?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
powerpc: Convert cmp to cmpd in idle enter sequence
[linux/fpc-iii.git] / fs / btrfs / ioctl.c
blobff742d30ba60655a3d306be6e2ce35f046232afb
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
23 #include <linux/fs.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include <linux/btrfs.h>
46 #include <linux/uaccess.h>
47 #include "ctree.h"
48 #include "disk-io.h"
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
52 #include "volumes.h"
53 #include "locking.h"
54 #include "inode-map.h"
55 #include "backref.h"
56 #include "rcu-string.h"
57 #include "send.h"
58 #include "dev-replace.h"
59 #include "props.h"
60 #include "sysfs.h"
61 #include "qgroup.h"
62
63 #ifdef CONFIG_64BIT
64 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
68 */
69 struct btrfs_ioctl_timespec_32 {
70 __u64 sec;
71 __u32 nsec;
72 } __attribute__ ((__packed__));
73
74 struct btrfs_ioctl_received_subvol_args_32 {
75 char uuid[BTRFS_UUID_SIZE]; /* in */
76 __u64 stransid; /* in */
77 __u64 rtransid; /* out */
78 struct btrfs_ioctl_timespec_32 stime; /* in */
79 struct btrfs_ioctl_timespec_32 rtime; /* out */
80 __u64 flags; /* in */
81 __u64 reserved[16]; /* in */
82 } __attribute__ ((__packed__));
83
84 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
86 #endif
87
88
89 static int btrfs_clone(struct inode *src, struct inode *inode,
90 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
91
92 /* Mask out flags that are inappropriate for the given type of inode. */
93 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
94 {
95 if (S_ISDIR(mode))
96 return flags;
97 else if (S_ISREG(mode))
98 return flags & ~FS_DIRSYNC_FL;
99 else
100 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
101 }
102
103 /*
104 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
105 */
106 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
107 {
108 unsigned int iflags = 0;
109
110 if (flags & BTRFS_INODE_SYNC)
111 iflags |= FS_SYNC_FL;
112 if (flags & BTRFS_INODE_IMMUTABLE)
113 iflags |= FS_IMMUTABLE_FL;
114 if (flags & BTRFS_INODE_APPEND)
115 iflags |= FS_APPEND_FL;
116 if (flags & BTRFS_INODE_NODUMP)
117 iflags |= FS_NODUMP_FL;
118 if (flags & BTRFS_INODE_NOATIME)
119 iflags |= FS_NOATIME_FL;
120 if (flags & BTRFS_INODE_DIRSYNC)
121 iflags |= FS_DIRSYNC_FL;
122 if (flags & BTRFS_INODE_NODATACOW)
123 iflags |= FS_NOCOW_FL;
124
125 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
126 iflags |= FS_COMPR_FL;
127 else if (flags & BTRFS_INODE_NOCOMPRESS)
128 iflags |= FS_NOCOMP_FL;
129
130 return iflags;
131 }
132
133 /*
134 * Update inode->i_flags based on the btrfs internal flags.
135 */
136 void btrfs_update_iflags(struct inode *inode)
137 {
138 struct btrfs_inode *ip = BTRFS_I(inode);
139 unsigned int new_fl = 0;
140
141 if (ip->flags & BTRFS_INODE_SYNC)
142 new_fl |= S_SYNC;
143 if (ip->flags & BTRFS_INODE_IMMUTABLE)
144 new_fl |= S_IMMUTABLE;
145 if (ip->flags & BTRFS_INODE_APPEND)
146 new_fl |= S_APPEND;
147 if (ip->flags & BTRFS_INODE_NOATIME)
148 new_fl |= S_NOATIME;
149 if (ip->flags & BTRFS_INODE_DIRSYNC)
150 new_fl |= S_DIRSYNC;
151
152 set_mask_bits(&inode->i_flags,
153 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
154 new_fl);
155 }
156
157 /*
158 * Inherit flags from the parent inode.
159 *
160 * Currently only the compression flags and the cow flags are inherited.
161 */
162 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
163 {
164 unsigned int flags;
165
166 if (!dir)
167 return;
168
169 flags = BTRFS_I(dir)->flags;
170
171 if (flags & BTRFS_INODE_NOCOMPRESS) {
172 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
173 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
174 } else if (flags & BTRFS_INODE_COMPRESS) {
175 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
176 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
177 }
178
179 if (flags & BTRFS_INODE_NODATACOW) {
180 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
181 if (S_ISREG(inode->i_mode))
182 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
183 }
184
185 btrfs_update_iflags(inode);
186 }
187
188 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
189 {
190 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
191 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
192
193 if (copy_to_user(arg, &flags, sizeof(flags)))
194 return -EFAULT;
195 return 0;
196 }
197
198 static int check_flags(unsigned int flags)
199 {
200 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
201 FS_NOATIME_FL | FS_NODUMP_FL | \
202 FS_SYNC_FL | FS_DIRSYNC_FL | \
203 FS_NOCOMP_FL | FS_COMPR_FL |
204 FS_NOCOW_FL))
205 return -EOPNOTSUPP;
206
207 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
208 return -EINVAL;
209
210 return 0;
211 }
212
213 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
214 {
215 struct inode *inode = file_inode(file);
216 struct btrfs_inode *ip = BTRFS_I(inode);
217 struct btrfs_root *root = ip->root;
218 struct btrfs_trans_handle *trans;
219 unsigned int flags, oldflags;
220 int ret;
221 u64 ip_oldflags;
222 unsigned int i_oldflags;
223 umode_t mode;
224
225 if (!inode_owner_or_capable(inode))
226 return -EPERM;
227
228 if (btrfs_root_readonly(root))
229 return -EROFS;
230
231 if (copy_from_user(&flags, arg, sizeof(flags)))
232 return -EFAULT;
233
234 ret = check_flags(flags);
235 if (ret)
236 return ret;
237
238 ret = mnt_want_write_file(file);
239 if (ret)
240 return ret;
241
242 mutex_lock(&inode->i_mutex);
243
244 ip_oldflags = ip->flags;
245 i_oldflags = inode->i_flags;
246 mode = inode->i_mode;
247
248 flags = btrfs_mask_flags(inode->i_mode, flags);
249 oldflags = btrfs_flags_to_ioctl(ip->flags);
250 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
251 if (!capable(CAP_LINUX_IMMUTABLE)) {
252 ret = -EPERM;
253 goto out_unlock;
254 }
255 }
256
257 if (flags & FS_SYNC_FL)
258 ip->flags |= BTRFS_INODE_SYNC;
259 else
260 ip->flags &= ~BTRFS_INODE_SYNC;
261 if (flags & FS_IMMUTABLE_FL)
262 ip->flags |= BTRFS_INODE_IMMUTABLE;
263 else
264 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
265 if (flags & FS_APPEND_FL)
266 ip->flags |= BTRFS_INODE_APPEND;
267 else
268 ip->flags &= ~BTRFS_INODE_APPEND;
269 if (flags & FS_NODUMP_FL)
270 ip->flags |= BTRFS_INODE_NODUMP;
271 else
272 ip->flags &= ~BTRFS_INODE_NODUMP;
273 if (flags & FS_NOATIME_FL)
274 ip->flags |= BTRFS_INODE_NOATIME;
275 else
276 ip->flags &= ~BTRFS_INODE_NOATIME;
277 if (flags & FS_DIRSYNC_FL)
278 ip->flags |= BTRFS_INODE_DIRSYNC;
279 else
280 ip->flags &= ~BTRFS_INODE_DIRSYNC;
281 if (flags & FS_NOCOW_FL) {
282 if (S_ISREG(mode)) {
283 /*
284 * It's safe to turn csums off here, no extents exist.
285 * Otherwise we want the flag to reflect the real COW
286 * status of the file and will not set it.
287 */
288 if (inode->i_size == 0)
289 ip->flags |= BTRFS_INODE_NODATACOW
290 | BTRFS_INODE_NODATASUM;
291 } else {
292 ip->flags |= BTRFS_INODE_NODATACOW;
293 }
294 } else {
295 /*
296 * Revert back under same assuptions as above
297 */
298 if (S_ISREG(mode)) {
299 if (inode->i_size == 0)
300 ip->flags &= ~(BTRFS_INODE_NODATACOW
301 | BTRFS_INODE_NODATASUM);
302 } else {
303 ip->flags &= ~BTRFS_INODE_NODATACOW;
304 }
305 }
306
307 /*
308 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
309 * flag may be changed automatically if compression code won't make
310 * things smaller.
311 */
312 if (flags & FS_NOCOMP_FL) {
313 ip->flags &= ~BTRFS_INODE_COMPRESS;
314 ip->flags |= BTRFS_INODE_NOCOMPRESS;
315
316 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
317 if (ret && ret != -ENODATA)
318 goto out_drop;
319 } else if (flags & FS_COMPR_FL) {
320 const char *comp;
321
322 ip->flags |= BTRFS_INODE_COMPRESS;
323 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
324
325 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
326 comp = "lzo";
327 else
328 comp = "zlib";
329 ret = btrfs_set_prop(inode, "btrfs.compression",
330 comp, strlen(comp), 0);
331 if (ret)
332 goto out_drop;
333
334 } else {
335 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
336 if (ret && ret != -ENODATA)
337 goto out_drop;
338 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
339 }
340
341 trans = btrfs_start_transaction(root, 1);
342 if (IS_ERR(trans)) {
343 ret = PTR_ERR(trans);
344 goto out_drop;
345 }
346
347 btrfs_update_iflags(inode);
348 inode_inc_iversion(inode);
349 inode->i_ctime = CURRENT_TIME;
350 ret = btrfs_update_inode(trans, root, inode);
351
352 btrfs_end_transaction(trans, root);
353 out_drop:
354 if (ret) {
355 ip->flags = ip_oldflags;
356 inode->i_flags = i_oldflags;
357 }
358
359 out_unlock:
360 mutex_unlock(&inode->i_mutex);
361 mnt_drop_write_file(file);
362 return ret;
363 }
364
365 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
366 {
367 struct inode *inode = file_inode(file);
368
369 return put_user(inode->i_generation, arg);
370 }
371
372 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
373 {
374 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
375 struct btrfs_device *device;
376 struct request_queue *q;
377 struct fstrim_range range;
378 u64 minlen = ULLONG_MAX;
379 u64 num_devices = 0;
380 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
381 int ret;
382
383 if (!capable(CAP_SYS_ADMIN))
384 return -EPERM;
385
386 rcu_read_lock();
387 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
388 dev_list) {
389 if (!device->bdev)
390 continue;
391 q = bdev_get_queue(device->bdev);
392 if (blk_queue_discard(q)) {
393 num_devices++;
394 minlen = min((u64)q->limits.discard_granularity,
395 minlen);
396 }
397 }
398 rcu_read_unlock();
399
400 if (!num_devices)
401 return -EOPNOTSUPP;
402 if (copy_from_user(&range, arg, sizeof(range)))
403 return -EFAULT;
404 if (range.start > total_bytes ||
405 range.len < fs_info->sb->s_blocksize)
406 return -EINVAL;
407
408 range.len = min(range.len, total_bytes - range.start);
409 range.minlen = max(range.minlen, minlen);
410 ret = btrfs_trim_fs(fs_info->tree_root, &range);
411 if (ret < 0)
412 return ret;
413
414 if (copy_to_user(arg, &range, sizeof(range)))
415 return -EFAULT;
416
417 return 0;
418 }
419
420 int btrfs_is_empty_uuid(u8 *uuid)
421 {
422 int i;
423
424 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
425 if (uuid[i])
426 return 0;
427 }
428 return 1;
429 }
430
431 static noinline int create_subvol(struct inode *dir,
432 struct dentry *dentry,
433 char *name, int namelen,
434 u64 *async_transid,
435 struct btrfs_qgroup_inherit *inherit)
436 {
437 struct btrfs_trans_handle *trans;
438 struct btrfs_key key;
439 struct btrfs_root_item root_item;
440 struct btrfs_inode_item *inode_item;
441 struct extent_buffer *leaf;
442 struct btrfs_root *root = BTRFS_I(dir)->root;
443 struct btrfs_root *new_root;
444 struct btrfs_block_rsv block_rsv;
445 struct timespec cur_time = CURRENT_TIME;
446 struct inode *inode;
447 int ret;
448 int err;
449 u64 objectid;
450 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
451 u64 index = 0;
452 u64 qgroup_reserved;
453 uuid_le new_uuid;
454
455 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
456 if (ret)
457 return ret;
458
459 /*
460 * Don't create subvolume whose level is not zero. Or qgroup will be
461 * screwed up since it assume subvolme qgroup's level to be 0.
462 */
463 if (btrfs_qgroup_level(objectid))
464 return -ENOSPC;
465
466 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
467 /*
468 * The same as the snapshot creation, please see the comment
469 * of create_snapshot().
470 */
471 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
472 8, &qgroup_reserved, false);
473 if (ret)
474 return ret;
475
476 trans = btrfs_start_transaction(root, 0);
477 if (IS_ERR(trans)) {
478 ret = PTR_ERR(trans);
479 btrfs_subvolume_release_metadata(root, &block_rsv,
480 qgroup_reserved);
481 return ret;
482 }
483 trans->block_rsv = &block_rsv;
484 trans->bytes_reserved = block_rsv.size;
485
486 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
487 if (ret)
488 goto fail;
489
490 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
491 if (IS_ERR(leaf)) {
492 ret = PTR_ERR(leaf);
493 goto fail;
494 }
495
496 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
497 btrfs_set_header_bytenr(leaf, leaf->start);
498 btrfs_set_header_generation(leaf, trans->transid);
499 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
500 btrfs_set_header_owner(leaf, objectid);
501
502 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
503 BTRFS_FSID_SIZE);
504 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
505 btrfs_header_chunk_tree_uuid(leaf),
506 BTRFS_UUID_SIZE);
507 btrfs_mark_buffer_dirty(leaf);
508
509 memset(&root_item, 0, sizeof(root_item));
510
511 inode_item = &root_item.inode;
512 btrfs_set_stack_inode_generation(inode_item, 1);
513 btrfs_set_stack_inode_size(inode_item, 3);
514 btrfs_set_stack_inode_nlink(inode_item, 1);
515 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
516 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
517
518 btrfs_set_root_flags(&root_item, 0);
519 btrfs_set_root_limit(&root_item, 0);
520 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
521
522 btrfs_set_root_bytenr(&root_item, leaf->start);
523 btrfs_set_root_generation(&root_item, trans->transid);
524 btrfs_set_root_level(&root_item, 0);
525 btrfs_set_root_refs(&root_item, 1);
526 btrfs_set_root_used(&root_item, leaf->len);
527 btrfs_set_root_last_snapshot(&root_item, 0);
528
529 btrfs_set_root_generation_v2(&root_item,
530 btrfs_root_generation(&root_item));
531 uuid_le_gen(&new_uuid);
532 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
533 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
534 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
535 root_item.ctime = root_item.otime;
536 btrfs_set_root_ctransid(&root_item, trans->transid);
537 btrfs_set_root_otransid(&root_item, trans->transid);
538
539 btrfs_tree_unlock(leaf);
540 free_extent_buffer(leaf);
541 leaf = NULL;
542
543 btrfs_set_root_dirid(&root_item, new_dirid);
544
545 key.objectid = objectid;
546 key.offset = 0;
547 key.type = BTRFS_ROOT_ITEM_KEY;
548 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
549 &root_item);
550 if (ret)
551 goto fail;
552
553 key.offset = (u64)-1;
554 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
555 if (IS_ERR(new_root)) {
556 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
557 ret = PTR_ERR(new_root);
558 goto fail;
559 }
560
561 btrfs_record_root_in_trans(trans, new_root);
562
563 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
564 if (ret) {
565 /* We potentially lose an unused inode item here */
566 btrfs_abort_transaction(trans, root, ret);
567 goto fail;
568 }
569
570 mutex_lock(&new_root->objectid_mutex);
571 new_root->highest_objectid = new_dirid;
572 mutex_unlock(&new_root->objectid_mutex);
573
574 /*
575 * insert the directory item
576 */
577 ret = btrfs_set_inode_index(dir, &index);
578 if (ret) {
579 btrfs_abort_transaction(trans, root, ret);
580 goto fail;
581 }
582
583 ret = btrfs_insert_dir_item(trans, root,
584 name, namelen, dir, &key,
585 BTRFS_FT_DIR, index);
586 if (ret) {
587 btrfs_abort_transaction(trans, root, ret);
588 goto fail;
589 }
590
591 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
592 ret = btrfs_update_inode(trans, root, dir);
593 BUG_ON(ret);
594
595 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
596 objectid, root->root_key.objectid,
597 btrfs_ino(dir), index, name, namelen);
598 BUG_ON(ret);
599
600 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
601 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
602 objectid);
603 if (ret)
604 btrfs_abort_transaction(trans, root, ret);
605
606 fail:
607 trans->block_rsv = NULL;
608 trans->bytes_reserved = 0;
609 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
610
611 if (async_transid) {
612 *async_transid = trans->transid;
613 err = btrfs_commit_transaction_async(trans, root, 1);
614 if (err)
615 err = btrfs_commit_transaction(trans, root);
616 } else {
617 err = btrfs_commit_transaction(trans, root);
618 }
619 if (err && !ret)
620 ret = err;
621
622 if (!ret) {
623 inode = btrfs_lookup_dentry(dir, dentry);
624 if (IS_ERR(inode))
625 return PTR_ERR(inode);
626 d_instantiate(dentry, inode);
627 }
628 return ret;
629 }
630
631 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
632 {
633 s64 writers;
634 DEFINE_WAIT(wait);
635
636 do {
637 prepare_to_wait(&root->subv_writers->wait, &wait,
638 TASK_UNINTERRUPTIBLE);
639
640 writers = percpu_counter_sum(&root->subv_writers->counter);
641 if (writers)
642 schedule();
643
644 finish_wait(&root->subv_writers->wait, &wait);
645 } while (writers);
646 }
647
648 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
649 struct dentry *dentry, char *name, int namelen,
650 u64 *async_transid, bool readonly,
651 struct btrfs_qgroup_inherit *inherit)
652 {
653 struct inode *inode;
654 struct btrfs_pending_snapshot *pending_snapshot;
655 struct btrfs_trans_handle *trans;
656 int ret;
657
658 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
659 return -EINVAL;
660
661 atomic_inc(&root->will_be_snapshoted);
662 smp_mb__after_atomic();
663 btrfs_wait_for_no_snapshoting_writes(root);
664
665 ret = btrfs_start_delalloc_inodes(root, 0);
666 if (ret)
667 goto out;
668
669 btrfs_wait_ordered_extents(root, -1);
670
671 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
672 if (!pending_snapshot) {
673 ret = -ENOMEM;
674 goto out;
675 }
676
677 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
678 BTRFS_BLOCK_RSV_TEMP);
679 /*
680 * 1 - parent dir inode
681 * 2 - dir entries
682 * 1 - root item
683 * 2 - root ref/backref
684 * 1 - root of snapshot
685 * 1 - UUID item
686 */
687 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
688 &pending_snapshot->block_rsv, 8,
689 &pending_snapshot->qgroup_reserved,
690 false);
691 if (ret)
692 goto free;
693
694 pending_snapshot->dentry = dentry;
695 pending_snapshot->root = root;
696 pending_snapshot->readonly = readonly;
697 pending_snapshot->dir = dir;
698 pending_snapshot->inherit = inherit;
699
700 trans = btrfs_start_transaction(root, 0);
701 if (IS_ERR(trans)) {
702 ret = PTR_ERR(trans);
703 goto fail;
704 }
705
706 spin_lock(&root->fs_info->trans_lock);
707 list_add(&pending_snapshot->list,
708 &trans->transaction->pending_snapshots);
709 spin_unlock(&root->fs_info->trans_lock);
710 if (async_transid) {
711 *async_transid = trans->transid;
712 ret = btrfs_commit_transaction_async(trans,
713 root->fs_info->extent_root, 1);
714 if (ret)
715 ret = btrfs_commit_transaction(trans, root);
716 } else {
717 ret = btrfs_commit_transaction(trans,
718 root->fs_info->extent_root);
719 }
720 if (ret)
721 goto fail;
722
723 ret = pending_snapshot->error;
724 if (ret)
725 goto fail;
726
727 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
728 if (ret)
729 goto fail;
730
731 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
732 if (IS_ERR(inode)) {
733 ret = PTR_ERR(inode);
734 goto fail;
735 }
736
737 d_instantiate(dentry, inode);
738 ret = 0;
739 fail:
740 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
741 &pending_snapshot->block_rsv,
742 pending_snapshot->qgroup_reserved);
743 free:
744 kfree(pending_snapshot);
745 out:
746 if (atomic_dec_and_test(&root->will_be_snapshoted))
747 wake_up_atomic_t(&root->will_be_snapshoted);
748 return ret;
749 }
750
751 /* copy of may_delete in fs/namei.c()
752 * Check whether we can remove a link victim from directory dir, check
753 * whether the type of victim is right.
754 * 1. We can't do it if dir is read-only (done in permission())
755 * 2. We should have write and exec permissions on dir
756 * 3. We can't remove anything from append-only dir
757 * 4. We can't do anything with immutable dir (done in permission())
758 * 5. If the sticky bit on dir is set we should either
759 * a. be owner of dir, or
760 * b. be owner of victim, or
761 * c. have CAP_FOWNER capability
762 * 6. If the victim is append-only or immutable we can't do antyhing with
763 * links pointing to it.
764 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
765 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
766 * 9. We can't remove a root or mountpoint.
767 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
768 * nfs_async_unlink().
769 */
770
771 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
772 {
773 int error;
774
775 if (d_really_is_negative(victim))
776 return -ENOENT;
777
778 BUG_ON(d_inode(victim->d_parent) != dir);
779 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
780
781 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
782 if (error)
783 return error;
784 if (IS_APPEND(dir))
785 return -EPERM;
786 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
787 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
788 return -EPERM;
789 if (isdir) {
790 if (!d_is_dir(victim))
791 return -ENOTDIR;
792 if (IS_ROOT(victim))
793 return -EBUSY;
794 } else if (d_is_dir(victim))
795 return -EISDIR;
796 if (IS_DEADDIR(dir))
797 return -ENOENT;
798 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
799 return -EBUSY;
800 return 0;
801 }
802
803 /* copy of may_create in fs/namei.c() */
804 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
805 {
806 if (d_really_is_positive(child))
807 return -EEXIST;
808 if (IS_DEADDIR(dir))
809 return -ENOENT;
810 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
811 }
812
813 /*
814 * Create a new subvolume below @parent. This is largely modeled after
815 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
816 * inside this filesystem so it's quite a bit simpler.
817 */
818 static noinline int btrfs_mksubvol(struct path *parent,
819 char *name, int namelen,
820 struct btrfs_root *snap_src,
821 u64 *async_transid, bool readonly,
822 struct btrfs_qgroup_inherit *inherit)
823 {
824 struct inode *dir = d_inode(parent->dentry);
825 struct dentry *dentry;
826 int error;
827
828 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
829 if (error == -EINTR)
830 return error;
831
832 dentry = lookup_one_len(name, parent->dentry, namelen);
833 error = PTR_ERR(dentry);
834 if (IS_ERR(dentry))
835 goto out_unlock;
836
837 error = -EEXIST;
838 if (d_really_is_positive(dentry))
839 goto out_dput;
840
841 error = btrfs_may_create(dir, dentry);
842 if (error)
843 goto out_dput;
844
845 /*
846 * even if this name doesn't exist, we may get hash collisions.
847 * check for them now when we can safely fail
848 */
849 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
850 dir->i_ino, name,
851 namelen);
852 if (error)
853 goto out_dput;
854
855 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
856
857 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
858 goto out_up_read;
859
860 if (snap_src) {
861 error = create_snapshot(snap_src, dir, dentry, name, namelen,
862 async_transid, readonly, inherit);
863 } else {
864 error = create_subvol(dir, dentry, name, namelen,
865 async_transid, inherit);
866 }
867 if (!error)
868 fsnotify_mkdir(dir, dentry);
869 out_up_read:
870 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
871 out_dput:
872 dput(dentry);
873 out_unlock:
874 mutex_unlock(&dir->i_mutex);
875 return error;
876 }
877
878 /*
879 * When we're defragging a range, we don't want to kick it off again
880 * if it is really just waiting for delalloc to send it down.
881 * If we find a nice big extent or delalloc range for the bytes in the
882 * file you want to defrag, we return 0 to let you know to skip this
883 * part of the file
884 */
885 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
886 {
887 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
888 struct extent_map *em = NULL;
889 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
890 u64 end;
891
892 read_lock(&em_tree->lock);
893 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
894 read_unlock(&em_tree->lock);
895
896 if (em) {
897 end = extent_map_end(em);
898 free_extent_map(em);
899 if (end - offset > thresh)
900 return 0;
901 }
902 /* if we already have a nice delalloc here, just stop */
903 thresh /= 2;
904 end = count_range_bits(io_tree, &offset, offset + thresh,
905 thresh, EXTENT_DELALLOC, 1);
906 if (end >= thresh)
907 return 0;
908 return 1;
909 }
910
911 /*
912 * helper function to walk through a file and find extents
913 * newer than a specific transid, and smaller than thresh.
914 *
915 * This is used by the defragging code to find new and small
916 * extents
917 */
918 static int find_new_extents(struct btrfs_root *root,
919 struct inode *inode, u64 newer_than,
920 u64 *off, u32 thresh)
921 {
922 struct btrfs_path *path;
923 struct btrfs_key min_key;
924 struct extent_buffer *leaf;
925 struct btrfs_file_extent_item *extent;
926 int type;
927 int ret;
928 u64 ino = btrfs_ino(inode);
929
930 path = btrfs_alloc_path();
931 if (!path)
932 return -ENOMEM;
933
934 min_key.objectid = ino;
935 min_key.type = BTRFS_EXTENT_DATA_KEY;
936 min_key.offset = *off;
937
938 while (1) {
939 ret = btrfs_search_forward(root, &min_key, path, newer_than);
940 if (ret != 0)
941 goto none;
942 process_slot:
943 if (min_key.objectid != ino)
944 goto none;
945 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
946 goto none;
947
948 leaf = path->nodes[0];
949 extent = btrfs_item_ptr(leaf, path->slots[0],
950 struct btrfs_file_extent_item);
951
952 type = btrfs_file_extent_type(leaf, extent);
953 if (type == BTRFS_FILE_EXTENT_REG &&
954 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
955 check_defrag_in_cache(inode, min_key.offset, thresh)) {
956 *off = min_key.offset;
957 btrfs_free_path(path);
958 return 0;
959 }
960
961 path->slots[0]++;
962 if (path->slots[0] < btrfs_header_nritems(leaf)) {
963 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
964 goto process_slot;
965 }
966
967 if (min_key.offset == (u64)-1)
968 goto none;
969
970 min_key.offset++;
971 btrfs_release_path(path);
972 }
973 none:
974 btrfs_free_path(path);
975 return -ENOENT;
976 }
977
978 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
979 {
980 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
981 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
982 struct extent_map *em;
983 u64 len = PAGE_CACHE_SIZE;
984
985 /*
986 * hopefully we have this extent in the tree already, try without
987 * the full extent lock
988 */
989 read_lock(&em_tree->lock);
990 em = lookup_extent_mapping(em_tree, start, len);
991 read_unlock(&em_tree->lock);
992
993 if (!em) {
994 struct extent_state *cached = NULL;
995 u64 end = start + len - 1;
996
997 /* get the big lock and read metadata off disk */
998 lock_extent_bits(io_tree, start, end, 0, &cached);
999 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1000 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1001
1002 if (IS_ERR(em))
1003 return NULL;
1004 }
1005
1006 return em;
1007 }
1008
1009 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1010 {
1011 struct extent_map *next;
1012 bool ret = true;
1013
1014 /* this is the last extent */
1015 if (em->start + em->len >= i_size_read(inode))
1016 return false;
1017
1018 next = defrag_lookup_extent(inode, em->start + em->len);
1019 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1020 ret = false;
1021 else if ((em->block_start + em->block_len == next->block_start) &&
1022 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
1023 ret = false;
1024
1025 free_extent_map(next);
1026 return ret;
1027 }
1028
1029 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1030 u64 *last_len, u64 *skip, u64 *defrag_end,
1031 int compress)
1032 {
1033 struct extent_map *em;
1034 int ret = 1;
1035 bool next_mergeable = true;
1036
1037 /*
1038 * make sure that once we start defragging an extent, we keep on
1039 * defragging it
1040 */
1041 if (start < *defrag_end)
1042 return 1;
1043
1044 *skip = 0;
1045
1046 em = defrag_lookup_extent(inode, start);
1047 if (!em)
1048 return 0;
1049
1050 /* this will cover holes, and inline extents */
1051 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1052 ret = 0;
1053 goto out;
1054 }
1055
1056 next_mergeable = defrag_check_next_extent(inode, em);
1057 /*
1058 * we hit a real extent, if it is big or the next extent is not a
1059 * real extent, don't bother defragging it
1060 */
1061 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1062 (em->len >= thresh || !next_mergeable))
1063 ret = 0;
1064 out:
1065 /*
1066 * last_len ends up being a counter of how many bytes we've defragged.
1067 * every time we choose not to defrag an extent, we reset *last_len
1068 * so that the next tiny extent will force a defrag.
1069 *
1070 * The end result of this is that tiny extents before a single big
1071 * extent will force at least part of that big extent to be defragged.
1072 */
1073 if (ret) {
1074 *defrag_end = extent_map_end(em);
1075 } else {
1076 *last_len = 0;
1077 *skip = extent_map_end(em);
1078 *defrag_end = 0;
1079 }
1080
1081 free_extent_map(em);
1082 return ret;
1083 }
1084
1085 /*
1086 * it doesn't do much good to defrag one or two pages
1087 * at a time. This pulls in a nice chunk of pages
1088 * to COW and defrag.
1089 *
1090 * It also makes sure the delalloc code has enough
1091 * dirty data to avoid making new small extents as part
1092 * of the defrag
1093 *
1094 * It's a good idea to start RA on this range
1095 * before calling this.
1096 */
1097 static int cluster_pages_for_defrag(struct inode *inode,
1098 struct page **pages,
1099 unsigned long start_index,
1100 unsigned long num_pages)
1101 {
1102 unsigned long file_end;
1103 u64 isize = i_size_read(inode);
1104 u64 page_start;
1105 u64 page_end;
1106 u64 page_cnt;
1107 int ret;
1108 int i;
1109 int i_done;
1110 struct btrfs_ordered_extent *ordered;
1111 struct extent_state *cached_state = NULL;
1112 struct extent_io_tree *tree;
1113 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1114
1115 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1116 if (!isize || start_index > file_end)
1117 return 0;
1118
1119 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1120
1121 ret = btrfs_delalloc_reserve_space(inode,
1122 page_cnt << PAGE_CACHE_SHIFT);
1123 if (ret)
1124 return ret;
1125 i_done = 0;
1126 tree = &BTRFS_I(inode)->io_tree;
1127
1128 /* step one, lock all the pages */
1129 for (i = 0; i < page_cnt; i++) {
1130 struct page *page;
1131 again:
1132 page = find_or_create_page(inode->i_mapping,
1133 start_index + i, mask);
1134 if (!page)
1135 break;
1136
1137 page_start = page_offset(page);
1138 page_end = page_start + PAGE_CACHE_SIZE - 1;
1139 while (1) {
1140 lock_extent_bits(tree, page_start, page_end,
1141 0, &cached_state);
1142 ordered = btrfs_lookup_ordered_extent(inode,
1143 page_start);
1144 unlock_extent_cached(tree, page_start, page_end,
1145 &cached_state, GFP_NOFS);
1146 if (!ordered)
1147 break;
1148
1149 unlock_page(page);
1150 btrfs_start_ordered_extent(inode, ordered, 1);
1151 btrfs_put_ordered_extent(ordered);
1152 lock_page(page);
1153 /*
1154 * we unlocked the page above, so we need check if
1155 * it was released or not.
1156 */
1157 if (page->mapping != inode->i_mapping) {
1158 unlock_page(page);
1159 page_cache_release(page);
1160 goto again;
1161 }
1162 }
1163
1164 if (!PageUptodate(page)) {
1165 btrfs_readpage(NULL, page);
1166 lock_page(page);
1167 if (!PageUptodate(page)) {
1168 unlock_page(page);
1169 page_cache_release(page);
1170 ret = -EIO;
1171 break;
1172 }
1173 }
1174
1175 if (page->mapping != inode->i_mapping) {
1176 unlock_page(page);
1177 page_cache_release(page);
1178 goto again;
1179 }
1180
1181 pages[i] = page;
1182 i_done++;
1183 }
1184 if (!i_done || ret)
1185 goto out;
1186
1187 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1188 goto out;
1189
1190 /*
1191 * so now we have a nice long stream of locked
1192 * and up to date pages, lets wait on them
1193 */
1194 for (i = 0; i < i_done; i++)
1195 wait_on_page_writeback(pages[i]);
1196
1197 page_start = page_offset(pages[0]);
1198 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1199
1200 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1201 page_start, page_end - 1, 0, &cached_state);
1202 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1203 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1204 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1205 &cached_state, GFP_NOFS);
1206
1207 if (i_done != page_cnt) {
1208 spin_lock(&BTRFS_I(inode)->lock);
1209 BTRFS_I(inode)->outstanding_extents++;
1210 spin_unlock(&BTRFS_I(inode)->lock);
1211 btrfs_delalloc_release_space(inode,
1212 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1213 }
1214
1215
1216 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1217 &cached_state, GFP_NOFS);
1218
1219 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1220 page_start, page_end - 1, &cached_state,
1221 GFP_NOFS);
1222
1223 for (i = 0; i < i_done; i++) {
1224 clear_page_dirty_for_io(pages[i]);
1225 ClearPageChecked(pages[i]);
1226 set_page_extent_mapped(pages[i]);
1227 set_page_dirty(pages[i]);
1228 unlock_page(pages[i]);
1229 page_cache_release(pages[i]);
1230 }
1231 return i_done;
1232 out:
1233 for (i = 0; i < i_done; i++) {
1234 unlock_page(pages[i]);
1235 page_cache_release(pages[i]);
1236 }
1237 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1238 return ret;
1239
1240 }
1241
1242 int btrfs_defrag_file(struct inode *inode, struct file *file,
1243 struct btrfs_ioctl_defrag_range_args *range,
1244 u64 newer_than, unsigned long max_to_defrag)
1245 {
1246 struct btrfs_root *root = BTRFS_I(inode)->root;
1247 struct file_ra_state *ra = NULL;
1248 unsigned long last_index;
1249 u64 isize = i_size_read(inode);
1250 u64 last_len = 0;
1251 u64 skip = 0;
1252 u64 defrag_end = 0;
1253 u64 newer_off = range->start;
1254 unsigned long i;
1255 unsigned long ra_index = 0;
1256 int ret;
1257 int defrag_count = 0;
1258 int compress_type = BTRFS_COMPRESS_ZLIB;
1259 u32 extent_thresh = range->extent_thresh;
1260 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1261 unsigned long cluster = max_cluster;
1262 u64 new_align = ~((u64)128 * 1024 - 1);
1263 struct page **pages = NULL;
1264
1265 if (isize == 0)
1266 return 0;
1267
1268 if (range->start >= isize)
1269 return -EINVAL;
1270
1271 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1272 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1273 return -EINVAL;
1274 if (range->compress_type)
1275 compress_type = range->compress_type;
1276 }
1277
1278 if (extent_thresh == 0)
1279 extent_thresh = 256 * 1024;
1280
1281 /*
1282 * if we were not given a file, allocate a readahead
1283 * context
1284 */
1285 if (!file) {
1286 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1287 if (!ra)
1288 return -ENOMEM;
1289 file_ra_state_init(ra, inode->i_mapping);
1290 } else {
1291 ra = &file->f_ra;
1292 }
1293
1294 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1295 GFP_NOFS);
1296 if (!pages) {
1297 ret = -ENOMEM;
1298 goto out_ra;
1299 }
1300
1301 /* find the last page to defrag */
1302 if (range->start + range->len > range->start) {
1303 last_index = min_t(u64, isize - 1,
1304 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1305 } else {
1306 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1307 }
1308
1309 if (newer_than) {
1310 ret = find_new_extents(root, inode, newer_than,
1311 &newer_off, 64 * 1024);
1312 if (!ret) {
1313 range->start = newer_off;
1314 /*
1315 * we always align our defrag to help keep
1316 * the extents in the file evenly spaced
1317 */
1318 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1319 } else
1320 goto out_ra;
1321 } else {
1322 i = range->start >> PAGE_CACHE_SHIFT;
1323 }
1324 if (!max_to_defrag)
1325 max_to_defrag = last_index + 1;
1326
1327 /*
1328 * make writeback starts from i, so the defrag range can be
1329 * written sequentially.
1330 */
1331 if (i < inode->i_mapping->writeback_index)
1332 inode->i_mapping->writeback_index = i;
1333
1334 while (i <= last_index && defrag_count < max_to_defrag &&
1335 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
1336 /*
1337 * make sure we stop running if someone unmounts
1338 * the FS
1339 */
1340 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1341 break;
1342
1343 if (btrfs_defrag_cancelled(root->fs_info)) {
1344 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1345 ret = -EAGAIN;
1346 break;
1347 }
1348
1349 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1350 extent_thresh, &last_len, &skip,
1351 &defrag_end, range->flags &
1352 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1353 unsigned long next;
1354 /*
1355 * the should_defrag function tells us how much to skip
1356 * bump our counter by the suggested amount
1357 */
1358 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
1359 i = max(i + 1, next);
1360 continue;
1361 }
1362
1363 if (!newer_than) {
1364 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1365 PAGE_CACHE_SHIFT) - i;
1366 cluster = min(cluster, max_cluster);
1367 } else {
1368 cluster = max_cluster;
1369 }
1370
1371 if (i + cluster > ra_index) {
1372 ra_index = max(i, ra_index);
1373 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1374 cluster);
1375 ra_index += max_cluster;
1376 }
1377
1378 mutex_lock(&inode->i_mutex);
1379 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1380 BTRFS_I(inode)->force_compress = compress_type;
1381 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1382 if (ret < 0) {
1383 mutex_unlock(&inode->i_mutex);
1384 goto out_ra;
1385 }
1386
1387 defrag_count += ret;
1388 balance_dirty_pages_ratelimited(inode->i_mapping);
1389 mutex_unlock(&inode->i_mutex);
1390
1391 if (newer_than) {
1392 if (newer_off == (u64)-1)
1393 break;
1394
1395 if (ret > 0)
1396 i += ret;
1397
1398 newer_off = max(newer_off + 1,
1399 (u64)i << PAGE_CACHE_SHIFT);
1400
1401 ret = find_new_extents(root, inode,
1402 newer_than, &newer_off,
1403 64 * 1024);
1404 if (!ret) {
1405 range->start = newer_off;
1406 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1407 } else {
1408 break;
1409 }
1410 } else {
1411 if (ret > 0) {
1412 i += ret;
1413 last_len += ret << PAGE_CACHE_SHIFT;
1414 } else {
1415 i++;
1416 last_len = 0;
1417 }
1418 }
1419 }
1420
1421 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1422 filemap_flush(inode->i_mapping);
1423 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1424 &BTRFS_I(inode)->runtime_flags))
1425 filemap_flush(inode->i_mapping);
1426 }
1427
1428 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1429 /* the filemap_flush will queue IO into the worker threads, but
1430 * we have to make sure the IO is actually started and that
1431 * ordered extents get created before we return
1432 */
1433 atomic_inc(&root->fs_info->async_submit_draining);
1434 while (atomic_read(&root->fs_info->nr_async_submits) ||
1435 atomic_read(&root->fs_info->async_delalloc_pages)) {
1436 wait_event(root->fs_info->async_submit_wait,
1437 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1438 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1439 }
1440 atomic_dec(&root->fs_info->async_submit_draining);
1441 }
1442
1443 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1444 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1445 }
1446
1447 ret = defrag_count;
1448
1449 out_ra:
1450 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1451 mutex_lock(&inode->i_mutex);
1452 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1453 mutex_unlock(&inode->i_mutex);
1454 }
1455 if (!file)
1456 kfree(ra);
1457 kfree(pages);
1458 return ret;
1459 }
1460
1461 static noinline int btrfs_ioctl_resize(struct file *file,
1462 void __user *arg)
1463 {
1464 u64 new_size;
1465 u64 old_size;
1466 u64 devid = 1;
1467 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1468 struct btrfs_ioctl_vol_args *vol_args;
1469 struct btrfs_trans_handle *trans;
1470 struct btrfs_device *device = NULL;
1471 char *sizestr;
1472 char *retptr;
1473 char *devstr = NULL;
1474 int ret = 0;
1475 int mod = 0;
1476
1477 if (!capable(CAP_SYS_ADMIN))
1478 return -EPERM;
1479
1480 ret = mnt_want_write_file(file);
1481 if (ret)
1482 return ret;
1483
1484 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1485 1)) {
1486 mnt_drop_write_file(file);
1487 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1488 }
1489
1490 mutex_lock(&root->fs_info->volume_mutex);
1491 vol_args = memdup_user(arg, sizeof(*vol_args));
1492 if (IS_ERR(vol_args)) {
1493 ret = PTR_ERR(vol_args);
1494 goto out;
1495 }
1496
1497 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1498
1499 sizestr = vol_args->name;
1500 devstr = strchr(sizestr, ':');
1501 if (devstr) {
1502 sizestr = devstr + 1;
1503 *devstr = '\0';
1504 devstr = vol_args->name;
1505 ret = kstrtoull(devstr, 10, &devid);
1506 if (ret)
1507 goto out_free;
1508 if (!devid) {
1509 ret = -EINVAL;
1510 goto out_free;
1511 }
1512 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1513 }
1514
1515 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1516 if (!device) {
1517 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1518 devid);
1519 ret = -ENODEV;
1520 goto out_free;
1521 }
1522
1523 if (!device->writeable) {
1524 btrfs_info(root->fs_info,
1525 "resizer unable to apply on readonly device %llu",
1526 devid);
1527 ret = -EPERM;
1528 goto out_free;
1529 }
1530
1531 if (!strcmp(sizestr, "max"))
1532 new_size = device->bdev->bd_inode->i_size;
1533 else {
1534 if (sizestr[0] == '-') {
1535 mod = -1;
1536 sizestr++;
1537 } else if (sizestr[0] == '+') {
1538 mod = 1;
1539 sizestr++;
1540 }
1541 new_size = memparse(sizestr, &retptr);
1542 if (*retptr != '\0' || new_size == 0) {
1543 ret = -EINVAL;
1544 goto out_free;
1545 }
1546 }
1547
1548 if (device->is_tgtdev_for_dev_replace) {
1549 ret = -EPERM;
1550 goto out_free;
1551 }
1552
1553 old_size = btrfs_device_get_total_bytes(device);
1554
1555 if (mod < 0) {
1556 if (new_size > old_size) {
1557 ret = -EINVAL;
1558 goto out_free;
1559 }
1560 new_size = old_size - new_size;
1561 } else if (mod > 0) {
1562 if (new_size > ULLONG_MAX - old_size) {
1563 ret = -ERANGE;
1564 goto out_free;
1565 }
1566 new_size = old_size + new_size;
1567 }
1568
1569 if (new_size < 256 * 1024 * 1024) {
1570 ret = -EINVAL;
1571 goto out_free;
1572 }
1573 if (new_size > device->bdev->bd_inode->i_size) {
1574 ret = -EFBIG;
1575 goto out_free;
1576 }
1577
1578 new_size = div_u64(new_size, root->sectorsize);
1579 new_size *= root->sectorsize;
1580
1581 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1582 rcu_str_deref(device->name), new_size);
1583
1584 if (new_size > old_size) {
1585 trans = btrfs_start_transaction(root, 0);
1586 if (IS_ERR(trans)) {
1587 ret = PTR_ERR(trans);
1588 goto out_free;
1589 }
1590 ret = btrfs_grow_device(trans, device, new_size);
1591 btrfs_commit_transaction(trans, root);
1592 } else if (new_size < old_size) {
1593 ret = btrfs_shrink_device(device, new_size);
1594 } /* equal, nothing need to do */
1595
1596 out_free:
1597 kfree(vol_args);
1598 out:
1599 mutex_unlock(&root->fs_info->volume_mutex);
1600 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1601 mnt_drop_write_file(file);
1602 return ret;
1603 }
1604
1605 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1606 char *name, unsigned long fd, int subvol,
1607 u64 *transid, bool readonly,
1608 struct btrfs_qgroup_inherit *inherit)
1609 {
1610 int namelen;
1611 int ret = 0;
1612
1613 if (!S_ISDIR(file_inode(file)->i_mode))
1614 return -ENOTDIR;
1615
1616 ret = mnt_want_write_file(file);
1617 if (ret)
1618 goto out;
1619
1620 namelen = strlen(name);
1621 if (strchr(name, '/')) {
1622 ret = -EINVAL;
1623 goto out_drop_write;
1624 }
1625
1626 if (name[0] == '.' &&
1627 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1628 ret = -EEXIST;
1629 goto out_drop_write;
1630 }
1631
1632 if (subvol) {
1633 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1634 NULL, transid, readonly, inherit);
1635 } else {
1636 struct fd src = fdget(fd);
1637 struct inode *src_inode;
1638 if (!src.file) {
1639 ret = -EINVAL;
1640 goto out_drop_write;
1641 }
1642
1643 src_inode = file_inode(src.file);
1644 if (src_inode->i_sb != file_inode(file)->i_sb) {
1645 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1646 "Snapshot src from another FS");
1647 ret = -EXDEV;
1648 } else if (!inode_owner_or_capable(src_inode)) {
1649 /*
1650 * Subvolume creation is not restricted, but snapshots
1651 * are limited to own subvolumes only
1652 */
1653 ret = -EPERM;
1654 } else {
1655 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1656 BTRFS_I(src_inode)->root,
1657 transid, readonly, inherit);
1658 }
1659 fdput(src);
1660 }
1661 out_drop_write:
1662 mnt_drop_write_file(file);
1663 out:
1664 return ret;
1665 }
1666
1667 static noinline int btrfs_ioctl_snap_create(struct file *file,
1668 void __user *arg, int subvol)
1669 {
1670 struct btrfs_ioctl_vol_args *vol_args;
1671 int ret;
1672
1673 if (!S_ISDIR(file_inode(file)->i_mode))
1674 return -ENOTDIR;
1675
1676 vol_args = memdup_user(arg, sizeof(*vol_args));
1677 if (IS_ERR(vol_args))
1678 return PTR_ERR(vol_args);
1679 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1680
1681 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1682 vol_args->fd, subvol,
1683 NULL, false, NULL);
1684
1685 kfree(vol_args);
1686 return ret;
1687 }
1688
1689 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1690 void __user *arg, int subvol)
1691 {
1692 struct btrfs_ioctl_vol_args_v2 *vol_args;
1693 int ret;
1694 u64 transid = 0;
1695 u64 *ptr = NULL;
1696 bool readonly = false;
1697 struct btrfs_qgroup_inherit *inherit = NULL;
1698
1699 if (!S_ISDIR(file_inode(file)->i_mode))
1700 return -ENOTDIR;
1701
1702 vol_args = memdup_user(arg, sizeof(*vol_args));
1703 if (IS_ERR(vol_args))
1704 return PTR_ERR(vol_args);
1705 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1706
1707 if (vol_args->flags &
1708 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1709 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1710 ret = -EOPNOTSUPP;
1711 goto free_args;
1712 }
1713
1714 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1715 ptr = &transid;
1716 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1717 readonly = true;
1718 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1719 if (vol_args->size > PAGE_CACHE_SIZE) {
1720 ret = -EINVAL;
1721 goto free_args;
1722 }
1723 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1724 if (IS_ERR(inherit)) {
1725 ret = PTR_ERR(inherit);
1726 goto free_args;
1727 }
1728 }
1729
1730 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1731 vol_args->fd, subvol, ptr,
1732 readonly, inherit);
1733 if (ret)
1734 goto free_inherit;
1735
1736 if (ptr && copy_to_user(arg +
1737 offsetof(struct btrfs_ioctl_vol_args_v2,
1738 transid),
1739 ptr, sizeof(*ptr)))
1740 ret = -EFAULT;
1741
1742 free_inherit:
1743 kfree(inherit);
1744 free_args:
1745 kfree(vol_args);
1746 return ret;
1747 }
1748
1749 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1750 void __user *arg)
1751 {
1752 struct inode *inode = file_inode(file);
1753 struct btrfs_root *root = BTRFS_I(inode)->root;
1754 int ret = 0;
1755 u64 flags = 0;
1756
1757 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1758 return -EINVAL;
1759
1760 down_read(&root->fs_info->subvol_sem);
1761 if (btrfs_root_readonly(root))
1762 flags |= BTRFS_SUBVOL_RDONLY;
1763 up_read(&root->fs_info->subvol_sem);
1764
1765 if (copy_to_user(arg, &flags, sizeof(flags)))
1766 ret = -EFAULT;
1767
1768 return ret;
1769 }
1770
1771 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1772 void __user *arg)
1773 {
1774 struct inode *inode = file_inode(file);
1775 struct btrfs_root *root = BTRFS_I(inode)->root;
1776 struct btrfs_trans_handle *trans;
1777 u64 root_flags;
1778 u64 flags;
1779 int ret = 0;
1780
1781 if (!inode_owner_or_capable(inode))
1782 return -EPERM;
1783
1784 ret = mnt_want_write_file(file);
1785 if (ret)
1786 goto out;
1787
1788 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1789 ret = -EINVAL;
1790 goto out_drop_write;
1791 }
1792
1793 if (copy_from_user(&flags, arg, sizeof(flags))) {
1794 ret = -EFAULT;
1795 goto out_drop_write;
1796 }
1797
1798 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1799 ret = -EINVAL;
1800 goto out_drop_write;
1801 }
1802
1803 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1804 ret = -EOPNOTSUPP;
1805 goto out_drop_write;
1806 }
1807
1808 down_write(&root->fs_info->subvol_sem);
1809
1810 /* nothing to do */
1811 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1812 goto out_drop_sem;
1813
1814 root_flags = btrfs_root_flags(&root->root_item);
1815 if (flags & BTRFS_SUBVOL_RDONLY) {
1816 btrfs_set_root_flags(&root->root_item,
1817 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1818 } else {
1819 /*
1820 * Block RO -> RW transition if this subvolume is involved in
1821 * send
1822 */
1823 spin_lock(&root->root_item_lock);
1824 if (root->send_in_progress == 0) {
1825 btrfs_set_root_flags(&root->root_item,
1826 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1827 spin_unlock(&root->root_item_lock);
1828 } else {
1829 spin_unlock(&root->root_item_lock);
1830 btrfs_warn(root->fs_info,
1831 "Attempt to set subvolume %llu read-write during send",
1832 root->root_key.objectid);
1833 ret = -EPERM;
1834 goto out_drop_sem;
1835 }
1836 }
1837
1838 trans = btrfs_start_transaction(root, 1);
1839 if (IS_ERR(trans)) {
1840 ret = PTR_ERR(trans);
1841 goto out_reset;
1842 }
1843
1844 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1845 &root->root_key, &root->root_item);
1846
1847 btrfs_commit_transaction(trans, root);
1848 out_reset:
1849 if (ret)
1850 btrfs_set_root_flags(&root->root_item, root_flags);
1851 out_drop_sem:
1852 up_write(&root->fs_info->subvol_sem);
1853 out_drop_write:
1854 mnt_drop_write_file(file);
1855 out:
1856 return ret;
1857 }
1858
1859 /*
1860 * helper to check if the subvolume references other subvolumes
1861 */
1862 static noinline int may_destroy_subvol(struct btrfs_root *root)
1863 {
1864 struct btrfs_path *path;
1865 struct btrfs_dir_item *di;
1866 struct btrfs_key key;
1867 u64 dir_id;
1868 int ret;
1869
1870 path = btrfs_alloc_path();
1871 if (!path)
1872 return -ENOMEM;
1873
1874 /* Make sure this root isn't set as the default subvol */
1875 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1876 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1877 dir_id, "default", 7, 0);
1878 if (di && !IS_ERR(di)) {
1879 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1880 if (key.objectid == root->root_key.objectid) {
1881 ret = -EPERM;
1882 btrfs_err(root->fs_info, "deleting default subvolume "
1883 "%llu is not allowed", key.objectid);
1884 goto out;
1885 }
1886 btrfs_release_path(path);
1887 }
1888
1889 key.objectid = root->root_key.objectid;
1890 key.type = BTRFS_ROOT_REF_KEY;
1891 key.offset = (u64)-1;
1892
1893 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1894 &key, path, 0, 0);
1895 if (ret < 0)
1896 goto out;
1897 BUG_ON(ret == 0);
1898
1899 ret = 0;
1900 if (path->slots[0] > 0) {
1901 path->slots[0]--;
1902 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1903 if (key.objectid == root->root_key.objectid &&
1904 key.type == BTRFS_ROOT_REF_KEY)
1905 ret = -ENOTEMPTY;
1906 }
1907 out:
1908 btrfs_free_path(path);
1909 return ret;
1910 }
1911
1912 static noinline int key_in_sk(struct btrfs_key *key,
1913 struct btrfs_ioctl_search_key *sk)
1914 {
1915 struct btrfs_key test;
1916 int ret;
1917
1918 test.objectid = sk->min_objectid;
1919 test.type = sk->min_type;
1920 test.offset = sk->min_offset;
1921
1922 ret = btrfs_comp_cpu_keys(key, &test);
1923 if (ret < 0)
1924 return 0;
1925
1926 test.objectid = sk->max_objectid;
1927 test.type = sk->max_type;
1928 test.offset = sk->max_offset;
1929
1930 ret = btrfs_comp_cpu_keys(key, &test);
1931 if (ret > 0)
1932 return 0;
1933 return 1;
1934 }
1935
1936 static noinline int copy_to_sk(struct btrfs_root *root,
1937 struct btrfs_path *path,
1938 struct btrfs_key *key,
1939 struct btrfs_ioctl_search_key *sk,
1940 size_t *buf_size,
1941 char __user *ubuf,
1942 unsigned long *sk_offset,
1943 int *num_found)
1944 {
1945 u64 found_transid;
1946 struct extent_buffer *leaf;
1947 struct btrfs_ioctl_search_header sh;
1948 unsigned long item_off;
1949 unsigned long item_len;
1950 int nritems;
1951 int i;
1952 int slot;
1953 int ret = 0;
1954
1955 leaf = path->nodes[0];
1956 slot = path->slots[0];
1957 nritems = btrfs_header_nritems(leaf);
1958
1959 if (btrfs_header_generation(leaf) > sk->max_transid) {
1960 i = nritems;
1961 goto advance_key;
1962 }
1963 found_transid = btrfs_header_generation(leaf);
1964
1965 for (i = slot; i < nritems; i++) {
1966 item_off = btrfs_item_ptr_offset(leaf, i);
1967 item_len = btrfs_item_size_nr(leaf, i);
1968
1969 btrfs_item_key_to_cpu(leaf, key, i);
1970 if (!key_in_sk(key, sk))
1971 continue;
1972
1973 if (sizeof(sh) + item_len > *buf_size) {
1974 if (*num_found) {
1975 ret = 1;
1976 goto out;
1977 }
1978
1979 /*
1980 * return one empty item back for v1, which does not
1981 * handle -EOVERFLOW
1982 */
1983
1984 *buf_size = sizeof(sh) + item_len;
1985 item_len = 0;
1986 ret = -EOVERFLOW;
1987 }
1988
1989 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1990 ret = 1;
1991 goto out;
1992 }
1993
1994 sh.objectid = key->objectid;
1995 sh.offset = key->offset;
1996 sh.type = key->type;
1997 sh.len = item_len;
1998 sh.transid = found_transid;
1999
2000 /* copy search result header */
2001 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2002 ret = -EFAULT;
2003 goto out;
2004 }
2005
2006 *sk_offset += sizeof(sh);
2007
2008 if (item_len) {
2009 char __user *up = ubuf + *sk_offset;
2010 /* copy the item */
2011 if (read_extent_buffer_to_user(leaf, up,
2012 item_off, item_len)) {
2013 ret = -EFAULT;
2014 goto out;
2015 }
2016
2017 *sk_offset += item_len;
2018 }
2019 (*num_found)++;
2020
2021 if (ret) /* -EOVERFLOW from above */
2022 goto out;
2023
2024 if (*num_found >= sk->nr_items) {
2025 ret = 1;
2026 goto out;
2027 }
2028 }
2029 advance_key:
2030 ret = 0;
2031 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
2032 key->offset++;
2033 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2034 key->offset = 0;
2035 key->type++;
2036 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2037 key->offset = 0;
2038 key->type = 0;
2039 key->objectid++;
2040 } else
2041 ret = 1;
2042 out:
2043 /*
2044 * 0: all items from this leaf copied, continue with next
2045 * 1: * more items can be copied, but unused buffer is too small
2046 * * all items were found
2047 * Either way, it will stops the loop which iterates to the next
2048 * leaf
2049 * -EOVERFLOW: item was to large for buffer
2050 * -EFAULT: could not copy extent buffer back to userspace
2051 */
2052 return ret;
2053 }
2054
2055 static noinline int search_ioctl(struct inode *inode,
2056 struct btrfs_ioctl_search_key *sk,
2057 size_t *buf_size,
2058 char __user *ubuf)
2059 {
2060 struct btrfs_root *root;
2061 struct btrfs_key key;
2062 struct btrfs_path *path;
2063 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2064 int ret;
2065 int num_found = 0;
2066 unsigned long sk_offset = 0;
2067
2068 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2069 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2070 return -EOVERFLOW;
2071 }
2072
2073 path = btrfs_alloc_path();
2074 if (!path)
2075 return -ENOMEM;
2076
2077 if (sk->tree_id == 0) {
2078 /* search the root of the inode that was passed */
2079 root = BTRFS_I(inode)->root;
2080 } else {
2081 key.objectid = sk->tree_id;
2082 key.type = BTRFS_ROOT_ITEM_KEY;
2083 key.offset = (u64)-1;
2084 root = btrfs_read_fs_root_no_name(info, &key);
2085 if (IS_ERR(root)) {
2086 printk(KERN_ERR "BTRFS: could not find root %llu\n",
2087 sk->tree_id);
2088 btrfs_free_path(path);
2089 return -ENOENT;
2090 }
2091 }
2092
2093 key.objectid = sk->min_objectid;
2094 key.type = sk->min_type;
2095 key.offset = sk->min_offset;
2096
2097 while (1) {
2098 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2099 if (ret != 0) {
2100 if (ret > 0)
2101 ret = 0;
2102 goto err;
2103 }
2104 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2105 &sk_offset, &num_found);
2106 btrfs_release_path(path);
2107 if (ret)
2108 break;
2109
2110 }
2111 if (ret > 0)
2112 ret = 0;
2113 err:
2114 sk->nr_items = num_found;
2115 btrfs_free_path(path);
2116 return ret;
2117 }
2118
2119 static noinline int btrfs_ioctl_tree_search(struct file *file,
2120 void __user *argp)
2121 {
2122 struct btrfs_ioctl_search_args __user *uargs;
2123 struct btrfs_ioctl_search_key sk;
2124 struct inode *inode;
2125 int ret;
2126 size_t buf_size;
2127
2128 if (!capable(CAP_SYS_ADMIN))
2129 return -EPERM;
2130
2131 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2132
2133 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2134 return -EFAULT;
2135
2136 buf_size = sizeof(uargs->buf);
2137
2138 inode = file_inode(file);
2139 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2140
2141 /*
2142 * In the origin implementation an overflow is handled by returning a
2143 * search header with a len of zero, so reset ret.
2144 */
2145 if (ret == -EOVERFLOW)
2146 ret = 0;
2147
2148 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2149 ret = -EFAULT;
2150 return ret;
2151 }
2152
2153 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2154 void __user *argp)
2155 {
2156 struct btrfs_ioctl_search_args_v2 __user *uarg;
2157 struct btrfs_ioctl_search_args_v2 args;
2158 struct inode *inode;
2159 int ret;
2160 size_t buf_size;
2161 const size_t buf_limit = 16 * 1024 * 1024;
2162
2163 if (!capable(CAP_SYS_ADMIN))
2164 return -EPERM;
2165
2166 /* copy search header and buffer size */
2167 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2168 if (copy_from_user(&args, uarg, sizeof(args)))
2169 return -EFAULT;
2170
2171 buf_size = args.buf_size;
2172
2173 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2174 return -EOVERFLOW;
2175
2176 /* limit result size to 16MB */
2177 if (buf_size > buf_limit)
2178 buf_size = buf_limit;
2179
2180 inode = file_inode(file);
2181 ret = search_ioctl(inode, &args.key, &buf_size,
2182 (char *)(&uarg->buf[0]));
2183 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2184 ret = -EFAULT;
2185 else if (ret == -EOVERFLOW &&
2186 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2187 ret = -EFAULT;
2188
2189 return ret;
2190 }
2191
2192 /*
2193 * Search INODE_REFs to identify path name of 'dirid' directory
2194 * in a 'tree_id' tree. and sets path name to 'name'.
2195 */
2196 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2197 u64 tree_id, u64 dirid, char *name)
2198 {
2199 struct btrfs_root *root;
2200 struct btrfs_key key;
2201 char *ptr;
2202 int ret = -1;
2203 int slot;
2204 int len;
2205 int total_len = 0;
2206 struct btrfs_inode_ref *iref;
2207 struct extent_buffer *l;
2208 struct btrfs_path *path;
2209
2210 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2211 name[0]='\0';
2212 return 0;
2213 }
2214
2215 path = btrfs_alloc_path();
2216 if (!path)
2217 return -ENOMEM;
2218
2219 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2220
2221 key.objectid = tree_id;
2222 key.type = BTRFS_ROOT_ITEM_KEY;
2223 key.offset = (u64)-1;
2224 root = btrfs_read_fs_root_no_name(info, &key);
2225 if (IS_ERR(root)) {
2226 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2227 ret = -ENOENT;
2228 goto out;
2229 }
2230
2231 key.objectid = dirid;
2232 key.type = BTRFS_INODE_REF_KEY;
2233 key.offset = (u64)-1;
2234
2235 while (1) {
2236 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2237 if (ret < 0)
2238 goto out;
2239 else if (ret > 0) {
2240 ret = btrfs_previous_item(root, path, dirid,
2241 BTRFS_INODE_REF_KEY);
2242 if (ret < 0)
2243 goto out;
2244 else if (ret > 0) {
2245 ret = -ENOENT;
2246 goto out;
2247 }
2248 }
2249
2250 l = path->nodes[0];
2251 slot = path->slots[0];
2252 btrfs_item_key_to_cpu(l, &key, slot);
2253
2254 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2255 len = btrfs_inode_ref_name_len(l, iref);
2256 ptr -= len + 1;
2257 total_len += len + 1;
2258 if (ptr < name) {
2259 ret = -ENAMETOOLONG;
2260 goto out;
2261 }
2262
2263 *(ptr + len) = '/';
2264 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2265
2266 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2267 break;
2268
2269 btrfs_release_path(path);
2270 key.objectid = key.offset;
2271 key.offset = (u64)-1;
2272 dirid = key.objectid;
2273 }
2274 memmove(name, ptr, total_len);
2275 name[total_len] = '\0';
2276 ret = 0;
2277 out:
2278 btrfs_free_path(path);
2279 return ret;
2280 }
2281
2282 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2283 void __user *argp)
2284 {
2285 struct btrfs_ioctl_ino_lookup_args *args;
2286 struct inode *inode;
2287 int ret;
2288
2289 if (!capable(CAP_SYS_ADMIN))
2290 return -EPERM;
2291
2292 args = memdup_user(argp, sizeof(*args));
2293 if (IS_ERR(args))
2294 return PTR_ERR(args);
2295
2296 inode = file_inode(file);
2297
2298 if (args->treeid == 0)
2299 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2300
2301 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2302 args->treeid, args->objectid,
2303 args->name);
2304
2305 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2306 ret = -EFAULT;
2307
2308 kfree(args);
2309 return ret;
2310 }
2311
2312 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2313 void __user *arg)
2314 {
2315 struct dentry *parent = file->f_path.dentry;
2316 struct dentry *dentry;
2317 struct inode *dir = d_inode(parent);
2318 struct inode *inode;
2319 struct btrfs_root *root = BTRFS_I(dir)->root;
2320 struct btrfs_root *dest = NULL;
2321 struct btrfs_ioctl_vol_args *vol_args;
2322 struct btrfs_trans_handle *trans;
2323 struct btrfs_block_rsv block_rsv;
2324 u64 root_flags;
2325 u64 qgroup_reserved;
2326 int namelen;
2327 int ret;
2328 int err = 0;
2329
2330 if (!S_ISDIR(dir->i_mode))
2331 return -ENOTDIR;
2332
2333 vol_args = memdup_user(arg, sizeof(*vol_args));
2334 if (IS_ERR(vol_args))
2335 return PTR_ERR(vol_args);
2336
2337 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2338 namelen = strlen(vol_args->name);
2339 if (strchr(vol_args->name, '/') ||
2340 strncmp(vol_args->name, "..", namelen) == 0) {
2341 err = -EINVAL;
2342 goto out;
2343 }
2344
2345 err = mnt_want_write_file(file);
2346 if (err)
2347 goto out;
2348
2349
2350 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2351 if (err == -EINTR)
2352 goto out_drop_write;
2353 dentry = lookup_one_len(vol_args->name, parent, namelen);
2354 if (IS_ERR(dentry)) {
2355 err = PTR_ERR(dentry);
2356 goto out_unlock_dir;
2357 }
2358
2359 if (d_really_is_negative(dentry)) {
2360 err = -ENOENT;
2361 goto out_dput;
2362 }
2363
2364 inode = d_inode(dentry);
2365 dest = BTRFS_I(inode)->root;
2366 if (!capable(CAP_SYS_ADMIN)) {
2367 /*
2368 * Regular user. Only allow this with a special mount
2369 * option, when the user has write+exec access to the
2370 * subvol root, and when rmdir(2) would have been
2371 * allowed.
2372 *
2373 * Note that this is _not_ check that the subvol is
2374 * empty or doesn't contain data that we wouldn't
2375 * otherwise be able to delete.
2376 *
2377 * Users who want to delete empty subvols should try
2378 * rmdir(2).
2379 */
2380 err = -EPERM;
2381 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2382 goto out_dput;
2383
2384 /*
2385 * Do not allow deletion if the parent dir is the same
2386 * as the dir to be deleted. That means the ioctl
2387 * must be called on the dentry referencing the root
2388 * of the subvol, not a random directory contained
2389 * within it.
2390 */
2391 err = -EINVAL;
2392 if (root == dest)
2393 goto out_dput;
2394
2395 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2396 if (err)
2397 goto out_dput;
2398 }
2399
2400 /* check if subvolume may be deleted by a user */
2401 err = btrfs_may_delete(dir, dentry, 1);
2402 if (err)
2403 goto out_dput;
2404
2405 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2406 err = -EINVAL;
2407 goto out_dput;
2408 }
2409
2410 mutex_lock(&inode->i_mutex);
2411
2412 /*
2413 * Don't allow to delete a subvolume with send in progress. This is
2414 * inside the i_mutex so the error handling that has to drop the bit
2415 * again is not run concurrently.
2416 */
2417 spin_lock(&dest->root_item_lock);
2418 root_flags = btrfs_root_flags(&dest->root_item);
2419 if (dest->send_in_progress == 0) {
2420 btrfs_set_root_flags(&dest->root_item,
2421 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2422 spin_unlock(&dest->root_item_lock);
2423 } else {
2424 spin_unlock(&dest->root_item_lock);
2425 btrfs_warn(root->fs_info,
2426 "Attempt to delete subvolume %llu during send",
2427 dest->root_key.objectid);
2428 err = -EPERM;
2429 goto out_unlock_inode;
2430 }
2431
2432 down_write(&root->fs_info->subvol_sem);
2433
2434 err = may_destroy_subvol(dest);
2435 if (err)
2436 goto out_up_write;
2437
2438 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2439 /*
2440 * One for dir inode, two for dir entries, two for root
2441 * ref/backref.
2442 */
2443 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2444 5, &qgroup_reserved, true);
2445 if (err)
2446 goto out_up_write;
2447
2448 trans = btrfs_start_transaction(root, 0);
2449 if (IS_ERR(trans)) {
2450 err = PTR_ERR(trans);
2451 goto out_release;
2452 }
2453 trans->block_rsv = &block_rsv;
2454 trans->bytes_reserved = block_rsv.size;
2455
2456 ret = btrfs_unlink_subvol(trans, root, dir,
2457 dest->root_key.objectid,
2458 dentry->d_name.name,
2459 dentry->d_name.len);
2460 if (ret) {
2461 err = ret;
2462 btrfs_abort_transaction(trans, root, ret);
2463 goto out_end_trans;
2464 }
2465
2466 btrfs_record_root_in_trans(trans, dest);
2467
2468 memset(&dest->root_item.drop_progress, 0,
2469 sizeof(dest->root_item.drop_progress));
2470 dest->root_item.drop_level = 0;
2471 btrfs_set_root_refs(&dest->root_item, 0);
2472
2473 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2474 ret = btrfs_insert_orphan_item(trans,
2475 root->fs_info->tree_root,
2476 dest->root_key.objectid);
2477 if (ret) {
2478 btrfs_abort_transaction(trans, root, ret);
2479 err = ret;
2480 goto out_end_trans;
2481 }
2482 }
2483
2484 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2485 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2486 dest->root_key.objectid);
2487 if (ret && ret != -ENOENT) {
2488 btrfs_abort_transaction(trans, root, ret);
2489 err = ret;
2490 goto out_end_trans;
2491 }
2492 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2493 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2494 dest->root_item.received_uuid,
2495 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2496 dest->root_key.objectid);
2497 if (ret && ret != -ENOENT) {
2498 btrfs_abort_transaction(trans, root, ret);
2499 err = ret;
2500 goto out_end_trans;
2501 }
2502 }
2503
2504 out_end_trans:
2505 trans->block_rsv = NULL;
2506 trans->bytes_reserved = 0;
2507 ret = btrfs_end_transaction(trans, root);
2508 if (ret && !err)
2509 err = ret;
2510 inode->i_flags |= S_DEAD;
2511 out_release:
2512 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2513 out_up_write:
2514 up_write(&root->fs_info->subvol_sem);
2515 if (err) {
2516 spin_lock(&dest->root_item_lock);
2517 root_flags = btrfs_root_flags(&dest->root_item);
2518 btrfs_set_root_flags(&dest->root_item,
2519 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2520 spin_unlock(&dest->root_item_lock);
2521 }
2522 out_unlock_inode:
2523 mutex_unlock(&inode->i_mutex);
2524 if (!err) {
2525 d_invalidate(dentry);
2526 btrfs_invalidate_inodes(dest);
2527 d_delete(dentry);
2528 ASSERT(dest->send_in_progress == 0);
2529
2530 /* the last ref */
2531 if (dest->ino_cache_inode) {
2532 iput(dest->ino_cache_inode);
2533 dest->ino_cache_inode = NULL;
2534 }
2535 }
2536 out_dput:
2537 dput(dentry);
2538 out_unlock_dir:
2539 mutex_unlock(&dir->i_mutex);
2540 out_drop_write:
2541 mnt_drop_write_file(file);
2542 out:
2543 kfree(vol_args);
2544 return err;
2545 }
2546
2547 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2548 {
2549 struct inode *inode = file_inode(file);
2550 struct btrfs_root *root = BTRFS_I(inode)->root;
2551 struct btrfs_ioctl_defrag_range_args *range;
2552 int ret;
2553
2554 ret = mnt_want_write_file(file);
2555 if (ret)
2556 return ret;
2557
2558 if (btrfs_root_readonly(root)) {
2559 ret = -EROFS;
2560 goto out;
2561 }
2562
2563 switch (inode->i_mode & S_IFMT) {
2564 case S_IFDIR:
2565 if (!capable(CAP_SYS_ADMIN)) {
2566 ret = -EPERM;
2567 goto out;
2568 }
2569 ret = btrfs_defrag_root(root);
2570 if (ret)
2571 goto out;
2572 ret = btrfs_defrag_root(root->fs_info->extent_root);
2573 break;
2574 case S_IFREG:
2575 if (!(file->f_mode & FMODE_WRITE)) {
2576 ret = -EINVAL;
2577 goto out;
2578 }
2579
2580 range = kzalloc(sizeof(*range), GFP_KERNEL);
2581 if (!range) {
2582 ret = -ENOMEM;
2583 goto out;
2584 }
2585
2586 if (argp) {
2587 if (copy_from_user(range, argp,
2588 sizeof(*range))) {
2589 ret = -EFAULT;
2590 kfree(range);
2591 goto out;
2592 }
2593 /* compression requires us to start the IO */
2594 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2595 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2596 range->extent_thresh = (u32)-1;
2597 }
2598 } else {
2599 /* the rest are all set to zero by kzalloc */
2600 range->len = (u64)-1;
2601 }
2602 ret = btrfs_defrag_file(file_inode(file), file,
2603 range, 0, 0);
2604 if (ret > 0)
2605 ret = 0;
2606 kfree(range);
2607 break;
2608 default:
2609 ret = -EINVAL;
2610 }
2611 out:
2612 mnt_drop_write_file(file);
2613 return ret;
2614 }
2615
2616 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2617 {
2618 struct btrfs_ioctl_vol_args *vol_args;
2619 int ret;
2620
2621 if (!capable(CAP_SYS_ADMIN))
2622 return -EPERM;
2623
2624 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2625 1)) {
2626 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2627 }
2628
2629 mutex_lock(&root->fs_info->volume_mutex);
2630 vol_args = memdup_user(arg, sizeof(*vol_args));
2631 if (IS_ERR(vol_args)) {
2632 ret = PTR_ERR(vol_args);
2633 goto out;
2634 }
2635
2636 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2637 ret = btrfs_init_new_device(root, vol_args->name);
2638
2639 if (!ret)
2640 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2641
2642 kfree(vol_args);
2643 out:
2644 mutex_unlock(&root->fs_info->volume_mutex);
2645 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2646 return ret;
2647 }
2648
2649 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2650 {
2651 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2652 struct btrfs_ioctl_vol_args *vol_args;
2653 int ret;
2654
2655 if (!capable(CAP_SYS_ADMIN))
2656 return -EPERM;
2657
2658 ret = mnt_want_write_file(file);
2659 if (ret)
2660 return ret;
2661
2662 vol_args = memdup_user(arg, sizeof(*vol_args));
2663 if (IS_ERR(vol_args)) {
2664 ret = PTR_ERR(vol_args);
2665 goto err_drop;
2666 }
2667
2668 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2669
2670 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2671 1)) {
2672 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2673 goto out;
2674 }
2675
2676 mutex_lock(&root->fs_info->volume_mutex);
2677 ret = btrfs_rm_device(root, vol_args->name);
2678 mutex_unlock(&root->fs_info->volume_mutex);
2679 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2680
2681 if (!ret)
2682 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2683
2684 out:
2685 kfree(vol_args);
2686 err_drop:
2687 mnt_drop_write_file(file);
2688 return ret;
2689 }
2690
2691 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2692 {
2693 struct btrfs_ioctl_fs_info_args *fi_args;
2694 struct btrfs_device *device;
2695 struct btrfs_device *next;
2696 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2697 int ret = 0;
2698
2699 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2700 if (!fi_args)
2701 return -ENOMEM;
2702
2703 mutex_lock(&fs_devices->device_list_mutex);
2704 fi_args->num_devices = fs_devices->num_devices;
2705 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2706
2707 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2708 if (device->devid > fi_args->max_id)
2709 fi_args->max_id = device->devid;
2710 }
2711 mutex_unlock(&fs_devices->device_list_mutex);
2712
2713 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2714 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2715 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2716
2717 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2718 ret = -EFAULT;
2719
2720 kfree(fi_args);
2721 return ret;
2722 }
2723
2724 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2725 {
2726 struct btrfs_ioctl_dev_info_args *di_args;
2727 struct btrfs_device *dev;
2728 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2729 int ret = 0;
2730 char *s_uuid = NULL;
2731
2732 di_args = memdup_user(arg, sizeof(*di_args));
2733 if (IS_ERR(di_args))
2734 return PTR_ERR(di_args);
2735
2736 if (!btrfs_is_empty_uuid(di_args->uuid))
2737 s_uuid = di_args->uuid;
2738
2739 mutex_lock(&fs_devices->device_list_mutex);
2740 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2741
2742 if (!dev) {
2743 ret = -ENODEV;
2744 goto out;
2745 }
2746
2747 di_args->devid = dev->devid;
2748 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2749 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2750 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2751 if (dev->name) {
2752 struct rcu_string *name;
2753
2754 rcu_read_lock();
2755 name = rcu_dereference(dev->name);
2756 strncpy(di_args->path, name->str, sizeof(di_args->path));
2757 rcu_read_unlock();
2758 di_args->path[sizeof(di_args->path) - 1] = 0;
2759 } else {
2760 di_args->path[0] = '\0';
2761 }
2762
2763 out:
2764 mutex_unlock(&fs_devices->device_list_mutex);
2765 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2766 ret = -EFAULT;
2767
2768 kfree(di_args);
2769 return ret;
2770 }
2771
2772 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2773 {
2774 struct page *page;
2775 pgoff_t index;
2776 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2777
2778 index = off >> PAGE_CACHE_SHIFT;
2779
2780 page = grab_cache_page(inode->i_mapping, index);
2781 if (!page)
2782 return NULL;
2783
2784 if (!PageUptodate(page)) {
2785 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2786 0))
2787 return NULL;
2788 lock_page(page);
2789 if (!PageUptodate(page)) {
2790 unlock_page(page);
2791 page_cache_release(page);
2792 return NULL;
2793 }
2794 }
2795 unlock_page(page);
2796
2797 return page;
2798 }
2799
2800 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2801 {
2802 /* do any pending delalloc/csum calc on src, one way or
2803 another, and lock file content */
2804 while (1) {
2805 struct btrfs_ordered_extent *ordered;
2806 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2807 ordered = btrfs_lookup_first_ordered_extent(inode,
2808 off + len - 1);
2809 if ((!ordered ||
2810 ordered->file_offset + ordered->len <= off ||
2811 ordered->file_offset >= off + len) &&
2812 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2813 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2814 if (ordered)
2815 btrfs_put_ordered_extent(ordered);
2816 break;
2817 }
2818 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2819 if (ordered)
2820 btrfs_put_ordered_extent(ordered);
2821 btrfs_wait_ordered_range(inode, off, len);
2822 }
2823 }
2824
2825 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2826 struct inode *inode2, u64 loff2, u64 len)
2827 {
2828 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2829 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2830
2831 mutex_unlock(&inode1->i_mutex);
2832 mutex_unlock(&inode2->i_mutex);
2833 }
2834
2835 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2836 struct inode *inode2, u64 loff2, u64 len)
2837 {
2838 if (inode1 < inode2) {
2839 swap(inode1, inode2);
2840 swap(loff1, loff2);
2841 }
2842
2843 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2844 lock_extent_range(inode1, loff1, len);
2845 if (inode1 != inode2) {
2846 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2847 lock_extent_range(inode2, loff2, len);
2848 }
2849 }
2850
2851 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2852 u64 dst_loff, u64 len)
2853 {
2854 int ret = 0;
2855 struct page *src_page, *dst_page;
2856 unsigned int cmp_len = PAGE_CACHE_SIZE;
2857 void *addr, *dst_addr;
2858
2859 while (len) {
2860 if (len < PAGE_CACHE_SIZE)
2861 cmp_len = len;
2862
2863 src_page = extent_same_get_page(src, loff);
2864 if (!src_page)
2865 return -EINVAL;
2866 dst_page = extent_same_get_page(dst, dst_loff);
2867 if (!dst_page) {
2868 page_cache_release(src_page);
2869 return -EINVAL;
2870 }
2871 addr = kmap_atomic(src_page);
2872 dst_addr = kmap_atomic(dst_page);
2873
2874 flush_dcache_page(src_page);
2875 flush_dcache_page(dst_page);
2876
2877 if (memcmp(addr, dst_addr, cmp_len))
2878 ret = BTRFS_SAME_DATA_DIFFERS;
2879
2880 kunmap_atomic(addr);
2881 kunmap_atomic(dst_addr);
2882 page_cache_release(src_page);
2883 page_cache_release(dst_page);
2884
2885 if (ret)
2886 break;
2887
2888 loff += cmp_len;
2889 dst_loff += cmp_len;
2890 len -= cmp_len;
2891 }
2892
2893 return ret;
2894 }
2895
2896 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2897 {
2898 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2899
2900 if (off + len > inode->i_size || off + len < off)
2901 return -EINVAL;
2902 /* Check that we are block aligned - btrfs_clone() requires this */
2903 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2904 return -EINVAL;
2905
2906 return 0;
2907 }
2908
2909 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2910 struct inode *dst, u64 dst_loff)
2911 {
2912 int ret;
2913
2914 /*
2915 * btrfs_clone() can't handle extents in the same file
2916 * yet. Once that works, we can drop this check and replace it
2917 * with a check for the same inode, but overlapping extents.
2918 */
2919 if (src == dst)
2920 return -EINVAL;
2921
2922 if (len == 0)
2923 return 0;
2924
2925 btrfs_double_lock(src, loff, dst, dst_loff, len);
2926
2927 ret = extent_same_check_offsets(src, loff, len);
2928 if (ret)
2929 goto out_unlock;
2930
2931 ret = extent_same_check_offsets(dst, dst_loff, len);
2932 if (ret)
2933 goto out_unlock;
2934
2935 /* don't make the dst file partly checksummed */
2936 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2937 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2938 ret = -EINVAL;
2939 goto out_unlock;
2940 }
2941
2942 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2943 if (ret == 0)
2944 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2945
2946 out_unlock:
2947 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2948
2949 return ret;
2950 }
2951
2952 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2953
2954 static long btrfs_ioctl_file_extent_same(struct file *file,
2955 struct btrfs_ioctl_same_args __user *argp)
2956 {
2957 struct btrfs_ioctl_same_args *same = NULL;
2958 struct btrfs_ioctl_same_extent_info *info;
2959 struct inode *src = file_inode(file);
2960 u64 off;
2961 u64 len;
2962 int i;
2963 int ret;
2964 unsigned long size;
2965 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2966 bool is_admin = capable(CAP_SYS_ADMIN);
2967 u16 count;
2968
2969 if (!(file->f_mode & FMODE_READ))
2970 return -EINVAL;
2971
2972 ret = mnt_want_write_file(file);
2973 if (ret)
2974 return ret;
2975
2976 if (get_user(count, &argp->dest_count)) {
2977 ret = -EFAULT;
2978 goto out;
2979 }
2980
2981 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
2982
2983 same = memdup_user(argp, size);
2984
2985 if (IS_ERR(same)) {
2986 ret = PTR_ERR(same);
2987 same = NULL;
2988 goto out;
2989 }
2990
2991 off = same->logical_offset;
2992 len = same->length;
2993
2994 /*
2995 * Limit the total length we will dedupe for each operation.
2996 * This is intended to bound the total time spent in this
2997 * ioctl to something sane.
2998 */
2999 if (len > BTRFS_MAX_DEDUPE_LEN)
3000 len = BTRFS_MAX_DEDUPE_LEN;
3001
3002 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
3003 /*
3004 * Btrfs does not support blocksize < page_size. As a
3005 * result, btrfs_cmp_data() won't correctly handle
3006 * this situation without an update.
3007 */
3008 ret = -EINVAL;
3009 goto out;
3010 }
3011
3012 ret = -EISDIR;
3013 if (S_ISDIR(src->i_mode))
3014 goto out;
3015
3016 ret = -EACCES;
3017 if (!S_ISREG(src->i_mode))
3018 goto out;
3019
3020 /* pre-format output fields to sane values */
3021 for (i = 0; i < count; i++) {
3022 same->info[i].bytes_deduped = 0ULL;
3023 same->info[i].status = 0;
3024 }
3025
3026 for (i = 0, info = same->info; i < count; i++, info++) {
3027 struct inode *dst;
3028 struct fd dst_file = fdget(info->fd);
3029 if (!dst_file.file) {
3030 info->status = -EBADF;
3031 continue;
3032 }
3033 dst = file_inode(dst_file.file);
3034
3035 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
3036 info->status = -EINVAL;
3037 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3038 info->status = -EXDEV;
3039 } else if (S_ISDIR(dst->i_mode)) {
3040 info->status = -EISDIR;
3041 } else if (!S_ISREG(dst->i_mode)) {
3042 info->status = -EACCES;
3043 } else {
3044 info->status = btrfs_extent_same(src, off, len, dst,
3045 info->logical_offset);
3046 if (info->status == 0)
3047 info->bytes_deduped += len;
3048 }
3049 fdput(dst_file);
3050 }
3051
3052 ret = copy_to_user(argp, same, size);
3053 if (ret)
3054 ret = -EFAULT;
3055
3056 out:
3057 mnt_drop_write_file(file);
3058 kfree(same);
3059 return ret;
3060 }
3061
3062 /* Helper to check and see if this root currently has a ref on the given disk
3063 * bytenr. If it does then we need to update the quota for this root. This
3064 * doesn't do anything if quotas aren't enabled.
3065 */
3066 static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3067 u64 disko)
3068 {
3069 struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem);
3070 struct ulist *roots;
3071 struct ulist_iterator uiter;
3072 struct ulist_node *root_node = NULL;
3073 int ret;
3074
3075 if (!root->fs_info->quota_enabled)
3076 return 1;
3077
3078 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3079 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3080 tree_mod_seq_elem.seq, &roots);
3081 if (ret < 0)
3082 goto out;
3083 ret = 0;
3084 ULIST_ITER_INIT(&uiter);
3085 while ((root_node = ulist_next(roots, &uiter))) {
3086 if (root_node->val == root->objectid) {
3087 ret = 1;
3088 break;
3089 }
3090 }
3091 ulist_free(roots);
3092 out:
3093 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3094 return ret;
3095 }
3096
3097 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3098 struct inode *inode,
3099 u64 endoff,
3100 const u64 destoff,
3101 const u64 olen)
3102 {
3103 struct btrfs_root *root = BTRFS_I(inode)->root;
3104 int ret;
3105
3106 inode_inc_iversion(inode);
3107 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3108 /*
3109 * We round up to the block size at eof when determining which
3110 * extents to clone above, but shouldn't round up the file size.
3111 */
3112 if (endoff > destoff + olen)
3113 endoff = destoff + olen;
3114 if (endoff > inode->i_size)
3115 btrfs_i_size_write(inode, endoff);
3116
3117 ret = btrfs_update_inode(trans, root, inode);
3118 if (ret) {
3119 btrfs_abort_transaction(trans, root, ret);
3120 btrfs_end_transaction(trans, root);
3121 goto out;
3122 }
3123 ret = btrfs_end_transaction(trans, root);
3124 out:
3125 return ret;
3126 }
3127
3128 static void clone_update_extent_map(struct inode *inode,
3129 const struct btrfs_trans_handle *trans,
3130 const struct btrfs_path *path,
3131 const u64 hole_offset,
3132 const u64 hole_len)
3133 {
3134 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3135 struct extent_map *em;
3136 int ret;
3137
3138 em = alloc_extent_map();
3139 if (!em) {
3140 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3141 &BTRFS_I(inode)->runtime_flags);
3142 return;
3143 }
3144
3145 if (path) {
3146 struct btrfs_file_extent_item *fi;
3147
3148 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3149 struct btrfs_file_extent_item);
3150 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3151 em->generation = -1;
3152 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3153 BTRFS_FILE_EXTENT_INLINE)
3154 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3155 &BTRFS_I(inode)->runtime_flags);
3156 } else {
3157 em->start = hole_offset;
3158 em->len = hole_len;
3159 em->ram_bytes = em->len;
3160 em->orig_start = hole_offset;
3161 em->block_start = EXTENT_MAP_HOLE;
3162 em->block_len = 0;
3163 em->orig_block_len = 0;
3164 em->compress_type = BTRFS_COMPRESS_NONE;
3165 em->generation = trans->transid;
3166 }
3167
3168 while (1) {
3169 write_lock(&em_tree->lock);
3170 ret = add_extent_mapping(em_tree, em, 1);
3171 write_unlock(&em_tree->lock);
3172 if (ret != -EEXIST) {
3173 free_extent_map(em);
3174 break;
3175 }
3176 btrfs_drop_extent_cache(inode, em->start,
3177 em->start + em->len - 1, 0);
3178 }
3179
3180 if (ret)
3181 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3182 &BTRFS_I(inode)->runtime_flags);
3183 }
3184
3185 /*
3186 * Make sure we do not end up inserting an inline extent into a file that has
3187 * already other (non-inline) extents. If a file has an inline extent it can
3188 * not have any other extents and the (single) inline extent must start at the
3189 * file offset 0. Failing to respect these rules will lead to file corruption,
3190 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3191 *
3192 * We can have extents that have been already written to disk or we can have
3193 * dirty ranges still in delalloc, in which case the extent maps and items are
3194 * created only when we run delalloc, and the delalloc ranges might fall outside
3195 * the range we are currently locking in the inode's io tree. So we check the
3196 * inode's i_size because of that (i_size updates are done while holding the
3197 * i_mutex, which we are holding here).
3198 * We also check to see if the inode has a size not greater than "datal" but has
3199 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3200 * protected against such concurrent fallocate calls by the i_mutex).
3201 *
3202 * If the file has no extents but a size greater than datal, do not allow the
3203 * copy because we would need turn the inline extent into a non-inline one (even
3204 * with NO_HOLES enabled). If we find our destination inode only has one inline
3205 * extent, just overwrite it with the source inline extent if its size is less
3206 * than the source extent's size, or we could copy the source inline extent's
3207 * data into the destination inode's inline extent if the later is greater then
3208 * the former.
3209 */
3210 static int clone_copy_inline_extent(struct inode *src,
3211 struct inode *dst,
3212 struct btrfs_trans_handle *trans,
3213 struct btrfs_path *path,
3214 struct btrfs_key *new_key,
3215 const u64 drop_start,
3216 const u64 datal,
3217 const u64 skip,
3218 const u64 size,
3219 char *inline_data)
3220 {
3221 struct btrfs_root *root = BTRFS_I(dst)->root;
3222 const u64 aligned_end = ALIGN(new_key->offset + datal,
3223 root->sectorsize);
3224 int ret;
3225 struct btrfs_key key;
3226
3227 if (new_key->offset > 0)
3228 return -EOPNOTSUPP;
3229
3230 key.objectid = btrfs_ino(dst);
3231 key.type = BTRFS_EXTENT_DATA_KEY;
3232 key.offset = 0;
3233 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3234 if (ret < 0) {
3235 return ret;
3236 } else if (ret > 0) {
3237 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3238 ret = btrfs_next_leaf(root, path);
3239 if (ret < 0)
3240 return ret;
3241 else if (ret > 0)
3242 goto copy_inline_extent;
3243 }
3244 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3245 if (key.objectid == btrfs_ino(dst) &&
3246 key.type == BTRFS_EXTENT_DATA_KEY) {
3247 ASSERT(key.offset > 0);
3248 return -EOPNOTSUPP;
3249 }
3250 } else if (i_size_read(dst) <= datal) {
3251 struct btrfs_file_extent_item *ei;
3252 u64 ext_len;
3253
3254 /*
3255 * If the file size is <= datal, make sure there are no other
3256 * extents following (can happen do to an fallocate call with
3257 * the flag FALLOC_FL_KEEP_SIZE).
3258 */
3259 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3260 struct btrfs_file_extent_item);
3261 /*
3262 * If it's an inline extent, it can not have other extents
3263 * following it.
3264 */
3265 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3266 BTRFS_FILE_EXTENT_INLINE)
3267 goto copy_inline_extent;
3268
3269 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3270 if (ext_len > aligned_end)
3271 return -EOPNOTSUPP;
3272
3273 ret = btrfs_next_item(root, path);
3274 if (ret < 0) {
3275 return ret;
3276 } else if (ret == 0) {
3277 btrfs_item_key_to_cpu(path->nodes[0], &key,
3278 path->slots[0]);
3279 if (key.objectid == btrfs_ino(dst) &&
3280 key.type == BTRFS_EXTENT_DATA_KEY)
3281 return -EOPNOTSUPP;
3282 }
3283 }
3284
3285 copy_inline_extent:
3286 /*
3287 * We have no extent items, or we have an extent at offset 0 which may
3288 * or may not be inlined. All these cases are dealt the same way.
3289 */
3290 if (i_size_read(dst) > datal) {
3291 /*
3292 * If the destination inode has an inline extent...
3293 * This would require copying the data from the source inline
3294 * extent into the beginning of the destination's inline extent.
3295 * But this is really complex, both extents can be compressed
3296 * or just one of them, which would require decompressing and
3297 * re-compressing data (which could increase the new compressed
3298 * size, not allowing the compressed data to fit anymore in an
3299 * inline extent).
3300 * So just don't support this case for now (it should be rare,
3301 * we are not really saving space when cloning inline extents).
3302 */
3303 return -EOPNOTSUPP;
3304 }
3305
3306 btrfs_release_path(path);
3307 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3308 if (ret)
3309 return ret;
3310 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3311 if (ret)
3312 return ret;
3313
3314 if (skip) {
3315 const u32 start = btrfs_file_extent_calc_inline_size(0);
3316
3317 memmove(inline_data + start, inline_data + start + skip, datal);
3318 }
3319
3320 write_extent_buffer(path->nodes[0], inline_data,
3321 btrfs_item_ptr_offset(path->nodes[0],
3322 path->slots[0]),
3323 size);
3324 inode_add_bytes(dst, datal);
3325
3326 return 0;
3327 }
3328
3329 /**
3330 * btrfs_clone() - clone a range from inode file to another
3331 *
3332 * @src: Inode to clone from
3333 * @inode: Inode to clone to
3334 * @off: Offset within source to start clone from
3335 * @olen: Original length, passed by user, of range to clone
3336 * @olen_aligned: Block-aligned value of olen, extent_same uses
3337 * identical values here
3338 * @destoff: Offset within @inode to start clone
3339 */
3340 static int btrfs_clone(struct inode *src, struct inode *inode,
3341 const u64 off, const u64 olen, const u64 olen_aligned,
3342 const u64 destoff)
3343 {
3344 struct btrfs_root *root = BTRFS_I(inode)->root;
3345 struct btrfs_path *path = NULL;
3346 struct extent_buffer *leaf;
3347 struct btrfs_trans_handle *trans;
3348 char *buf = NULL;
3349 struct btrfs_key key;
3350 u32 nritems;
3351 int slot;
3352 int ret;
3353 int no_quota;
3354 const u64 len = olen_aligned;
3355 u64 last_disko = 0;
3356 u64 last_dest_end = destoff;
3357
3358 ret = -ENOMEM;
3359 buf = vmalloc(root->nodesize);
3360 if (!buf)
3361 return ret;
3362
3363 path = btrfs_alloc_path();
3364 if (!path) {
3365 vfree(buf);
3366 return ret;
3367 }
3368
3369 path->reada = 2;
3370 /* clone data */
3371 key.objectid = btrfs_ino(src);
3372 key.type = BTRFS_EXTENT_DATA_KEY;
3373 key.offset = off;
3374
3375 while (1) {
3376 u64 next_key_min_offset = key.offset + 1;
3377
3378 /*
3379 * note the key will change type as we walk through the
3380 * tree.
3381 */
3382 path->leave_spinning = 1;
3383 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3384 0, 0);
3385 if (ret < 0)
3386 goto out;
3387 /*
3388 * First search, if no extent item that starts at offset off was
3389 * found but the previous item is an extent item, it's possible
3390 * it might overlap our target range, therefore process it.
3391 */
3392 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3393 btrfs_item_key_to_cpu(path->nodes[0], &key,
3394 path->slots[0] - 1);
3395 if (key.type == BTRFS_EXTENT_DATA_KEY)
3396 path->slots[0]--;
3397 }
3398
3399 nritems = btrfs_header_nritems(path->nodes[0]);
3400 process_slot:
3401 no_quota = 1;
3402 if (path->slots[0] >= nritems) {
3403 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3404 if (ret < 0)
3405 goto out;
3406 if (ret > 0)
3407 break;
3408 nritems = btrfs_header_nritems(path->nodes[0]);
3409 }
3410 leaf = path->nodes[0];
3411 slot = path->slots[0];
3412
3413 btrfs_item_key_to_cpu(leaf, &key, slot);
3414 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3415 key.objectid != btrfs_ino(src))
3416 break;
3417
3418 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3419 struct btrfs_file_extent_item *extent;
3420 int type;
3421 u32 size;
3422 struct btrfs_key new_key;
3423 u64 disko = 0, diskl = 0;
3424 u64 datao = 0, datal = 0;
3425 u8 comp;
3426 u64 drop_start;
3427
3428 extent = btrfs_item_ptr(leaf, slot,
3429 struct btrfs_file_extent_item);
3430 comp = btrfs_file_extent_compression(leaf, extent);
3431 type = btrfs_file_extent_type(leaf, extent);
3432 if (type == BTRFS_FILE_EXTENT_REG ||
3433 type == BTRFS_FILE_EXTENT_PREALLOC) {
3434 disko = btrfs_file_extent_disk_bytenr(leaf,
3435 extent);
3436 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3437 extent);
3438 datao = btrfs_file_extent_offset(leaf, extent);
3439 datal = btrfs_file_extent_num_bytes(leaf,
3440 extent);
3441 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3442 /* take upper bound, may be compressed */
3443 datal = btrfs_file_extent_ram_bytes(leaf,
3444 extent);
3445 }
3446
3447 /*
3448 * The first search might have left us at an extent
3449 * item that ends before our target range's start, can
3450 * happen if we have holes and NO_HOLES feature enabled.
3451 */
3452 if (key.offset + datal <= off) {
3453 path->slots[0]++;
3454 goto process_slot;
3455 } else if (key.offset >= off + len) {
3456 break;
3457 }
3458 next_key_min_offset = key.offset + datal;
3459 size = btrfs_item_size_nr(leaf, slot);
3460 read_extent_buffer(leaf, buf,
3461 btrfs_item_ptr_offset(leaf, slot),
3462 size);
3463
3464 btrfs_release_path(path);
3465 path->leave_spinning = 0;
3466
3467 memcpy(&new_key, &key, sizeof(new_key));
3468 new_key.objectid = btrfs_ino(inode);
3469 if (off <= key.offset)
3470 new_key.offset = key.offset + destoff - off;
3471 else
3472 new_key.offset = destoff;
3473
3474 /*
3475 * Deal with a hole that doesn't have an extent item
3476 * that represents it (NO_HOLES feature enabled).
3477 * This hole is either in the middle of the cloning
3478 * range or at the beginning (fully overlaps it or
3479 * partially overlaps it).
3480 */
3481 if (new_key.offset != last_dest_end)
3482 drop_start = last_dest_end;
3483 else
3484 drop_start = new_key.offset;
3485
3486 /*
3487 * 1 - adjusting old extent (we may have to split it)
3488 * 1 - add new extent
3489 * 1 - inode update
3490 */
3491 trans = btrfs_start_transaction(root, 3);
3492 if (IS_ERR(trans)) {
3493 ret = PTR_ERR(trans);
3494 goto out;
3495 }
3496
3497 if (type == BTRFS_FILE_EXTENT_REG ||
3498 type == BTRFS_FILE_EXTENT_PREALLOC) {
3499 /*
3500 * a | --- range to clone ---| b
3501 * | ------------- extent ------------- |
3502 */
3503
3504 /* subtract range b */
3505 if (key.offset + datal > off + len)
3506 datal = off + len - key.offset;
3507
3508 /* subtract range a */
3509 if (off > key.offset) {
3510 datao += off - key.offset;
3511 datal -= off - key.offset;
3512 }
3513
3514 ret = btrfs_drop_extents(trans, root, inode,
3515 drop_start,
3516 new_key.offset + datal,
3517 1);
3518 if (ret) {
3519 if (ret != -EOPNOTSUPP)
3520 btrfs_abort_transaction(trans,
3521 root, ret);
3522 btrfs_end_transaction(trans, root);
3523 goto out;
3524 }
3525
3526 ret = btrfs_insert_empty_item(trans, root, path,
3527 &new_key, size);
3528 if (ret) {
3529 btrfs_abort_transaction(trans, root,
3530 ret);
3531 btrfs_end_transaction(trans, root);
3532 goto out;
3533 }
3534
3535 leaf = path->nodes[0];
3536 slot = path->slots[0];
3537 write_extent_buffer(leaf, buf,
3538 btrfs_item_ptr_offset(leaf, slot),
3539 size);
3540
3541 extent = btrfs_item_ptr(leaf, slot,
3542 struct btrfs_file_extent_item);
3543
3544 /* disko == 0 means it's a hole */
3545 if (!disko)
3546 datao = 0;
3547
3548 btrfs_set_file_extent_offset(leaf, extent,
3549 datao);
3550 btrfs_set_file_extent_num_bytes(leaf, extent,
3551 datal);
3552
3553 /*
3554 * We need to look up the roots that point at
3555 * this bytenr and see if the new root does. If
3556 * it does not we need to make sure we update
3557 * quotas appropriately.
3558 */
3559 if (disko && root != BTRFS_I(src)->root &&
3560 disko != last_disko) {
3561 no_quota = check_ref(trans, root,
3562 disko);
3563 if (no_quota < 0) {
3564 btrfs_abort_transaction(trans,
3565 root,
3566 ret);
3567 btrfs_end_transaction(trans,
3568 root);
3569 ret = no_quota;
3570 goto out;
3571 }
3572 }
3573
3574 if (disko) {
3575 inode_add_bytes(inode, datal);
3576 ret = btrfs_inc_extent_ref(trans, root,
3577 disko, diskl, 0,
3578 root->root_key.objectid,
3579 btrfs_ino(inode),
3580 new_key.offset - datao,
3581 no_quota);
3582 if (ret) {
3583 btrfs_abort_transaction(trans,
3584 root,
3585 ret);
3586 btrfs_end_transaction(trans,
3587 root);
3588 goto out;
3589
3590 }
3591 }
3592 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3593 u64 skip = 0;
3594 u64 trim = 0;
3595
3596 if (off > key.offset) {
3597 skip = off - key.offset;
3598 new_key.offset += skip;
3599 }
3600
3601 if (key.offset + datal > off + len)
3602 trim = key.offset + datal - (off + len);
3603
3604 if (comp && (skip || trim)) {
3605 ret = -EINVAL;
3606 btrfs_end_transaction(trans, root);
3607 goto out;
3608 }
3609 size -= skip + trim;
3610 datal -= skip + trim;
3611
3612 ret = clone_copy_inline_extent(src, inode,
3613 trans, path,
3614 &new_key,
3615 drop_start,
3616 datal,
3617 skip, size, buf);
3618 if (ret) {
3619 if (ret != -EOPNOTSUPP)
3620 btrfs_abort_transaction(trans,
3621 root,
3622 ret);
3623 btrfs_end_transaction(trans, root);
3624 goto out;
3625 }
3626 leaf = path->nodes[0];
3627 slot = path->slots[0];
3628 }
3629
3630 /* If we have an implicit hole (NO_HOLES feature). */
3631 if (drop_start < new_key.offset)
3632 clone_update_extent_map(inode, trans,
3633 NULL, drop_start,
3634 new_key.offset - drop_start);
3635
3636 clone_update_extent_map(inode, trans, path, 0, 0);
3637
3638 btrfs_mark_buffer_dirty(leaf);
3639 btrfs_release_path(path);
3640
3641 last_dest_end = ALIGN(new_key.offset + datal,
3642 root->sectorsize);
3643 ret = clone_finish_inode_update(trans, inode,
3644 last_dest_end,
3645 destoff, olen);
3646 if (ret)
3647 goto out;
3648 if (new_key.offset + datal >= destoff + len)
3649 break;
3650 }
3651 btrfs_release_path(path);
3652 key.offset = next_key_min_offset;
3653 }
3654 ret = 0;
3655
3656 if (last_dest_end < destoff + len) {
3657 /*
3658 * We have an implicit hole (NO_HOLES feature is enabled) that
3659 * fully or partially overlaps our cloning range at its end.
3660 */
3661 btrfs_release_path(path);
3662
3663 /*
3664 * 1 - remove extent(s)
3665 * 1 - inode update
3666 */
3667 trans = btrfs_start_transaction(root, 2);
3668 if (IS_ERR(trans)) {
3669 ret = PTR_ERR(trans);
3670 goto out;
3671 }
3672 ret = btrfs_drop_extents(trans, root, inode,
3673 last_dest_end, destoff + len, 1);
3674 if (ret) {
3675 if (ret != -EOPNOTSUPP)
3676 btrfs_abort_transaction(trans, root, ret);
3677 btrfs_end_transaction(trans, root);
3678 goto out;
3679 }
3680 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3681 destoff + len - last_dest_end);
3682 ret = clone_finish_inode_update(trans, inode, destoff + len,
3683 destoff, olen);
3684 }
3685
3686 out:
3687 btrfs_free_path(path);
3688 vfree(buf);
3689 return ret;
3690 }
3691
3692 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3693 u64 off, u64 olen, u64 destoff)
3694 {
3695 struct inode *inode = file_inode(file);
3696 struct btrfs_root *root = BTRFS_I(inode)->root;
3697 struct fd src_file;
3698 struct inode *src;
3699 int ret;
3700 u64 len = olen;
3701 u64 bs = root->fs_info->sb->s_blocksize;
3702 int same_inode = 0;
3703
3704 /*
3705 * TODO:
3706 * - split compressed inline extents. annoying: we need to
3707 * decompress into destination's address_space (the file offset
3708 * may change, so source mapping won't do), then recompress (or
3709 * otherwise reinsert) a subrange.
3710 *
3711 * - split destination inode's inline extents. The inline extents can
3712 * be either compressed or non-compressed.
3713 */
3714
3715 /* the destination must be opened for writing */
3716 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3717 return -EINVAL;
3718
3719 if (btrfs_root_readonly(root))
3720 return -EROFS;
3721
3722 ret = mnt_want_write_file(file);
3723 if (ret)
3724 return ret;
3725
3726 src_file = fdget(srcfd);
3727 if (!src_file.file) {
3728 ret = -EBADF;
3729 goto out_drop_write;
3730 }
3731
3732 ret = -EXDEV;
3733 if (src_file.file->f_path.mnt != file->f_path.mnt)
3734 goto out_fput;
3735
3736 src = file_inode(src_file.file);
3737
3738 ret = -EINVAL;
3739 if (src == inode)
3740 same_inode = 1;
3741
3742 /* the src must be open for reading */
3743 if (!(src_file.file->f_mode & FMODE_READ))
3744 goto out_fput;
3745
3746 /* don't make the dst file partly checksummed */
3747 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3748 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3749 goto out_fput;
3750
3751 ret = -EISDIR;
3752 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3753 goto out_fput;
3754
3755 ret = -EXDEV;
3756 if (src->i_sb != inode->i_sb)
3757 goto out_fput;
3758
3759 if (!same_inode) {
3760 if (inode < src) {
3761 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3762 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3763 } else {
3764 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3765 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3766 }
3767 } else {
3768 mutex_lock(&src->i_mutex);
3769 }
3770
3771 /* determine range to clone */
3772 ret = -EINVAL;
3773 if (off + len > src->i_size || off + len < off)
3774 goto out_unlock;
3775 if (len == 0)
3776 olen = len = src->i_size - off;
3777 /* if we extend to eof, continue to block boundary */
3778 if (off + len == src->i_size)
3779 len = ALIGN(src->i_size, bs) - off;
3780
3781 if (len == 0) {
3782 ret = 0;
3783 goto out_unlock;
3784 }
3785
3786 /* verify the end result is block aligned */
3787 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3788 !IS_ALIGNED(destoff, bs))
3789 goto out_unlock;
3790
3791 /* verify if ranges are overlapped within the same file */
3792 if (same_inode) {
3793 if (destoff + len > off && destoff < off + len)
3794 goto out_unlock;
3795 }
3796
3797 if (destoff > inode->i_size) {
3798 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3799 if (ret)
3800 goto out_unlock;
3801 }
3802
3803 /*
3804 * Lock the target range too. Right after we replace the file extent
3805 * items in the fs tree (which now point to the cloned data), we might
3806 * have a worker replace them with extent items relative to a write
3807 * operation that was issued before this clone operation (i.e. confront
3808 * with inode.c:btrfs_finish_ordered_io).
3809 */
3810 if (same_inode) {
3811 u64 lock_start = min_t(u64, off, destoff);
3812 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3813
3814 lock_extent_range(src, lock_start, lock_len);
3815 } else {
3816 lock_extent_range(src, off, len);
3817 lock_extent_range(inode, destoff, len);
3818 }
3819
3820 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3821
3822 if (same_inode) {
3823 u64 lock_start = min_t(u64, off, destoff);
3824 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3825
3826 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3827 } else {
3828 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3829 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3830 destoff + len - 1);
3831 }
3832 /*
3833 * Truncate page cache pages so that future reads will see the cloned
3834 * data immediately and not the previous data.
3835 */
3836 truncate_inode_pages_range(&inode->i_data, destoff,
3837 PAGE_CACHE_ALIGN(destoff + len) - 1);
3838 out_unlock:
3839 if (!same_inode) {
3840 if (inode < src) {
3841 mutex_unlock(&src->i_mutex);
3842 mutex_unlock(&inode->i_mutex);
3843 } else {
3844 mutex_unlock(&inode->i_mutex);
3845 mutex_unlock(&src->i_mutex);
3846 }
3847 } else {
3848 mutex_unlock(&src->i_mutex);
3849 }
3850 out_fput:
3851 fdput(src_file);
3852 out_drop_write:
3853 mnt_drop_write_file(file);
3854 return ret;
3855 }
3856
3857 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3858 {
3859 struct btrfs_ioctl_clone_range_args args;
3860
3861 if (copy_from_user(&args, argp, sizeof(args)))
3862 return -EFAULT;
3863 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3864 args.src_length, args.dest_offset);
3865 }
3866
3867 /*
3868 * there are many ways the trans_start and trans_end ioctls can lead
3869 * to deadlocks. They should only be used by applications that
3870 * basically own the machine, and have a very in depth understanding
3871 * of all the possible deadlocks and enospc problems.
3872 */
3873 static long btrfs_ioctl_trans_start(struct file *file)
3874 {
3875 struct inode *inode = file_inode(file);
3876 struct btrfs_root *root = BTRFS_I(inode)->root;
3877 struct btrfs_trans_handle *trans;
3878 int ret;
3879
3880 ret = -EPERM;
3881 if (!capable(CAP_SYS_ADMIN))
3882 goto out;
3883
3884 ret = -EINPROGRESS;
3885 if (file->private_data)
3886 goto out;
3887
3888 ret = -EROFS;
3889 if (btrfs_root_readonly(root))
3890 goto out;
3891
3892 ret = mnt_want_write_file(file);
3893 if (ret)
3894 goto out;
3895
3896 atomic_inc(&root->fs_info->open_ioctl_trans);
3897
3898 ret = -ENOMEM;
3899 trans = btrfs_start_ioctl_transaction(root);
3900 if (IS_ERR(trans))
3901 goto out_drop;
3902
3903 file->private_data = trans;
3904 return 0;
3905
3906 out_drop:
3907 atomic_dec(&root->fs_info->open_ioctl_trans);
3908 mnt_drop_write_file(file);
3909 out:
3910 return ret;
3911 }
3912
3913 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3914 {
3915 struct inode *inode = file_inode(file);
3916 struct btrfs_root *root = BTRFS_I(inode)->root;
3917 struct btrfs_root *new_root;
3918 struct btrfs_dir_item *di;
3919 struct btrfs_trans_handle *trans;
3920 struct btrfs_path *path;
3921 struct btrfs_key location;
3922 struct btrfs_disk_key disk_key;
3923 u64 objectid = 0;
3924 u64 dir_id;
3925 int ret;
3926
3927 if (!capable(CAP_SYS_ADMIN))
3928 return -EPERM;
3929
3930 ret = mnt_want_write_file(file);
3931 if (ret)
3932 return ret;
3933
3934 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3935 ret = -EFAULT;
3936 goto out;
3937 }
3938
3939 if (!objectid)
3940 objectid = BTRFS_FS_TREE_OBJECTID;
3941
3942 location.objectid = objectid;
3943 location.type = BTRFS_ROOT_ITEM_KEY;
3944 location.offset = (u64)-1;
3945
3946 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3947 if (IS_ERR(new_root)) {
3948 ret = PTR_ERR(new_root);
3949 goto out;
3950 }
3951
3952 path = btrfs_alloc_path();
3953 if (!path) {
3954 ret = -ENOMEM;
3955 goto out;
3956 }
3957 path->leave_spinning = 1;
3958
3959 trans = btrfs_start_transaction(root, 1);
3960 if (IS_ERR(trans)) {
3961 btrfs_free_path(path);
3962 ret = PTR_ERR(trans);
3963 goto out;
3964 }
3965
3966 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3967 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3968 dir_id, "default", 7, 1);
3969 if (IS_ERR_OR_NULL(di)) {
3970 btrfs_free_path(path);
3971 btrfs_end_transaction(trans, root);
3972 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3973 "item, this isn't going to work");
3974 ret = -ENOENT;
3975 goto out;
3976 }
3977
3978 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3979 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3980 btrfs_mark_buffer_dirty(path->nodes[0]);
3981 btrfs_free_path(path);
3982
3983 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3984 btrfs_end_transaction(trans, root);
3985 out:
3986 mnt_drop_write_file(file);
3987 return ret;
3988 }
3989
3990 void btrfs_get_block_group_info(struct list_head *groups_list,
3991 struct btrfs_ioctl_space_info *space)
3992 {
3993 struct btrfs_block_group_cache *block_group;
3994
3995 space->total_bytes = 0;
3996 space->used_bytes = 0;
3997 space->flags = 0;
3998 list_for_each_entry(block_group, groups_list, list) {
3999 space->flags = block_group->flags;
4000 space->total_bytes += block_group->key.offset;
4001 space->used_bytes +=
4002 btrfs_block_group_used(&block_group->item);
4003 }
4004 }
4005
4006 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
4007 {
4008 struct btrfs_ioctl_space_args space_args;
4009 struct btrfs_ioctl_space_info space;
4010 struct btrfs_ioctl_space_info *dest;
4011 struct btrfs_ioctl_space_info *dest_orig;
4012 struct btrfs_ioctl_space_info __user *user_dest;
4013 struct btrfs_space_info *info;
4014 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4015 BTRFS_BLOCK_GROUP_SYSTEM,
4016 BTRFS_BLOCK_GROUP_METADATA,
4017 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4018 int num_types = 4;
4019 int alloc_size;
4020 int ret = 0;
4021 u64 slot_count = 0;
4022 int i, c;
4023
4024 if (copy_from_user(&space_args,
4025 (struct btrfs_ioctl_space_args __user *)arg,
4026 sizeof(space_args)))
4027 return -EFAULT;
4028
4029 for (i = 0; i < num_types; i++) {
4030 struct btrfs_space_info *tmp;
4031
4032 info = NULL;
4033 rcu_read_lock();
4034 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4035 list) {
4036 if (tmp->flags == types[i]) {
4037 info = tmp;
4038 break;
4039 }
4040 }
4041 rcu_read_unlock();
4042
4043 if (!info)
4044 continue;
4045
4046 down_read(&info->groups_sem);
4047 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4048 if (!list_empty(&info->block_groups[c]))
4049 slot_count++;
4050 }
4051 up_read(&info->groups_sem);
4052 }
4053
4054 /*
4055 * Global block reserve, exported as a space_info
4056 */
4057 slot_count++;
4058
4059 /* space_slots == 0 means they are asking for a count */
4060 if (space_args.space_slots == 0) {
4061 space_args.total_spaces = slot_count;
4062 goto out;
4063 }
4064
4065 slot_count = min_t(u64, space_args.space_slots, slot_count);
4066
4067 alloc_size = sizeof(*dest) * slot_count;
4068
4069 /* we generally have at most 6 or so space infos, one for each raid
4070 * level. So, a whole page should be more than enough for everyone
4071 */
4072 if (alloc_size > PAGE_CACHE_SIZE)
4073 return -ENOMEM;
4074
4075 space_args.total_spaces = 0;
4076 dest = kmalloc(alloc_size, GFP_NOFS);
4077 if (!dest)
4078 return -ENOMEM;
4079 dest_orig = dest;
4080
4081 /* now we have a buffer to copy into */
4082 for (i = 0; i < num_types; i++) {
4083 struct btrfs_space_info *tmp;
4084
4085 if (!slot_count)
4086 break;
4087
4088 info = NULL;
4089 rcu_read_lock();
4090 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4091 list) {
4092 if (tmp->flags == types[i]) {
4093 info = tmp;
4094 break;
4095 }
4096 }
4097 rcu_read_unlock();
4098
4099 if (!info)
4100 continue;
4101 down_read(&info->groups_sem);
4102 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4103 if (!list_empty(&info->block_groups[c])) {
4104 btrfs_get_block_group_info(
4105 &info->block_groups[c], &space);
4106 memcpy(dest, &space, sizeof(space));
4107 dest++;
4108 space_args.total_spaces++;
4109 slot_count--;
4110 }
4111 if (!slot_count)
4112 break;
4113 }
4114 up_read(&info->groups_sem);
4115 }
4116
4117 /*
4118 * Add global block reserve
4119 */
4120 if (slot_count) {
4121 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4122
4123 spin_lock(&block_rsv->lock);
4124 space.total_bytes = block_rsv->size;
4125 space.used_bytes = block_rsv->size - block_rsv->reserved;
4126 spin_unlock(&block_rsv->lock);
4127 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4128 memcpy(dest, &space, sizeof(space));
4129 space_args.total_spaces++;
4130 }
4131
4132 user_dest = (struct btrfs_ioctl_space_info __user *)
4133 (arg + sizeof(struct btrfs_ioctl_space_args));
4134
4135 if (copy_to_user(user_dest, dest_orig, alloc_size))
4136 ret = -EFAULT;
4137
4138 kfree(dest_orig);
4139 out:
4140 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4141 ret = -EFAULT;
4142
4143 return ret;
4144 }
4145
4146 /*
4147 * there are many ways the trans_start and trans_end ioctls can lead
4148 * to deadlocks. They should only be used by applications that
4149 * basically own the machine, and have a very in depth understanding
4150 * of all the possible deadlocks and enospc problems.
4151 */
4152 long btrfs_ioctl_trans_end(struct file *file)
4153 {
4154 struct inode *inode = file_inode(file);
4155 struct btrfs_root *root = BTRFS_I(inode)->root;
4156 struct btrfs_trans_handle *trans;
4157
4158 trans = file->private_data;
4159 if (!trans)
4160 return -EINVAL;
4161 file->private_data = NULL;
4162
4163 btrfs_end_transaction(trans, root);
4164
4165 atomic_dec(&root->fs_info->open_ioctl_trans);
4166
4167 mnt_drop_write_file(file);
4168 return 0;
4169 }
4170
4171 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4172 void __user *argp)
4173 {
4174 struct btrfs_trans_handle *trans;
4175 u64 transid;
4176 int ret;
4177
4178 trans = btrfs_attach_transaction_barrier(root);
4179 if (IS_ERR(trans)) {
4180 if (PTR_ERR(trans) != -ENOENT)
4181 return PTR_ERR(trans);
4182
4183 /* No running transaction, don't bother */
4184 transid = root->fs_info->last_trans_committed;
4185 goto out;
4186 }
4187 transid = trans->transid;
4188 ret = btrfs_commit_transaction_async(trans, root, 0);
4189 if (ret) {
4190 btrfs_end_transaction(trans, root);
4191 return ret;
4192 }
4193 out:
4194 if (argp)
4195 if (copy_to_user(argp, &transid, sizeof(transid)))
4196 return -EFAULT;
4197 return 0;
4198 }
4199
4200 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4201 void __user *argp)
4202 {
4203 u64 transid;
4204
4205 if (argp) {
4206 if (copy_from_user(&transid, argp, sizeof(transid)))
4207 return -EFAULT;
4208 } else {
4209 transid = 0; /* current trans */
4210 }
4211 return btrfs_wait_for_commit(root, transid);
4212 }
4213
4214 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4215 {
4216 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4217 struct btrfs_ioctl_scrub_args *sa;
4218 int ret;
4219
4220 if (!capable(CAP_SYS_ADMIN))
4221 return -EPERM;
4222
4223 sa = memdup_user(arg, sizeof(*sa));
4224 if (IS_ERR(sa))
4225 return PTR_ERR(sa);
4226
4227 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4228 ret = mnt_want_write_file(file);
4229 if (ret)
4230 goto out;
4231 }
4232
4233 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4234 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4235 0);
4236
4237 if (copy_to_user(arg, sa, sizeof(*sa)))
4238 ret = -EFAULT;
4239
4240 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4241 mnt_drop_write_file(file);
4242 out:
4243 kfree(sa);
4244 return ret;
4245 }
4246
4247 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4248 {
4249 if (!capable(CAP_SYS_ADMIN))
4250 return -EPERM;
4251
4252 return btrfs_scrub_cancel(root->fs_info);
4253 }
4254
4255 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4256 void __user *arg)
4257 {
4258 struct btrfs_ioctl_scrub_args *sa;
4259 int ret;
4260
4261 if (!capable(CAP_SYS_ADMIN))
4262 return -EPERM;
4263
4264 sa = memdup_user(arg, sizeof(*sa));
4265 if (IS_ERR(sa))
4266 return PTR_ERR(sa);
4267
4268 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4269
4270 if (copy_to_user(arg, sa, sizeof(*sa)))
4271 ret = -EFAULT;
4272
4273 kfree(sa);
4274 return ret;
4275 }
4276
4277 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4278 void __user *arg)
4279 {
4280 struct btrfs_ioctl_get_dev_stats *sa;
4281 int ret;
4282
4283 sa = memdup_user(arg, sizeof(*sa));
4284 if (IS_ERR(sa))
4285 return PTR_ERR(sa);
4286
4287 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4288 kfree(sa);
4289 return -EPERM;
4290 }
4291
4292 ret = btrfs_get_dev_stats(root, sa);
4293
4294 if (copy_to_user(arg, sa, sizeof(*sa)))
4295 ret = -EFAULT;
4296
4297 kfree(sa);
4298 return ret;
4299 }
4300
4301 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4302 {
4303 struct btrfs_ioctl_dev_replace_args *p;
4304 int ret;
4305
4306 if (!capable(CAP_SYS_ADMIN))
4307 return -EPERM;
4308
4309 p = memdup_user(arg, sizeof(*p));
4310 if (IS_ERR(p))
4311 return PTR_ERR(p);
4312
4313 switch (p->cmd) {
4314 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4315 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4316 ret = -EROFS;
4317 goto out;
4318 }
4319 if (atomic_xchg(
4320 &root->fs_info->mutually_exclusive_operation_running,
4321 1)) {
4322 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4323 } else {
4324 ret = btrfs_dev_replace_start(root, p);
4325 atomic_set(
4326 &root->fs_info->mutually_exclusive_operation_running,
4327 0);
4328 }
4329 break;
4330 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4331 btrfs_dev_replace_status(root->fs_info, p);
4332 ret = 0;
4333 break;
4334 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4335 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4336 break;
4337 default:
4338 ret = -EINVAL;
4339 break;
4340 }
4341
4342 if (copy_to_user(arg, p, sizeof(*p)))
4343 ret = -EFAULT;
4344 out:
4345 kfree(p);
4346 return ret;
4347 }
4348
4349 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4350 {
4351 int ret = 0;
4352 int i;
4353 u64 rel_ptr;
4354 int size;
4355 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4356 struct inode_fs_paths *ipath = NULL;
4357 struct btrfs_path *path;
4358
4359 if (!capable(CAP_DAC_READ_SEARCH))
4360 return -EPERM;
4361
4362 path = btrfs_alloc_path();
4363 if (!path) {
4364 ret = -ENOMEM;
4365 goto out;
4366 }
4367
4368 ipa = memdup_user(arg, sizeof(*ipa));
4369 if (IS_ERR(ipa)) {
4370 ret = PTR_ERR(ipa);
4371 ipa = NULL;
4372 goto out;
4373 }
4374
4375 size = min_t(u32, ipa->size, 4096);
4376 ipath = init_ipath(size, root, path);
4377 if (IS_ERR(ipath)) {
4378 ret = PTR_ERR(ipath);
4379 ipath = NULL;
4380 goto out;
4381 }
4382
4383 ret = paths_from_inode(ipa->inum, ipath);
4384 if (ret < 0)
4385 goto out;
4386
4387 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4388 rel_ptr = ipath->fspath->val[i] -
4389 (u64)(unsigned long)ipath->fspath->val;
4390 ipath->fspath->val[i] = rel_ptr;
4391 }
4392
4393 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4394 (void *)(unsigned long)ipath->fspath, size);
4395 if (ret) {
4396 ret = -EFAULT;
4397 goto out;
4398 }
4399
4400 out:
4401 btrfs_free_path(path);
4402 free_ipath(ipath);
4403 kfree(ipa);
4404
4405 return ret;
4406 }
4407
4408 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4409 {
4410 struct btrfs_data_container *inodes = ctx;
4411 const size_t c = 3 * sizeof(u64);
4412
4413 if (inodes->bytes_left >= c) {
4414 inodes->bytes_left -= c;
4415 inodes->val[inodes->elem_cnt] = inum;
4416 inodes->val[inodes->elem_cnt + 1] = offset;
4417 inodes->val[inodes->elem_cnt + 2] = root;
4418 inodes->elem_cnt += 3;
4419 } else {
4420 inodes->bytes_missing += c - inodes->bytes_left;
4421 inodes->bytes_left = 0;
4422 inodes->elem_missed += 3;
4423 }
4424
4425 return 0;
4426 }
4427
4428 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4429 void __user *arg)
4430 {
4431 int ret = 0;
4432 int size;
4433 struct btrfs_ioctl_logical_ino_args *loi;
4434 struct btrfs_data_container *inodes = NULL;
4435 struct btrfs_path *path = NULL;
4436
4437 if (!capable(CAP_SYS_ADMIN))
4438 return -EPERM;
4439
4440 loi = memdup_user(arg, sizeof(*loi));
4441 if (IS_ERR(loi)) {
4442 ret = PTR_ERR(loi);
4443 loi = NULL;
4444 goto out;
4445 }
4446
4447 path = btrfs_alloc_path();
4448 if (!path) {
4449 ret = -ENOMEM;
4450 goto out;
4451 }
4452
4453 size = min_t(u32, loi->size, 64 * 1024);
4454 inodes = init_data_container(size);
4455 if (IS_ERR(inodes)) {
4456 ret = PTR_ERR(inodes);
4457 inodes = NULL;
4458 goto out;
4459 }
4460
4461 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4462 build_ino_list, inodes);
4463 if (ret == -EINVAL)
4464 ret = -ENOENT;
4465 if (ret < 0)
4466 goto out;
4467
4468 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4469 (void *)(unsigned long)inodes, size);
4470 if (ret)
4471 ret = -EFAULT;
4472
4473 out:
4474 btrfs_free_path(path);
4475 vfree(inodes);
4476 kfree(loi);
4477
4478 return ret;
4479 }
4480
4481 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4482 struct btrfs_ioctl_balance_args *bargs)
4483 {
4484 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4485
4486 bargs->flags = bctl->flags;
4487
4488 if (atomic_read(&fs_info->balance_running))
4489 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4490 if (atomic_read(&fs_info->balance_pause_req))
4491 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4492 if (atomic_read(&fs_info->balance_cancel_req))
4493 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4494
4495 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4496 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4497 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4498
4499 if (lock) {
4500 spin_lock(&fs_info->balance_lock);
4501 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4502 spin_unlock(&fs_info->balance_lock);
4503 } else {
4504 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4505 }
4506 }
4507
4508 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4509 {
4510 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4511 struct btrfs_fs_info *fs_info = root->fs_info;
4512 struct btrfs_ioctl_balance_args *bargs;
4513 struct btrfs_balance_control *bctl;
4514 bool need_unlock; /* for mut. excl. ops lock */
4515 int ret;
4516
4517 if (!capable(CAP_SYS_ADMIN))
4518 return -EPERM;
4519
4520 ret = mnt_want_write_file(file);
4521 if (ret)
4522 return ret;
4523
4524 again:
4525 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4526 mutex_lock(&fs_info->volume_mutex);
4527 mutex_lock(&fs_info->balance_mutex);
4528 need_unlock = true;
4529 goto locked;
4530 }
4531
4532 /*
4533 * mut. excl. ops lock is locked. Three possibilites:
4534 * (1) some other op is running
4535 * (2) balance is running
4536 * (3) balance is paused -- special case (think resume)
4537 */
4538 mutex_lock(&fs_info->balance_mutex);
4539 if (fs_info->balance_ctl) {
4540 /* this is either (2) or (3) */
4541 if (!atomic_read(&fs_info->balance_running)) {
4542 mutex_unlock(&fs_info->balance_mutex);
4543 if (!mutex_trylock(&fs_info->volume_mutex))
4544 goto again;
4545 mutex_lock(&fs_info->balance_mutex);
4546
4547 if (fs_info->balance_ctl &&
4548 !atomic_read(&fs_info->balance_running)) {
4549 /* this is (3) */
4550 need_unlock = false;
4551 goto locked;
4552 }
4553
4554 mutex_unlock(&fs_info->balance_mutex);
4555 mutex_unlock(&fs_info->volume_mutex);
4556 goto again;
4557 } else {
4558 /* this is (2) */
4559 mutex_unlock(&fs_info->balance_mutex);
4560 ret = -EINPROGRESS;
4561 goto out;
4562 }
4563 } else {
4564 /* this is (1) */
4565 mutex_unlock(&fs_info->balance_mutex);
4566 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4567 goto out;
4568 }
4569
4570 locked:
4571 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4572
4573 if (arg) {
4574 bargs = memdup_user(arg, sizeof(*bargs));
4575 if (IS_ERR(bargs)) {
4576 ret = PTR_ERR(bargs);
4577 goto out_unlock;
4578 }
4579
4580 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4581 if (!fs_info->balance_ctl) {
4582 ret = -ENOTCONN;
4583 goto out_bargs;
4584 }
4585
4586 bctl = fs_info->balance_ctl;
4587 spin_lock(&fs_info->balance_lock);
4588 bctl->flags |= BTRFS_BALANCE_RESUME;
4589 spin_unlock(&fs_info->balance_lock);
4590
4591 goto do_balance;
4592 }
4593 } else {
4594 bargs = NULL;
4595 }
4596
4597 if (fs_info->balance_ctl) {
4598 ret = -EINPROGRESS;
4599 goto out_bargs;
4600 }
4601
4602 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4603 if (!bctl) {
4604 ret = -ENOMEM;
4605 goto out_bargs;
4606 }
4607
4608 bctl->fs_info = fs_info;
4609 if (arg) {
4610 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4611 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4612 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4613
4614 bctl->flags = bargs->flags;
4615 } else {
4616 /* balance everything - no filters */
4617 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4618 }
4619
4620 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4621 ret = -EINVAL;
4622 goto out_bctl;
4623 }
4624
4625 do_balance:
4626 /*
4627 * Ownership of bctl and mutually_exclusive_operation_running
4628 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4629 * or, if restriper was paused all the way until unmount, in
4630 * free_fs_info. mutually_exclusive_operation_running is
4631 * cleared in __cancel_balance.
4632 */
4633 need_unlock = false;
4634
4635 ret = btrfs_balance(bctl, bargs);
4636 bctl = NULL;
4637
4638 if (arg) {
4639 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4640 ret = -EFAULT;
4641 }
4642
4643 out_bctl:
4644 kfree(bctl);
4645 out_bargs:
4646 kfree(bargs);
4647 out_unlock:
4648 mutex_unlock(&fs_info->balance_mutex);
4649 mutex_unlock(&fs_info->volume_mutex);
4650 if (need_unlock)
4651 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4652 out:
4653 mnt_drop_write_file(file);
4654 return ret;
4655 }
4656
4657 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4658 {
4659 if (!capable(CAP_SYS_ADMIN))
4660 return -EPERM;
4661
4662 switch (cmd) {
4663 case BTRFS_BALANCE_CTL_PAUSE:
4664 return btrfs_pause_balance(root->fs_info);
4665 case BTRFS_BALANCE_CTL_CANCEL:
4666 return btrfs_cancel_balance(root->fs_info);
4667 }
4668
4669 return -EINVAL;
4670 }
4671
4672 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4673 void __user *arg)
4674 {
4675 struct btrfs_fs_info *fs_info = root->fs_info;
4676 struct btrfs_ioctl_balance_args *bargs;
4677 int ret = 0;
4678
4679 if (!capable(CAP_SYS_ADMIN))
4680 return -EPERM;
4681
4682 mutex_lock(&fs_info->balance_mutex);
4683 if (!fs_info->balance_ctl) {
4684 ret = -ENOTCONN;
4685 goto out;
4686 }
4687
4688 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4689 if (!bargs) {
4690 ret = -ENOMEM;
4691 goto out;
4692 }
4693
4694 update_ioctl_balance_args(fs_info, 1, bargs);
4695
4696 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4697 ret = -EFAULT;
4698
4699 kfree(bargs);
4700 out:
4701 mutex_unlock(&fs_info->balance_mutex);
4702 return ret;
4703 }
4704
4705 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4706 {
4707 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4708 struct btrfs_ioctl_quota_ctl_args *sa;
4709 struct btrfs_trans_handle *trans = NULL;
4710 int ret;
4711 int err;
4712
4713 if (!capable(CAP_SYS_ADMIN))
4714 return -EPERM;
4715
4716 ret = mnt_want_write_file(file);
4717 if (ret)
4718 return ret;
4719
4720 sa = memdup_user(arg, sizeof(*sa));
4721 if (IS_ERR(sa)) {
4722 ret = PTR_ERR(sa);
4723 goto drop_write;
4724 }
4725
4726 down_write(&root->fs_info->subvol_sem);
4727 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4728 if (IS_ERR(trans)) {
4729 ret = PTR_ERR(trans);
4730 goto out;
4731 }
4732
4733 switch (sa->cmd) {
4734 case BTRFS_QUOTA_CTL_ENABLE:
4735 ret = btrfs_quota_enable(trans, root->fs_info);
4736 break;
4737 case BTRFS_QUOTA_CTL_DISABLE:
4738 ret = btrfs_quota_disable(trans, root->fs_info);
4739 break;
4740 default:
4741 ret = -EINVAL;
4742 break;
4743 }
4744
4745 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4746 if (err && !ret)
4747 ret = err;
4748 out:
4749 kfree(sa);
4750 up_write(&root->fs_info->subvol_sem);
4751 drop_write:
4752 mnt_drop_write_file(file);
4753 return ret;
4754 }
4755
4756 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4757 {
4758 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4759 struct btrfs_ioctl_qgroup_assign_args *sa;
4760 struct btrfs_trans_handle *trans;
4761 int ret;
4762 int err;
4763
4764 if (!capable(CAP_SYS_ADMIN))
4765 return -EPERM;
4766
4767 ret = mnt_want_write_file(file);
4768 if (ret)
4769 return ret;
4770
4771 sa = memdup_user(arg, sizeof(*sa));
4772 if (IS_ERR(sa)) {
4773 ret = PTR_ERR(sa);
4774 goto drop_write;
4775 }
4776
4777 trans = btrfs_join_transaction(root);
4778 if (IS_ERR(trans)) {
4779 ret = PTR_ERR(trans);
4780 goto out;
4781 }
4782
4783 /* FIXME: check if the IDs really exist */
4784 if (sa->assign) {
4785 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4786 sa->src, sa->dst);
4787 } else {
4788 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4789 sa->src, sa->dst);
4790 }
4791
4792 /* update qgroup status and info */
4793 err = btrfs_run_qgroups(trans, root->fs_info);
4794 if (err < 0)
4795 btrfs_error(root->fs_info, ret,
4796 "failed to update qgroup status and info\n");
4797 err = btrfs_end_transaction(trans, root);
4798 if (err && !ret)
4799 ret = err;
4800
4801 out:
4802 kfree(sa);
4803 drop_write:
4804 mnt_drop_write_file(file);
4805 return ret;
4806 }
4807
4808 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4809 {
4810 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4811 struct btrfs_ioctl_qgroup_create_args *sa;
4812 struct btrfs_trans_handle *trans;
4813 int ret;
4814 int err;
4815
4816 if (!capable(CAP_SYS_ADMIN))
4817 return -EPERM;
4818
4819 ret = mnt_want_write_file(file);
4820 if (ret)
4821 return ret;
4822
4823 sa = memdup_user(arg, sizeof(*sa));
4824 if (IS_ERR(sa)) {
4825 ret = PTR_ERR(sa);
4826 goto drop_write;
4827 }
4828
4829 if (!sa->qgroupid) {
4830 ret = -EINVAL;
4831 goto out;
4832 }
4833
4834 trans = btrfs_join_transaction(root);
4835 if (IS_ERR(trans)) {
4836 ret = PTR_ERR(trans);
4837 goto out;
4838 }
4839
4840 /* FIXME: check if the IDs really exist */
4841 if (sa->create) {
4842 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
4843 } else {
4844 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4845 }
4846
4847 err = btrfs_end_transaction(trans, root);
4848 if (err && !ret)
4849 ret = err;
4850
4851 out:
4852 kfree(sa);
4853 drop_write:
4854 mnt_drop_write_file(file);
4855 return ret;
4856 }
4857
4858 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4859 {
4860 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4861 struct btrfs_ioctl_qgroup_limit_args *sa;
4862 struct btrfs_trans_handle *trans;
4863 int ret;
4864 int err;
4865 u64 qgroupid;
4866
4867 if (!capable(CAP_SYS_ADMIN))
4868 return -EPERM;
4869
4870 ret = mnt_want_write_file(file);
4871 if (ret)
4872 return ret;
4873
4874 sa = memdup_user(arg, sizeof(*sa));
4875 if (IS_ERR(sa)) {
4876 ret = PTR_ERR(sa);
4877 goto drop_write;
4878 }
4879
4880 trans = btrfs_join_transaction(root);
4881 if (IS_ERR(trans)) {
4882 ret = PTR_ERR(trans);
4883 goto out;
4884 }
4885
4886 qgroupid = sa->qgroupid;
4887 if (!qgroupid) {
4888 /* take the current subvol as qgroup */
4889 qgroupid = root->root_key.objectid;
4890 }
4891
4892 /* FIXME: check if the IDs really exist */
4893 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4894
4895 err = btrfs_end_transaction(trans, root);
4896 if (err && !ret)
4897 ret = err;
4898
4899 out:
4900 kfree(sa);
4901 drop_write:
4902 mnt_drop_write_file(file);
4903 return ret;
4904 }
4905
4906 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4907 {
4908 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4909 struct btrfs_ioctl_quota_rescan_args *qsa;
4910 int ret;
4911
4912 if (!capable(CAP_SYS_ADMIN))
4913 return -EPERM;
4914
4915 ret = mnt_want_write_file(file);
4916 if (ret)
4917 return ret;
4918
4919 qsa = memdup_user(arg, sizeof(*qsa));
4920 if (IS_ERR(qsa)) {
4921 ret = PTR_ERR(qsa);
4922 goto drop_write;
4923 }
4924
4925 if (qsa->flags) {
4926 ret = -EINVAL;
4927 goto out;
4928 }
4929
4930 ret = btrfs_qgroup_rescan(root->fs_info);
4931
4932 out:
4933 kfree(qsa);
4934 drop_write:
4935 mnt_drop_write_file(file);
4936 return ret;
4937 }
4938
4939 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4940 {
4941 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4942 struct btrfs_ioctl_quota_rescan_args *qsa;
4943 int ret = 0;
4944
4945 if (!capable(CAP_SYS_ADMIN))
4946 return -EPERM;
4947
4948 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4949 if (!qsa)
4950 return -ENOMEM;
4951
4952 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4953 qsa->flags = 1;
4954 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4955 }
4956
4957 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4958 ret = -EFAULT;
4959
4960 kfree(qsa);
4961 return ret;
4962 }
4963
4964 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4965 {
4966 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4967
4968 if (!capable(CAP_SYS_ADMIN))
4969 return -EPERM;
4970
4971 return btrfs_qgroup_wait_for_completion(root->fs_info);
4972 }
4973
4974 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4975 struct btrfs_ioctl_received_subvol_args *sa)
4976 {
4977 struct inode *inode = file_inode(file);
4978 struct btrfs_root *root = BTRFS_I(inode)->root;
4979 struct btrfs_root_item *root_item = &root->root_item;
4980 struct btrfs_trans_handle *trans;
4981 struct timespec ct = CURRENT_TIME;
4982 int ret = 0;
4983 int received_uuid_changed;
4984
4985 if (!inode_owner_or_capable(inode))
4986 return -EPERM;
4987
4988 ret = mnt_want_write_file(file);
4989 if (ret < 0)
4990 return ret;
4991
4992 down_write(&root->fs_info->subvol_sem);
4993
4994 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4995 ret = -EINVAL;
4996 goto out;
4997 }
4998
4999 if (btrfs_root_readonly(root)) {
5000 ret = -EROFS;
5001 goto out;
5002 }
5003
5004 /*
5005 * 1 - root item
5006 * 2 - uuid items (received uuid + subvol uuid)
5007 */
5008 trans = btrfs_start_transaction(root, 3);
5009 if (IS_ERR(trans)) {
5010 ret = PTR_ERR(trans);
5011 trans = NULL;
5012 goto out;
5013 }
5014
5015 sa->rtransid = trans->transid;
5016 sa->rtime.sec = ct.tv_sec;
5017 sa->rtime.nsec = ct.tv_nsec;
5018
5019 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5020 BTRFS_UUID_SIZE);
5021 if (received_uuid_changed &&
5022 !btrfs_is_empty_uuid(root_item->received_uuid))
5023 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
5024 root_item->received_uuid,
5025 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5026 root->root_key.objectid);
5027 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5028 btrfs_set_root_stransid(root_item, sa->stransid);
5029 btrfs_set_root_rtransid(root_item, sa->rtransid);
5030 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5031 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5032 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5033 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5034
5035 ret = btrfs_update_root(trans, root->fs_info->tree_root,
5036 &root->root_key, &root->root_item);
5037 if (ret < 0) {
5038 btrfs_end_transaction(trans, root);
5039 goto out;
5040 }
5041 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5042 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
5043 sa->uuid,
5044 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5045 root->root_key.objectid);
5046 if (ret < 0 && ret != -EEXIST) {
5047 btrfs_abort_transaction(trans, root, ret);
5048 goto out;
5049 }
5050 }
5051 ret = btrfs_commit_transaction(trans, root);
5052 if (ret < 0) {
5053 btrfs_abort_transaction(trans, root, ret);
5054 goto out;
5055 }
5056
5057 out:
5058 up_write(&root->fs_info->subvol_sem);
5059 mnt_drop_write_file(file);
5060 return ret;
5061 }
5062
5063 #ifdef CONFIG_64BIT
5064 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5065 void __user *arg)
5066 {
5067 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5068 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5069 int ret = 0;
5070
5071 args32 = memdup_user(arg, sizeof(*args32));
5072 if (IS_ERR(args32)) {
5073 ret = PTR_ERR(args32);
5074 args32 = NULL;
5075 goto out;
5076 }
5077
5078 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
5079 if (!args64) {
5080 ret = -ENOMEM;
5081 goto out;
5082 }
5083
5084 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5085 args64->stransid = args32->stransid;
5086 args64->rtransid = args32->rtransid;
5087 args64->stime.sec = args32->stime.sec;
5088 args64->stime.nsec = args32->stime.nsec;
5089 args64->rtime.sec = args32->rtime.sec;
5090 args64->rtime.nsec = args32->rtime.nsec;
5091 args64->flags = args32->flags;
5092
5093 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5094 if (ret)
5095 goto out;
5096
5097 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5098 args32->stransid = args64->stransid;
5099 args32->rtransid = args64->rtransid;
5100 args32->stime.sec = args64->stime.sec;
5101 args32->stime.nsec = args64->stime.nsec;
5102 args32->rtime.sec = args64->rtime.sec;
5103 args32->rtime.nsec = args64->rtime.nsec;
5104 args32->flags = args64->flags;
5105
5106 ret = copy_to_user(arg, args32, sizeof(*args32));
5107 if (ret)
5108 ret = -EFAULT;
5109
5110 out:
5111 kfree(args32);
5112 kfree(args64);
5113 return ret;
5114 }
5115 #endif
5116
5117 static long btrfs_ioctl_set_received_subvol(struct file *file,
5118 void __user *arg)
5119 {
5120 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5121 int ret = 0;
5122
5123 sa = memdup_user(arg, sizeof(*sa));
5124 if (IS_ERR(sa)) {
5125 ret = PTR_ERR(sa);
5126 sa = NULL;
5127 goto out;
5128 }
5129
5130 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5131
5132 if (ret)
5133 goto out;
5134
5135 ret = copy_to_user(arg, sa, sizeof(*sa));
5136 if (ret)
5137 ret = -EFAULT;
5138
5139 out:
5140 kfree(sa);
5141 return ret;
5142 }
5143
5144 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5145 {
5146 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5147 size_t len;
5148 int ret;
5149 char label[BTRFS_LABEL_SIZE];
5150
5151 spin_lock(&root->fs_info->super_lock);
5152 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5153 spin_unlock(&root->fs_info->super_lock);
5154
5155 len = strnlen(label, BTRFS_LABEL_SIZE);
5156
5157 if (len == BTRFS_LABEL_SIZE) {
5158 btrfs_warn(root->fs_info,
5159 "label is too long, return the first %zu bytes", --len);
5160 }
5161
5162 ret = copy_to_user(arg, label, len);
5163
5164 return ret ? -EFAULT : 0;
5165 }
5166
5167 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5168 {
5169 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5170 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5171 struct btrfs_trans_handle *trans;
5172 char label[BTRFS_LABEL_SIZE];
5173 int ret;
5174
5175 if (!capable(CAP_SYS_ADMIN))
5176 return -EPERM;
5177
5178 if (copy_from_user(label, arg, sizeof(label)))
5179 return -EFAULT;
5180
5181 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5182 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5183 BTRFS_LABEL_SIZE - 1);
5184 return -EINVAL;
5185 }
5186
5187 ret = mnt_want_write_file(file);
5188 if (ret)
5189 return ret;
5190
5191 trans = btrfs_start_transaction(root, 0);
5192 if (IS_ERR(trans)) {
5193 ret = PTR_ERR(trans);
5194 goto out_unlock;
5195 }
5196
5197 spin_lock(&root->fs_info->super_lock);
5198 strcpy(super_block->label, label);
5199 spin_unlock(&root->fs_info->super_lock);
5200 ret = btrfs_commit_transaction(trans, root);
5201
5202 out_unlock:
5203 mnt_drop_write_file(file);
5204 return ret;
5205 }
5206
5207 #define INIT_FEATURE_FLAGS(suffix) \
5208 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5209 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5210 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5211
5212 static int btrfs_ioctl_get_supported_features(struct file *file,
5213 void __user *arg)
5214 {
5215 static struct btrfs_ioctl_feature_flags features[3] = {
5216 INIT_FEATURE_FLAGS(SUPP),
5217 INIT_FEATURE_FLAGS(SAFE_SET),
5218 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5219 };
5220
5221 if (copy_to_user(arg, &features, sizeof(features)))
5222 return -EFAULT;
5223
5224 return 0;
5225 }
5226
5227 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5228 {
5229 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5230 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5231 struct btrfs_ioctl_feature_flags features;
5232
5233 features.compat_flags = btrfs_super_compat_flags(super_block);
5234 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5235 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5236
5237 if (copy_to_user(arg, &features, sizeof(features)))
5238 return -EFAULT;
5239
5240 return 0;
5241 }
5242
5243 static int check_feature_bits(struct btrfs_root *root,
5244 enum btrfs_feature_set set,
5245 u64 change_mask, u64 flags, u64 supported_flags,
5246 u64 safe_set, u64 safe_clear)
5247 {
5248 const char *type = btrfs_feature_set_names[set];
5249 char *names;
5250 u64 disallowed, unsupported;
5251 u64 set_mask = flags & change_mask;
5252 u64 clear_mask = ~flags & change_mask;
5253
5254 unsupported = set_mask & ~supported_flags;
5255 if (unsupported) {
5256 names = btrfs_printable_features(set, unsupported);
5257 if (names) {
5258 btrfs_warn(root->fs_info,
5259 "this kernel does not support the %s feature bit%s",
5260 names, strchr(names, ',') ? "s" : "");
5261 kfree(names);
5262 } else
5263 btrfs_warn(root->fs_info,
5264 "this kernel does not support %s bits 0x%llx",
5265 type, unsupported);
5266 return -EOPNOTSUPP;
5267 }
5268
5269 disallowed = set_mask & ~safe_set;
5270 if (disallowed) {
5271 names = btrfs_printable_features(set, disallowed);
5272 if (names) {
5273 btrfs_warn(root->fs_info,
5274 "can't set the %s feature bit%s while mounted",
5275 names, strchr(names, ',') ? "s" : "");
5276 kfree(names);
5277 } else
5278 btrfs_warn(root->fs_info,
5279 "can't set %s bits 0x%llx while mounted",
5280 type, disallowed);
5281 return -EPERM;
5282 }
5283
5284 disallowed = clear_mask & ~safe_clear;
5285 if (disallowed) {
5286 names = btrfs_printable_features(set, disallowed);
5287 if (names) {
5288 btrfs_warn(root->fs_info,
5289 "can't clear the %s feature bit%s while mounted",
5290 names, strchr(names, ',') ? "s" : "");
5291 kfree(names);
5292 } else
5293 btrfs_warn(root->fs_info,
5294 "can't clear %s bits 0x%llx while mounted",
5295 type, disallowed);
5296 return -EPERM;
5297 }
5298
5299 return 0;
5300 }
5301
5302 #define check_feature(root, change_mask, flags, mask_base) \
5303 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5304 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5305 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5306 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5307
5308 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5309 {
5310 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5311 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5312 struct btrfs_ioctl_feature_flags flags[2];
5313 struct btrfs_trans_handle *trans;
5314 u64 newflags;
5315 int ret;
5316
5317 if (!capable(CAP_SYS_ADMIN))
5318 return -EPERM;
5319
5320 if (copy_from_user(flags, arg, sizeof(flags)))
5321 return -EFAULT;
5322
5323 /* Nothing to do */
5324 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5325 !flags[0].incompat_flags)
5326 return 0;
5327
5328 ret = check_feature(root, flags[0].compat_flags,
5329 flags[1].compat_flags, COMPAT);
5330 if (ret)
5331 return ret;
5332
5333 ret = check_feature(root, flags[0].compat_ro_flags,
5334 flags[1].compat_ro_flags, COMPAT_RO);
5335 if (ret)
5336 return ret;
5337
5338 ret = check_feature(root, flags[0].incompat_flags,
5339 flags[1].incompat_flags, INCOMPAT);
5340 if (ret)
5341 return ret;
5342
5343 trans = btrfs_start_transaction(root, 0);
5344 if (IS_ERR(trans))
5345 return PTR_ERR(trans);
5346
5347 spin_lock(&root->fs_info->super_lock);
5348 newflags = btrfs_super_compat_flags(super_block);
5349 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5350 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5351 btrfs_set_super_compat_flags(super_block, newflags);
5352
5353 newflags = btrfs_super_compat_ro_flags(super_block);
5354 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5355 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5356 btrfs_set_super_compat_ro_flags(super_block, newflags);
5357
5358 newflags = btrfs_super_incompat_flags(super_block);
5359 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5360 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5361 btrfs_set_super_incompat_flags(super_block, newflags);
5362 spin_unlock(&root->fs_info->super_lock);
5363
5364 return btrfs_commit_transaction(trans, root);
5365 }
5366
5367 long btrfs_ioctl(struct file *file, unsigned int
5368 cmd, unsigned long arg)
5369 {
5370 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5371 void __user *argp = (void __user *)arg;
5372
5373 switch (cmd) {
5374 case FS_IOC_GETFLAGS:
5375 return btrfs_ioctl_getflags(file, argp);
5376 case FS_IOC_SETFLAGS:
5377 return btrfs_ioctl_setflags(file, argp);
5378 case FS_IOC_GETVERSION:
5379 return btrfs_ioctl_getversion(file, argp);
5380 case FITRIM:
5381 return btrfs_ioctl_fitrim(file, argp);
5382 case BTRFS_IOC_SNAP_CREATE:
5383 return btrfs_ioctl_snap_create(file, argp, 0);
5384 case BTRFS_IOC_SNAP_CREATE_V2:
5385 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5386 case BTRFS_IOC_SUBVOL_CREATE:
5387 return btrfs_ioctl_snap_create(file, argp, 1);
5388 case BTRFS_IOC_SUBVOL_CREATE_V2:
5389 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5390 case BTRFS_IOC_SNAP_DESTROY:
5391 return btrfs_ioctl_snap_destroy(file, argp);
5392 case BTRFS_IOC_SUBVOL_GETFLAGS:
5393 return btrfs_ioctl_subvol_getflags(file, argp);
5394 case BTRFS_IOC_SUBVOL_SETFLAGS:
5395 return btrfs_ioctl_subvol_setflags(file, argp);
5396 case BTRFS_IOC_DEFAULT_SUBVOL:
5397 return btrfs_ioctl_default_subvol(file, argp);
5398 case BTRFS_IOC_DEFRAG:
5399 return btrfs_ioctl_defrag(file, NULL);
5400 case BTRFS_IOC_DEFRAG_RANGE:
5401 return btrfs_ioctl_defrag(file, argp);
5402 case BTRFS_IOC_RESIZE:
5403 return btrfs_ioctl_resize(file, argp);
5404 case BTRFS_IOC_ADD_DEV:
5405 return btrfs_ioctl_add_dev(root, argp);
5406 case BTRFS_IOC_RM_DEV:
5407 return btrfs_ioctl_rm_dev(file, argp);
5408 case BTRFS_IOC_FS_INFO:
5409 return btrfs_ioctl_fs_info(root, argp);
5410 case BTRFS_IOC_DEV_INFO:
5411 return btrfs_ioctl_dev_info(root, argp);
5412 case BTRFS_IOC_BALANCE:
5413 return btrfs_ioctl_balance(file, NULL);
5414 case BTRFS_IOC_CLONE:
5415 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5416 case BTRFS_IOC_CLONE_RANGE:
5417 return btrfs_ioctl_clone_range(file, argp);
5418 case BTRFS_IOC_TRANS_START:
5419 return btrfs_ioctl_trans_start(file);
5420 case BTRFS_IOC_TRANS_END:
5421 return btrfs_ioctl_trans_end(file);
5422 case BTRFS_IOC_TREE_SEARCH:
5423 return btrfs_ioctl_tree_search(file, argp);
5424 case BTRFS_IOC_TREE_SEARCH_V2:
5425 return btrfs_ioctl_tree_search_v2(file, argp);
5426 case BTRFS_IOC_INO_LOOKUP:
5427 return btrfs_ioctl_ino_lookup(file, argp);
5428 case BTRFS_IOC_INO_PATHS:
5429 return btrfs_ioctl_ino_to_path(root, argp);
5430 case BTRFS_IOC_LOGICAL_INO:
5431 return btrfs_ioctl_logical_to_ino(root, argp);
5432 case BTRFS_IOC_SPACE_INFO:
5433 return btrfs_ioctl_space_info(root, argp);
5434 case BTRFS_IOC_SYNC: {
5435 int ret;
5436
5437 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5438 if (ret)
5439 return ret;
5440 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
5441 /*
5442 * The transaction thread may want to do more work,
5443 * namely it pokes the cleaner ktread that will start
5444 * processing uncleaned subvols.
5445 */
5446 wake_up_process(root->fs_info->transaction_kthread);
5447 return ret;
5448 }
5449 case BTRFS_IOC_START_SYNC:
5450 return btrfs_ioctl_start_sync(root, argp);
5451 case BTRFS_IOC_WAIT_SYNC:
5452 return btrfs_ioctl_wait_sync(root, argp);
5453 case BTRFS_IOC_SCRUB:
5454 return btrfs_ioctl_scrub(file, argp);
5455 case BTRFS_IOC_SCRUB_CANCEL:
5456 return btrfs_ioctl_scrub_cancel(root, argp);
5457 case BTRFS_IOC_SCRUB_PROGRESS:
5458 return btrfs_ioctl_scrub_progress(root, argp);
5459 case BTRFS_IOC_BALANCE_V2:
5460 return btrfs_ioctl_balance(file, argp);
5461 case BTRFS_IOC_BALANCE_CTL:
5462 return btrfs_ioctl_balance_ctl(root, arg);
5463 case BTRFS_IOC_BALANCE_PROGRESS:
5464 return btrfs_ioctl_balance_progress(root, argp);
5465 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5466 return btrfs_ioctl_set_received_subvol(file, argp);
5467 #ifdef CONFIG_64BIT
5468 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5469 return btrfs_ioctl_set_received_subvol_32(file, argp);
5470 #endif
5471 case BTRFS_IOC_SEND:
5472 return btrfs_ioctl_send(file, argp);
5473 case BTRFS_IOC_GET_DEV_STATS:
5474 return btrfs_ioctl_get_dev_stats(root, argp);
5475 case BTRFS_IOC_QUOTA_CTL:
5476 return btrfs_ioctl_quota_ctl(file, argp);
5477 case BTRFS_IOC_QGROUP_ASSIGN:
5478 return btrfs_ioctl_qgroup_assign(file, argp);
5479 case BTRFS_IOC_QGROUP_CREATE:
5480 return btrfs_ioctl_qgroup_create(file, argp);
5481 case BTRFS_IOC_QGROUP_LIMIT:
5482 return btrfs_ioctl_qgroup_limit(file, argp);
5483 case BTRFS_IOC_QUOTA_RESCAN:
5484 return btrfs_ioctl_quota_rescan(file, argp);
5485 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5486 return btrfs_ioctl_quota_rescan_status(file, argp);
5487 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5488 return btrfs_ioctl_quota_rescan_wait(file, argp);
5489 case BTRFS_IOC_DEV_REPLACE:
5490 return btrfs_ioctl_dev_replace(root, argp);
5491 case BTRFS_IOC_GET_FSLABEL:
5492 return btrfs_ioctl_get_fslabel(file, argp);
5493 case BTRFS_IOC_SET_FSLABEL:
5494 return btrfs_ioctl_set_fslabel(file, argp);
5495 case BTRFS_IOC_FILE_EXTENT_SAME:
5496 return btrfs_ioctl_file_extent_same(file, argp);
5497 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5498 return btrfs_ioctl_get_supported_features(file, argp);
5499 case BTRFS_IOC_GET_FEATURES:
5500 return btrfs_ioctl_get_features(file, argp);
5501 case BTRFS_IOC_SET_FEATURES:
5502 return btrfs_ioctl_set_features(file, argp);
5503 }
5504
5505 return -ENOTTY;
5506 }
5507
5508 #ifdef CONFIG_COMPAT
5509 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5510 {
5511 switch (cmd) {
5512 case FS_IOC32_GETFLAGS:
5513 cmd = FS_IOC_GETFLAGS;
5514 break;
5515 case FS_IOC32_SETFLAGS:
5516 cmd = FS_IOC_SETFLAGS;
5517 break;
5518 case FS_IOC32_GETVERSION:
5519 cmd = FS_IOC_GETVERSION;
5520 break;
5521 default:
5522 return -ENOIOCTLCMD;
5523 }
5524
5525 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5526 }
5527 #endif
Linux with added FPC-III support