size */
3160 ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp);
3161 if (ret == 0)
3162 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3163
3164 if (same_inode)
3165 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3166 same_lock_start + same_lock_len - 1);
3167 else
3168 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3169
3170 btrfs_cmp_data_free(&cmp);
3171 out_unlock:
3172 if (same_inode)
3173 inode_unlock(src);
3174 else
3175 btrfs_double_inode_unlock(src, dst);
3176
3177 return ret;
3178 }
3179
3180 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3181
3182 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3183 struct file *dst_file, u64 dst_loff)
3184 {
3185 struct inode *src = file_inode(src_file);
3186 struct inode *dst = file_inode(dst_file);
3187 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3188 ssize_t res;
3189
3190 if (olen > BTRFS_MAX_DEDUPE_LEN)
3191 olen = BTRFS_MAX_DEDUPE_LEN;
3192
3193 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
3194 /*
3195 * Btrfs does not support blocksize < page_size. As a
3196 * result, btrfs_cmp_data() won't correctly handle
3197 * this situation without an update.
3198 */
3199 return -EINVAL;
3200 }
3201
3202 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3203 if (res)
3204 return res;
3205 return olen;
3206 }
3207
3208 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3209 struct inode *inode,
3210 u64 endoff,
3211 const u64 destoff,
3212 const u64 olen,
3213 int no_time_update)
3214 {
3215 struct btrfs_root *root = BTRFS_I(inode)->root;
3216 int ret;
3217
3218 inode_inc_iversion(inode);
3219 if (!no_time_update)
3220 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3221 /*
3222 * We round up to the block size at eof when determining which
3223 * extents to clone above, but shouldn't round up the file size.
3224 */
3225 if (endoff > destoff + olen)
3226 endoff = destoff + olen;
3227 if (endoff > inode->i_size)
3228 btrfs_i_size_write(inode, endoff);
3229
3230 ret = btrfs_update_inode(trans, root, inode);
3231 if (ret) {
3232 btrfs_abort_transaction(trans, root, ret);
3233 btrfs_end_transaction(trans, root);
3234 goto out;
3235 }
3236 ret = btrfs_end_transaction(trans, root);
3237 out:
3238 return ret;
3239 }
3240
3241 static void clone_update_extent_map(struct inode *inode,
3242 const struct btrfs_trans_handle *trans,
3243 const struct btrfs_path *path,
3244 const u64 hole_offset,
3245 const u64 hole_len)
3246 {
3247 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3248 struct extent_map *em;
3249 int ret;
3250
3251 em = alloc_extent_map();
3252 if (!em) {
3253 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3254 &BTRFS_I(inode)->runtime_flags);
3255 return;
3256 }
3257
3258 if (path) {
3259 struct btrfs_file_extent_item *fi;
3260
3261 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3262 struct btrfs_file_extent_item);
3263 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3264 em->generation = -1;
3265 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3266 BTRFS_FILE_EXTENT_INLINE)
3267 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3268 &BTRFS_I(inode)->runtime_flags);
3269 } else {
3270 em->start = hole_offset;
3271 em->len = hole_len;
3272 em->ram_bytes = em->len;
3273 em->orig_start = hole_offset;
3274 em->block_start = EXTENT_MAP_HOLE;
3275 em->block_len = 0;
3276 em->orig_block_len = 0;
3277 em->compress_type = BTRFS_COMPRESS_NONE;
3278 em->generation = trans->transid;
3279 }
3280
3281 while (1) {
3282 write_lock(&em_tree->lock);
3283 ret = add_extent_mapping(em_tree, em, 1);
3284 write_unlock(&em_tree->lock);
3285 if (ret != -EEXIST) {
3286 free_extent_map(em);
3287 break;
3288 }
3289 btrfs_drop_extent_cache(inode, em->start,
3290 em->start + em->len - 1, 0);
3291 }
3292
3293 if (ret)
3294 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3295 &BTRFS_I(inode)->runtime_flags);
3296 }
3297
3298 /*
3299 * Make sure we do not end up inserting an inline extent into a file that has
3300 * already other (non-inline) extents. If a file has an inline extent it can
3301 * not have any other extents and the (single) inline extent must start at the
3302 * file offset 0. Failing to respect these rules will lead to file corruption,
3303 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3304 *
3305 * We can have extents that have been already written to disk or we can have
3306 * dirty ranges still in delalloc, in which case the extent maps and items are
3307 * created only when we run delalloc, and the delalloc ranges might fall outside
3308 * the range we are currently locking in the inode's io tree. So we check the
3309 * inode's i_size because of that (i_size updates are done while holding the
3310 * i_mutex, which we are holding here).
3311 * We also check to see if the inode has a size not greater than "datal" but has
3312 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3313 * protected against such concurrent fallocate calls by the i_mutex).
3314 *
3315 * If the file has no extents but a size greater than datal, do not allow the
3316 * copy because we would need turn the inline extent into a non-inline one (even
3317 * with NO_HOLES enabled). If we find our destination inode only has one inline
3318 * extent, just overwrite it with the source inline extent if its size is less
3319 * than the source extent's size, or we could copy the source inline extent's
3320 * data into the destination inode's inline extent if the later is greater then
3321 * the former.
3322 */
3323 static int clone_copy_inline_extent(struct inode *src,
3324 struct inode *dst,
3325 struct btrfs_trans_handle *trans,
3326 struct btrfs_path *path,
3327 struct btrfs_key *new_key,
3328 const u64 drop_start,
3329 const u64 datal,
3330 const u64 skip,
3331 const u64 size,
3332 char *inline_data)
3333 {
3334 struct btrfs_root *root = BTRFS_I(dst)->root;
3335 const u64 aligned_end = ALIGN(new_key->offset + datal,
3336 root->sectorsize);
3337 int ret;
3338 struct btrfs_key key;
3339
3340 if (new_key->offset > 0)
3341 return -EOPNOTSUPP;
3342
3343 key.objectid = btrfs_ino(dst);
3344 key.type = BTRFS_EXTENT_DATA_KEY;
3345 key.offset = 0;
3346 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3347 if (ret < 0) {
3348 return ret;
3349 } else if (ret > 0) {
3350 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3351 ret = btrfs_next_leaf(root, path);
3352 if (ret < 0)
3353 return ret;
3354 else if (ret > 0)
3355 goto copy_inline_extent;
3356 }
3357 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3358 if (key.objectid == btrfs_ino(dst) &&
3359 key.type == BTRFS_EXTENT_DATA_KEY) {
3360 ASSERT(key.offset > 0);
3361 return -EOPNOTSUPP;
3362 }
3363 } else if (i_size_read(dst) <= datal) {
3364 struct btrfs_file_extent_item *ei;
3365 u64 ext_len;
3366
3367 /*
3368 * If the file size is <= datal, make sure there are no other
3369 * extents following (can happen do to an fallocate call with
3370 * the flag FALLOC_FL_KEEP_SIZE).
3371 */
3372 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3373 struct btrfs_file_extent_item);
3374 /*
3375 * If it's an inline extent, it can not have other extents
3376 * following it.
3377 */
3378 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3379 BTRFS_FILE_EXTENT_INLINE)
3380 goto copy_inline_extent;
3381
3382 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3383 if (ext_len > aligned_end)
3384 return -EOPNOTSUPP;
3385
3386 ret = btrfs_next_item(root, path);
3387 if (ret < 0) {
3388 return ret;
3389 } else if (ret == 0) {
3390 btrfs_item_key_to_cpu(path->nodes[0], &key,
3391 path->slots[0]);
3392 if (key.objectid == btrfs_ino(dst) &&
3393 key.type == BTRFS_EXTENT_DATA_KEY)
3394 return -EOPNOTSUPP;
3395 }
3396 }
3397
3398 copy_inline_extent:
3399 /*
3400 * We have no extent items, or we have an extent at offset 0 which may
3401 * or may not be inlined. All these cases are dealt the same way.
3402 */
3403 if (i_size_read(dst) > datal) {
3404 /*
3405 * If the destination inode has an inline extent...
3406 * This would require copying the data from the source inline
3407 * extent into the beginning of the destination's inline extent.
3408 * But this is really complex, both extents can be compressed
3409 * or just one of them, which would require decompressing and
3410 * re-compressing data (which could increase the new compressed
3411 * size, not allowing the compressed data to fit anymore in an
3412 * inline extent).
3413 * So just don't support this case for now (it should be rare,
3414 * we are not really saving space when cloning inline extents).
3415 */
3416 return -EOPNOTSUPP;
3417 }
3418
3419 btrfs_release_path(path);
3420 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3421 if (ret)
3422 return ret;
3423 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3424 if (ret)
3425 return ret;
3426
3427 if (skip) {
3428 const u32 start = btrfs_file_extent_calc_inline_size(0);
3429
3430 memmove(inline_data + start, inline_data + start + skip, datal);
3431 }
3432
3433 write_extent_buffer(path->nodes[0], inline_data,
3434 btrfs_item_ptr_offset(path->nodes[0],
3435 path->slots[0]),
3436 size);
3437 inode_add_bytes(dst, datal);
3438
3439 return 0;
3440 }
3441
3442 /**
3443 * btrfs_clone() - clone a range from inode file to another
3444 *
3445 * @src: Inode to clone from
3446 * @inode: Inode to clone to
3447 * @off: Offset within source to start clone from
3448 * @olen: Original length, passed by user, of range to clone
3449 * @olen_aligned: Block-aligned value of olen
3450 * @destoff: Offset within @inode to start clone
3451 * @no_time_update: Whether to update mtime/ctime on the target inode
3452 */
3453 static int btrfs_clone(struct inode *src, struct inode *inode,
3454 const u64 off, const u64 olen, const u64 olen_aligned,
3455 const u64 destoff, int no_time_update)
3456 {
3457 struct btrfs_root *root = BTRFS_I(inode)->root;
3458 struct btrfs_path *path = NULL;
3459 struct extent_buffer *leaf;
3460 struct btrfs_trans_handle *trans;
3461 char *buf = NULL;
3462 struct btrfs_key key;
3463 u32 nritems;
3464 int slot;
3465 int ret;
3466 const u64 len = olen_aligned;
3467 u64 last_dest_end = destoff;
3468
3469 ret = -ENOMEM;
3470 buf = vmalloc(root->nodesize);
3471 if (!buf)
3472 return ret;
3473
3474 path = btrfs_alloc_path();
3475 if (!path) {
3476 vfree(buf);
3477 return ret;
3478 }
3479
3480 path->reada = READA_FORWARD;
3481 /* clone data */
3482 key.objectid = btrfs_ino(src);
3483 key.type = BTRFS_EXTENT_DATA_KEY;
3484 key.offset = off;
3485
3486 while (1) {
3487 u64 next_key_min_offset = key.offset + 1;
3488
3489 /*
3490 * note the key will change type as we walk through the
3491 * tree.
3492 */
3493 path->leave_spinning = 1;
3494 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3495 0, 0);
3496 if (ret < 0)
3497 goto out;
3498 /*
3499 * First search, if no extent item that starts at offset off was
3500 * found but the previous item is an extent item, it's possible
3501 * it might overlap our target range, therefore process it.
3502 */
3503 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3504 btrfs_item_key_to_cpu(path->nodes[0], &key,
3505 path->slots[0] - 1);
3506 if (key.type == BTRFS_EXTENT_DATA_KEY)
3507 path->slots[0]--;
3508 }
3509
3510 nritems = btrfs_header_nritems(path->nodes[0]);
3511 process_slot:
3512 if (path->slots[0] >= nritems) {
3513 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3514 if (ret < 0)
3515 goto out;
3516 if (ret > 0)
3517 break;
3518 nritems = btrfs_header_nritems(path->nodes[0]);
3519 }
3520 leaf = path->nodes[0];
3521 slot = path->slots[0];
3522
3523 btrfs_item_key_to_cpu(leaf, &key, slot);
3524 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3525 key.objectid != btrfs_ino(src))
3526 break;
3527
3528 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3529 struct btrfs_file_extent_item *extent;
3530 int type;
3531 u32 size;
3532 struct btrfs_key new_key;
3533 u64 disko = 0, diskl = 0;
3534 u64 datao = 0, datal = 0;
3535 u8 comp;
3536 u64 drop_start;
3537
3538 extent = btrfs_item_ptr(leaf, slot,
3539 struct btrfs_file_extent_item);
3540 comp = btrfs_file_extent_compression(leaf, extent);
3541 type = btrfs_file_extent_type(leaf, extent);
3542 if (type == BTRFS_FILE_EXTENT_REG ||
3543 type == BTRFS_FILE_EXTENT_PREALLOC) {
3544 disko = btrfs_file_extent_disk_bytenr(leaf,
3545 extent);
3546 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3547 extent);
3548 datao = btrfs_file_extent_offset(leaf, extent);
3549 datal = btrfs_file_extent_num_bytes(leaf,
3550 extent);
3551 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3552 /* take upper bound, may be compressed */
3553 datal = btrfs_file_extent_ram_bytes(leaf,
3554 extent);
3555 }
3556
3557 /*
3558 * The first search might have left us at an extent
3559 * item that ends before our target range's start, can
3560 * happen if we have holes and NO_HOLES feature enabled.
3561 */
3562 if (key.offset + datal <= off) {
3563 path->slots[0]++;
3564 goto process_slot;
3565 } else if (key.offset >= off + len) {
3566 break;
3567 }
3568 next_key_min_offset = key.offset + datal;
3569 size = btrfs_item_size_nr(leaf, slot);
3570 read_extent_buffer(leaf, buf,
3571 btrfs_item_ptr_offset(leaf, slot),
3572 size);
3573
3574 btrfs_release_path(path);
3575 path->leave_spinning = 0;
3576
3577 memcpy(&new_key, &key, sizeof(new_key));
3578 new_key.objectid = btrfs_ino(inode);
3579 if (off <= key.offset)
3580 new_key.offset = key.offset + destoff - off;
3581 else
3582 new_key.offset = destoff;
3583
3584 /*
3585 * Deal with a hole that doesn't have an extent item
3586 * that represents it (NO_HOLES feature enabled).
3587 * This hole is either in the middle of the cloning
3588 * range or at the beginning (fully overlaps it or
3589 * partially overlaps it).
3590 */
3591 if (new_key.offset != last_dest_end)
3592 drop_start = last_dest_end;
3593 else
3594 drop_start = new_key.offset;
3595
3596 /*
3597 * 1 - adjusting old extent (we may have to split it)
3598 * 1 - add new extent
3599 * 1 - inode update
3600 */
3601 trans = btrfs_start_transaction(root, 3);
3602 if (IS_ERR(trans)) {
3603 ret = PTR_ERR(trans);
3604 goto out;
3605 }
3606
3607 if (type == BTRFS_FILE_EXTENT_REG ||
3608 type == BTRFS_FILE_EXTENT_PREALLOC) {
3609 /*
3610 * a | --- range to clone ---| b
3611 * | ------------- extent ------------- |
3612 */
3613
3614 /* subtract range b */
3615 if (key.offset + datal > off + len)
3616 datal = off + len - key.offset;
3617
3618 /* subtract range a */
3619 if (off > key.offset) {
3620 datao += off - key.offset;
3621 datal -= off - key.offset;
3622 }
3623
3624 ret = btrfs_drop_extents(trans, root, inode,
3625 drop_start,
3626 new_key.offset + datal,
3627 1);
3628 if (ret) {
3629 if (ret != -EOPNOTSUPP)
3630 btrfs_abort_transaction(trans,
3631 root, ret);
3632 btrfs_end_transaction(trans, root);
3633 goto out;
3634 }
3635
3636 ret = btrfs_insert_empty_item(trans, root, path,
3637 &new_key, size);
3638 if (ret) {
3639 btrfs_abort_transaction(trans, root,
3640 ret);
3641 btrfs_end_transaction(trans, root);
3642 goto out;
3643 }
3644
3645 leaf = path->nodes[0];
3646 slot = path->slots[0];
3647 write_extent_buffer(leaf, buf,
3648 btrfs_item_ptr_offset(leaf, slot),
3649 size);
3650
3651 extent = btrfs_item_ptr(leaf, slot,
3652 struct btrfs_file_extent_item);
3653
3654 /* disko == 0 means it's a hole */
3655 if (!disko)
3656 datao = 0;
3657
3658 btrfs_set_file_extent_offset(leaf, extent,
3659 datao);
3660 btrfs_set_file_extent_num_bytes(leaf, extent,
3661 datal);
3662
3663 if (disko) {
3664 inode_add_bytes(inode, datal);
3665 ret = btrfs_inc_extent_ref(trans, root,
3666 disko, diskl, 0,
3667 root->root_key.objectid,
3668 btrfs_ino(inode),
3669 new_key.offset - datao);
3670 if (ret) {
3671 btrfs_abort_transaction(trans,
3672 root,
3673 ret);
3674 btrfs_end_transaction(trans,
3675 root);
3676 goto out;
3677
3678 }
3679 }
3680 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3681 u64 skip = 0;
3682 u64 trim = 0;
3683
3684 if (off > key.offset) {
3685 skip = off - key.offset;
3686 new_key.offset += skip;
3687 }
3688
3689 if (key.offset + datal > off + len)
3690 trim = key.offset + datal - (off + len);
3691
3692 if (comp && (skip || trim)) {
3693 ret = -EINVAL;
3694 btrfs_end_transaction(trans, root);
3695 goto out;
3696 }
3697 size -= skip + trim;
3698 datal -= skip + trim;
3699
3700 ret = clone_copy_inline_extent(src, inode,
3701 trans, path,
3702 &new_key,
3703 drop_start,
3704 datal,
3705 skip, size, buf);
3706 if (ret) {
3707 if (ret != -EOPNOTSUPP)
3708 btrfs_abort_transaction(trans,
3709 root,
3710 ret);
3711 btrfs_end_transaction(trans, root);
3712 goto out;
3713 }
3714 leaf = path->nodes[0];
3715 slot = path->slots[0];
3716 }
3717
3718 /* If we have an implicit hole (NO_HOLES feature). */
3719 if (drop_start < new_key.offset)
3720 clone_update_extent_map(inode, trans,
3721 NULL, drop_start,
3722 new_key.offset - drop_start);
3723
3724 clone_update_extent_map(inode, trans, path, 0, 0);
3725
3726 btrfs_mark_buffer_dirty(leaf);
3727 btrfs_release_path(path);
3728
3729 last_dest_end = ALIGN(new_key.offset + datal,
3730 root->sectorsize);
3731 ret = clone_finish_inode_update(trans, inode,
3732 last_dest_end,
3733 destoff, olen,
3734 no_time_update);
3735 if (ret)
3736 goto out;
3737 if (new_key.offset + datal >= destoff + len)
3738 break;
3739 }
3740 btrfs_release_path(path);
3741 key.offset = next_key_min_offset;
3742 }
3743 ret = 0;
3744
3745 if (last_dest_end < destoff + len) {
3746 /*
3747 * We have an implicit hole (NO_HOLES feature is enabled) that
3748 * fully or partially overlaps our cloning range at its end.
3749 */
3750 btrfs_release_path(path);
3751
3752 /*
3753 * 1 - remove extent(s)
3754 * 1 - inode update
3755 */
3756 trans = btrfs_start_transaction(root, 2);
3757 if (IS_ERR(trans)) {
3758 ret = PTR_ERR(trans);
3759 goto out;
3760 }
3761 ret = btrfs_drop_extents(trans, root, inode,
3762 last_dest_end, destoff + len, 1);
3763 if (ret) {
3764 if (ret != -EOPNOTSUPP)
3765 btrfs_abort_transaction(trans, root, ret);
3766 btrfs_end_transaction(trans, root);
3767 goto out;
3768 }
3769 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3770 destoff + len - last_dest_end);
3771 ret = clone_finish_inode_update(trans, inode, destoff + len,
3772 destoff, olen, no_time_update);
3773 }
3774
3775 out:
3776 btrfs_free_path(path);
3777 vfree(buf);
3778 return ret;
3779 }
3780
3781 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3782 u64 off, u64 olen, u64 destoff)
3783 {
3784 struct inode *inode = file_inode(file);
3785 struct inode *src = file_inode(file_src);
3786 struct btrfs_root *root = BTRFS_I(inode)->root;
3787 int ret;
3788 u64 len = olen;
3789 u64 bs = root->fs_info->sb->s_blocksize;
3790 int same_inode = src == inode;
3791
3792 /*
3793 * TODO:
3794 * - split compressed inline extents. annoying: we need to
3795 * decompress into destination's address_space (the file offset
3796 * may change, so source mapping won't do), then recompress (or
3797 * otherwise reinsert) a subrange.
3798 *
3799 * - split destination inode's inline extents. The inline extents can
3800 * be either compressed or non-compressed.
3801 */
3802
3803 if (btrfs_root_readonly(root))
3804 return -EROFS;
3805
3806 if (file_src->f_path.mnt != file->f_path.mnt ||
3807 src->i_sb != inode->i_sb)
3808 return -EXDEV;
3809
3810 /* don't make the dst file partly checksummed */
3811 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3812 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3813 return -EINVAL;
3814
3815 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3816 return -EISDIR;
3817
3818 if (!same_inode) {
3819 btrfs_double_inode_lock(src, inode);
3820 } else {
3821 inode_lock(src);
3822 }
3823
3824 /* determine range to clone */
3825 ret = -EINVAL;
3826 if (off + len > src->i_size || off + len < off)
3827 goto out_unlock;
3828 if (len == 0)
3829 olen = len = src->i_size - off;
3830 /* if we extend to eof, continue to block boundary */
3831 if (off + len == src->i_size)
3832 len = ALIGN(src->i_size, bs) - off;
3833
3834 if (len == 0) {
3835 ret = 0;
3836 goto out_unlock;
3837 }
3838
3839 /* verify the end result is block aligned */
3840 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3841 !IS_ALIGNED(destoff, bs))
3842 goto out_unlock;
3843
3844 /* verify if ranges are overlapped within the same file */
3845 if (same_inode) {
3846 if (destoff + len > off && destoff < off + len)
3847 goto out_unlock;
3848 }
3849
3850 if (destoff > inode->i_size) {
3851 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3852 if (ret)
3853 goto out_unlock;
3854 }
3855
3856 /*
3857 * Lock the target range too. Right after we replace the file extent
3858 * items in the fs tree (which now point to the cloned data), we might
3859 * have a worker replace them with extent items relative to a write
3860 * operation that was issued before this clone operation (i.e. confront
3861 * with inode.c:btrfs_finish_ordered_io).
3862 */
3863 if (same_inode) {
3864 u64 lock_start = min_t(u64, off, destoff);
3865 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3866
3867 ret = lock_extent_range(src, lock_start, lock_len, true);
3868 } else {
3869 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3870 true);
3871 }
3872 ASSERT(ret == 0);
3873 if (WARN_ON(ret)) {
3874 /* ranges in the io trees already unlocked */
3875 goto out_unlock;
3876 }
3877
3878 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3879
3880 if (same_inode) {
3881 u64 lock_start = min_t(u64, off, destoff);
3882 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3883
3884 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3885 } else {
3886 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3887 }
3888 /*
3889 * Truncate page cache pages so that future reads will see the cloned
3890 * data immediately and not the previous data.
3891 */
3892 truncate_inode_pages_range(&inode->i_data, destoff,
3893 PAGE_CACHE_ALIGN(destoff + len) - 1);
3894 out_unlock:
3895 if (!same_inode)
3896 btrfs_double_inode_unlock(src, inode);
3897 else
3898 inode_unlock(src);
3899 return ret;
3900 }
3901
3902 ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in,
3903 struct file *file_out, loff_t pos_out,
3904 size_t len, unsigned int flags)
3905 {
3906 ssize_t ret;
3907
3908 ret = btrfs_clone_files(file_out, file_in, pos_in, len, pos_out);
3909 if (ret == 0)
3910 ret = len;
3911 return ret;
3912 }
3913
3914 int btrfs_clone_file_range(struct file *src_file, loff_t off,
3915 struct file *dst_file, loff_t destoff, u64 len)
3916 {
3917 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
3918 }
3919
3920 /*
3921 * there are many ways the trans_start and trans_end ioctls can lead
3922 * to deadlocks. They should only be used by applications that
3923 * basically own the machine, and have a very in depth understanding
3924 * of all the possible deadlocks and enospc problems.
3925 */
3926 static long btrfs_ioctl_trans_start(struct file *file)
3927 {
3928 struct inode *inode = file_inode(file);
3929 struct btrfs_root *root = BTRFS_I(inode)->root;
3930 struct btrfs_trans_handle *trans;
3931 int ret;
3932
3933 ret = -EPERM;
3934 if (!capable(CAP_SYS_ADMIN))
3935 goto out;
3936
3937 ret = -EINPROGRESS;
3938 if (file->private_data)
3939 goto out;
3940
3941 ret = -EROFS;
3942 if (btrfs_root_readonly(root))
3943 goto out;
3944
3945 ret = mnt_want_write_file(file);
3946 if (ret)
3947 goto out;
3948
3949 atomic_inc(&root->fs_info->open_ioctl_trans);
3950
3951 ret = -ENOMEM;
3952 trans = btrfs_start_ioctl_transaction(root);
3953 if (IS_ERR(trans))
3954 goto out_drop;
3955
3956 file->private_data = trans;
3957 return 0;
3958
3959 out_drop:
3960 atomic_dec(&root->fs_info->open_ioctl_trans);
3961 mnt_drop_write_file(file);
3962 out:
3963 return ret;
3964 }
3965
3966 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3967 {
3968 struct inode *inode = file_inode(file);
3969 struct btrfs_root *root = BTRFS_I(inode)->root;
3970 struct btrfs_root *new_root;
3971 struct btrfs_dir_item *di;
3972 struct btrfs_trans_handle *trans;
3973 struct btrfs_path *path;
3974 struct btrfs_key location;
3975 struct btrfs_disk_key disk_key;
3976 u64 objectid = 0;
3977 u64 dir_id;
3978 int ret;
3979
3980 if (!capable(CAP_SYS_ADMIN))
3981 return -EPERM;
3982
3983 ret = mnt_want_write_file(file);
3984 if (ret)
3985 return ret;
3986
3987 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3988 ret = -EFAULT;
3989 goto out;
3990 }
3991
3992 if (!objectid)
3993 objectid = BTRFS_FS_TREE_OBJECTID;
3994
3995 location.objectid = objectid;
3996 location.type = BTRFS_ROOT_ITEM_KEY;
3997 location.offset = (u64)-1;
3998
3999 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
4000 if (IS_ERR(new_root)) {
4001 ret = PTR_ERR(new_root);
4002 goto out;
4003 }
4004
4005 path = btrfs_alloc_path();
4006 if (!path) {
4007 ret = -ENOMEM;
4008 goto out;
4009 }
4010 path->leave_spinning = 1;
4011
4012 trans = btrfs_start_transaction(root, 1);
4013 if (IS_ERR(trans)) {
4014 btrfs_free_path(path);
4015 ret = PTR_ERR(trans);
4016 goto out;
4017 }
4018
4019 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
4020 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
4021 dir_id, "default", 7, 1);
4022 if (IS_ERR_OR_NULL(di)) {
4023 btrfs_free_path(path);
4024 btrfs_end_transaction(trans, root);
4025 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
4026 "item, this isn't going to work");
4027 ret = -ENOENT;
4028 goto out;
4029 }
4030
4031 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4032 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4033 btrfs_mark_buffer_dirty(path->nodes[0]);
4034 btrfs_free_path(path);
4035
4036 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
4037 btrfs_end_transaction(trans, root);
4038 out:
4039 mnt_drop_write_file(file);
4040 return ret;
4041 }
4042
4043 void btrfs_get_block_group_info(struct list_head *groups_list,
4044 struct btrfs_ioctl_space_info *space)
4045 {
4046 struct btrfs_block_group_cache *block_group;
4047
4048 space->total_bytes = 0;
4049 space->used_bytes = 0;
4050 space->flags = 0;
4051 list_for_each_entry(block_group, groups_list, list) {
4052 space->flags = block_group->flags;
4053 space->total_bytes += block_group->key.offset;
4054 space->used_bytes +=
4055 btrfs_block_group_used(&block_group->item);
4056 }
4057 }
4058
4059 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
4060 {
4061 struct btrfs_ioctl_space_args space_args;
4062 struct btrfs_ioctl_space_info space;
4063 struct btrfs_ioctl_space_info *dest;
4064 struct btrfs_ioctl_space_info *dest_orig;
4065 struct btrfs_ioctl_space_info __user *user_dest;
4066 struct btrfs_space_info *info;
4067 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4068 BTRFS_BLOCK_GROUP_SYSTEM,
4069 BTRFS_BLOCK_GROUP_METADATA,
4070 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4071 int num_types = 4;
4072 int alloc_size;
4073 int ret = 0;
4074 u64 slot_count = 0;
4075 int i, c;
4076
4077 if (copy_from_user(&space_args,
4078 (struct btrfs_ioctl_space_args __user *)arg,
4079 sizeof(space_args)))
4080 return -EFAULT;
4081
4082 for (i = 0; i < num_types; i++) {
4083 struct btrfs_space_info *tmp;
4084
4085 info = NULL;
4086 rcu_read_lock();
4087 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4088 list) {
4089 if (tmp->flags == types[i]) {
4090 info = tmp;
4091 break;
4092 }
4093 }
4094 rcu_read_unlock();
4095
4096 if (!info)
4097 continue;
4098
4099 down_read(&info->groups_sem);
4100 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4101 if (!list_empty(&info->block_groups[c]))
4102 slot_count++;
4103 }
4104 up_read(&info->groups_sem);
4105 }
4106
4107 /*
4108 * Global block reserve, exported as a space_info
4109 */
4110 slot_count++;
4111
4112 /* space_slots == 0 means they are asking for a count */
4113 if (space_args.space_slots == 0) {
4114 space_args.total_spaces = slot_count;
4115 goto out;
4116 }
4117
4118 slot_count = min_t(u64, space_args.space_slots, slot_count);
4119
4120 alloc_size = sizeof(*dest) * slot_count;
4121
4122 /* we generally have at most 6 or so space infos, one for each raid
4123 * level. So, a whole page should be more than enough for everyone
4124 */
4125 if (alloc_size > PAGE_CACHE_SIZE)
4126 return -ENOMEM;
4127
4128 space_args.total_spaces = 0;
4129 dest = kmalloc(alloc_size, GFP_KERNEL);
4130 if (!dest)
4131 return -ENOMEM;
4132 dest_orig = dest;
4133
4134 /* now we have a buffer to copy into */
4135 for (i = 0; i < num_types; i++) {
4136 struct btrfs_space_info *tmp;
4137
4138 if (!slot_count)
4139 break;
4140
4141 info = NULL;
4142 rcu_read_lock();
4143 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4144 list) {
4145 if (tmp->flags == types[i]) {
4146 info = tmp;
4147 break;
4148 }
4149 }
4150 rcu_read_unlock();
4151
4152 if (!info)
4153 continue;
4154 down_read(&info->groups_sem);
4155 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4156 if (!list_empty(&info->block_groups[c])) {
4157 btrfs_get_block_group_info(
4158 &info->block_groups[c], &space);
4159 memcpy(dest, &space, sizeof(space));
4160 dest++;
4161 space_args.total_spaces++;
4162 slot_count--;
4163 }
4164 if (!slot_count)
4165 break;
4166 }
4167 up_read(&info->groups_sem);
4168 }
4169
4170 /*
4171 * Add global block reserve
4172 */
4173 if (slot_count) {
4174 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4175
4176 spin_lock(&block_rsv->lock);
4177 space.total_bytes = block_rsv->size;
4178 space.used_bytes = block_rsv->size - block_rsv->reserved;
4179 spin_unlock(&block_rsv->lock);
4180 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4181 memcpy(dest, &space, sizeof(space));
4182 space_args.total_spaces++;
4183 }
4184
4185 user_dest = (struct btrfs_ioctl_space_info __user *)
4186 (arg + sizeof(struct btrfs_ioctl_space_args));
4187
4188 if (copy_to_user(user_dest, dest_orig, alloc_size))
4189 ret = -EFAULT;
4190
4191 kfree(dest_orig);
4192 out:
4193 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4194 ret = -EFAULT;
4195
4196 return ret;
4197 }
4198
4199 /*
4200 * there are many ways the trans_start and trans_end ioctls can lead
4201 * to deadlocks. They should only be used by applications that
4202 * basically own the machine, and have a very in depth understanding
4203 * of all the possible deadlocks and enospc problems.
4204 */
4205 long btrfs_ioctl_trans_end(struct file *file)
4206 {
4207 struct inode *inode = file_inode(file);
4208 struct btrfs_root *root = BTRFS_I(inode)->root;
4209 struct btrfs_trans_handle *trans;
4210
4211 trans = file->private_data;
4212 if (!trans)
4213 return -EINVAL;
4214 file->private_data = NULL;
4215
4216 btrfs_end_transaction(trans, root);
4217
4218 atomic_dec(&root->fs_info->open_ioctl_trans);
4219
4220 mnt_drop_write_file(file);
4221 return 0;
4222 }
4223
4224 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4225 void __user *argp)
4226 {
4227 struct btrfs_trans_handle *trans;
4228 u64 transid;
4229 int ret;
4230
4231 trans = btrfs_attach_transaction_barrier(root);
4232 if (IS_ERR(trans)) {
4233 if (PTR_ERR(trans) != -ENOENT)
4234 return PTR_ERR(trans);
4235
4236 /* No running transaction, don't bother */
4237 transid = root->fs_info->last_trans_committed;
4238 goto out;
4239 }
4240 transid = trans->transid;
4241 ret = btrfs_commit_transaction_async(trans, root, 0);
4242 if (ret) {
4243 btrfs_end_transaction(trans, root);
4244 return ret;
4245 }
4246 out:
4247 if (argp)
4248 if (copy_to_user(argp, &transid, sizeof(transid)))
4249 return -EFAULT;
4250 return 0;
4251 }
4252
4253 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4254 void __user *argp)
4255 {
4256 u64 transid;
4257
4258 if (argp) {
4259 if (copy_from_user(&transid, argp, sizeof(transid)))
4260 return -EFAULT;
4261 } else {
4262 transid = 0; /* current trans */
4263 }
4264 return btrfs_wait_for_commit(root, transid);
4265 }
4266
4267 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4268 {
4269 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4270 struct btrfs_ioctl_scrub_args *sa;
4271 int ret;
4272
4273 if (!capable(CAP_SYS_ADMIN))
4274 return -EPERM;
4275
4276 sa = memdup_user(arg, sizeof(*sa));
4277 if (IS_ERR(sa))
4278 return PTR_ERR(sa);
4279
4280 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4281 ret = mnt_want_write_file(file);
4282 if (ret)
4283 goto out;
4284 }
4285
4286 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4287 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4288 0);
4289
4290 if (copy_to_user(arg, sa, sizeof(*sa)))
4291 ret = -EFAULT;
4292
4293 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4294 mnt_drop_write_file(file);
4295 out:
4296 kfree(sa);
4297 return ret;
4298 }
4299
4300 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4301 {
4302 if (!capable(CAP_SYS_ADMIN))
4303 return -EPERM;
4304
4305 return btrfs_scrub_cancel(root->fs_info);
4306 }
4307
4308 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4309 void __user *arg)
4310 {
4311 struct btrfs_ioctl_scrub_args *sa;
4312 int ret;
4313
4314 if (!capable(CAP_SYS_ADMIN))
4315 return -EPERM;
4316
4317 sa = memdup_user(arg, sizeof(*sa));
4318 if (IS_ERR(sa))
4319 return PTR_ERR(sa);
4320
4321 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4322
4323 if (copy_to_user(arg, sa, sizeof(*sa)))
4324 ret = -EFAULT;
4325
4326 kfree(sa);
4327 return ret;
4328 }
4329
4330 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4331 void __user *arg)
4332 {
4333 struct btrfs_ioctl_get_dev_stats *sa;
4334 int ret;
4335
4336 sa = memdup_user(arg, sizeof(*sa));
4337 if (IS_ERR(sa))
4338 return PTR_ERR(sa);
4339
4340 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4341 kfree(sa);
4342 return -EPERM;
4343 }
4344
4345 ret = btrfs_get_dev_stats(root, sa);
4346
4347 if (copy_to_user(arg, sa, sizeof(*sa)))
4348 ret = -EFAULT;
4349
4350 kfree(sa);
4351 return ret;
4352 }
4353
4354 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4355 {
4356 struct btrfs_ioctl_dev_replace_args *p;
4357 int ret;
4358
4359 if (!capable(CAP_SYS_ADMIN))
4360 return -EPERM;
4361
4362 p = memdup_user(arg, sizeof(*p));
4363 if (IS_ERR(p))
4364 return PTR_ERR(p);
4365
4366 switch (p->cmd) {
4367 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4368 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4369 ret = -EROFS;
4370 goto out;
4371 }
4372 if (atomic_xchg(
4373 &root->fs_info->mutually_exclusive_operation_running,
4374 1)) {
4375 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4376 } else {
4377 ret = btrfs_dev_replace_start(root, p);
4378 atomic_set(
4379 &root->fs_info->mutually_exclusive_operation_running,
4380 0);
4381 }
4382 break;
4383 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4384 btrfs_dev_replace_status(root->fs_info, p);
4385 ret = 0;
4386 break;
4387 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4388 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4389 break;
4390 default:
4391 ret = -EINVAL;
4392 break;
4393 }
4394
4395 if (copy_to_user(arg, p, sizeof(*p)))
4396 ret = -EFAULT;
4397 out:
4398 kfree(p);
4399 return ret;
4400 }
4401
4402 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4403 {
4404 int ret = 0;
4405 int i;
4406 u64 rel_ptr;
4407 int size;
4408 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4409 struct inode_fs_paths *ipath = NULL;
4410 struct btrfs_path *path;
4411
4412 if (!capable(CAP_DAC_READ_SEARCH))
4413 return -EPERM;
4414
4415 path = btrfs_alloc_path();
4416 if (!path) {
4417 ret = -ENOMEM;
4418 goto out;
4419 }
4420
4421 ipa = memdup_user(arg, sizeof(*ipa));
4422 if (IS_ERR(ipa)) {
4423 ret = PTR_ERR(ipa);
4424 ipa = NULL;
4425 goto out;
4426 }
4427
4428 size = min_t(u32, ipa->size, 4096);
4429 ipath = init_ipath(size, root, path);
4430 if (IS_ERR(ipath)) {
4431 ret = PTR_ERR(ipath);
4432 ipath = NULL;
4433 goto out;
4434 }
4435
4436 ret = paths_from_inode(ipa->inum, ipath);
4437 if (ret < 0)
4438 goto out;
4439
4440 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4441 rel_ptr = ipath->fspath->val[i] -
4442 (u64)(unsigned long)ipath->fspath->val;
4443 ipath->fspath->val[i] = rel_ptr;
4444 }
4445
4446 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4447 (void *)(unsigned long)ipath->fspath, size);
4448 if (ret) {
4449 ret = -EFAULT;
4450 goto out;
4451 }
4452
4453 out:
4454 btrfs_free_path(path);
4455 free_ipath(ipath);
4456 kfree(ipa);
4457
4458 return ret;
4459 }
4460
4461 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4462 {
4463 struct btrfs_data_container *inodes = ctx;
4464 const size_t c = 3 * sizeof(u64);
4465
4466 if (inodes->bytes_left >= c) {
4467 inodes->bytes_left -= c;
4468 inodes->val[inodes->elem_cnt] = inum;
4469 inodes->val[inodes->elem_cnt + 1] = offset;
4470 inodes->val[inodes->elem_cnt + 2] = root;
4471 inodes->elem_cnt += 3;
4472 } else {
4473 inodes->bytes_missing += c - inodes->bytes_left;
4474 inodes->bytes_left = 0;
4475 inodes->elem_missed += 3;
4476 }
4477
4478 return 0;
4479 }
4480
4481 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4482 void __user *arg)
4483 {
4484 int ret = 0;
4485 int size;
4486 struct btrfs_ioctl_logical_ino_args *loi;
4487 struct btrfs_data_container *inodes = NULL;
4488 struct btrfs_path *path = NULL;
4489
4490 if (!capable(CAP_SYS_ADMIN))
4491 return -EPERM;
4492
4493 loi = memdup_user(arg, sizeof(*loi));
4494 if (IS_ERR(loi)) {
4495 ret = PTR_ERR(loi);
4496 loi = NULL;
4497 goto out;
4498 }
4499
4500 path = btrfs_alloc_path();
4501 if (!path) {
4502 ret = -ENOMEM;
4503 goto out;
4504 }
4505
4506 size = min_t(u32, loi->size, SZ_64K);
4507 inodes = init_data_container(size);
4508 if (IS_ERR(inodes)) {
4509 ret = PTR_ERR(inodes);
4510 inodes = NULL;
4511 goto out;
4512 }
4513
4514 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4515 build_ino_list, inodes);
4516 if (ret == -EINVAL)
4517 ret = -ENOENT;
4518 if (ret < 0)
4519 goto out;
4520
4521 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4522 (void *)(unsigned long)inodes, size);
4523 if (ret)
4524 ret = -EFAULT;
4525
4526 out:
4527 btrfs_free_path(path);
4528 vfree(inodes);
4529 kfree(loi);
4530
4531 return ret;
4532 }
4533
4534 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4535 struct btrfs_ioctl_balance_args *bargs)
4536 {
4537 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4538
4539 bargs->flags = bctl->flags;
4540
4541 if (atomic_read(&fs_info->balance_running))
4542 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4543 if (atomic_read(&fs_info->balance_pause_req))
4544 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4545 if (atomic_read(&fs_info->balance_cancel_req))
4546 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4547
4548 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4549 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4550 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4551
4552 if (lock) {
4553 spin_lock(&fs_info->balance_lock);
4554 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4555 spin_unlock(&fs_info->balance_lock);
4556 } else {
4557 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4558 }
4559 }
4560
4561 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4562 {
4563 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4564 struct btrfs_fs_info *fs_info = root->fs_info;
4565 struct btrfs_ioctl_balance_args *bargs;
4566 struct btrfs_balance_control *bctl;
4567 bool need_unlock; /* for mut. excl. ops lock */
4568 int ret;
4569
4570 if (!capable(CAP_SYS_ADMIN))
4571 return -EPERM;
4572
4573 ret = mnt_want_write_file(file);
4574 if (ret)
4575 return ret;
4576
4577 again:
4578 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4579 mutex_lock(&fs_info->volume_mutex);
4580 mutex_lock(&fs_info->balance_mutex);
4581 need_unlock = true;
4582 goto locked;
4583 }
4584
4585 /*
4586 * mut. excl. ops lock is locked. Three possibilites:
4587 * (1) some other op is running
4588 * (2) balance is running
4589 * (3) balance is paused -- special case (think resume)
4590 */
4591 mutex_lock(&fs_info->balance_mutex);
4592 if (fs_info->balance_ctl) {
4593 /* this is either (2) or (3) */
4594 if (!atomic_read(&fs_info->balance_running)) {
4595 mutex_unlock(&fs_info->balance_mutex);
4596 if (!mutex_trylock(&fs_info->volume_mutex))
4597 goto again;
4598 mutex_lock(&fs_info->balance_mutex);
4599
4600 if (fs_info->balance_ctl &&
4601 !atomic_read(&fs_info->balance_running)) {
4602 /* this is (3) */
4603 need_unlock = false;
4604 goto locked;
4605 }
4606
4607 mutex_unlock(&fs_info->balance_mutex);
4608 mutex_unlock(&fs_info->volume_mutex);
4609 goto again;
4610 } else {
4611 /* this is (2) */
4612 mutex_unlock(&fs_info->balance_mutex);
4613 ret = -EINPROGRESS;
4614 goto out;
4615 }
4616 } else {
4617 /* this is (1) */
4618 mutex_unlock(&fs_info->balance_mutex);
4619 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4620 goto out;
4621 }
4622
4623 locked:
4624 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4625
4626 if (arg) {
4627 bargs = memdup_user(arg, sizeof(*bargs));
4628 if (IS_ERR(bargs)) {
4629 ret = PTR_ERR(bargs);
4630 goto out_unlock;
4631 }
4632
4633 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4634 if (!fs_info->balance_ctl) {
4635 ret = -ENOTCONN;
4636 goto out_bargs;
4637 }
4638
4639 bctl = fs_info->balance_ctl;
4640 spin_lock(&fs_info->balance_lock);
4641 bctl->flags |= BTRFS_BALANCE_RESUME;
4642 spin_unlock(&fs_info->balance_lock);
4643
4644 goto do_balance;
4645 }
4646 } else {
4647 bargs = NULL;
4648 }
4649
4650 if (fs_info->balance_ctl) {
4651 ret = -EINPROGRESS;
4652 goto out_bargs;
4653 }
4654
4655 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4656 if (!bctl) {
4657 ret = -ENOMEM;
4658 goto out_bargs;
4659 }
4660
4661 bctl->fs_info = fs_info;
4662 if (arg) {
4663 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4664 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4665 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4666
4667 bctl->flags = bargs->flags;
4668 } else {
4669 /* balance everything - no filters */
4670 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4671 }
4672
4673 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4674 ret = -EINVAL;
4675 goto out_bctl;
4676 }
4677
4678 do_balance:
4679 /*
4680 * Ownership of bctl and mutually_exclusive_operation_running
4681 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4682 * or, if restriper was paused all the way until unmount, in
4683 * free_fs_info. mutually_exclusive_operation_running is
4684 * cleared in __cancel_balance.
4685 */
4686 need_unlock = false;
4687
4688 ret = btrfs_balance(bctl, bargs);
4689 bctl = NULL;
4690
4691 if (arg) {
4692 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4693 ret = -EFAULT;
4694 }
4695
4696 out_bctl:
4697 kfree(bctl);
4698 out_bargs:
4699 kfree(bargs);
4700 out_unlock:
4701 mutex_unlock(&fs_info->balance_mutex);
4702 mutex_unlock(&fs_info->volume_mutex);
4703 if (need_unlock)
4704 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4705 out:
4706 mnt_drop_write_file(file);
4707 return ret;
4708 }
4709
4710 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4711 {
4712 if (!capable(CAP_SYS_ADMIN))
4713 return -EPERM;
4714
4715 switch (cmd) {
4716 case BTRFS_BALANCE_CTL_PAUSE:
4717 return btrfs_pause_balance(root->fs_info);
4718 case BTRFS_BALANCE_CTL_CANCEL:
4719 return btrfs_cancel_balance(root->fs_info);
4720 }
4721
4722 return -EINVAL;
4723 }
4724
4725 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4726 void __user *arg)
4727 {
4728 struct btrfs_fs_info *fs_info = root->fs_info;
4729 struct btrfs_ioctl_balance_args *bargs;
4730 int ret = 0;
4731
4732 if (!capable(CAP_SYS_ADMIN))
4733 return -EPERM;
4734
4735 mutex_lock(&fs_info->balance_mutex);
4736 if (!fs_info->balance_ctl) {
4737 ret = -ENOTCONN;
4738 goto out;
4739 }
4740
4741 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4742 if (!bargs) {
4743 ret = -ENOMEM;
4744 goto out;
4745 }
4746
4747 update_ioctl_balance_args(fs_info, 1, bargs);
4748
4749 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4750 ret = -EFAULT;
4751
4752 kfree(bargs);
4753 out:
4754 mutex_unlock(&fs_info->balance_mutex);
4755 return ret;
4756 }
4757
4758 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4759 {
4760 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4761 struct btrfs_ioctl_quota_ctl_args *sa;
4762 struct btrfs_trans_handle *trans = NULL;
4763 int ret;
4764 int err;
4765
4766 if (!capable(CAP_SYS_ADMIN))
4767 return -EPERM;
4768
4769 ret = mnt_want_write_file(file);
4770 if (ret)
4771 return ret;
4772
4773 sa = memdup_user(arg, sizeof(*sa));
4774 if (IS_ERR(sa)) {
4775 ret = PTR_ERR(sa);
4776 goto drop_write;
4777 }
4778
4779 down_write(&root->fs_info->subvol_sem);
4780 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4781 if (IS_ERR(trans)) {
4782 ret = PTR_ERR(trans);
4783 goto out;
4784 }
4785
4786 switch (sa->cmd) {
4787 case BTRFS_QUOTA_CTL_ENABLE:
4788 ret = btrfs_quota_enable(trans, root->fs_info);
4789 break;
4790 case BTRFS_QUOTA_CTL_DISABLE:
4791 ret = btrfs_quota_disable(trans, root->fs_info);
4792 break;
4793 default:
4794 ret = -EINVAL;
4795 break;
4796 }
4797
4798 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4799 if (err && !ret)
4800 ret = err;
4801 out:
4802 kfree(sa);
4803 up_write(&root->fs_info->subvol_sem);
4804 drop_write:
4805 mnt_drop_write_file(file);
4806 return ret;
4807 }
4808
4809 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4810 {
4811 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4812 struct btrfs_ioctl_qgroup_assign_args *sa;
4813 struct btrfs_trans_handle *trans;
4814 int ret;
4815 int err;
4816
4817 if (!capable(CAP_SYS_ADMIN))
4818 return -EPERM;
4819
4820 ret = mnt_want_write_file(file);
4821 if (ret)
4822 return ret;
4823
4824 sa = memdup_user(arg, sizeof(*sa));
4825 if (IS_ERR(sa)) {
4826 ret = PTR_ERR(sa);
4827 goto drop_write;
4828 }
4829
4830 trans = btrfs_join_transaction(root);
4831 if (IS_ERR(trans)) {
4832 ret = PTR_ERR(trans);
4833 goto out;
4834 }
4835
4836 /* FIXME: check if the IDs really exist */
4837 if (sa->assign) {
4838 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4839 sa->src, sa->dst);
4840 } else {
4841 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4842 sa->src, sa->dst);
4843 }
4844
4845 /* update qgroup status and info */
4846 err = btrfs_run_qgroups(trans, root->fs_info);
4847 if (err < 0)
4848 btrfs_std_error(root->fs_info, ret,
4849 "failed to update qgroup status and info\n");
4850 err = btrfs_end_transaction(trans, root);
4851 if (err && !ret)
4852 ret = err;
4853
4854 out:
4855 kfree(sa);
4856 drop_write:
4857 mnt_drop_write_file(file);
4858 return ret;
4859 }
4860
4861 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4862 {
4863 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4864 struct btrfs_ioctl_qgroup_create_args *sa;
4865 struct btrfs_trans_handle *trans;
4866 int ret;
4867 int err;
4868
4869 if (!capable(CAP_SYS_ADMIN))
4870 return -EPERM;
4871
4872 ret = mnt_want_write_file(file);
4873 if (ret)
4874 return ret;
4875
4876 sa = memdup_user(arg, sizeof(*sa));
4877 if (IS_ERR(sa)) {
4878 ret = PTR_ERR(sa);
4879 goto drop_write;
4880 }
4881
4882 if (!sa->qgroupid) {
4883 ret = -EINVAL;
4884 goto out;
4885 }
4886
4887 trans = btrfs_join_transaction(root);
4888 if (IS_ERR(trans)) {
4889 ret = PTR_ERR(trans);
4890 goto out;
4891 }
4892
4893 /* FIXME: check if the IDs really exist */
4894 if (sa->create) {
4895 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
4896 } else {
4897 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4898 }
4899
4900 err = btrfs_end_transaction(trans, root);
4901 if (err && !ret)
4902 ret = err;
4903
4904 out:
4905 kfree(sa);
4906 drop_write:
4907 mnt_drop_write_file(file);
4908 return ret;
4909 }
4910
4911 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4912 {
4913 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4914 struct btrfs_ioctl_qgroup_limit_args *sa;
4915 struct btrfs_trans_handle *trans;
4916 int ret;
4917 int err;
4918 u64 qgroupid;
4919
4920 if (!capable(CAP_SYS_ADMIN))
4921 return -EPERM;
4922
4923 ret = mnt_want_write_file(file);
4924 if (ret)
4925 return ret;
4926
4927 sa = memdup_user(arg, sizeof(*sa));
4928 if (IS_ERR(sa)) {
4929 ret = PTR_ERR(sa);
4930 goto drop_write;
4931 }
4932
4933 trans = btrfs_join_transaction(root);
4934 if (IS_ERR(trans)) {
4935 ret = PTR_ERR(trans);
4936 goto out;
4937 }
4938
4939 qgroupid = sa->qgroupid;
4940 if (!qgroupid) {
4941 /* take the current subvol as qgroup */
4942 qgroupid = root->root_key.objectid;
4943 }
4944
4945 /* FIXME: check if the IDs really exist */
4946 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4947
4948 err = btrfs_end_transaction(trans, root);
4949 if (err && !ret)
4950 ret = err;
4951
4952 out:
4953 kfree(sa);
4954 drop_write:
4955 mnt_drop_write_file(file);
4956 return ret;
4957 }
4958
4959 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4960 {
4961 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4962 struct btrfs_ioctl_quota_rescan_args *qsa;
4963 int ret;
4964
4965 if (!capable(CAP_SYS_ADMIN))
4966 return -EPERM;
4967
4968 ret = mnt_want_write_file(file);
4969 if (ret)
4970 return ret;
4971
4972 qsa = memdup_user(arg, sizeof(*qsa));
4973 if (IS_ERR(qsa)) {
4974 ret = PTR_ERR(qsa);
4975 goto drop_write;
4976 }
4977
4978 if (qsa->flags) {
4979 ret = -EINVAL;
4980 goto out;
4981 }
4982
4983 ret = btrfs_qgroup_rescan(root->fs_info);
4984
4985 out:
4986 kfree(qsa);
4987 drop_write:
4988 mnt_drop_write_file(file);
4989 return ret;
4990 }
4991
4992 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4993 {
4994 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4995 struct btrfs_ioctl_quota_rescan_args *qsa;
4996 int ret = 0;
4997
4998 if (!capable(CAP_SYS_ADMIN))
4999 return -EPERM;
5000
5001 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
5002 if (!qsa)
5003 return -ENOMEM;
5004
5005 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5006 qsa->flags = 1;
5007 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
5008 }
5009
5010 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5011 ret = -EFAULT;
5012
5013 kfree(qsa);
5014 return ret;
5015 }
5016
5017 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5018 {
5019 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5020
5021 if (!capable(CAP_SYS_ADMIN))
5022 return -EPERM;
5023
5024 return btrfs_qgroup_wait_for_completion(root->fs_info);
5025 }
5026
5027 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5028 struct btrfs_ioctl_received_subvol_args *sa)
5029 {
5030 struct inode *inode = file_inode(file);
5031 struct btrfs_root *root = BTRFS_I(inode)->root;
5032 struct btrfs_root_item *root_item = &root->root_item;
5033 struct btrfs_trans_handle *trans;
5034 struct timespec ct = CURRENT_TIME;
5035 int ret = 0;
5036 int received_uuid_changed;
5037
5038 if (!inode_owner_or_capable(inode))
5039 return -EPERM;
5040
5041 ret = mnt_want_write_file(file);
5042 if (ret < 0)
5043 return ret;
5044
5045 down_write(&root->fs_info->subvol_sem);
5046
5047 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
5048 ret = -EINVAL;
5049 goto out;
5050 }
5051
5052 if (btrfs_root_readonly(root)) {
5053 ret = -EROFS;
5054 goto out;
5055 }
5056
5057 /*
5058 * 1 - root item
5059 * 2 - uuid items (received uuid + subvol uuid)
5060 */
5061 trans = btrfs_start_transaction(root, 3);
5062 if (IS_ERR(trans)) {
5063 ret = PTR_ERR(trans);
5064 trans = NULL;
5065 goto out;
5066 }
5067
5068 sa->rtransid = trans->transid;
5069 sa->rtime.sec = ct.tv_sec;
5070 sa->rtime.nsec = ct.tv_nsec;
5071
5072 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5073 BTRFS_UUID_SIZE);
5074 if (received_uuid_changed &&
5075 !btrfs_is_empty_uuid(root_item->received_uuid))
5076 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
5077 root_item->received_uuid,
5078 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5079 root->root_key.objectid);
5080 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5081 btrfs_set_root_stransid(root_item, sa->stransid);
5082 btrfs_set_root_rtransid(root_item, sa->rtransid);
5083 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5084 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5085 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5086 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5087
5088 ret = btrfs_update_root(trans, root->fs_info->tree_root,
5089 &root->root_key, &root->root_item);
5090 if (ret < 0) {
5091 btrfs_end_transaction(trans, root);
5092 goto out;
5093 }
5094 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5095 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
5096 sa->uuid,
5097 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5098 root->root_key.objectid);
5099 if (ret < 0 && ret != -EEXIST) {
5100 btrfs_abort_transaction(trans, root, ret);
5101 goto out;
5102 }
5103 }
5104 ret = btrfs_commit_transaction(trans, root);
5105 if (ret < 0) {
5106 btrfs_abort_transaction(trans, root, ret);
5107 goto out;
5108 }
5109
5110 out:
5111 up_write(&root->fs_info->subvol_sem);
5112 mnt_drop_write_file(file);
5113 return ret;
5114 }
5115
5116 #ifdef CONFIG_64BIT
5117 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5118 void __user *arg)
5119 {
5120 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5121 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5122 int ret = 0;
5123
5124 args32 = memdup_user(arg, sizeof(*args32));
5125 if (IS_ERR(args32)) {
5126 ret = PTR_ERR(args32);
5127 args32 = NULL;
5128 goto out;
5129 }
5130
5131 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5132 if (!args64) {
5133 ret = -ENOMEM;
5134 goto out;
5135 }
5136
5137 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5138 args64->stransid = args32->stransid;
5139 args64->rtransid = args32->rtransid;
5140 args64->stime.sec = args32->stime.sec;
5141 args64->stime.nsec = args32->stime.nsec;
5142 args64->rtime.sec = args32->rtime.sec;
5143 args64->rtime.nsec = args32->rtime.nsec;
5144 args64->flags = args32->flags;
5145
5146 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5147 if (ret)
5148 goto out;
5149
5150 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5151 args32->stransid = args64->stransid;
5152 args32->rtransid = args64->rtransid;
5153 args32->stime.sec = args64->stime.sec;
5154 args32->stime.nsec = args64->stime.nsec;
5155 args32->rtime.sec = args64->rtime.sec;
5156 args32->rtime.nsec = args64->rtime.nsec;
5157 args32->flags = args64->flags;
5158
5159 ret = copy_to_user(arg, args32, sizeof(*args32));
5160 if (ret)
5161 ret = -EFAULT;
5162
5163 out:
5164 kfree(args32);
5165 kfree(args64);
5166 return ret;
5167 }
5168 #endif
5169
5170 static long btrfs_ioctl_set_received_subvol(struct file *file,
5171 void __user *arg)
5172 {
5173 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5174 int ret = 0;
5175
5176 sa = memdup_user(arg, sizeof(*sa));
5177 if (IS_ERR(sa)) {
5178 ret = PTR_ERR(sa);
5179 sa = NULL;
5180 goto out;
5181 }
5182
5183 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5184
5185 if (ret)
5186 goto out;
5187
5188 ret = copy_to_user(arg, sa, sizeof(*sa));
5189 if (ret)
5190 ret = -EFAULT;
5191
5192 out:
5193 kfree(sa);
5194 return ret;
5195 }
5196
5197 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5198 {
5199 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5200 size_t len;
5201 int ret;
5202 char label[BTRFS_LABEL_SIZE];
5203
5204 spin_lock(&root->fs_info->super_lock);
5205 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5206 spin_unlock(&root->fs_info->super_lock);
5207
5208 len = strnlen(label, BTRFS_LABEL_SIZE);
5209
5210 if (len == BTRFS_LABEL_SIZE) {
5211 btrfs_warn(root->fs_info,
5212 "label is too long, return the first %zu bytes", --len);
5213 }
5214
5215 ret = copy_to_user(arg, label, len);
5216
5217 return ret ? -EFAULT : 0;
5218 }
5219
5220 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5221 {
5222 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5223 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5224 struct btrfs_trans_handle *trans;
5225 char label[BTRFS_LABEL_SIZE];
5226 int ret;
5227
5228 if (!capable(CAP_SYS_ADMIN))
5229 return -EPERM;
5230
5231 if (copy_from_user(label, arg, sizeof(label)))
5232 return -EFAULT;
5233
5234 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5235 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5236 BTRFS_LABEL_SIZE - 1);
5237 return -EINVAL;
5238 }
5239
5240 ret = mnt_want_write_file(file);
5241 if (ret)
5242 return ret;
5243
5244 trans = btrfs_start_transaction(root, 0);
5245 if (IS_ERR(trans)) {
5246 ret = PTR_ERR(trans);
5247 goto out_unlock;
5248 }
5249
5250 spin_lock(&root->fs_info->super_lock);
5251 strcpy(super_block->label, label);
5252 spin_unlock(&root->fs_info->super_lock);
5253 ret = btrfs_commit_transaction(trans, root);
5254
5255 out_unlock:
5256 mnt_drop_write_file(file);
5257 return ret;
5258 }
5259
5260 #define INIT_FEATURE_FLAGS(suffix) \
5261 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5262 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5263 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5264
5265 static int btrfs_ioctl_get_supported_features(struct file *file,
5266 void __user *arg)
5267 {
5268 static const struct btrfs_ioctl_feature_flags features[3] = {
5269 INIT_FEATURE_FLAGS(SUPP),
5270 INIT_FEATURE_FLAGS(SAFE_SET),
5271 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5272 };
5273
5274 if (copy_to_user(arg, &features, sizeof(features)))
5275 return -EFAULT;
5276
5277 return 0;
5278 }
5279
5280 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5281 {
5282 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5283 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5284 struct btrfs_ioctl_feature_flags features;
5285
5286 features.compat_flags = btrfs_super_compat_flags(super_block);
5287 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5288 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5289
5290 if (copy_to_user(arg, &features, sizeof(features)))
5291 return -EFAULT;
5292
5293 return 0;
5294 }
5295
5296 static int check_feature_bits(struct btrfs_root *root,
5297 enum btrfs_feature_set set,
5298 u64 change_mask, u64 flags, u64 supported_flags,
5299 u64 safe_set, u64 safe_clear)
5300 {
5301 const char *type = btrfs_feature_set_names[set];
5302 char *names;
5303 u64 disallowed, unsupported;
5304 u64 set_mask = flags & change_mask;
5305 u64 clear_mask = ~flags & change_mask;
5306
5307 unsupported = set_mask & ~supported_flags;
5308 if (unsupported) {
5309 names = btrfs_printable_features(set, unsupported);
5310 if (names) {
5311 btrfs_warn(root->fs_info,
5312 "this kernel does not support the %s feature bit%s",
5313 names, strchr(names, ',') ? "s" : "");
5314 kfree(names);
5315 } else
5316 btrfs_warn(root->fs_info,
5317 "this kernel does not support %s bits 0x%llx",
5318 type, unsupported);
5319 return -EOPNOTSUPP;
5320 }
5321
5322 disallowed = set_mask & ~safe_set;
5323 if (disallowed) {
5324 names = btrfs_printable_features(set, disallowed);
5325 if (names) {
5326 btrfs_warn(root->fs_info,
5327 "can't set the %s feature bit%s while mounted",
5328 names, strchr(names, ',') ? "s" : "");
5329 kfree(names);
5330 } else
5331 btrfs_warn(root->fs_info,
5332 "can't set %s bits 0x%llx while mounted",
5333 type, disallowed);
5334 return -EPERM;
5335 }
5336
5337 disallowed = clear_mask & ~safe_clear;
5338 if (disallowed) {
5339 names = btrfs_printable_features(set, disallowed);
5340 if (names) {
5341 btrfs_warn(root->fs_info,
5342 "can't clear the %s feature bit%s while mounted",
5343 names, strchr(names, ',') ? "s" : "");
5344 kfree(names);
5345 } else
5346 btrfs_warn(root->fs_info,
5347 "can't clear %s bits 0x%llx while mounted",
5348 type, disallowed);
5349 return -EPERM;
5350 }
5351
5352 return 0;
5353 }
5354
5355 #define check_feature(root, change_mask, flags, mask_base) \
5356 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5357 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5358 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5359 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5360
5361 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5362 {
5363 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5364 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5365 struct btrfs_ioctl_feature_flags flags[2];
5366 struct btrfs_trans_handle *trans;
5367 u64 newflags;
5368 int ret;
5369
5370 if (!capable(CAP_SYS_ADMIN))
5371 return -EPERM;
5372
5373 if (copy_from_user(flags, arg, sizeof(flags)))
5374 return -EFAULT;
5375
5376 /* Nothing to do */
5377 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5378 !flags[0].incompat_flags)
5379 return 0;
5380
5381 ret = check_feature(root, flags[0].compat_flags,
5382 flags[1].compat_flags, COMPAT);
5383 if (ret)
5384 return ret;
5385
5386 ret = check_feature(root, flags[0].compat_ro_flags,
5387 flags[1].compat_ro_flags, COMPAT_RO);
5388 if (ret)
5389 return ret;
5390
5391 ret = check_feature(root, flags[0].incompat_flags,
5392 flags[1].incompat_flags, INCOMPAT);
5393 if (ret)
5394 return ret;
5395
5396 trans = btrfs_start_transaction(root, 0);
5397 if (IS_ERR(trans))
5398 return PTR_ERR(trans);
5399
5400 spin_lock(&root->fs_info->super_lock);
5401 newflags = btrfs_super_compat_flags(super_block);
5402 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5403 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5404 btrfs_set_super_compat_flags(super_block, newflags);
5405
5406 newflags = btrfs_super_compat_ro_flags(super_block);
5407 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5408 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5409 btrfs_set_super_compat_ro_flags(super_block, newflags);
5410
5411 newflags = btrfs_super_incompat_flags(super_block);
5412 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5413 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5414 btrfs_set_super_incompat_flags(super_block, newflags);
5415 spin_unlock(&root->fs_info->super_lock);
5416
5417 return btrfs_commit_transaction(trans, root);
5418 }
5419
5420 long btrfs_ioctl(struct file *file, unsigned int
5421 cmd, unsigned long arg)
5422 {
5423 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5424 void __user *argp = (void __user *)arg;
5425
5426 switch (cmd) {
5427 case FS_IOC_GETFLAGS:
5428 return btrfs_ioctl_getflags(file, argp);
5429 case FS_IOC_SETFLAGS:
5430 return btrfs_ioctl_setflags(file, argp);
5431 case FS_IOC_GETVERSION:
5432 return btrfs_ioctl_getversion(file, argp);
5433 case FITRIM:
5434 return btrfs_ioctl_fitrim(file, argp);
5435 case BTRFS_IOC_SNAP_CREATE:
5436 return btrfs_ioctl_snap_create(file, argp, 0);
5437 case BTRFS_IOC_SNAP_CREATE_V2:
5438 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5439 case BTRFS_IOC_SUBVOL_CREATE:
5440 return btrfs_ioctl_snap_create(file, argp, 1);
5441 case BTRFS_IOC_SUBVOL_CREATE_V2:
5442 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5443 case BTRFS_IOC_SNAP_DESTROY:
5444 return btrfs_ioctl_snap_destroy(file, argp);
5445 case BTRFS_IOC_SUBVOL_GETFLAGS:
5446 return btrfs_ioctl_subvol_getflags(file, argp);
5447 case BTRFS_IOC_SUBVOL_SETFLAGS:
5448 return btrfs_ioctl_subvol_setflags(file, argp);
5449 case BTRFS_IOC_DEFAULT_SUBVOL:
5450 return btrfs_ioctl_default_subvol(file, argp);
5451 case BTRFS_IOC_DEFRAG:
5452 return btrfs_ioctl_defrag(file, NULL);
5453 case BTRFS_IOC_DEFRAG_RANGE:
5454 return btrfs_ioctl_defrag(file, argp);
5455 case BTRFS_IOC_RESIZE:
5456 return btrfs_ioctl_resize(file, argp);
5457 case BTRFS_IOC_ADD_DEV:
5458 return btrfs_ioctl_add_dev(root, argp);
5459 case BTRFS_IOC_RM_DEV:
5460 return btrfs_ioctl_rm_dev(file, argp);
5461 case BTRFS_IOC_FS_INFO:
5462 return btrfs_ioctl_fs_info(root, argp);
5463 case BTRFS_IOC_DEV_INFO:
5464 return btrfs_ioctl_dev_info(root, argp);
5465 case BTRFS_IOC_BALANCE:
5466 return btrfs_ioctl_balance(file, NULL);
5467 case BTRFS_IOC_TRANS_START:
5468 return btrfs_ioctl_trans_start(file);
5469 case BTRFS_IOC_TRANS_END:
5470 return btrfs_ioctl_trans_end(file);
5471 case BTRFS_IOC_TREE_SEARCH:
5472 return btrfs_ioctl_tree_search(file, argp);
5473 case BTRFS_IOC_TREE_SEARCH_V2:
5474 return btrfs_ioctl_tree_search_v2(file, argp);
5475 case BTRFS_IOC_INO_LOOKUP:
5476 return btrfs_ioctl_ino_lookup(file, argp);
5477 case BTRFS_IOC_INO_PATHS:
5478 return btrfs_ioctl_ino_to_path(root, argp);
5479 case BTRFS_IOC_LOGICAL_INO:
5480 return btrfs_ioctl_logical_to_ino(root, argp);
5481 case BTRFS_IOC_SPACE_INFO:
5482 return btrfs_ioctl_space_info(root, argp);
5483 case BTRFS_IOC_SYNC: {
5484 int ret;
5485
5486 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5487 if (ret)
5488 return ret;
5489 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
5490 /*
5491 * The transaction thread may want to do more work,
5492 * namely it pokes the cleaner ktread that will start
5493 * processing uncleaned subvols.
5494 */
5495 wake_up_process(root->fs_info->transaction_kthread);
5496 return ret;
5497 }
5498 case BTRFS_IOC_START_SYNC:
5499 return btrfs_ioctl_start_sync(root, argp);
5500 case BTRFS_IOC_WAIT_SYNC:
5501 return btrfs_ioctl_wait_sync(root, argp);
5502 case BTRFS_IOC_SCRUB:
5503 return btrfs_ioctl_scrub(file, argp);
5504 case BTRFS_IOC_SCRUB_CANCEL:
5505 return btrfs_ioctl_scrub_cancel(root, argp);
5506 case BTRFS_IOC_SCRUB_PROGRESS:
5507 return btrfs_ioctl_scrub_progress(root, argp);
5508 case BTRFS_IOC_BALANCE_V2:
5509 return btrfs_ioctl_balance(file, argp);
5510 case BTRFS_IOC_BALANCE_CTL:
5511 return btrfs_ioctl_balance_ctl(root, arg);
5512 case BTRFS_IOC_BALANCE_PROGRESS:
5513 return btrfs_ioctl_balance_progress(root, argp);
5514 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5515 return btrfs_ioctl_set_received_subvol(file, argp);
5516 #ifdef CONFIG_64BIT
5517 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5518 return btrfs_ioctl_set_received_subvol_32(file, argp);
5519 #endif
5520 case BTRFS_IOC_SEND:
5521 return btrfs_ioctl_send(file, argp);
5522 case BTRFS_IOC_GET_DEV_STATS:
5523 return btrfs_ioctl_get_dev_stats(root, argp);
5524 case BTRFS_IOC_QUOTA_CTL:
5525 return btrfs_ioctl_quota_ctl(file, argp);
5526 case BTRFS_IOC_QGROUP_ASSIGN:
5527 return btrfs_ioctl_qgroup_assign(file, argp);
5528 case BTRFS_IOC_QGROUP_CREATE:
5529 return btrfs_ioctl_qgroup_create(file, argp);
5530 case BTRFS_IOC_QGROUP_LIMIT:
5531 return btrfs_ioctl_qgroup_limit(file, argp);
5532 case BTRFS_IOC_QUOTA_RESCAN:
5533 return btrfs_ioctl_quota_rescan(file, argp);
5534 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5535 return btrfs_ioctl_quota_rescan_status(file, argp);
5536 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5537 return btrfs_ioctl_quota_rescan_wait(file, argp);
5538 case BTRFS_IOC_DEV_REPLACE:
5539 return btrfs_ioctl_dev_replace(root, argp);
5540 case BTRFS_IOC_GET_FSLABEL:
5541 return btrfs_ioctl_get_fslabel(file, argp);
5542 case BTRFS_IOC_SET_FSLABEL:
5543 return btrfs_ioctl_set_fslabel(file, argp);
5544 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5545 return btrfs_ioctl_get_supported_features(file, argp);
5546 case BTRFS_IOC_GET_FEATURES:
5547 return btrfs_ioctl_get_features(file, argp);
5548 case BTRFS_IOC_SET_FEATURES:
5549 return btrfs_ioctl_set_features(file, argp);
5550 }
5551
5552 return -ENOTTY;
5553 }
Linux kernel - Deliverables
This page took 0.150157 seconds and 5 git commands to generate.