]> git.ipfire.org Git - thirdparty/linux.git/blob - fs/btrfs/ioctl.c
projects / thirdparty / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
btrfs: export btrfs_set_prop
[thirdparty/linux.git] / fs / btrfs / ioctl.c
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Copyright (C) 2007 Oracle. All rights reserved.
4 */
5
6 #include <linux/kernel.h>
7 #include <linux/bio.h>
8 #include <linux/file.h>
9 #include <linux/fs.h>
10 #include <linux/fsnotify.h>
11 #include <linux/pagemap.h>
12 #include <linux/highmem.h>
13 #include <linux/time.h>
14 #include <linux/string.h>
15 #include <linux/backing-dev.h>
16 #include <linux/mount.h>
17 #include <linux/namei.h>
18 #include <linux/writeback.h>
19 #include <linux/compat.h>
20 #include <linux/security.h>
21 #include <linux/xattr.h>
22 #include <linux/mm.h>
23 #include <linux/slab.h>
24 #include <linux/blkdev.h>
25 #include <linux/uuid.h>
26 #include <linux/btrfs.h>
27 #include <linux/uaccess.h>
28 #include <linux/iversion.h>
29 #include "ctree.h"
30 #include "disk-io.h"
31 #include "transaction.h"
32 #include "btrfs_inode.h"
33 #include "print-tree.h"
34 #include "volumes.h"
35 #include "locking.h"
36 #include "inode-map.h"
37 #include "backref.h"
38 #include "rcu-string.h"
39 #include "send.h"
40 #include "dev-replace.h"
41 #include "props.h"
42 #include "sysfs.h"
43 #include "qgroup.h"
44 #include "tree-log.h"
45 #include "compression.h"
46
47 #ifdef CONFIG_64BIT
48 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
49 * structures are incorrect, as the timespec structure from userspace
50 * is 4 bytes too small. We define these alternatives here to teach
51 * the kernel about the 32-bit struct packing.
52 */
53 struct btrfs_ioctl_timespec_32 {
54 __u64 sec;
55 __u32 nsec;
56 } __attribute__ ((__packed__));
57
58 struct btrfs_ioctl_received_subvol_args_32 {
59 char uuid[BTRFS_UUID_SIZE]; /* in */
60 __u64 stransid; /* in */
61 __u64 rtransid; /* out */
62 struct btrfs_ioctl_timespec_32 stime; /* in */
63 struct btrfs_ioctl_timespec_32 rtime; /* out */
64 __u64 flags; /* in */
65 __u64 reserved[16]; /* in */
66 } __attribute__ ((__packed__));
67
68 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
69 struct btrfs_ioctl_received_subvol_args_32)
70 #endif
71
72 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
73 struct btrfs_ioctl_send_args_32 {
74 __s64 send_fd; /* in */
75 __u64 clone_sources_count; /* in */
76 compat_uptr_t clone_sources; /* in */
77 __u64 parent_root; /* in */
78 __u64 flags; /* in */
79 __u64 reserved[4]; /* in */
80 } __attribute__ ((__packed__));
81
82 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
83 struct btrfs_ioctl_send_args_32)
84 #endif
85
86 static int btrfs_clone(struct inode *src, struct inode *inode,
87 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
88 int no_time_update);
89
90 /* Mask out flags that are inappropriate for the given type of inode. */
91 static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
92 unsigned int flags)
93 {
94 if (S_ISDIR(inode->i_mode))
95 return flags;
96 else if (S_ISREG(inode->i_mode))
97 return flags & ~FS_DIRSYNC_FL;
98 else
99 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
100 }
101
102 /*
103 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
104 * ioctl.
105 */
106 static unsigned int btrfs_inode_flags_to_fsflags(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_NOCOMPRESS)
126 iflags |= FS_NOCOMP_FL;
127 else if (flags & BTRFS_INODE_COMPRESS)
128 iflags |= FS_COMPR_FL;
129
130 return iflags;
131 }
132
133 /*
134 * Update inode->i_flags based on the btrfs internal flags.
135 */
136 void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
137 {
138 struct btrfs_inode *binode = BTRFS_I(inode);
139 unsigned int new_fl = 0;
140
141 if (binode->flags & BTRFS_INODE_SYNC)
142 new_fl |= S_SYNC;
143 if (binode->flags & BTRFS_INODE_IMMUTABLE)
144 new_fl |= S_IMMUTABLE;
145 if (binode->flags & BTRFS_INODE_APPEND)
146 new_fl |= S_APPEND;
147 if (binode->flags & BTRFS_INODE_NOATIME)
148 new_fl |= S_NOATIME;
149 if (binode->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 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
158 {
159 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
160 unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags);
161
162 if (copy_to_user(arg, &flags, sizeof(flags)))
163 return -EFAULT;
164 return 0;
165 }
166
167 /* Check if @flags are a supported and valid set of FS_*_FL flags */
168 static int check_fsflags(unsigned int flags)
169 {
170 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
171 FS_NOATIME_FL | FS_NODUMP_FL | \
172 FS_SYNC_FL | FS_DIRSYNC_FL | \
173 FS_NOCOMP_FL | FS_COMPR_FL |
174 FS_NOCOW_FL))
175 return -EOPNOTSUPP;
176
177 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
178 return -EINVAL;
179
180 return 0;
181 }
182
183 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
184 {
185 struct inode *inode = file_inode(file);
186 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
187 struct btrfs_inode *binode = BTRFS_I(inode);
188 struct btrfs_root *root = binode->root;
189 struct btrfs_trans_handle *trans;
190 unsigned int fsflags, old_fsflags;
191 int ret;
192 u64 old_flags;
193 unsigned int old_i_flags;
194 umode_t mode;
195
196 if (!inode_owner_or_capable(inode))
197 return -EPERM;
198
199 if (btrfs_root_readonly(root))
200 return -EROFS;
201
202 if (copy_from_user(&fsflags, arg, sizeof(fsflags)))
203 return -EFAULT;
204
205 ret = check_fsflags(fsflags);
206 if (ret)
207 return ret;
208
209 ret = mnt_want_write_file(file);
210 if (ret)
211 return ret;
212
213 inode_lock(inode);
214
215 old_flags = binode->flags;
216 old_i_flags = inode->i_flags;
217 mode = inode->i_mode;
218
219 fsflags = btrfs_mask_fsflags_for_type(inode, fsflags);
220 old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags);
221 if ((fsflags ^ old_fsflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
222 if (!capable(CAP_LINUX_IMMUTABLE)) {
223 ret = -EPERM;
224 goto out_unlock;
225 }
226 }
227
228 if (fsflags & FS_SYNC_FL)
229 binode->flags |= BTRFS_INODE_SYNC;
230 else
231 binode->flags &= ~BTRFS_INODE_SYNC;
232 if (fsflags & FS_IMMUTABLE_FL)
233 binode->flags |= BTRFS_INODE_IMMUTABLE;
234 else
235 binode->flags &= ~BTRFS_INODE_IMMUTABLE;
236 if (fsflags & FS_APPEND_FL)
237 binode->flags |= BTRFS_INODE_APPEND;
238 else
239 binode->flags &= ~BTRFS_INODE_APPEND;
240 if (fsflags & FS_NODUMP_FL)
241 binode->flags |= BTRFS_INODE_NODUMP;
242 else
243 binode->flags &= ~BTRFS_INODE_NODUMP;
244 if (fsflags & FS_NOATIME_FL)
245 binode->flags |= BTRFS_INODE_NOATIME;
246 else
247 binode->flags &= ~BTRFS_INODE_NOATIME;
248 if (fsflags & FS_DIRSYNC_FL)
249 binode->flags |= BTRFS_INODE_DIRSYNC;
250 else
251 binode->flags &= ~BTRFS_INODE_DIRSYNC;
252 if (fsflags & FS_NOCOW_FL) {
253 if (S_ISREG(mode)) {
254 /*
255 * It's safe to turn csums off here, no extents exist.
256 * Otherwise we want the flag to reflect the real COW
257 * status of the file and will not set it.
258 */
259 if (inode->i_size == 0)
260 binode->flags |= BTRFS_INODE_NODATACOW
261 | BTRFS_INODE_NODATASUM;
262 } else {
263 binode->flags |= BTRFS_INODE_NODATACOW;
264 }
265 } else {
266 /*
267 * Revert back under same assumptions as above
268 */
269 if (S_ISREG(mode)) {
270 if (inode->i_size == 0)
271 binode->flags &= ~(BTRFS_INODE_NODATACOW
272 | BTRFS_INODE_NODATASUM);
273 } else {
274 binode->flags &= ~BTRFS_INODE_NODATACOW;
275 }
276 }
277
278 /*
279 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
280 * flag may be changed automatically if compression code won't make
281 * things smaller.
282 */
283 if (fsflags & FS_NOCOMP_FL) {
284 binode->flags &= ~BTRFS_INODE_COMPRESS;
285 binode->flags |= BTRFS_INODE_NOCOMPRESS;
286
287 /* set no-compression no need to validate prop here */
288 ret = btrfs_set_prop_trans(inode, "btrfs.compression", NULL,
289 0, 0);
290 if (ret && ret != -ENODATA)
291 goto out_drop;
292 } else if (fsflags & FS_COMPR_FL) {
293 const char *comp;
294
295 if (IS_SWAPFILE(inode)) {
296 ret = -ETXTBSY;
297 goto out_unlock;
298 }
299
300 binode->flags |= BTRFS_INODE_COMPRESS;
301 binode->flags &= ~BTRFS_INODE_NOCOMPRESS;
302
303 /* compress_type is already validated during mount options */
304 comp = btrfs_compress_type2str(fs_info->compress_type);
305 if (!comp || comp[0] == 0)
306 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
307
308 ret = btrfs_set_prop_trans(inode, "btrfs.compression", comp,
309 strlen(comp), 0);
310 if (ret)
311 goto out_drop;
312
313 } else {
314 /* reset prop, no need of validate prop here */
315 ret = btrfs_set_prop_trans(inode, "btrfs.compression", NULL,
316 0, 0);
317 if (ret && ret != -ENODATA)
318 goto out_drop;
319 binode->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
320 }
321
322 trans = btrfs_start_transaction(root, 1);
323 if (IS_ERR(trans)) {
324 ret = PTR_ERR(trans);
325 goto out_drop;
326 }
327
328 btrfs_sync_inode_flags_to_i_flags(inode);
329 inode_inc_iversion(inode);
330 inode->i_ctime = current_time(inode);
331 ret = btrfs_update_inode(trans, root, inode);
332
333 btrfs_end_transaction(trans);
334 out_drop:
335 if (ret) {
336 binode->flags = old_flags;
337 inode->i_flags = old_i_flags;
338 }
339
340 out_unlock:
341 inode_unlock(inode);
342 mnt_drop_write_file(file);
343 return ret;
344 }
345
346 /*
347 * Translate btrfs internal inode flags to xflags as expected by the
348 * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are
349 * silently dropped.
350 */
351 static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags)
352 {
353 unsigned int xflags = 0;
354
355 if (flags & BTRFS_INODE_APPEND)
356 xflags |= FS_XFLAG_APPEND;
357 if (flags & BTRFS_INODE_IMMUTABLE)
358 xflags |= FS_XFLAG_IMMUTABLE;
359 if (flags & BTRFS_INODE_NOATIME)
360 xflags |= FS_XFLAG_NOATIME;
361 if (flags & BTRFS_INODE_NODUMP)
362 xflags |= FS_XFLAG_NODUMP;
363 if (flags & BTRFS_INODE_SYNC)
364 xflags |= FS_XFLAG_SYNC;
365
366 return xflags;
367 }
368
369 /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */
370 static int check_xflags(unsigned int flags)
371 {
372 if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME |
373 FS_XFLAG_NODUMP | FS_XFLAG_SYNC))
374 return -EOPNOTSUPP;
375 return 0;
376 }
377
378 /*
379 * Set the xflags from the internal inode flags. The remaining items of fsxattr
380 * are zeroed.
381 */
382 static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg)
383 {
384 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
385 struct fsxattr fa;
386
387 memset(&fa, 0, sizeof(fa));
388 fa.fsx_xflags = btrfs_inode_flags_to_xflags(binode->flags);
389
390 if (copy_to_user(arg, &fa, sizeof(fa)))
391 return -EFAULT;
392
393 return 0;
394 }
395
396 static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg)
397 {
398 struct inode *inode = file_inode(file);
399 struct btrfs_inode *binode = BTRFS_I(inode);
400 struct btrfs_root *root = binode->root;
401 struct btrfs_trans_handle *trans;
402 struct fsxattr fa;
403 unsigned old_flags;
404 unsigned old_i_flags;
405 int ret = 0;
406
407 if (!inode_owner_or_capable(inode))
408 return -EPERM;
409
410 if (btrfs_root_readonly(root))
411 return -EROFS;
412
413 memset(&fa, 0, sizeof(fa));
414 if (copy_from_user(&fa, arg, sizeof(fa)))
415 return -EFAULT;
416
417 ret = check_xflags(fa.fsx_xflags);
418 if (ret)
419 return ret;
420
421 if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0)
422 return -EOPNOTSUPP;
423
424 ret = mnt_want_write_file(file);
425 if (ret)
426 return ret;
427
428 inode_lock(inode);
429
430 old_flags = binode->flags;
431 old_i_flags = inode->i_flags;
432
433 /* We need the capabilities to change append-only or immutable inode */
434 if (((old_flags & (BTRFS_INODE_APPEND | BTRFS_INODE_IMMUTABLE)) ||
435 (fa.fsx_xflags & (FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE))) &&
436 !capable(CAP_LINUX_IMMUTABLE)) {
437 ret = -EPERM;
438 goto out_unlock;
439 }
440
441 if (fa.fsx_xflags & FS_XFLAG_SYNC)
442 binode->flags |= BTRFS_INODE_SYNC;
443 else
444 binode->flags &= ~BTRFS_INODE_SYNC;
445 if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE)
446 binode->flags |= BTRFS_INODE_IMMUTABLE;
447 else
448 binode->flags &= ~BTRFS_INODE_IMMUTABLE;
449 if (fa.fsx_xflags & FS_XFLAG_APPEND)
450 binode->flags |= BTRFS_INODE_APPEND;
451 else
452 binode->flags &= ~BTRFS_INODE_APPEND;
453 if (fa.fsx_xflags & FS_XFLAG_NODUMP)
454 binode->flags |= BTRFS_INODE_NODUMP;
455 else
456 binode->flags &= ~BTRFS_INODE_NODUMP;
457 if (fa.fsx_xflags & FS_XFLAG_NOATIME)
458 binode->flags |= BTRFS_INODE_NOATIME;
459 else
460 binode->flags &= ~BTRFS_INODE_NOATIME;
461
462 /* 1 item for the inode */
463 trans = btrfs_start_transaction(root, 1);
464 if (IS_ERR(trans)) {
465 ret = PTR_ERR(trans);
466 goto out_unlock;
467 }
468
469 btrfs_sync_inode_flags_to_i_flags(inode);
470 inode_inc_iversion(inode);
471 inode->i_ctime = current_time(inode);
472 ret = btrfs_update_inode(trans, root, inode);
473
474 btrfs_end_transaction(trans);
475
476 out_unlock:
477 if (ret) {
478 binode->flags = old_flags;
479 inode->i_flags = old_i_flags;
480 }
481
482 inode_unlock(inode);
483 mnt_drop_write_file(file);
484
485 return ret;
486 }
487
488 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
489 {
490 struct inode *inode = file_inode(file);
491
492 return put_user(inode->i_generation, arg);
493 }
494
495 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
496 {
497 struct inode *inode = file_inode(file);
498 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
499 struct btrfs_device *device;
500 struct request_queue *q;
501 struct fstrim_range range;
502 u64 minlen = ULLONG_MAX;
503 u64 num_devices = 0;
504 int ret;
505
506 if (!capable(CAP_SYS_ADMIN))
507 return -EPERM;
508
509 /*
510 * If the fs is mounted with nologreplay, which requires it to be
511 * mounted in RO mode as well, we can not allow discard on free space
512 * inside block groups, because log trees refer to extents that are not
513 * pinned in a block group's free space cache (pinning the extents is
514 * precisely the first phase of replaying a log tree).
515 */
516 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
517 return -EROFS;
518
519 rcu_read_lock();
520 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
521 dev_list) {
522 if (!device->bdev)
523 continue;
524 q = bdev_get_queue(device->bdev);
525 if (blk_queue_discard(q)) {
526 num_devices++;
527 minlen = min_t(u64, q->limits.discard_granularity,
528 minlen);
529 }
530 }
531 rcu_read_unlock();
532
533 if (!num_devices)
534 return -EOPNOTSUPP;
535 if (copy_from_user(&range, arg, sizeof(range)))
536 return -EFAULT;
537
538 /*
539 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
540 * block group is in the logical address space, which can be any
541 * sectorsize aligned bytenr in the range [0, U64_MAX].
542 */
543 if (range.len < fs_info->sb->s_blocksize)
544 return -EINVAL;
545
546 range.minlen = max(range.minlen, minlen);
547 ret = btrfs_trim_fs(fs_info, &range);
548 if (ret < 0)
549 return ret;
550
551 if (copy_to_user(arg, &range, sizeof(range)))
552 return -EFAULT;
553
554 return 0;
555 }
556
557 int btrfs_is_empty_uuid(u8 *uuid)
558 {
559 int i;
560
561 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
562 if (uuid[i])
563 return 0;
564 }
565 return 1;
566 }
567
568 static noinline int create_subvol(struct inode *dir,
569 struct dentry *dentry,
570 const char *name, int namelen,
571 u64 *async_transid,
572 struct btrfs_qgroup_inherit *inherit)
573 {
574 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
575 struct btrfs_trans_handle *trans;
576 struct btrfs_key key;
577 struct btrfs_root_item *root_item;
578 struct btrfs_inode_item *inode_item;
579 struct extent_buffer *leaf;
580 struct btrfs_root *root = BTRFS_I(dir)->root;
581 struct btrfs_root *new_root;
582 struct btrfs_block_rsv block_rsv;
583 struct timespec64 cur_time = current_time(dir);
584 struct inode *inode;
585 int ret;
586 int err;
587 u64 objectid;
588 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
589 u64 index = 0;
590 uuid_le new_uuid;
591
592 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
593 if (!root_item)
594 return -ENOMEM;
595
596 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
597 if (ret)
598 goto fail_free;
599
600 /*
601 * Don't create subvolume whose level is not zero. Or qgroup will be
602 * screwed up since it assumes subvolume qgroup's level to be 0.
603 */
604 if (btrfs_qgroup_level(objectid)) {
605 ret = -ENOSPC;
606 goto fail_free;
607 }
608
609 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
610 /*
611 * The same as the snapshot creation, please see the comment
612 * of create_snapshot().
613 */
614 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false);
615 if (ret)
616 goto fail_free;
617
618 trans = btrfs_start_transaction(root, 0);
619 if (IS_ERR(trans)) {
620 ret = PTR_ERR(trans);
621 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
622 goto fail_free;
623 }
624 trans->block_rsv = &block_rsv;
625 trans->bytes_reserved = block_rsv.size;
626
627 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
628 if (ret)
629 goto fail;
630
631 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
632 if (IS_ERR(leaf)) {
633 ret = PTR_ERR(leaf);
634 goto fail;
635 }
636
637 btrfs_mark_buffer_dirty(leaf);
638
639 inode_item = &root_item->inode;
640 btrfs_set_stack_inode_generation(inode_item, 1);
641 btrfs_set_stack_inode_size(inode_item, 3);
642 btrfs_set_stack_inode_nlink(inode_item, 1);
643 btrfs_set_stack_inode_nbytes(inode_item,
644 fs_info->nodesize);
645 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
646
647 btrfs_set_root_flags(root_item, 0);
648 btrfs_set_root_limit(root_item, 0);
649 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
650
651 btrfs_set_root_bytenr(root_item, leaf->start);
652 btrfs_set_root_generation(root_item, trans->transid);
653 btrfs_set_root_level(root_item, 0);
654 btrfs_set_root_refs(root_item, 1);
655 btrfs_set_root_used(root_item, leaf->len);
656 btrfs_set_root_last_snapshot(root_item, 0);
657
658 btrfs_set_root_generation_v2(root_item,
659 btrfs_root_generation(root_item));
660 uuid_le_gen(&new_uuid);
661 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
662 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
663 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
664 root_item->ctime = root_item->otime;
665 btrfs_set_root_ctransid(root_item, trans->transid);
666 btrfs_set_root_otransid(root_item, trans->transid);
667
668 btrfs_tree_unlock(leaf);
669 free_extent_buffer(leaf);
670 leaf = NULL;
671
672 btrfs_set_root_dirid(root_item, new_dirid);
673
674 key.objectid = objectid;
675 key.offset = 0;
676 key.type = BTRFS_ROOT_ITEM_KEY;
677 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
678 root_item);
679 if (ret)
680 goto fail;
681
682 key.offset = (u64)-1;
683 new_root = btrfs_read_fs_root_no_name(fs_info, &key);
684 if (IS_ERR(new_root)) {
685 ret = PTR_ERR(new_root);
686 btrfs_abort_transaction(trans, ret);
687 goto fail;
688 }
689
690 btrfs_record_root_in_trans(trans, new_root);
691
692 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
693 if (ret) {
694 /* We potentially lose an unused inode item here */
695 btrfs_abort_transaction(trans, ret);
696 goto fail;
697 }
698
699 mutex_lock(&new_root->objectid_mutex);
700 new_root->highest_objectid = new_dirid;
701 mutex_unlock(&new_root->objectid_mutex);
702
703 /*
704 * insert the directory item
705 */
706 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
707 if (ret) {
708 btrfs_abort_transaction(trans, ret);
709 goto fail;
710 }
711
712 ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
713 BTRFS_FT_DIR, index);
714 if (ret) {
715 btrfs_abort_transaction(trans, ret);
716 goto fail;
717 }
718
719 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
720 ret = btrfs_update_inode(trans, root, dir);
721 BUG_ON(ret);
722
723 ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
724 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
725 BUG_ON(ret);
726
727 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
728 BTRFS_UUID_KEY_SUBVOL, objectid);
729 if (ret)
730 btrfs_abort_transaction(trans, ret);
731
732 fail:
733 kfree(root_item);
734 trans->block_rsv = NULL;
735 trans->bytes_reserved = 0;
736 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
737
738 if (async_transid) {
739 *async_transid = trans->transid;
740 err = btrfs_commit_transaction_async(trans, 1);
741 if (err)
742 err = btrfs_commit_transaction(trans);
743 } else {
744 err = btrfs_commit_transaction(trans);
745 }
746 if (err && !ret)
747 ret = err;
748
749 if (!ret) {
750 inode = btrfs_lookup_dentry(dir, dentry);
751 if (IS_ERR(inode))
752 return PTR_ERR(inode);
753 d_instantiate(dentry, inode);
754 }
755 return ret;
756
757 fail_free:
758 kfree(root_item);
759 return ret;
760 }
761
762 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
763 struct dentry *dentry,
764 u64 *async_transid, bool readonly,
765 struct btrfs_qgroup_inherit *inherit)
766 {
767 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
768 struct inode *inode;
769 struct btrfs_pending_snapshot *pending_snapshot;
770 struct btrfs_trans_handle *trans;
771 int ret;
772 bool snapshot_force_cow = false;
773
774 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
775 return -EINVAL;
776
777 if (atomic_read(&root->nr_swapfiles)) {
778 btrfs_warn(fs_info,
779 "cannot snapshot subvolume with active swapfile");
780 return -ETXTBSY;
781 }
782
783 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
784 if (!pending_snapshot)
785 return -ENOMEM;
786
787 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
788 GFP_KERNEL);
789 pending_snapshot->path = btrfs_alloc_path();
790 if (!pending_snapshot->root_item || !pending_snapshot->path) {
791 ret = -ENOMEM;
792 goto free_pending;
793 }
794
795 /*
796 * Force new buffered writes to reserve space even when NOCOW is
797 * possible. This is to avoid later writeback (running dealloc) to
798 * fallback to COW mode and unexpectedly fail with ENOSPC.
799 */
800 atomic_inc(&root->will_be_snapshotted);
801 smp_mb__after_atomic();
802 /* wait for no snapshot writes */
803 wait_event(root->subv_writers->wait,
804 percpu_counter_sum(&root->subv_writers->counter) == 0);
805
806 ret = btrfs_start_delalloc_snapshot(root);
807 if (ret)
808 goto dec_and_free;
809
810 /*
811 * All previous writes have started writeback in NOCOW mode, so now
812 * we force future writes to fallback to COW mode during snapshot
813 * creation.
814 */
815 atomic_inc(&root->snapshot_force_cow);
816 snapshot_force_cow = true;
817
818 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
819
820 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
821 BTRFS_BLOCK_RSV_TEMP);
822 /*
823 * 1 - parent dir inode
824 * 2 - dir entries
825 * 1 - root item
826 * 2 - root ref/backref
827 * 1 - root of snapshot
828 * 1 - UUID item
829 */
830 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
831 &pending_snapshot->block_rsv, 8,
832 false);
833 if (ret)
834 goto dec_and_free;
835
836 pending_snapshot->dentry = dentry;
837 pending_snapshot->root = root;
838 pending_snapshot->readonly = readonly;
839 pending_snapshot->dir = dir;
840 pending_snapshot->inherit = inherit;
841
842 trans = btrfs_start_transaction(root, 0);
843 if (IS_ERR(trans)) {
844 ret = PTR_ERR(trans);
845 goto fail;
846 }
847
848 spin_lock(&fs_info->trans_lock);
849 list_add(&pending_snapshot->list,
850 &trans->transaction->pending_snapshots);
851 spin_unlock(&fs_info->trans_lock);
852 if (async_transid) {
853 *async_transid = trans->transid;
854 ret = btrfs_commit_transaction_async(trans, 1);
855 if (ret)
856 ret = btrfs_commit_transaction(trans);
857 } else {
858 ret = btrfs_commit_transaction(trans);
859 }
860 if (ret)
861 goto fail;
862
863 ret = pending_snapshot->error;
864 if (ret)
865 goto fail;
866
867 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
868 if (ret)
869 goto fail;
870
871 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
872 if (IS_ERR(inode)) {
873 ret = PTR_ERR(inode);
874 goto fail;
875 }
876
877 d_instantiate(dentry, inode);
878 ret = 0;
879 fail:
880 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
881 dec_and_free:
882 if (snapshot_force_cow)
883 atomic_dec(&root->snapshot_force_cow);
884 if (atomic_dec_and_test(&root->will_be_snapshotted))
885 wake_up_var(&root->will_be_snapshotted);
886 free_pending:
887 kfree(pending_snapshot->root_item);
888 btrfs_free_path(pending_snapshot->path);
889 kfree(pending_snapshot);
890
891 return ret;
892 }
893
894 /* copy of may_delete in fs/namei.c()
895 * Check whether we can remove a link victim from directory dir, check
896 * whether the type of victim is right.
897 * 1. We can't do it if dir is read-only (done in permission())
898 * 2. We should have write and exec permissions on dir
899 * 3. We can't remove anything from append-only dir
900 * 4. We can't do anything with immutable dir (done in permission())
901 * 5. If the sticky bit on dir is set we should either
902 * a. be owner of dir, or
903 * b. be owner of victim, or
904 * c. have CAP_FOWNER capability
905 * 6. If the victim is append-only or immutable we can't do anything with
906 * links pointing to it.
907 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
908 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
909 * 9. We can't remove a root or mountpoint.
910 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
911 * nfs_async_unlink().
912 */
913
914 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
915 {
916 int error;
917
918 if (d_really_is_negative(victim))
919 return -ENOENT;
920
921 BUG_ON(d_inode(victim->d_parent) != dir);
922 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
923
924 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
925 if (error)
926 return error;
927 if (IS_APPEND(dir))
928 return -EPERM;
929 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
930 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
931 return -EPERM;
932 if (isdir) {
933 if (!d_is_dir(victim))
934 return -ENOTDIR;
935 if (IS_ROOT(victim))
936 return -EBUSY;
937 } else if (d_is_dir(victim))
938 return -EISDIR;
939 if (IS_DEADDIR(dir))
940 return -ENOENT;
941 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
942 return -EBUSY;
943 return 0;
944 }
945
946 /* copy of may_create in fs/namei.c() */
947 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
948 {
949 if (d_really_is_positive(child))
950 return -EEXIST;
951 if (IS_DEADDIR(dir))
952 return -ENOENT;
953 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
954 }
955
956 /*
957 * Create a new subvolume below @parent. This is largely modeled after
958 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
959 * inside this filesystem so it's quite a bit simpler.
960 */
961 static noinline int btrfs_mksubvol(const struct path *parent,
962 const char *name, int namelen,
963 struct btrfs_root *snap_src,
964 u64 *async_transid, bool readonly,
965 struct btrfs_qgroup_inherit *inherit)
966 {
967 struct inode *dir = d_inode(parent->dentry);
968 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
969 struct dentry *dentry;
970 int error;
971
972 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
973 if (error == -EINTR)
974 return error;
975
976 dentry = lookup_one_len(name, parent->dentry, namelen);
977 error = PTR_ERR(dentry);
978 if (IS_ERR(dentry))
979 goto out_unlock;
980
981 error = btrfs_may_create(dir, dentry);
982 if (error)
983 goto out_dput;
984
985 /*
986 * even if this name doesn't exist, we may get hash collisions.
987 * check for them now when we can safely fail
988 */
989 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
990 dir->i_ino, name,
991 namelen);
992 if (error)
993 goto out_dput;
994
995 down_read(&fs_info->subvol_sem);
996
997 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
998 goto out_up_read;
999
1000 if (snap_src) {
1001 error = create_snapshot(snap_src, dir, dentry,
1002 async_transid, readonly, inherit);
1003 } else {
1004 error = create_subvol(dir, dentry, name, namelen,
1005 async_transid, inherit);
1006 }
1007 if (!error)
1008 fsnotify_mkdir(dir, dentry);
1009 out_up_read:
1010 up_read(&fs_info->subvol_sem);
1011 out_dput:
1012 dput(dentry);
1013 out_unlock:
1014 inode_unlock(dir);
1015 return error;
1016 }
1017
1018 /*
1019 * When we're defragging a range, we don't want to kick it off again
1020 * if it is really just waiting for delalloc to send it down.
1021 * If we find a nice big extent or delalloc range for the bytes in the
1022 * file you want to defrag, we return 0 to let you know to skip this
1023 * part of the file
1024 */
1025 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
1026 {
1027 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1028 struct extent_map *em = NULL;
1029 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1030 u64 end;
1031
1032 read_lock(&em_tree->lock);
1033 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
1034 read_unlock(&em_tree->lock);
1035
1036 if (em) {
1037 end = extent_map_end(em);
1038 free_extent_map(em);
1039 if (end - offset > thresh)
1040 return 0;
1041 }
1042 /* if we already have a nice delalloc here, just stop */
1043 thresh /= 2;
1044 end = count_range_bits(io_tree, &offset, offset + thresh,
1045 thresh, EXTENT_DELALLOC, 1);
1046 if (end >= thresh)
1047 return 0;
1048 return 1;
1049 }
1050
1051 /*
1052 * helper function to walk through a file and find extents
1053 * newer than a specific transid, and smaller than thresh.
1054 *
1055 * This is used by the defragging code to find new and small
1056 * extents
1057 */
1058 static int find_new_extents(struct btrfs_root *root,
1059 struct inode *inode, u64 newer_than,
1060 u64 *off, u32 thresh)
1061 {
1062 struct btrfs_path *path;
1063 struct btrfs_key min_key;
1064 struct extent_buffer *leaf;
1065 struct btrfs_file_extent_item *extent;
1066 int type;
1067 int ret;
1068 u64 ino = btrfs_ino(BTRFS_I(inode));
1069
1070 path = btrfs_alloc_path();
1071 if (!path)
1072 return -ENOMEM;
1073
1074 min_key.objectid = ino;
1075 min_key.type = BTRFS_EXTENT_DATA_KEY;
1076 min_key.offset = *off;
1077
1078 while (1) {
1079 ret = btrfs_search_forward(root, &min_key, path, newer_than);
1080 if (ret != 0)
1081 goto none;
1082 process_slot:
1083 if (min_key.objectid != ino)
1084 goto none;
1085 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1086 goto none;
1087
1088 leaf = path->nodes[0];
1089 extent = btrfs_item_ptr(leaf, path->slots[0],
1090 struct btrfs_file_extent_item);
1091
1092 type = btrfs_file_extent_type(leaf, extent);
1093 if (type == BTRFS_FILE_EXTENT_REG &&
1094 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1095 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1096 *off = min_key.offset;
1097 btrfs_free_path(path);
1098 return 0;
1099 }
1100
1101 path->slots[0]++;
1102 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1103 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1104 goto process_slot;
1105 }
1106
1107 if (min_key.offset == (u64)-1)
1108 goto none;
1109
1110 min_key.offset++;
1111 btrfs_release_path(path);
1112 }
1113 none:
1114 btrfs_free_path(path);
1115 return -ENOENT;
1116 }
1117
1118 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1119 {
1120 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1121 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1122 struct extent_map *em;
1123 u64 len = PAGE_SIZE;
1124
1125 /*
1126 * hopefully we have this extent in the tree already, try without
1127 * the full extent lock
1128 */
1129 read_lock(&em_tree->lock);
1130 em = lookup_extent_mapping(em_tree, start, len);
1131 read_unlock(&em_tree->lock);
1132
1133 if (!em) {
1134 struct extent_state *cached = NULL;
1135 u64 end = start + len - 1;
1136
1137 /* get the big lock and read metadata off disk */
1138 lock_extent_bits(io_tree, start, end, &cached);
1139 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0);
1140 unlock_extent_cached(io_tree, start, end, &cached);
1141
1142 if (IS_ERR(em))
1143 return NULL;
1144 }
1145
1146 return em;
1147 }
1148
1149 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1150 {
1151 struct extent_map *next;
1152 bool ret = true;
1153
1154 /* this is the last extent */
1155 if (em->start + em->len >= i_size_read(inode))
1156 return false;
1157
1158 next = defrag_lookup_extent(inode, em->start + em->len);
1159 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1160 ret = false;
1161 else if ((em->block_start + em->block_len == next->block_start) &&
1162 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1163 ret = false;
1164
1165 free_extent_map(next);
1166 return ret;
1167 }
1168
1169 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1170 u64 *last_len, u64 *skip, u64 *defrag_end,
1171 int compress)
1172 {
1173 struct extent_map *em;
1174 int ret = 1;
1175 bool next_mergeable = true;
1176 bool prev_mergeable = true;
1177
1178 /*
1179 * make sure that once we start defragging an extent, we keep on
1180 * defragging it
1181 */
1182 if (start < *defrag_end)
1183 return 1;
1184
1185 *skip = 0;
1186
1187 em = defrag_lookup_extent(inode, start);
1188 if (!em)
1189 return 0;
1190
1191 /* this will cover holes, and inline extents */
1192 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1193 ret = 0;
1194 goto out;
1195 }
1196
1197 if (!*defrag_end)
1198 prev_mergeable = false;
1199
1200 next_mergeable = defrag_check_next_extent(inode, em);
1201 /*
1202 * we hit a real extent, if it is big or the next extent is not a
1203 * real extent, don't bother defragging it
1204 */
1205 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1206 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1207 ret = 0;
1208 out:
1209 /*
1210 * last_len ends up being a counter of how many bytes we've defragged.
1211 * every time we choose not to defrag an extent, we reset *last_len
1212 * so that the next tiny extent will force a defrag.
1213 *
1214 * The end result of this is that tiny extents before a single big
1215 * extent will force at least part of that big extent to be defragged.
1216 */
1217 if (ret) {
1218 *defrag_end = extent_map_end(em);
1219 } else {
1220 *last_len = 0;
1221 *skip = extent_map_end(em);
1222 *defrag_end = 0;
1223 }
1224
1225 free_extent_map(em);
1226 return ret;
1227 }
1228
1229 /*
1230 * it doesn't do much good to defrag one or two pages
1231 * at a time. This pulls in a nice chunk of pages
1232 * to COW and defrag.
1233 *
1234 * It also makes sure the delalloc code has enough
1235 * dirty data to avoid making new small extents as part
1236 * of the defrag
1237 *
1238 * It's a good idea to start RA on this range
1239 * before calling this.
1240 */
1241 static int cluster_pages_for_defrag(struct inode *inode,
1242 struct page **pages,
1243 unsigned long start_index,
1244 unsigned long num_pages)
1245 {
1246 unsigned long file_end;
1247 u64 isize = i_size_read(inode);
1248 u64 page_start;
1249 u64 page_end;
1250 u64 page_cnt;
1251 int ret;
1252 int i;
1253 int i_done;
1254 struct btrfs_ordered_extent *ordered;
1255 struct extent_state *cached_state = NULL;
1256 struct extent_io_tree *tree;
1257 struct extent_changeset *data_reserved = NULL;
1258 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1259
1260 file_end = (isize - 1) >> PAGE_SHIFT;
1261 if (!isize || start_index > file_end)
1262 return 0;
1263
1264 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1265
1266 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
1267 start_index << PAGE_SHIFT,
1268 page_cnt << PAGE_SHIFT);
1269 if (ret)
1270 return ret;
1271 i_done = 0;
1272 tree = &BTRFS_I(inode)->io_tree;
1273
1274 /* step one, lock all the pages */
1275 for (i = 0; i < page_cnt; i++) {
1276 struct page *page;
1277 again:
1278 page = find_or_create_page(inode->i_mapping,
1279 start_index + i, mask);
1280 if (!page)
1281 break;
1282
1283 page_start = page_offset(page);
1284 page_end = page_start + PAGE_SIZE - 1;
1285 while (1) {
1286 lock_extent_bits(tree, page_start, page_end,
1287 &cached_state);
1288 ordered = btrfs_lookup_ordered_extent(inode,
1289 page_start);
1290 unlock_extent_cached(tree, page_start, page_end,
1291 &cached_state);
1292 if (!ordered)
1293 break;
1294
1295 unlock_page(page);
1296 btrfs_start_ordered_extent(inode, ordered, 1);
1297 btrfs_put_ordered_extent(ordered);
1298 lock_page(page);
1299 /*
1300 * we unlocked the page above, so we need check if
1301 * it was released or not.
1302 */
1303 if (page->mapping != inode->i_mapping) {
1304 unlock_page(page);
1305 put_page(page);
1306 goto again;
1307 }
1308 }
1309
1310 if (!PageUptodate(page)) {
1311 btrfs_readpage(NULL, page);
1312 lock_page(page);
1313 if (!PageUptodate(page)) {
1314 unlock_page(page);
1315 put_page(page);
1316 ret = -EIO;
1317 break;
1318 }
1319 }
1320
1321 if (page->mapping != inode->i_mapping) {
1322 unlock_page(page);
1323 put_page(page);
1324 goto again;
1325 }
1326
1327 pages[i] = page;
1328 i_done++;
1329 }
1330 if (!i_done || ret)
1331 goto out;
1332
1333 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1334 goto out;
1335
1336 /*
1337 * so now we have a nice long stream of locked
1338 * and up to date pages, lets wait on them
1339 */
1340 for (i = 0; i < i_done; i++)
1341 wait_on_page_writeback(pages[i]);
1342
1343 page_start = page_offset(pages[0]);
1344 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1345
1346 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1347 page_start, page_end - 1, &cached_state);
1348 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1349 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1350 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1351 &cached_state);
1352
1353 if (i_done != page_cnt) {
1354 spin_lock(&BTRFS_I(inode)->lock);
1355 btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
1356 spin_unlock(&BTRFS_I(inode)->lock);
1357 btrfs_delalloc_release_space(inode, data_reserved,
1358 start_index << PAGE_SHIFT,
1359 (page_cnt - i_done) << PAGE_SHIFT, true);
1360 }
1361
1362
1363 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1364 &cached_state);
1365
1366 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1367 page_start, page_end - 1, &cached_state);
1368
1369 for (i = 0; i < i_done; i++) {
1370 clear_page_dirty_for_io(pages[i]);
1371 ClearPageChecked(pages[i]);
1372 set_page_extent_mapped(pages[i]);
1373 set_page_dirty(pages[i]);
1374 unlock_page(pages[i]);
1375 put_page(pages[i]);
1376 }
1377 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1378 false);
1379 extent_changeset_free(data_reserved);
1380 return i_done;
1381 out:
1382 for (i = 0; i < i_done; i++) {
1383 unlock_page(pages[i]);
1384 put_page(pages[i]);
1385 }
1386 btrfs_delalloc_release_space(inode, data_reserved,
1387 start_index << PAGE_SHIFT,
1388 page_cnt << PAGE_SHIFT, true);
1389 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1390 true);
1391 extent_changeset_free(data_reserved);
1392 return ret;
1393
1394 }
1395
1396 int btrfs_defrag_file(struct inode *inode, struct file *file,
1397 struct btrfs_ioctl_defrag_range_args *range,
1398 u64 newer_than, unsigned long max_to_defrag)
1399 {
1400 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1401 struct btrfs_root *root = BTRFS_I(inode)->root;
1402 struct file_ra_state *ra = NULL;
1403 unsigned long last_index;
1404 u64 isize = i_size_read(inode);
1405 u64 last_len = 0;
1406 u64 skip = 0;
1407 u64 defrag_end = 0;
1408 u64 newer_off = range->start;
1409 unsigned long i;
1410 unsigned long ra_index = 0;
1411 int ret;
1412 int defrag_count = 0;
1413 int compress_type = BTRFS_COMPRESS_ZLIB;
1414 u32 extent_thresh = range->extent_thresh;
1415 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1416 unsigned long cluster = max_cluster;
1417 u64 new_align = ~((u64)SZ_128K - 1);
1418 struct page **pages = NULL;
1419 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1420
1421 if (isize == 0)
1422 return 0;
1423
1424 if (range->start >= isize)
1425 return -EINVAL;
1426
1427 if (do_compress) {
1428 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1429 return -EINVAL;
1430 if (range->compress_type)
1431 compress_type = range->compress_type;
1432 }
1433
1434 if (extent_thresh == 0)
1435 extent_thresh = SZ_256K;
1436
1437 /*
1438 * If we were not given a file, allocate a readahead context. As
1439 * readahead is just an optimization, defrag will work without it so
1440 * we don't error out.
1441 */
1442 if (!file) {
1443 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1444 if (ra)
1445 file_ra_state_init(ra, inode->i_mapping);
1446 } else {
1447 ra = &file->f_ra;
1448 }
1449
1450 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1451 if (!pages) {
1452 ret = -ENOMEM;
1453 goto out_ra;
1454 }
1455
1456 /* find the last page to defrag */
1457 if (range->start + range->len > range->start) {
1458 last_index = min_t(u64, isize - 1,
1459 range->start + range->len - 1) >> PAGE_SHIFT;
1460 } else {
1461 last_index = (isize - 1) >> PAGE_SHIFT;
1462 }
1463
1464 if (newer_than) {
1465 ret = find_new_extents(root, inode, newer_than,
1466 &newer_off, SZ_64K);
1467 if (!ret) {
1468 range->start = newer_off;
1469 /*
1470 * we always align our defrag to help keep
1471 * the extents in the file evenly spaced
1472 */
1473 i = (newer_off & new_align) >> PAGE_SHIFT;
1474 } else
1475 goto out_ra;
1476 } else {
1477 i = range->start >> PAGE_SHIFT;
1478 }
1479 if (!max_to_defrag)
1480 max_to_defrag = last_index - i + 1;
1481
1482 /*
1483 * make writeback starts from i, so the defrag range can be
1484 * written sequentially.
1485 */
1486 if (i < inode->i_mapping->writeback_index)
1487 inode->i_mapping->writeback_index = i;
1488
1489 while (i <= last_index && defrag_count < max_to_defrag &&
1490 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1491 /*
1492 * make sure we stop running if someone unmounts
1493 * the FS
1494 */
1495 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1496 break;
1497
1498 if (btrfs_defrag_cancelled(fs_info)) {
1499 btrfs_debug(fs_info, "defrag_file cancelled");
1500 ret = -EAGAIN;
1501 break;
1502 }
1503
1504 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1505 extent_thresh, &last_len, &skip,
1506 &defrag_end, do_compress)){
1507 unsigned long next;
1508 /*
1509 * the should_defrag function tells us how much to skip
1510 * bump our counter by the suggested amount
1511 */
1512 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1513 i = max(i + 1, next);
1514 continue;
1515 }
1516
1517 if (!newer_than) {
1518 cluster = (PAGE_ALIGN(defrag_end) >>
1519 PAGE_SHIFT) - i;
1520 cluster = min(cluster, max_cluster);
1521 } else {
1522 cluster = max_cluster;
1523 }
1524
1525 if (i + cluster > ra_index) {
1526 ra_index = max(i, ra_index);
1527 if (ra)
1528 page_cache_sync_readahead(inode->i_mapping, ra,
1529 file, ra_index, cluster);
1530 ra_index += cluster;
1531 }
1532
1533 inode_lock(inode);
1534 if (IS_SWAPFILE(inode)) {
1535 ret = -ETXTBSY;
1536 } else {
1537 if (do_compress)
1538 BTRFS_I(inode)->defrag_compress = compress_type;
1539 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1540 }
1541 if (ret < 0) {
1542 inode_unlock(inode);
1543 goto out_ra;
1544 }
1545
1546 defrag_count += ret;
1547 balance_dirty_pages_ratelimited(inode->i_mapping);
1548 inode_unlock(inode);
1549
1550 if (newer_than) {
1551 if (newer_off == (u64)-1)
1552 break;
1553
1554 if (ret > 0)
1555 i += ret;
1556
1557 newer_off = max(newer_off + 1,
1558 (u64)i << PAGE_SHIFT);
1559
1560 ret = find_new_extents(root, inode, newer_than,
1561 &newer_off, SZ_64K);
1562 if (!ret) {
1563 range->start = newer_off;
1564 i = (newer_off & new_align) >> PAGE_SHIFT;
1565 } else {
1566 break;
1567 }
1568 } else {
1569 if (ret > 0) {
1570 i += ret;
1571 last_len += ret << PAGE_SHIFT;
1572 } else {
1573 i++;
1574 last_len = 0;
1575 }
1576 }
1577 }
1578
1579 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1580 filemap_flush(inode->i_mapping);
1581 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1582 &BTRFS_I(inode)->runtime_flags))
1583 filemap_flush(inode->i_mapping);
1584 }
1585
1586 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1587 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1588 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1589 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1590 }
1591
1592 ret = defrag_count;
1593
1594 out_ra:
1595 if (do_compress) {
1596 inode_lock(inode);
1597 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1598 inode_unlock(inode);
1599 }
1600 if (!file)
1601 kfree(ra);
1602 kfree(pages);
1603 return ret;
1604 }
1605
1606 static noinline int btrfs_ioctl_resize(struct file *file,
1607 void __user *arg)
1608 {
1609 struct inode *inode = file_inode(file);
1610 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1611 u64 new_size;
1612 u64 old_size;
1613 u64 devid = 1;
1614 struct btrfs_root *root = BTRFS_I(inode)->root;
1615 struct btrfs_ioctl_vol_args *vol_args;
1616 struct btrfs_trans_handle *trans;
1617 struct btrfs_device *device = NULL;
1618 char *sizestr;
1619 char *retptr;
1620 char *devstr = NULL;
1621 int ret = 0;
1622 int mod = 0;
1623
1624 if (!capable(CAP_SYS_ADMIN))
1625 return -EPERM;
1626
1627 ret = mnt_want_write_file(file);
1628 if (ret)
1629 return ret;
1630
1631 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
1632 mnt_drop_write_file(file);
1633 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1634 }
1635
1636 vol_args = memdup_user(arg, sizeof(*vol_args));
1637 if (IS_ERR(vol_args)) {
1638 ret = PTR_ERR(vol_args);
1639 goto out;
1640 }
1641
1642 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1643
1644 sizestr = vol_args->name;
1645 devstr = strchr(sizestr, ':');
1646 if (devstr) {
1647 sizestr = devstr + 1;
1648 *devstr = '\0';
1649 devstr = vol_args->name;
1650 ret = kstrtoull(devstr, 10, &devid);
1651 if (ret)
1652 goto out_free;
1653 if (!devid) {
1654 ret = -EINVAL;
1655 goto out_free;
1656 }
1657 btrfs_info(fs_info, "resizing devid %llu", devid);
1658 }
1659
1660 device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true);
1661 if (!device) {
1662 btrfs_info(fs_info, "resizer unable to find device %llu",
1663 devid);
1664 ret = -ENODEV;
1665 goto out_free;
1666 }
1667
1668 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1669 btrfs_info(fs_info,
1670 "resizer unable to apply on readonly device %llu",
1671 devid);
1672 ret = -EPERM;
1673 goto out_free;
1674 }
1675
1676 if (!strcmp(sizestr, "max"))
1677 new_size = device->bdev->bd_inode->i_size;
1678 else {
1679 if (sizestr[0] == '-') {
1680 mod = -1;
1681 sizestr++;
1682 } else if (sizestr[0] == '+') {
1683 mod = 1;
1684 sizestr++;
1685 }
1686 new_size = memparse(sizestr, &retptr);
1687 if (*retptr != '\0' || new_size == 0) {
1688 ret = -EINVAL;
1689 goto out_free;
1690 }
1691 }
1692
1693 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1694 ret = -EPERM;
1695 goto out_free;
1696 }
1697
1698 old_size = btrfs_device_get_total_bytes(device);
1699
1700 if (mod < 0) {
1701 if (new_size > old_size) {
1702 ret = -EINVAL;
1703 goto out_free;
1704 }
1705 new_size = old_size - new_size;
1706 } else if (mod > 0) {
1707 if (new_size > ULLONG_MAX - old_size) {
1708 ret = -ERANGE;
1709 goto out_free;
1710 }
1711 new_size = old_size + new_size;
1712 }
1713
1714 if (new_size < SZ_256M) {
1715 ret = -EINVAL;
1716 goto out_free;
1717 }
1718 if (new_size > device->bdev->bd_inode->i_size) {
1719 ret = -EFBIG;
1720 goto out_free;
1721 }
1722
1723 new_size = round_down(new_size, fs_info->sectorsize);
1724
1725 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1726 rcu_str_deref(device->name), new_size);
1727
1728 if (new_size > old_size) {
1729 trans = btrfs_start_transaction(root, 0);
1730 if (IS_ERR(trans)) {
1731 ret = PTR_ERR(trans);
1732 goto out_free;
1733 }
1734 ret = btrfs_grow_device(trans, device, new_size);
1735 btrfs_commit_transaction(trans);
1736 } else if (new_size < old_size) {
1737 ret = btrfs_shrink_device(device, new_size);
1738 } /* equal, nothing need to do */
1739
1740 out_free:
1741 kfree(vol_args);
1742 out:
1743 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
1744 mnt_drop_write_file(file);
1745 return ret;
1746 }
1747
1748 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1749 const char *name, unsigned long fd, int subvol,
1750 u64 *transid, bool readonly,
1751 struct btrfs_qgroup_inherit *inherit)
1752 {
1753 int namelen;
1754 int ret = 0;
1755
1756 if (!S_ISDIR(file_inode(file)->i_mode))
1757 return -ENOTDIR;
1758
1759 ret = mnt_want_write_file(file);
1760 if (ret)
1761 goto out;
1762
1763 namelen = strlen(name);
1764 if (strchr(name, '/')) {
1765 ret = -EINVAL;
1766 goto out_drop_write;
1767 }
1768
1769 if (name[0] == '.' &&
1770 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1771 ret = -EEXIST;
1772 goto out_drop_write;
1773 }
1774
1775 if (subvol) {
1776 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1777 NULL, transid, readonly, inherit);
1778 } else {
1779 struct fd src = fdget(fd);
1780 struct inode *src_inode;
1781 if (!src.file) {
1782 ret = -EINVAL;
1783 goto out_drop_write;
1784 }
1785
1786 src_inode = file_inode(src.file);
1787 if (src_inode->i_sb != file_inode(file)->i_sb) {
1788 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1789 "Snapshot src from another FS");
1790 ret = -EXDEV;
1791 } else if (!inode_owner_or_capable(src_inode)) {
1792 /*
1793 * Subvolume creation is not restricted, but snapshots
1794 * are limited to own subvolumes only
1795 */
1796 ret = -EPERM;
1797 } else {
1798 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1799 BTRFS_I(src_inode)->root,
1800 transid, readonly, inherit);
1801 }
1802 fdput(src);
1803 }
1804 out_drop_write:
1805 mnt_drop_write_file(file);
1806 out:
1807 return ret;
1808 }
1809
1810 static noinline int btrfs_ioctl_snap_create(struct file *file,
1811 void __user *arg, int subvol)
1812 {
1813 struct btrfs_ioctl_vol_args *vol_args;
1814 int ret;
1815
1816 if (!S_ISDIR(file_inode(file)->i_mode))
1817 return -ENOTDIR;
1818
1819 vol_args = memdup_user(arg, sizeof(*vol_args));
1820 if (IS_ERR(vol_args))
1821 return PTR_ERR(vol_args);
1822 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1823
1824 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1825 vol_args->fd, subvol,
1826 NULL, false, NULL);
1827
1828 kfree(vol_args);
1829 return ret;
1830 }
1831
1832 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1833 void __user *arg, int subvol)
1834 {
1835 struct btrfs_ioctl_vol_args_v2 *vol_args;
1836 int ret;
1837 u64 transid = 0;
1838 u64 *ptr = NULL;
1839 bool readonly = false;
1840 struct btrfs_qgroup_inherit *inherit = NULL;
1841
1842 if (!S_ISDIR(file_inode(file)->i_mode))
1843 return -ENOTDIR;
1844
1845 vol_args = memdup_user(arg, sizeof(*vol_args));
1846 if (IS_ERR(vol_args))
1847 return PTR_ERR(vol_args);
1848 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1849
1850 if (vol_args->flags &
1851 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1852 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1853 ret = -EOPNOTSUPP;
1854 goto free_args;
1855 }
1856
1857 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1858 ptr = &transid;
1859 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1860 readonly = true;
1861 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1862 if (vol_args->size > PAGE_SIZE) {
1863 ret = -EINVAL;
1864 goto free_args;
1865 }
1866 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1867 if (IS_ERR(inherit)) {
1868 ret = PTR_ERR(inherit);
1869 goto free_args;
1870 }
1871 }
1872
1873 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1874 vol_args->fd, subvol, ptr,
1875 readonly, inherit);
1876 if (ret)
1877 goto free_inherit;
1878
1879 if (ptr && copy_to_user(arg +
1880 offsetof(struct btrfs_ioctl_vol_args_v2,
1881 transid),
1882 ptr, sizeof(*ptr)))
1883 ret = -EFAULT;
1884
1885 free_inherit:
1886 kfree(inherit);
1887 free_args:
1888 kfree(vol_args);
1889 return ret;
1890 }
1891
1892 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1893 void __user *arg)
1894 {
1895 struct inode *inode = file_inode(file);
1896 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1897 struct btrfs_root *root = BTRFS_I(inode)->root;
1898 int ret = 0;
1899 u64 flags = 0;
1900
1901 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1902 return -EINVAL;
1903
1904 down_read(&fs_info->subvol_sem);
1905 if (btrfs_root_readonly(root))
1906 flags |= BTRFS_SUBVOL_RDONLY;
1907 up_read(&fs_info->subvol_sem);
1908
1909 if (copy_to_user(arg, &flags, sizeof(flags)))
1910 ret = -EFAULT;
1911
1912 return ret;
1913 }
1914
1915 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1916 void __user *arg)
1917 {
1918 struct inode *inode = file_inode(file);
1919 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1920 struct btrfs_root *root = BTRFS_I(inode)->root;
1921 struct btrfs_trans_handle *trans;
1922 u64 root_flags;
1923 u64 flags;
1924 int ret = 0;
1925
1926 if (!inode_owner_or_capable(inode))
1927 return -EPERM;
1928
1929 ret = mnt_want_write_file(file);
1930 if (ret)
1931 goto out;
1932
1933 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1934 ret = -EINVAL;
1935 goto out_drop_write;
1936 }
1937
1938 if (copy_from_user(&flags, arg, sizeof(flags))) {
1939 ret = -EFAULT;
1940 goto out_drop_write;
1941 }
1942
1943 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1944 ret = -EINVAL;
1945 goto out_drop_write;
1946 }
1947
1948 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1949 ret = -EOPNOTSUPP;
1950 goto out_drop_write;
1951 }
1952
1953 down_write(&fs_info->subvol_sem);
1954
1955 /* nothing to do */
1956 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1957 goto out_drop_sem;
1958
1959 root_flags = btrfs_root_flags(&root->root_item);
1960 if (flags & BTRFS_SUBVOL_RDONLY) {
1961 btrfs_set_root_flags(&root->root_item,
1962 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1963 } else {
1964 /*
1965 * Block RO -> RW transition if this subvolume is involved in
1966 * send
1967 */
1968 spin_lock(&root->root_item_lock);
1969 if (root->send_in_progress == 0) {
1970 btrfs_set_root_flags(&root->root_item,
1971 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1972 spin_unlock(&root->root_item_lock);
1973 } else {
1974 spin_unlock(&root->root_item_lock);
1975 btrfs_warn(fs_info,
1976 "Attempt to set subvolume %llu read-write during send",
1977 root->root_key.objectid);
1978 ret = -EPERM;
1979 goto out_drop_sem;
1980 }
1981 }
1982
1983 trans = btrfs_start_transaction(root, 1);
1984 if (IS_ERR(trans)) {
1985 ret = PTR_ERR(trans);
1986 goto out_reset;
1987 }
1988
1989 ret = btrfs_update_root(trans, fs_info->tree_root,
1990 &root->root_key, &root->root_item);
1991 if (ret < 0) {
1992 btrfs_end_transaction(trans);
1993 goto out_reset;
1994 }
1995
1996 ret = btrfs_commit_transaction(trans);
1997
1998 out_reset:
1999 if (ret)
2000 btrfs_set_root_flags(&root->root_item, root_flags);
2001 out_drop_sem:
2002 up_write(&fs_info->subvol_sem);
2003 out_drop_write:
2004 mnt_drop_write_file(file);
2005 out:
2006 return ret;
2007 }
2008
2009 static noinline int key_in_sk(struct btrfs_key *key,
2010 struct btrfs_ioctl_search_key *sk)
2011 {
2012 struct btrfs_key test;
2013 int ret;
2014
2015 test.objectid = sk->min_objectid;
2016 test.type = sk->min_type;
2017 test.offset = sk->min_offset;
2018
2019 ret = btrfs_comp_cpu_keys(key, &test);
2020 if (ret < 0)
2021 return 0;
2022
2023 test.objectid = sk->max_objectid;
2024 test.type = sk->max_type;
2025 test.offset = sk->max_offset;
2026
2027 ret = btrfs_comp_cpu_keys(key, &test);
2028 if (ret > 0)
2029 return 0;
2030 return 1;
2031 }
2032
2033 static noinline int copy_to_sk(struct btrfs_path *path,
2034 struct btrfs_key *key,
2035 struct btrfs_ioctl_search_key *sk,
2036 size_t *buf_size,
2037 char __user *ubuf,
2038 unsigned long *sk_offset,
2039 int *num_found)
2040 {
2041 u64 found_transid;
2042 struct extent_buffer *leaf;
2043 struct btrfs_ioctl_search_header sh;
2044 struct btrfs_key test;
2045 unsigned long item_off;
2046 unsigned long item_len;
2047 int nritems;
2048 int i;
2049 int slot;
2050 int ret = 0;
2051
2052 leaf = path->nodes[0];
2053 slot = path->slots[0];
2054 nritems = btrfs_header_nritems(leaf);
2055
2056 if (btrfs_header_generation(leaf) > sk->max_transid) {
2057 i = nritems;
2058 goto advance_key;
2059 }
2060 found_transid = btrfs_header_generation(leaf);
2061
2062 for (i = slot; i < nritems; i++) {
2063 item_off = btrfs_item_ptr_offset(leaf, i);
2064 item_len = btrfs_item_size_nr(leaf, i);
2065
2066 btrfs_item_key_to_cpu(leaf, key, i);
2067 if (!key_in_sk(key, sk))
2068 continue;
2069
2070 if (sizeof(sh) + item_len > *buf_size) {
2071 if (*num_found) {
2072 ret = 1;
2073 goto out;
2074 }
2075
2076 /*
2077 * return one empty item back for v1, which does not
2078 * handle -EOVERFLOW
2079 */
2080
2081 *buf_size = sizeof(sh) + item_len;
2082 item_len = 0;
2083 ret = -EOVERFLOW;
2084 }
2085
2086 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2087 ret = 1;
2088 goto out;
2089 }
2090
2091 sh.objectid = key->objectid;
2092 sh.offset = key->offset;
2093 sh.type = key->type;
2094 sh.len = item_len;
2095 sh.transid = found_transid;
2096
2097 /* copy search result header */
2098 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2099 ret = -EFAULT;
2100 goto out;
2101 }
2102
2103 *sk_offset += sizeof(sh);
2104
2105 if (item_len) {
2106 char __user *up = ubuf + *sk_offset;
2107 /* copy the item */
2108 if (read_extent_buffer_to_user(leaf, up,
2109 item_off, item_len)) {
2110 ret = -EFAULT;
2111 goto out;
2112 }
2113
2114 *sk_offset += item_len;
2115 }
2116 (*num_found)++;
2117
2118 if (ret) /* -EOVERFLOW from above */
2119 goto out;
2120
2121 if (*num_found >= sk->nr_items) {
2122 ret = 1;
2123 goto out;
2124 }
2125 }
2126 advance_key:
2127 ret = 0;
2128 test.objectid = sk->max_objectid;
2129 test.type = sk->max_type;
2130 test.offset = sk->max_offset;
2131 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2132 ret = 1;
2133 else if (key->offset < (u64)-1)
2134 key->offset++;
2135 else if (key->type < (u8)-1) {
2136 key->offset = 0;
2137 key->type++;
2138 } else if (key->objectid < (u64)-1) {
2139 key->offset = 0;
2140 key->type = 0;
2141 key->objectid++;
2142 } else
2143 ret = 1;
2144 out:
2145 /*
2146 * 0: all items from this leaf copied, continue with next
2147 * 1: * more items can be copied, but unused buffer is too small
2148 * * all items were found
2149 * Either way, it will stops the loop which iterates to the next
2150 * leaf
2151 * -EOVERFLOW: item was to large for buffer
2152 * -EFAULT: could not copy extent buffer back to userspace
2153 */
2154 return ret;
2155 }
2156
2157 static noinline int search_ioctl(struct inode *inode,
2158 struct btrfs_ioctl_search_key *sk,
2159 size_t *buf_size,
2160 char __user *ubuf)
2161 {
2162 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2163 struct btrfs_root *root;
2164 struct btrfs_key key;
2165 struct btrfs_path *path;
2166 int ret;
2167 int num_found = 0;
2168 unsigned long sk_offset = 0;
2169
2170 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2171 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2172 return -EOVERFLOW;
2173 }
2174
2175 path = btrfs_alloc_path();
2176 if (!path)
2177 return -ENOMEM;
2178
2179 if (sk->tree_id == 0) {
2180 /* search the root of the inode that was passed */
2181 root = BTRFS_I(inode)->root;
2182 } else {
2183 key.objectid = sk->tree_id;
2184 key.type = BTRFS_ROOT_ITEM_KEY;
2185 key.offset = (u64)-1;
2186 root = btrfs_read_fs_root_no_name(info, &key);
2187 if (IS_ERR(root)) {
2188 btrfs_free_path(path);
2189 return PTR_ERR(root);
2190 }
2191 }
2192
2193 key.objectid = sk->min_objectid;
2194 key.type = sk->min_type;
2195 key.offset = sk->min_offset;
2196
2197 while (1) {
2198 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2199 if (ret != 0) {
2200 if (ret > 0)
2201 ret = 0;
2202 goto err;
2203 }
2204 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2205 &sk_offset, &num_found);
2206 btrfs_release_path(path);
2207 if (ret)
2208 break;
2209
2210 }
2211 if (ret > 0)
2212 ret = 0;
2213 err:
2214 sk->nr_items = num_found;
2215 btrfs_free_path(path);
2216 return ret;
2217 }
2218
2219 static noinline int btrfs_ioctl_tree_search(struct file *file,
2220 void __user *argp)
2221 {
2222 struct btrfs_ioctl_search_args __user *uargs;
2223 struct btrfs_ioctl_search_key sk;
2224 struct inode *inode;
2225 int ret;
2226 size_t buf_size;
2227
2228 if (!capable(CAP_SYS_ADMIN))
2229 return -EPERM;
2230
2231 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2232
2233 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2234 return -EFAULT;
2235
2236 buf_size = sizeof(uargs->buf);
2237
2238 inode = file_inode(file);
2239 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2240
2241 /*
2242 * In the origin implementation an overflow is handled by returning a
2243 * search header with a len of zero, so reset ret.
2244 */
2245 if (ret == -EOVERFLOW)
2246 ret = 0;
2247
2248 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2249 ret = -EFAULT;
2250 return ret;
2251 }
2252
2253 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2254 void __user *argp)
2255 {
2256 struct btrfs_ioctl_search_args_v2 __user *uarg;
2257 struct btrfs_ioctl_search_args_v2 args;
2258 struct inode *inode;
2259 int ret;
2260 size_t buf_size;
2261 const size_t buf_limit = SZ_16M;
2262
2263 if (!capable(CAP_SYS_ADMIN))
2264 return -EPERM;
2265
2266 /* copy search header and buffer size */
2267 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2268 if (copy_from_user(&args, uarg, sizeof(args)))
2269 return -EFAULT;
2270
2271 buf_size = args.buf_size;
2272
2273 /* limit result size to 16MB */
2274 if (buf_size > buf_limit)
2275 buf_size = buf_limit;
2276
2277 inode = file_inode(file);
2278 ret = search_ioctl(inode, &args.key, &buf_size,
2279 (char __user *)(&uarg->buf[0]));
2280 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2281 ret = -EFAULT;
2282 else if (ret == -EOVERFLOW &&
2283 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2284 ret = -EFAULT;
2285
2286 return ret;
2287 }
2288
2289 /*
2290 * Search INODE_REFs to identify path name of 'dirid' directory
2291 * in a 'tree_id' tree. and sets path name to 'name'.
2292 */
2293 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2294 u64 tree_id, u64 dirid, char *name)
2295 {
2296 struct btrfs_root *root;
2297 struct btrfs_key key;
2298 char *ptr;
2299 int ret = -1;
2300 int slot;
2301 int len;
2302 int total_len = 0;
2303 struct btrfs_inode_ref *iref;
2304 struct extent_buffer *l;
2305 struct btrfs_path *path;
2306
2307 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2308 name[0]='\0';
2309 return 0;
2310 }
2311
2312 path = btrfs_alloc_path();
2313 if (!path)
2314 return -ENOMEM;
2315
2316 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2317
2318 key.objectid = tree_id;
2319 key.type = BTRFS_ROOT_ITEM_KEY;
2320 key.offset = (u64)-1;
2321 root = btrfs_read_fs_root_no_name(info, &key);
2322 if (IS_ERR(root)) {
2323 ret = PTR_ERR(root);
2324 goto out;
2325 }
2326
2327 key.objectid = dirid;
2328 key.type = BTRFS_INODE_REF_KEY;
2329 key.offset = (u64)-1;
2330
2331 while (1) {
2332 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2333 if (ret < 0)
2334 goto out;
2335 else if (ret > 0) {
2336 ret = btrfs_previous_item(root, path, dirid,
2337 BTRFS_INODE_REF_KEY);
2338 if (ret < 0)
2339 goto out;
2340 else if (ret > 0) {
2341 ret = -ENOENT;
2342 goto out;
2343 }
2344 }
2345
2346 l = path->nodes[0];
2347 slot = path->slots[0];
2348 btrfs_item_key_to_cpu(l, &key, slot);
2349
2350 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2351 len = btrfs_inode_ref_name_len(l, iref);
2352 ptr -= len + 1;
2353 total_len += len + 1;
2354 if (ptr < name) {
2355 ret = -ENAMETOOLONG;
2356 goto out;
2357 }
2358
2359 *(ptr + len) = '/';
2360 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2361
2362 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2363 break;
2364
2365 btrfs_release_path(path);
2366 key.objectid = key.offset;
2367 key.offset = (u64)-1;
2368 dirid = key.objectid;
2369 }
2370 memmove(name, ptr, total_len);
2371 name[total_len] = '\0';
2372 ret = 0;
2373 out:
2374 btrfs_free_path(path);
2375 return ret;
2376 }
2377
2378 static int btrfs_search_path_in_tree_user(struct inode *inode,
2379 struct btrfs_ioctl_ino_lookup_user_args *args)
2380 {
2381 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2382 struct super_block *sb = inode->i_sb;
2383 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2384 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2385 u64 dirid = args->dirid;
2386 unsigned long item_off;
2387 unsigned long item_len;
2388 struct btrfs_inode_ref *iref;
2389 struct btrfs_root_ref *rref;
2390 struct btrfs_root *root;
2391 struct btrfs_path *path;
2392 struct btrfs_key key, key2;
2393 struct extent_buffer *leaf;
2394 struct inode *temp_inode;
2395 char *ptr;
2396 int slot;
2397 int len;
2398 int total_len = 0;
2399 int ret;
2400
2401 path = btrfs_alloc_path();
2402 if (!path)
2403 return -ENOMEM;
2404
2405 /*
2406 * If the bottom subvolume does not exist directly under upper_limit,
2407 * construct the path in from the bottom up.
2408 */
2409 if (dirid != upper_limit.objectid) {
2410 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2411
2412 key.objectid = treeid;
2413 key.type = BTRFS_ROOT_ITEM_KEY;
2414 key.offset = (u64)-1;
2415 root = btrfs_read_fs_root_no_name(fs_info, &key);
2416 if (IS_ERR(root)) {
2417 ret = PTR_ERR(root);
2418 goto out;
2419 }
2420
2421 key.objectid = dirid;
2422 key.type = BTRFS_INODE_REF_KEY;
2423 key.offset = (u64)-1;
2424 while (1) {
2425 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2426 if (ret < 0) {
2427 goto out;
2428 } else if (ret > 0) {
2429 ret = btrfs_previous_item(root, path, dirid,
2430 BTRFS_INODE_REF_KEY);
2431 if (ret < 0) {
2432 goto out;
2433 } else if (ret > 0) {
2434 ret = -ENOENT;
2435 goto out;
2436 }
2437 }
2438
2439 leaf = path->nodes[0];
2440 slot = path->slots[0];
2441 btrfs_item_key_to_cpu(leaf, &key, slot);
2442
2443 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2444 len = btrfs_inode_ref_name_len(leaf, iref);
2445 ptr -= len + 1;
2446 total_len += len + 1;
2447 if (ptr < args->path) {
2448 ret = -ENAMETOOLONG;
2449 goto out;
2450 }
2451
2452 *(ptr + len) = '/';
2453 read_extent_buffer(leaf, ptr,
2454 (unsigned long)(iref + 1), len);
2455
2456 /* Check the read+exec permission of this directory */
2457 ret = btrfs_previous_item(root, path, dirid,
2458 BTRFS_INODE_ITEM_KEY);
2459 if (ret < 0) {
2460 goto out;
2461 } else if (ret > 0) {
2462 ret = -ENOENT;
2463 goto out;
2464 }
2465
2466 leaf = path->nodes[0];
2467 slot = path->slots[0];
2468 btrfs_item_key_to_cpu(leaf, &key2, slot);
2469 if (key2.objectid != dirid) {
2470 ret = -ENOENT;
2471 goto out;
2472 }
2473
2474 temp_inode = btrfs_iget(sb, &key2, root, NULL);
2475 if (IS_ERR(temp_inode)) {
2476 ret = PTR_ERR(temp_inode);
2477 goto out;
2478 }
2479 ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC);
2480 iput(temp_inode);
2481 if (ret) {
2482 ret = -EACCES;
2483 goto out;
2484 }
2485
2486 if (key.offset == upper_limit.objectid)
2487 break;
2488 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2489 ret = -EACCES;
2490 goto out;
2491 }
2492
2493 btrfs_release_path(path);
2494 key.objectid = key.offset;
2495 key.offset = (u64)-1;
2496 dirid = key.objectid;
2497 }
2498
2499 memmove(args->path, ptr, total_len);
2500 args->path[total_len] = '\0';
2501 btrfs_release_path(path);
2502 }
2503
2504 /* Get the bottom subvolume's name from ROOT_REF */
2505 root = fs_info->tree_root;
2506 key.objectid = treeid;
2507 key.type = BTRFS_ROOT_REF_KEY;
2508 key.offset = args->treeid;
2509 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2510 if (ret < 0) {
2511 goto out;
2512 } else if (ret > 0) {
2513 ret = -ENOENT;
2514 goto out;
2515 }
2516
2517 leaf = path->nodes[0];
2518 slot = path->slots[0];
2519 btrfs_item_key_to_cpu(leaf, &key, slot);
2520
2521 item_off = btrfs_item_ptr_offset(leaf, slot);
2522 item_len = btrfs_item_size_nr(leaf, slot);
2523 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2524 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2525 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2526 ret = -EINVAL;
2527 goto out;
2528 }
2529
2530 /* Copy subvolume's name */
2531 item_off += sizeof(struct btrfs_root_ref);
2532 item_len -= sizeof(struct btrfs_root_ref);
2533 read_extent_buffer(leaf, args->name, item_off, item_len);
2534 args->name[item_len] = 0;
2535
2536 out:
2537 btrfs_free_path(path);
2538 return ret;
2539 }
2540
2541 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2542 void __user *argp)
2543 {
2544 struct btrfs_ioctl_ino_lookup_args *args;
2545 struct inode *inode;
2546 int ret = 0;
2547
2548 args = memdup_user(argp, sizeof(*args));
2549 if (IS_ERR(args))
2550 return PTR_ERR(args);
2551
2552 inode = file_inode(file);
2553
2554 /*
2555 * Unprivileged query to obtain the containing subvolume root id. The
2556 * path is reset so it's consistent with btrfs_search_path_in_tree.
2557 */
2558 if (args->treeid == 0)
2559 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2560
2561 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2562 args->name[0] = 0;
2563 goto out;
2564 }
2565
2566 if (!capable(CAP_SYS_ADMIN)) {
2567 ret = -EPERM;
2568 goto out;
2569 }
2570
2571 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2572 args->treeid, args->objectid,
2573 args->name);
2574
2575 out:
2576 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2577 ret = -EFAULT;
2578
2579 kfree(args);
2580 return ret;
2581 }
2582
2583 /*
2584 * Version of ino_lookup ioctl (unprivileged)
2585 *
2586 * The main differences from ino_lookup ioctl are:
2587 *
2588 * 1. Read + Exec permission will be checked using inode_permission() during
2589 * path construction. -EACCES will be returned in case of failure.
2590 * 2. Path construction will be stopped at the inode number which corresponds
2591 * to the fd with which this ioctl is called. If constructed path does not
2592 * exist under fd's inode, -EACCES will be returned.
2593 * 3. The name of bottom subvolume is also searched and filled.
2594 */
2595 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2596 {
2597 struct btrfs_ioctl_ino_lookup_user_args *args;
2598 struct inode *inode;
2599 int ret;
2600
2601 args = memdup_user(argp, sizeof(*args));
2602 if (IS_ERR(args))
2603 return PTR_ERR(args);
2604
2605 inode = file_inode(file);
2606
2607 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2608 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2609 /*
2610 * The subvolume does not exist under fd with which this is
2611 * called
2612 */
2613 kfree(args);
2614 return -EACCES;
2615 }
2616
2617 ret = btrfs_search_path_in_tree_user(inode, args);
2618
2619 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2620 ret = -EFAULT;
2621
2622 kfree(args);
2623 return ret;
2624 }
2625
2626 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2627 static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2628 {
2629 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2630 struct btrfs_fs_info *fs_info;
2631 struct btrfs_root *root;
2632 struct btrfs_path *path;
2633 struct btrfs_key key;
2634 struct btrfs_root_item *root_item;
2635 struct btrfs_root_ref *rref;
2636 struct extent_buffer *leaf;
2637 unsigned long item_off;
2638 unsigned long item_len;
2639 struct inode *inode;
2640 int slot;
2641 int ret = 0;
2642
2643 path = btrfs_alloc_path();
2644 if (!path)
2645 return -ENOMEM;
2646
2647 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2648 if (!subvol_info) {
2649 btrfs_free_path(path);
2650 return -ENOMEM;
2651 }
2652
2653 inode = file_inode(file);
2654 fs_info = BTRFS_I(inode)->root->fs_info;
2655
2656 /* Get root_item of inode's subvolume */
2657 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2658 key.type = BTRFS_ROOT_ITEM_KEY;
2659 key.offset = (u64)-1;
2660 root = btrfs_read_fs_root_no_name(fs_info, &key);
2661 if (IS_ERR(root)) {
2662 ret = PTR_ERR(root);
2663 goto out;
2664 }
2665 root_item = &root->root_item;
2666
2667 subvol_info->treeid = key.objectid;
2668
2669 subvol_info->generation = btrfs_root_generation(root_item);
2670 subvol_info->flags = btrfs_root_flags(root_item);
2671
2672 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2673 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2674 BTRFS_UUID_SIZE);
2675 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2676 BTRFS_UUID_SIZE);
2677
2678 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2679 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2680 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2681
2682 subvol_info->otransid = btrfs_root_otransid(root_item);
2683 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2684 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2685
2686 subvol_info->stransid = btrfs_root_stransid(root_item);
2687 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2688 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2689
2690 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2691 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2692 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2693
2694 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2695 /* Search root tree for ROOT_BACKREF of this subvolume */
2696 root = fs_info->tree_root;
2697
2698 key.type = BTRFS_ROOT_BACKREF_KEY;
2699 key.offset = 0;
2700 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2701 if (ret < 0) {
2702 goto out;
2703 } else if (path->slots[0] >=
2704 btrfs_header_nritems(path->nodes[0])) {
2705 ret = btrfs_next_leaf(root, path);
2706 if (ret < 0) {
2707 goto out;
2708 } else if (ret > 0) {
2709 ret = -EUCLEAN;
2710 goto out;
2711 }
2712 }
2713
2714 leaf = path->nodes[0];
2715 slot = path->slots[0];
2716 btrfs_item_key_to_cpu(leaf, &key, slot);
2717 if (key.objectid == subvol_info->treeid &&
2718 key.type == BTRFS_ROOT_BACKREF_KEY) {
2719 subvol_info->parent_id = key.offset;
2720
2721 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2722 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2723
2724 item_off = btrfs_item_ptr_offset(leaf, slot)
2725 + sizeof(struct btrfs_root_ref);
2726 item_len = btrfs_item_size_nr(leaf, slot)
2727 - sizeof(struct btrfs_root_ref);
2728 read_extent_buffer(leaf, subvol_info->name,
2729 item_off, item_len);
2730 } else {
2731 ret = -ENOENT;
2732 goto out;
2733 }
2734 }
2735
2736 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2737 ret = -EFAULT;
2738
2739 out:
2740 btrfs_free_path(path);
2741 kzfree(subvol_info);
2742 return ret;
2743 }
2744
2745 /*
2746 * Return ROOT_REF information of the subvolume containing this inode
2747 * except the subvolume name.
2748 */
2749 static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2750 {
2751 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2752 struct btrfs_root_ref *rref;
2753 struct btrfs_root *root;
2754 struct btrfs_path *path;
2755 struct btrfs_key key;
2756 struct extent_buffer *leaf;
2757 struct inode *inode;
2758 u64 objectid;
2759 int slot;
2760 int ret;
2761 u8 found;
2762
2763 path = btrfs_alloc_path();
2764 if (!path)
2765 return -ENOMEM;
2766
2767 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2768 if (IS_ERR(rootrefs)) {
2769 btrfs_free_path(path);
2770 return PTR_ERR(rootrefs);
2771 }
2772
2773 inode = file_inode(file);
2774 root = BTRFS_I(inode)->root->fs_info->tree_root;
2775 objectid = BTRFS_I(inode)->root->root_key.objectid;
2776
2777 key.objectid = objectid;
2778 key.type = BTRFS_ROOT_REF_KEY;
2779 key.offset = rootrefs->min_treeid;
2780 found = 0;
2781
2782 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2783 if (ret < 0) {
2784 goto out;
2785 } else if (path->slots[0] >=
2786 btrfs_header_nritems(path->nodes[0])) {
2787 ret = btrfs_next_leaf(root, path);
2788 if (ret < 0) {
2789 goto out;
2790 } else if (ret > 0) {
2791 ret = -EUCLEAN;
2792 goto out;
2793 }
2794 }
2795 while (1) {
2796 leaf = path->nodes[0];
2797 slot = path->slots[0];
2798
2799 btrfs_item_key_to_cpu(leaf, &key, slot);
2800 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2801 ret = 0;
2802 goto out;
2803 }
2804
2805 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2806 ret = -EOVERFLOW;
2807 goto out;
2808 }
2809
2810 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2811 rootrefs->rootref[found].treeid = key.offset;
2812 rootrefs->rootref[found].dirid =
2813 btrfs_root_ref_dirid(leaf, rref);
2814 found++;
2815
2816 ret = btrfs_next_item(root, path);
2817 if (ret < 0) {
2818 goto out;
2819 } else if (ret > 0) {
2820 ret = -EUCLEAN;
2821 goto out;
2822 }
2823 }
2824
2825 out:
2826 if (!ret || ret == -EOVERFLOW) {
2827 rootrefs->num_items = found;
2828 /* update min_treeid for next search */
2829 if (found)
2830 rootrefs->min_treeid =
2831 rootrefs->rootref[found - 1].treeid + 1;
2832 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2833 ret = -EFAULT;
2834 }
2835
2836 kfree(rootrefs);
2837 btrfs_free_path(path);
2838
2839 return ret;
2840 }
2841
2842 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2843 void __user *arg)
2844 {
2845 struct dentry *parent = file->f_path.dentry;
2846 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2847 struct dentry *dentry;
2848 struct inode *dir = d_inode(parent);
2849 struct inode *inode;
2850 struct btrfs_root *root = BTRFS_I(dir)->root;
2851 struct btrfs_root *dest = NULL;
2852 struct btrfs_ioctl_vol_args *vol_args;
2853 int namelen;
2854 int err = 0;
2855
2856 if (!S_ISDIR(dir->i_mode))
2857 return -ENOTDIR;
2858
2859 vol_args = memdup_user(arg, sizeof(*vol_args));
2860 if (IS_ERR(vol_args))
2861 return PTR_ERR(vol_args);
2862
2863 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2864 namelen = strlen(vol_args->name);
2865 if (strchr(vol_args->name, '/') ||
2866 strncmp(vol_args->name, "..", namelen) == 0) {
2867 err = -EINVAL;
2868 goto out;
2869 }
2870
2871 err = mnt_want_write_file(file);
2872 if (err)
2873 goto out;
2874
2875
2876 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2877 if (err == -EINTR)
2878 goto out_drop_write;
2879 dentry = lookup_one_len(vol_args->name, parent, namelen);
2880 if (IS_ERR(dentry)) {
2881 err = PTR_ERR(dentry);
2882 goto out_unlock_dir;
2883 }
2884
2885 if (d_really_is_negative(dentry)) {
2886 err = -ENOENT;
2887 goto out_dput;
2888 }
2889
2890 inode = d_inode(dentry);
2891 dest = BTRFS_I(inode)->root;
2892 if (!capable(CAP_SYS_ADMIN)) {
2893 /*
2894 * Regular user. Only allow this with a special mount
2895 * option, when the user has write+exec access to the
2896 * subvol root, and when rmdir(2) would have been
2897 * allowed.
2898 *
2899 * Note that this is _not_ check that the subvol is
2900 * empty or doesn't contain data that we wouldn't
2901 * otherwise be able to delete.
2902 *
2903 * Users who want to delete empty subvols should try
2904 * rmdir(2).
2905 */
2906 err = -EPERM;
2907 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2908 goto out_dput;
2909
2910 /*
2911 * Do not allow deletion if the parent dir is the same
2912 * as the dir to be deleted. That means the ioctl
2913 * must be called on the dentry referencing the root
2914 * of the subvol, not a random directory contained
2915 * within it.
2916 */
2917 err = -EINVAL;
2918 if (root == dest)
2919 goto out_dput;
2920
2921 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2922 if (err)
2923 goto out_dput;
2924 }
2925
2926 /* check if subvolume may be deleted by a user */
2927 err = btrfs_may_delete(dir, dentry, 1);
2928 if (err)
2929 goto out_dput;
2930
2931 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2932 err = -EINVAL;
2933 goto out_dput;
2934 }
2935
2936 inode_lock(inode);
2937 err = btrfs_delete_subvolume(dir, dentry);
2938 inode_unlock(inode);
2939 if (!err)
2940 d_delete(dentry);
2941
2942 out_dput:
2943 dput(dentry);
2944 out_unlock_dir:
2945 inode_unlock(dir);
2946 out_drop_write:
2947 mnt_drop_write_file(file);
2948 out:
2949 kfree(vol_args);
2950 return err;
2951 }
2952
2953 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2954 {
2955 struct inode *inode = file_inode(file);
2956 struct btrfs_root *root = BTRFS_I(inode)->root;
2957 struct btrfs_ioctl_defrag_range_args *range;
2958 int ret;
2959
2960 ret = mnt_want_write_file(file);
2961 if (ret)
2962 return ret;
2963
2964 if (btrfs_root_readonly(root)) {
2965 ret = -EROFS;
2966 goto out;
2967 }
2968
2969 switch (inode->i_mode & S_IFMT) {
2970 case S_IFDIR:
2971 if (!capable(CAP_SYS_ADMIN)) {
2972 ret = -EPERM;
2973 goto out;
2974 }
2975 ret = btrfs_defrag_root(root);
2976 break;
2977 case S_IFREG:
2978 /*
2979 * Note that this does not check the file descriptor for write
2980 * access. This prevents defragmenting executables that are
2981 * running and allows defrag on files open in read-only mode.
2982 */
2983 if (!capable(CAP_SYS_ADMIN) &&
2984 inode_permission(inode, MAY_WRITE)) {
2985 ret = -EPERM;
2986 goto out;
2987 }
2988
2989 range = kzalloc(sizeof(*range), GFP_KERNEL);
2990 if (!range) {
2991 ret = -ENOMEM;
2992 goto out;
2993 }
2994
2995 if (argp) {
2996 if (copy_from_user(range, argp,
2997 sizeof(*range))) {
2998 ret = -EFAULT;
2999 kfree(range);
3000 goto out;
3001 }
3002 /* compression requires us to start the IO */
3003 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3004 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
3005 range->extent_thresh = (u32)-1;
3006 }
3007 } else {
3008 /* the rest are all set to zero by kzalloc */
3009 range->len = (u64)-1;
3010 }
3011 ret = btrfs_defrag_file(file_inode(file), file,
3012 range, BTRFS_OLDEST_GENERATION, 0);
3013 if (ret > 0)
3014 ret = 0;
3015 kfree(range);
3016 break;
3017 default:
3018 ret = -EINVAL;
3019 }
3020 out:
3021 mnt_drop_write_file(file);
3022 return ret;
3023 }
3024
3025 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
3026 {
3027 struct btrfs_ioctl_vol_args *vol_args;
3028 int ret;
3029
3030 if (!capable(CAP_SYS_ADMIN))
3031 return -EPERM;
3032
3033 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
3034 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3035
3036 vol_args = memdup_user(arg, sizeof(*vol_args));
3037 if (IS_ERR(vol_args)) {
3038 ret = PTR_ERR(vol_args);
3039 goto out;
3040 }
3041
3042 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3043 ret = btrfs_init_new_device(fs_info, vol_args->name);
3044
3045 if (!ret)
3046 btrfs_info(fs_info, "disk added %s", vol_args->name);
3047
3048 kfree(vol_args);
3049 out:
3050 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3051 return ret;
3052 }
3053
3054 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
3055 {
3056 struct inode *inode = file_inode(file);
3057 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3058 struct btrfs_ioctl_vol_args_v2 *vol_args;
3059 int ret;
3060
3061 if (!capable(CAP_SYS_ADMIN))
3062 return -EPERM;
3063
3064 ret = mnt_want_write_file(file);
3065 if (ret)
3066 return ret;
3067
3068 vol_args = memdup_user(arg, sizeof(*vol_args));
3069 if (IS_ERR(vol_args)) {
3070 ret = PTR_ERR(vol_args);
3071 goto err_drop;
3072 }
3073
3074 /* Check for compatibility reject unknown flags */
3075 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) {
3076 ret = -EOPNOTSUPP;
3077 goto out;
3078 }
3079
3080 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3081 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3082 goto out;
3083 }
3084
3085 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
3086 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
3087 } else {
3088 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3089 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3090 }
3091 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3092
3093 if (!ret) {
3094 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3095 btrfs_info(fs_info, "device deleted: id %llu",
3096 vol_args->devid);
3097 else
3098 btrfs_info(fs_info, "device deleted: %s",
3099 vol_args->name);
3100 }
3101 out:
3102 kfree(vol_args);
3103 err_drop:
3104 mnt_drop_write_file(file);
3105 return ret;
3106 }
3107
3108 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3109 {
3110 struct inode *inode = file_inode(file);
3111 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3112 struct btrfs_ioctl_vol_args *vol_args;
3113 int ret;
3114
3115 if (!capable(CAP_SYS_ADMIN))
3116 return -EPERM;
3117
3118 ret = mnt_want_write_file(file);
3119 if (ret)
3120 return ret;
3121
3122 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3123 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3124 goto out_drop_write;
3125 }
3126
3127 vol_args = memdup_user(arg, sizeof(*vol_args));
3128 if (IS_ERR(vol_args)) {
3129 ret = PTR_ERR(vol_args);
3130 goto out;
3131 }
3132
3133 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3134 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3135
3136 if (!ret)
3137 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3138 kfree(vol_args);
3139 out:
3140 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3141 out_drop_write:
3142 mnt_drop_write_file(file);
3143
3144 return ret;
3145 }
3146
3147 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3148 void __user *arg)
3149 {
3150 struct btrfs_ioctl_fs_info_args *fi_args;
3151 struct btrfs_device *device;
3152 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3153 int ret = 0;
3154
3155 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
3156 if (!fi_args)
3157 return -ENOMEM;
3158
3159 rcu_read_lock();
3160 fi_args->num_devices = fs_devices->num_devices;
3161
3162 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3163 if (device->devid > fi_args->max_id)
3164 fi_args->max_id = device->devid;
3165 }
3166 rcu_read_unlock();
3167
3168 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
3169 fi_args->nodesize = fs_info->nodesize;
3170 fi_args->sectorsize = fs_info->sectorsize;
3171 fi_args->clone_alignment = fs_info->sectorsize;
3172
3173 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3174 ret = -EFAULT;
3175
3176 kfree(fi_args);
3177 return ret;
3178 }
3179
3180 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3181 void __user *arg)
3182 {
3183 struct btrfs_ioctl_dev_info_args *di_args;
3184 struct btrfs_device *dev;
3185 int ret = 0;
3186 char *s_uuid = NULL;
3187
3188 di_args = memdup_user(arg, sizeof(*di_args));
3189 if (IS_ERR(di_args))
3190 return PTR_ERR(di_args);
3191
3192 if (!btrfs_is_empty_uuid(di_args->uuid))
3193 s_uuid = di_args->uuid;
3194
3195 rcu_read_lock();
3196 dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid,
3197 NULL, true);
3198
3199 if (!dev) {
3200 ret = -ENODEV;
3201 goto out;
3202 }
3203
3204 di_args->devid = dev->devid;
3205 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3206 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3207 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3208 if (dev->name) {
3209 strncpy(di_args->path, rcu_str_deref(dev->name),
3210 sizeof(di_args->path) - 1);
3211 di_args->path[sizeof(di_args->path) - 1] = 0;
3212 } else {
3213 di_args->path[0] = '\0';
3214 }
3215
3216 out:
3217 rcu_read_unlock();
3218 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3219 ret = -EFAULT;
3220
3221 kfree(di_args);
3222 return ret;
3223 }
3224
3225 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
3226 struct inode *inode2, u64 loff2, u64 len)
3227 {
3228 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
3229 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
3230 }
3231
3232 static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
3233 struct inode *inode2, u64 loff2, u64 len)
3234 {
3235 if (inode1 < inode2) {
3236 swap(inode1, inode2);
3237 swap(loff1, loff2);
3238 } else if (inode1 == inode2 && loff2 < loff1) {
3239 swap(loff1, loff2);
3240 }
3241 lock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
3242 lock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
3243 }
3244
3245 static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 len,
3246 struct inode *dst, u64 dst_loff)
3247 {
3248 int ret;
3249
3250 /*
3251 * Lock destination range to serialize with concurrent readpages() and
3252 * source range to serialize with relocation.
3253 */
3254 btrfs_double_extent_lock(src, loff, dst, dst_loff, len);
3255 ret = btrfs_clone(src, dst, loff, len, len, dst_loff, 1);
3256 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3257
3258 return ret;
3259 }
3260
3261 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3262
3263 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3264 struct inode *dst, u64 dst_loff)
3265 {
3266 int ret;
3267 u64 i, tail_len, chunk_count;
3268 struct btrfs_root *root_dst = BTRFS_I(dst)->root;
3269
3270 spin_lock(&root_dst->root_item_lock);
3271 if (root_dst->send_in_progress) {
3272 btrfs_warn_rl(root_dst->fs_info,
3273 "cannot deduplicate to root %llu while send operations are using it (%d in progress)",
3274 root_dst->root_key.objectid,
3275 root_dst->send_in_progress);
3276 spin_unlock(&root_dst->root_item_lock);
3277 return -EAGAIN;
3278 }
3279 root_dst->dedupe_in_progress++;
3280 spin_unlock(&root_dst->root_item_lock);
3281
3282 tail_len = olen % BTRFS_MAX_DEDUPE_LEN;
3283 chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN);
3284
3285 for (i = 0; i < chunk_count; i++) {
3286 ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN,
3287 dst, dst_loff);
3288 if (ret)
3289 goto out;
3290
3291 loff += BTRFS_MAX_DEDUPE_LEN;
3292 dst_loff += BTRFS_MAX_DEDUPE_LEN;
3293 }
3294
3295 if (tail_len > 0)
3296 ret = btrfs_extent_same_range(src, loff, tail_len, dst,
3297 dst_loff);
3298 out:
3299 spin_lock(&root_dst->root_item_lock);
3300 root_dst->dedupe_in_progress--;
3301 spin_unlock(&root_dst->root_item_lock);
3302
3303 return ret;
3304 }
3305
3306 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3307 struct inode *inode,
3308 u64 endoff,
3309 const u64 destoff,
3310 const u64 olen,
3311 int no_time_update)
3312 {
3313 struct btrfs_root *root = BTRFS_I(inode)->root;
3314 int ret;
3315
3316 inode_inc_iversion(inode);
3317 if (!no_time_update)
3318 inode->i_mtime = inode->i_ctime = current_time(inode);
3319 /*
3320 * We round up to the block size at eof when determining which
3321 * extents to clone above, but shouldn't round up the file size.
3322 */
3323 if (endoff > destoff + olen)
3324 endoff = destoff + olen;
3325 if (endoff > inode->i_size)
3326 btrfs_i_size_write(BTRFS_I(inode), endoff);
3327
3328 ret = btrfs_update_inode(trans, root, inode);
3329 if (ret) {
3330 btrfs_abort_transaction(trans, ret);
3331 btrfs_end_transaction(trans);
3332 goto out;
3333 }
3334 ret = btrfs_end_transaction(trans);
3335 out:
3336 return ret;
3337 }
3338
3339 static void clone_update_extent_map(struct btrfs_inode *inode,
3340 const struct btrfs_trans_handle *trans,
3341 const struct btrfs_path *path,
3342 const u64 hole_offset,
3343 const u64 hole_len)
3344 {
3345 struct extent_map_tree *em_tree = &inode->extent_tree;
3346 struct extent_map *em;
3347 int ret;
3348
3349 em = alloc_extent_map();
3350 if (!em) {
3351 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3352 return;
3353 }
3354
3355 if (path) {
3356 struct btrfs_file_extent_item *fi;
3357
3358 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3359 struct btrfs_file_extent_item);
3360 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3361 em->generation = -1;
3362 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3363 BTRFS_FILE_EXTENT_INLINE)
3364 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3365 &inode->runtime_flags);
3366 } else {
3367 em->start = hole_offset;
3368 em->len = hole_len;
3369 em->ram_bytes = em->len;
3370 em->orig_start = hole_offset;
3371 em->block_start = EXTENT_MAP_HOLE;
3372 em->block_len = 0;
3373 em->orig_block_len = 0;
3374 em->compress_type = BTRFS_COMPRESS_NONE;
3375 em->generation = trans->transid;
3376 }
3377
3378 while (1) {
3379 write_lock(&em_tree->lock);
3380 ret = add_extent_mapping(em_tree, em, 1);
3381 write_unlock(&em_tree->lock);
3382 if (ret != -EEXIST) {
3383 free_extent_map(em);
3384 break;
3385 }
3386 btrfs_drop_extent_cache(inode, em->start,
3387 em->start + em->len - 1, 0);
3388 }
3389
3390 if (ret)
3391 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3392 }
3393
3394 /*
3395 * Make sure we do not end up inserting an inline extent into a file that has
3396 * already other (non-inline) extents. If a file has an inline extent it can
3397 * not have any other extents and the (single) inline extent must start at the
3398 * file offset 0. Failing to respect these rules will lead to file corruption,
3399 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3400 *
3401 * We can have extents that have been already written to disk or we can have
3402 * dirty ranges still in delalloc, in which case the extent maps and items are
3403 * created only when we run delalloc, and the delalloc ranges might fall outside
3404 * the range we are currently locking in the inode's io tree. So we check the
3405 * inode's i_size because of that (i_size updates are done while holding the
3406 * i_mutex, which we are holding here).
3407 * We also check to see if the inode has a size not greater than "datal" but has
3408 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3409 * protected against such concurrent fallocate calls by the i_mutex).
3410 *
3411 * If the file has no extents but a size greater than datal, do not allow the
3412 * copy because we would need turn the inline extent into a non-inline one (even
3413 * with NO_HOLES enabled). If we find our destination inode only has one inline
3414 * extent, just overwrite it with the source inline extent if its size is less
3415 * than the source extent's size, or we could copy the source inline extent's
3416 * data into the destination inode's inline extent if the later is greater then
3417 * the former.
3418 */
3419 static int clone_copy_inline_extent(struct inode *dst,
3420 struct btrfs_trans_handle *trans,
3421 struct btrfs_path *path,
3422 struct btrfs_key *new_key,
3423 const u64 drop_start,
3424 const u64 datal,
3425 const u64 skip,
3426 const u64 size,
3427 char *inline_data)
3428 {
3429 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
3430 struct btrfs_root *root = BTRFS_I(dst)->root;
3431 const u64 aligned_end = ALIGN(new_key->offset + datal,
3432 fs_info->sectorsize);
3433 int ret;
3434 struct btrfs_key key;
3435
3436 if (new_key->offset > 0)
3437 return -EOPNOTSUPP;
3438
3439 key.objectid = btrfs_ino(BTRFS_I(dst));
3440 key.type = BTRFS_EXTENT_DATA_KEY;
3441 key.offset = 0;
3442 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3443 if (ret < 0) {
3444 return ret;
3445 } else if (ret > 0) {
3446 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3447 ret = btrfs_next_leaf(root, path);
3448 if (ret < 0)
3449 return ret;
3450 else if (ret > 0)
3451 goto copy_inline_extent;
3452 }
3453 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3454 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3455 key.type == BTRFS_EXTENT_DATA_KEY) {
3456 ASSERT(key.offset > 0);
3457 return -EOPNOTSUPP;
3458 }
3459 } else if (i_size_read(dst) <= datal) {
3460 struct btrfs_file_extent_item *ei;
3461 u64 ext_len;
3462
3463 /*
3464 * If the file size is <= datal, make sure there are no other
3465 * extents following (can happen do to an fallocate call with
3466 * the flag FALLOC_FL_KEEP_SIZE).
3467 */
3468 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3469 struct btrfs_file_extent_item);
3470 /*
3471 * If it's an inline extent, it can not have other extents
3472 * following it.
3473 */
3474 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3475 BTRFS_FILE_EXTENT_INLINE)
3476 goto copy_inline_extent;
3477
3478 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3479 if (ext_len > aligned_end)
3480 return -EOPNOTSUPP;
3481
3482 ret = btrfs_next_item(root, path);
3483 if (ret < 0) {
3484 return ret;
3485 } else if (ret == 0) {
3486 btrfs_item_key_to_cpu(path->nodes[0], &key,
3487 path->slots[0]);
3488 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3489 key.type == BTRFS_EXTENT_DATA_KEY)
3490 return -EOPNOTSUPP;
3491 }
3492 }
3493
3494 copy_inline_extent:
3495 /*
3496 * We have no extent items, or we have an extent at offset 0 which may
3497 * or may not be inlined. All these cases are dealt the same way.
3498 */
3499 if (i_size_read(dst) > datal) {
3500 /*
3501 * If the destination inode has an inline extent...
3502 * This would require copying the data from the source inline
3503 * extent into the beginning of the destination's inline extent.
3504 * But this is really complex, both extents can be compressed
3505 * or just one of them, which would require decompressing and
3506 * re-compressing data (which could increase the new compressed
3507 * size, not allowing the compressed data to fit anymore in an
3508 * inline extent).
3509 * So just don't support this case for now (it should be rare,
3510 * we are not really saving space when cloning inline extents).
3511 */
3512 return -EOPNOTSUPP;
3513 }
3514
3515 btrfs_release_path(path);
3516 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3517 if (ret)
3518 return ret;
3519 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3520 if (ret)
3521 return ret;
3522
3523 if (skip) {
3524 const u32 start = btrfs_file_extent_calc_inline_size(0);
3525
3526 memmove(inline_data + start, inline_data + start + skip, datal);
3527 }
3528
3529 write_extent_buffer(path->nodes[0], inline_data,
3530 btrfs_item_ptr_offset(path->nodes[0],
3531 path->slots[0]),
3532 size);
3533 inode_add_bytes(dst, datal);
3534
3535 return 0;
3536 }
3537
3538 /**
3539 * btrfs_clone() - clone a range from inode file to another
3540 *
3541 * @src: Inode to clone from
3542 * @inode: Inode to clone to
3543 * @off: Offset within source to start clone from
3544 * @olen: Original length, passed by user, of range to clone
3545 * @olen_aligned: Block-aligned value of olen
3546 * @destoff: Offset within @inode to start clone
3547 * @no_time_update: Whether to update mtime/ctime on the target inode
3548 */
3549 static int btrfs_clone(struct inode *src, struct inode *inode,
3550 const u64 off, const u64 olen, const u64 olen_aligned,
3551 const u64 destoff, int no_time_update)
3552 {
3553 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3554 struct btrfs_root *root = BTRFS_I(inode)->root;
3555 struct btrfs_path *path = NULL;
3556 struct extent_buffer *leaf;
3557 struct btrfs_trans_handle *trans;
3558 char *buf = NULL;
3559 struct btrfs_key key;
3560 u32 nritems;
3561 int slot;
3562 int ret;
3563 const u64 len = olen_aligned;
3564 u64 last_dest_end = destoff;
3565
3566 ret = -ENOMEM;
3567 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3568 if (!buf)
3569 return ret;
3570
3571 path = btrfs_alloc_path();
3572 if (!path) {
3573 kvfree(buf);
3574 return ret;
3575 }
3576
3577 path->reada = READA_FORWARD;
3578 /* clone data */
3579 key.objectid = btrfs_ino(BTRFS_I(src));
3580 key.type = BTRFS_EXTENT_DATA_KEY;
3581 key.offset = off;
3582
3583 while (1) {
3584 u64 next_key_min_offset = key.offset + 1;
3585
3586 /*
3587 * note the key will change type as we walk through the
3588 * tree.
3589 */
3590 path->leave_spinning = 1;
3591 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3592 0, 0);
3593 if (ret < 0)
3594 goto out;
3595 /*
3596 * First search, if no extent item that starts at offset off was
3597 * found but the previous item is an extent item, it's possible
3598 * it might overlap our target range, therefore process it.
3599 */
3600 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3601 btrfs_item_key_to_cpu(path->nodes[0], &key,
3602 path->slots[0] - 1);
3603 if (key.type == BTRFS_EXTENT_DATA_KEY)
3604 path->slots[0]--;
3605 }
3606
3607 nritems = btrfs_header_nritems(path->nodes[0]);
3608 process_slot:
3609 if (path->slots[0] >= nritems) {
3610 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3611 if (ret < 0)
3612 goto out;
3613 if (ret > 0)
3614 break;
3615 nritems = btrfs_header_nritems(path->nodes[0]);
3616 }
3617 leaf = path->nodes[0];
3618 slot = path->slots[0];
3619
3620 btrfs_item_key_to_cpu(leaf, &key, slot);
3621 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3622 key.objectid != btrfs_ino(BTRFS_I(src)))
3623 break;
3624
3625 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3626 struct btrfs_file_extent_item *extent;
3627 int type;
3628 u32 size;
3629 struct btrfs_key new_key;
3630 u64 disko = 0, diskl = 0;
3631 u64 datao = 0, datal = 0;
3632 u8 comp;
3633 u64 drop_start;
3634
3635 extent = btrfs_item_ptr(leaf, slot,
3636 struct btrfs_file_extent_item);
3637 comp = btrfs_file_extent_compression(leaf, extent);
3638 type = btrfs_file_extent_type(leaf, extent);
3639 if (type == BTRFS_FILE_EXTENT_REG ||
3640 type == BTRFS_FILE_EXTENT_PREALLOC) {
3641 disko = btrfs_file_extent_disk_bytenr(leaf,
3642 extent);
3643 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3644 extent);
3645 datao = btrfs_file_extent_offset(leaf, extent);
3646 datal = btrfs_file_extent_num_bytes(leaf,
3647 extent);
3648 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3649 /* take upper bound, may be compressed */
3650 datal = btrfs_file_extent_ram_bytes(leaf,
3651 extent);
3652 }
3653
3654 /*
3655 * The first search might have left us at an extent
3656 * item that ends before our target range's start, can
3657 * happen if we have holes and NO_HOLES feature enabled.
3658 */
3659 if (key.offset + datal <= off) {
3660 path->slots[0]++;
3661 goto process_slot;
3662 } else if (key.offset >= off + len) {
3663 break;
3664 }
3665 next_key_min_offset = key.offset + datal;
3666 size = btrfs_item_size_nr(leaf, slot);
3667 read_extent_buffer(leaf, buf,
3668 btrfs_item_ptr_offset(leaf, slot),
3669 size);
3670
3671 btrfs_release_path(path);
3672 path->leave_spinning = 0;
3673
3674 memcpy(&new_key, &key, sizeof(new_key));
3675 new_key.objectid = btrfs_ino(BTRFS_I(inode));
3676 if (off <= key.offset)
3677 new_key.offset = key.offset + destoff - off;
3678 else
3679 new_key.offset = destoff;
3680
3681 /*
3682 * Deal with a hole that doesn't have an extent item
3683 * that represents it (NO_HOLES feature enabled).
3684 * This hole is either in the middle of the cloning
3685 * range or at the beginning (fully overlaps it or
3686 * partially overlaps it).
3687 */
3688 if (new_key.offset != last_dest_end)
3689 drop_start = last_dest_end;
3690 else
3691 drop_start = new_key.offset;
3692
3693 /*
3694 * 1 - adjusting old extent (we may have to split it)
3695 * 1 - add new extent
3696 * 1 - inode update
3697 */
3698 trans = btrfs_start_transaction(root, 3);
3699 if (IS_ERR(trans)) {
3700 ret = PTR_ERR(trans);
3701 goto out;
3702 }
3703
3704 if (type == BTRFS_FILE_EXTENT_REG ||
3705 type == BTRFS_FILE_EXTENT_PREALLOC) {
3706 /*
3707 * a | --- range to clone ---| b
3708 * | ------------- extent ------------- |
3709 */
3710
3711 /* subtract range b */
3712 if (key.offset + datal > off + len)
3713 datal = off + len - key.offset;
3714
3715 /* subtract range a */
3716 if (off > key.offset) {
3717 datao += off - key.offset;
3718 datal -= off - key.offset;
3719 }
3720
3721 ret = btrfs_drop_extents(trans, root, inode,
3722 drop_start,
3723 new_key.offset + datal,
3724 1);
3725 if (ret) {
3726 if (ret != -EOPNOTSUPP)
3727 btrfs_abort_transaction(trans,
3728 ret);
3729 btrfs_end_transaction(trans);
3730 goto out;
3731 }
3732
3733 ret = btrfs_insert_empty_item(trans, root, path,
3734 &new_key, size);
3735 if (ret) {
3736 btrfs_abort_transaction(trans, ret);
3737 btrfs_end_transaction(trans);
3738 goto out;
3739 }
3740
3741 leaf = path->nodes[0];
3742 slot = path->slots[0];
3743 write_extent_buffer(leaf, buf,
3744 btrfs_item_ptr_offset(leaf, slot),
3745 size);
3746
3747 extent = btrfs_item_ptr(leaf, slot,
3748 struct btrfs_file_extent_item);
3749
3750 /* disko == 0 means it's a hole */
3751 if (!disko)
3752 datao = 0;
3753
3754 btrfs_set_file_extent_offset(leaf, extent,
3755 datao);
3756 btrfs_set_file_extent_num_bytes(leaf, extent,
3757 datal);
3758
3759 if (disko) {
3760 struct btrfs_ref ref = { 0 };
3761 inode_add_bytes(inode, datal);
3762 btrfs_init_generic_ref(&ref,
3763 BTRFS_ADD_DELAYED_REF, disko,
3764 diskl, 0);
3765 btrfs_init_data_ref(&ref,
3766 root->root_key.objectid,
3767 btrfs_ino(BTRFS_I(inode)),
3768 new_key.offset - datao);
3769 ret = btrfs_inc_extent_ref(trans, &ref);
3770 if (ret) {
3771 btrfs_abort_transaction(trans,
3772 ret);
3773 btrfs_end_transaction(trans);
3774 goto out;
3775
3776 }
3777 }
3778 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3779 u64 skip = 0;
3780 u64 trim = 0;
3781
3782 if (off > key.offset) {
3783 skip = off - key.offset;
3784 new_key.offset += skip;
3785 }
3786
3787 if (key.offset + datal > off + len)
3788 trim = key.offset + datal - (off + len);
3789
3790 if (comp && (skip || trim)) {
3791 ret = -EINVAL;
3792 btrfs_end_transaction(trans);
3793 goto out;
3794 }
3795 size -= skip + trim;
3796 datal -= skip + trim;
3797
3798 ret = clone_copy_inline_extent(inode,
3799 trans, path,
3800 &new_key,
3801 drop_start,
3802 datal,
3803 skip, size, buf);
3804 if (ret) {
3805 if (ret != -EOPNOTSUPP)
3806 btrfs_abort_transaction(trans,
3807 ret);
3808 btrfs_end_transaction(trans);
3809 goto out;
3810 }
3811 leaf = path->nodes[0];
3812 slot = path->slots[0];
3813 }
3814
3815 /* If we have an implicit hole (NO_HOLES feature). */
3816 if (drop_start < new_key.offset)
3817 clone_update_extent_map(BTRFS_I(inode), trans,
3818 NULL, drop_start,
3819 new_key.offset - drop_start);
3820
3821 clone_update_extent_map(BTRFS_I(inode), trans,
3822 path, 0, 0);
3823
3824 btrfs_mark_buffer_dirty(leaf);
3825 btrfs_release_path(path);
3826
3827 last_dest_end = ALIGN(new_key.offset + datal,
3828 fs_info->sectorsize);
3829 ret = clone_finish_inode_update(trans, inode,
3830 last_dest_end,
3831 destoff, olen,
3832 no_time_update);
3833 if (ret)
3834 goto out;
3835 if (new_key.offset + datal >= destoff + len)
3836 break;
3837 }
3838 btrfs_release_path(path);
3839 key.offset = next_key_min_offset;
3840
3841 if (fatal_signal_pending(current)) {
3842 ret = -EINTR;
3843 goto out;
3844 }
3845 }
3846 ret = 0;
3847
3848 if (last_dest_end < destoff + len) {
3849 /*
3850 * We have an implicit hole (NO_HOLES feature is enabled) that
3851 * fully or partially overlaps our cloning range at its end.
3852 */
3853 btrfs_release_path(path);
3854
3855 /*
3856 * 1 - remove extent(s)
3857 * 1 - inode update
3858 */
3859 trans = btrfs_start_transaction(root, 2);
3860 if (IS_ERR(trans)) {
3861 ret = PTR_ERR(trans);
3862 goto out;
3863 }
3864 ret = btrfs_drop_extents(trans, root, inode,
3865 last_dest_end, destoff + len, 1);
3866 if (ret) {
3867 if (ret != -EOPNOTSUPP)
3868 btrfs_abort_transaction(trans, ret);
3869 btrfs_end_transaction(trans);
3870 goto out;
3871 }
3872 clone_update_extent_map(BTRFS_I(inode), trans, NULL,
3873 last_dest_end,
3874 destoff + len - last_dest_end);
3875 ret = clone_finish_inode_update(trans, inode, destoff + len,
3876 destoff, olen, no_time_update);
3877 }
3878
3879 out:
3880 btrfs_free_path(path);
3881 kvfree(buf);
3882 return ret;
3883 }
3884
3885 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3886 u64 off, u64 olen, u64 destoff)
3887 {
3888 struct inode *inode = file_inode(file);
3889 struct inode *src = file_inode(file_src);
3890 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3891 int ret;
3892 u64 len = olen;
3893 u64 bs = fs_info->sb->s_blocksize;
3894
3895 /*
3896 * TODO:
3897 * - split compressed inline extents. annoying: we need to
3898 * decompress into destination's address_space (the file offset
3899 * may change, so source mapping won't do), then recompress (or
3900 * otherwise reinsert) a subrange.
3901 *
3902 * - split destination inode's inline extents. The inline extents can
3903 * be either compressed or non-compressed.
3904 */
3905
3906 /*
3907 * VFS's generic_remap_file_range_prep() protects us from cloning the
3908 * eof block into the middle of a file, which would result in corruption
3909 * if the file size is not blocksize aligned. So we don't need to check
3910 * for that case here.
3911 */
3912 if (off + len == src->i_size)
3913 len = ALIGN(src->i_size, bs) - off;
3914
3915 if (destoff > inode->i_size) {
3916 const u64 wb_start = ALIGN_DOWN(inode->i_size, bs);
3917
3918 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3919 if (ret)
3920 return ret;
3921 /*
3922 * We may have truncated the last block if the inode's size is
3923 * not sector size aligned, so we need to wait for writeback to
3924 * complete before proceeding further, otherwise we can race
3925 * with cloning and attempt to increment a reference to an
3926 * extent that no longer exists (writeback completed right after
3927 * we found the previous extent covering eof and before we
3928 * attempted to increment its reference count).
3929 */
3930 ret = btrfs_wait_ordered_range(inode, wb_start,
3931 destoff - wb_start);
3932 if (ret)
3933 return ret;
3934 }
3935
3936 /*
3937 * Lock destination range to serialize with concurrent readpages() and
3938 * source range to serialize with relocation.
3939 */
3940 btrfs_double_extent_lock(src, off, inode, destoff, len);
3941 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3942 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3943 /*
3944 * Truncate page cache pages so that future reads will see the cloned
3945 * data immediately and not the previous data.
3946 */
3947 truncate_inode_pages_range(&inode->i_data,
3948 round_down(destoff, PAGE_SIZE),
3949 round_up(destoff + len, PAGE_SIZE) - 1);
3950
3951 return ret;
3952 }
3953
3954 static int btrfs_remap_file_range_prep(struct file *file_in, loff_t pos_in,
3955 struct file *file_out, loff_t pos_out,
3956 loff_t *len, unsigned int remap_flags)
3957 {
3958 struct inode *inode_in = file_inode(file_in);
3959 struct inode *inode_out = file_inode(file_out);
3960 u64 bs = BTRFS_I(inode_out)->root->fs_info->sb->s_blocksize;
3961 bool same_inode = inode_out == inode_in;
3962 u64 wb_len;
3963 int ret;
3964
3965 if (!(remap_flags & REMAP_FILE_DEDUP)) {
3966 struct btrfs_root *root_out = BTRFS_I(inode_out)->root;
3967
3968 if (btrfs_root_readonly(root_out))
3969 return -EROFS;
3970
3971 if (file_in->f_path.mnt != file_out->f_path.mnt ||
3972 inode_in->i_sb != inode_out->i_sb)
3973 return -EXDEV;
3974 }
3975
3976 /* don't make the dst file partly checksummed */
3977 if ((BTRFS_I(inode_in)->flags & BTRFS_INODE_NODATASUM) !=
3978 (BTRFS_I(inode_out)->flags & BTRFS_INODE_NODATASUM)) {
3979 return -EINVAL;
3980 }
3981
3982 /*
3983 * Now that the inodes are locked, we need to start writeback ourselves
3984 * and can not rely on the writeback from the VFS's generic helper
3985 * generic_remap_file_range_prep() because:
3986 *
3987 * 1) For compression we must call filemap_fdatawrite_range() range
3988 * twice (btrfs_fdatawrite_range() does it for us), and the generic
3989 * helper only calls it once;
3990 *
3991 * 2) filemap_fdatawrite_range(), called by the generic helper only
3992 * waits for the writeback to complete, i.e. for IO to be done, and
3993 * not for the ordered extents to complete. We need to wait for them
3994 * to complete so that new file extent items are in the fs tree.
3995 */
3996 if (*len == 0 && !(remap_flags & REMAP_FILE_DEDUP))
3997 wb_len = ALIGN(inode_in->i_size, bs) - ALIGN_DOWN(pos_in, bs);
3998 else
3999 wb_len = ALIGN(*len, bs);
4000
4001 /*
4002 * Since we don't lock ranges, wait for ongoing lockless dio writes (as
4003 * any in progress could create its ordered extents after we wait for
4004 * existing ordered extents below).
4005 */
4006 inode_dio_wait(inode_in);
4007 if (!same_inode)
4008 inode_dio_wait(inode_out);
4009
4010 ret = btrfs_wait_ordered_range(inode_in, ALIGN_DOWN(pos_in, bs),
4011 wb_len);
4012 if (ret < 0)
4013 return ret;
4014 ret = btrfs_wait_ordered_range(inode_out, ALIGN_DOWN(pos_out, bs),
4015 wb_len);
4016 if (ret < 0)
4017 return ret;
4018
4019 return generic_remap_file_range_prep(file_in, pos_in, file_out, pos_out,
4020 len, remap_flags);
4021 }
4022
4023 loff_t btrfs_remap_file_range(struct file *src_file, loff_t off,
4024 struct file *dst_file, loff_t destoff, loff_t len,
4025 unsigned int remap_flags)
4026 {
4027 struct inode *src_inode = file_inode(src_file);
4028 struct inode *dst_inode = file_inode(dst_file);
4029 bool same_inode = dst_inode == src_inode;
4030 int ret;
4031
4032 if (remap_flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_ADVISORY))
4033 return -EINVAL;
4034
4035 if (same_inode)
4036 inode_lock(src_inode);
4037 else
4038 lock_two_nondirectories(src_inode, dst_inode);
4039
4040 ret = btrfs_remap_file_range_prep(src_file, off, dst_file, destoff,
4041 &len, remap_flags);
4042 if (ret < 0 || len == 0)
4043 goto out_unlock;
4044
4045 if (remap_flags & REMAP_FILE_DEDUP)
4046 ret = btrfs_extent_same(src_inode, off, len, dst_inode, destoff);
4047 else
4048 ret = btrfs_clone_files(dst_file, src_file, off, len, destoff);
4049
4050 out_unlock:
4051 if (same_inode)
4052 inode_unlock(src_inode);
4053 else
4054 unlock_two_nondirectories(src_inode, dst_inode);
4055
4056 return ret < 0 ? ret : len;
4057 }
4058
4059 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
4060 {
4061 struct inode *inode = file_inode(file);
4062 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4063 struct btrfs_root *root = BTRFS_I(inode)->root;
4064 struct btrfs_root *new_root;
4065 struct btrfs_dir_item *di;
4066 struct btrfs_trans_handle *trans;
4067 struct btrfs_path *path;
4068 struct btrfs_key location;
4069 struct btrfs_disk_key disk_key;
4070 u64 objectid = 0;
4071 u64 dir_id;
4072 int ret;
4073
4074 if (!capable(CAP_SYS_ADMIN))
4075 return -EPERM;
4076
4077 ret = mnt_want_write_file(file);
4078 if (ret)
4079 return ret;
4080
4081 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4082 ret = -EFAULT;
4083 goto out;
4084 }
4085
4086 if (!objectid)
4087 objectid = BTRFS_FS_TREE_OBJECTID;
4088
4089 location.objectid = objectid;
4090 location.type = BTRFS_ROOT_ITEM_KEY;
4091 location.offset = (u64)-1;
4092
4093 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
4094 if (IS_ERR(new_root)) {
4095 ret = PTR_ERR(new_root);
4096 goto out;
4097 }
4098 if (!is_fstree(new_root->root_key.objectid)) {
4099 ret = -ENOENT;
4100 goto out;
4101 }
4102
4103 path = btrfs_alloc_path();
4104 if (!path) {
4105 ret = -ENOMEM;
4106 goto out;
4107 }
4108 path->leave_spinning = 1;
4109
4110 trans = btrfs_start_transaction(root, 1);
4111 if (IS_ERR(trans)) {
4112 btrfs_free_path(path);
4113 ret = PTR_ERR(trans);
4114 goto out;
4115 }
4116
4117 dir_id = btrfs_super_root_dir(fs_info->super_copy);
4118 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
4119 dir_id, "default", 7, 1);
4120 if (IS_ERR_OR_NULL(di)) {
4121 btrfs_free_path(path);
4122 btrfs_end_transaction(trans);
4123 btrfs_err(fs_info,
4124 "Umm, you don't have the default diritem, this isn't going to work");
4125 ret = -ENOENT;
4126 goto out;
4127 }
4128
4129 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4130 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4131 btrfs_mark_buffer_dirty(path->nodes[0]);
4132 btrfs_free_path(path);
4133
4134 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
4135 btrfs_end_transaction(trans);
4136 out:
4137 mnt_drop_write_file(file);
4138 return ret;
4139 }
4140
4141 static void get_block_group_info(struct list_head *groups_list,
4142 struct btrfs_ioctl_space_info *space)
4143 {
4144 struct btrfs_block_group_cache *block_group;
4145
4146 space->total_bytes = 0;
4147 space->used_bytes = 0;
4148 space->flags = 0;
4149 list_for_each_entry(block_group, groups_list, list) {
4150 space->flags = block_group->flags;
4151 space->total_bytes += block_group->key.offset;
4152 space->used_bytes +=
4153 btrfs_block_group_used(&block_group->item);
4154 }
4155 }
4156
4157 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
4158 void __user *arg)
4159 {
4160 struct btrfs_ioctl_space_args space_args;
4161 struct btrfs_ioctl_space_info space;
4162 struct btrfs_ioctl_space_info *dest;
4163 struct btrfs_ioctl_space_info *dest_orig;
4164 struct btrfs_ioctl_space_info __user *user_dest;
4165 struct btrfs_space_info *info;
4166 static const u64 types[] = {
4167 BTRFS_BLOCK_GROUP_DATA,
4168 BTRFS_BLOCK_GROUP_SYSTEM,
4169 BTRFS_BLOCK_GROUP_METADATA,
4170 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
4171 };
4172 int num_types = 4;
4173 int alloc_size;
4174 int ret = 0;
4175 u64 slot_count = 0;
4176 int i, c;
4177
4178 if (copy_from_user(&space_args,
4179 (struct btrfs_ioctl_space_args __user *)arg,
4180 sizeof(space_args)))
4181 return -EFAULT;
4182
4183 for (i = 0; i < num_types; i++) {
4184 struct btrfs_space_info *tmp;
4185
4186 info = NULL;
4187 rcu_read_lock();
4188 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4189 list) {
4190 if (tmp->flags == types[i]) {
4191 info = tmp;
4192 break;
4193 }
4194 }
4195 rcu_read_unlock();
4196
4197 if (!info)
4198 continue;
4199
4200 down_read(&info->groups_sem);
4201 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4202 if (!list_empty(&info->block_groups[c]))
4203 slot_count++;
4204 }
4205 up_read(&info->groups_sem);
4206 }
4207
4208 /*
4209 * Global block reserve, exported as a space_info
4210 */
4211 slot_count++;
4212
4213 /* space_slots == 0 means they are asking for a count */
4214 if (space_args.space_slots == 0) {
4215 space_args.total_spaces = slot_count;
4216 goto out;
4217 }
4218
4219 slot_count = min_t(u64, space_args.space_slots, slot_count);
4220
4221 alloc_size = sizeof(*dest) * slot_count;
4222
4223 /* we generally have at most 6 or so space infos, one for each raid
4224 * level. So, a whole page should be more than enough for everyone
4225 */
4226 if (alloc_size > PAGE_SIZE)
4227 return -ENOMEM;
4228
4229 space_args.total_spaces = 0;
4230 dest = kmalloc(alloc_size, GFP_KERNEL);
4231 if (!dest)
4232 return -ENOMEM;
4233 dest_orig = dest;
4234
4235 /* now we have a buffer to copy into */
4236 for (i = 0; i < num_types; i++) {
4237 struct btrfs_space_info *tmp;
4238
4239 if (!slot_count)
4240 break;
4241
4242 info = NULL;
4243 rcu_read_lock();
4244 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4245 list) {
4246 if (tmp->flags == types[i]) {
4247 info = tmp;
4248 break;
4249 }
4250 }
4251 rcu_read_unlock();
4252
4253 if (!info)
4254 continue;
4255 down_read(&info->groups_sem);
4256 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4257 if (!list_empty(&info->block_groups[c])) {
4258 get_block_group_info(&info->block_groups[c],
4259 &space);
4260 memcpy(dest, &space, sizeof(space));
4261 dest++;
4262 space_args.total_spaces++;
4263 slot_count--;
4264 }
4265 if (!slot_count)
4266 break;
4267 }
4268 up_read(&info->groups_sem);
4269 }
4270
4271 /*
4272 * Add global block reserve
4273 */
4274 if (slot_count) {
4275 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
4276
4277 spin_lock(&block_rsv->lock);
4278 space.total_bytes = block_rsv->size;
4279 space.used_bytes = block_rsv->size - block_rsv->reserved;
4280 spin_unlock(&block_rsv->lock);
4281 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4282 memcpy(dest, &space, sizeof(space));
4283 space_args.total_spaces++;
4284 }
4285
4286 user_dest = (struct btrfs_ioctl_space_info __user *)
4287 (arg + sizeof(struct btrfs_ioctl_space_args));
4288
4289 if (copy_to_user(user_dest, dest_orig, alloc_size))
4290 ret = -EFAULT;
4291
4292 kfree(dest_orig);
4293 out:
4294 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4295 ret = -EFAULT;
4296
4297 return ret;
4298 }
4299
4300 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4301 void __user *argp)
4302 {
4303 struct btrfs_trans_handle *trans;
4304 u64 transid;
4305 int ret;
4306
4307 trans = btrfs_attach_transaction_barrier(root);
4308 if (IS_ERR(trans)) {
4309 if (PTR_ERR(trans) != -ENOENT)
4310 return PTR_ERR(trans);
4311
4312 /* No running transaction, don't bother */
4313 transid = root->fs_info->last_trans_committed;
4314 goto out;
4315 }
4316 transid = trans->transid;
4317 ret = btrfs_commit_transaction_async(trans, 0);
4318 if (ret) {
4319 btrfs_end_transaction(trans);
4320 return ret;
4321 }
4322 out:
4323 if (argp)
4324 if (copy_to_user(argp, &transid, sizeof(transid)))
4325 return -EFAULT;
4326 return 0;
4327 }
4328
4329 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
4330 void __user *argp)
4331 {
4332 u64 transid;
4333
4334 if (argp) {
4335 if (copy_from_user(&transid, argp, sizeof(transid)))
4336 return -EFAULT;
4337 } else {
4338 transid = 0; /* current trans */
4339 }
4340 return btrfs_wait_for_commit(fs_info, transid);
4341 }
4342
4343 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4344 {
4345 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4346 struct btrfs_ioctl_scrub_args *sa;
4347 int ret;
4348
4349 if (!capable(CAP_SYS_ADMIN))
4350 return -EPERM;
4351
4352 sa = memdup_user(arg, sizeof(*sa));
4353 if (IS_ERR(sa))
4354 return PTR_ERR(sa);
4355
4356 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4357 ret = mnt_want_write_file(file);
4358 if (ret)
4359 goto out;
4360 }
4361
4362 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4363 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4364 0);
4365
4366 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
4367 ret = -EFAULT;
4368
4369 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4370 mnt_drop_write_file(file);
4371 out:
4372 kfree(sa);
4373 return ret;
4374 }
4375
4376 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
4377 {
4378 if (!capable(CAP_SYS_ADMIN))
4379 return -EPERM;
4380
4381 return btrfs_scrub_cancel(fs_info);
4382 }
4383
4384 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
4385 void __user *arg)
4386 {
4387 struct btrfs_ioctl_scrub_args *sa;
4388 int ret;
4389
4390 if (!capable(CAP_SYS_ADMIN))
4391 return -EPERM;
4392
4393 sa = memdup_user(arg, sizeof(*sa));
4394 if (IS_ERR(sa))
4395 return PTR_ERR(sa);
4396
4397 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
4398
4399 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
4400 ret = -EFAULT;
4401
4402 kfree(sa);
4403 return ret;
4404 }
4405
4406 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
4407 void __user *arg)
4408 {
4409 struct btrfs_ioctl_get_dev_stats *sa;
4410 int ret;
4411
4412 sa = memdup_user(arg, sizeof(*sa));
4413 if (IS_ERR(sa))
4414 return PTR_ERR(sa);
4415
4416 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4417 kfree(sa);
4418 return -EPERM;
4419 }
4420
4421 ret = btrfs_get_dev_stats(fs_info, sa);
4422
4423 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
4424 ret = -EFAULT;
4425
4426 kfree(sa);
4427 return ret;
4428 }
4429
4430 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4431 void __user *arg)
4432 {
4433 struct btrfs_ioctl_dev_replace_args *p;
4434 int ret;
4435
4436 if (!capable(CAP_SYS_ADMIN))
4437 return -EPERM;
4438
4439 p = memdup_user(arg, sizeof(*p));
4440 if (IS_ERR(p))
4441 return PTR_ERR(p);
4442
4443 switch (p->cmd) {
4444 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4445 if (sb_rdonly(fs_info->sb)) {
4446 ret = -EROFS;
4447 goto out;
4448 }
4449 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4450 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4451 } else {
4452 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
4453 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4454 }
4455 break;
4456 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4457 btrfs_dev_replace_status(fs_info, p);
4458 ret = 0;
4459 break;
4460 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4461 p->result = btrfs_dev_replace_cancel(fs_info);
4462 ret = 0;
4463 break;
4464 default:
4465 ret = -EINVAL;
4466 break;
4467 }
4468
4469 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
4470 ret = -EFAULT;
4471 out:
4472 kfree(p);
4473 return ret;
4474 }
4475
4476 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4477 {
4478 int ret = 0;
4479 int i;
4480 u64 rel_ptr;
4481 int size;
4482 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4483 struct inode_fs_paths *ipath = NULL;
4484 struct btrfs_path *path;
4485
4486 if (!capable(CAP_DAC_READ_SEARCH))
4487 return -EPERM;
4488
4489 path = btrfs_alloc_path();
4490 if (!path) {
4491 ret = -ENOMEM;
4492 goto out;
4493 }
4494
4495 ipa = memdup_user(arg, sizeof(*ipa));
4496 if (IS_ERR(ipa)) {
4497 ret = PTR_ERR(ipa);
4498 ipa = NULL;
4499 goto out;
4500 }
4501
4502 size = min_t(u32, ipa->size, 4096);
4503 ipath = init_ipath(size, root, path);
4504 if (IS_ERR(ipath)) {
4505 ret = PTR_ERR(ipath);
4506 ipath = NULL;
4507 goto out;
4508 }
4509
4510 ret = paths_from_inode(ipa->inum, ipath);
4511 if (ret < 0)
4512 goto out;
4513
4514 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4515 rel_ptr = ipath->fspath->val[i] -
4516 (u64)(unsigned long)ipath->fspath->val;
4517 ipath->fspath->val[i] = rel_ptr;
4518 }
4519
4520 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
4521 ipath->fspath, size);
4522 if (ret) {
4523 ret = -EFAULT;
4524 goto out;
4525 }
4526
4527 out:
4528 btrfs_free_path(path);
4529 free_ipath(ipath);
4530 kfree(ipa);
4531
4532 return ret;
4533 }
4534
4535 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4536 {
4537 struct btrfs_data_container *inodes = ctx;
4538 const size_t c = 3 * sizeof(u64);
4539
4540 if (inodes->bytes_left >= c) {
4541 inodes->bytes_left -= c;
4542 inodes->val[inodes->elem_cnt] = inum;
4543 inodes->val[inodes->elem_cnt + 1] = offset;
4544 inodes->val[inodes->elem_cnt + 2] = root;
4545 inodes->elem_cnt += 3;
4546 } else {
4547 inodes->bytes_missing += c - inodes->bytes_left;
4548 inodes->bytes_left = 0;
4549 inodes->elem_missed += 3;
4550 }
4551
4552 return 0;
4553 }
4554
4555 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
4556 void __user *arg, int version)
4557 {
4558 int ret = 0;
4559 int size;
4560 struct btrfs_ioctl_logical_ino_args *loi;
4561 struct btrfs_data_container *inodes = NULL;
4562 struct btrfs_path *path = NULL;
4563 bool ignore_offset;
4564
4565 if (!capable(CAP_SYS_ADMIN))
4566 return -EPERM;
4567
4568 loi = memdup_user(arg, sizeof(*loi));
4569 if (IS_ERR(loi))
4570 return PTR_ERR(loi);
4571
4572 if (version == 1) {
4573 ignore_offset = false;
4574 size = min_t(u32, loi->size, SZ_64K);
4575 } else {
4576 /* All reserved bits must be 0 for now */
4577 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
4578 ret = -EINVAL;
4579 goto out_loi;
4580 }
4581 /* Only accept flags we have defined so far */
4582 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
4583 ret = -EINVAL;
4584 goto out_loi;
4585 }
4586 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
4587 size = min_t(u32, loi->size, SZ_16M);
4588 }
4589
4590 path = btrfs_alloc_path();
4591 if (!path) {
4592 ret = -ENOMEM;
4593 goto out;
4594 }
4595
4596 inodes = init_data_container(size);
4597 if (IS_ERR(inodes)) {
4598 ret = PTR_ERR(inodes);
4599 inodes = NULL;
4600 goto out;
4601 }
4602
4603 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
4604 build_ino_list, inodes, ignore_offset);
4605 if (ret == -EINVAL)
4606 ret = -ENOENT;
4607 if (ret < 0)
4608 goto out;
4609
4610 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
4611 size);
4612 if (ret)
4613 ret = -EFAULT;
4614
4615 out:
4616 btrfs_free_path(path);
4617 kvfree(inodes);
4618 out_loi:
4619 kfree(loi);
4620
4621 return ret;
4622 }
4623
4624 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
4625 struct btrfs_ioctl_balance_args *bargs)
4626 {
4627 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4628
4629 bargs->flags = bctl->flags;
4630
4631 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
4632 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4633 if (atomic_read(&fs_info->balance_pause_req))
4634 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4635 if (atomic_read(&fs_info->balance_cancel_req))
4636 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4637
4638 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4639 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4640 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4641
4642 spin_lock(&fs_info->balance_lock);
4643 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4644 spin_unlock(&fs_info->balance_lock);
4645 }
4646
4647 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4648 {
4649 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4650 struct btrfs_fs_info *fs_info = root->fs_info;
4651 struct btrfs_ioctl_balance_args *bargs;
4652 struct btrfs_balance_control *bctl;
4653 bool need_unlock; /* for mut. excl. ops lock */
4654 int ret;
4655
4656 if (!capable(CAP_SYS_ADMIN))
4657 return -EPERM;
4658
4659 ret = mnt_want_write_file(file);
4660 if (ret)
4661 return ret;
4662
4663 again:
4664 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4665 mutex_lock(&fs_info->balance_mutex);
4666 need_unlock = true;
4667 goto locked;
4668 }
4669
4670 /*
4671 * mut. excl. ops lock is locked. Three possibilities:
4672 * (1) some other op is running
4673 * (2) balance is running
4674 * (3) balance is paused -- special case (think resume)
4675 */
4676 mutex_lock(&fs_info->balance_mutex);
4677 if (fs_info->balance_ctl) {
4678 /* this is either (2) or (3) */
4679 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4680 mutex_unlock(&fs_info->balance_mutex);
4681 /*
4682 * Lock released to allow other waiters to continue,
4683 * we'll reexamine the status again.
4684 */
4685 mutex_lock(&fs_info->balance_mutex);
4686
4687 if (fs_info->balance_ctl &&
4688 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4689 /* this is (3) */
4690 need_unlock = false;
4691 goto locked;
4692 }
4693
4694 mutex_unlock(&fs_info->balance_mutex);
4695 goto again;
4696 } else {
4697 /* this is (2) */
4698 mutex_unlock(&fs_info->balance_mutex);
4699 ret = -EINPROGRESS;
4700 goto out;
4701 }
4702 } else {
4703 /* this is (1) */
4704 mutex_unlock(&fs_info->balance_mutex);
4705 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4706 goto out;
4707 }
4708
4709 locked:
4710 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
4711
4712 if (arg) {
4713 bargs = memdup_user(arg, sizeof(*bargs));
4714 if (IS_ERR(bargs)) {
4715 ret = PTR_ERR(bargs);
4716 goto out_unlock;
4717 }
4718
4719 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4720 if (!fs_info->balance_ctl) {
4721 ret = -ENOTCONN;
4722 goto out_bargs;
4723 }
4724
4725 bctl = fs_info->balance_ctl;
4726 spin_lock(&fs_info->balance_lock);
4727 bctl->flags |= BTRFS_BALANCE_RESUME;
4728 spin_unlock(&fs_info->balance_lock);
4729
4730 goto do_balance;
4731 }
4732 } else {
4733 bargs = NULL;
4734 }
4735
4736 if (fs_info->balance_ctl) {
4737 ret = -EINPROGRESS;
4738 goto out_bargs;
4739 }
4740
4741 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4742 if (!bctl) {
4743 ret = -ENOMEM;
4744 goto out_bargs;
4745 }
4746
4747 if (arg) {
4748 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4749 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4750 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4751
4752 bctl->flags = bargs->flags;
4753 } else {
4754 /* balance everything - no filters */
4755 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4756 }
4757
4758 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4759 ret = -EINVAL;
4760 goto out_bctl;
4761 }
4762
4763 do_balance:
4764 /*
4765 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to
4766 * btrfs_balance. bctl is freed in reset_balance_state, or, if
4767 * restriper was paused all the way until unmount, in free_fs_info.
4768 * The flag should be cleared after reset_balance_state.
4769 */
4770 need_unlock = false;
4771
4772 ret = btrfs_balance(fs_info, bctl, bargs);
4773 bctl = NULL;
4774
4775 if ((ret == 0 || ret == -ECANCELED) && arg) {
4776 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4777 ret = -EFAULT;
4778 }
4779
4780 out_bctl:
4781 kfree(bctl);
4782 out_bargs:
4783 kfree(bargs);
4784 out_unlock:
4785 mutex_unlock(&fs_info->balance_mutex);
4786 if (need_unlock)
4787 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4788 out:
4789 mnt_drop_write_file(file);
4790 return ret;
4791 }
4792
4793 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4794 {
4795 if (!capable(CAP_SYS_ADMIN))
4796 return -EPERM;
4797
4798 switch (cmd) {
4799 case BTRFS_BALANCE_CTL_PAUSE:
4800 return btrfs_pause_balance(fs_info);
4801 case BTRFS_BALANCE_CTL_CANCEL:
4802 return btrfs_cancel_balance(fs_info);
4803 }
4804
4805 return -EINVAL;
4806 }
4807
4808 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4809 void __user *arg)
4810 {
4811 struct btrfs_ioctl_balance_args *bargs;
4812 int ret = 0;
4813
4814 if (!capable(CAP_SYS_ADMIN))
4815 return -EPERM;
4816
4817 mutex_lock(&fs_info->balance_mutex);
4818 if (!fs_info->balance_ctl) {
4819 ret = -ENOTCONN;
4820 goto out;
4821 }
4822
4823 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4824 if (!bargs) {
4825 ret = -ENOMEM;
4826 goto out;
4827 }
4828
4829 btrfs_update_ioctl_balance_args(fs_info, bargs);
4830
4831 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4832 ret = -EFAULT;
4833
4834 kfree(bargs);
4835 out:
4836 mutex_unlock(&fs_info->balance_mutex);
4837 return ret;
4838 }
4839
4840 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4841 {
4842 struct inode *inode = file_inode(file);
4843 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4844 struct btrfs_ioctl_quota_ctl_args *sa;
4845 int ret;
4846
4847 if (!capable(CAP_SYS_ADMIN))
4848 return -EPERM;
4849
4850 ret = mnt_want_write_file(file);
4851 if (ret)
4852 return ret;
4853
4854 sa = memdup_user(arg, sizeof(*sa));
4855 if (IS_ERR(sa)) {
4856 ret = PTR_ERR(sa);
4857 goto drop_write;
4858 }
4859
4860 down_write(&fs_info->subvol_sem);
4861
4862 switch (sa->cmd) {
4863 case BTRFS_QUOTA_CTL_ENABLE:
4864 ret = btrfs_quota_enable(fs_info);
4865 break;
4866 case BTRFS_QUOTA_CTL_DISABLE:
4867 ret = btrfs_quota_disable(fs_info);
4868 break;
4869 default:
4870 ret = -EINVAL;
4871 break;
4872 }
4873
4874 kfree(sa);
4875 up_write(&fs_info->subvol_sem);
4876 drop_write:
4877 mnt_drop_write_file(file);
4878 return ret;
4879 }
4880
4881 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4882 {
4883 struct inode *inode = file_inode(file);
4884 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4885 struct btrfs_root *root = BTRFS_I(inode)->root;
4886 struct btrfs_ioctl_qgroup_assign_args *sa;
4887 struct btrfs_trans_handle *trans;
4888 int ret;
4889 int err;
4890
4891 if (!capable(CAP_SYS_ADMIN))
4892 return -EPERM;
4893
4894 ret = mnt_want_write_file(file);
4895 if (ret)
4896 return ret;
4897
4898 sa = memdup_user(arg, sizeof(*sa));
4899 if (IS_ERR(sa)) {
4900 ret = PTR_ERR(sa);
4901 goto drop_write;
4902 }
4903
4904 trans = btrfs_join_transaction(root);
4905 if (IS_ERR(trans)) {
4906 ret = PTR_ERR(trans);
4907 goto out;
4908 }
4909
4910 if (sa->assign) {
4911 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
4912 } else {
4913 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
4914 }
4915
4916 /* update qgroup status and info */
4917 err = btrfs_run_qgroups(trans);
4918 if (err < 0)
4919 btrfs_handle_fs_error(fs_info, err,
4920 "failed to update qgroup status and info");
4921 err = btrfs_end_transaction(trans);
4922 if (err && !ret)
4923 ret = err;
4924
4925 out:
4926 kfree(sa);
4927 drop_write:
4928 mnt_drop_write_file(file);
4929 return ret;
4930 }
4931
4932 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4933 {
4934 struct inode *inode = file_inode(file);
4935 struct btrfs_root *root = BTRFS_I(inode)->root;
4936 struct btrfs_ioctl_qgroup_create_args *sa;
4937 struct btrfs_trans_handle *trans;
4938 int ret;
4939 int err;
4940
4941 if (!capable(CAP_SYS_ADMIN))
4942 return -EPERM;
4943
4944 ret = mnt_want_write_file(file);
4945 if (ret)
4946 return ret;
4947
4948 sa = memdup_user(arg, sizeof(*sa));
4949 if (IS_ERR(sa)) {
4950 ret = PTR_ERR(sa);
4951 goto drop_write;
4952 }
4953
4954 if (!sa->qgroupid) {
4955 ret = -EINVAL;
4956 goto out;
4957 }
4958
4959 trans = btrfs_join_transaction(root);
4960 if (IS_ERR(trans)) {
4961 ret = PTR_ERR(trans);
4962 goto out;
4963 }
4964
4965 if (sa->create) {
4966 ret = btrfs_create_qgroup(trans, sa->qgroupid);
4967 } else {
4968 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
4969 }
4970
4971 err = btrfs_end_transaction(trans);
4972 if (err && !ret)
4973 ret = err;
4974
4975 out:
4976 kfree(sa);
4977 drop_write:
4978 mnt_drop_write_file(file);
4979 return ret;
4980 }
4981
4982 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4983 {
4984 struct inode *inode = file_inode(file);
4985 struct btrfs_root *root = BTRFS_I(inode)->root;
4986 struct btrfs_ioctl_qgroup_limit_args *sa;
4987 struct btrfs_trans_handle *trans;
4988 int ret;
4989 int err;
4990 u64 qgroupid;
4991
4992 if (!capable(CAP_SYS_ADMIN))
4993 return -EPERM;
4994
4995 ret = mnt_want_write_file(file);
4996 if (ret)
4997 return ret;
4998
4999 sa = memdup_user(arg, sizeof(*sa));
5000 if (IS_ERR(sa)) {
5001 ret = PTR_ERR(sa);
5002 goto drop_write;
5003 }
5004
5005 trans = btrfs_join_transaction(root);
5006 if (IS_ERR(trans)) {
5007 ret = PTR_ERR(trans);
5008 goto out;
5009 }
5010
5011 qgroupid = sa->qgroupid;
5012 if (!qgroupid) {
5013 /* take the current subvol as qgroup */
5014 qgroupid = root->root_key.objectid;
5015 }
5016
5017 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
5018
5019 err = btrfs_end_transaction(trans);
5020 if (err && !ret)
5021 ret = err;
5022
5023 out:
5024 kfree(sa);
5025 drop_write:
5026 mnt_drop_write_file(file);
5027 return ret;
5028 }
5029
5030 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
5031 {
5032 struct inode *inode = file_inode(file);
5033 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5034 struct btrfs_ioctl_quota_rescan_args *qsa;
5035 int ret;
5036
5037 if (!capable(CAP_SYS_ADMIN))
5038 return -EPERM;
5039
5040 ret = mnt_want_write_file(file);
5041 if (ret)
5042 return ret;
5043
5044 qsa = memdup_user(arg, sizeof(*qsa));
5045 if (IS_ERR(qsa)) {
5046 ret = PTR_ERR(qsa);
5047 goto drop_write;
5048 }
5049
5050 if (qsa->flags) {
5051 ret = -EINVAL;
5052 goto out;
5053 }
5054
5055 ret = btrfs_qgroup_rescan(fs_info);
5056
5057 out:
5058 kfree(qsa);
5059 drop_write:
5060 mnt_drop_write_file(file);
5061 return ret;
5062 }
5063
5064 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5065 {
5066 struct inode *inode = file_inode(file);
5067 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5068 struct btrfs_ioctl_quota_rescan_args *qsa;
5069 int ret = 0;
5070
5071 if (!capable(CAP_SYS_ADMIN))
5072 return -EPERM;
5073
5074 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
5075 if (!qsa)
5076 return -ENOMEM;
5077
5078 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5079 qsa->flags = 1;
5080 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
5081 }
5082
5083 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5084 ret = -EFAULT;
5085
5086 kfree(qsa);
5087 return ret;
5088 }
5089
5090 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5091 {
5092 struct inode *inode = file_inode(file);
5093 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5094
5095 if (!capable(CAP_SYS_ADMIN))
5096 return -EPERM;
5097
5098 return btrfs_qgroup_wait_for_completion(fs_info, true);
5099 }
5100
5101 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5102 struct btrfs_ioctl_received_subvol_args *sa)
5103 {
5104 struct inode *inode = file_inode(file);
5105 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5106 struct btrfs_root *root = BTRFS_I(inode)->root;
5107 struct btrfs_root_item *root_item = &root->root_item;
5108 struct btrfs_trans_handle *trans;
5109 struct timespec64 ct = current_time(inode);
5110 int ret = 0;
5111 int received_uuid_changed;
5112
5113 if (!inode_owner_or_capable(inode))
5114 return -EPERM;
5115
5116 ret = mnt_want_write_file(file);
5117 if (ret < 0)
5118 return ret;
5119
5120 down_write(&fs_info->subvol_sem);
5121
5122 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
5123 ret = -EINVAL;
5124 goto out;
5125 }
5126
5127 if (btrfs_root_readonly(root)) {
5128 ret = -EROFS;
5129 goto out;
5130 }
5131
5132 /*
5133 * 1 - root item
5134 * 2 - uuid items (received uuid + subvol uuid)
5135 */
5136 trans = btrfs_start_transaction(root, 3);
5137 if (IS_ERR(trans)) {
5138 ret = PTR_ERR(trans);
5139 trans = NULL;
5140 goto out;
5141 }
5142
5143 sa->rtransid = trans->transid;
5144 sa->rtime.sec = ct.tv_sec;
5145 sa->rtime.nsec = ct.tv_nsec;
5146
5147 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5148 BTRFS_UUID_SIZE);
5149 if (received_uuid_changed &&
5150 !btrfs_is_empty_uuid(root_item->received_uuid)) {
5151 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
5152 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5153 root->root_key.objectid);
5154 if (ret && ret != -ENOENT) {
5155 btrfs_abort_transaction(trans, ret);
5156 btrfs_end_transaction(trans);
5157 goto out;
5158 }
5159 }
5160 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5161 btrfs_set_root_stransid(root_item, sa->stransid);
5162 btrfs_set_root_rtransid(root_item, sa->rtransid);
5163 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5164 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5165 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5166 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5167
5168 ret = btrfs_update_root(trans, fs_info->tree_root,
5169 &root->root_key, &root->root_item);
5170 if (ret < 0) {
5171 btrfs_end_transaction(trans);
5172 goto out;
5173 }
5174 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5175 ret = btrfs_uuid_tree_add(trans, sa->uuid,
5176 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5177 root->root_key.objectid);
5178 if (ret < 0 && ret != -EEXIST) {
5179 btrfs_abort_transaction(trans, ret);
5180 btrfs_end_transaction(trans);
5181 goto out;
5182 }
5183 }
5184 ret = btrfs_commit_transaction(trans);
5185 out:
5186 up_write(&fs_info->subvol_sem);
5187 mnt_drop_write_file(file);
5188 return ret;
5189 }
5190
5191 #ifdef CONFIG_64BIT
5192 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5193 void __user *arg)
5194 {
5195 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5196 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5197 int ret = 0;
5198
5199 args32 = memdup_user(arg, sizeof(*args32));
5200 if (IS_ERR(args32))
5201 return PTR_ERR(args32);
5202
5203 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5204 if (!args64) {
5205 ret = -ENOMEM;
5206 goto out;
5207 }
5208
5209 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5210 args64->stransid = args32->stransid;
5211 args64->rtransid = args32->rtransid;
5212 args64->stime.sec = args32->stime.sec;
5213 args64->stime.nsec = args32->stime.nsec;
5214 args64->rtime.sec = args32->rtime.sec;
5215 args64->rtime.nsec = args32->rtime.nsec;
5216 args64->flags = args32->flags;
5217
5218 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5219 if (ret)
5220 goto out;
5221
5222 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5223 args32->stransid = args64->stransid;
5224 args32->rtransid = args64->rtransid;
5225 args32->stime.sec = args64->stime.sec;
5226 args32->stime.nsec = args64->stime.nsec;
5227 args32->rtime.sec = args64->rtime.sec;
5228 args32->rtime.nsec = args64->rtime.nsec;
5229 args32->flags = args64->flags;
5230
5231 ret = copy_to_user(arg, args32, sizeof(*args32));
5232 if (ret)
5233 ret = -EFAULT;
5234
5235 out:
5236 kfree(args32);
5237 kfree(args64);
5238 return ret;
5239 }
5240 #endif
5241
5242 static long btrfs_ioctl_set_received_subvol(struct file *file,
5243 void __user *arg)
5244 {
5245 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5246 int ret = 0;
5247
5248 sa = memdup_user(arg, sizeof(*sa));
5249 if (IS_ERR(sa))
5250 return PTR_ERR(sa);
5251
5252 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5253
5254 if (ret)
5255 goto out;
5256
5257 ret = copy_to_user(arg, sa, sizeof(*sa));
5258 if (ret)
5259 ret = -EFAULT;
5260
5261 out:
5262 kfree(sa);
5263 return ret;
5264 }
5265
5266 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5267 {
5268 struct inode *inode = file_inode(file);
5269 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5270 size_t len;
5271 int ret;
5272 char label[BTRFS_LABEL_SIZE];
5273
5274 spin_lock(&fs_info->super_lock);
5275 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5276 spin_unlock(&fs_info->super_lock);
5277
5278 len = strnlen(label, BTRFS_LABEL_SIZE);
5279
5280 if (len == BTRFS_LABEL_SIZE) {
5281 btrfs_warn(fs_info,
5282 "label is too long, return the first %zu bytes",
5283 --len);
5284 }
5285
5286 ret = copy_to_user(arg, label, len);
5287
5288 return ret ? -EFAULT : 0;
5289 }
5290
5291 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5292 {
5293 struct inode *inode = file_inode(file);
5294 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5295 struct btrfs_root *root = BTRFS_I(inode)->root;
5296 struct btrfs_super_block *super_block = fs_info->super_copy;
5297 struct btrfs_trans_handle *trans;
5298 char label[BTRFS_LABEL_SIZE];
5299 int ret;
5300
5301 if (!capable(CAP_SYS_ADMIN))
5302 return -EPERM;
5303
5304 if (copy_from_user(label, arg, sizeof(label)))
5305 return -EFAULT;
5306
5307 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5308 btrfs_err(fs_info,
5309 "unable to set label with more than %d bytes",
5310 BTRFS_LABEL_SIZE - 1);
5311 return -EINVAL;
5312 }
5313
5314 ret = mnt_want_write_file(file);
5315 if (ret)
5316 return ret;
5317
5318 trans = btrfs_start_transaction(root, 0);
5319 if (IS_ERR(trans)) {
5320 ret = PTR_ERR(trans);
5321 goto out_unlock;
5322 }
5323
5324 spin_lock(&fs_info->super_lock);
5325 strcpy(super_block->label, label);
5326 spin_unlock(&fs_info->super_lock);
5327 ret = btrfs_commit_transaction(trans);
5328
5329 out_unlock:
5330 mnt_drop_write_file(file);
5331 return ret;
5332 }
5333
5334 #define INIT_FEATURE_FLAGS(suffix) \
5335 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5336 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5337 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5338
5339 int btrfs_ioctl_get_supported_features(void __user *arg)
5340 {
5341 static const struct btrfs_ioctl_feature_flags features[3] = {
5342 INIT_FEATURE_FLAGS(SUPP),
5343 INIT_FEATURE_FLAGS(SAFE_SET),
5344 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5345 };
5346
5347 if (copy_to_user(arg, &features, sizeof(features)))
5348 return -EFAULT;
5349
5350 return 0;
5351 }
5352
5353 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5354 {
5355 struct inode *inode = file_inode(file);
5356 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5357 struct btrfs_super_block *super_block = fs_info->super_copy;
5358 struct btrfs_ioctl_feature_flags features;
5359
5360 features.compat_flags = btrfs_super_compat_flags(super_block);
5361 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5362 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5363
5364 if (copy_to_user(arg, &features, sizeof(features)))
5365 return -EFAULT;
5366
5367 return 0;
5368 }
5369
5370 static int check_feature_bits(struct btrfs_fs_info *fs_info,
5371 enum btrfs_feature_set set,
5372 u64 change_mask, u64 flags, u64 supported_flags,
5373 u64 safe_set, u64 safe_clear)
5374 {
5375 const char *type = btrfs_feature_set_names[set];
5376 char *names;
5377 u64 disallowed, unsupported;
5378 u64 set_mask = flags & change_mask;
5379 u64 clear_mask = ~flags & change_mask;
5380
5381 unsupported = set_mask & ~supported_flags;
5382 if (unsupported) {
5383 names = btrfs_printable_features(set, unsupported);
5384 if (names) {
5385 btrfs_warn(fs_info,
5386 "this kernel does not support the %s feature bit%s",
5387 names, strchr(names, ',') ? "s" : "");
5388 kfree(names);
5389 } else
5390 btrfs_warn(fs_info,
5391 "this kernel does not support %s bits 0x%llx",
5392 type, unsupported);
5393 return -EOPNOTSUPP;
5394 }
5395
5396 disallowed = set_mask & ~safe_set;
5397 if (disallowed) {
5398 names = btrfs_printable_features(set, disallowed);
5399 if (names) {
5400 btrfs_warn(fs_info,
5401 "can't set the %s feature bit%s while mounted",
5402 names, strchr(names, ',') ? "s" : "");
5403 kfree(names);
5404 } else
5405 btrfs_warn(fs_info,
5406 "can't set %s bits 0x%llx while mounted",
5407 type, disallowed);
5408 return -EPERM;
5409 }
5410
5411 disallowed = clear_mask & ~safe_clear;
5412 if (disallowed) {
5413 names = btrfs_printable_features(set, disallowed);
5414 if (names) {
5415 btrfs_warn(fs_info,
5416 "can't clear the %s feature bit%s while mounted",
5417 names, strchr(names, ',') ? "s" : "");
5418 kfree(names);
5419 } else
5420 btrfs_warn(fs_info,
5421 "can't clear %s bits 0x%llx while mounted",
5422 type, disallowed);
5423 return -EPERM;
5424 }
5425
5426 return 0;
5427 }
5428
5429 #define check_feature(fs_info, change_mask, flags, mask_base) \
5430 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5431 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5432 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5433 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5434
5435 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5436 {
5437 struct inode *inode = file_inode(file);
5438 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5439 struct btrfs_root *root = BTRFS_I(inode)->root;
5440 struct btrfs_super_block *super_block = fs_info->super_copy;
5441 struct btrfs_ioctl_feature_flags flags[2];
5442 struct btrfs_trans_handle *trans;
5443 u64 newflags;
5444 int ret;
5445
5446 if (!capable(CAP_SYS_ADMIN))
5447 return -EPERM;
5448
5449 if (copy_from_user(flags, arg, sizeof(flags)))
5450 return -EFAULT;
5451
5452 /* Nothing to do */
5453 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5454 !flags[0].incompat_flags)
5455 return 0;
5456
5457 ret = check_feature(fs_info, flags[0].compat_flags,
5458 flags[1].compat_flags, COMPAT);
5459 if (ret)
5460 return ret;
5461
5462 ret = check_feature(fs_info, flags[0].compat_ro_flags,
5463 flags[1].compat_ro_flags, COMPAT_RO);
5464 if (ret)
5465 return ret;
5466
5467 ret = check_feature(fs_info, flags[0].incompat_flags,
5468 flags[1].incompat_flags, INCOMPAT);
5469 if (ret)
5470 return ret;
5471
5472 ret = mnt_want_write_file(file);
5473 if (ret)
5474 return ret;
5475
5476 trans = btrfs_start_transaction(root, 0);
5477 if (IS_ERR(trans)) {
5478 ret = PTR_ERR(trans);
5479 goto out_drop_write;
5480 }
5481
5482 spin_lock(&fs_info->super_lock);
5483 newflags = btrfs_super_compat_flags(super_block);
5484 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5485 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5486 btrfs_set_super_compat_flags(super_block, newflags);
5487
5488 newflags = btrfs_super_compat_ro_flags(super_block);
5489 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5490 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5491 btrfs_set_super_compat_ro_flags(super_block, newflags);
5492
5493 newflags = btrfs_super_incompat_flags(super_block);
5494 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5495 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5496 btrfs_set_super_incompat_flags(super_block, newflags);
5497 spin_unlock(&fs_info->super_lock);
5498
5499 ret = btrfs_commit_transaction(trans);
5500 out_drop_write:
5501 mnt_drop_write_file(file);
5502
5503 return ret;
5504 }
5505
5506 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
5507 {
5508 struct btrfs_ioctl_send_args *arg;
5509 int ret;
5510
5511 if (compat) {
5512 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5513 struct btrfs_ioctl_send_args_32 args32;
5514
5515 ret = copy_from_user(&args32, argp, sizeof(args32));
5516 if (ret)
5517 return -EFAULT;
5518 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
5519 if (!arg)
5520 return -ENOMEM;
5521 arg->send_fd = args32.send_fd;
5522 arg->clone_sources_count = args32.clone_sources_count;
5523 arg->clone_sources = compat_ptr(args32.clone_sources);
5524 arg->parent_root = args32.parent_root;
5525 arg->flags = args32.flags;
5526 memcpy(arg->reserved, args32.reserved,
5527 sizeof(args32.reserved));
5528 #else
5529 return -ENOTTY;
5530 #endif
5531 } else {
5532 arg = memdup_user(argp, sizeof(*arg));
5533 if (IS_ERR(arg))
5534 return PTR_ERR(arg);
5535 }
5536 ret = btrfs_ioctl_send(file, arg);
5537 kfree(arg);
5538 return ret;
5539 }
5540
5541 long btrfs_ioctl(struct file *file, unsigned int
5542 cmd, unsigned long arg)
5543 {
5544 struct inode *inode = file_inode(file);
5545 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5546 struct btrfs_root *root = BTRFS_I(inode)->root;
5547 void __user *argp = (void __user *)arg;
5548
5549 switch (cmd) {
5550 case FS_IOC_GETFLAGS:
5551 return btrfs_ioctl_getflags(file, argp);
5552 case FS_IOC_SETFLAGS:
5553 return btrfs_ioctl_setflags(file, argp);
5554 case FS_IOC_GETVERSION:
5555 return btrfs_ioctl_getversion(file, argp);
5556 case FITRIM:
5557 return btrfs_ioctl_fitrim(file, argp);
5558 case BTRFS_IOC_SNAP_CREATE:
5559 return btrfs_ioctl_snap_create(file, argp, 0);
5560 case BTRFS_IOC_SNAP_CREATE_V2:
5561 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5562 case BTRFS_IOC_SUBVOL_CREATE:
5563 return btrfs_ioctl_snap_create(file, argp, 1);
5564 case BTRFS_IOC_SUBVOL_CREATE_V2:
5565 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5566 case BTRFS_IOC_SNAP_DESTROY:
5567 return btrfs_ioctl_snap_destroy(file, argp);
5568 case BTRFS_IOC_SUBVOL_GETFLAGS:
5569 return btrfs_ioctl_subvol_getflags(file, argp);
5570 case BTRFS_IOC_SUBVOL_SETFLAGS:
5571 return btrfs_ioctl_subvol_setflags(file, argp);
5572 case BTRFS_IOC_DEFAULT_SUBVOL:
5573 return btrfs_ioctl_default_subvol(file, argp);
5574 case BTRFS_IOC_DEFRAG:
5575 return btrfs_ioctl_defrag(file, NULL);
5576 case BTRFS_IOC_DEFRAG_RANGE:
5577 return btrfs_ioctl_defrag(file, argp);
5578 case BTRFS_IOC_RESIZE:
5579 return btrfs_ioctl_resize(file, argp);
5580 case BTRFS_IOC_ADD_DEV:
5581 return btrfs_ioctl_add_dev(fs_info, argp);
5582 case BTRFS_IOC_RM_DEV:
5583 return btrfs_ioctl_rm_dev(file, argp);
5584 case BTRFS_IOC_RM_DEV_V2:
5585 return btrfs_ioctl_rm_dev_v2(file, argp);
5586 case BTRFS_IOC_FS_INFO:
5587 return btrfs_ioctl_fs_info(fs_info, argp);
5588 case BTRFS_IOC_DEV_INFO:
5589 return btrfs_ioctl_dev_info(fs_info, argp);
5590 case BTRFS_IOC_BALANCE:
5591 return btrfs_ioctl_balance(file, NULL);
5592 case BTRFS_IOC_TREE_SEARCH:
5593 return btrfs_ioctl_tree_search(file, argp);
5594 case BTRFS_IOC_TREE_SEARCH_V2:
5595 return btrfs_ioctl_tree_search_v2(file, argp);
5596 case BTRFS_IOC_INO_LOOKUP:
5597 return btrfs_ioctl_ino_lookup(file, argp);
5598 case BTRFS_IOC_INO_PATHS:
5599 return btrfs_ioctl_ino_to_path(root, argp);
5600 case BTRFS_IOC_LOGICAL_INO:
5601 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
5602 case BTRFS_IOC_LOGICAL_INO_V2:
5603 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
5604 case BTRFS_IOC_SPACE_INFO:
5605 return btrfs_ioctl_space_info(fs_info, argp);
5606 case BTRFS_IOC_SYNC: {
5607 int ret;
5608
5609 ret = btrfs_start_delalloc_roots(fs_info, -1);
5610 if (ret)
5611 return ret;
5612 ret = btrfs_sync_fs(inode->i_sb, 1);
5613 /*
5614 * The transaction thread may want to do more work,
5615 * namely it pokes the cleaner kthread that will start
5616 * processing uncleaned subvols.
5617 */
5618 wake_up_process(fs_info->transaction_kthread);
5619 return ret;
5620 }
5621 case BTRFS_IOC_START_SYNC:
5622 return btrfs_ioctl_start_sync(root, argp);
5623 case BTRFS_IOC_WAIT_SYNC:
5624 return btrfs_ioctl_wait_sync(fs_info, argp);
5625 case BTRFS_IOC_SCRUB:
5626 return btrfs_ioctl_scrub(file, argp);
5627 case BTRFS_IOC_SCRUB_CANCEL:
5628 return btrfs_ioctl_scrub_cancel(fs_info);
5629 case BTRFS_IOC_SCRUB_PROGRESS:
5630 return btrfs_ioctl_scrub_progress(fs_info, argp);
5631 case BTRFS_IOC_BALANCE_V2:
5632 return btrfs_ioctl_balance(file, argp);
5633 case BTRFS_IOC_BALANCE_CTL:
5634 return btrfs_ioctl_balance_ctl(fs_info, arg);
5635 case BTRFS_IOC_BALANCE_PROGRESS:
5636 return btrfs_ioctl_balance_progress(fs_info, argp);
5637 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5638 return btrfs_ioctl_set_received_subvol(file, argp);
5639 #ifdef CONFIG_64BIT
5640 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5641 return btrfs_ioctl_set_received_subvol_32(file, argp);
5642 #endif
5643 case BTRFS_IOC_SEND:
5644 return _btrfs_ioctl_send(file, argp, false);
5645 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5646 case BTRFS_IOC_SEND_32:
5647 return _btrfs_ioctl_send(file, argp, true);
5648 #endif
5649 case BTRFS_IOC_GET_DEV_STATS:
5650 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5651 case BTRFS_IOC_QUOTA_CTL:
5652 return btrfs_ioctl_quota_ctl(file, argp);
5653 case BTRFS_IOC_QGROUP_ASSIGN:
5654 return btrfs_ioctl_qgroup_assign(file, argp);
5655 case BTRFS_IOC_QGROUP_CREATE:
5656 return btrfs_ioctl_qgroup_create(file, argp);
5657 case BTRFS_IOC_QGROUP_LIMIT:
5658 return btrfs_ioctl_qgroup_limit(file, argp);
5659 case BTRFS_IOC_QUOTA_RESCAN:
5660 return btrfs_ioctl_quota_rescan(file, argp);
5661 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5662 return btrfs_ioctl_quota_rescan_status(file, argp);
5663 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5664 return btrfs_ioctl_quota_rescan_wait(file, argp);
5665 case BTRFS_IOC_DEV_REPLACE:
5666 return btrfs_ioctl_dev_replace(fs_info, argp);
5667 case BTRFS_IOC_GET_FSLABEL:
5668 return btrfs_ioctl_get_fslabel(file, argp);
5669 case BTRFS_IOC_SET_FSLABEL:
5670 return btrfs_ioctl_set_fslabel(file, argp);
5671 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5672 return btrfs_ioctl_get_supported_features(argp);
5673 case BTRFS_IOC_GET_FEATURES:
5674 return btrfs_ioctl_get_features(file, argp);
5675 case BTRFS_IOC_SET_FEATURES:
5676 return btrfs_ioctl_set_features(file, argp);
5677 case FS_IOC_FSGETXATTR:
5678 return btrfs_ioctl_fsgetxattr(file, argp);
5679 case FS_IOC_FSSETXATTR:
5680 return btrfs_ioctl_fssetxattr(file, argp);
5681 case BTRFS_IOC_GET_SUBVOL_INFO:
5682 return btrfs_ioctl_get_subvol_info(file, argp);
5683 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5684 return btrfs_ioctl_get_subvol_rootref(file, argp);
5685 case BTRFS_IOC_INO_LOOKUP_USER:
5686 return btrfs_ioctl_ino_lookup_user(file, argp);
5687 }
5688
5689 return -ENOTTY;
5690 }
5691
5692 #ifdef CONFIG_COMPAT
5693 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5694 {
5695 /*
5696 * These all access 32-bit values anyway so no further
5697 * handling is necessary.
5698 */
5699 switch (cmd) {
5700 case FS_IOC32_GETFLAGS:
5701 cmd = FS_IOC_GETFLAGS;
5702 break;
5703 case FS_IOC32_SETFLAGS:
5704 cmd = FS_IOC_SETFLAGS;
5705 break;
5706 case FS_IOC32_GETVERSION:
5707 cmd = FS_IOC_GETVERSION;
5708 break;
5709 }
5710
5711 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5712 }
5713 #endif
A mirror of Linus' kernel repository