projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
btrfs: qgroup: Avoid calling btrfs_free_reserved_data_space in clear_bit_hook
[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 mutex_lock(&inode->i_mutex);
244
245 ip_oldflags = ip->flags;
246 i_oldflags = inode->i_flags;
247 mode = inode->i_mode;
248
249 flags = btrfs_mask_flags(inode->i_mode, flags);
250 oldflags = btrfs_flags_to_ioctl(ip->flags);
251 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
252 if (!capable(CAP_LINUX_IMMUTABLE)) {
253 ret = -EPERM;
254 goto out_unlock;
255 }
256 }
257
258 if (flags & FS_SYNC_FL)
259 ip->flags |= BTRFS_INODE_SYNC;
260 else
261 ip->flags &= ~BTRFS_INODE_SYNC;
262 if (flags & FS_IMMUTABLE_FL)
263 ip->flags |= BTRFS_INODE_IMMUTABLE;
264 else
265 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
266 if (flags & FS_APPEND_FL)
267 ip->flags |= BTRFS_INODE_APPEND;
268 else
269 ip->flags &= ~BTRFS_INODE_APPEND;
270 if (flags & FS_NODUMP_FL)
271 ip->flags |= BTRFS_INODE_NODUMP;
272 else
273 ip->flags &= ~BTRFS_INODE_NODUMP;
274 if (flags & FS_NOATIME_FL)
275 ip->flags |= BTRFS_INODE_NOATIME;
276 else
277 ip->flags &= ~BTRFS_INODE_NOATIME;
278 if (flags & FS_DIRSYNC_FL)
279 ip->flags |= BTRFS_INODE_DIRSYNC;
280 else
281 ip->flags &= ~BTRFS_INODE_DIRSYNC;
282 if (flags & FS_NOCOW_FL) {
283 if (S_ISREG(mode)) {
284 /*
285 * It's safe to turn csums off here, no extents exist.
286 * Otherwise we want the flag to reflect the real COW
287 * status of the file and will not set it.
288 */
289 if (inode->i_size == 0)
290 ip->flags |= BTRFS_INODE_NODATACOW
291 | BTRFS_INODE_NODATASUM;
292 } else {
293 ip->flags |= BTRFS_INODE_NODATACOW;
294 }
295 } else {
296 /*
297 * Revert back under same assuptions as above
298 */
299 if (S_ISREG(mode)) {
300 if (inode->i_size == 0)
301 ip->flags &= ~(BTRFS_INODE_NODATACOW
302 | BTRFS_INODE_NODATASUM);
303 } else {
304 ip->flags &= ~BTRFS_INODE_NODATACOW;
305 }
306 }
307
308 /*
309 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
310 * flag may be changed automatically if compression code won't make
311 * things smaller.
312 */
313 if (flags & FS_NOCOMP_FL) {
314 ip->flags &= ~BTRFS_INODE_COMPRESS;
315 ip->flags |= BTRFS_INODE_NOCOMPRESS;
316
317 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
318 if (ret && ret != -ENODATA)
319 goto out_drop;
320 } else if (flags & FS_COMPR_FL) {
321 const char *comp;
322
323 ip->flags |= BTRFS_INODE_COMPRESS;
324 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
325
326 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
327 comp = "lzo";
328 else
329 comp = "zlib";
330 ret = btrfs_set_prop(inode, "btrfs.compression",
331 comp, strlen(comp), 0);
332 if (ret)
333 goto out_drop;
334
335 } else {
336 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
337 if (ret && ret != -ENODATA)
338 goto out_drop;
339 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
340 }
341
342 trans = btrfs_start_transaction(root, 1);
343 if (IS_ERR(trans)) {
344 ret = PTR_ERR(trans);
345 goto out_drop;
346 }
347
348 btrfs_update_iflags(inode);
349 inode_inc_iversion(inode);
350 inode->i_ctime = CURRENT_TIME;
351 ret = btrfs_update_inode(trans, root, inode);
352
353 btrfs_end_transaction(trans, root);
354 out_drop:
355 if (ret) {
356 ip->flags = ip_oldflags;
357 inode->i_flags = i_oldflags;
358 }
359
360 out_unlock:
361 mutex_unlock(&inode->i_mutex);
362 mnt_drop_write_file(file);
363 return ret;
364 }
365
366 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
367 {
368 struct inode *inode = file_inode(file);
369
370 return put_user(inode->i_generation, arg);
371 }
372
373 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
374 {
375 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
376 struct btrfs_device *device;
377 struct request_queue *q;
378 struct fstrim_range range;
379 u64 minlen = ULLONG_MAX;
380 u64 num_devices = 0;
381 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
382 int ret;
383
384 if (!capable(CAP_SYS_ADMIN))
385 return -EPERM;
386
387 rcu_read_lock();
388 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
389 dev_list) {
390 if (!device->bdev)
391 continue;
392 q = bdev_get_queue(device->bdev);
393 if (blk_queue_discard(q)) {
394 num_devices++;
395 minlen = min((u64)q->limits.discard_granularity,
396 minlen);
397 }
398 }
399 rcu_read_unlock();
400
401 if (!num_devices)
402 return -EOPNOTSUPP;
403 if (copy_from_user(&range, arg, sizeof(range)))
404 return -EFAULT;
405 if (range.start > total_bytes ||
406 range.len < fs_info->sb->s_blocksize)
407 return -EINVAL;
408
409 range.len = min(range.len, total_bytes - range.start);
410 range.minlen = max(range.minlen, minlen);
411 ret = btrfs_trim_fs(fs_info->tree_root, &range);
412 if (ret < 0)
413 return ret;
414
415 if (copy_to_user(arg, &range, sizeof(range)))
416 return -EFAULT;
417
418 return 0;
419 }
420
421 int btrfs_is_empty_uuid(u8 *uuid)
422 {
423 int i;
424
425 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
426 if (uuid[i])
427 return 0;
428 }
429 return 1;
430 }
431
432 static noinline int create_subvol(struct inode *dir,
433 struct dentry *dentry,
434 char *name, int namelen,
435 u64 *async_transid,
436 struct btrfs_qgroup_inherit *inherit)
437 {
438 struct btrfs_trans_handle *trans;
439 struct btrfs_key key;
440 struct btrfs_root_item root_item;
441 struct btrfs_inode_item *inode_item;
442 struct extent_buffer *leaf;
443 struct btrfs_root *root = BTRFS_I(dir)->root;
444 struct btrfs_root *new_root;
445 struct btrfs_block_rsv block_rsv;
446 struct timespec cur_time = CURRENT_TIME;
447 struct inode *inode;
448 int ret;
449 int err;
450 u64 objectid;
451 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
452 u64 index = 0;
453 u64 qgroup_reserved;
454 uuid_le new_uuid;
455
456 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
457 if (ret)
458 return ret;
459
460 /*
461 * Don't create subvolume whose level is not zero. Or qgroup will be
462 * screwed up since it assume subvolme qgroup's level to be 0.
463 */
464 if (btrfs_qgroup_level(objectid))
465 return -ENOSPC;
466
467 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
468 /*
469 * The same as the snapshot creation, please see the comment
470 * of create_snapshot().
471 */
472 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
473 8, &qgroup_reserved, false);
474 if (ret)
475 return ret;
476
477 trans = btrfs_start_transaction(root, 0);
478 if (IS_ERR(trans)) {
479 ret = PTR_ERR(trans);
480 btrfs_subvolume_release_metadata(root, &block_rsv,
481 qgroup_reserved);
482 return ret;
483 }
484 trans->block_rsv = &block_rsv;
485 trans->bytes_reserved = block_rsv.size;
486
487 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
488 if (ret)
489 goto fail;
490
491 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
492 if (IS_ERR(leaf)) {
493 ret = PTR_ERR(leaf);
494 goto fail;
495 }
496
497 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
498 btrfs_set_header_bytenr(leaf, leaf->start);
499 btrfs_set_header_generation(leaf, trans->transid);
500 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
501 btrfs_set_header_owner(leaf, objectid);
502
503 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
504 BTRFS_FSID_SIZE);
505 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
506 btrfs_header_chunk_tree_uuid(leaf),
507 BTRFS_UUID_SIZE);
508 btrfs_mark_buffer_dirty(leaf);
509
510 memset(&root_item, 0, sizeof(root_item));
511
512 inode_item = &root_item.inode;
513 btrfs_set_stack_inode_generation(inode_item, 1);
514 btrfs_set_stack_inode_size(inode_item, 3);
515 btrfs_set_stack_inode_nlink(inode_item, 1);
516 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
517 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
518
519 btrfs_set_root_flags(&root_item, 0);
520 btrfs_set_root_limit(&root_item, 0);
521 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
522
523 btrfs_set_root_bytenr(&root_item, leaf->start);
524 btrfs_set_root_generation(&root_item, trans->transid);
525 btrfs_set_root_level(&root_item, 0);
526 btrfs_set_root_refs(&root_item, 1);
527 btrfs_set_root_used(&root_item, leaf->len);
528 btrfs_set_root_last_snapshot(&root_item, 0);
529
530 btrfs_set_root_generation_v2(&root_item,
531 btrfs_root_generation(&root_item));
532 uuid_le_gen(&new_uuid);
533 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
534 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
535 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
536 root_item.ctime = root_item.otime;
537 btrfs_set_root_ctransid(&root_item, trans->transid);
538 btrfs_set_root_otransid(&root_item, trans->transid);
539
540 btrfs_tree_unlock(leaf);
541 free_extent_buffer(leaf);
542 leaf = NULL;
543
544 btrfs_set_root_dirid(&root_item, new_dirid);
545
546 key.objectid = objectid;
547 key.offset = 0;
548 key.type = BTRFS_ROOT_ITEM_KEY;
549 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
550 &root_item);
551 if (ret)
552 goto fail;
553
554 key.offset = (u64)-1;
555 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
556 if (IS_ERR(new_root)) {
557 ret = PTR_ERR(new_root);
558 btrfs_abort_transaction(trans, root, ret);
559 goto fail;
560 }
561
562 btrfs_record_root_in_trans(trans, new_root);
563
564 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
565 if (ret) {
566 /* We potentially lose an unused inode item here */
567 btrfs_abort_transaction(trans, root, ret);
568 goto fail;
569 }
570
571 /*
572 * insert the directory item
573 */
574 ret = btrfs_set_inode_index(dir, &index);
575 if (ret) {
576 btrfs_abort_transaction(trans, root, ret);
577 goto fail;
578 }
579
580 ret = btrfs_insert_dir_item(trans, root,
581 name, namelen, dir, &key,
582 BTRFS_FT_DIR, index);
583 if (ret) {
584 btrfs_abort_transaction(trans, root, ret);
585 goto fail;
586 }
587
588 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
589 ret = btrfs_update_inode(trans, root, dir);
590 BUG_ON(ret);
591
592 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
593 objectid, root->root_key.objectid,
594 btrfs_ino(dir), index, name, namelen);
595 BUG_ON(ret);
596
597 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
598 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
599 objectid);
600 if (ret)
601 btrfs_abort_transaction(trans, root, ret);
602
603 fail:
604 trans->block_rsv = NULL;
605 trans->bytes_reserved = 0;
606 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
607
608 if (async_transid) {
609 *async_transid = trans->transid;
610 err = btrfs_commit_transaction_async(trans, root, 1);
611 if (err)
612 err = btrfs_commit_transaction(trans, root);
613 } else {
614 err = btrfs_commit_transaction(trans, root);
615 }
616 if (err && !ret)
617 ret = err;
618
619 if (!ret) {
620 inode = btrfs_lookup_dentry(dir, dentry);
621 if (IS_ERR(inode))
622 return PTR_ERR(inode);
623 d_instantiate(dentry, inode);
624 }
625 return ret;
626 }
627
628 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
629 {
630 s64 writers;
631 DEFINE_WAIT(wait);
632
633 do {
634 prepare_to_wait(&root->subv_writers->wait, &wait,
635 TASK_UNINTERRUPTIBLE);
636
637 writers = percpu_counter_sum(&root->subv_writers->counter);
638 if (writers)
639 schedule();
640
641 finish_wait(&root->subv_writers->wait, &wait);
642 } while (writers);
643 }
644
645 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
646 struct dentry *dentry, char *name, int namelen,
647 u64 *async_transid, bool readonly,
648 struct btrfs_qgroup_inherit *inherit)
649 {
650 struct inode *inode;
651 struct btrfs_pending_snapshot *pending_snapshot;
652 struct btrfs_trans_handle *trans;
653 int ret;
654
655 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
656 return -EINVAL;
657
658 atomic_inc(&root->will_be_snapshoted);
659 smp_mb__after_atomic();
660 btrfs_wait_for_no_snapshoting_writes(root);
661
662 ret = btrfs_start_delalloc_inodes(root, 0);
663 if (ret)
664 goto out;
665
666 btrfs_wait_ordered_extents(root, -1);
667
668 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
669 if (!pending_snapshot) {
670 ret = -ENOMEM;
671 goto out;
672 }
673
674 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
675 BTRFS_BLOCK_RSV_TEMP);
676 /*
677 * 1 - parent dir inode
678 * 2 - dir entries
679 * 1 - root item
680 * 2 - root ref/backref
681 * 1 - root of snapshot
682 * 1 - UUID item
683 */
684 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
685 &pending_snapshot->block_rsv, 8,
686 &pending_snapshot->qgroup_reserved,
687 false);
688 if (ret)
689 goto free;
690
691 pending_snapshot->dentry = dentry;
692 pending_snapshot->root = root;
693 pending_snapshot->readonly = readonly;
694 pending_snapshot->dir = dir;
695 pending_snapshot->inherit = inherit;
696
697 trans = btrfs_start_transaction(root, 0);
698 if (IS_ERR(trans)) {
699 ret = PTR_ERR(trans);
700 goto fail;
701 }
702
703 spin_lock(&root->fs_info->trans_lock);
704 list_add(&pending_snapshot->list,
705 &trans->transaction->pending_snapshots);
706 spin_unlock(&root->fs_info->trans_lock);
707 if (async_transid) {
708 *async_transid = trans->transid;
709 ret = btrfs_commit_transaction_async(trans,
710 root->fs_info->extent_root, 1);
711 if (ret)
712 ret = btrfs_commit_transaction(trans, root);
713 } else {
714 ret = btrfs_commit_transaction(trans,
715 root->fs_info->extent_root);
716 }
717 if (ret)
718 goto fail;
719
720 ret = pending_snapshot->error;
721 if (ret)
722 goto fail;
723
724 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
725 if (ret)
726 goto fail;
727
728 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
729 if (IS_ERR(inode)) {
730 ret = PTR_ERR(inode);
731 goto fail;
732 }
733
734 d_instantiate(dentry, inode);
735 ret = 0;
736 fail:
737 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
738 &pending_snapshot->block_rsv,
739 pending_snapshot->qgroup_reserved);
740 free:
741 kfree(pending_snapshot);
742 out:
743 if (atomic_dec_and_test(&root->will_be_snapshoted))
744 wake_up_atomic_t(&root->will_be_snapshoted);
745 return ret;
746 }
747
748 /* copy of may_delete in fs/namei.c()
749 * Check whether we can remove a link victim from directory dir, check
750 * whether the type of victim is right.
751 * 1. We can't do it if dir is read-only (done in permission())
752 * 2. We should have write and exec permissions on dir
753 * 3. We can't remove anything from append-only dir
754 * 4. We can't do anything with immutable dir (done in permission())
755 * 5. If the sticky bit on dir is set we should either
756 * a. be owner of dir, or
757 * b. be owner of victim, or
758 * c. have CAP_FOWNER capability
759 * 6. If the victim is append-only or immutable we can't do antyhing with
760 * links pointing to it.
761 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
762 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
763 * 9. We can't remove a root or mountpoint.
764 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
765 * nfs_async_unlink().
766 */
767
768 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
769 {
770 int error;
771
772 if (d_really_is_negative(victim))
773 return -ENOENT;
774
775 BUG_ON(d_inode(victim->d_parent) != dir);
776 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
777
778 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
779 if (error)
780 return error;
781 if (IS_APPEND(dir))
782 return -EPERM;
783 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
784 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
785 return -EPERM;
786 if (isdir) {
787 if (!d_is_dir(victim))
788 return -ENOTDIR;
789 if (IS_ROOT(victim))
790 return -EBUSY;
791 } else if (d_is_dir(victim))
792 return -EISDIR;
793 if (IS_DEADDIR(dir))
794 return -ENOENT;
795 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
796 return -EBUSY;
797 return 0;
798 }
799
800 /* copy of may_create in fs/namei.c() */
801 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
802 {
803 if (d_really_is_positive(child))
804 return -EEXIST;
805 if (IS_DEADDIR(dir))
806 return -ENOENT;
807 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
808 }
809
810 /*
811 * Create a new subvolume below @parent. This is largely modeled after
812 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
813 * inside this filesystem so it's quite a bit simpler.
814 */
815 static noinline int btrfs_mksubvol(struct path *parent,
816 char *name, int namelen,
817 struct btrfs_root *snap_src,
818 u64 *async_transid, bool readonly,
819 struct btrfs_qgroup_inherit *inherit)
820 {
821 struct inode *dir = d_inode(parent->dentry);
822 struct dentry *dentry;
823 int error;
824
825 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
826 if (error == -EINTR)
827 return error;
828
829 dentry = lookup_one_len(name, parent->dentry, namelen);
830 error = PTR_ERR(dentry);
831 if (IS_ERR(dentry))
832 goto out_unlock;
833
834 error = -EEXIST;
835 if (d_really_is_positive(dentry))
836 goto out_dput;
837
838 error = btrfs_may_create(dir, dentry);
839 if (error)
840 goto out_dput;
841
842 /*
843 * even if this name doesn't exist, we may get hash collisions.
844 * check for them now when we can safely fail
845 */
846 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
847 dir->i_ino, name,
848 namelen);
849 if (error)
850 goto out_dput;
851
852 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
853
854 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
855 goto out_up_read;
856
857 if (snap_src) {
858 error = create_snapshot(snap_src, dir, dentry, name, namelen,
859 async_transid, readonly, inherit);
860 } else {
861 error = create_subvol(dir, dentry, name, namelen,
862 async_transid, inherit);
863 }
864 if (!error)
865 fsnotify_mkdir(dir, dentry);
866 out_up_read:
867 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
868 out_dput:
869 dput(dentry);
870 out_unlock:
871 mutex_unlock(&dir->i_mutex);
872 return error;
873 }
874
875 /*
876 * When we're defragging a range, we don't want to kick it off again
877 * if it is really just waiting for delalloc to send it down.
878 * If we find a nice big extent or delalloc range for the bytes in the
879 * file you want to defrag, we return 0 to let you know to skip this
880 * part of the file
881 */
882 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
883 {
884 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
885 struct extent_map *em = NULL;
886 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
887 u64 end;
888
889 read_lock(&em_tree->lock);
890 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
891 read_unlock(&em_tree->lock);
892
893 if (em) {
894 end = extent_map_end(em);
895 free_extent_map(em);
896 if (end - offset > thresh)
897 return 0;
898 }
899 /* if we already have a nice delalloc here, just stop */
900 thresh /= 2;
901 end = count_range_bits(io_tree, &offset, offset + thresh,
902 thresh, EXTENT_DELALLOC, 1);
903 if (end >= thresh)
904 return 0;
905 return 1;
906 }
907
908 /*
909 * helper function to walk through a file and find extents
910 * newer than a specific transid, and smaller than thresh.
911 *
912 * This is used by the defragging code to find new and small
913 * extents
914 */
915 static int find_new_extents(struct btrfs_root *root,
916 struct inode *inode, u64 newer_than,
917 u64 *off, u32 thresh)
918 {
919 struct btrfs_path *path;
920 struct btrfs_key min_key;
921 struct extent_buffer *leaf;
922 struct btrfs_file_extent_item *extent;
923 int type;
924 int ret;
925 u64 ino = btrfs_ino(inode);
926
927 path = btrfs_alloc_path();
928 if (!path)
929 return -ENOMEM;
930
931 min_key.objectid = ino;
932 min_key.type = BTRFS_EXTENT_DATA_KEY;
933 min_key.offset = *off;
934
935 while (1) {
936 ret = btrfs_search_forward(root, &min_key, path, newer_than);
937 if (ret != 0)
938 goto none;
939 process_slot:
940 if (min_key.objectid != ino)
941 goto none;
942 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
943 goto none;
944
945 leaf = path->nodes[0];
946 extent = btrfs_item_ptr(leaf, path->slots[0],
947 struct btrfs_file_extent_item);
948
949 type = btrfs_file_extent_type(leaf, extent);
950 if (type == BTRFS_FILE_EXTENT_REG &&
951 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
952 check_defrag_in_cache(inode, min_key.offset, thresh)) {
953 *off = min_key.offset;
954 btrfs_free_path(path);
955 return 0;
956 }
957
958 path->slots[0]++;
959 if (path->slots[0] < btrfs_header_nritems(leaf)) {
960 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
961 goto process_slot;
962 }
963
964 if (min_key.offset == (u64)-1)
965 goto none;
966
967 min_key.offset++;
968 btrfs_release_path(path);
969 }
970 none:
971 btrfs_free_path(path);
972 return -ENOENT;
973 }
974
975 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
976 {
977 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
978 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
979 struct extent_map *em;
980 u64 len = PAGE_CACHE_SIZE;
981
982 /*
983 * hopefully we have this extent in the tree already, try without
984 * the full extent lock
985 */
986 read_lock(&em_tree->lock);
987 em = lookup_extent_mapping(em_tree, start, len);
988 read_unlock(&em_tree->lock);
989
990 if (!em) {
991 struct extent_state *cached = NULL;
992 u64 end = start + len - 1;
993
994 /* get the big lock and read metadata off disk */
995 lock_extent_bits(io_tree, start, end, 0, &cached);
996 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
997 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
998
999 if (IS_ERR(em))
1000 return NULL;
1001 }
1002
1003 return em;
1004 }
1005
1006 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1007 {
1008 struct extent_map *next;
1009 bool ret = true;
1010
1011 /* this is the last extent */
1012 if (em->start + em->len >= i_size_read(inode))
1013 return false;
1014
1015 next = defrag_lookup_extent(inode, em->start + em->len);
1016 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1017 ret = false;
1018 else if ((em->block_start + em->block_len == next->block_start) &&
1019 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
1020 ret = false;
1021
1022 free_extent_map(next);
1023 return ret;
1024 }
1025
1026 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1027 u64 *last_len, u64 *skip, u64 *defrag_end,
1028 int compress)
1029 {
1030 struct extent_map *em;
1031 int ret = 1;
1032 bool next_mergeable = true;
1033 bool prev_mergeable = true;
1034
1035 /*
1036 * make sure that once we start defragging an extent, we keep on
1037 * defragging it
1038 */
1039 if (start < *defrag_end)
1040 return 1;
1041
1042 *skip = 0;
1043
1044 em = defrag_lookup_extent(inode, start);
1045 if (!em)
1046 return 0;
1047
1048 /* this will cover holes, and inline extents */
1049 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1050 ret = 0;
1051 goto out;
1052 }
1053
1054 if (!*defrag_end)
1055 prev_mergeable = false;
1056
1057 next_mergeable = defrag_check_next_extent(inode, em);
1058 /*
1059 * we hit a real extent, if it is big or the next extent is not a
1060 * real extent, don't bother defragging it
1061 */
1062 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1063 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1064 ret = 0;
1065 out:
1066 /*
1067 * last_len ends up being a counter of how many bytes we've defragged.
1068 * every time we choose not to defrag an extent, we reset *last_len
1069 * so that the next tiny extent will force a defrag.
1070 *
1071 * The end result of this is that tiny extents before a single big
1072 * extent will force at least part of that big extent to be defragged.
1073 */
1074 if (ret) {
1075 *defrag_end = extent_map_end(em);
1076 } else {
1077 *last_len = 0;
1078 *skip = extent_map_end(em);
1079 *defrag_end = 0;
1080 }
1081
1082 free_extent_map(em);
1083 return ret;
1084 }
1085
1086 /*
1087 * it doesn't do much good to defrag one or two pages
1088 * at a time. This pulls in a nice chunk of pages
1089 * to COW and defrag.
1090 *
1091 * It also makes sure the delalloc code has enough
1092 * dirty data to avoid making new small extents as part
1093 * of the defrag
1094 *
1095 * It's a good idea to start RA on this range
1096 * before calling this.
1097 */
1098 static int cluster_pages_for_defrag(struct inode *inode,
1099 struct page **pages,
1100 unsigned long start_index,
1101 unsigned long num_pages)
1102 {
1103 unsigned long file_end;
1104 u64 isize = i_size_read(inode);
1105 u64 page_start;
1106 u64 page_end;
1107 u64 page_cnt;
1108 int ret;
1109 int i;
1110 int i_done;
1111 struct btrfs_ordered_extent *ordered;
1112 struct extent_state *cached_state = NULL;
1113 struct extent_io_tree *tree;
1114 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1115
1116 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1117 if (!isize || start_index > file_end)
1118 return 0;
1119
1120 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1121
1122 ret = btrfs_delalloc_reserve_space(inode,
1123 start_index << PAGE_CACHE_SHIFT,
1124 page_cnt << PAGE_CACHE_SHIFT);
1125 if (ret)
1126 return ret;
1127 i_done = 0;
1128 tree = &BTRFS_I(inode)->io_tree;
1129
1130 /* step one, lock all the pages */
1131 for (i = 0; i < page_cnt; i++) {
1132 struct page *page;
1133 again:
1134 page = find_or_create_page(inode->i_mapping,
1135 start_index + i, mask);
1136 if (!page)
1137 break;
1138
1139 page_start = page_offset(page);
1140 page_end = page_start + PAGE_CACHE_SIZE - 1;
1141 while (1) {
1142 lock_extent_bits(tree, page_start, page_end,
1143 0, &cached_state);
1144 ordered = btrfs_lookup_ordered_extent(inode,
1145 page_start);
1146 unlock_extent_cached(tree, page_start, page_end,
1147 &cached_state, GFP_NOFS);
1148 if (!ordered)
1149 break;
1150
1151 unlock_page(page);
1152 btrfs_start_ordered_extent(inode, ordered, 1);
1153 btrfs_put_ordered_extent(ordered);
1154 lock_page(page);
1155 /*
1156 * we unlocked the page above, so we need check if
1157 * it was released or not.
1158 */
1159 if (page->mapping != inode->i_mapping) {
1160 unlock_page(page);
1161 page_cache_release(page);
1162 goto again;
1163 }
1164 }
1165
1166 if (!PageUptodate(page)) {
1167 btrfs_readpage(NULL, page);
1168 lock_page(page);
1169 if (!PageUptodate(page)) {
1170 unlock_page(page);
1171 page_cache_release(page);
1172 ret = -EIO;
1173 break;
1174 }
1175 }
1176
1177 if (page->mapping != inode->i_mapping) {
1178 unlock_page(page);
1179 page_cache_release(page);
1180 goto again;
1181 }
1182
1183 pages[i] = page;
1184 i_done++;
1185 }
1186 if (!i_done || ret)
1187 goto out;
1188
1189 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1190 goto out;
1191
1192 /*
1193 * so now we have a nice long stream of locked
1194 * and up to date pages, lets wait on them
1195 */
1196 for (i = 0; i < i_done; i++)
1197 wait_on_page_writeback(pages[i]);
1198
1199 page_start = page_offset(pages[0]);
1200 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1201
1202 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1203 page_start, page_end - 1, 0, &cached_state);
1204 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1205 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1206 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1207 &cached_state, GFP_NOFS);
1208
1209 if (i_done != page_cnt) {
1210 spin_lock(&BTRFS_I(inode)->lock);
1211 BTRFS_I(inode)->outstanding_extents++;
1212 spin_unlock(&BTRFS_I(inode)->lock);
1213 btrfs_delalloc_release_space(inode,
1214 start_index << PAGE_CACHE_SHIFT,
1215 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1216 }
1217
1218
1219 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1220 &cached_state, GFP_NOFS);
1221
1222 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1223 page_start, page_end - 1, &cached_state,
1224 GFP_NOFS);
1225
1226 for (i = 0; i < i_done; i++) {
1227 clear_page_dirty_for_io(pages[i]);
1228 ClearPageChecked(pages[i]);
1229 set_page_extent_mapped(pages[i]);
1230 set_page_dirty(pages[i]);
1231 unlock_page(pages[i]);
1232 page_cache_release(pages[i]);
1233 }
1234 return i_done;
1235 out:
1236 for (i = 0; i < i_done; i++) {
1237 unlock_page(pages[i]);
1238 page_cache_release(pages[i]);
1239 }
1240 btrfs_delalloc_release_space(inode,
1241 start_index << PAGE_CACHE_SHIFT,
1242 page_cnt << PAGE_CACHE_SHIFT);
1243 return ret;
1244
1245 }
1246
1247 int btrfs_defrag_file(struct inode *inode, struct file *file,
1248 struct btrfs_ioctl_defrag_range_args *range,
1249 u64 newer_than, unsigned long max_to_defrag)
1250 {
1251 struct btrfs_root *root = BTRFS_I(inode)->root;
1252 struct file_ra_state *ra = NULL;
1253 unsigned long last_index;
1254 u64 isize = i_size_read(inode);
1255 u64 last_len = 0;
1256 u64 skip = 0;
1257 u64 defrag_end = 0;
1258 u64 newer_off = range->start;
1259 unsigned long i;
1260 unsigned long ra_index = 0;
1261 int ret;
1262 int defrag_count = 0;
1263 int compress_type = BTRFS_COMPRESS_ZLIB;
1264 u32 extent_thresh = range->extent_thresh;
1265 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1266 unsigned long cluster = max_cluster;
1267 u64 new_align = ~((u64)128 * 1024 - 1);
1268 struct page **pages = NULL;
1269
1270 if (isize == 0)
1271 return 0;
1272
1273 if (range->start >= isize)
1274 return -EINVAL;
1275
1276 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1277 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1278 return -EINVAL;
1279 if (range->compress_type)
1280 compress_type = range->compress_type;
1281 }
1282
1283 if (extent_thresh == 0)
1284 extent_thresh = 256 * 1024;
1285
1286 /*
1287 * if we were not given a file, allocate a readahead
1288 * context
1289 */
1290 if (!file) {
1291 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1292 if (!ra)
1293 return -ENOMEM;
1294 file_ra_state_init(ra, inode->i_mapping);
1295 } else {
1296 ra = &file->f_ra;
1297 }
1298
1299 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1300 GFP_NOFS);
1301 if (!pages) {
1302 ret = -ENOMEM;
1303 goto out_ra;
1304 }
1305
1306 /* find the last page to defrag */
1307 if (range->start + range->len > range->start) {
1308 last_index = min_t(u64, isize - 1,
1309 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1310 } else {
1311 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1312 }
1313
1314 if (newer_than) {
1315 ret = find_new_extents(root, inode, newer_than,
1316 &newer_off, 64 * 1024);
1317 if (!ret) {
1318 range->start = newer_off;
1319 /*
1320 * we always align our defrag to help keep
1321 * the extents in the file evenly spaced
1322 */
1323 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1324 } else
1325 goto out_ra;
1326 } else {
1327 i = range->start >> PAGE_CACHE_SHIFT;
1328 }
1329 if (!max_to_defrag)
1330 max_to_defrag = last_index - i + 1;
1331
1332 /*
1333 * make writeback starts from i, so the defrag range can be
1334 * written sequentially.
1335 */
1336 if (i < inode->i_mapping->writeback_index)
1337 inode->i_mapping->writeback_index = i;
1338
1339 while (i <= last_index && defrag_count < max_to_defrag &&
1340 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
1341 /*
1342 * make sure we stop running if someone unmounts
1343 * the FS
1344 */
1345 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1346 break;
1347
1348 if (btrfs_defrag_cancelled(root->fs_info)) {
1349 btrfs_debug(root->fs_info, "defrag_file cancelled");
1350 ret = -EAGAIN;
1351 break;
1352 }
1353
1354 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1355 extent_thresh, &last_len, &skip,
1356 &defrag_end, range->flags &
1357 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1358 unsigned long next;
1359 /*
1360 * the should_defrag function tells us how much to skip
1361 * bump our counter by the suggested amount
1362 */
1363 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
1364 i = max(i + 1, next);
1365 continue;
1366 }
1367
1368 if (!newer_than) {
1369 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1370 PAGE_CACHE_SHIFT) - i;
1371 cluster = min(cluster, max_cluster);
1372 } else {
1373 cluster = max_cluster;
1374 }
1375
1376 if (i + cluster > ra_index) {
1377 ra_index = max(i, ra_index);
1378 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1379 cluster);
1380 ra_index += cluster;
1381 }
1382
1383 mutex_lock(&inode->i_mutex);
1384 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1385 BTRFS_I(inode)->force_compress = compress_type;
1386 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1387 if (ret < 0) {
1388 mutex_unlock(&inode->i_mutex);
1389 goto out_ra;
1390 }
1391
1392 defrag_count += ret;
1393 balance_dirty_pages_ratelimited(inode->i_mapping);
1394 mutex_unlock(&inode->i_mutex);
1395
1396 if (newer_than) {
1397 if (newer_off == (u64)-1)
1398 break;
1399
1400 if (ret > 0)
1401 i += ret;
1402
1403 newer_off = max(newer_off + 1,
1404 (u64)i << PAGE_CACHE_SHIFT);
1405
1406 ret = find_new_extents(root, inode,
1407 newer_than, &newer_off,
1408 64 * 1024);
1409 if (!ret) {
1410 range->start = newer_off;
1411 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1412 } else {
1413 break;
1414 }
1415 } else {
1416 if (ret > 0) {
1417 i += ret;
1418 last_len += ret << PAGE_CACHE_SHIFT;
1419 } else {
1420 i++;
1421 last_len = 0;
1422 }
1423 }
1424 }
1425
1426 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1427 filemap_flush(inode->i_mapping);
1428 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1429 &BTRFS_I(inode)->runtime_flags))
1430 filemap_flush(inode->i_mapping);
1431 }
1432
1433 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1434 /* the filemap_flush will queue IO into the worker threads, but
1435 * we have to make sure the IO is actually started and that
1436 * ordered extents get created before we return
1437 */
1438 atomic_inc(&root->fs_info->async_submit_draining);
1439 while (atomic_read(&root->fs_info->nr_async_submits) ||
1440 atomic_read(&root->fs_info->async_delalloc_pages)) {
1441 wait_event(root->fs_info->async_submit_wait,
1442 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1443 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1444 }
1445 atomic_dec(&root->fs_info->async_submit_draining);
1446 }
1447
1448 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1449 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1450 }
1451
1452 ret = defrag_count;
1453
1454 out_ra:
1455 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1456 mutex_lock(&inode->i_mutex);
1457 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1458 mutex_unlock(&inode->i_mutex);
1459 }
1460 if (!file)
1461 kfree(ra);
1462 kfree(pages);
1463 return ret;
1464 }
1465
1466 static noinline int btrfs_ioctl_resize(struct file *file,
1467 void __user *arg)
1468 {
1469 u64 new_size;
1470 u64 old_size;
1471 u64 devid = 1;
1472 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1473 struct btrfs_ioctl_vol_args *vol_args;
1474 struct btrfs_trans_handle *trans;
1475 struct btrfs_device *device = NULL;
1476 char *sizestr;
1477 char *retptr;
1478 char *devstr = NULL;
1479 int ret = 0;
1480 int mod = 0;
1481
1482 if (!capable(CAP_SYS_ADMIN))
1483 return -EPERM;
1484
1485 ret = mnt_want_write_file(file);
1486 if (ret)
1487 return ret;
1488
1489 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1490 1)) {
1491 mnt_drop_write_file(file);
1492 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1493 }
1494
1495 mutex_lock(&root->fs_info->volume_mutex);
1496 vol_args = memdup_user(arg, sizeof(*vol_args));
1497 if (IS_ERR(vol_args)) {
1498 ret = PTR_ERR(vol_args);
1499 goto out;
1500 }
1501
1502 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1503
1504 sizestr = vol_args->name;
1505 devstr = strchr(sizestr, ':');
1506 if (devstr) {
1507 sizestr = devstr + 1;
1508 *devstr = '\0';
1509 devstr = vol_args->name;
1510 ret = kstrtoull(devstr, 10, &devid);
1511 if (ret)
1512 goto out_free;
1513 if (!devid) {
1514 ret = -EINVAL;
1515 goto out_free;
1516 }
1517 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1518 }
1519
1520 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1521 if (!device) {
1522 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1523 devid);
1524 ret = -ENODEV;
1525 goto out_free;
1526 }
1527
1528 if (!device->writeable) {
1529 btrfs_info(root->fs_info,
1530 "resizer unable to apply on readonly device %llu",
1531 devid);
1532 ret = -EPERM;
1533 goto out_free;
1534 }
1535
1536 if (!strcmp(sizestr, "max"))
1537 new_size = device->bdev->bd_inode->i_size;
1538 else {
1539 if (sizestr[0] == '-') {
1540 mod = -1;
1541 sizestr++;
1542 } else if (sizestr[0] == '+') {
1543 mod = 1;
1544 sizestr++;
1545 }
1546 new_size = memparse(sizestr, &retptr);
1547 if (*retptr != '\0' || new_size == 0) {
1548 ret = -EINVAL;
1549 goto out_free;
1550 }
1551 }
1552
1553 if (device->is_tgtdev_for_dev_replace) {
1554 ret = -EPERM;
1555 goto out_free;
1556 }
1557
1558 old_size = btrfs_device_get_total_bytes(device);
1559
1560 if (mod < 0) {
1561 if (new_size > old_size) {
1562 ret = -EINVAL;
1563 goto out_free;
1564 }
1565 new_size = old_size - new_size;
1566 } else if (mod > 0) {
1567 if (new_size > ULLONG_MAX - old_size) {
1568 ret = -ERANGE;
1569 goto out_free;
1570 }
1571 new_size = old_size + new_size;
1572 }
1573
1574 if (new_size < 256 * 1024 * 1024) {
1575 ret = -EINVAL;
1576 goto out_free;
1577 }
1578 if (new_size > device->bdev->bd_inode->i_size) {
1579 ret = -EFBIG;
1580 goto out_free;
1581 }
1582
1583 new_size = div_u64(new_size, root->sectorsize);
1584 new_size *= root->sectorsize;
1585
1586 btrfs_info_in_rcu(root->fs_info, "new size for %s is %llu",
1587 rcu_str_deref(device->name), new_size);
1588
1589 if (new_size > old_size) {
1590 trans = btrfs_start_transaction(root, 0);
1591 if (IS_ERR(trans)) {
1592 ret = PTR_ERR(trans);
1593 goto out_free;
1594 }
1595 ret = btrfs_grow_device(trans, device, new_size);
1596 btrfs_commit_transaction(trans, root);
1597 } else if (new_size < old_size) {
1598 ret = btrfs_shrink_device(device, new_size);
1599 } /* equal, nothing need to do */
1600
1601 out_free:
1602 kfree(vol_args);
1603 out:
1604 mutex_unlock(&root->fs_info->volume_mutex);
1605 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1606 mnt_drop_write_file(file);
1607 return ret;
1608 }
1609
1610 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1611 char *name, unsigned long fd, int subvol,
1612 u64 *transid, bool readonly,
1613 struct btrfs_qgroup_inherit *inherit)
1614 {
1615 int namelen;
1616 int ret = 0;
1617
1618 ret = mnt_want_write_file(file);
1619 if (ret)
1620 goto out;
1621
1622 namelen = strlen(name);
1623 if (strchr(name, '/')) {
1624 ret = -EINVAL;
1625 goto out_drop_write;
1626 }
1627
1628 if (name[0] == '.' &&
1629 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1630 ret = -EEXIST;
1631 goto out_drop_write;
1632 }
1633
1634 if (subvol) {
1635 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1636 NULL, transid, readonly, inherit);
1637 } else {
1638 struct fd src = fdget(fd);
1639 struct inode *src_inode;
1640 if (!src.file) {
1641 ret = -EINVAL;
1642 goto out_drop_write;
1643 }
1644
1645 src_inode = file_inode(src.file);
1646 if (src_inode->i_sb != file_inode(file)->i_sb) {
1647 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1648 "Snapshot src from another FS");
1649 ret = -EXDEV;
1650 } else if (!inode_owner_or_capable(src_inode)) {
1651 /*
1652 * Subvolume creation is not restricted, but snapshots
1653 * are limited to own subvolumes only
1654 */
1655 ret = -EPERM;
1656 } else {
1657 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1658 BTRFS_I(src_inode)->root,
1659 transid, readonly, inherit);
1660 }
1661 fdput(src);
1662 }
1663 out_drop_write:
1664 mnt_drop_write_file(file);
1665 out:
1666 return ret;
1667 }
1668
1669 static noinline int btrfs_ioctl_snap_create(struct file *file,
1670 void __user *arg, int subvol)
1671 {
1672 struct btrfs_ioctl_vol_args *vol_args;
1673 int ret;
1674
1675 vol_args = memdup_user(arg, sizeof(*vol_args));
1676 if (IS_ERR(vol_args))
1677 return PTR_ERR(vol_args);
1678 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1679
1680 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1681 vol_args->fd, subvol,
1682 NULL, false, NULL);
1683
1684 kfree(vol_args);
1685 return ret;
1686 }
1687
1688 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1689 void __user *arg, int subvol)
1690 {
1691 struct btrfs_ioctl_vol_args_v2 *vol_args;
1692 int ret;
1693 u64 transid = 0;
1694 u64 *ptr = NULL;
1695 bool readonly = false;
1696 struct btrfs_qgroup_inherit *inherit = NULL;
1697
1698 vol_args = memdup_user(arg, sizeof(*vol_args));
1699 if (IS_ERR(vol_args))
1700 return PTR_ERR(vol_args);
1701 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1702
1703 if (vol_args->flags &
1704 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1705 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1706 ret = -EOPNOTSUPP;
1707 goto free_args;
1708 }
1709
1710 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1711 ptr = &transid;
1712 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1713 readonly = true;
1714 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1715 if (vol_args->size > PAGE_CACHE_SIZE) {
1716 ret = -EINVAL;
1717 goto free_args;
1718 }
1719 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1720 if (IS_ERR(inherit)) {
1721 ret = PTR_ERR(inherit);
1722 goto free_args;
1723 }
1724 }
1725
1726 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1727 vol_args->fd, subvol, ptr,
1728 readonly, inherit);
1729 if (ret)
1730 goto free_inherit;
1731
1732 if (ptr && copy_to_user(arg +
1733 offsetof(struct btrfs_ioctl_vol_args_v2,
1734 transid),
1735 ptr, sizeof(*ptr)))
1736 ret = -EFAULT;
1737
1738 free_inherit:
1739 kfree(inherit);
1740 free_args:
1741 kfree(vol_args);
1742 return ret;
1743 }
1744
1745 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1746 void __user *arg)
1747 {
1748 struct inode *inode = file_inode(file);
1749 struct btrfs_root *root = BTRFS_I(inode)->root;
1750 int ret = 0;
1751 u64 flags = 0;
1752
1753 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1754 return -EINVAL;
1755
1756 down_read(&root->fs_info->subvol_sem);
1757 if (btrfs_root_readonly(root))
1758 flags |= BTRFS_SUBVOL_RDONLY;
1759 up_read(&root->fs_info->subvol_sem);
1760
1761 if (copy_to_user(arg, &flags, sizeof(flags)))
1762 ret = -EFAULT;
1763
1764 return ret;
1765 }
1766
1767 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1768 void __user *arg)
1769 {
1770 struct inode *inode = file_inode(file);
1771 struct btrfs_root *root = BTRFS_I(inode)->root;
1772 struct btrfs_trans_handle *trans;
1773 u64 root_flags;
1774 u64 flags;
1775 int ret = 0;
1776
1777 if (!inode_owner_or_capable(inode))
1778 return -EPERM;
1779
1780 ret = mnt_want_write_file(file);
1781 if (ret)
1782 goto out;
1783
1784 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1785 ret = -EINVAL;
1786 goto out_drop_write;
1787 }
1788
1789 if (copy_from_user(&flags, arg, sizeof(flags))) {
1790 ret = -EFAULT;
1791 goto out_drop_write;
1792 }
1793
1794 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1795 ret = -EINVAL;
1796 goto out_drop_write;
1797 }
1798
1799 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1800 ret = -EOPNOTSUPP;
1801 goto out_drop_write;
1802 }
1803
1804 down_write(&root->fs_info->subvol_sem);
1805
1806 /* nothing to do */
1807 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1808 goto out_drop_sem;
1809
1810 root_flags = btrfs_root_flags(&root->root_item);
1811 if (flags & BTRFS_SUBVOL_RDONLY) {
1812 btrfs_set_root_flags(&root->root_item,
1813 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1814 } else {
1815 /*
1816 * Block RO -> RW transition if this subvolume is involved in
1817 * send
1818 */
1819 spin_lock(&root->root_item_lock);
1820 if (root->send_in_progress == 0) {
1821 btrfs_set_root_flags(&root->root_item,
1822 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1823 spin_unlock(&root->root_item_lock);
1824 } else {
1825 spin_unlock(&root->root_item_lock);
1826 btrfs_warn(root->fs_info,
1827 "Attempt to set subvolume %llu read-write during send",
1828 root->root_key.objectid);
1829 ret = -EPERM;
1830 goto out_drop_sem;
1831 }
1832 }
1833
1834 trans = btrfs_start_transaction(root, 1);
1835 if (IS_ERR(trans)) {
1836 ret = PTR_ERR(trans);
1837 goto out_reset;
1838 }
1839
1840 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1841 &root->root_key, &root->root_item);
1842
1843 btrfs_commit_transaction(trans, root);
1844 out_reset:
1845 if (ret)
1846 btrfs_set_root_flags(&root->root_item, root_flags);
1847 out_drop_sem:
1848 up_write(&root->fs_info->subvol_sem);
1849 out_drop_write:
1850 mnt_drop_write_file(file);
1851 out:
1852 return ret;
1853 }
1854
1855 /*
1856 * helper to check if the subvolume references other subvolumes
1857 */
1858 static noinline int may_destroy_subvol(struct btrfs_root *root)
1859 {
1860 struct btrfs_path *path;
1861 struct btrfs_dir_item *di;
1862 struct btrfs_key key;
1863 u64 dir_id;
1864 int ret;
1865
1866 path = btrfs_alloc_path();
1867 if (!path)
1868 return -ENOMEM;
1869
1870 /* Make sure this root isn't set as the default subvol */
1871 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1872 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1873 dir_id, "default", 7, 0);
1874 if (di && !IS_ERR(di)) {
1875 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1876 if (key.objectid == root->root_key.objectid) {
1877 ret = -EPERM;
1878 btrfs_err(root->fs_info, "deleting default subvolume "
1879 "%llu is not allowed", key.objectid);
1880 goto out;
1881 }
1882 btrfs_release_path(path);
1883 }
1884
1885 key.objectid = root->root_key.objectid;
1886 key.type = BTRFS_ROOT_REF_KEY;
1887 key.offset = (u64)-1;
1888
1889 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1890 &key, path, 0, 0);
1891 if (ret < 0)
1892 goto out;
1893 BUG_ON(ret == 0);
1894
1895 ret = 0;
1896 if (path->slots[0] > 0) {
1897 path->slots[0]--;
1898 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1899 if (key.objectid == root->root_key.objectid &&
1900 key.type == BTRFS_ROOT_REF_KEY)
1901 ret = -ENOTEMPTY;
1902 }
1903 out:
1904 btrfs_free_path(path);
1905 return ret;
1906 }
1907
1908 static noinline int key_in_sk(struct btrfs_key *key,
1909 struct btrfs_ioctl_search_key *sk)
1910 {
1911 struct btrfs_key test;
1912 int ret;
1913
1914 test.objectid = sk->min_objectid;
1915 test.type = sk->min_type;
1916 test.offset = sk->min_offset;
1917
1918 ret = btrfs_comp_cpu_keys(key, &test);
1919 if (ret < 0)
1920 return 0;
1921
1922 test.objectid = sk->max_objectid;
1923 test.type = sk->max_type;
1924 test.offset = sk->max_offset;
1925
1926 ret = btrfs_comp_cpu_keys(key, &test);
1927 if (ret > 0)
1928 return 0;
1929 return 1;
1930 }
1931
1932 static noinline int copy_to_sk(struct btrfs_root *root,
1933 struct btrfs_path *path,
1934 struct btrfs_key *key,
1935 struct btrfs_ioctl_search_key *sk,
1936 size_t *buf_size,
1937 char __user *ubuf,
1938 unsigned long *sk_offset,
1939 int *num_found)
1940 {
1941 u64 found_transid;
1942 struct extent_buffer *leaf;
1943 struct btrfs_ioctl_search_header sh;
1944 struct btrfs_key test;
1945 unsigned long item_off;
1946 unsigned long item_len;
1947 int nritems;
1948 int i;
1949 int slot;
1950 int ret = 0;
1951
1952 leaf = path->nodes[0];
1953 slot = path->slots[0];
1954 nritems = btrfs_header_nritems(leaf);
1955
1956 if (btrfs_header_generation(leaf) > sk->max_transid) {
1957 i = nritems;
1958 goto advance_key;
1959 }
1960 found_transid = btrfs_header_generation(leaf);
1961
1962 for (i = slot; i < nritems; i++) {
1963 item_off = btrfs_item_ptr_offset(leaf, i);
1964 item_len = btrfs_item_size_nr(leaf, i);
1965
1966 btrfs_item_key_to_cpu(leaf, key, i);
1967 if (!key_in_sk(key, sk))
1968 continue;
1969
1970 if (sizeof(sh) + item_len > *buf_size) {
1971 if (*num_found) {
1972 ret = 1;
1973 goto out;
1974 }
1975
1976 /*
1977 * return one empty item back for v1, which does not
1978 * handle -EOVERFLOW
1979 */
1980
1981 *buf_size = sizeof(sh) + item_len;
1982 item_len = 0;
1983 ret = -EOVERFLOW;
1984 }
1985
1986 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1987 ret = 1;
1988 goto out;
1989 }
1990
1991 sh.objectid = key->objectid;
1992 sh.offset = key->offset;
1993 sh.type = key->type;
1994 sh.len = item_len;
1995 sh.transid = found_transid;
1996
1997 /* copy search result header */
1998 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1999 ret = -EFAULT;
2000 goto out;
2001 }
2002
2003 *sk_offset += sizeof(sh);
2004
2005 if (item_len) {
2006 char __user *up = ubuf + *sk_offset;
2007 /* copy the item */
2008 if (read_extent_buffer_to_user(leaf, up,
2009 item_off, item_len)) {
2010 ret = -EFAULT;
2011 goto out;
2012 }
2013
2014 *sk_offset += item_len;
2015 }
2016 (*num_found)++;
2017
2018 if (ret) /* -EOVERFLOW from above */
2019 goto out;
2020
2021 if (*num_found >= sk->nr_items) {
2022 ret = 1;
2023 goto out;
2024 }
2025 }
2026 advance_key:
2027 ret = 0;
2028 test.objectid = sk->max_objectid;
2029 test.type = sk->max_type;
2030 test.offset = sk->max_offset;
2031 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2032 ret = 1;
2033 else if (key->offset < (u64)-1)
2034 key->offset++;
2035 else if (key->type < (u8)-1) {
2036 key->offset = 0;
2037 key->type++;
2038 } else if (key->objectid < (u64)-1) {
2039 key->offset = 0;
2040 key->type = 0;
2041 key->objectid++;
2042 } else
2043 ret = 1;
2044 out:
2045 /*
2046 * 0: all items from this leaf copied, continue with next
2047 * 1: * more items can be copied, but unused buffer is too small
2048 * * all items were found
2049 * Either way, it will stops the loop which iterates to the next
2050 * leaf
2051 * -EOVERFLOW: item was to large for buffer
2052 * -EFAULT: could not copy extent buffer back to userspace
2053 */
2054 return ret;
2055 }
2056
2057 static noinline int search_ioctl(struct inode *inode,
2058 struct btrfs_ioctl_search_key *sk,
2059 size_t *buf_size,
2060 char __user *ubuf)
2061 {
2062 struct btrfs_root *root;
2063 struct btrfs_key key;
2064 struct btrfs_path *path;
2065 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2066 int ret;
2067 int num_found = 0;
2068 unsigned long sk_offset = 0;
2069
2070 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2071 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2072 return -EOVERFLOW;
2073 }
2074
2075 path = btrfs_alloc_path();
2076 if (!path)
2077 return -ENOMEM;
2078
2079 if (sk->tree_id == 0) {
2080 /* search the root of the inode that was passed */
2081 root = BTRFS_I(inode)->root;
2082 } else {
2083 key.objectid = sk->tree_id;
2084 key.type = BTRFS_ROOT_ITEM_KEY;
2085 key.offset = (u64)-1;
2086 root = btrfs_read_fs_root_no_name(info, &key);
2087 if (IS_ERR(root)) {
2088 btrfs_err(info, "could not find root %llu",
2089 sk->tree_id);
2090 btrfs_free_path(path);
2091 return -ENOENT;
2092 }
2093 }
2094
2095 key.objectid = sk->min_objectid;
2096 key.type = sk->min_type;
2097 key.offset = sk->min_offset;
2098
2099 while (1) {
2100 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2101 if (ret != 0) {
2102 if (ret > 0)
2103 ret = 0;
2104 goto err;
2105 }
2106 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2107 &sk_offset, &num_found);
2108 btrfs_release_path(path);
2109 if (ret)
2110 break;
2111
2112 }
2113 if (ret > 0)
2114 ret = 0;
2115 err:
2116 sk->nr_items = num_found;
2117 btrfs_free_path(path);
2118 return ret;
2119 }
2120
2121 static noinline int btrfs_ioctl_tree_search(struct file *file,
2122 void __user *argp)
2123 {
2124 struct btrfs_ioctl_search_args __user *uargs;
2125 struct btrfs_ioctl_search_key sk;
2126 struct inode *inode;
2127 int ret;
2128 size_t buf_size;
2129
2130 if (!capable(CAP_SYS_ADMIN))
2131 return -EPERM;
2132
2133 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2134
2135 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2136 return -EFAULT;
2137
2138 buf_size = sizeof(uargs->buf);
2139
2140 inode = file_inode(file);
2141 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2142
2143 /*
2144 * In the origin implementation an overflow is handled by returning a
2145 * search header with a len of zero, so reset ret.
2146 */
2147 if (ret == -EOVERFLOW)
2148 ret = 0;
2149
2150 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2151 ret = -EFAULT;
2152 return ret;
2153 }
2154
2155 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2156 void __user *argp)
2157 {
2158 struct btrfs_ioctl_search_args_v2 __user *uarg;
2159 struct btrfs_ioctl_search_args_v2 args;
2160 struct inode *inode;
2161 int ret;
2162 size_t buf_size;
2163 const size_t buf_limit = 16 * 1024 * 1024;
2164
2165 if (!capable(CAP_SYS_ADMIN))
2166 return -EPERM;
2167
2168 /* copy search header and buffer size */
2169 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2170 if (copy_from_user(&args, uarg, sizeof(args)))
2171 return -EFAULT;
2172
2173 buf_size = args.buf_size;
2174
2175 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2176 return -EOVERFLOW;
2177
2178 /* limit result size to 16MB */
2179 if (buf_size > buf_limit)
2180 buf_size = buf_limit;
2181
2182 inode = file_inode(file);
2183 ret = search_ioctl(inode, &args.key, &buf_size,
2184 (char *)(&uarg->buf[0]));
2185 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2186 ret = -EFAULT;
2187 else if (ret == -EOVERFLOW &&
2188 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2189 ret = -EFAULT;
2190
2191 return ret;
2192 }
2193
2194 /*
2195 * Search INODE_REFs to identify path name of 'dirid' directory
2196 * in a 'tree_id' tree. and sets path name to 'name'.
2197 */
2198 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2199 u64 tree_id, u64 dirid, char *name)
2200 {
2201 struct btrfs_root *root;
2202 struct btrfs_key key;
2203 char *ptr;
2204 int ret = -1;
2205 int slot;
2206 int len;
2207 int total_len = 0;
2208 struct btrfs_inode_ref *iref;
2209 struct extent_buffer *l;
2210 struct btrfs_path *path;
2211
2212 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2213 name[0]='\0';
2214 return 0;
2215 }
2216
2217 path = btrfs_alloc_path();
2218 if (!path)
2219 return -ENOMEM;
2220
2221 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2222
2223 key.objectid = tree_id;
2224 key.type = BTRFS_ROOT_ITEM_KEY;
2225 key.offset = (u64)-1;
2226 root = btrfs_read_fs_root_no_name(info, &key);
2227 if (IS_ERR(root)) {
2228 btrfs_err(info, "could not find root %llu", tree_id);
2229 ret = -ENOENT;
2230 goto out;
2231 }
2232
2233 key.objectid = dirid;
2234 key.type = BTRFS_INODE_REF_KEY;
2235 key.offset = (u64)-1;
2236
2237 while (1) {
2238 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2239 if (ret < 0)
2240 goto out;
2241 else if (ret > 0) {
2242 ret = btrfs_previous_item(root, path, dirid,
2243 BTRFS_INODE_REF_KEY);
2244 if (ret < 0)
2245 goto out;
2246 else if (ret > 0) {
2247 ret = -ENOENT;
2248 goto out;
2249 }
2250 }
2251
2252 l = path->nodes[0];
2253 slot = path->slots[0];
2254 btrfs_item_key_to_cpu(l, &key, slot);
2255
2256 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2257 len = btrfs_inode_ref_name_len(l, iref);
2258 ptr -= len + 1;
2259 total_len += len + 1;
2260 if (ptr < name) {
2261 ret = -ENAMETOOLONG;
2262 goto out;
2263 }
2264
2265 *(ptr + len) = '/';
2266 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2267
2268 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2269 break;
2270
2271 btrfs_release_path(path);
2272 key.objectid = key.offset;
2273 key.offset = (u64)-1;
2274 dirid = key.objectid;
2275 }
2276 memmove(name, ptr, total_len);
2277 name[total_len] = '\0';
2278 ret = 0;
2279 out:
2280 btrfs_free_path(path);
2281 return ret;
2282 }
2283
2284 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2285 void __user *argp)
2286 {
2287 struct btrfs_ioctl_ino_lookup_args *args;
2288 struct inode *inode;
2289 int ret = 0;
2290
2291 args = memdup_user(argp, sizeof(*args));
2292 if (IS_ERR(args))
2293 return PTR_ERR(args);
2294
2295 inode = file_inode(file);
2296
2297 /*
2298 * Unprivileged query to obtain the containing subvolume root id. The
2299 * path is reset so it's consistent with btrfs_search_path_in_tree.
2300 */
2301 if (args->treeid == 0)
2302 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2303
2304 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2305 args->name[0] = 0;
2306 goto out;
2307 }
2308
2309 if (!capable(CAP_SYS_ADMIN)) {
2310 ret = -EPERM;
2311 goto out;
2312 }
2313
2314 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2315 args->treeid, args->objectid,
2316 args->name);
2317
2318 out:
2319 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2320 ret = -EFAULT;
2321
2322 kfree(args);
2323 return ret;
2324 }
2325
2326 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2327 void __user *arg)
2328 {
2329 struct dentry *parent = file->f_path.dentry;
2330 struct dentry *dentry;
2331 struct inode *dir = d_inode(parent);
2332 struct inode *inode;
2333 struct btrfs_root *root = BTRFS_I(dir)->root;
2334 struct btrfs_root *dest = NULL;
2335 struct btrfs_ioctl_vol_args *vol_args;
2336 struct btrfs_trans_handle *trans;
2337 struct btrfs_block_rsv block_rsv;
2338 u64 root_flags;
2339 u64 qgroup_reserved;
2340 int namelen;
2341 int ret;
2342 int err = 0;
2343
2344 vol_args = memdup_user(arg, sizeof(*vol_args));
2345 if (IS_ERR(vol_args))
2346 return PTR_ERR(vol_args);
2347
2348 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2349 namelen = strlen(vol_args->name);
2350 if (strchr(vol_args->name, '/') ||
2351 strncmp(vol_args->name, "..", namelen) == 0) {
2352 err = -EINVAL;
2353 goto out;
2354 }
2355
2356 err = mnt_want_write_file(file);
2357 if (err)
2358 goto out;
2359
2360
2361 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2362 if (err == -EINTR)
2363 goto out_drop_write;
2364 dentry = lookup_one_len(vol_args->name, parent, namelen);
2365 if (IS_ERR(dentry)) {
2366 err = PTR_ERR(dentry);
2367 goto out_unlock_dir;
2368 }
2369
2370 if (d_really_is_negative(dentry)) {
2371 err = -ENOENT;
2372 goto out_dput;
2373 }
2374
2375 inode = d_inode(dentry);
2376 dest = BTRFS_I(inode)->root;
2377 if (!capable(CAP_SYS_ADMIN)) {
2378 /*
2379 * Regular user. Only allow this with a special mount
2380 * option, when the user has write+exec access to the
2381 * subvol root, and when rmdir(2) would have been
2382 * allowed.
2383 *
2384 * Note that this is _not_ check that the subvol is
2385 * empty or doesn't contain data that we wouldn't
2386 * otherwise be able to delete.
2387 *
2388 * Users who want to delete empty subvols should try
2389 * rmdir(2).
2390 */
2391 err = -EPERM;
2392 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2393 goto out_dput;
2394
2395 /*
2396 * Do not allow deletion if the parent dir is the same
2397 * as the dir to be deleted. That means the ioctl
2398 * must be called on the dentry referencing the root
2399 * of the subvol, not a random directory contained
2400 * within it.
2401 */
2402 err = -EINVAL;
2403 if (root == dest)
2404 goto out_dput;
2405
2406 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2407 if (err)
2408 goto out_dput;
2409 }
2410
2411 /* check if subvolume may be deleted by a user */
2412 err = btrfs_may_delete(dir, dentry, 1);
2413 if (err)
2414 goto out_dput;
2415
2416 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2417 err = -EINVAL;
2418 goto out_dput;
2419 }
2420
2421 mutex_lock(&inode->i_mutex);
2422
2423 /*
2424 * Don't allow to delete a subvolume with send in progress. This is
2425 * inside the i_mutex so the error handling that has to drop the bit
2426 * again is not run concurrently.
2427 */
2428 spin_lock(&dest->root_item_lock);
2429 root_flags = btrfs_root_flags(&dest->root_item);
2430 if (dest->send_in_progress == 0) {
2431 btrfs_set_root_flags(&dest->root_item,
2432 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2433 spin_unlock(&dest->root_item_lock);
2434 } else {
2435 spin_unlock(&dest->root_item_lock);
2436 btrfs_warn(root->fs_info,
2437 "Attempt to delete subvolume %llu during send",
2438 dest->root_key.objectid);
2439 err = -EPERM;
2440 goto out_unlock_inode;
2441 }
2442
2443 down_write(&root->fs_info->subvol_sem);
2444
2445 err = may_destroy_subvol(dest);
2446 if (err)
2447 goto out_up_write;
2448
2449 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2450 /*
2451 * One for dir inode, two for dir entries, two for root
2452 * ref/backref.
2453 */
2454 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2455 5, &qgroup_reserved, true);
2456 if (err)
2457 goto out_up_write;
2458
2459 trans = btrfs_start_transaction(root, 0);
2460 if (IS_ERR(trans)) {
2461 err = PTR_ERR(trans);
2462 goto out_release;
2463 }
2464 trans->block_rsv = &block_rsv;
2465 trans->bytes_reserved = block_rsv.size;
2466
2467 ret = btrfs_unlink_subvol(trans, root, dir,
2468 dest->root_key.objectid,
2469 dentry->d_name.name,
2470 dentry->d_name.len);
2471 if (ret) {
2472 err = ret;
2473 btrfs_abort_transaction(trans, root, ret);
2474 goto out_end_trans;
2475 }
2476
2477 btrfs_record_root_in_trans(trans, dest);
2478
2479 memset(&dest->root_item.drop_progress, 0,
2480 sizeof(dest->root_item.drop_progress));
2481 dest->root_item.drop_level = 0;
2482 btrfs_set_root_refs(&dest->root_item, 0);
2483
2484 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2485 ret = btrfs_insert_orphan_item(trans,
2486 root->fs_info->tree_root,
2487 dest->root_key.objectid);
2488 if (ret) {
2489 btrfs_abort_transaction(trans, root, ret);
2490 err = ret;
2491 goto out_end_trans;
2492 }
2493 }
2494
2495 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2496 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2497 dest->root_key.objectid);
2498 if (ret && ret != -ENOENT) {
2499 btrfs_abort_transaction(trans, root, ret);
2500 err = ret;
2501 goto out_end_trans;
2502 }
2503 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2504 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2505 dest->root_item.received_uuid,
2506 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2507 dest->root_key.objectid);
2508 if (ret && ret != -ENOENT) {
2509 btrfs_abort_transaction(trans, root, ret);
2510 err = ret;
2511 goto out_end_trans;
2512 }
2513 }
2514
2515 out_end_trans:
2516 trans->block_rsv = NULL;
2517 trans->bytes_reserved = 0;
2518 ret = btrfs_end_transaction(trans, root);
2519 if (ret && !err)
2520 err = ret;
2521 inode->i_flags |= S_DEAD;
2522 out_release:
2523 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2524 out_up_write:
2525 up_write(&root->fs_info->subvol_sem);
2526 if (err) {
2527 spin_lock(&dest->root_item_lock);
2528 root_flags = btrfs_root_flags(&dest->root_item);
2529 btrfs_set_root_flags(&dest->root_item,
2530 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2531 spin_unlock(&dest->root_item_lock);
2532 }
2533 out_unlock_inode:
2534 mutex_unlock(&inode->i_mutex);
2535 if (!err) {
2536 d_invalidate(dentry);
2537 btrfs_invalidate_inodes(dest);
2538 d_delete(dentry);
2539 ASSERT(dest->send_in_progress == 0);
2540
2541 /* the last ref */
2542 if (dest->ino_cache_inode) {
2543 iput(dest->ino_cache_inode);
2544 dest->ino_cache_inode = NULL;
2545 }
2546 }
2547 out_dput:
2548 dput(dentry);
2549 out_unlock_dir:
2550 mutex_unlock(&dir->i_mutex);
2551 out_drop_write:
2552 mnt_drop_write_file(file);
2553 out:
2554 kfree(vol_args);
2555 return err;
2556 }
2557
2558 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2559 {
2560 struct inode *inode = file_inode(file);
2561 struct btrfs_root *root = BTRFS_I(inode)->root;
2562 struct btrfs_ioctl_defrag_range_args *range;
2563 int ret;
2564
2565 ret = mnt_want_write_file(file);
2566 if (ret)
2567 return ret;
2568
2569 if (btrfs_root_readonly(root)) {
2570 ret = -EROFS;
2571 goto out;
2572 }
2573
2574 switch (inode->i_mode & S_IFMT) {
2575 case S_IFDIR:
2576 if (!capable(CAP_SYS_ADMIN)) {
2577 ret = -EPERM;
2578 goto out;
2579 }
2580 ret = btrfs_defrag_root(root);
2581 if (ret)
2582 goto out;
2583 ret = btrfs_defrag_root(root->fs_info->extent_root);
2584 break;
2585 case S_IFREG:
2586 if (!(file->f_mode & FMODE_WRITE)) {
2587 ret = -EINVAL;
2588 goto out;
2589 }
2590
2591 range = kzalloc(sizeof(*range), GFP_KERNEL);
2592 if (!range) {
2593 ret = -ENOMEM;
2594 goto out;
2595 }
2596
2597 if (argp) {
2598 if (copy_from_user(range, argp,
2599 sizeof(*range))) {
2600 ret = -EFAULT;
2601 kfree(range);
2602 goto out;
2603 }
2604 /* compression requires us to start the IO */
2605 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2606 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2607 range->extent_thresh = (u32)-1;
2608 }
2609 } else {
2610 /* the rest are all set to zero by kzalloc */
2611 range->len = (u64)-1;
2612 }
2613 ret = btrfs_defrag_file(file_inode(file), file,
2614 range, 0, 0);
2615 if (ret > 0)
2616 ret = 0;
2617 kfree(range);
2618 break;
2619 default:
2620 ret = -EINVAL;
2621 }
2622 out:
2623 mnt_drop_write_file(file);
2624 return ret;
2625 }
2626
2627 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2628 {
2629 struct btrfs_ioctl_vol_args *vol_args;
2630 int ret;
2631
2632 if (!capable(CAP_SYS_ADMIN))
2633 return -EPERM;
2634
2635 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2636 1)) {
2637 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2638 }
2639
2640 mutex_lock(&root->fs_info->volume_mutex);
2641 vol_args = memdup_user(arg, sizeof(*vol_args));
2642 if (IS_ERR(vol_args)) {
2643 ret = PTR_ERR(vol_args);
2644 goto out;
2645 }
2646
2647 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2648 ret = btrfs_init_new_device(root, vol_args->name);
2649
2650 if (!ret)
2651 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2652
2653 kfree(vol_args);
2654 out:
2655 mutex_unlock(&root->fs_info->volume_mutex);
2656 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2657 return ret;
2658 }
2659
2660 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2661 {
2662 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2663 struct btrfs_ioctl_vol_args *vol_args;
2664 int ret;
2665
2666 if (!capable(CAP_SYS_ADMIN))
2667 return -EPERM;
2668
2669 ret = mnt_want_write_file(file);
2670 if (ret)
2671 return ret;
2672
2673 vol_args = memdup_user(arg, sizeof(*vol_args));
2674 if (IS_ERR(vol_args)) {
2675 ret = PTR_ERR(vol_args);
2676 goto err_drop;
2677 }
2678
2679 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2680
2681 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2682 1)) {
2683 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2684 goto out;
2685 }
2686
2687 mutex_lock(&root->fs_info->volume_mutex);
2688 ret = btrfs_rm_device(root, vol_args->name);
2689 mutex_unlock(&root->fs_info->volume_mutex);
2690 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2691
2692 if (!ret)
2693 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2694
2695 out:
2696 kfree(vol_args);
2697 err_drop:
2698 mnt_drop_write_file(file);
2699 return ret;
2700 }
2701
2702 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2703 {
2704 struct btrfs_ioctl_fs_info_args *fi_args;
2705 struct btrfs_device *device;
2706 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2707 int ret = 0;
2708
2709 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2710 if (!fi_args)
2711 return -ENOMEM;
2712
2713 mutex_lock(&fs_devices->device_list_mutex);
2714 fi_args->num_devices = fs_devices->num_devices;
2715 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2716
2717 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2718 if (device->devid > fi_args->max_id)
2719 fi_args->max_id = device->devid;
2720 }
2721 mutex_unlock(&fs_devices->device_list_mutex);
2722
2723 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2724 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2725 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2726
2727 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2728 ret = -EFAULT;
2729
2730 kfree(fi_args);
2731 return ret;
2732 }
2733
2734 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2735 {
2736 struct btrfs_ioctl_dev_info_args *di_args;
2737 struct btrfs_device *dev;
2738 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2739 int ret = 0;
2740 char *s_uuid = NULL;
2741
2742 di_args = memdup_user(arg, sizeof(*di_args));
2743 if (IS_ERR(di_args))
2744 return PTR_ERR(di_args);
2745
2746 if (!btrfs_is_empty_uuid(di_args->uuid))
2747 s_uuid = di_args->uuid;
2748
2749 mutex_lock(&fs_devices->device_list_mutex);
2750 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2751
2752 if (!dev) {
2753 ret = -ENODEV;
2754 goto out;
2755 }
2756
2757 di_args->devid = dev->devid;
2758 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2759 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2760 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2761 if (dev->name) {
2762 struct rcu_string *name;
2763
2764 rcu_read_lock();
2765 name = rcu_dereference(dev->name);
2766 strncpy(di_args->path, name->str, sizeof(di_args->path));
2767 rcu_read_unlock();
2768 di_args->path[sizeof(di_args->path) - 1] = 0;
2769 } else {
2770 di_args->path[0] = '\0';
2771 }
2772
2773 out:
2774 mutex_unlock(&fs_devices->device_list_mutex);
2775 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2776 ret = -EFAULT;
2777
2778 kfree(di_args);
2779 return ret;
2780 }
2781
2782 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2783 {
2784 struct page *page;
2785 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2786
2787 page = grab_cache_page(inode->i_mapping, index);
2788 if (!page)
2789 return NULL;
2790
2791 if (!PageUptodate(page)) {
2792 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2793 0))
2794 return NULL;
2795 lock_page(page);
2796 if (!PageUptodate(page)) {
2797 unlock_page(page);
2798 page_cache_release(page);
2799 return NULL;
2800 }
2801 }
2802 unlock_page(page);
2803
2804 return page;
2805 }
2806
2807 static int gather_extent_pages(struct inode *inode, struct page **pages,
2808 int num_pages, u64 off)
2809 {
2810 int i;
2811 pgoff_t index = off >> PAGE_CACHE_SHIFT;
2812
2813 for (i = 0; i < num_pages; i++) {
2814 pages[i] = extent_same_get_page(inode, index + i);
2815 if (!pages[i])
2816 return -ENOMEM;
2817 }
2818 return 0;
2819 }
2820
2821 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2822 {
2823 /* do any pending delalloc/csum calc on src, one way or
2824 another, and lock file content */
2825 while (1) {
2826 struct btrfs_ordered_extent *ordered;
2827 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2828 ordered = btrfs_lookup_first_ordered_extent(inode,
2829 off + len - 1);
2830 if ((!ordered ||
2831 ordered->file_offset + ordered->len <= off ||
2832 ordered->file_offset >= off + len) &&
2833 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2834 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2835 if (ordered)
2836 btrfs_put_ordered_extent(ordered);
2837 break;
2838 }
2839 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2840 if (ordered)
2841 btrfs_put_ordered_extent(ordered);
2842 btrfs_wait_ordered_range(inode, off, len);
2843 }
2844 }
2845
2846 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2847 {
2848 mutex_unlock(&inode1->i_mutex);
2849 mutex_unlock(&inode2->i_mutex);
2850 }
2851
2852 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2853 {
2854 if (inode1 < inode2)
2855 swap(inode1, inode2);
2856
2857 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2858 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2859 }
2860
2861 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2862 struct inode *inode2, u64 loff2, u64 len)
2863 {
2864 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2865 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2866 }
2867
2868 static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2869 struct inode *inode2, u64 loff2, u64 len)
2870 {
2871 if (inode1 < inode2) {
2872 swap(inode1, inode2);
2873 swap(loff1, loff2);
2874 }
2875 lock_extent_range(inode1, loff1, len);
2876 lock_extent_range(inode2, loff2, len);
2877 }
2878
2879 struct cmp_pages {
2880 int num_pages;
2881 struct page **src_pages;
2882 struct page **dst_pages;
2883 };
2884
2885 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2886 {
2887 int i;
2888 struct page *pg;
2889
2890 for (i = 0; i < cmp->num_pages; i++) {
2891 pg = cmp->src_pages[i];
2892 if (pg)
2893 page_cache_release(pg);
2894 pg = cmp->dst_pages[i];
2895 if (pg)
2896 page_cache_release(pg);
2897 }
2898 kfree(cmp->src_pages);
2899 kfree(cmp->dst_pages);
2900 }
2901
2902 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2903 struct inode *dst, u64 dst_loff,
2904 u64 len, struct cmp_pages *cmp)
2905 {
2906 int ret;
2907 int num_pages = PAGE_CACHE_ALIGN(len) >> PAGE_CACHE_SHIFT;
2908 struct page **src_pgarr, **dst_pgarr;
2909
2910 /*
2911 * We must gather up all the pages before we initiate our
2912 * extent locking. We use an array for the page pointers. Size
2913 * of the array is bounded by len, which is in turn bounded by
2914 * BTRFS_MAX_DEDUPE_LEN.
2915 */
2916 src_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2917 dst_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2918 if (!src_pgarr || !dst_pgarr) {
2919 kfree(src_pgarr);
2920 kfree(dst_pgarr);
2921 return -ENOMEM;
2922 }
2923 cmp->num_pages = num_pages;
2924 cmp->src_pages = src_pgarr;
2925 cmp->dst_pages = dst_pgarr;
2926
2927 ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff);
2928 if (ret)
2929 goto out;
2930
2931 ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff);
2932
2933 out:
2934 if (ret)
2935 btrfs_cmp_data_free(cmp);
2936 return 0;
2937 }
2938
2939 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2940 u64 dst_loff, u64 len, struct cmp_pages *cmp)
2941 {
2942 int ret = 0;
2943 int i;
2944 struct page *src_page, *dst_page;
2945 unsigned int cmp_len = PAGE_CACHE_SIZE;
2946 void *addr, *dst_addr;
2947
2948 i = 0;
2949 while (len) {
2950 if (len < PAGE_CACHE_SIZE)
2951 cmp_len = len;
2952
2953 BUG_ON(i >= cmp->num_pages);
2954
2955 src_page = cmp->src_pages[i];
2956 dst_page = cmp->dst_pages[i];
2957
2958 addr = kmap_atomic(src_page);
2959 dst_addr = kmap_atomic(dst_page);
2960
2961 flush_dcache_page(src_page);
2962 flush_dcache_page(dst_page);
2963
2964 if (memcmp(addr, dst_addr, cmp_len))
2965 ret = BTRFS_SAME_DATA_DIFFERS;
2966
2967 kunmap_atomic(addr);
2968 kunmap_atomic(dst_addr);
2969
2970 if (ret)
2971 break;
2972
2973 len -= cmp_len;
2974 i++;
2975 }
2976
2977 return ret;
2978 }
2979
2980 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
2981 u64 olen)
2982 {
2983 u64 len = *plen;
2984 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2985
2986 if (off + olen > inode->i_size || off + olen < off)
2987 return -EINVAL;
2988
2989 /* if we extend to eof, continue to block boundary */
2990 if (off + len == inode->i_size)
2991 *plen = len = ALIGN(inode->i_size, bs) - off;
2992
2993 /* Check that we are block aligned - btrfs_clone() requires this */
2994 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2995 return -EINVAL;
2996
2997 return 0;
2998 }
2999
3000 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3001 struct inode *dst, u64 dst_loff)
3002 {
3003 int ret;
3004 u64 len = olen;
3005 struct cmp_pages cmp;
3006 int same_inode = 0;
3007 u64 same_lock_start = 0;
3008 u64 same_lock_len = 0;
3009
3010 if (src == dst)
3011 same_inode = 1;
3012
3013 if (len == 0)
3014 return 0;
3015
3016 if (same_inode) {
3017 mutex_lock(&src->i_mutex);
3018
3019 ret = extent_same_check_offsets(src, loff, &len, olen);
3020 if (ret)
3021 goto out_unlock;
3022
3023 /*
3024 * Single inode case wants the same checks, except we
3025 * don't want our length pushed out past i_size as
3026 * comparing that data range makes no sense.
3027 *
3028 * extent_same_check_offsets() will do this for an
3029 * unaligned length at i_size, so catch it here and
3030 * reject the request.
3031 *
3032 * This effectively means we require aligned extents
3033 * for the single-inode case, whereas the other cases
3034 * allow an unaligned length so long as it ends at
3035 * i_size.
3036 */
3037 if (len != olen) {
3038 ret = -EINVAL;
3039 goto out_unlock;
3040 }
3041
3042 /* Check for overlapping ranges */
3043 if (dst_loff + len > loff && dst_loff < loff + len) {
3044 ret = -EINVAL;
3045 goto out_unlock;
3046 }
3047
3048 same_lock_start = min_t(u64, loff, dst_loff);
3049 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3050 } else {
3051 btrfs_double_inode_lock(src, dst);
3052
3053 ret = extent_same_check_offsets(src, loff, &len, olen);
3054 if (ret)
3055 goto out_unlock;
3056
3057 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3058 if (ret)
3059 goto out_unlock;
3060 }
3061
3062 /* don't make the dst file partly checksummed */
3063 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3064 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3065 ret = -EINVAL;
3066 goto out_unlock;
3067 }
3068
3069 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3070 if (ret)
3071 goto out_unlock;
3072
3073 if (same_inode)
3074 lock_extent_range(src, same_lock_start, same_lock_len);
3075 else
3076 btrfs_double_extent_lock(src, loff, dst, dst_loff, len);
3077
3078 /* pass original length for comparison so we stay within i_size */
3079 ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp);
3080 if (ret == 0)
3081 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3082
3083 if (same_inode)
3084 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3085 same_lock_start + same_lock_len - 1);
3086 else
3087 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3088
3089 btrfs_cmp_data_free(&cmp);
3090 out_unlock:
3091 if (same_inode)
3092 mutex_unlock(&src->i_mutex);
3093 else
3094 btrfs_double_inode_unlock(src, dst);
3095
3096 return ret;
3097 }
3098
3099 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
3100
3101 static long btrfs_ioctl_file_extent_same(struct file *file,
3102 struct btrfs_ioctl_same_args __user *argp)
3103 {
3104 struct btrfs_ioctl_same_args *same = NULL;
3105 struct btrfs_ioctl_same_extent_info *info;
3106 struct inode *src = file_inode(file);
3107 u64 off;
3108 u64 len;
3109 int i;
3110 int ret;
3111 unsigned long size;
3112 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3113 bool is_admin = capable(CAP_SYS_ADMIN);
3114 u16 count;
3115
3116 if (!(file->f_mode & FMODE_READ))
3117 return -EINVAL;
3118
3119 ret = mnt_want_write_file(file);
3120 if (ret)
3121 return ret;
3122
3123 if (get_user(count, &argp->dest_count)) {
3124 ret = -EFAULT;
3125 goto out;
3126 }
3127
3128 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
3129
3130 same = memdup_user(argp, size);
3131
3132 if (IS_ERR(same)) {
3133 ret = PTR_ERR(same);
3134 same = NULL;
3135 goto out;
3136 }
3137
3138 off = same->logical_offset;
3139 len = same->length;
3140
3141 /*
3142 * Limit the total length we will dedupe for each operation.
3143 * This is intended to bound the total time spent in this
3144 * ioctl to something sane.
3145 */
3146 if (len > BTRFS_MAX_DEDUPE_LEN)
3147 len = BTRFS_MAX_DEDUPE_LEN;
3148
3149 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
3150 /*
3151 * Btrfs does not support blocksize < page_size. As a
3152 * result, btrfs_cmp_data() won't correctly handle
3153 * this situation without an update.
3154 */
3155 ret = -EINVAL;
3156 goto out;
3157 }
3158
3159 ret = -EISDIR;
3160 if (S_ISDIR(src->i_mode))
3161 goto out;
3162
3163 ret = -EACCES;
3164 if (!S_ISREG(src->i_mode))
3165 goto out;
3166
3167 /* pre-format output fields to sane values */
3168 for (i = 0; i < count; i++) {
3169 same->info[i].bytes_deduped = 0ULL;
3170 same->info[i].status = 0;
3171 }
3172
3173 for (i = 0, info = same->info; i < count; i++, info++) {
3174 struct inode *dst;
3175 struct fd dst_file = fdget(info->fd);
3176 if (!dst_file.file) {
3177 info->status = -EBADF;
3178 continue;
3179 }
3180 dst = file_inode(dst_file.file);
3181
3182 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
3183 info->status = -EINVAL;
3184 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3185 info->status = -EXDEV;
3186 } else if (S_ISDIR(dst->i_mode)) {
3187 info->status = -EISDIR;
3188 } else if (!S_ISREG(dst->i_mode)) {
3189 info->status = -EACCES;
3190 } else {
3191 info->status = btrfs_extent_same(src, off, len, dst,
3192 info->logical_offset);
3193 if (info->status == 0)
3194 info->bytes_deduped += len;
3195 }
3196 fdput(dst_file);
3197 }
3198
3199 ret = copy_to_user(argp, same, size);
3200 if (ret)
3201 ret = -EFAULT;
3202
3203 out:
3204 mnt_drop_write_file(file);
3205 kfree(same);
3206 return ret;
3207 }
3208
3209 /* Helper to check and see if this root currently has a ref on the given disk
3210 * bytenr. If it does then we need to update the quota for this root. This
3211 * doesn't do anything if quotas aren't enabled.
3212 */
3213 static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3214 u64 disko)
3215 {
3216 struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem);
3217 struct ulist *roots;
3218 struct ulist_iterator uiter;
3219 struct ulist_node *root_node = NULL;
3220 int ret;
3221
3222 if (!root->fs_info->quota_enabled)
3223 return 1;
3224
3225 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3226 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3227 tree_mod_seq_elem.seq, &roots);
3228 if (ret < 0)
3229 goto out;
3230 ret = 0;
3231 ULIST_ITER_INIT(&uiter);
3232 while ((root_node = ulist_next(roots, &uiter))) {
3233 if (root_node->val == root->objectid) {
3234 ret = 1;
3235 break;
3236 }
3237 }
3238 ulist_free(roots);
3239 out:
3240 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3241 return ret;
3242 }
3243
3244 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3245 struct inode *inode,
3246 u64 endoff,
3247 const u64 destoff,
3248 const u64 olen,
3249 int no_time_update)
3250 {
3251 struct btrfs_root *root = BTRFS_I(inode)->root;
3252 int ret;
3253
3254 inode_inc_iversion(inode);
3255 if (!no_time_update)
3256 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3257 /*
3258 * We round up to the block size at eof when determining which
3259 * extents to clone above, but shouldn't round up the file size.
3260 */
3261 if (endoff > destoff + olen)
3262 endoff = destoff + olen;
3263 if (endoff > inode->i_size)
3264 btrfs_i_size_write(inode, endoff);
3265
3266 ret = btrfs_update_inode(trans, root, inode);
3267 if (ret) {
3268 btrfs_abort_transaction(trans, root, ret);
3269 btrfs_end_transaction(trans, root);
3270 goto out;
3271 }
3272 ret = btrfs_end_transaction(trans, root);
3273 out:
3274 return ret;
3275 }
3276
3277 static void clone_update_extent_map(struct inode *inode,
3278 const struct btrfs_trans_handle *trans,
3279 const struct btrfs_path *path,
3280 const u64 hole_offset,
3281 const u64 hole_len)
3282 {
3283 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3284 struct extent_map *em;
3285 int ret;
3286
3287 em = alloc_extent_map();
3288 if (!em) {
3289 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3290 &BTRFS_I(inode)->runtime_flags);
3291 return;
3292 }
3293
3294 if (path) {
3295 struct btrfs_file_extent_item *fi;
3296
3297 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3298 struct btrfs_file_extent_item);
3299 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3300 em->generation = -1;
3301 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3302 BTRFS_FILE_EXTENT_INLINE)
3303 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3304 &BTRFS_I(inode)->runtime_flags);
3305 } else {
3306 em->start = hole_offset;
3307 em->len = hole_len;
3308 em->ram_bytes = em->len;
3309 em->orig_start = hole_offset;
3310 em->block_start = EXTENT_MAP_HOLE;
3311 em->block_len = 0;
3312 em->orig_block_len = 0;
3313 em->compress_type = BTRFS_COMPRESS_NONE;
3314 em->generation = trans->transid;
3315 }
3316
3317 while (1) {
3318 write_lock(&em_tree->lock);
3319 ret = add_extent_mapping(em_tree, em, 1);
3320 write_unlock(&em_tree->lock);
3321 if (ret != -EEXIST) {
3322 free_extent_map(em);
3323 break;
3324 }
3325 btrfs_drop_extent_cache(inode, em->start,
3326 em->start + em->len - 1, 0);
3327 }
3328
3329 if (ret)
3330 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3331 &BTRFS_I(inode)->runtime_flags);
3332 }
3333
3334 /*
3335 * Make sure we do not end up inserting an inline extent into a file that has
3336 * already other (non-inline) extents. If a file has an inline extent it can
3337 * not have any other extents and the (single) inline extent must start at the
3338 * file offset 0. Failing to respect these rules will lead to file corruption,
3339 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3340 *
3341 * We can have extents that have been already written to disk or we can have
3342 * dirty ranges still in delalloc, in which case the extent maps and items are
3343 * created only when we run delalloc, and the delalloc ranges might fall outside
3344 * the range we are currently locking in the inode's io tree. So we check the
3345 * inode's i_size because of that (i_size updates are done while holding the
3346 * i_mutex, which we are holding here).
3347 * We also check to see if the inode has a size not greater than "datal" but has
3348 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3349 * protected against such concurrent fallocate calls by the i_mutex).
3350 *
3351 * If the file has no extents but a size greater than datal, do not allow the
3352 * copy because we would need turn the inline extent into a non-inline one (even
3353 * with NO_HOLES enabled). If we find our destination inode only has one inline
3354 * extent, just overwrite it with the source inline extent if its size is less
3355 * than the source extent's size, or we could copy the source inline extent's
3356 * data into the destination inode's inline extent if the later is greater then
3357 * the former.
3358 */
3359 static int clone_copy_inline_extent(struct inode *src,
3360 struct inode *dst,
3361 struct btrfs_trans_handle *trans,
3362 struct btrfs_path *path,
3363 struct btrfs_key *new_key,
3364 const u64 drop_start,
3365 const u64 datal,
3366 const u64 skip,
3367 const u64 size,
3368 char *inline_data)
3369 {
3370 struct btrfs_root *root = BTRFS_I(dst)->root;
3371 const u64 aligned_end = ALIGN(new_key->offset + datal,
3372 root->sectorsize);
3373 int ret;
3374 struct btrfs_key key;
3375
3376 if (new_key->offset > 0)
3377 return -EOPNOTSUPP;
3378
3379 key.objectid = btrfs_ino(dst);
3380 key.type = BTRFS_EXTENT_DATA_KEY;
3381 key.offset = 0;
3382 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3383 if (ret < 0) {
3384 return ret;
3385 } else if (ret > 0) {
3386 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3387 ret = btrfs_next_leaf(root, path);
3388 if (ret < 0)
3389 return ret;
3390 else if (ret > 0)
3391 goto copy_inline_extent;
3392 }
3393 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3394 if (key.objectid == btrfs_ino(dst) &&
3395 key.type == BTRFS_EXTENT_DATA_KEY) {
3396 ASSERT(key.offset > 0);
3397 return -EOPNOTSUPP;
3398 }
3399 } else if (i_size_read(dst) <= datal) {
3400 struct btrfs_file_extent_item *ei;
3401 u64 ext_len;
3402
3403 /*
3404 * If the file size is <= datal, make sure there are no other
3405 * extents following (can happen do to an fallocate call with
3406 * the flag FALLOC_FL_KEEP_SIZE).
3407 */
3408 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3409 struct btrfs_file_extent_item);
3410 /*
3411 * If it's an inline extent, it can not have other extents
3412 * following it.
3413 */
3414 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3415 BTRFS_FILE_EXTENT_INLINE)
3416 goto copy_inline_extent;
3417
3418 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3419 if (ext_len > aligned_end)
3420 return -EOPNOTSUPP;
3421
3422 ret = btrfs_next_item(root, path);
3423 if (ret < 0) {
3424 return ret;
3425 } else if (ret == 0) {
3426 btrfs_item_key_to_cpu(path->nodes[0], &key,
3427 path->slots[0]);
3428 if (key.objectid == btrfs_ino(dst) &&
3429 key.type == BTRFS_EXTENT_DATA_KEY)
3430 return -EOPNOTSUPP;
3431 }
3432 }
3433
3434 copy_inline_extent:
3435 /*
3436 * We have no extent items, or we have an extent at offset 0 which may
3437 * or may not be inlined. All these cases are dealt the same way.
3438 */
3439 if (i_size_read(dst) > datal) {
3440 /*
3441 * If the destination inode has an inline extent...
3442 * This would require copying the data from the source inline
3443 * extent into the beginning of the destination's inline extent.
3444 * But this is really complex, both extents can be compressed
3445 * or just one of them, which would require decompressing and
3446 * re-compressing data (which could increase the new compressed
3447 * size, not allowing the compressed data to fit anymore in an
3448 * inline extent).
3449 * So just don't support this case for now (it should be rare,
3450 * we are not really saving space when cloning inline extents).
3451 */
3452 return -EOPNOTSUPP;
3453 }
3454
3455 btrfs_release_path(path);
3456 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3457 if (ret)
3458 return ret;
3459 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3460 if (ret)
3461 return ret;
3462
3463 if (skip) {
3464 const u32 start = btrfs_file_extent_calc_inline_size(0);
3465
3466 memmove(inline_data + start, inline_data + start + skip, datal);
3467 }
3468
3469 write_extent_buffer(path->nodes[0], inline_data,
3470 btrfs_item_ptr_offset(path->nodes[0],
3471 path->slots[0]),
3472 size);
3473 inode_add_bytes(dst, datal);
3474
3475 return 0;
3476 }
3477
3478 /**
3479 * btrfs_clone() - clone a range from inode file to another
3480 *
3481 * @src: Inode to clone from
3482 * @inode: Inode to clone to
3483 * @off: Offset within source to start clone from
3484 * @olen: Original length, passed by user, of range to clone
3485 * @olen_aligned: Block-aligned value of olen
3486 * @destoff: Offset within @inode to start clone
3487 * @no_time_update: Whether to update mtime/ctime on the target inode
3488 */
3489 static int btrfs_clone(struct inode *src, struct inode *inode,
3490 const u64 off, const u64 olen, const u64 olen_aligned,
3491 const u64 destoff, int no_time_update)
3492 {
3493 struct btrfs_root *root = BTRFS_I(inode)->root;
3494 struct btrfs_path *path = NULL;
3495 struct extent_buffer *leaf;
3496 struct btrfs_trans_handle *trans;
3497 char *buf = NULL;
3498 struct btrfs_key key;
3499 u32 nritems;
3500 int slot;
3501 int ret;
3502 int no_quota;
3503 const u64 len = olen_aligned;
3504 u64 last_disko = 0;
3505 u64 last_dest_end = destoff;
3506
3507 ret = -ENOMEM;
3508 buf = vmalloc(root->nodesize);
3509 if (!buf)
3510 return ret;
3511
3512 path = btrfs_alloc_path();
3513 if (!path) {
3514 vfree(buf);
3515 return ret;
3516 }
3517
3518 path->reada = 2;
3519 /* clone data */
3520 key.objectid = btrfs_ino(src);
3521 key.type = BTRFS_EXTENT_DATA_KEY;
3522 key.offset = off;
3523
3524 while (1) {
3525 u64 next_key_min_offset = key.offset + 1;
3526
3527 /*
3528 * note the key will change type as we walk through the
3529 * tree.
3530 */
3531 path->leave_spinning = 1;
3532 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3533 0, 0);
3534 if (ret < 0)
3535 goto out;
3536 /*
3537 * First search, if no extent item that starts at offset off was
3538 * found but the previous item is an extent item, it's possible
3539 * it might overlap our target range, therefore process it.
3540 */
3541 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3542 btrfs_item_key_to_cpu(path->nodes[0], &key,
3543 path->slots[0] - 1);
3544 if (key.type == BTRFS_EXTENT_DATA_KEY)
3545 path->slots[0]--;
3546 }
3547
3548 nritems = btrfs_header_nritems(path->nodes[0]);
3549 process_slot:
3550 no_quota = 1;
3551 if (path->slots[0] >= nritems) {
3552 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3553 if (ret < 0)
3554 goto out;
3555 if (ret > 0)
3556 break;
3557 nritems = btrfs_header_nritems(path->nodes[0]);
3558 }
3559 leaf = path->nodes[0];
3560 slot = path->slots[0];
3561
3562 btrfs_item_key_to_cpu(leaf, &key, slot);
3563 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3564 key.objectid != btrfs_ino(src))
3565 break;
3566
3567 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3568 struct btrfs_file_extent_item *extent;
3569 int type;
3570 u32 size;
3571 struct btrfs_key new_key;
3572 u64 disko = 0, diskl = 0;
3573 u64 datao = 0, datal = 0;
3574 u8 comp;
3575 u64 drop_start;
3576
3577 extent = btrfs_item_ptr(leaf, slot,
3578 struct btrfs_file_extent_item);
3579 comp = btrfs_file_extent_compression(leaf, extent);
3580 type = btrfs_file_extent_type(leaf, extent);
3581 if (type == BTRFS_FILE_EXTENT_REG ||
3582 type == BTRFS_FILE_EXTENT_PREALLOC) {
3583 disko = btrfs_file_extent_disk_bytenr(leaf,
3584 extent);
3585 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3586 extent);
3587 datao = btrfs_file_extent_offset(leaf, extent);
3588 datal = btrfs_file_extent_num_bytes(leaf,
3589 extent);
3590 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3591 /* take upper bound, may be compressed */
3592 datal = btrfs_file_extent_ram_bytes(leaf,
3593 extent);
3594 }
3595
3596 /*
3597 * The first search might have left us at an extent
3598 * item that ends before our target range's start, can
3599 * happen if we have holes and NO_HOLES feature enabled.
3600 */
3601 if (key.offset + datal <= off) {
3602 path->slots[0]++;
3603 goto process_slot;
3604 } else if (key.offset >= off + len) {
3605 break;
3606 }
3607 next_key_min_offset = key.offset + datal;
3608 size = btrfs_item_size_nr(leaf, slot);
3609 read_extent_buffer(leaf, buf,
3610 btrfs_item_ptr_offset(leaf, slot),
3611 size);
3612
3613 btrfs_release_path(path);
3614 path->leave_spinning = 0;
3615
3616 memcpy(&new_key, &key, sizeof(new_key));
3617 new_key.objectid = btrfs_ino(inode);
3618 if (off <= key.offset)
3619 new_key.offset = key.offset + destoff - off;
3620 else
3621 new_key.offset = destoff;
3622
3623 /*
3624 * Deal with a hole that doesn't have an extent item
3625 * that represents it (NO_HOLES feature enabled).
3626 * This hole is either in the middle of the cloning
3627 * range or at the beginning (fully overlaps it or
3628 * partially overlaps it).
3629 */
3630 if (new_key.offset != last_dest_end)
3631 drop_start = last_dest_end;
3632 else
3633 drop_start = new_key.offset;
3634
3635 /*
3636 * 1 - adjusting old extent (we may have to split it)
3637 * 1 - add new extent
3638 * 1 - inode update
3639 */
3640 trans = btrfs_start_transaction(root, 3);
3641 if (IS_ERR(trans)) {
3642 ret = PTR_ERR(trans);
3643 goto out;
3644 }
3645
3646 if (type == BTRFS_FILE_EXTENT_REG ||
3647 type == BTRFS_FILE_EXTENT_PREALLOC) {
3648 /*
3649 * a | --- range to clone ---| b
3650 * | ------------- extent ------------- |
3651 */
3652
3653 /* subtract range b */
3654 if (key.offset + datal > off + len)
3655 datal = off + len - key.offset;
3656
3657 /* subtract range a */
3658 if (off > key.offset) {
3659 datao += off - key.offset;
3660 datal -= off - key.offset;
3661 }
3662
3663 ret = btrfs_drop_extents(trans, root, inode,
3664 drop_start,
3665 new_key.offset + datal,
3666 1);
3667 if (ret) {
3668 if (ret != -EOPNOTSUPP)
3669 btrfs_abort_transaction(trans,
3670 root, ret);
3671 btrfs_end_transaction(trans, root);
3672 goto out;
3673 }
3674
3675 ret = btrfs_insert_empty_item(trans, root, path,
3676 &new_key, size);
3677 if (ret) {
3678 btrfs_abort_transaction(trans, root,
3679 ret);
3680 btrfs_end_transaction(trans, root);
3681 goto out;
3682 }
3683
3684 leaf = path->nodes[0];
3685 slot = path->slots[0];
3686 write_extent_buffer(leaf, buf,
3687 btrfs_item_ptr_offset(leaf, slot),
3688 size);
3689
3690 extent = btrfs_item_ptr(leaf, slot,
3691 struct btrfs_file_extent_item);
3692
3693 /* disko == 0 means it's a hole */
3694 if (!disko)
3695 datao = 0;
3696
3697 btrfs_set_file_extent_offset(leaf, extent,
3698 datao);
3699 btrfs_set_file_extent_num_bytes(leaf, extent,
3700 datal);
3701
3702 /*
3703 * We need to look up the roots that point at
3704 * this bytenr and see if the new root does. If
3705 * it does not we need to make sure we update
3706 * quotas appropriately.
3707 */
3708 if (disko && root != BTRFS_I(src)->root &&
3709 disko != last_disko) {
3710 no_quota = check_ref(trans, root,
3711 disko);
3712 if (no_quota < 0) {
3713 btrfs_abort_transaction(trans,
3714 root,
3715 ret);
3716 btrfs_end_transaction(trans,
3717 root);
3718 ret = no_quota;
3719 goto out;
3720 }
3721 }
3722
3723 if (disko) {
3724 inode_add_bytes(inode, datal);
3725 ret = btrfs_inc_extent_ref(trans, root,
3726 disko, diskl, 0,
3727 root->root_key.objectid,
3728 btrfs_ino(inode),
3729 new_key.offset - datao,
3730 no_quota);
3731 if (ret) {
3732 btrfs_abort_transaction(trans,
3733 root,
3734 ret);
3735 btrfs_end_transaction(trans,
3736 root);
3737 goto out;
3738
3739 }
3740 }
3741 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3742 u64 skip = 0;
3743 u64 trim = 0;
3744
3745 if (off > key.offset) {
3746 skip = off - key.offset;
3747 new_key.offset += skip;
3748 }
3749
3750 if (key.offset + datal > off + len)
3751 trim = key.offset + datal - (off + len);
3752
3753 if (comp && (skip || trim)) {
3754 ret = -EINVAL;
3755 btrfs_end_transaction(trans, root);
3756 goto out;
3757 }
3758 size -= skip + trim;
3759 datal -= skip + trim;
3760
3761 ret = clone_copy_inline_extent(src, inode,
3762 trans, path,
3763 &new_key,
3764 drop_start,
3765 datal,
3766 skip, size, buf);
3767 if (ret) {
3768 if (ret != -EOPNOTSUPP)
3769 btrfs_abort_transaction(trans,
3770 root,
3771 ret);
3772 btrfs_end_transaction(trans, root);
3773 goto out;
3774 }
3775 leaf = path->nodes[0];
3776 slot = path->slots[0];
3777 }
3778
3779 /* If we have an implicit hole (NO_HOLES feature). */
3780 if (drop_start < new_key.offset)
3781 clone_update_extent_map(inode, trans,
3782 NULL, drop_start,
3783 new_key.offset - drop_start);
3784
3785 clone_update_extent_map(inode, trans, path, 0, 0);
3786
3787 btrfs_mark_buffer_dirty(leaf);
3788 btrfs_release_path(path);
3789
3790 last_dest_end = ALIGN(new_key.offset + datal,
3791 root->sectorsize);
3792 ret = clone_finish_inode_update(trans, inode,
3793 last_dest_end,
3794 destoff, olen,
3795 no_time_update);
3796 if (ret)
3797 goto out;
3798 if (new_key.offset + datal >= destoff + len)
3799 break;
3800 }
3801 btrfs_release_path(path);
3802 key.offset = next_key_min_offset;
3803 }
3804 ret = 0;
3805
3806 if (last_dest_end < destoff + len) {
3807 /*
3808 * We have an implicit hole (NO_HOLES feature is enabled) that
3809 * fully or partially overlaps our cloning range at its end.
3810 */
3811 btrfs_release_path(path);
3812
3813 /*
3814 * 1 - remove extent(s)
3815 * 1 - inode update
3816 */
3817 trans = btrfs_start_transaction(root, 2);
3818 if (IS_ERR(trans)) {
3819 ret = PTR_ERR(trans);
3820 goto out;
3821 }
3822 ret = btrfs_drop_extents(trans, root, inode,
3823 last_dest_end, destoff + len, 1);
3824 if (ret) {
3825 if (ret != -EOPNOTSUPP)
3826 btrfs_abort_transaction(trans, root, ret);
3827 btrfs_end_transaction(trans, root);
3828 goto out;
3829 }
3830 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3831 destoff + len - last_dest_end);
3832 ret = clone_finish_inode_update(trans, inode, destoff + len,
3833 destoff, olen, no_time_update);
3834 }
3835
3836 out:
3837 btrfs_free_path(path);
3838 vfree(buf);
3839 return ret;
3840 }
3841
3842 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3843 u64 off, u64 olen, u64 destoff)
3844 {
3845 struct inode *inode = file_inode(file);
3846 struct btrfs_root *root = BTRFS_I(inode)->root;
3847 struct fd src_file;
3848 struct inode *src;
3849 int ret;
3850 u64 len = olen;
3851 u64 bs = root->fs_info->sb->s_blocksize;
3852 int same_inode = 0;
3853
3854 /*
3855 * TODO:
3856 * - split compressed inline extents. annoying: we need to
3857 * decompress into destination's address_space (the file offset
3858 * may change, so source mapping won't do), then recompress (or
3859 * otherwise reinsert) a subrange.
3860 *
3861 * - split destination inode's inline extents. The inline extents can
3862 * be either compressed or non-compressed.
3863 */
3864
3865 /* the destination must be opened for writing */
3866 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3867 return -EINVAL;
3868
3869 if (btrfs_root_readonly(root))
3870 return -EROFS;
3871
3872 ret = mnt_want_write_file(file);
3873 if (ret)
3874 return ret;
3875
3876 src_file = fdget(srcfd);
3877 if (!src_file.file) {
3878 ret = -EBADF;
3879 goto out_drop_write;
3880 }
3881
3882 ret = -EXDEV;
3883 if (src_file.file->f_path.mnt != file->f_path.mnt)
3884 goto out_fput;
3885
3886 src = file_inode(src_file.file);
3887
3888 ret = -EINVAL;
3889 if (src == inode)
3890 same_inode = 1;
3891
3892 /* the src must be open for reading */
3893 if (!(src_file.file->f_mode & FMODE_READ))
3894 goto out_fput;
3895
3896 /* don't make the dst file partly checksummed */
3897 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3898 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3899 goto out_fput;
3900
3901 ret = -EISDIR;
3902 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3903 goto out_fput;
3904
3905 ret = -EXDEV;
3906 if (src->i_sb != inode->i_sb)
3907 goto out_fput;
3908
3909 if (!same_inode) {
3910 btrfs_double_inode_lock(src, inode);
3911 } else {
3912 mutex_lock(&src->i_mutex);
3913 }
3914
3915 /* determine range to clone */
3916 ret = -EINVAL;
3917 if (off + len > src->i_size || off + len < off)
3918 goto out_unlock;
3919 if (len == 0)
3920 olen = len = src->i_size - off;
3921 /* if we extend to eof, continue to block boundary */
3922 if (off + len == src->i_size)
3923 len = ALIGN(src->i_size, bs) - off;
3924
3925 if (len == 0) {
3926 ret = 0;
3927 goto out_unlock;
3928 }
3929
3930 /* verify the end result is block aligned */
3931 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3932 !IS_ALIGNED(destoff, bs))
3933 goto out_unlock;
3934
3935 /* verify if ranges are overlapped within the same file */
3936 if (same_inode) {
3937 if (destoff + len > off && destoff < off + len)
3938 goto out_unlock;
3939 }
3940
3941 if (destoff > inode->i_size) {
3942 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3943 if (ret)
3944 goto out_unlock;
3945 }
3946
3947 /*
3948 * Lock the target range too. Right after we replace the file extent
3949 * items in the fs tree (which now point to the cloned data), we might
3950 * have a worker replace them with extent items relative to a write
3951 * operation that was issued before this clone operation (i.e. confront
3952 * with inode.c:btrfs_finish_ordered_io).
3953 */
3954 if (same_inode) {
3955 u64 lock_start = min_t(u64, off, destoff);
3956 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3957
3958 lock_extent_range(src, lock_start, lock_len);
3959 } else {
3960 btrfs_double_extent_lock(src, off, inode, destoff, len);
3961 }
3962
3963 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3964
3965 if (same_inode) {
3966 u64 lock_start = min_t(u64, off, destoff);
3967 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3968
3969 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3970 } else {
3971 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3972 }
3973 /*
3974 * Truncate page cache pages so that future reads will see the cloned
3975 * data immediately and not the previous data.
3976 */
3977 truncate_inode_pages_range(&inode->i_data, destoff,
3978 PAGE_CACHE_ALIGN(destoff + len) - 1);
3979 out_unlock:
3980 if (!same_inode)
3981 btrfs_double_inode_unlock(src, inode);
3982 else
3983 mutex_unlock(&src->i_mutex);
3984 out_fput:
3985 fdput(src_file);
3986 out_drop_write:
3987 mnt_drop_write_file(file);
3988 return ret;
3989 }
3990
3991 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3992 {
3993 struct btrfs_ioctl_clone_range_args args;
3994
3995 if (copy_from_user(&args, argp, sizeof(args)))
3996 return -EFAULT;
3997 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3998 args.src_length, args.dest_offset);
3999 }
4000
4001 /*
4002 * there are many ways the trans_start and trans_end ioctls can lead
4003 * to deadlocks. They should only be used by applications that
4004 * basically own the machine, and have a very in depth understanding
4005 * of all the possible deadlocks and enospc problems.
4006 */
4007 static long btrfs_ioctl_trans_start(struct file *file)
4008 {
4009 struct inode *inode = file_inode(file);
4010 struct btrfs_root *root = BTRFS_I(inode)->root;
4011 struct btrfs_trans_handle *trans;
4012 int ret;
4013
4014 ret = -EPERM;
4015 if (!capable(CAP_SYS_ADMIN))
4016 goto out;
4017
4018 ret = -EINPROGRESS;
4019 if (file->private_data)
4020 goto out;
4021
4022 ret = -EROFS;
4023 if (btrfs_root_readonly(root))
4024 goto out;
4025
4026 ret = mnt_want_write_file(file);
4027 if (ret)
4028 goto out;
4029
4030 atomic_inc(&root->fs_info->open_ioctl_trans);
4031
4032 ret = -ENOMEM;
4033 trans = btrfs_start_ioctl_transaction(root);
4034 if (IS_ERR(trans))
4035 goto out_drop;
4036
4037 file->private_data = trans;
4038 return 0;
4039
4040 out_drop:
4041 atomic_dec(&root->fs_info->open_ioctl_trans);
4042 mnt_drop_write_file(file);
4043 out:
4044 return ret;
4045 }
4046
4047 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
4048 {
4049 struct inode *inode = file_inode(file);
4050 struct btrfs_root *root = BTRFS_I(inode)->root;
4051 struct btrfs_root *new_root;
4052 struct btrfs_dir_item *di;
4053 struct btrfs_trans_handle *trans;
4054 struct btrfs_path *path;
4055 struct btrfs_key location;
4056 struct btrfs_disk_key disk_key;
4057 u64 objectid = 0;
4058 u64 dir_id;
4059 int ret;
4060
4061 if (!capable(CAP_SYS_ADMIN))
4062 return -EPERM;
4063
4064 ret = mnt_want_write_file(file);
4065 if (ret)
4066 return ret;
4067
4068 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4069 ret = -EFAULT;
4070 goto out;
4071 }
4072
4073 if (!objectid)
4074 objectid = BTRFS_FS_TREE_OBJECTID;
4075
4076 location.objectid = objectid;
4077 location.type = BTRFS_ROOT_ITEM_KEY;
4078 location.offset = (u64)-1;
4079
4080 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
4081 if (IS_ERR(new_root)) {
4082 ret = PTR_ERR(new_root);
4083 goto out;
4084 }
4085
4086 path = btrfs_alloc_path();
4087 if (!path) {
4088 ret = -ENOMEM;
4089 goto out;
4090 }
4091 path->leave_spinning = 1;
4092
4093 trans = btrfs_start_transaction(root, 1);
4094 if (IS_ERR(trans)) {
4095 btrfs_free_path(path);
4096 ret = PTR_ERR(trans);
4097 goto out;
4098 }
4099
4100 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
4101 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
4102 dir_id, "default", 7, 1);
4103 if (IS_ERR_OR_NULL(di)) {
4104 btrfs_free_path(path);
4105 btrfs_end_transaction(trans, root);
4106 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
4107 "item, this isn't going to work");
4108 ret = -ENOENT;
4109 goto out;
4110 }
4111
4112 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4113 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4114 btrfs_mark_buffer_dirty(path->nodes[0]);
4115 btrfs_free_path(path);
4116
4117 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
4118 btrfs_end_transaction(trans, root);
4119 out:
4120 mnt_drop_write_file(file);
4121 return ret;
4122 }
4123
4124 void btrfs_get_block_group_info(struct list_head *groups_list,
4125 struct btrfs_ioctl_space_info *space)
4126 {
4127 struct btrfs_block_group_cache *block_group;
4128
4129 space->total_bytes = 0;
4130 space->used_bytes = 0;
4131 space->flags = 0;
4132 list_for_each_entry(block_group, groups_list, list) {
4133 space->flags = block_group->flags;
4134 space->total_bytes += block_group->key.offset;
4135 space->used_bytes +=
4136 btrfs_block_group_used(&block_group->item);
4137 }
4138 }
4139
4140 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
4141 {
4142 struct btrfs_ioctl_space_args space_args;
4143 struct btrfs_ioctl_space_info space;
4144 struct btrfs_ioctl_space_info *dest;
4145 struct btrfs_ioctl_space_info *dest_orig;
4146 struct btrfs_ioctl_space_info __user *user_dest;
4147 struct btrfs_space_info *info;
4148 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4149 BTRFS_BLOCK_GROUP_SYSTEM,
4150 BTRFS_BLOCK_GROUP_METADATA,
4151 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4152 int num_types = 4;
4153 int alloc_size;
4154 int ret = 0;
4155 u64 slot_count = 0;
4156 int i, c;
4157
4158 if (copy_from_user(&space_args,
4159 (struct btrfs_ioctl_space_args __user *)arg,
4160 sizeof(space_args)))
4161 return -EFAULT;
4162
4163 for (i = 0; i < num_types; i++) {
4164 struct btrfs_space_info *tmp;
4165
4166 info = NULL;
4167 rcu_read_lock();
4168 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4169 list) {
4170 if (tmp->flags == types[i]) {
4171 info = tmp;
4172 break;
4173 }
4174 }
4175 rcu_read_unlock();
4176
4177 if (!info)
4178 continue;
4179
4180 down_read(&info->groups_sem);
4181 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4182 if (!list_empty(&info->block_groups[c]))
4183 slot_count++;
4184 }
4185 up_read(&info->groups_sem);
4186 }
4187
4188 /*
4189 * Global block reserve, exported as a space_info
4190 */
4191 slot_count++;
4192
4193 /* space_slots == 0 means they are asking for a count */
4194 if (space_args.space_slots == 0) {
4195 space_args.total_spaces = slot_count;
4196 goto out;
4197 }
4198
4199 slot_count = min_t(u64, space_args.space_slots, slot_count);
4200
4201 alloc_size = sizeof(*dest) * slot_count;
4202
4203 /* we generally have at most 6 or so space infos, one for each raid
4204 * level. So, a whole page should be more than enough for everyone
4205 */
4206 if (alloc_size > PAGE_CACHE_SIZE)
4207 return -ENOMEM;
4208
4209 space_args.total_spaces = 0;
4210 dest = kmalloc(alloc_size, GFP_NOFS);
4211 if (!dest)
4212 return -ENOMEM;
4213 dest_orig = dest;
4214
4215 /* now we have a buffer to copy into */
4216 for (i = 0; i < num_types; i++) {
4217 struct btrfs_space_info *tmp;
4218
4219 if (!slot_count)
4220 break;
4221
4222 info = NULL;
4223 rcu_read_lock();
4224 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4225 list) {
4226 if (tmp->flags == types[i]) {
4227 info = tmp;
4228 break;
4229 }
4230 }
4231 rcu_read_unlock();
4232
4233 if (!info)
4234 continue;
4235 down_read(&info->groups_sem);
4236 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4237 if (!list_empty(&info->block_groups[c])) {
4238 btrfs_get_block_group_info(
4239 &info->block_groups[c], &space);
4240 memcpy(dest, &space, sizeof(space));
4241 dest++;
4242 space_args.total_spaces++;
4243 slot_count--;
4244 }
4245 if (!slot_count)
4246 break;
4247 }
4248 up_read(&info->groups_sem);
4249 }
4250
4251 /*
4252 * Add global block reserve
4253 */
4254 if (slot_count) {
4255 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4256
4257 spin_lock(&block_rsv->lock);
4258 space.total_bytes = block_rsv->size;
4259 space.used_bytes = block_rsv->size - block_rsv->reserved;
4260 spin_unlock(&block_rsv->lock);
4261 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4262 memcpy(dest, &space, sizeof(space));
4263 space_args.total_spaces++;
4264 }
4265
4266 user_dest = (struct btrfs_ioctl_space_info __user *)
4267 (arg + sizeof(struct btrfs_ioctl_space_args));
4268
4269 if (copy_to_user(user_dest, dest_orig, alloc_size))
4270 ret = -EFAULT;
4271
4272 kfree(dest_orig);
4273 out:
4274 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4275 ret = -EFAULT;
4276
4277 return ret;
4278 }
4279
4280 /*
4281 * there are many ways the trans_start and trans_end ioctls can lead
4282 * to deadlocks. They should only be used by applications that
4283 * basically own the machine, and have a very in depth understanding
4284 * of all the possible deadlocks and enospc problems.
4285 */
4286 long btrfs_ioctl_trans_end(struct file *file)
4287 {
4288 struct inode *inode = file_inode(file);
4289 struct btrfs_root *root = BTRFS_I(inode)->root;
4290 struct btrfs_trans_handle *trans;
4291
4292 trans = file->private_data;
4293 if (!trans)
4294 return -EINVAL;
4295 file->private_data = NULL;
4296
4297 btrfs_end_transaction(trans, root);
4298
4299 atomic_dec(&root->fs_info->open_ioctl_trans);
4300
4301 mnt_drop_write_file(file);
4302 return 0;
4303 }
4304
4305 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4306 void __user *argp)
4307 {
4308 struct btrfs_trans_handle *trans;
4309 u64 transid;
4310 int ret;
4311
4312 trans = btrfs_attach_transaction_barrier(root);
4313 if (IS_ERR(trans)) {
4314 if (PTR_ERR(trans) != -ENOENT)
4315 return PTR_ERR(trans);
4316
4317 /* No running transaction, don't bother */
4318 transid = root->fs_info->last_trans_committed;
4319 goto out;
4320 }
4321 transid = trans->transid;
4322 ret = btrfs_commit_transaction_async(trans, root, 0);
4323 if (ret) {
4324 btrfs_end_transaction(trans, root);
4325 return ret;
4326 }
4327 out:
4328 if (argp)
4329 if (copy_to_user(argp, &transid, sizeof(transid)))
4330 return -EFAULT;
4331 return 0;
4332 }
4333
4334 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4335 void __user *argp)
4336 {
4337 u64 transid;
4338
4339 if (argp) {
4340 if (copy_from_user(&transid, argp, sizeof(transid)))
4341 return -EFAULT;
4342 } else {
4343 transid = 0; /* current trans */
4344 }
4345 return btrfs_wait_for_commit(root, transid);
4346 }
4347
4348 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4349 {
4350 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4351 struct btrfs_ioctl_scrub_args *sa;
4352 int ret;
4353
4354 if (!capable(CAP_SYS_ADMIN))
4355 return -EPERM;
4356
4357 sa = memdup_user(arg, sizeof(*sa));
4358 if (IS_ERR(sa))
4359 return PTR_ERR(sa);
4360
4361 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4362 ret = mnt_want_write_file(file);
4363 if (ret)
4364 goto out;
4365 }
4366
4367 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4368 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4369 0);
4370
4371 if (copy_to_user(arg, sa, sizeof(*sa)))
4372 ret = -EFAULT;
4373
4374 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4375 mnt_drop_write_file(file);
4376 out:
4377 kfree(sa);
4378 return ret;
4379 }
4380
4381 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4382 {
4383 if (!capable(CAP_SYS_ADMIN))
4384 return -EPERM;
4385
4386 return btrfs_scrub_cancel(root->fs_info);
4387 }
4388
4389 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4390 void __user *arg)
4391 {
4392 struct btrfs_ioctl_scrub_args *sa;
4393 int ret;
4394
4395 if (!capable(CAP_SYS_ADMIN))
4396 return -EPERM;
4397
4398 sa = memdup_user(arg, sizeof(*sa));
4399 if (IS_ERR(sa))
4400 return PTR_ERR(sa);
4401
4402 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4403
4404 if (copy_to_user(arg, sa, sizeof(*sa)))
4405 ret = -EFAULT;
4406
4407 kfree(sa);
4408 return ret;
4409 }
4410
4411 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4412 void __user *arg)
4413 {
4414 struct btrfs_ioctl_get_dev_stats *sa;
4415 int ret;
4416
4417 sa = memdup_user(arg, sizeof(*sa));
4418 if (IS_ERR(sa))
4419 return PTR_ERR(sa);
4420
4421 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4422 kfree(sa);
4423 return -EPERM;
4424 }
4425
4426 ret = btrfs_get_dev_stats(root, sa);
4427
4428 if (copy_to_user(arg, sa, sizeof(*sa)))
4429 ret = -EFAULT;
4430
4431 kfree(sa);
4432 return ret;
4433 }
4434
4435 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4436 {
4437 struct btrfs_ioctl_dev_replace_args *p;
4438 int ret;
4439
4440 if (!capable(CAP_SYS_ADMIN))
4441 return -EPERM;
4442
4443 p = memdup_user(arg, sizeof(*p));
4444 if (IS_ERR(p))
4445 return PTR_ERR(p);
4446
4447 switch (p->cmd) {
4448 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4449 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4450 ret = -EROFS;
4451 goto out;
4452 }
4453 if (atomic_xchg(
4454 &root->fs_info->mutually_exclusive_operation_running,
4455 1)) {
4456 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4457 } else {
4458 ret = btrfs_dev_replace_start(root, p);
4459 atomic_set(
4460 &root->fs_info->mutually_exclusive_operation_running,
4461 0);
4462 }
4463 break;
4464 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4465 btrfs_dev_replace_status(root->fs_info, p);
4466 ret = 0;
4467 break;
4468 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4469 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4470 break;
4471 default:
4472 ret = -EINVAL;
4473 break;
4474 }
4475
4476 if (copy_to_user(arg, p, sizeof(*p)))
4477 ret = -EFAULT;
4478 out:
4479 kfree(p);
4480 return ret;
4481 }
4482
4483 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4484 {
4485 int ret = 0;
4486 int i;
4487 u64 rel_ptr;
4488 int size;
4489 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4490 struct inode_fs_paths *ipath = NULL;
4491 struct btrfs_path *path;
4492
4493 if (!capable(CAP_DAC_READ_SEARCH))
4494 return -EPERM;
4495
4496 path = btrfs_alloc_path();
4497 if (!path) {
4498 ret = -ENOMEM;
4499 goto out;
4500 }
4501
4502 ipa = memdup_user(arg, sizeof(*ipa));
4503 if (IS_ERR(ipa)) {
4504 ret = PTR_ERR(ipa);
4505 ipa = NULL;
4506 goto out;
4507 }
4508
4509 size = min_t(u32, ipa->size, 4096);
4510 ipath = init_ipath(size, root, path);
4511 if (IS_ERR(ipath)) {
4512 ret = PTR_ERR(ipath);
4513 ipath = NULL;
4514 goto out;
4515 }
4516
4517 ret = paths_from_inode(ipa->inum, ipath);
4518 if (ret < 0)
4519 goto out;
4520
4521 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4522 rel_ptr = ipath->fspath->val[i] -
4523 (u64)(unsigned long)ipath->fspath->val;
4524 ipath->fspath->val[i] = rel_ptr;
4525 }
4526
4527 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4528 (void *)(unsigned long)ipath->fspath, size);
4529 if (ret) {
4530 ret = -EFAULT;
4531 goto out;
4532 }
4533
4534 out:
4535 btrfs_free_path(path);
4536 free_ipath(ipath);
4537 kfree(ipa);
4538
4539 return ret;
4540 }
4541
4542 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4543 {
4544 struct btrfs_data_container *inodes = ctx;
4545 const size_t c = 3 * sizeof(u64);
4546
4547 if (inodes->bytes_left >= c) {
4548 inodes->bytes_left -= c;
4549 inodes->val[inodes->elem_cnt] = inum;
4550 inodes->val[inodes->elem_cnt + 1] = offset;
4551 inodes->val[inodes->elem_cnt + 2] = root;
4552 inodes->elem_cnt += 3;
4553 } else {
4554 inodes->bytes_missing += c - inodes->bytes_left;
4555 inodes->bytes_left = 0;
4556 inodes->elem_missed += 3;
4557 }
4558
4559 return 0;
4560 }
4561
4562 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4563 void __user *arg)
4564 {
4565 int ret = 0;
4566 int size;
4567 struct btrfs_ioctl_logical_ino_args *loi;
4568 struct btrfs_data_container *inodes = NULL;
4569 struct btrfs_path *path = NULL;
4570
4571 if (!capable(CAP_SYS_ADMIN))
4572 return -EPERM;
4573
4574 loi = memdup_user(arg, sizeof(*loi));
4575 if (IS_ERR(loi)) {
4576 ret = PTR_ERR(loi);
4577 loi = NULL;
4578 goto out;
4579 }
4580
4581 path = btrfs_alloc_path();
4582 if (!path) {
4583 ret = -ENOMEM;
4584 goto out;
4585 }
4586
4587 size = min_t(u32, loi->size, 64 * 1024);
4588 inodes = init_data_container(size);
4589 if (IS_ERR(inodes)) {
4590 ret = PTR_ERR(inodes);
4591 inodes = NULL;
4592 goto out;
4593 }
4594
4595 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4596 build_ino_list, inodes);
4597 if (ret == -EINVAL)
4598 ret = -ENOENT;
4599 if (ret < 0)
4600 goto out;
4601
4602 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4603 (void *)(unsigned long)inodes, size);
4604 if (ret)
4605 ret = -EFAULT;
4606
4607 out:
4608 btrfs_free_path(path);
4609 vfree(inodes);
4610 kfree(loi);
4611
4612 return ret;
4613 }
4614
4615 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4616 struct btrfs_ioctl_balance_args *bargs)
4617 {
4618 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4619
4620 bargs->flags = bctl->flags;
4621
4622 if (atomic_read(&fs_info->balance_running))
4623 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4624 if (atomic_read(&fs_info->balance_pause_req))
4625 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4626 if (atomic_read(&fs_info->balance_cancel_req))
4627 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4628
4629 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4630 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4631 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4632
4633 if (lock) {
4634 spin_lock(&fs_info->balance_lock);
4635 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4636 spin_unlock(&fs_info->balance_lock);
4637 } else {
4638 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4639 }
4640 }
4641
4642 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4643 {
4644 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4645 struct btrfs_fs_info *fs_info = root->fs_info;
4646 struct btrfs_ioctl_balance_args *bargs;
4647 struct btrfs_balance_control *bctl;
4648 bool need_unlock; /* for mut. excl. ops lock */
4649 int ret;
4650
4651 if (!capable(CAP_SYS_ADMIN))
4652 return -EPERM;
4653
4654 ret = mnt_want_write_file(file);
4655 if (ret)
4656 return ret;
4657
4658 again:
4659 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4660 mutex_lock(&fs_info->volume_mutex);
4661 mutex_lock(&fs_info->balance_mutex);
4662 need_unlock = true;
4663 goto locked;
4664 }
4665
4666 /*
4667 * mut. excl. ops lock is locked. Three possibilites:
4668 * (1) some other op is running
4669 * (2) balance is running
4670 * (3) balance is paused -- special case (think resume)
4671 */
4672 mutex_lock(&fs_info->balance_mutex);
4673 if (fs_info->balance_ctl) {
4674 /* this is either (2) or (3) */
4675 if (!atomic_read(&fs_info->balance_running)) {
4676 mutex_unlock(&fs_info->balance_mutex);
4677 if (!mutex_trylock(&fs_info->volume_mutex))
4678 goto again;
4679 mutex_lock(&fs_info->balance_mutex);
4680
4681 if (fs_info->balance_ctl &&
4682 !atomic_read(&fs_info->balance_running)) {
4683 /* this is (3) */
4684 need_unlock = false;
4685 goto locked;
4686 }
4687
4688 mutex_unlock(&fs_info->balance_mutex);
4689 mutex_unlock(&fs_info->volume_mutex);
4690 goto again;
4691 } else {
4692 /* this is (2) */
4693 mutex_unlock(&fs_info->balance_mutex);
4694 ret = -EINPROGRESS;
4695 goto out;
4696 }
4697 } else {
4698 /* this is (1) */
4699 mutex_unlock(&fs_info->balance_mutex);
4700 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4701 goto out;
4702 }
4703
4704 locked:
4705 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4706
4707 if (arg) {
4708 bargs = memdup_user(arg, sizeof(*bargs));
4709 if (IS_ERR(bargs)) {
4710 ret = PTR_ERR(bargs);
4711 goto out_unlock;
4712 }
4713
4714 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4715 if (!fs_info->balance_ctl) {
4716 ret = -ENOTCONN;
4717 goto out_bargs;
4718 }
4719
4720 bctl = fs_info->balance_ctl;
4721 spin_lock(&fs_info->balance_lock);
4722 bctl->flags |= BTRFS_BALANCE_RESUME;
4723 spin_unlock(&fs_info->balance_lock);
4724
4725 goto do_balance;
4726 }
4727 } else {
4728 bargs = NULL;
4729 }
4730
4731 if (fs_info->balance_ctl) {
4732 ret = -EINPROGRESS;
4733 goto out_bargs;
4734 }
4735
4736 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4737 if (!bctl) {
4738 ret = -ENOMEM;
4739 goto out_bargs;
4740 }
4741
4742 bctl->fs_info = fs_info;
4743 if (arg) {
4744 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4745 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4746 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4747
4748 bctl->flags = bargs->flags;
4749 } else {
4750 /* balance everything - no filters */
4751 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4752 }
4753
4754 do_balance:
4755 /*
4756 * Ownership of bctl and mutually_exclusive_operation_running
4757 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4758 * or, if restriper was paused all the way until unmount, in
4759 * free_fs_info. mutually_exclusive_operation_running is
4760 * cleared in __cancel_balance.
4761 */
4762 need_unlock = false;
4763
4764 ret = btrfs_balance(bctl, bargs);
4765
4766 if (arg) {
4767 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4768 ret = -EFAULT;
4769 }
4770
4771 out_bargs:
4772 kfree(bargs);
4773 out_unlock:
4774 mutex_unlock(&fs_info->balance_mutex);
4775 mutex_unlock(&fs_info->volume_mutex);
4776 if (need_unlock)
4777 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4778 out:
4779 mnt_drop_write_file(file);
4780 return ret;
4781 }
4782
4783 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4784 {
4785 if (!capable(CAP_SYS_ADMIN))
4786 return -EPERM;
4787
4788 switch (cmd) {
4789 case BTRFS_BALANCE_CTL_PAUSE:
4790 return btrfs_pause_balance(root->fs_info);
4791 case BTRFS_BALANCE_CTL_CANCEL:
4792 return btrfs_cancel_balance(root->fs_info);
4793 }
4794
4795 return -EINVAL;
4796 }
4797
4798 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4799 void __user *arg)
4800 {
4801 struct btrfs_fs_info *fs_info = root->fs_info;
4802 struct btrfs_ioctl_balance_args *bargs;
4803 int ret = 0;
4804
4805 if (!capable(CAP_SYS_ADMIN))
4806 return -EPERM;
4807
4808 mutex_lock(&fs_info->balance_mutex);
4809 if (!fs_info->balance_ctl) {
4810 ret = -ENOTCONN;
4811 goto out;
4812 }
4813
4814 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4815 if (!bargs) {
4816 ret = -ENOMEM;
4817 goto out;
4818 }
4819
4820 update_ioctl_balance_args(fs_info, 1, bargs);
4821
4822 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4823 ret = -EFAULT;
4824
4825 kfree(bargs);
4826 out:
4827 mutex_unlock(&fs_info->balance_mutex);
4828 return ret;
4829 }
4830
4831 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4832 {
4833 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4834 struct btrfs_ioctl_quota_ctl_args *sa;
4835 struct btrfs_trans_handle *trans = NULL;
4836 int ret;
4837 int err;
4838
4839 if (!capable(CAP_SYS_ADMIN))
4840 return -EPERM;
4841
4842 ret = mnt_want_write_file(file);
4843 if (ret)
4844 return ret;
4845
4846 sa = memdup_user(arg, sizeof(*sa));
4847 if (IS_ERR(sa)) {
4848 ret = PTR_ERR(sa);
4849 goto drop_write;
4850 }
4851
4852 down_write(&root->fs_info->subvol_sem);
4853 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4854 if (IS_ERR(trans)) {
4855 ret = PTR_ERR(trans);
4856 goto out;
4857 }
4858
4859 switch (sa->cmd) {
4860 case BTRFS_QUOTA_CTL_ENABLE:
4861 ret = btrfs_quota_enable(trans, root->fs_info);
4862 break;
4863 case BTRFS_QUOTA_CTL_DISABLE:
4864 ret = btrfs_quota_disable(trans, root->fs_info);
4865 break;
4866 default:
4867 ret = -EINVAL;
4868 break;
4869 }
4870
4871 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4872 if (err && !ret)
4873 ret = err;
4874 out:
4875 kfree(sa);
4876 up_write(&root->fs_info->subvol_sem);
4877 drop_write:
4878 mnt_drop_write_file(file);
4879 return ret;
4880 }
4881
4882 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4883 {
4884 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4885 struct btrfs_ioctl_qgroup_assign_args *sa;
4886 struct btrfs_trans_handle *trans;
4887 int ret;
4888 int err;
4889
4890 if (!capable(CAP_SYS_ADMIN))
4891 return -EPERM;
4892
4893 ret = mnt_want_write_file(file);
4894 if (ret)
4895 return ret;
4896
4897 sa = memdup_user(arg, sizeof(*sa));
4898 if (IS_ERR(sa)) {
4899 ret = PTR_ERR(sa);
4900 goto drop_write;
4901 }
4902
4903 trans = btrfs_join_transaction(root);
4904 if (IS_ERR(trans)) {
4905 ret = PTR_ERR(trans);
4906 goto out;
4907 }
4908
4909 /* FIXME: check if the IDs really exist */
4910 if (sa->assign) {
4911 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4912 sa->src, sa->dst);
4913 } else {
4914 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4915 sa->src, sa->dst);
4916 }
4917
4918 /* update qgroup status and info */
4919 err = btrfs_run_qgroups(trans, root->fs_info);
4920 if (err < 0)
4921 btrfs_std_error(root->fs_info, ret,
4922 "failed to update qgroup status and info\n");
4923 err = btrfs_end_transaction(trans, root);
4924 if (err && !ret)
4925 ret = err;
4926
4927 out:
4928 kfree(sa);
4929 drop_write:
4930 mnt_drop_write_file(file);
4931 return ret;
4932 }
4933
4934 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4935 {
4936 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4937 struct btrfs_ioctl_qgroup_create_args *sa;
4938 struct btrfs_trans_handle *trans;
4939 int ret;
4940 int err;
4941
4942 if (!capable(CAP_SYS_ADMIN))
4943 return -EPERM;
4944
4945 ret = mnt_want_write_file(file);
4946 if (ret)
4947 return ret;
4948
4949 sa = memdup_user(arg, sizeof(*sa));
4950 if (IS_ERR(sa)) {
4951 ret = PTR_ERR(sa);
4952 goto drop_write;
4953 }
4954
4955 if (!sa->qgroupid) {
4956 ret = -EINVAL;
4957 goto out;
4958 }
4959
4960 trans = btrfs_join_transaction(root);
4961 if (IS_ERR(trans)) {
4962 ret = PTR_ERR(trans);
4963 goto out;
4964 }
4965
4966 /* FIXME: check if the IDs really exist */
4967 if (sa->create) {
4968 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
4969 } else {
4970 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4971 }
4972
4973 err = btrfs_end_transaction(trans, root);
4974 if (err && !ret)
4975 ret = err;
4976
4977 out:
4978 kfree(sa);
4979 drop_write:
4980 mnt_drop_write_file(file);
4981 return ret;
4982 }
4983
4984 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4985 {
4986 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4987 struct btrfs_ioctl_qgroup_limit_args *sa;
4988 struct btrfs_trans_handle *trans;
4989 int ret;
4990 int err;
4991 u64 qgroupid;
4992
4993 if (!capable(CAP_SYS_ADMIN))
4994 return -EPERM;
4995
4996 ret = mnt_want_write_file(file);
4997 if (ret)
4998 return ret;
4999
5000 sa = memdup_user(arg, sizeof(*sa));
5001 if (IS_ERR(sa)) {
5002 ret = PTR_ERR(sa);
5003 goto drop_write;
5004 }
5005
5006 trans = btrfs_join_transaction(root);
5007 if (IS_ERR(trans)) {
5008 ret = PTR_ERR(trans);
5009 goto out;
5010 }
5011
5012 qgroupid = sa->qgroupid;
5013 if (!qgroupid) {
5014 /* take the current subvol as qgroup */
5015 qgroupid = root->root_key.objectid;
5016 }
5017
5018 /* FIXME: check if the IDs really exist */
5019 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
5020
5021 err = btrfs_end_transaction(trans, root);
5022 if (err && !ret)
5023 ret = err;
5024
5025 out:
5026 kfree(sa);
5027 drop_write:
5028 mnt_drop_write_file(file);
5029 return ret;
5030 }
5031
5032 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
5033 {
5034 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5035 struct btrfs_ioctl_quota_rescan_args *qsa;
5036 int ret;
5037
5038 if (!capable(CAP_SYS_ADMIN))
5039 return -EPERM;
5040
5041 ret = mnt_want_write_file(file);
5042 if (ret)
5043 return ret;
5044
5045 qsa = memdup_user(arg, sizeof(*qsa));
5046 if (IS_ERR(qsa)) {
5047 ret = PTR_ERR(qsa);
5048 goto drop_write;
5049 }
5050
5051 if (qsa->flags) {
5052 ret = -EINVAL;
5053 goto out;
5054 }
5055
5056 ret = btrfs_qgroup_rescan(root->fs_info);
5057
5058 out:
5059 kfree(qsa);
5060 drop_write:
5061 mnt_drop_write_file(file);
5062 return ret;
5063 }
5064
5065 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5066 {
5067 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5068 struct btrfs_ioctl_quota_rescan_args *qsa;
5069 int ret = 0;
5070
5071 if (!capable(CAP_SYS_ADMIN))
5072 return -EPERM;
5073
5074 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
5075 if (!qsa)
5076 return -ENOMEM;
5077
5078 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5079 qsa->flags = 1;
5080 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
5081 }
5082
5083 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5084 ret = -EFAULT;
5085
5086 kfree(qsa);
5087 return ret;
5088 }
5089
5090 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5091 {
5092 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5093
5094 if (!capable(CAP_SYS_ADMIN))
5095 return -EPERM;
5096
5097 return btrfs_qgroup_wait_for_completion(root->fs_info);
5098 }
5099
5100 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5101 struct btrfs_ioctl_received_subvol_args *sa)
5102 {
5103 struct inode *inode = file_inode(file);
5104 struct btrfs_root *root = BTRFS_I(inode)->root;
5105 struct btrfs_root_item *root_item = &root->root_item;
5106 struct btrfs_trans_handle *trans;
5107 struct timespec ct = CURRENT_TIME;
5108 int ret = 0;
5109 int received_uuid_changed;
5110
5111 if (!inode_owner_or_capable(inode))
5112 return -EPERM;
5113
5114 ret = mnt_want_write_file(file);
5115 if (ret < 0)
5116 return ret;
5117
5118 down_write(&root->fs_info->subvol_sem);
5119
5120 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
5121 ret = -EINVAL;
5122 goto out;
5123 }
5124
5125 if (btrfs_root_readonly(root)) {
5126 ret = -EROFS;
5127 goto out;
5128 }
5129
5130 /*
5131 * 1 - root item
5132 * 2 - uuid items (received uuid + subvol uuid)
5133 */
5134 trans = btrfs_start_transaction(root, 3);
5135 if (IS_ERR(trans)) {
5136 ret = PTR_ERR(trans);
5137 trans = NULL;
5138 goto out;
5139 }
5140
5141 sa->rtransid = trans->transid;
5142 sa->rtime.sec = ct.tv_sec;
5143 sa->rtime.nsec = ct.tv_nsec;
5144
5145 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5146 BTRFS_UUID_SIZE);
5147 if (received_uuid_changed &&
5148 !btrfs_is_empty_uuid(root_item->received_uuid))
5149 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
5150 root_item->received_uuid,
5151 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5152 root->root_key.objectid);
5153 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5154 btrfs_set_root_stransid(root_item, sa->stransid);
5155 btrfs_set_root_rtransid(root_item, sa->rtransid);
5156 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5157 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5158 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5159 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5160
5161 ret = btrfs_update_root(trans, root->fs_info->tree_root,
5162 &root->root_key, &root->root_item);
5163 if (ret < 0) {
5164 btrfs_end_transaction(trans, root);
5165 goto out;
5166 }
5167 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5168 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
5169 sa->uuid,
5170 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5171 root->root_key.objectid);
5172 if (ret < 0 && ret != -EEXIST) {
5173 btrfs_abort_transaction(trans, root, ret);
5174 goto out;
5175 }
5176 }
5177 ret = btrfs_commit_transaction(trans, root);
5178 if (ret < 0) {
5179 btrfs_abort_transaction(trans, root, ret);
5180 goto out;
5181 }
5182
5183 out:
5184 up_write(&root->fs_info->subvol_sem);
5185 mnt_drop_write_file(file);
5186 return ret;
5187 }
5188
5189 #ifdef CONFIG_64BIT
5190 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5191 void __user *arg)
5192 {
5193 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5194 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5195 int ret = 0;
5196
5197 args32 = memdup_user(arg, sizeof(*args32));
5198 if (IS_ERR(args32)) {
5199 ret = PTR_ERR(args32);
5200 args32 = NULL;
5201 goto out;
5202 }
5203
5204 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
5205 if (!args64) {
5206 ret = -ENOMEM;
5207 goto out;
5208 }
5209
5210 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5211 args64->stransid = args32->stransid;
5212 args64->rtransid = args32->rtransid;
5213 args64->stime.sec = args32->stime.sec;
5214 args64->stime.nsec = args32->stime.nsec;
5215 args64->rtime.sec = args32->rtime.sec;
5216 args64->rtime.nsec = args32->rtime.nsec;
5217 args64->flags = args32->flags;
5218
5219 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5220 if (ret)
5221 goto out;
5222
5223 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5224 args32->stransid = args64->stransid;
5225 args32->rtransid = args64->rtransid;
5226 args32->stime.sec = args64->stime.sec;
5227 args32->stime.nsec = args64->stime.nsec;
5228 args32->rtime.sec = args64->rtime.sec;
5229 args32->rtime.nsec = args64->rtime.nsec;
5230 args32->flags = args64->flags;
5231
5232 ret = copy_to_user(arg, args32, sizeof(*args32));
5233 if (ret)
5234 ret = -EFAULT;
5235
5236 out:
5237 kfree(args32);
5238 kfree(args64);
5239 return ret;
5240 }
5241 #endif
5242
5243 static long btrfs_ioctl_set_received_subvol(struct file *file,
5244 void __user *arg)
5245 {
5246 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5247 int ret = 0;
5248
5249 sa = memdup_user(arg, sizeof(*sa));
5250 if (IS_ERR(sa)) {
5251 ret = PTR_ERR(sa);
5252 sa = NULL;
5253 goto out;
5254 }
5255
5256 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5257
5258 if (ret)
5259 goto out;
5260
5261 ret = copy_to_user(arg, sa, sizeof(*sa));
5262 if (ret)
5263 ret = -EFAULT;
5264
5265 out:
5266 kfree(sa);
5267 return ret;
5268 }
5269
5270 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5271 {
5272 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5273 size_t len;
5274 int ret;
5275 char label[BTRFS_LABEL_SIZE];
5276
5277 spin_lock(&root->fs_info->super_lock);
5278 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5279 spin_unlock(&root->fs_info->super_lock);
5280
5281 len = strnlen(label, BTRFS_LABEL_SIZE);
5282
5283 if (len == BTRFS_LABEL_SIZE) {
5284 btrfs_warn(root->fs_info,
5285 "label is too long, return the first %zu bytes", --len);
5286 }
5287
5288 ret = copy_to_user(arg, label, len);
5289
5290 return ret ? -EFAULT : 0;
5291 }
5292
5293 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5294 {
5295 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5296 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5297 struct btrfs_trans_handle *trans;
5298 char label[BTRFS_LABEL_SIZE];
5299 int ret;
5300
5301 if (!capable(CAP_SYS_ADMIN))
5302 return -EPERM;
5303
5304 if (copy_from_user(label, arg, sizeof(label)))
5305 return -EFAULT;
5306
5307 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5308 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5309 BTRFS_LABEL_SIZE - 1);
5310 return -EINVAL;
5311 }
5312
5313 ret = mnt_want_write_file(file);
5314 if (ret)
5315 return ret;
5316
5317 trans = btrfs_start_transaction(root, 0);
5318 if (IS_ERR(trans)) {
5319 ret = PTR_ERR(trans);
5320 goto out_unlock;
5321 }
5322
5323 spin_lock(&root->fs_info->super_lock);
5324 strcpy(super_block->label, label);
5325 spin_unlock(&root->fs_info->super_lock);
5326 ret = btrfs_commit_transaction(trans, root);
5327
5328 out_unlock:
5329 mnt_drop_write_file(file);
5330 return ret;
5331 }
5332
5333 #define INIT_FEATURE_FLAGS(suffix) \
5334 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5335 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5336 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5337
5338 static int btrfs_ioctl_get_supported_features(struct file *file,
5339 void __user *arg)
5340 {
5341 static struct btrfs_ioctl_feature_flags features[3] = {
5342 INIT_FEATURE_FLAGS(SUPP),
5343 INIT_FEATURE_FLAGS(SAFE_SET),
5344 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5345 };
5346
5347 if (copy_to_user(arg, &features, sizeof(features)))
5348 return -EFAULT;
5349
5350 return 0;
5351 }
5352
5353 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5354 {
5355 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5356 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5357 struct btrfs_ioctl_feature_flags features;
5358
5359 features.compat_flags = btrfs_super_compat_flags(super_block);
5360 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5361 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5362
5363 if (copy_to_user(arg, &features, sizeof(features)))
5364 return -EFAULT;
5365
5366 return 0;
5367 }
5368
5369 static int check_feature_bits(struct btrfs_root *root,
5370 enum btrfs_feature_set set,
5371 u64 change_mask, u64 flags, u64 supported_flags,
5372 u64 safe_set, u64 safe_clear)
5373 {
5374 const char *type = btrfs_feature_set_names[set];
5375 char *names;
5376 u64 disallowed, unsupported;
5377 u64 set_mask = flags & change_mask;
5378 u64 clear_mask = ~flags & change_mask;
5379
5380 unsupported = set_mask & ~supported_flags;
5381 if (unsupported) {
5382 names = btrfs_printable_features(set, unsupported);
5383 if (names) {
5384 btrfs_warn(root->fs_info,
5385 "this kernel does not support the %s feature bit%s",
5386 names, strchr(names, ',') ? "s" : "");
5387 kfree(names);
5388 } else
5389 btrfs_warn(root->fs_info,
5390 "this kernel does not support %s bits 0x%llx",
5391 type, unsupported);
5392 return -EOPNOTSUPP;
5393 }
5394
5395 disallowed = set_mask & ~safe_set;
5396 if (disallowed) {
5397 names = btrfs_printable_features(set, disallowed);
5398 if (names) {
5399 btrfs_warn(root->fs_info,
5400 "can't set the %s feature bit%s while mounted",
5401 names, strchr(names, ',') ? "s" : "");
5402 kfree(names);
5403 } else
5404 btrfs_warn(root->fs_info,
5405 "can't set %s bits 0x%llx while mounted",
5406 type, disallowed);
5407 return -EPERM;
5408 }
5409
5410 disallowed = clear_mask & ~safe_clear;
5411 if (disallowed) {
5412 names = btrfs_printable_features(set, disallowed);
5413 if (names) {
5414 btrfs_warn(root->fs_info,
5415 "can't clear the %s feature bit%s while mounted",
5416 names, strchr(names, ',') ? "s" : "");
5417 kfree(names);
5418 } else
5419 btrfs_warn(root->fs_info,
5420 "can't clear %s bits 0x%llx while mounted",
5421 type, disallowed);
5422 return -EPERM;
5423 }
5424
5425 return 0;
5426 }
5427
5428 #define check_feature(root, change_mask, flags, mask_base) \
5429 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5430 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5431 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5432 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5433
5434 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5435 {
5436 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5437 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5438 struct btrfs_ioctl_feature_flags flags[2];
5439 struct btrfs_trans_handle *trans;
5440 u64 newflags;
5441 int ret;
5442
5443 if (!capable(CAP_SYS_ADMIN))
5444 return -EPERM;
5445
5446 if (copy_from_user(flags, arg, sizeof(flags)))
5447 return -EFAULT;
5448
5449 /* Nothing to do */
5450 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5451 !flags[0].incompat_flags)
5452 return 0;
5453
5454 ret = check_feature(root, flags[0].compat_flags,
5455 flags[1].compat_flags, COMPAT);
5456 if (ret)
5457 return ret;
5458
5459 ret = check_feature(root, flags[0].compat_ro_flags,
5460 flags[1].compat_ro_flags, COMPAT_RO);
5461 if (ret)
5462 return ret;
5463
5464 ret = check_feature(root, flags[0].incompat_flags,
5465 flags[1].incompat_flags, INCOMPAT);
5466 if (ret)
5467 return ret;
5468
5469 trans = btrfs_start_transaction(root, 0);
5470 if (IS_ERR(trans))
5471 return PTR_ERR(trans);
5472
5473 spin_lock(&root->fs_info->super_lock);
5474 newflags = btrfs_super_compat_flags(super_block);
5475 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5476 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5477 btrfs_set_super_compat_flags(super_block, newflags);
5478
5479 newflags = btrfs_super_compat_ro_flags(super_block);
5480 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5481 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5482 btrfs_set_super_compat_ro_flags(super_block, newflags);
5483
5484 newflags = btrfs_super_incompat_flags(super_block);
5485 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5486 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5487 btrfs_set_super_incompat_flags(super_block, newflags);
5488 spin_unlock(&root->fs_info->super_lock);
5489
5490 return btrfs_commit_transaction(trans, root);
5491 }
5492
5493 long btrfs_ioctl(struct file *file, unsigned int
5494 cmd, unsigned long arg)
5495 {
5496 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5497 void __user *argp = (void __user *)arg;
5498
5499 switch (cmd) {
5500 case FS_IOC_GETFLAGS:
5501 return btrfs_ioctl_getflags(file, argp);
5502 case FS_IOC_SETFLAGS:
5503 return btrfs_ioctl_setflags(file, argp);
5504 case FS_IOC_GETVERSION:
5505 return btrfs_ioctl_getversion(file, argp);
5506 case FITRIM:
5507 return btrfs_ioctl_fitrim(file, argp);
5508 case BTRFS_IOC_SNAP_CREATE:
5509 return btrfs_ioctl_snap_create(file, argp, 0);
5510 case BTRFS_IOC_SNAP_CREATE_V2:
5511 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5512 case BTRFS_IOC_SUBVOL_CREATE:
5513 return btrfs_ioctl_snap_create(file, argp, 1);
5514 case BTRFS_IOC_SUBVOL_CREATE_V2:
5515 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5516 case BTRFS_IOC_SNAP_DESTROY:
5517 return btrfs_ioctl_snap_destroy(file, argp);
5518 case BTRFS_IOC_SUBVOL_GETFLAGS:
5519 return btrfs_ioctl_subvol_getflags(file, argp);
5520 case BTRFS_IOC_SUBVOL_SETFLAGS:
5521 return btrfs_ioctl_subvol_setflags(file, argp);
5522 case BTRFS_IOC_DEFAULT_SUBVOL:
5523 return btrfs_ioctl_default_subvol(file, argp);
5524 case BTRFS_IOC_DEFRAG:
5525 return btrfs_ioctl_defrag(file, NULL);
5526 case BTRFS_IOC_DEFRAG_RANGE:
5527 return btrfs_ioctl_defrag(file, argp);
5528 case BTRFS_IOC_RESIZE:
5529 return btrfs_ioctl_resize(file, argp);
5530 case BTRFS_IOC_ADD_DEV:
5531 return btrfs_ioctl_add_dev(root, argp);
5532 case BTRFS_IOC_RM_DEV:
5533 return btrfs_ioctl_rm_dev(file, argp);
5534 case BTRFS_IOC_FS_INFO:
5535 return btrfs_ioctl_fs_info(root, argp);
5536 case BTRFS_IOC_DEV_INFO:
5537 return btrfs_ioctl_dev_info(root, argp);
5538 case BTRFS_IOC_BALANCE:
5539 return btrfs_ioctl_balance(file, NULL);
5540 case BTRFS_IOC_CLONE:
5541 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5542 case BTRFS_IOC_CLONE_RANGE:
5543 return btrfs_ioctl_clone_range(file, argp);
5544 case BTRFS_IOC_TRANS_START:
5545 return btrfs_ioctl_trans_start(file);
5546 case BTRFS_IOC_TRANS_END:
5547 return btrfs_ioctl_trans_end(file);
5548 case BTRFS_IOC_TREE_SEARCH:
5549 return btrfs_ioctl_tree_search(file, argp);
5550 case BTRFS_IOC_TREE_SEARCH_V2:
5551 return btrfs_ioctl_tree_search_v2(file, argp);
5552 case BTRFS_IOC_INO_LOOKUP:
5553 return btrfs_ioctl_ino_lookup(file, argp);
5554 case BTRFS_IOC_INO_PATHS:
5555 return btrfs_ioctl_ino_to_path(root, argp);
5556 case BTRFS_IOC_LOGICAL_INO:
5557 return btrfs_ioctl_logical_to_ino(root, argp);
5558 case BTRFS_IOC_SPACE_INFO:
5559 return btrfs_ioctl_space_info(root, argp);
5560 case BTRFS_IOC_SYNC: {
5561 int ret;
5562
5563 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5564 if (ret)
5565 return ret;
5566 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
5567 /*
5568 * The transaction thread may want to do more work,
5569 * namely it pokes the cleaner ktread that will start
5570 * processing uncleaned subvols.
5571 */
5572 wake_up_process(root->fs_info->transaction_kthread);
5573 return ret;
5574 }
5575 case BTRFS_IOC_START_SYNC:
5576 return btrfs_ioctl_start_sync(root, argp);
5577 case BTRFS_IOC_WAIT_SYNC:
5578 return btrfs_ioctl_wait_sync(root, argp);
5579 case BTRFS_IOC_SCRUB:
5580 return btrfs_ioctl_scrub(file, argp);
5581 case BTRFS_IOC_SCRUB_CANCEL:
5582 return btrfs_ioctl_scrub_cancel(root, argp);
5583 case BTRFS_IOC_SCRUB_PROGRESS:
5584 return btrfs_ioctl_scrub_progress(root, argp);
5585 case BTRFS_IOC_BALANCE_V2:
5586 return btrfs_ioctl_balance(file, argp);
5587 case BTRFS_IOC_BALANCE_CTL:
5588 return btrfs_ioctl_balance_ctl(root, arg);
5589 case BTRFS_IOC_BALANCE_PROGRESS:
5590 return btrfs_ioctl_balance_progress(root, argp);
5591 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5592 return btrfs_ioctl_set_received_subvol(file, argp);
5593 #ifdef CONFIG_64BIT
5594 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5595 return btrfs_ioctl_set_received_subvol_32(file, argp);
5596 #endif
5597 case BTRFS_IOC_SEND:
5598 return btrfs_ioctl_send(file, argp);
5599 case BTRFS_IOC_GET_DEV_STATS:
5600 return btrfs_ioctl_get_dev_stats(root, argp);
5601 case BTRFS_IOC_QUOTA_CTL:
5602 return btrfs_ioctl_quota_ctl(file, argp);
5603 case BTRFS_IOC_QGROUP_ASSIGN:
5604 return btrfs_ioctl_qgroup_assign(file, argp);
5605 case BTRFS_IOC_QGROUP_CREATE:
5606 return btrfs_ioctl_qgroup_create(file, argp);
5607 case BTRFS_IOC_QGROUP_LIMIT:
5608 return btrfs_ioctl_qgroup_limit(file, argp);
5609 case BTRFS_IOC_QUOTA_RESCAN:
5610 return btrfs_ioctl_quota_rescan(file, argp);
5611 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5612 return btrfs_ioctl_quota_rescan_status(file, argp);
5613 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5614 return btrfs_ioctl_quota_rescan_wait(file, argp);
5615 case BTRFS_IOC_DEV_REPLACE:
5616 return btrfs_ioctl_dev_replace(root, argp);
5617 case BTRFS_IOC_GET_FSLABEL:
5618 return btrfs_ioctl_get_fslabel(file, argp);
5619 case BTRFS_IOC_SET_FSLABEL:
5620 return btrfs_ioctl_set_fslabel(file, argp);
5621 case BTRFS_IOC_FILE_EXTENT_SAME:
5622 return btrfs_ioctl_file_extent_same(file, argp);
5623 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5624 return btrfs_ioctl_get_supported_features(file, argp);
5625 case BTRFS_IOC_GET_FEATURES:
5626 return btrfs_ioctl_get_features(file, argp);
5627 case BTRFS_IOC_SET_FEATURES:
5628 return btrfs_ioctl_set_features(file, argp);
5629 }
5630
5631 return -ENOTTY;
5632 }
Linux kernel - Deliverables
This page took 0.165179 seconds and 5 git commands to generate.