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 <linux/fileattr.h>
30 #include <linux/fsverity.h>
31 #include <linux/sched/xacct.h>
32 #include "ctree.h"
33 #include "disk-io.h"
34 #include "export.h"
35 #include "transaction.h"
36 #include "btrfs_inode.h"
37 #include "print-tree.h"
38 #include "volumes.h"
39 #include "locking.h"
40 #include "backref.h"
41 #include "rcu-string.h"
42 #include "send.h"
43 #include "dev-replace.h"
44 #include "props.h"
45 #include "sysfs.h"
46 #include "qgroup.h"
47 #include "tree-log.h"
48 #include "compression.h"
49 #include "space-info.h"
50 #include "delalloc-space.h"
51 #include "block-group.h"
52 #include "subpage.h"
53 #include "fs.h"
54 #include "accessors.h"
55 #include "extent-tree.h"
56 #include "root-tree.h"
57 #include "defrag.h"
58 #include "dir-item.h"
59 #include "uuid-tree.h"
60 #include "ioctl.h"
61 #include "file.h"
62 #include "scrub.h"
63 #include "super.h"
64
65 #ifdef CONFIG_64BIT
66 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
67 * structures are incorrect, as the timespec structure from userspace
68 * is 4 bytes too small. We define these alternatives here to teach
69 * the kernel about the 32-bit struct packing.
70 */
71 struct btrfs_ioctl_timespec_32 {
72 __u64 sec;
73 __u32 nsec;
74 } __attribute__ ((__packed__));
75
76 struct btrfs_ioctl_received_subvol_args_32 {
77 char uuid[BTRFS_UUID_SIZE]; /* in */
78 __u64 stransid; /* in */
79 __u64 rtransid; /* out */
80 struct btrfs_ioctl_timespec_32 stime; /* in */
81 struct btrfs_ioctl_timespec_32 rtime; /* out */
82 __u64 flags; /* in */
83 __u64 reserved[16]; /* in */
84 } __attribute__ ((__packed__));
85
86 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
87 struct btrfs_ioctl_received_subvol_args_32)
88 #endif
89
90 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
91 struct btrfs_ioctl_send_args_32 {
92 __s64 send_fd; /* in */
93 __u64 clone_sources_count; /* in */
94 compat_uptr_t clone_sources; /* in */
95 __u64 parent_root; /* in */
96 __u64 flags; /* in */
97 __u32 version; /* in */
98 __u8 reserved[28]; /* in */
99 } __attribute__ ((__packed__));
100
101 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
102 struct btrfs_ioctl_send_args_32)
103
104 struct btrfs_ioctl_encoded_io_args_32 {
105 compat_uptr_t iov;
106 compat_ulong_t iovcnt;
107 __s64 offset;
108 __u64 flags;
109 __u64 len;
110 __u64 unencoded_len;
111 __u64 unencoded_offset;
112 __u32 compression;
113 __u32 encryption;
114 __u8 reserved[64];
115 };
116
117 #define BTRFS_IOC_ENCODED_READ_32 _IOR(BTRFS_IOCTL_MAGIC, 64, \
118 struct btrfs_ioctl_encoded_io_args_32)
119 #define BTRFS_IOC_ENCODED_WRITE_32 _IOW(BTRFS_IOCTL_MAGIC, 64, \
120 struct btrfs_ioctl_encoded_io_args_32)
121 #endif
122
123 /* Mask out flags that are inappropriate for the given type of inode. */
btrfs_mask_fsflags_for_type(struct inode * inode,unsigned int flags)124 static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
125 unsigned int flags)
126 {
127 if (S_ISDIR(inode->i_mode))
128 return flags;
129 else if (S_ISREG(inode->i_mode))
130 return flags & ~FS_DIRSYNC_FL;
131 else
132 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
133 }
134
135 /*
136 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
137 * ioctl.
138 */
btrfs_inode_flags_to_fsflags(struct btrfs_inode * binode)139 static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode)
140 {
141 unsigned int iflags = 0;
142 u32 flags = binode->flags;
143 u32 ro_flags = binode->ro_flags;
144
145 if (flags & BTRFS_INODE_SYNC)
146 iflags |= FS_SYNC_FL;
147 if (flags & BTRFS_INODE_IMMUTABLE)
148 iflags |= FS_IMMUTABLE_FL;
149 if (flags & BTRFS_INODE_APPEND)
150 iflags |= FS_APPEND_FL;
151 if (flags & BTRFS_INODE_NODUMP)
152 iflags |= FS_NODUMP_FL;
153 if (flags & BTRFS_INODE_NOATIME)
154 iflags |= FS_NOATIME_FL;
155 if (flags & BTRFS_INODE_DIRSYNC)
156 iflags |= FS_DIRSYNC_FL;
157 if (flags & BTRFS_INODE_NODATACOW)
158 iflags |= FS_NOCOW_FL;
159 if (ro_flags & BTRFS_INODE_RO_VERITY)
160 iflags |= FS_VERITY_FL;
161
162 if (flags & BTRFS_INODE_NOCOMPRESS)
163 iflags |= FS_NOCOMP_FL;
164 else if (flags & BTRFS_INODE_COMPRESS)
165 iflags |= FS_COMPR_FL;
166
167 return iflags;
168 }
169
170 /*
171 * Update inode->i_flags based on the btrfs internal flags.
172 */
btrfs_sync_inode_flags_to_i_flags(struct inode * inode)173 void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
174 {
175 struct btrfs_inode *binode = BTRFS_I(inode);
176 unsigned int new_fl = 0;
177
178 if (binode->flags & BTRFS_INODE_SYNC)
179 new_fl |= S_SYNC;
180 if (binode->flags & BTRFS_INODE_IMMUTABLE)
181 new_fl |= S_IMMUTABLE;
182 if (binode->flags & BTRFS_INODE_APPEND)
183 new_fl |= S_APPEND;
184 if (binode->flags & BTRFS_INODE_NOATIME)
185 new_fl |= S_NOATIME;
186 if (binode->flags & BTRFS_INODE_DIRSYNC)
187 new_fl |= S_DIRSYNC;
188 if (binode->ro_flags & BTRFS_INODE_RO_VERITY)
189 new_fl |= S_VERITY;
190
191 set_mask_bits(&inode->i_flags,
192 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC |
193 S_VERITY, new_fl);
194 }
195
196 /*
197 * Check if @flags are a supported and valid set of FS_*_FL flags and that
198 * the old and new flags are not conflicting
199 */
check_fsflags(unsigned int old_flags,unsigned int flags)200 static int check_fsflags(unsigned int old_flags, unsigned int flags)
201 {
202 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
203 FS_NOATIME_FL | FS_NODUMP_FL | \
204 FS_SYNC_FL | FS_DIRSYNC_FL | \
205 FS_NOCOMP_FL | FS_COMPR_FL |
206 FS_NOCOW_FL))
207 return -EOPNOTSUPP;
208
209 /* COMPR and NOCOMP on new/old are valid */
210 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
211 return -EINVAL;
212
213 if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
214 return -EINVAL;
215
216 /* NOCOW and compression options are mutually exclusive */
217 if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
218 return -EINVAL;
219 if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
220 return -EINVAL;
221
222 return 0;
223 }
224
check_fsflags_compatible(struct btrfs_fs_info * fs_info,unsigned int flags)225 static int check_fsflags_compatible(struct btrfs_fs_info *fs_info,
226 unsigned int flags)
227 {
228 if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL))
229 return -EPERM;
230
231 return 0;
232 }
233
234 /*
235 * Set flags/xflags from the internal inode flags. The remaining items of
236 * fsxattr are zeroed.
237 */
btrfs_fileattr_get(struct dentry * dentry,struct fileattr * fa)238 int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
239 {
240 struct btrfs_inode *binode = BTRFS_I(d_inode(dentry));
241
242 fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode));
243 return 0;
244 }
245
btrfs_fileattr_set(struct mnt_idmap * idmap,struct dentry * dentry,struct fileattr * fa)246 int btrfs_fileattr_set(struct mnt_idmap *idmap,
247 struct dentry *dentry, struct fileattr *fa)
248 {
249 struct inode *inode = d_inode(dentry);
250 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
251 struct btrfs_inode *binode = BTRFS_I(inode);
252 struct btrfs_root *root = binode->root;
253 struct btrfs_trans_handle *trans;
254 unsigned int fsflags, old_fsflags;
255 int ret;
256 const char *comp = NULL;
257 u32 binode_flags;
258
259 if (btrfs_root_readonly(root))
260 return -EROFS;
261
262 if (fileattr_has_fsx(fa))
263 return -EOPNOTSUPP;
264
265 fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags);
266 old_fsflags = btrfs_inode_flags_to_fsflags(binode);
267 ret = check_fsflags(old_fsflags, fsflags);
268 if (ret)
269 return ret;
270
271 ret = check_fsflags_compatible(fs_info, fsflags);
272 if (ret)
273 return ret;
274
275 binode_flags = binode->flags;
276 if (fsflags & FS_SYNC_FL)
277 binode_flags |= BTRFS_INODE_SYNC;
278 else
279 binode_flags &= ~BTRFS_INODE_SYNC;
280 if (fsflags & FS_IMMUTABLE_FL)
281 binode_flags |= BTRFS_INODE_IMMUTABLE;
282 else
283 binode_flags &= ~BTRFS_INODE_IMMUTABLE;
284 if (fsflags & FS_APPEND_FL)
285 binode_flags |= BTRFS_INODE_APPEND;
286 else
287 binode_flags &= ~BTRFS_INODE_APPEND;
288 if (fsflags & FS_NODUMP_FL)
289 binode_flags |= BTRFS_INODE_NODUMP;
290 else
291 binode_flags &= ~BTRFS_INODE_NODUMP;
292 if (fsflags & FS_NOATIME_FL)
293 binode_flags |= BTRFS_INODE_NOATIME;
294 else
295 binode_flags &= ~BTRFS_INODE_NOATIME;
296
297 /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */
298 if (!fa->flags_valid) {
299 /* 1 item for the inode */
300 trans = btrfs_start_transaction(root, 1);
301 if (IS_ERR(trans))
302 return PTR_ERR(trans);
303 goto update_flags;
304 }
305
306 if (fsflags & FS_DIRSYNC_FL)
307 binode_flags |= BTRFS_INODE_DIRSYNC;
308 else
309 binode_flags &= ~BTRFS_INODE_DIRSYNC;
310 if (fsflags & FS_NOCOW_FL) {
311 if (S_ISREG(inode->i_mode)) {
312 /*
313 * It's safe to turn csums off here, no extents exist.
314 * Otherwise we want the flag to reflect the real COW
315 * status of the file and will not set it.
316 */
317 if (inode->i_size == 0)
318 binode_flags |= BTRFS_INODE_NODATACOW |
319 BTRFS_INODE_NODATASUM;
320 } else {
321 binode_flags |= BTRFS_INODE_NODATACOW;
322 }
323 } else {
324 /*
325 * Revert back under same assumptions as above
326 */
327 if (S_ISREG(inode->i_mode)) {
328 if (inode->i_size == 0)
329 binode_flags &= ~(BTRFS_INODE_NODATACOW |
330 BTRFS_INODE_NODATASUM);
331 } else {
332 binode_flags &= ~BTRFS_INODE_NODATACOW;
333 }
334 }
335
336 /*
337 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
338 * flag may be changed automatically if compression code won't make
339 * things smaller.
340 */
341 if (fsflags & FS_NOCOMP_FL) {
342 binode_flags &= ~BTRFS_INODE_COMPRESS;
343 binode_flags |= BTRFS_INODE_NOCOMPRESS;
344 } else if (fsflags & FS_COMPR_FL) {
345
346 if (IS_SWAPFILE(inode))
347 return -ETXTBSY;
348
349 binode_flags |= BTRFS_INODE_COMPRESS;
350 binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
351
352 comp = btrfs_compress_type2str(fs_info->compress_type);
353 if (!comp || comp[0] == 0)
354 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
355 } else {
356 binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
357 }
358
359 /*
360 * 1 for inode item
361 * 2 for properties
362 */
363 trans = btrfs_start_transaction(root, 3);
364 if (IS_ERR(trans))
365 return PTR_ERR(trans);
366
367 if (comp) {
368 ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
369 strlen(comp), 0);
370 if (ret) {
371 btrfs_abort_transaction(trans, ret);
372 goto out_end_trans;
373 }
374 } else {
375 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
376 0, 0);
377 if (ret && ret != -ENODATA) {
378 btrfs_abort_transaction(trans, ret);
379 goto out_end_trans;
380 }
381 }
382
383 update_flags:
384 binode->flags = binode_flags;
385 btrfs_sync_inode_flags_to_i_flags(inode);
386 inode_inc_iversion(inode);
387 inode_set_ctime_current(inode);
388 ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
389
390 out_end_trans:
391 btrfs_end_transaction(trans);
392 return ret;
393 }
394
395 /*
396 * Start exclusive operation @type, return true on success
397 */
btrfs_exclop_start(struct btrfs_fs_info * fs_info,enum btrfs_exclusive_operation type)398 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
399 enum btrfs_exclusive_operation type)
400 {
401 bool ret = false;
402
403 spin_lock(&fs_info->super_lock);
404 if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) {
405 fs_info->exclusive_operation = type;
406 ret = true;
407 }
408 spin_unlock(&fs_info->super_lock);
409
410 return ret;
411 }
412
413 /*
414 * Conditionally allow to enter the exclusive operation in case it's compatible
415 * with the running one. This must be paired with btrfs_exclop_start_unlock and
416 * btrfs_exclop_finish.
417 *
418 * Compatibility:
419 * - the same type is already running
420 * - when trying to add a device and balance has been paused
421 * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller
422 * must check the condition first that would allow none -> @type
423 */
btrfs_exclop_start_try_lock(struct btrfs_fs_info * fs_info,enum btrfs_exclusive_operation type)424 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
425 enum btrfs_exclusive_operation type)
426 {
427 spin_lock(&fs_info->super_lock);
428 if (fs_info->exclusive_operation == type ||
429 (fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED &&
430 type == BTRFS_EXCLOP_DEV_ADD))
431 return true;
432
433 spin_unlock(&fs_info->super_lock);
434 return false;
435 }
436
btrfs_exclop_start_unlock(struct btrfs_fs_info * fs_info)437 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info)
438 {
439 spin_unlock(&fs_info->super_lock);
440 }
441
btrfs_exclop_finish(struct btrfs_fs_info * fs_info)442 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info)
443 {
444 spin_lock(&fs_info->super_lock);
445 WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE);
446 spin_unlock(&fs_info->super_lock);
447 sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation");
448 }
449
btrfs_exclop_balance(struct btrfs_fs_info * fs_info,enum btrfs_exclusive_operation op)450 void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
451 enum btrfs_exclusive_operation op)
452 {
453 switch (op) {
454 case BTRFS_EXCLOP_BALANCE_PAUSED:
455 spin_lock(&fs_info->super_lock);
456 ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE ||
457 fs_info->exclusive_operation == BTRFS_EXCLOP_DEV_ADD ||
458 fs_info->exclusive_operation == BTRFS_EXCLOP_NONE ||
459 fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
460 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED;
461 spin_unlock(&fs_info->super_lock);
462 break;
463 case BTRFS_EXCLOP_BALANCE:
464 spin_lock(&fs_info->super_lock);
465 ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
466 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE;
467 spin_unlock(&fs_info->super_lock);
468 break;
469 default:
470 btrfs_warn(fs_info,
471 "invalid exclop balance operation %d requested", op);
472 }
473 }
474
btrfs_ioctl_getversion(struct inode * inode,int __user * arg)475 static int btrfs_ioctl_getversion(struct inode *inode, int __user *arg)
476 {
477 return put_user(inode->i_generation, arg);
478 }
479
btrfs_ioctl_fitrim(struct btrfs_fs_info * fs_info,void __user * arg)480 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
481 void __user *arg)
482 {
483 struct btrfs_device *device;
484 struct fstrim_range range;
485 u64 minlen = ULLONG_MAX;
486 u64 num_devices = 0;
487 int ret;
488
489 if (!capable(CAP_SYS_ADMIN))
490 return -EPERM;
491
492 /*
493 * btrfs_trim_block_group() depends on space cache, which is not
494 * available in zoned filesystem. So, disallow fitrim on a zoned
495 * filesystem for now.
496 */
497 if (btrfs_is_zoned(fs_info))
498 return -EOPNOTSUPP;
499
500 /*
501 * If the fs is mounted with nologreplay, which requires it to be
502 * mounted in RO mode as well, we can not allow discard on free space
503 * inside block groups, because log trees refer to extents that are not
504 * pinned in a block group's free space cache (pinning the extents is
505 * precisely the first phase of replaying a log tree).
506 */
507 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
508 return -EROFS;
509
510 rcu_read_lock();
511 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
512 dev_list) {
513 if (!device->bdev || !bdev_max_discard_sectors(device->bdev))
514 continue;
515 num_devices++;
516 minlen = min_t(u64, bdev_discard_granularity(device->bdev),
517 minlen);
518 }
519 rcu_read_unlock();
520
521 if (!num_devices)
522 return -EOPNOTSUPP;
523 if (copy_from_user(&range, arg, sizeof(range)))
524 return -EFAULT;
525
526 /*
527 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
528 * block group is in the logical address space, which can be any
529 * sectorsize aligned bytenr in the range [0, U64_MAX].
530 */
531 if (range.len < fs_info->sectorsize)
532 return -EINVAL;
533
534 range.minlen = max(range.minlen, minlen);
535 ret = btrfs_trim_fs(fs_info, &range);
536 if (ret < 0)
537 return ret;
538
539 if (copy_to_user(arg, &range, sizeof(range)))
540 return -EFAULT;
541
542 return 0;
543 }
544
btrfs_is_empty_uuid(u8 * uuid)545 int __pure btrfs_is_empty_uuid(u8 *uuid)
546 {
547 int i;
548
549 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
550 if (uuid[i])
551 return 0;
552 }
553 return 1;
554 }
555
556 /*
557 * Calculate the number of transaction items to reserve for creating a subvolume
558 * or snapshot, not including the inode, directory entries, or parent directory.
559 */
create_subvol_num_items(struct btrfs_qgroup_inherit * inherit)560 static unsigned int create_subvol_num_items(struct btrfs_qgroup_inherit *inherit)
561 {
562 /*
563 * 1 to add root block
564 * 1 to add root item
565 * 1 to add root ref
566 * 1 to add root backref
567 * 1 to add UUID item
568 * 1 to add qgroup info
569 * 1 to add qgroup limit
570 *
571 * Ideally the last two would only be accounted if qgroups are enabled,
572 * but that can change between now and the time we would insert them.
573 */
574 unsigned int num_items = 7;
575
576 if (inherit) {
577 /* 2 to add qgroup relations for each inherited qgroup */
578 num_items += 2 * inherit->num_qgroups;
579 }
580 return num_items;
581 }
582
create_subvol(struct mnt_idmap * idmap,struct inode * dir,struct dentry * dentry,struct btrfs_qgroup_inherit * inherit)583 static noinline int create_subvol(struct mnt_idmap *idmap,
584 struct inode *dir, struct dentry *dentry,
585 struct btrfs_qgroup_inherit *inherit)
586 {
587 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
588 struct btrfs_trans_handle *trans;
589 struct btrfs_key key;
590 struct btrfs_root_item *root_item;
591 struct btrfs_inode_item *inode_item;
592 struct extent_buffer *leaf;
593 struct btrfs_root *root = BTRFS_I(dir)->root;
594 struct btrfs_root *new_root;
595 struct btrfs_block_rsv block_rsv;
596 struct timespec64 cur_time = current_time(dir);
597 struct btrfs_new_inode_args new_inode_args = {
598 .dir = dir,
599 .dentry = dentry,
600 .subvol = true,
601 };
602 unsigned int trans_num_items;
603 int ret;
604 dev_t anon_dev;
605 u64 objectid;
606 u64 qgroup_reserved = 0;
607
608 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
609 if (!root_item)
610 return -ENOMEM;
611
612 ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
613 if (ret)
614 goto out_root_item;
615
616 /*
617 * Don't create subvolume whose level is not zero. Or qgroup will be
618 * screwed up since it assumes subvolume qgroup's level to be 0.
619 */
620 if (btrfs_qgroup_level(objectid)) {
621 ret = -ENOSPC;
622 goto out_root_item;
623 }
624
625 ret = get_anon_bdev(&anon_dev);
626 if (ret < 0)
627 goto out_root_item;
628
629 new_inode_args.inode = btrfs_new_subvol_inode(idmap, dir);
630 if (!new_inode_args.inode) {
631 ret = -ENOMEM;
632 goto out_anon_dev;
633 }
634 ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items);
635 if (ret)
636 goto out_inode;
637 trans_num_items += create_subvol_num_items(inherit);
638
639 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
640 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
641 trans_num_items, false);
642 if (ret)
643 goto out_new_inode_args;
644 qgroup_reserved = block_rsv.qgroup_rsv_reserved;
645
646 trans = btrfs_start_transaction(root, 0);
647 if (IS_ERR(trans)) {
648 ret = PTR_ERR(trans);
649 goto out_release_rsv;
650 }
651 ret = btrfs_record_root_in_trans(trans, BTRFS_I(dir)->root);
652 if (ret)
653 goto out;
654 btrfs_qgroup_convert_reserved_meta(root, qgroup_reserved);
655 qgroup_reserved = 0;
656 trans->block_rsv = &block_rsv;
657 trans->bytes_reserved = block_rsv.size;
658 /* Tree log can't currently deal with an inode which is a new root. */
659 btrfs_set_log_full_commit(trans);
660
661 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
662 if (ret)
663 goto out;
664
665 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
666 BTRFS_NESTING_NORMAL);
667 if (IS_ERR(leaf)) {
668 ret = PTR_ERR(leaf);
669 goto out;
670 }
671
672 btrfs_mark_buffer_dirty(trans, leaf);
673
674 inode_item = &root_item->inode;
675 btrfs_set_stack_inode_generation(inode_item, 1);
676 btrfs_set_stack_inode_size(inode_item, 3);
677 btrfs_set_stack_inode_nlink(inode_item, 1);
678 btrfs_set_stack_inode_nbytes(inode_item,
679 fs_info->nodesize);
680 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
681
682 btrfs_set_root_flags(root_item, 0);
683 btrfs_set_root_limit(root_item, 0);
684 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
685
686 btrfs_set_root_bytenr(root_item, leaf->start);
687 btrfs_set_root_generation(root_item, trans->transid);
688 btrfs_set_root_level(root_item, 0);
689 btrfs_set_root_refs(root_item, 1);
690 btrfs_set_root_used(root_item, leaf->len);
691 btrfs_set_root_last_snapshot(root_item, 0);
692
693 btrfs_set_root_generation_v2(root_item,
694 btrfs_root_generation(root_item));
695 generate_random_guid(root_item->uuid);
696 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
697 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
698 root_item->ctime = root_item->otime;
699 btrfs_set_root_ctransid(root_item, trans->transid);
700 btrfs_set_root_otransid(root_item, trans->transid);
701
702 btrfs_tree_unlock(leaf);
703
704 btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
705
706 key.objectid = objectid;
707 key.offset = 0;
708 key.type = BTRFS_ROOT_ITEM_KEY;
709 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
710 root_item);
711 if (ret) {
712 /*
713 * Since we don't abort the transaction in this case, free the
714 * tree block so that we don't leak space and leave the
715 * filesystem in an inconsistent state (an extent item in the
716 * extent tree with a backreference for a root that does not
717 * exists).
718 */
719 btrfs_tree_lock(leaf);
720 btrfs_clear_buffer_dirty(trans, leaf);
721 btrfs_tree_unlock(leaf);
722 btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
723 free_extent_buffer(leaf);
724 goto out;
725 }
726
727 free_extent_buffer(leaf);
728 leaf = NULL;
729
730 new_root = btrfs_get_new_fs_root(fs_info, objectid, &anon_dev);
731 if (IS_ERR(new_root)) {
732 ret = PTR_ERR(new_root);
733 btrfs_abort_transaction(trans, ret);
734 goto out;
735 }
736 /* anon_dev is owned by new_root now. */
737 anon_dev = 0;
738 BTRFS_I(new_inode_args.inode)->root = new_root;
739 /* ... and new_root is owned by new_inode_args.inode now. */
740
741 ret = btrfs_record_root_in_trans(trans, new_root);
742 if (ret) {
743 btrfs_abort_transaction(trans, ret);
744 goto out;
745 }
746
747 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
748 BTRFS_UUID_KEY_SUBVOL, objectid);
749 if (ret) {
750 btrfs_abort_transaction(trans, ret);
751 goto out;
752 }
753
754 ret = btrfs_create_new_inode(trans, &new_inode_args);
755 if (ret) {
756 btrfs_abort_transaction(trans, ret);
757 goto out;
758 }
759
760 d_instantiate_new(dentry, new_inode_args.inode);
761 new_inode_args.inode = NULL;
762
763 out:
764 trans->block_rsv = NULL;
765 trans->bytes_reserved = 0;
766 btrfs_end_transaction(trans);
767 out_release_rsv:
768 btrfs_block_rsv_release(fs_info, &block_rsv, (u64)-1, NULL);
769 if (qgroup_reserved)
770 btrfs_qgroup_free_meta_prealloc(root, qgroup_reserved);
771 out_new_inode_args:
772 btrfs_new_inode_args_destroy(&new_inode_args);
773 out_inode:
774 iput(new_inode_args.inode);
775 out_anon_dev:
776 if (anon_dev)
777 free_anon_bdev(anon_dev);
778 out_root_item:
779 kfree(root_item);
780 return ret;
781 }
782
create_snapshot(struct btrfs_root * root,struct inode * dir,struct dentry * dentry,bool readonly,struct btrfs_qgroup_inherit * inherit)783 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
784 struct dentry *dentry, bool readonly,
785 struct btrfs_qgroup_inherit *inherit)
786 {
787 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
788 struct inode *inode;
789 struct btrfs_pending_snapshot *pending_snapshot;
790 unsigned int trans_num_items;
791 struct btrfs_trans_handle *trans;
792 struct btrfs_block_rsv *block_rsv;
793 u64 qgroup_reserved = 0;
794 int ret;
795
796 /* We do not support snapshotting right now. */
797 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
798 btrfs_warn(fs_info,
799 "extent tree v2 doesn't support snapshotting yet");
800 return -EOPNOTSUPP;
801 }
802
803 if (btrfs_root_refs(&root->root_item) == 0)
804 return -ENOENT;
805
806 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
807 return -EINVAL;
808
809 if (atomic_read(&root->nr_swapfiles)) {
810 btrfs_warn(fs_info,
811 "cannot snapshot subvolume with active swapfile");
812 return -ETXTBSY;
813 }
814
815 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
816 if (!pending_snapshot)
817 return -ENOMEM;
818
819 ret = get_anon_bdev(&pending_snapshot->anon_dev);
820 if (ret < 0)
821 goto free_pending;
822 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
823 GFP_KERNEL);
824 pending_snapshot->path = btrfs_alloc_path();
825 if (!pending_snapshot->root_item || !pending_snapshot->path) {
826 ret = -ENOMEM;
827 goto free_pending;
828 }
829
830 block_rsv = &pending_snapshot->block_rsv;
831 btrfs_init_block_rsv(block_rsv, BTRFS_BLOCK_RSV_TEMP);
832 /*
833 * 1 to add dir item
834 * 1 to add dir index
835 * 1 to update parent inode item
836 */
837 trans_num_items = create_subvol_num_items(inherit) + 3;
838 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, block_rsv,
839 trans_num_items, false);
840 if (ret)
841 goto free_pending;
842 qgroup_reserved = block_rsv->qgroup_rsv_reserved;
843
844 pending_snapshot->dentry = dentry;
845 pending_snapshot->root = root;
846 pending_snapshot->readonly = readonly;
847 pending_snapshot->dir = dir;
848 pending_snapshot->inherit = inherit;
849
850 trans = btrfs_start_transaction(root, 0);
851 if (IS_ERR(trans)) {
852 ret = PTR_ERR(trans);
853 goto fail;
854 }
855 ret = btrfs_record_root_in_trans(trans, BTRFS_I(dir)->root);
856 if (ret) {
857 btrfs_end_transaction(trans);
858 goto fail;
859 }
860 btrfs_qgroup_convert_reserved_meta(root, qgroup_reserved);
861 qgroup_reserved = 0;
862
863 trans->pending_snapshot = pending_snapshot;
864
865 ret = btrfs_commit_transaction(trans);
866 if (ret)
867 goto fail;
868
869 ret = pending_snapshot->error;
870 if (ret)
871 goto fail;
872
873 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
874 if (ret)
875 goto fail;
876
877 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
878 if (IS_ERR(inode)) {
879 ret = PTR_ERR(inode);
880 goto fail;
881 }
882
883 d_instantiate(dentry, inode);
884 ret = 0;
885 pending_snapshot->anon_dev = 0;
886 fail:
887 /* Prevent double freeing of anon_dev */
888 if (ret && pending_snapshot->snap)
889 pending_snapshot->snap->anon_dev = 0;
890 btrfs_put_root(pending_snapshot->snap);
891 btrfs_block_rsv_release(fs_info, block_rsv, (u64)-1, NULL);
892 if (qgroup_reserved)
893 btrfs_qgroup_free_meta_prealloc(root, qgroup_reserved);
894 free_pending:
895 if (pending_snapshot->anon_dev)
896 free_anon_bdev(pending_snapshot->anon_dev);
897 kfree(pending_snapshot->root_item);
898 btrfs_free_path(pending_snapshot->path);
899 kfree(pending_snapshot);
900
901 return ret;
902 }
903
904 /* copy of may_delete in fs/namei.c()
905 * Check whether we can remove a link victim from directory dir, check
906 * whether the type of victim is right.
907 * 1. We can't do it if dir is read-only (done in permission())
908 * 2. We should have write and exec permissions on dir
909 * 3. We can't remove anything from append-only dir
910 * 4. We can't do anything with immutable dir (done in permission())
911 * 5. If the sticky bit on dir is set we should either
912 * a. be owner of dir, or
913 * b. be owner of victim, or
914 * c. have CAP_FOWNER capability
915 * 6. If the victim is append-only or immutable we can't do anything with
916 * links pointing to it.
917 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
918 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
919 * 9. We can't remove a root or mountpoint.
920 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
921 * nfs_async_unlink().
922 */
923
btrfs_may_delete(struct mnt_idmap * idmap,struct inode * dir,struct dentry * victim,int isdir)924 static int btrfs_may_delete(struct mnt_idmap *idmap,
925 struct inode *dir, struct dentry *victim, int isdir)
926 {
927 int error;
928
929 if (d_really_is_negative(victim))
930 return -ENOENT;
931
932 BUG_ON(d_inode(victim->d_parent) != dir);
933 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
934
935 error = inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
936 if (error)
937 return error;
938 if (IS_APPEND(dir))
939 return -EPERM;
940 if (check_sticky(idmap, dir, d_inode(victim)) ||
941 IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
942 IS_SWAPFILE(d_inode(victim)))
943 return -EPERM;
944 if (isdir) {
945 if (!d_is_dir(victim))
946 return -ENOTDIR;
947 if (IS_ROOT(victim))
948 return -EBUSY;
949 } else if (d_is_dir(victim))
950 return -EISDIR;
951 if (IS_DEADDIR(dir))
952 return -ENOENT;
953 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
954 return -EBUSY;
955 return 0;
956 }
957
958 /* copy of may_create in fs/namei.c() */
btrfs_may_create(struct mnt_idmap * idmap,struct inode * dir,struct dentry * child)959 static inline int btrfs_may_create(struct mnt_idmap *idmap,
960 struct inode *dir, struct dentry *child)
961 {
962 if (d_really_is_positive(child))
963 return -EEXIST;
964 if (IS_DEADDIR(dir))
965 return -ENOENT;
966 if (!fsuidgid_has_mapping(dir->i_sb, idmap))
967 return -EOVERFLOW;
968 return inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
969 }
970
971 /*
972 * Create a new subvolume below @parent. This is largely modeled after
973 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
974 * inside this filesystem so it's quite a bit simpler.
975 */
btrfs_mksubvol(const struct path * parent,struct mnt_idmap * idmap,const char * name,int namelen,struct btrfs_root * snap_src,bool readonly,struct btrfs_qgroup_inherit * inherit)976 static noinline int btrfs_mksubvol(const struct path *parent,
977 struct mnt_idmap *idmap,
978 const char *name, int namelen,
979 struct btrfs_root *snap_src,
980 bool readonly,
981 struct btrfs_qgroup_inherit *inherit)
982 {
983 struct inode *dir = d_inode(parent->dentry);
984 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
985 struct dentry *dentry;
986 struct fscrypt_str name_str = FSTR_INIT((char *)name, namelen);
987 int error;
988
989 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
990 if (error == -EINTR)
991 return error;
992
993 dentry = lookup_one(idmap, name, parent->dentry, namelen);
994 error = PTR_ERR(dentry);
995 if (IS_ERR(dentry))
996 goto out_unlock;
997
998 error = btrfs_may_create(idmap, dir, dentry);
999 if (error)
1000 goto out_dput;
1001
1002 /*
1003 * even if this name doesn't exist, we may get hash collisions.
1004 * check for them now when we can safely fail
1005 */
1006 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
1007 dir->i_ino, &name_str);
1008 if (error)
1009 goto out_dput;
1010
1011 down_read(&fs_info->subvol_sem);
1012
1013 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
1014 goto out_up_read;
1015
1016 if (snap_src)
1017 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
1018 else
1019 error = create_subvol(idmap, dir, dentry, inherit);
1020
1021 if (!error)
1022 fsnotify_mkdir(dir, dentry);
1023 out_up_read:
1024 up_read(&fs_info->subvol_sem);
1025 out_dput:
1026 dput(dentry);
1027 out_unlock:
1028 btrfs_inode_unlock(BTRFS_I(dir), 0);
1029 return error;
1030 }
1031
btrfs_mksnapshot(const struct path * parent,struct mnt_idmap * idmap,const char * name,int namelen,struct btrfs_root * root,bool readonly,struct btrfs_qgroup_inherit * inherit)1032 static noinline int btrfs_mksnapshot(const struct path *parent,
1033 struct mnt_idmap *idmap,
1034 const char *name, int namelen,
1035 struct btrfs_root *root,
1036 bool readonly,
1037 struct btrfs_qgroup_inherit *inherit)
1038 {
1039 int ret;
1040 bool snapshot_force_cow = false;
1041
1042 /*
1043 * Force new buffered writes to reserve space even when NOCOW is
1044 * possible. This is to avoid later writeback (running dealloc) to
1045 * fallback to COW mode and unexpectedly fail with ENOSPC.
1046 */
1047 btrfs_drew_read_lock(&root->snapshot_lock);
1048
1049 ret = btrfs_start_delalloc_snapshot(root, false);
1050 if (ret)
1051 goto out;
1052
1053 /*
1054 * All previous writes have started writeback in NOCOW mode, so now
1055 * we force future writes to fallback to COW mode during snapshot
1056 * creation.
1057 */
1058 atomic_inc(&root->snapshot_force_cow);
1059 snapshot_force_cow = true;
1060
1061 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
1062
1063 ret = btrfs_mksubvol(parent, idmap, name, namelen,
1064 root, readonly, inherit);
1065 out:
1066 if (snapshot_force_cow)
1067 atomic_dec(&root->snapshot_force_cow);
1068 btrfs_drew_read_unlock(&root->snapshot_lock);
1069 return ret;
1070 }
1071
1072 /*
1073 * Try to start exclusive operation @type or cancel it if it's running.
1074 *
1075 * Return:
1076 * 0 - normal mode, newly claimed op started
1077 * >0 - normal mode, something else is running,
1078 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1079 * ECANCELED - cancel mode, successful cancel
1080 * ENOTCONN - cancel mode, operation not running anymore
1081 */
exclop_start_or_cancel_reloc(struct btrfs_fs_info * fs_info,enum btrfs_exclusive_operation type,bool cancel)1082 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1083 enum btrfs_exclusive_operation type, bool cancel)
1084 {
1085 if (!cancel) {
1086 /* Start normal op */
1087 if (!btrfs_exclop_start(fs_info, type))
1088 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1089 /* Exclusive operation is now claimed */
1090 return 0;
1091 }
1092
1093 /* Cancel running op */
1094 if (btrfs_exclop_start_try_lock(fs_info, type)) {
1095 /*
1096 * This blocks any exclop finish from setting it to NONE, so we
1097 * request cancellation. Either it runs and we will wait for it,
1098 * or it has finished and no waiting will happen.
1099 */
1100 atomic_inc(&fs_info->reloc_cancel_req);
1101 btrfs_exclop_start_unlock(fs_info);
1102
1103 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1104 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1105 TASK_INTERRUPTIBLE);
1106
1107 return -ECANCELED;
1108 }
1109
1110 /* Something else is running or none */
1111 return -ENOTCONN;
1112 }
1113
btrfs_ioctl_resize(struct file * file,void __user * arg)1114 static noinline int btrfs_ioctl_resize(struct file *file,
1115 void __user *arg)
1116 {
1117 BTRFS_DEV_LOOKUP_ARGS(args);
1118 struct inode *inode = file_inode(file);
1119 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1120 u64 new_size;
1121 u64 old_size;
1122 u64 devid = 1;
1123 struct btrfs_root *root = BTRFS_I(inode)->root;
1124 struct btrfs_ioctl_vol_args *vol_args;
1125 struct btrfs_trans_handle *trans;
1126 struct btrfs_device *device = NULL;
1127 char *sizestr;
1128 char *retptr;
1129 char *devstr = NULL;
1130 int ret = 0;
1131 int mod = 0;
1132 bool cancel;
1133
1134 if (!capable(CAP_SYS_ADMIN))
1135 return -EPERM;
1136
1137 ret = mnt_want_write_file(file);
1138 if (ret)
1139 return ret;
1140
1141 /*
1142 * Read the arguments before checking exclusivity to be able to
1143 * distinguish regular resize and cancel
1144 */
1145 vol_args = memdup_user(arg, sizeof(*vol_args));
1146 if (IS_ERR(vol_args)) {
1147 ret = PTR_ERR(vol_args);
1148 goto out_drop;
1149 }
1150 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1151 sizestr = vol_args->name;
1152 cancel = (strcmp("cancel", sizestr) == 0);
1153 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1154 if (ret)
1155 goto out_free;
1156 /* Exclusive operation is now claimed */
1157
1158 devstr = strchr(sizestr, ':');
1159 if (devstr) {
1160 sizestr = devstr + 1;
1161 *devstr = '\0';
1162 devstr = vol_args->name;
1163 ret = kstrtoull(devstr, 10, &devid);
1164 if (ret)
1165 goto out_finish;
1166 if (!devid) {
1167 ret = -EINVAL;
1168 goto out_finish;
1169 }
1170 btrfs_info(fs_info, "resizing devid %llu", devid);
1171 }
1172
1173 args.devid = devid;
1174 device = btrfs_find_device(fs_info->fs_devices, &args);
1175 if (!device) {
1176 btrfs_info(fs_info, "resizer unable to find device %llu",
1177 devid);
1178 ret = -ENODEV;
1179 goto out_finish;
1180 }
1181
1182 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1183 btrfs_info(fs_info,
1184 "resizer unable to apply on readonly device %llu",
1185 devid);
1186 ret = -EPERM;
1187 goto out_finish;
1188 }
1189
1190 if (!strcmp(sizestr, "max"))
1191 new_size = bdev_nr_bytes(device->bdev);
1192 else {
1193 if (sizestr[0] == '-') {
1194 mod = -1;
1195 sizestr++;
1196 } else if (sizestr[0] == '+') {
1197 mod = 1;
1198 sizestr++;
1199 }
1200 new_size = memparse(sizestr, &retptr);
1201 if (*retptr != '\0' || new_size == 0) {
1202 ret = -EINVAL;
1203 goto out_finish;
1204 }
1205 }
1206
1207 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1208 ret = -EPERM;
1209 goto out_finish;
1210 }
1211
1212 old_size = btrfs_device_get_total_bytes(device);
1213
1214 if (mod < 0) {
1215 if (new_size > old_size) {
1216 ret = -EINVAL;
1217 goto out_finish;
1218 }
1219 new_size = old_size - new_size;
1220 } else if (mod > 0) {
1221 if (new_size > ULLONG_MAX - old_size) {
1222 ret = -ERANGE;
1223 goto out_finish;
1224 }
1225 new_size = old_size + new_size;
1226 }
1227
1228 if (new_size < SZ_256M) {
1229 ret = -EINVAL;
1230 goto out_finish;
1231 }
1232 if (new_size > bdev_nr_bytes(device->bdev)) {
1233 ret = -EFBIG;
1234 goto out_finish;
1235 }
1236
1237 new_size = round_down(new_size, fs_info->sectorsize);
1238
1239 if (new_size > old_size) {
1240 trans = btrfs_start_transaction(root, 0);
1241 if (IS_ERR(trans)) {
1242 ret = PTR_ERR(trans);
1243 goto out_finish;
1244 }
1245 ret = btrfs_grow_device(trans, device, new_size);
1246 btrfs_commit_transaction(trans);
1247 } else if (new_size < old_size) {
1248 ret = btrfs_shrink_device(device, new_size);
1249 } /* equal, nothing need to do */
1250
1251 if (ret == 0 && new_size != old_size)
1252 btrfs_info_in_rcu(fs_info,
1253 "resize device %s (devid %llu) from %llu to %llu",
1254 btrfs_dev_name(device), device->devid,
1255 old_size, new_size);
1256 out_finish:
1257 btrfs_exclop_finish(fs_info);
1258 out_free:
1259 kfree(vol_args);
1260 out_drop:
1261 mnt_drop_write_file(file);
1262 return ret;
1263 }
1264
__btrfs_ioctl_snap_create(struct file * file,struct mnt_idmap * idmap,const char * name,unsigned long fd,int subvol,bool readonly,struct btrfs_qgroup_inherit * inherit)1265 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1266 struct mnt_idmap *idmap,
1267 const char *name, unsigned long fd, int subvol,
1268 bool readonly,
1269 struct btrfs_qgroup_inherit *inherit)
1270 {
1271 int namelen;
1272 int ret = 0;
1273
1274 if (!S_ISDIR(file_inode(file)->i_mode))
1275 return -ENOTDIR;
1276
1277 ret = mnt_want_write_file(file);
1278 if (ret)
1279 goto out;
1280
1281 namelen = strlen(name);
1282 if (strchr(name, '/')) {
1283 ret = -EINVAL;
1284 goto out_drop_write;
1285 }
1286
1287 if (name[0] == '.' &&
1288 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1289 ret = -EEXIST;
1290 goto out_drop_write;
1291 }
1292
1293 if (subvol) {
1294 ret = btrfs_mksubvol(&file->f_path, idmap, name,
1295 namelen, NULL, readonly, inherit);
1296 } else {
1297 struct fd src = fdget(fd);
1298 struct inode *src_inode;
1299 if (!src.file) {
1300 ret = -EINVAL;
1301 goto out_drop_write;
1302 }
1303
1304 src_inode = file_inode(src.file);
1305 if (src_inode->i_sb != file_inode(file)->i_sb) {
1306 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1307 "Snapshot src from another FS");
1308 ret = -EXDEV;
1309 } else if (!inode_owner_or_capable(idmap, src_inode)) {
1310 /*
1311 * Subvolume creation is not restricted, but snapshots
1312 * are limited to own subvolumes only
1313 */
1314 ret = -EPERM;
1315 } else if (btrfs_ino(BTRFS_I(src_inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1316 /*
1317 * Snapshots must be made with the src_inode referring
1318 * to the subvolume inode, otherwise the permission
1319 * checking above is useless because we may have
1320 * permission on a lower directory but not the subvol
1321 * itself.
1322 */
1323 ret = -EINVAL;
1324 } else {
1325 ret = btrfs_mksnapshot(&file->f_path, idmap,
1326 name, namelen,
1327 BTRFS_I(src_inode)->root,
1328 readonly, inherit);
1329 }
1330 fdput(src);
1331 }
1332 out_drop_write:
1333 mnt_drop_write_file(file);
1334 out:
1335 return ret;
1336 }
1337
btrfs_ioctl_snap_create(struct file * file,void __user * arg,int subvol)1338 static noinline int btrfs_ioctl_snap_create(struct file *file,
1339 void __user *arg, int subvol)
1340 {
1341 struct btrfs_ioctl_vol_args *vol_args;
1342 int ret;
1343
1344 if (!S_ISDIR(file_inode(file)->i_mode))
1345 return -ENOTDIR;
1346
1347 vol_args = memdup_user(arg, sizeof(*vol_args));
1348 if (IS_ERR(vol_args))
1349 return PTR_ERR(vol_args);
1350 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1351
1352 ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1353 vol_args->name, vol_args->fd, subvol,
1354 false, NULL);
1355
1356 kfree(vol_args);
1357 return ret;
1358 }
1359
btrfs_ioctl_snap_create_v2(struct file * file,void __user * arg,int subvol)1360 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1361 void __user *arg, int subvol)
1362 {
1363 struct btrfs_ioctl_vol_args_v2 *vol_args;
1364 int ret;
1365 bool readonly = false;
1366 struct btrfs_qgroup_inherit *inherit = NULL;
1367
1368 if (!S_ISDIR(file_inode(file)->i_mode))
1369 return -ENOTDIR;
1370
1371 vol_args = memdup_user(arg, sizeof(*vol_args));
1372 if (IS_ERR(vol_args))
1373 return PTR_ERR(vol_args);
1374 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1375
1376 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1377 ret = -EOPNOTSUPP;
1378 goto free_args;
1379 }
1380
1381 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1382 readonly = true;
1383 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1384 u64 nums;
1385
1386 if (vol_args->size < sizeof(*inherit) ||
1387 vol_args->size > PAGE_SIZE) {
1388 ret = -EINVAL;
1389 goto free_args;
1390 }
1391 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1392 if (IS_ERR(inherit)) {
1393 ret = PTR_ERR(inherit);
1394 goto free_args;
1395 }
1396
1397 if (inherit->num_qgroups > PAGE_SIZE ||
1398 inherit->num_ref_copies > PAGE_SIZE ||
1399 inherit->num_excl_copies > PAGE_SIZE) {
1400 ret = -EINVAL;
1401 goto free_inherit;
1402 }
1403
1404 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
1405 2 * inherit->num_excl_copies;
1406 if (vol_args->size != struct_size(inherit, qgroups, nums)) {
1407 ret = -EINVAL;
1408 goto free_inherit;
1409 }
1410 }
1411
1412 ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1413 vol_args->name, vol_args->fd, subvol,
1414 readonly, inherit);
1415 if (ret)
1416 goto free_inherit;
1417 free_inherit:
1418 kfree(inherit);
1419 free_args:
1420 kfree(vol_args);
1421 return ret;
1422 }
1423
btrfs_ioctl_subvol_getflags(struct inode * inode,void __user * arg)1424 static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode,
1425 void __user *arg)
1426 {
1427 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1428 struct btrfs_root *root = BTRFS_I(inode)->root;
1429 int ret = 0;
1430 u64 flags = 0;
1431
1432 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1433 return -EINVAL;
1434
1435 down_read(&fs_info->subvol_sem);
1436 if (btrfs_root_readonly(root))
1437 flags |= BTRFS_SUBVOL_RDONLY;
1438 up_read(&fs_info->subvol_sem);
1439
1440 if (copy_to_user(arg, &flags, sizeof(flags)))
1441 ret = -EFAULT;
1442
1443 return ret;
1444 }
1445
btrfs_ioctl_subvol_setflags(struct file * file,void __user * arg)1446 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1447 void __user *arg)
1448 {
1449 struct inode *inode = file_inode(file);
1450 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1451 struct btrfs_root *root = BTRFS_I(inode)->root;
1452 struct btrfs_trans_handle *trans;
1453 u64 root_flags;
1454 u64 flags;
1455 int ret = 0;
1456
1457 if (!inode_owner_or_capable(file_mnt_idmap(file), inode))
1458 return -EPERM;
1459
1460 ret = mnt_want_write_file(file);
1461 if (ret)
1462 goto out;
1463
1464 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1465 ret = -EINVAL;
1466 goto out_drop_write;
1467 }
1468
1469 if (copy_from_user(&flags, arg, sizeof(flags))) {
1470 ret = -EFAULT;
1471 goto out_drop_write;
1472 }
1473
1474 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1475 ret = -EOPNOTSUPP;
1476 goto out_drop_write;
1477 }
1478
1479 down_write(&fs_info->subvol_sem);
1480
1481 /* nothing to do */
1482 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1483 goto out_drop_sem;
1484
1485 root_flags = btrfs_root_flags(&root->root_item);
1486 if (flags & BTRFS_SUBVOL_RDONLY) {
1487 btrfs_set_root_flags(&root->root_item,
1488 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1489 } else {
1490 /*
1491 * Block RO -> RW transition if this subvolume is involved in
1492 * send
1493 */
1494 spin_lock(&root->root_item_lock);
1495 if (root->send_in_progress == 0) {
1496 btrfs_set_root_flags(&root->root_item,
1497 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1498 spin_unlock(&root->root_item_lock);
1499 } else {
1500 spin_unlock(&root->root_item_lock);
1501 btrfs_warn(fs_info,
1502 "Attempt to set subvolume %llu read-write during send",
1503 root->root_key.objectid);
1504 ret = -EPERM;
1505 goto out_drop_sem;
1506 }
1507 }
1508
1509 trans = btrfs_start_transaction(root, 1);
1510 if (IS_ERR(trans)) {
1511 ret = PTR_ERR(trans);
1512 goto out_reset;
1513 }
1514
1515 ret = btrfs_update_root(trans, fs_info->tree_root,
1516 &root->root_key, &root->root_item);
1517 if (ret < 0) {
1518 btrfs_end_transaction(trans);
1519 goto out_reset;
1520 }
1521
1522 ret = btrfs_commit_transaction(trans);
1523
1524 out_reset:
1525 if (ret)
1526 btrfs_set_root_flags(&root->root_item, root_flags);
1527 out_drop_sem:
1528 up_write(&fs_info->subvol_sem);
1529 out_drop_write:
1530 mnt_drop_write_file(file);
1531 out:
1532 return ret;
1533 }
1534
key_in_sk(struct btrfs_key * key,struct btrfs_ioctl_search_key * sk)1535 static noinline int key_in_sk(struct btrfs_key *key,
1536 struct btrfs_ioctl_search_key *sk)
1537 {
1538 struct btrfs_key test;
1539 int ret;
1540
1541 test.objectid = sk->min_objectid;
1542 test.type = sk->min_type;
1543 test.offset = sk->min_offset;
1544
1545 ret = btrfs_comp_cpu_keys(key, &test);
1546 if (ret < 0)
1547 return 0;
1548
1549 test.objectid = sk->max_objectid;
1550 test.type = sk->max_type;
1551 test.offset = sk->max_offset;
1552
1553 ret = btrfs_comp_cpu_keys(key, &test);
1554 if (ret > 0)
1555 return 0;
1556 return 1;
1557 }
1558
copy_to_sk(struct btrfs_path * path,struct btrfs_key * key,struct btrfs_ioctl_search_key * sk,u64 * buf_size,char __user * ubuf,unsigned long * sk_offset,int * num_found)1559 static noinline int copy_to_sk(struct btrfs_path *path,
1560 struct btrfs_key *key,
1561 struct btrfs_ioctl_search_key *sk,
1562 u64 *buf_size,
1563 char __user *ubuf,
1564 unsigned long *sk_offset,
1565 int *num_found)
1566 {
1567 u64 found_transid;
1568 struct extent_buffer *leaf;
1569 struct btrfs_ioctl_search_header sh;
1570 struct btrfs_key test;
1571 unsigned long item_off;
1572 unsigned long item_len;
1573 int nritems;
1574 int i;
1575 int slot;
1576 int ret = 0;
1577
1578 leaf = path->nodes[0];
1579 slot = path->slots[0];
1580 nritems = btrfs_header_nritems(leaf);
1581
1582 if (btrfs_header_generation(leaf) > sk->max_transid) {
1583 i = nritems;
1584 goto advance_key;
1585 }
1586 found_transid = btrfs_header_generation(leaf);
1587
1588 for (i = slot; i < nritems; i++) {
1589 item_off = btrfs_item_ptr_offset(leaf, i);
1590 item_len = btrfs_item_size(leaf, i);
1591
1592 btrfs_item_key_to_cpu(leaf, key, i);
1593 if (!key_in_sk(key, sk))
1594 continue;
1595
1596 if (sizeof(sh) + item_len > *buf_size) {
1597 if (*num_found) {
1598 ret = 1;
1599 goto out;
1600 }
1601
1602 /*
1603 * return one empty item back for v1, which does not
1604 * handle -EOVERFLOW
1605 */
1606
1607 *buf_size = sizeof(sh) + item_len;
1608 item_len = 0;
1609 ret = -EOVERFLOW;
1610 }
1611
1612 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1613 ret = 1;
1614 goto out;
1615 }
1616
1617 sh.objectid = key->objectid;
1618 sh.offset = key->offset;
1619 sh.type = key->type;
1620 sh.len = item_len;
1621 sh.transid = found_transid;
1622
1623 /*
1624 * Copy search result header. If we fault then loop again so we
1625 * can fault in the pages and -EFAULT there if there's a
1626 * problem. Otherwise we'll fault and then copy the buffer in
1627 * properly this next time through
1628 */
1629 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
1630 ret = 0;
1631 goto out;
1632 }
1633
1634 *sk_offset += sizeof(sh);
1635
1636 if (item_len) {
1637 char __user *up = ubuf + *sk_offset;
1638 /*
1639 * Copy the item, same behavior as above, but reset the
1640 * * sk_offset so we copy the full thing again.
1641 */
1642 if (read_extent_buffer_to_user_nofault(leaf, up,
1643 item_off, item_len)) {
1644 ret = 0;
1645 *sk_offset -= sizeof(sh);
1646 goto out;
1647 }
1648
1649 *sk_offset += item_len;
1650 }
1651 (*num_found)++;
1652
1653 if (ret) /* -EOVERFLOW from above */
1654 goto out;
1655
1656 if (*num_found >= sk->nr_items) {
1657 ret = 1;
1658 goto out;
1659 }
1660 }
1661 advance_key:
1662 ret = 0;
1663 test.objectid = sk->max_objectid;
1664 test.type = sk->max_type;
1665 test.offset = sk->max_offset;
1666 if (btrfs_comp_cpu_keys(key, &test) >= 0)
1667 ret = 1;
1668 else if (key->offset < (u64)-1)
1669 key->offset++;
1670 else if (key->type < (u8)-1) {
1671 key->offset = 0;
1672 key->type++;
1673 } else if (key->objectid < (u64)-1) {
1674 key->offset = 0;
1675 key->type = 0;
1676 key->objectid++;
1677 } else
1678 ret = 1;
1679 out:
1680 /*
1681 * 0: all items from this leaf copied, continue with next
1682 * 1: * more items can be copied, but unused buffer is too small
1683 * * all items were found
1684 * Either way, it will stops the loop which iterates to the next
1685 * leaf
1686 * -EOVERFLOW: item was to large for buffer
1687 * -EFAULT: could not copy extent buffer back to userspace
1688 */
1689 return ret;
1690 }
1691
search_ioctl(struct inode * inode,struct btrfs_ioctl_search_key * sk,u64 * buf_size,char __user * ubuf)1692 static noinline int search_ioctl(struct inode *inode,
1693 struct btrfs_ioctl_search_key *sk,
1694 u64 *buf_size,
1695 char __user *ubuf)
1696 {
1697 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
1698 struct btrfs_root *root;
1699 struct btrfs_key key;
1700 struct btrfs_path *path;
1701 int ret;
1702 int num_found = 0;
1703 unsigned long sk_offset = 0;
1704
1705 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
1706 *buf_size = sizeof(struct btrfs_ioctl_search_header);
1707 return -EOVERFLOW;
1708 }
1709
1710 path = btrfs_alloc_path();
1711 if (!path)
1712 return -ENOMEM;
1713
1714 if (sk->tree_id == 0) {
1715 /* search the root of the inode that was passed */
1716 root = btrfs_grab_root(BTRFS_I(inode)->root);
1717 } else {
1718 root = btrfs_get_fs_root(info, sk->tree_id, true);
1719 if (IS_ERR(root)) {
1720 btrfs_free_path(path);
1721 return PTR_ERR(root);
1722 }
1723 }
1724
1725 key.objectid = sk->min_objectid;
1726 key.type = sk->min_type;
1727 key.offset = sk->min_offset;
1728
1729 while (1) {
1730 ret = -EFAULT;
1731 /*
1732 * Ensure that the whole user buffer is faulted in at sub-page
1733 * granularity, otherwise the loop may live-lock.
1734 */
1735 if (fault_in_subpage_writeable(ubuf + sk_offset,
1736 *buf_size - sk_offset))
1737 break;
1738
1739 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
1740 if (ret != 0) {
1741 if (ret > 0)
1742 ret = 0;
1743 goto err;
1744 }
1745 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
1746 &sk_offset, &num_found);
1747 btrfs_release_path(path);
1748 if (ret)
1749 break;
1750
1751 }
1752 if (ret > 0)
1753 ret = 0;
1754 err:
1755 sk->nr_items = num_found;
1756 btrfs_put_root(root);
1757 btrfs_free_path(path);
1758 return ret;
1759 }
1760
btrfs_ioctl_tree_search(struct inode * inode,void __user * argp)1761 static noinline int btrfs_ioctl_tree_search(struct inode *inode,
1762 void __user *argp)
1763 {
1764 struct btrfs_ioctl_search_args __user *uargs = argp;
1765 struct btrfs_ioctl_search_key sk;
1766 int ret;
1767 u64 buf_size;
1768
1769 if (!capable(CAP_SYS_ADMIN))
1770 return -EPERM;
1771
1772 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
1773 return -EFAULT;
1774
1775 buf_size = sizeof(uargs->buf);
1776
1777 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
1778
1779 /*
1780 * In the origin implementation an overflow is handled by returning a
1781 * search header with a len of zero, so reset ret.
1782 */
1783 if (ret == -EOVERFLOW)
1784 ret = 0;
1785
1786 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
1787 ret = -EFAULT;
1788 return ret;
1789 }
1790
btrfs_ioctl_tree_search_v2(struct inode * inode,void __user * argp)1791 static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode,
1792 void __user *argp)
1793 {
1794 struct btrfs_ioctl_search_args_v2 __user *uarg = argp;
1795 struct btrfs_ioctl_search_args_v2 args;
1796 int ret;
1797 u64 buf_size;
1798 const u64 buf_limit = SZ_16M;
1799
1800 if (!capable(CAP_SYS_ADMIN))
1801 return -EPERM;
1802
1803 /* copy search header and buffer size */
1804 if (copy_from_user(&args, uarg, sizeof(args)))
1805 return -EFAULT;
1806
1807 buf_size = args.buf_size;
1808
1809 /* limit result size to 16MB */
1810 if (buf_size > buf_limit)
1811 buf_size = buf_limit;
1812
1813 ret = search_ioctl(inode, &args.key, &buf_size,
1814 (char __user *)(&uarg->buf[0]));
1815 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
1816 ret = -EFAULT;
1817 else if (ret == -EOVERFLOW &&
1818 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
1819 ret = -EFAULT;
1820
1821 return ret;
1822 }
1823
1824 /*
1825 * Search INODE_REFs to identify path name of 'dirid' directory
1826 * in a 'tree_id' tree. and sets path name to 'name'.
1827 */
btrfs_search_path_in_tree(struct btrfs_fs_info * info,u64 tree_id,u64 dirid,char * name)1828 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1829 u64 tree_id, u64 dirid, char *name)
1830 {
1831 struct btrfs_root *root;
1832 struct btrfs_key key;
1833 char *ptr;
1834 int ret = -1;
1835 int slot;
1836 int len;
1837 int total_len = 0;
1838 struct btrfs_inode_ref *iref;
1839 struct extent_buffer *l;
1840 struct btrfs_path *path;
1841
1842 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1843 name[0]='\0';
1844 return 0;
1845 }
1846
1847 path = btrfs_alloc_path();
1848 if (!path)
1849 return -ENOMEM;
1850
1851 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
1852
1853 root = btrfs_get_fs_root(info, tree_id, true);
1854 if (IS_ERR(root)) {
1855 ret = PTR_ERR(root);
1856 root = NULL;
1857 goto out;
1858 }
1859
1860 key.objectid = dirid;
1861 key.type = BTRFS_INODE_REF_KEY;
1862 key.offset = (u64)-1;
1863
1864 while (1) {
1865 ret = btrfs_search_backwards(root, &key, path);
1866 if (ret < 0)
1867 goto out;
1868 else if (ret > 0) {
1869 ret = -ENOENT;
1870 goto out;
1871 }
1872
1873 l = path->nodes[0];
1874 slot = path->slots[0];
1875
1876 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1877 len = btrfs_inode_ref_name_len(l, iref);
1878 ptr -= len + 1;
1879 total_len += len + 1;
1880 if (ptr < name) {
1881 ret = -ENAMETOOLONG;
1882 goto out;
1883 }
1884
1885 *(ptr + len) = '/';
1886 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
1887
1888 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1889 break;
1890
1891 btrfs_release_path(path);
1892 key.objectid = key.offset;
1893 key.offset = (u64)-1;
1894 dirid = key.objectid;
1895 }
1896 memmove(name, ptr, total_len);
1897 name[total_len] = '\0';
1898 ret = 0;
1899 out:
1900 btrfs_put_root(root);
1901 btrfs_free_path(path);
1902 return ret;
1903 }
1904
btrfs_search_path_in_tree_user(struct mnt_idmap * idmap,struct inode * inode,struct btrfs_ioctl_ino_lookup_user_args * args)1905 static int btrfs_search_path_in_tree_user(struct mnt_idmap *idmap,
1906 struct inode *inode,
1907 struct btrfs_ioctl_ino_lookup_user_args *args)
1908 {
1909 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
1910 struct super_block *sb = inode->i_sb;
1911 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
1912 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
1913 u64 dirid = args->dirid;
1914 unsigned long item_off;
1915 unsigned long item_len;
1916 struct btrfs_inode_ref *iref;
1917 struct btrfs_root_ref *rref;
1918 struct btrfs_root *root = NULL;
1919 struct btrfs_path *path;
1920 struct btrfs_key key, key2;
1921 struct extent_buffer *leaf;
1922 struct inode *temp_inode;
1923 char *ptr;
1924 int slot;
1925 int len;
1926 int total_len = 0;
1927 int ret;
1928
1929 path = btrfs_alloc_path();
1930 if (!path)
1931 return -ENOMEM;
1932
1933 /*
1934 * If the bottom subvolume does not exist directly under upper_limit,
1935 * construct the path in from the bottom up.
1936 */
1937 if (dirid != upper_limit.objectid) {
1938 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
1939
1940 root = btrfs_get_fs_root(fs_info, treeid, true);
1941 if (IS_ERR(root)) {
1942 ret = PTR_ERR(root);
1943 goto out;
1944 }
1945
1946 key.objectid = dirid;
1947 key.type = BTRFS_INODE_REF_KEY;
1948 key.offset = (u64)-1;
1949 while (1) {
1950 ret = btrfs_search_backwards(root, &key, path);
1951 if (ret < 0)
1952 goto out_put;
1953 else if (ret > 0) {
1954 ret = -ENOENT;
1955 goto out_put;
1956 }
1957
1958 leaf = path->nodes[0];
1959 slot = path->slots[0];
1960
1961 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
1962 len = btrfs_inode_ref_name_len(leaf, iref);
1963 ptr -= len + 1;
1964 total_len += len + 1;
1965 if (ptr < args->path) {
1966 ret = -ENAMETOOLONG;
1967 goto out_put;
1968 }
1969
1970 *(ptr + len) = '/';
1971 read_extent_buffer(leaf, ptr,
1972 (unsigned long)(iref + 1), len);
1973
1974 /* Check the read+exec permission of this directory */
1975 ret = btrfs_previous_item(root, path, dirid,
1976 BTRFS_INODE_ITEM_KEY);
1977 if (ret < 0) {
1978 goto out_put;
1979 } else if (ret > 0) {
1980 ret = -ENOENT;
1981 goto out_put;
1982 }
1983
1984 leaf = path->nodes[0];
1985 slot = path->slots[0];
1986 btrfs_item_key_to_cpu(leaf, &key2, slot);
1987 if (key2.objectid != dirid) {
1988 ret = -ENOENT;
1989 goto out_put;
1990 }
1991
1992 /*
1993 * We don't need the path anymore, so release it and
1994 * avoid deadlocks and lockdep warnings in case
1995 * btrfs_iget() needs to lookup the inode from its root
1996 * btree and lock the same leaf.
1997 */
1998 btrfs_release_path(path);
1999 temp_inode = btrfs_iget(sb, key2.objectid, root);
2000 if (IS_ERR(temp_inode)) {
2001 ret = PTR_ERR(temp_inode);
2002 goto out_put;
2003 }
2004 ret = inode_permission(idmap, temp_inode,
2005 MAY_READ | MAY_EXEC);
2006 iput(temp_inode);
2007 if (ret) {
2008 ret = -EACCES;
2009 goto out_put;
2010 }
2011
2012 if (key.offset == upper_limit.objectid)
2013 break;
2014 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2015 ret = -EACCES;
2016 goto out_put;
2017 }
2018
2019 key.objectid = key.offset;
2020 key.offset = (u64)-1;
2021 dirid = key.objectid;
2022 }
2023
2024 memmove(args->path, ptr, total_len);
2025 args->path[total_len] = '\0';
2026 btrfs_put_root(root);
2027 root = NULL;
2028 btrfs_release_path(path);
2029 }
2030
2031 /* Get the bottom subvolume's name from ROOT_REF */
2032 key.objectid = treeid;
2033 key.type = BTRFS_ROOT_REF_KEY;
2034 key.offset = args->treeid;
2035 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2036 if (ret < 0) {
2037 goto out;
2038 } else if (ret > 0) {
2039 ret = -ENOENT;
2040 goto out;
2041 }
2042
2043 leaf = path->nodes[0];
2044 slot = path->slots[0];
2045 btrfs_item_key_to_cpu(leaf, &key, slot);
2046
2047 item_off = btrfs_item_ptr_offset(leaf, slot);
2048 item_len = btrfs_item_size(leaf, slot);
2049 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2050 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2051 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2052 ret = -EINVAL;
2053 goto out;
2054 }
2055
2056 /* Copy subvolume's name */
2057 item_off += sizeof(struct btrfs_root_ref);
2058 item_len -= sizeof(struct btrfs_root_ref);
2059 read_extent_buffer(leaf, args->name, item_off, item_len);
2060 args->name[item_len] = 0;
2061
2062 out_put:
2063 btrfs_put_root(root);
2064 out:
2065 btrfs_free_path(path);
2066 return ret;
2067 }
2068
btrfs_ioctl_ino_lookup(struct btrfs_root * root,void __user * argp)2069 static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root,
2070 void __user *argp)
2071 {
2072 struct btrfs_ioctl_ino_lookup_args *args;
2073 int ret = 0;
2074
2075 args = memdup_user(argp, sizeof(*args));
2076 if (IS_ERR(args))
2077 return PTR_ERR(args);
2078
2079 /*
2080 * Unprivileged query to obtain the containing subvolume root id. The
2081 * path is reset so it's consistent with btrfs_search_path_in_tree.
2082 */
2083 if (args->treeid == 0)
2084 args->treeid = root->root_key.objectid;
2085
2086 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2087 args->name[0] = 0;
2088 goto out;
2089 }
2090
2091 if (!capable(CAP_SYS_ADMIN)) {
2092 ret = -EPERM;
2093 goto out;
2094 }
2095
2096 ret = btrfs_search_path_in_tree(root->fs_info,
2097 args->treeid, args->objectid,
2098 args->name);
2099
2100 out:
2101 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2102 ret = -EFAULT;
2103
2104 kfree(args);
2105 return ret;
2106 }
2107
2108 /*
2109 * Version of ino_lookup ioctl (unprivileged)
2110 *
2111 * The main differences from ino_lookup ioctl are:
2112 *
2113 * 1. Read + Exec permission will be checked using inode_permission() during
2114 * path construction. -EACCES will be returned in case of failure.
2115 * 2. Path construction will be stopped at the inode number which corresponds
2116 * to the fd with which this ioctl is called. If constructed path does not
2117 * exist under fd's inode, -EACCES will be returned.
2118 * 3. The name of bottom subvolume is also searched and filled.
2119 */
btrfs_ioctl_ino_lookup_user(struct file * file,void __user * argp)2120 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2121 {
2122 struct btrfs_ioctl_ino_lookup_user_args *args;
2123 struct inode *inode;
2124 int ret;
2125
2126 args = memdup_user(argp, sizeof(*args));
2127 if (IS_ERR(args))
2128 return PTR_ERR(args);
2129
2130 inode = file_inode(file);
2131
2132 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2133 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2134 /*
2135 * The subvolume does not exist under fd with which this is
2136 * called
2137 */
2138 kfree(args);
2139 return -EACCES;
2140 }
2141
2142 ret = btrfs_search_path_in_tree_user(file_mnt_idmap(file), inode, args);
2143
2144 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2145 ret = -EFAULT;
2146
2147 kfree(args);
2148 return ret;
2149 }
2150
2151 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
btrfs_ioctl_get_subvol_info(struct inode * inode,void __user * argp)2152 static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp)
2153 {
2154 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2155 struct btrfs_fs_info *fs_info;
2156 struct btrfs_root *root;
2157 struct btrfs_path *path;
2158 struct btrfs_key key;
2159 struct btrfs_root_item *root_item;
2160 struct btrfs_root_ref *rref;
2161 struct extent_buffer *leaf;
2162 unsigned long item_off;
2163 unsigned long item_len;
2164 int slot;
2165 int ret = 0;
2166
2167 path = btrfs_alloc_path();
2168 if (!path)
2169 return -ENOMEM;
2170
2171 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2172 if (!subvol_info) {
2173 btrfs_free_path(path);
2174 return -ENOMEM;
2175 }
2176
2177 fs_info = BTRFS_I(inode)->root->fs_info;
2178
2179 /* Get root_item of inode's subvolume */
2180 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2181 root = btrfs_get_fs_root(fs_info, key.objectid, true);
2182 if (IS_ERR(root)) {
2183 ret = PTR_ERR(root);
2184 goto out_free;
2185 }
2186 root_item = &root->root_item;
2187
2188 subvol_info->treeid = key.objectid;
2189
2190 subvol_info->generation = btrfs_root_generation(root_item);
2191 subvol_info->flags = btrfs_root_flags(root_item);
2192
2193 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2194 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2195 BTRFS_UUID_SIZE);
2196 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2197 BTRFS_UUID_SIZE);
2198
2199 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2200 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2201 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2202
2203 subvol_info->otransid = btrfs_root_otransid(root_item);
2204 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2205 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2206
2207 subvol_info->stransid = btrfs_root_stransid(root_item);
2208 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2209 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2210
2211 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2212 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2213 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2214
2215 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2216 /* Search root tree for ROOT_BACKREF of this subvolume */
2217 key.type = BTRFS_ROOT_BACKREF_KEY;
2218 key.offset = 0;
2219 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2220 if (ret < 0) {
2221 goto out;
2222 } else if (path->slots[0] >=
2223 btrfs_header_nritems(path->nodes[0])) {
2224 ret = btrfs_next_leaf(fs_info->tree_root, path);
2225 if (ret < 0) {
2226 goto out;
2227 } else if (ret > 0) {
2228 ret = -EUCLEAN;
2229 goto out;
2230 }
2231 }
2232
2233 leaf = path->nodes[0];
2234 slot = path->slots[0];
2235 btrfs_item_key_to_cpu(leaf, &key, slot);
2236 if (key.objectid == subvol_info->treeid &&
2237 key.type == BTRFS_ROOT_BACKREF_KEY) {
2238 subvol_info->parent_id = key.offset;
2239
2240 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2241 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2242
2243 item_off = btrfs_item_ptr_offset(leaf, slot)
2244 + sizeof(struct btrfs_root_ref);
2245 item_len = btrfs_item_size(leaf, slot)
2246 - sizeof(struct btrfs_root_ref);
2247 read_extent_buffer(leaf, subvol_info->name,
2248 item_off, item_len);
2249 } else {
2250 ret = -ENOENT;
2251 goto out;
2252 }
2253 }
2254
2255 btrfs_free_path(path);
2256 path = NULL;
2257 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2258 ret = -EFAULT;
2259
2260 out:
2261 btrfs_put_root(root);
2262 out_free:
2263 btrfs_free_path(path);
2264 kfree(subvol_info);
2265 return ret;
2266 }
2267
2268 /*
2269 * Return ROOT_REF information of the subvolume containing this inode
2270 * except the subvolume name.
2271 */
btrfs_ioctl_get_subvol_rootref(struct btrfs_root * root,void __user * argp)2272 static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root,
2273 void __user *argp)
2274 {
2275 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2276 struct btrfs_root_ref *rref;
2277 struct btrfs_path *path;
2278 struct btrfs_key key;
2279 struct extent_buffer *leaf;
2280 u64 objectid;
2281 int slot;
2282 int ret;
2283 u8 found;
2284
2285 path = btrfs_alloc_path();
2286 if (!path)
2287 return -ENOMEM;
2288
2289 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2290 if (IS_ERR(rootrefs)) {
2291 btrfs_free_path(path);
2292 return PTR_ERR(rootrefs);
2293 }
2294
2295 objectid = root->root_key.objectid;
2296 key.objectid = objectid;
2297 key.type = BTRFS_ROOT_REF_KEY;
2298 key.offset = rootrefs->min_treeid;
2299 found = 0;
2300
2301 root = root->fs_info->tree_root;
2302 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2303 if (ret < 0) {
2304 goto out;
2305 } else if (path->slots[0] >=
2306 btrfs_header_nritems(path->nodes[0])) {
2307 ret = btrfs_next_leaf(root, path);
2308 if (ret < 0) {
2309 goto out;
2310 } else if (ret > 0) {
2311 ret = -EUCLEAN;
2312 goto out;
2313 }
2314 }
2315 while (1) {
2316 leaf = path->nodes[0];
2317 slot = path->slots[0];
2318
2319 btrfs_item_key_to_cpu(leaf, &key, slot);
2320 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2321 ret = 0;
2322 goto out;
2323 }
2324
2325 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2326 ret = -EOVERFLOW;
2327 goto out;
2328 }
2329
2330 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2331 rootrefs->rootref[found].treeid = key.offset;
2332 rootrefs->rootref[found].dirid =
2333 btrfs_root_ref_dirid(leaf, rref);
2334 found++;
2335
2336 ret = btrfs_next_item(root, path);
2337 if (ret < 0) {
2338 goto out;
2339 } else if (ret > 0) {
2340 ret = -EUCLEAN;
2341 goto out;
2342 }
2343 }
2344
2345 out:
2346 btrfs_free_path(path);
2347
2348 if (!ret || ret == -EOVERFLOW) {
2349 rootrefs->num_items = found;
2350 /* update min_treeid for next search */
2351 if (found)
2352 rootrefs->min_treeid =
2353 rootrefs->rootref[found - 1].treeid + 1;
2354 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2355 ret = -EFAULT;
2356 }
2357
2358 kfree(rootrefs);
2359
2360 return ret;
2361 }
2362
btrfs_ioctl_snap_destroy(struct file * file,void __user * arg,bool destroy_v2)2363 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2364 void __user *arg,
2365 bool destroy_v2)
2366 {
2367 struct dentry *parent = file->f_path.dentry;
2368 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2369 struct dentry *dentry;
2370 struct inode *dir = d_inode(parent);
2371 struct inode *inode;
2372 struct btrfs_root *root = BTRFS_I(dir)->root;
2373 struct btrfs_root *dest = NULL;
2374 struct btrfs_ioctl_vol_args *vol_args = NULL;
2375 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2376 struct mnt_idmap *idmap = file_mnt_idmap(file);
2377 char *subvol_name, *subvol_name_ptr = NULL;
2378 int subvol_namelen;
2379 int err = 0;
2380 bool destroy_parent = false;
2381
2382 /* We don't support snapshots with extent tree v2 yet. */
2383 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2384 btrfs_err(fs_info,
2385 "extent tree v2 doesn't support snapshot deletion yet");
2386 return -EOPNOTSUPP;
2387 }
2388
2389 if (destroy_v2) {
2390 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2391 if (IS_ERR(vol_args2))
2392 return PTR_ERR(vol_args2);
2393
2394 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2395 err = -EOPNOTSUPP;
2396 goto out;
2397 }
2398
2399 /*
2400 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2401 * name, same as v1 currently does.
2402 */
2403 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2404 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2405 subvol_name = vol_args2->name;
2406
2407 err = mnt_want_write_file(file);
2408 if (err)
2409 goto out;
2410 } else {
2411 struct inode *old_dir;
2412
2413 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2414 err = -EINVAL;
2415 goto out;
2416 }
2417
2418 err = mnt_want_write_file(file);
2419 if (err)
2420 goto out;
2421
2422 dentry = btrfs_get_dentry(fs_info->sb,
2423 BTRFS_FIRST_FREE_OBJECTID,
2424 vol_args2->subvolid, 0);
2425 if (IS_ERR(dentry)) {
2426 err = PTR_ERR(dentry);
2427 goto out_drop_write;
2428 }
2429
2430 /*
2431 * Change the default parent since the subvolume being
2432 * deleted can be outside of the current mount point.
2433 */
2434 parent = btrfs_get_parent(dentry);
2435
2436 /*
2437 * At this point dentry->d_name can point to '/' if the
2438 * subvolume we want to destroy is outsite of the
2439 * current mount point, so we need to release the
2440 * current dentry and execute the lookup to return a new
2441 * one with ->d_name pointing to the
2442 * <mount point>/subvol_name.
2443 */
2444 dput(dentry);
2445 if (IS_ERR(parent)) {
2446 err = PTR_ERR(parent);
2447 goto out_drop_write;
2448 }
2449 old_dir = dir;
2450 dir = d_inode(parent);
2451
2452 /*
2453 * If v2 was used with SPEC_BY_ID, a new parent was
2454 * allocated since the subvolume can be outside of the
2455 * current mount point. Later on we need to release this
2456 * new parent dentry.
2457 */
2458 destroy_parent = true;
2459
2460 /*
2461 * On idmapped mounts, deletion via subvolid is
2462 * restricted to subvolumes that are immediate
2463 * ancestors of the inode referenced by the file
2464 * descriptor in the ioctl. Otherwise the idmapping
2465 * could potentially be abused to delete subvolumes
2466 * anywhere in the filesystem the user wouldn't be able
2467 * to delete without an idmapped mount.
2468 */
2469 if (old_dir != dir && idmap != &nop_mnt_idmap) {
2470 err = -EOPNOTSUPP;
2471 goto free_parent;
2472 }
2473
2474 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2475 fs_info, vol_args2->subvolid);
2476 if (IS_ERR(subvol_name_ptr)) {
2477 err = PTR_ERR(subvol_name_ptr);
2478 goto free_parent;
2479 }
2480 /* subvol_name_ptr is already nul terminated */
2481 subvol_name = (char *)kbasename(subvol_name_ptr);
2482 }
2483 } else {
2484 vol_args = memdup_user(arg, sizeof(*vol_args));
2485 if (IS_ERR(vol_args))
2486 return PTR_ERR(vol_args);
2487
2488 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
2489 subvol_name = vol_args->name;
2490
2491 err = mnt_want_write_file(file);
2492 if (err)
2493 goto out;
2494 }
2495
2496 subvol_namelen = strlen(subvol_name);
2497
2498 if (strchr(subvol_name, '/') ||
2499 strncmp(subvol_name, "..", subvol_namelen) == 0) {
2500 err = -EINVAL;
2501 goto free_subvol_name;
2502 }
2503
2504 if (!S_ISDIR(dir->i_mode)) {
2505 err = -ENOTDIR;
2506 goto free_subvol_name;
2507 }
2508
2509 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2510 if (err == -EINTR)
2511 goto free_subvol_name;
2512 dentry = lookup_one(idmap, subvol_name, parent, subvol_namelen);
2513 if (IS_ERR(dentry)) {
2514 err = PTR_ERR(dentry);
2515 goto out_unlock_dir;
2516 }
2517
2518 if (d_really_is_negative(dentry)) {
2519 err = -ENOENT;
2520 goto out_dput;
2521 }
2522
2523 inode = d_inode(dentry);
2524 dest = BTRFS_I(inode)->root;
2525 if (!capable(CAP_SYS_ADMIN)) {
2526 /*
2527 * Regular user. Only allow this with a special mount
2528 * option, when the user has write+exec access to the
2529 * subvol root, and when rmdir(2) would have been
2530 * allowed.
2531 *
2532 * Note that this is _not_ check that the subvol is
2533 * empty or doesn't contain data that we wouldn't
2534 * otherwise be able to delete.
2535 *
2536 * Users who want to delete empty subvols should try
2537 * rmdir(2).
2538 */
2539 err = -EPERM;
2540 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2541 goto out_dput;
2542
2543 /*
2544 * Do not allow deletion if the parent dir is the same
2545 * as the dir to be deleted. That means the ioctl
2546 * must be called on the dentry referencing the root
2547 * of the subvol, not a random directory contained
2548 * within it.
2549 */
2550 err = -EINVAL;
2551 if (root == dest)
2552 goto out_dput;
2553
2554 err = inode_permission(idmap, inode, MAY_WRITE | MAY_EXEC);
2555 if (err)
2556 goto out_dput;
2557 }
2558
2559 /* check if subvolume may be deleted by a user */
2560 err = btrfs_may_delete(idmap, dir, dentry, 1);
2561 if (err)
2562 goto out_dput;
2563
2564 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2565 err = -EINVAL;
2566 goto out_dput;
2567 }
2568
2569 btrfs_inode_lock(BTRFS_I(inode), 0);
2570 err = btrfs_delete_subvolume(BTRFS_I(dir), dentry);
2571 btrfs_inode_unlock(BTRFS_I(inode), 0);
2572 if (!err)
2573 d_delete_notify(dir, dentry);
2574
2575 out_dput:
2576 dput(dentry);
2577 out_unlock_dir:
2578 btrfs_inode_unlock(BTRFS_I(dir), 0);
2579 free_subvol_name:
2580 kfree(subvol_name_ptr);
2581 free_parent:
2582 if (destroy_parent)
2583 dput(parent);
2584 out_drop_write:
2585 mnt_drop_write_file(file);
2586 out:
2587 kfree(vol_args2);
2588 kfree(vol_args);
2589 return err;
2590 }
2591
btrfs_ioctl_defrag(struct file * file,void __user * argp)2592 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2593 {
2594 struct inode *inode = file_inode(file);
2595 struct btrfs_root *root = BTRFS_I(inode)->root;
2596 struct btrfs_ioctl_defrag_range_args range = {0};
2597 int ret;
2598
2599 ret = mnt_want_write_file(file);
2600 if (ret)
2601 return ret;
2602
2603 if (btrfs_root_readonly(root)) {
2604 ret = -EROFS;
2605 goto out;
2606 }
2607
2608 switch (inode->i_mode & S_IFMT) {
2609 case S_IFDIR:
2610 if (!capable(CAP_SYS_ADMIN)) {
2611 ret = -EPERM;
2612 goto out;
2613 }
2614 ret = btrfs_defrag_root(root);
2615 break;
2616 case S_IFREG:
2617 /*
2618 * Note that this does not check the file descriptor for write
2619 * access. This prevents defragmenting executables that are
2620 * running and allows defrag on files open in read-only mode.
2621 */
2622 if (!capable(CAP_SYS_ADMIN) &&
2623 inode_permission(&nop_mnt_idmap, inode, MAY_WRITE)) {
2624 ret = -EPERM;
2625 goto out;
2626 }
2627
2628 if (argp) {
2629 if (copy_from_user(&range, argp, sizeof(range))) {
2630 ret = -EFAULT;
2631 goto out;
2632 }
2633 if (range.flags & ~BTRFS_DEFRAG_RANGE_FLAGS_SUPP) {
2634 ret = -EOPNOTSUPP;
2635 goto out;
2636 }
2637 /* compression requires us to start the IO */
2638 if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2639 range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
2640 range.extent_thresh = (u32)-1;
2641 }
2642 } else {
2643 /* the rest are all set to zero by kzalloc */
2644 range.len = (u64)-1;
2645 }
2646 ret = btrfs_defrag_file(file_inode(file), &file->f_ra,
2647 &range, BTRFS_OLDEST_GENERATION, 0);
2648 if (ret > 0)
2649 ret = 0;
2650 break;
2651 default:
2652 ret = -EINVAL;
2653 }
2654 out:
2655 mnt_drop_write_file(file);
2656 return ret;
2657 }
2658
btrfs_ioctl_add_dev(struct btrfs_fs_info * fs_info,void __user * arg)2659 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2660 {
2661 struct btrfs_ioctl_vol_args *vol_args;
2662 bool restore_op = false;
2663 int ret;
2664
2665 if (!capable(CAP_SYS_ADMIN))
2666 return -EPERM;
2667
2668 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2669 btrfs_err(fs_info, "device add not supported on extent tree v2 yet");
2670 return -EINVAL;
2671 }
2672
2673 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) {
2674 if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD))
2675 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2676
2677 /*
2678 * We can do the device add because we have a paused balanced,
2679 * change the exclusive op type and remember we should bring
2680 * back the paused balance
2681 */
2682 fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD;
2683 btrfs_exclop_start_unlock(fs_info);
2684 restore_op = true;
2685 }
2686
2687 vol_args = memdup_user(arg, sizeof(*vol_args));
2688 if (IS_ERR(vol_args)) {
2689 ret = PTR_ERR(vol_args);
2690 goto out;
2691 }
2692
2693 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2694 ret = btrfs_init_new_device(fs_info, vol_args->name);
2695
2696 if (!ret)
2697 btrfs_info(fs_info, "disk added %s", vol_args->name);
2698
2699 kfree(vol_args);
2700 out:
2701 if (restore_op)
2702 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
2703 else
2704 btrfs_exclop_finish(fs_info);
2705 return ret;
2706 }
2707
btrfs_ioctl_rm_dev_v2(struct file * file,void __user * arg)2708 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2709 {
2710 BTRFS_DEV_LOOKUP_ARGS(args);
2711 struct inode *inode = file_inode(file);
2712 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2713 struct btrfs_ioctl_vol_args_v2 *vol_args;
2714 struct block_device *bdev = NULL;
2715 void *holder;
2716 int ret;
2717 bool cancel = false;
2718
2719 if (!capable(CAP_SYS_ADMIN))
2720 return -EPERM;
2721
2722 vol_args = memdup_user(arg, sizeof(*vol_args));
2723 if (IS_ERR(vol_args))
2724 return PTR_ERR(vol_args);
2725
2726 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
2727 ret = -EOPNOTSUPP;
2728 goto out;
2729 }
2730
2731 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2732 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2733 args.devid = vol_args->devid;
2734 } else if (!strcmp("cancel", vol_args->name)) {
2735 cancel = true;
2736 } else {
2737 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2738 if (ret)
2739 goto out;
2740 }
2741
2742 ret = mnt_want_write_file(file);
2743 if (ret)
2744 goto out;
2745
2746 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2747 cancel);
2748 if (ret)
2749 goto err_drop;
2750
2751 /* Exclusive operation is now claimed */
2752 ret = btrfs_rm_device(fs_info, &args, &bdev, &holder);
2753
2754 btrfs_exclop_finish(fs_info);
2755
2756 if (!ret) {
2757 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2758 btrfs_info(fs_info, "device deleted: id %llu",
2759 vol_args->devid);
2760 else
2761 btrfs_info(fs_info, "device deleted: %s",
2762 vol_args->name);
2763 }
2764 err_drop:
2765 mnt_drop_write_file(file);
2766 if (bdev)
2767 blkdev_put(bdev, holder);
2768 out:
2769 btrfs_put_dev_args_from_path(&args);
2770 kfree(vol_args);
2771 return ret;
2772 }
2773
btrfs_ioctl_rm_dev(struct file * file,void __user * arg)2774 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2775 {
2776 BTRFS_DEV_LOOKUP_ARGS(args);
2777 struct inode *inode = file_inode(file);
2778 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2779 struct btrfs_ioctl_vol_args *vol_args;
2780 struct block_device *bdev = NULL;
2781 void *holder;
2782 int ret;
2783 bool cancel = false;
2784
2785 if (!capable(CAP_SYS_ADMIN))
2786 return -EPERM;
2787
2788 vol_args = memdup_user(arg, sizeof(*vol_args));
2789 if (IS_ERR(vol_args))
2790 return PTR_ERR(vol_args);
2791
2792 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2793 if (!strcmp("cancel", vol_args->name)) {
2794 cancel = true;
2795 } else {
2796 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2797 if (ret)
2798 goto out;
2799 }
2800
2801 ret = mnt_want_write_file(file);
2802 if (ret)
2803 goto out;
2804
2805 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2806 cancel);
2807 if (ret == 0) {
2808 ret = btrfs_rm_device(fs_info, &args, &bdev, &holder);
2809 if (!ret)
2810 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2811 btrfs_exclop_finish(fs_info);
2812 }
2813
2814 mnt_drop_write_file(file);
2815 if (bdev)
2816 blkdev_put(bdev, holder);
2817 out:
2818 btrfs_put_dev_args_from_path(&args);
2819 kfree(vol_args);
2820 return ret;
2821 }
2822
btrfs_ioctl_fs_info(struct btrfs_fs_info * fs_info,void __user * arg)2823 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2824 void __user *arg)
2825 {
2826 struct btrfs_ioctl_fs_info_args *fi_args;
2827 struct btrfs_device *device;
2828 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2829 u64 flags_in;
2830 int ret = 0;
2831
2832 fi_args = memdup_user(arg, sizeof(*fi_args));
2833 if (IS_ERR(fi_args))
2834 return PTR_ERR(fi_args);
2835
2836 flags_in = fi_args->flags;
2837 memset(fi_args, 0, sizeof(*fi_args));
2838
2839 rcu_read_lock();
2840 fi_args->num_devices = fs_devices->num_devices;
2841
2842 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2843 if (device->devid > fi_args->max_id)
2844 fi_args->max_id = device->devid;
2845 }
2846 rcu_read_unlock();
2847
2848 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
2849 fi_args->nodesize = fs_info->nodesize;
2850 fi_args->sectorsize = fs_info->sectorsize;
2851 fi_args->clone_alignment = fs_info->sectorsize;
2852
2853 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
2854 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
2855 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
2856 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
2857 }
2858
2859 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
2860 fi_args->generation = fs_info->generation;
2861 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
2862 }
2863
2864 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
2865 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
2866 sizeof(fi_args->metadata_uuid));
2867 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
2868 }
2869
2870 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2871 ret = -EFAULT;
2872
2873 kfree(fi_args);
2874 return ret;
2875 }
2876
btrfs_ioctl_dev_info(struct btrfs_fs_info * fs_info,void __user * arg)2877 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2878 void __user *arg)
2879 {
2880 BTRFS_DEV_LOOKUP_ARGS(args);
2881 struct btrfs_ioctl_dev_info_args *di_args;
2882 struct btrfs_device *dev;
2883 int ret = 0;
2884
2885 di_args = memdup_user(arg, sizeof(*di_args));
2886 if (IS_ERR(di_args))
2887 return PTR_ERR(di_args);
2888
2889 args.devid = di_args->devid;
2890 if (!btrfs_is_empty_uuid(di_args->uuid))
2891 args.uuid = di_args->uuid;
2892
2893 rcu_read_lock();
2894 dev = btrfs_find_device(fs_info->fs_devices, &args);
2895 if (!dev) {
2896 ret = -ENODEV;
2897 goto out;
2898 }
2899
2900 di_args->devid = dev->devid;
2901 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2902 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2903 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2904 memcpy(di_args->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
2905 if (dev->name)
2906 strscpy(di_args->path, btrfs_dev_name(dev), sizeof(di_args->path));
2907 else
2908 di_args->path[0] = '\0';
2909
2910 out:
2911 rcu_read_unlock();
2912 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2913 ret = -EFAULT;
2914
2915 kfree(di_args);
2916 return ret;
2917 }
2918
btrfs_ioctl_default_subvol(struct file * file,void __user * argp)2919 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2920 {
2921 struct inode *inode = file_inode(file);
2922 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2923 struct btrfs_root *root = BTRFS_I(inode)->root;
2924 struct btrfs_root *new_root;
2925 struct btrfs_dir_item *di;
2926 struct btrfs_trans_handle *trans;
2927 struct btrfs_path *path = NULL;
2928 struct btrfs_disk_key disk_key;
2929 struct fscrypt_str name = FSTR_INIT("default", 7);
2930 u64 objectid = 0;
2931 u64 dir_id;
2932 int ret;
2933
2934 if (!capable(CAP_SYS_ADMIN))
2935 return -EPERM;
2936
2937 ret = mnt_want_write_file(file);
2938 if (ret)
2939 return ret;
2940
2941 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2942 ret = -EFAULT;
2943 goto out;
2944 }
2945
2946 if (!objectid)
2947 objectid = BTRFS_FS_TREE_OBJECTID;
2948
2949 new_root = btrfs_get_fs_root(fs_info, objectid, true);
2950 if (IS_ERR(new_root)) {
2951 ret = PTR_ERR(new_root);
2952 goto out;
2953 }
2954 if (!is_fstree(new_root->root_key.objectid)) {
2955 ret = -ENOENT;
2956 goto out_free;
2957 }
2958
2959 path = btrfs_alloc_path();
2960 if (!path) {
2961 ret = -ENOMEM;
2962 goto out_free;
2963 }
2964
2965 trans = btrfs_start_transaction(root, 1);
2966 if (IS_ERR(trans)) {
2967 ret = PTR_ERR(trans);
2968 goto out_free;
2969 }
2970
2971 dir_id = btrfs_super_root_dir(fs_info->super_copy);
2972 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
2973 dir_id, &name, 1);
2974 if (IS_ERR_OR_NULL(di)) {
2975 btrfs_release_path(path);
2976 btrfs_end_transaction(trans);
2977 btrfs_err(fs_info,
2978 "Umm, you don't have the default diritem, this isn't going to work");
2979 ret = -ENOENT;
2980 goto out_free;
2981 }
2982
2983 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2984 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2985 btrfs_mark_buffer_dirty(trans, path->nodes[0]);
2986 btrfs_release_path(path);
2987
2988 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
2989 btrfs_end_transaction(trans);
2990 out_free:
2991 btrfs_put_root(new_root);
2992 btrfs_free_path(path);
2993 out:
2994 mnt_drop_write_file(file);
2995 return ret;
2996 }
2997
get_block_group_info(struct list_head * groups_list,struct btrfs_ioctl_space_info * space)2998 static void get_block_group_info(struct list_head *groups_list,
2999 struct btrfs_ioctl_space_info *space)
3000 {
3001 struct btrfs_block_group *block_group;
3002
3003 space->total_bytes = 0;
3004 space->used_bytes = 0;
3005 space->flags = 0;
3006 list_for_each_entry(block_group, groups_list, list) {
3007 space->flags = block_group->flags;
3008 space->total_bytes += block_group->length;
3009 space->used_bytes += block_group->used;
3010 }
3011 }
3012
btrfs_ioctl_space_info(struct btrfs_fs_info * fs_info,void __user * arg)3013 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3014 void __user *arg)
3015 {
3016 struct btrfs_ioctl_space_args space_args = { 0 };
3017 struct btrfs_ioctl_space_info space;
3018 struct btrfs_ioctl_space_info *dest;
3019 struct btrfs_ioctl_space_info *dest_orig;
3020 struct btrfs_ioctl_space_info __user *user_dest;
3021 struct btrfs_space_info *info;
3022 static const u64 types[] = {
3023 BTRFS_BLOCK_GROUP_DATA,
3024 BTRFS_BLOCK_GROUP_SYSTEM,
3025 BTRFS_BLOCK_GROUP_METADATA,
3026 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3027 };
3028 int num_types = 4;
3029 int alloc_size;
3030 int ret = 0;
3031 u64 slot_count = 0;
3032 int i, c;
3033
3034 if (copy_from_user(&space_args,
3035 (struct btrfs_ioctl_space_args __user *)arg,
3036 sizeof(space_args)))
3037 return -EFAULT;
3038
3039 for (i = 0; i < num_types; i++) {
3040 struct btrfs_space_info *tmp;
3041
3042 info = NULL;
3043 list_for_each_entry(tmp, &fs_info->space_info, list) {
3044 if (tmp->flags == types[i]) {
3045 info = tmp;
3046 break;
3047 }
3048 }
3049
3050 if (!info)
3051 continue;
3052
3053 down_read(&info->groups_sem);
3054 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3055 if (!list_empty(&info->block_groups[c]))
3056 slot_count++;
3057 }
3058 up_read(&info->groups_sem);
3059 }
3060
3061 /*
3062 * Global block reserve, exported as a space_info
3063 */
3064 slot_count++;
3065
3066 /* space_slots == 0 means they are asking for a count */
3067 if (space_args.space_slots == 0) {
3068 space_args.total_spaces = slot_count;
3069 goto out;
3070 }
3071
3072 slot_count = min_t(u64, space_args.space_slots, slot_count);
3073
3074 alloc_size = sizeof(*dest) * slot_count;
3075
3076 /* we generally have at most 6 or so space infos, one for each raid
3077 * level. So, a whole page should be more than enough for everyone
3078 */
3079 if (alloc_size > PAGE_SIZE)
3080 return -ENOMEM;
3081
3082 space_args.total_spaces = 0;
3083 dest = kmalloc(alloc_size, GFP_KERNEL);
3084 if (!dest)
3085 return -ENOMEM;
3086 dest_orig = dest;
3087
3088 /* now we have a buffer to copy into */
3089 for (i = 0; i < num_types; i++) {
3090 struct btrfs_space_info *tmp;
3091
3092 if (!slot_count)
3093 break;
3094
3095 info = NULL;
3096 list_for_each_entry(tmp, &fs_info->space_info, list) {
3097 if (tmp->flags == types[i]) {
3098 info = tmp;
3099 break;
3100 }
3101 }
3102
3103 if (!info)
3104 continue;
3105 down_read(&info->groups_sem);
3106 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3107 if (!list_empty(&info->block_groups[c])) {
3108 get_block_group_info(&info->block_groups[c],
3109 &space);
3110 memcpy(dest, &space, sizeof(space));
3111 dest++;
3112 space_args.total_spaces++;
3113 slot_count--;
3114 }
3115 if (!slot_count)
3116 break;
3117 }
3118 up_read(&info->groups_sem);
3119 }
3120
3121 /*
3122 * Add global block reserve
3123 */
3124 if (slot_count) {
3125 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3126
3127 spin_lock(&block_rsv->lock);
3128 space.total_bytes = block_rsv->size;
3129 space.used_bytes = block_rsv->size - block_rsv->reserved;
3130 spin_unlock(&block_rsv->lock);
3131 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3132 memcpy(dest, &space, sizeof(space));
3133 space_args.total_spaces++;
3134 }
3135
3136 user_dest = (struct btrfs_ioctl_space_info __user *)
3137 (arg + sizeof(struct btrfs_ioctl_space_args));
3138
3139 if (copy_to_user(user_dest, dest_orig, alloc_size))
3140 ret = -EFAULT;
3141
3142 kfree(dest_orig);
3143 out:
3144 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3145 ret = -EFAULT;
3146
3147 return ret;
3148 }
3149
btrfs_ioctl_start_sync(struct btrfs_root * root,void __user * argp)3150 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3151 void __user *argp)
3152 {
3153 struct btrfs_trans_handle *trans;
3154 u64 transid;
3155
3156 /*
3157 * Start orphan cleanup here for the given root in case it hasn't been
3158 * started already by other means. Errors are handled in the other
3159 * functions during transaction commit.
3160 */
3161 btrfs_orphan_cleanup(root);
3162
3163 trans = btrfs_attach_transaction_barrier(root);
3164 if (IS_ERR(trans)) {
3165 if (PTR_ERR(trans) != -ENOENT)
3166 return PTR_ERR(trans);
3167
3168 /* No running transaction, don't bother */
3169 transid = root->fs_info->last_trans_committed;
3170 goto out;
3171 }
3172 transid = trans->transid;
3173 btrfs_commit_transaction_async(trans);
3174 out:
3175 if (argp)
3176 if (copy_to_user(argp, &transid, sizeof(transid)))
3177 return -EFAULT;
3178 return 0;
3179 }
3180
btrfs_ioctl_wait_sync(struct btrfs_fs_info * fs_info,void __user * argp)3181 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3182 void __user *argp)
3183 {
3184 /* By default wait for the current transaction. */
3185 u64 transid = 0;
3186
3187 if (argp)
3188 if (copy_from_user(&transid, argp, sizeof(transid)))
3189 return -EFAULT;
3190
3191 return btrfs_wait_for_commit(fs_info, transid);
3192 }
3193
btrfs_ioctl_scrub(struct file * file,void __user * arg)3194 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3195 {
3196 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3197 struct btrfs_ioctl_scrub_args *sa;
3198 int ret;
3199
3200 if (!capable(CAP_SYS_ADMIN))
3201 return -EPERM;
3202
3203 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3204 btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet");
3205 return -EINVAL;
3206 }
3207
3208 sa = memdup_user(arg, sizeof(*sa));
3209 if (IS_ERR(sa))
3210 return PTR_ERR(sa);
3211
3212 if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) {
3213 ret = -EOPNOTSUPP;
3214 goto out;
3215 }
3216
3217 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3218 ret = mnt_want_write_file(file);
3219 if (ret)
3220 goto out;
3221 }
3222
3223 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3224 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3225 0);
3226
3227 /*
3228 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3229 * error. This is important as it allows user space to know how much
3230 * progress scrub has done. For example, if scrub is canceled we get
3231 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3232 * space. Later user space can inspect the progress from the structure
3233 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3234 * previously (btrfs-progs does this).
3235 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3236 * then return -EFAULT to signal the structure was not copied or it may
3237 * be corrupt and unreliable due to a partial copy.
3238 */
3239 if (copy_to_user(arg, sa, sizeof(*sa)))
3240 ret = -EFAULT;
3241
3242 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3243 mnt_drop_write_file(file);
3244 out:
3245 kfree(sa);
3246 return ret;
3247 }
3248
btrfs_ioctl_scrub_cancel(struct btrfs_fs_info * fs_info)3249 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3250 {
3251 if (!capable(CAP_SYS_ADMIN))
3252 return -EPERM;
3253
3254 return btrfs_scrub_cancel(fs_info);
3255 }
3256
btrfs_ioctl_scrub_progress(struct btrfs_fs_info * fs_info,void __user * arg)3257 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3258 void __user *arg)
3259 {
3260 struct btrfs_ioctl_scrub_args *sa;
3261 int ret;
3262
3263 if (!capable(CAP_SYS_ADMIN))
3264 return -EPERM;
3265
3266 sa = memdup_user(arg, sizeof(*sa));
3267 if (IS_ERR(sa))
3268 return PTR_ERR(sa);
3269
3270 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3271
3272 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3273 ret = -EFAULT;
3274
3275 kfree(sa);
3276 return ret;
3277 }
3278
btrfs_ioctl_get_dev_stats(struct btrfs_fs_info * fs_info,void __user * arg)3279 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3280 void __user *arg)
3281 {
3282 struct btrfs_ioctl_get_dev_stats *sa;
3283 int ret;
3284
3285 sa = memdup_user(arg, sizeof(*sa));
3286 if (IS_ERR(sa))
3287 return PTR_ERR(sa);
3288
3289 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3290 kfree(sa);
3291 return -EPERM;
3292 }
3293
3294 ret = btrfs_get_dev_stats(fs_info, sa);
3295
3296 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3297 ret = -EFAULT;
3298
3299 kfree(sa);
3300 return ret;
3301 }
3302
btrfs_ioctl_dev_replace(struct btrfs_fs_info * fs_info,void __user * arg)3303 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3304 void __user *arg)
3305 {
3306 struct btrfs_ioctl_dev_replace_args *p;
3307 int ret;
3308
3309 if (!capable(CAP_SYS_ADMIN))
3310 return -EPERM;
3311
3312 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3313 btrfs_err(fs_info, "device replace not supported on extent tree v2 yet");
3314 return -EINVAL;
3315 }
3316
3317 p = memdup_user(arg, sizeof(*p));
3318 if (IS_ERR(p))
3319 return PTR_ERR(p);
3320
3321 switch (p->cmd) {
3322 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3323 if (sb_rdonly(fs_info->sb)) {
3324 ret = -EROFS;
3325 goto out;
3326 }
3327 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3328 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3329 } else {
3330 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3331 btrfs_exclop_finish(fs_info);
3332 }
3333 break;
3334 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3335 btrfs_dev_replace_status(fs_info, p);
3336 ret = 0;
3337 break;
3338 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3339 p->result = btrfs_dev_replace_cancel(fs_info);
3340 ret = 0;
3341 break;
3342 default:
3343 ret = -EINVAL;
3344 break;
3345 }
3346
3347 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3348 ret = -EFAULT;
3349 out:
3350 kfree(p);
3351 return ret;
3352 }
3353
btrfs_ioctl_ino_to_path(struct btrfs_root * root,void __user * arg)3354 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3355 {
3356 int ret = 0;
3357 int i;
3358 u64 rel_ptr;
3359 int size;
3360 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3361 struct inode_fs_paths *ipath = NULL;
3362 struct btrfs_path *path;
3363
3364 if (!capable(CAP_DAC_READ_SEARCH))
3365 return -EPERM;
3366
3367 path = btrfs_alloc_path();
3368 if (!path) {
3369 ret = -ENOMEM;
3370 goto out;
3371 }
3372
3373 ipa = memdup_user(arg, sizeof(*ipa));
3374 if (IS_ERR(ipa)) {
3375 ret = PTR_ERR(ipa);
3376 ipa = NULL;
3377 goto out;
3378 }
3379
3380 size = min_t(u32, ipa->size, 4096);
3381 ipath = init_ipath(size, root, path);
3382 if (IS_ERR(ipath)) {
3383 ret = PTR_ERR(ipath);
3384 ipath = NULL;
3385 goto out;
3386 }
3387
3388 ret = paths_from_inode(ipa->inum, ipath);
3389 if (ret < 0)
3390 goto out;
3391
3392 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3393 rel_ptr = ipath->fspath->val[i] -
3394 (u64)(unsigned long)ipath->fspath->val;
3395 ipath->fspath->val[i] = rel_ptr;
3396 }
3397
3398 btrfs_free_path(path);
3399 path = NULL;
3400 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3401 ipath->fspath, size);
3402 if (ret) {
3403 ret = -EFAULT;
3404 goto out;
3405 }
3406
3407 out:
3408 btrfs_free_path(path);
3409 free_ipath(ipath);
3410 kfree(ipa);
3411
3412 return ret;
3413 }
3414
btrfs_ioctl_logical_to_ino(struct btrfs_fs_info * fs_info,void __user * arg,int version)3415 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3416 void __user *arg, int version)
3417 {
3418 int ret = 0;
3419 int size;
3420 struct btrfs_ioctl_logical_ino_args *loi;
3421 struct btrfs_data_container *inodes = NULL;
3422 struct btrfs_path *path = NULL;
3423 bool ignore_offset;
3424
3425 if (!capable(CAP_SYS_ADMIN))
3426 return -EPERM;
3427
3428 loi = memdup_user(arg, sizeof(*loi));
3429 if (IS_ERR(loi))
3430 return PTR_ERR(loi);
3431
3432 if (version == 1) {
3433 ignore_offset = false;
3434 size = min_t(u32, loi->size, SZ_64K);
3435 } else {
3436 /* All reserved bits must be 0 for now */
3437 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3438 ret = -EINVAL;
3439 goto out_loi;
3440 }
3441 /* Only accept flags we have defined so far */
3442 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3443 ret = -EINVAL;
3444 goto out_loi;
3445 }
3446 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3447 size = min_t(u32, loi->size, SZ_16M);
3448 }
3449
3450 inodes = init_data_container(size);
3451 if (IS_ERR(inodes)) {
3452 ret = PTR_ERR(inodes);
3453 goto out_loi;
3454 }
3455
3456 path = btrfs_alloc_path();
3457 if (!path) {
3458 ret = -ENOMEM;
3459 goto out;
3460 }
3461 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3462 inodes, ignore_offset);
3463 btrfs_free_path(path);
3464 if (ret == -EINVAL)
3465 ret = -ENOENT;
3466 if (ret < 0)
3467 goto out;
3468
3469 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3470 size);
3471 if (ret)
3472 ret = -EFAULT;
3473
3474 out:
3475 kvfree(inodes);
3476 out_loi:
3477 kfree(loi);
3478
3479 return ret;
3480 }
3481
btrfs_update_ioctl_balance_args(struct btrfs_fs_info * fs_info,struct btrfs_ioctl_balance_args * bargs)3482 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3483 struct btrfs_ioctl_balance_args *bargs)
3484 {
3485 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3486
3487 bargs->flags = bctl->flags;
3488
3489 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3490 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3491 if (atomic_read(&fs_info->balance_pause_req))
3492 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3493 if (atomic_read(&fs_info->balance_cancel_req))
3494 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3495
3496 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3497 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3498 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3499
3500 spin_lock(&fs_info->balance_lock);
3501 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3502 spin_unlock(&fs_info->balance_lock);
3503 }
3504
3505 /*
3506 * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as
3507 * required.
3508 *
3509 * @fs_info: the filesystem
3510 * @excl_acquired: ptr to boolean value which is set to false in case balance
3511 * is being resumed
3512 *
3513 * Return 0 on success in which case both fs_info::balance is acquired as well
3514 * as exclusive ops are blocked. In case of failure return an error code.
3515 */
btrfs_try_lock_balance(struct btrfs_fs_info * fs_info,bool * excl_acquired)3516 static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired)
3517 {
3518 int ret;
3519
3520 /*
3521 * Exclusive operation is locked. Three possibilities:
3522 * (1) some other op is running
3523 * (2) balance is running
3524 * (3) balance is paused -- special case (think resume)
3525 */
3526 while (1) {
3527 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
3528 *excl_acquired = true;
3529 mutex_lock(&fs_info->balance_mutex);
3530 return 0;
3531 }
3532
3533 mutex_lock(&fs_info->balance_mutex);
3534 if (fs_info->balance_ctl) {
3535 /* This is either (2) or (3) */
3536 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3537 /* This is (2) */
3538 ret = -EINPROGRESS;
3539 goto out_failure;
3540
3541 } else {
3542 mutex_unlock(&fs_info->balance_mutex);
3543 /*
3544 * Lock released to allow other waiters to
3545 * continue, we'll reexamine the status again.
3546 */
3547 mutex_lock(&fs_info->balance_mutex);
3548
3549 if (fs_info->balance_ctl &&
3550 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3551 /* This is (3) */
3552 *excl_acquired = false;
3553 return 0;
3554 }
3555 }
3556 } else {
3557 /* This is (1) */
3558 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3559 goto out_failure;
3560 }
3561
3562 mutex_unlock(&fs_info->balance_mutex);
3563 }
3564
3565 out_failure:
3566 mutex_unlock(&fs_info->balance_mutex);
3567 *excl_acquired = false;
3568 return ret;
3569 }
3570
btrfs_ioctl_balance(struct file * file,void __user * arg)3571 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3572 {
3573 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3574 struct btrfs_fs_info *fs_info = root->fs_info;
3575 struct btrfs_ioctl_balance_args *bargs;
3576 struct btrfs_balance_control *bctl;
3577 bool need_unlock = true;
3578 int ret;
3579
3580 if (!capable(CAP_SYS_ADMIN))
3581 return -EPERM;
3582
3583 ret = mnt_want_write_file(file);
3584 if (ret)
3585 return ret;
3586
3587 bargs = memdup_user(arg, sizeof(*bargs));
3588 if (IS_ERR(bargs)) {
3589 ret = PTR_ERR(bargs);
3590 bargs = NULL;
3591 goto out;
3592 }
3593
3594 ret = btrfs_try_lock_balance(fs_info, &need_unlock);
3595 if (ret)
3596 goto out;
3597
3598 lockdep_assert_held(&fs_info->balance_mutex);
3599
3600 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3601 if (!fs_info->balance_ctl) {
3602 ret = -ENOTCONN;
3603 goto out_unlock;
3604 }
3605
3606 bctl = fs_info->balance_ctl;
3607 spin_lock(&fs_info->balance_lock);
3608 bctl->flags |= BTRFS_BALANCE_RESUME;
3609 spin_unlock(&fs_info->balance_lock);
3610 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE);
3611
3612 goto do_balance;
3613 }
3614
3615 if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3616 ret = -EINVAL;
3617 goto out_unlock;
3618 }
3619
3620 if (fs_info->balance_ctl) {
3621 ret = -EINPROGRESS;
3622 goto out_unlock;
3623 }
3624
3625 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
3626 if (!bctl) {
3627 ret = -ENOMEM;
3628 goto out_unlock;
3629 }
3630
3631 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3632 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3633 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3634
3635 bctl->flags = bargs->flags;
3636 do_balance:
3637 /*
3638 * Ownership of bctl and exclusive operation goes to btrfs_balance.
3639 * bctl is freed in reset_balance_state, or, if restriper was paused
3640 * all the way until unmount, in free_fs_info. The flag should be
3641 * cleared after reset_balance_state.
3642 */
3643 need_unlock = false;
3644
3645 ret = btrfs_balance(fs_info, bctl, bargs);
3646 bctl = NULL;
3647
3648 if (ret == 0 || ret == -ECANCELED) {
3649 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3650 ret = -EFAULT;
3651 }
3652
3653 kfree(bctl);
3654 out_unlock:
3655 mutex_unlock(&fs_info->balance_mutex);
3656 if (need_unlock)
3657 btrfs_exclop_finish(fs_info);
3658 out:
3659 mnt_drop_write_file(file);
3660 kfree(bargs);
3661 return ret;
3662 }
3663
btrfs_ioctl_balance_ctl(struct btrfs_fs_info * fs_info,int cmd)3664 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
3665 {
3666 if (!capable(CAP_SYS_ADMIN))
3667 return -EPERM;
3668
3669 switch (cmd) {
3670 case BTRFS_BALANCE_CTL_PAUSE:
3671 return btrfs_pause_balance(fs_info);
3672 case BTRFS_BALANCE_CTL_CANCEL:
3673 return btrfs_cancel_balance(fs_info);
3674 }
3675
3676 return -EINVAL;
3677 }
3678
btrfs_ioctl_balance_progress(struct btrfs_fs_info * fs_info,void __user * arg)3679 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
3680 void __user *arg)
3681 {
3682 struct btrfs_ioctl_balance_args *bargs;
3683 int ret = 0;
3684
3685 if (!capable(CAP_SYS_ADMIN))
3686 return -EPERM;
3687
3688 mutex_lock(&fs_info->balance_mutex);
3689 if (!fs_info->balance_ctl) {
3690 ret = -ENOTCONN;
3691 goto out;
3692 }
3693
3694 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
3695 if (!bargs) {
3696 ret = -ENOMEM;
3697 goto out;
3698 }
3699
3700 btrfs_update_ioctl_balance_args(fs_info, bargs);
3701
3702 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3703 ret = -EFAULT;
3704
3705 kfree(bargs);
3706 out:
3707 mutex_unlock(&fs_info->balance_mutex);
3708 return ret;
3709 }
3710
btrfs_ioctl_quota_ctl(struct file * file,void __user * arg)3711 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3712 {
3713 struct inode *inode = file_inode(file);
3714 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3715 struct btrfs_ioctl_quota_ctl_args *sa;
3716 int ret;
3717
3718 if (!capable(CAP_SYS_ADMIN))
3719 return -EPERM;
3720
3721 ret = mnt_want_write_file(file);
3722 if (ret)
3723 return ret;
3724
3725 sa = memdup_user(arg, sizeof(*sa));
3726 if (IS_ERR(sa)) {
3727 ret = PTR_ERR(sa);
3728 goto drop_write;
3729 }
3730
3731 down_write(&fs_info->subvol_sem);
3732
3733 switch (sa->cmd) {
3734 case BTRFS_QUOTA_CTL_ENABLE:
3735 ret = btrfs_quota_enable(fs_info);
3736 break;
3737 case BTRFS_QUOTA_CTL_DISABLE:
3738 ret = btrfs_quota_disable(fs_info);
3739 break;
3740 default:
3741 ret = -EINVAL;
3742 break;
3743 }
3744
3745 kfree(sa);
3746 up_write(&fs_info->subvol_sem);
3747 drop_write:
3748 mnt_drop_write_file(file);
3749 return ret;
3750 }
3751
btrfs_ioctl_qgroup_assign(struct file * file,void __user * arg)3752 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3753 {
3754 struct inode *inode = file_inode(file);
3755 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3756 struct btrfs_root *root = BTRFS_I(inode)->root;
3757 struct btrfs_ioctl_qgroup_assign_args *sa;
3758 struct btrfs_trans_handle *trans;
3759 int ret;
3760 int err;
3761
3762 if (!capable(CAP_SYS_ADMIN))
3763 return -EPERM;
3764
3765 ret = mnt_want_write_file(file);
3766 if (ret)
3767 return ret;
3768
3769 sa = memdup_user(arg, sizeof(*sa));
3770 if (IS_ERR(sa)) {
3771 ret = PTR_ERR(sa);
3772 goto drop_write;
3773 }
3774
3775 trans = btrfs_join_transaction(root);
3776 if (IS_ERR(trans)) {
3777 ret = PTR_ERR(trans);
3778 goto out;
3779 }
3780
3781 if (sa->assign) {
3782 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
3783 } else {
3784 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
3785 }
3786
3787 /* update qgroup status and info */
3788 mutex_lock(&fs_info->qgroup_ioctl_lock);
3789 err = btrfs_run_qgroups(trans);
3790 mutex_unlock(&fs_info->qgroup_ioctl_lock);
3791 if (err < 0)
3792 btrfs_handle_fs_error(fs_info, err,
3793 "failed to update qgroup status and info");
3794 err = btrfs_end_transaction(trans);
3795 if (err && !ret)
3796 ret = err;
3797
3798 out:
3799 kfree(sa);
3800 drop_write:
3801 mnt_drop_write_file(file);
3802 return ret;
3803 }
3804
btrfs_ioctl_qgroup_create(struct file * file,void __user * arg)3805 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3806 {
3807 struct inode *inode = file_inode(file);
3808 struct btrfs_root *root = BTRFS_I(inode)->root;
3809 struct btrfs_ioctl_qgroup_create_args *sa;
3810 struct btrfs_trans_handle *trans;
3811 int ret;
3812 int err;
3813
3814 if (!capable(CAP_SYS_ADMIN))
3815 return -EPERM;
3816
3817 ret = mnt_want_write_file(file);
3818 if (ret)
3819 return ret;
3820
3821 sa = memdup_user(arg, sizeof(*sa));
3822 if (IS_ERR(sa)) {
3823 ret = PTR_ERR(sa);
3824 goto drop_write;
3825 }
3826
3827 if (!sa->qgroupid) {
3828 ret = -EINVAL;
3829 goto out;
3830 }
3831
3832 if (sa->create && is_fstree(sa->qgroupid)) {
3833 ret = -EINVAL;
3834 goto out;
3835 }
3836
3837 trans = btrfs_join_transaction(root);
3838 if (IS_ERR(trans)) {
3839 ret = PTR_ERR(trans);
3840 goto out;
3841 }
3842
3843 if (sa->create) {
3844 ret = btrfs_create_qgroup(trans, sa->qgroupid);
3845 } else {
3846 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
3847 }
3848
3849 err = btrfs_end_transaction(trans);
3850 if (err && !ret)
3851 ret = err;
3852
3853 out:
3854 kfree(sa);
3855 drop_write:
3856 mnt_drop_write_file(file);
3857 return ret;
3858 }
3859
btrfs_ioctl_qgroup_limit(struct file * file,void __user * arg)3860 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3861 {
3862 struct inode *inode = file_inode(file);
3863 struct btrfs_root *root = BTRFS_I(inode)->root;
3864 struct btrfs_ioctl_qgroup_limit_args *sa;
3865 struct btrfs_trans_handle *trans;
3866 int ret;
3867 int err;
3868 u64 qgroupid;
3869
3870 if (!capable(CAP_SYS_ADMIN))
3871 return -EPERM;
3872
3873 ret = mnt_want_write_file(file);
3874 if (ret)
3875 return ret;
3876
3877 sa = memdup_user(arg, sizeof(*sa));
3878 if (IS_ERR(sa)) {
3879 ret = PTR_ERR(sa);
3880 goto drop_write;
3881 }
3882
3883 trans = btrfs_join_transaction(root);
3884 if (IS_ERR(trans)) {
3885 ret = PTR_ERR(trans);
3886 goto out;
3887 }
3888
3889 qgroupid = sa->qgroupid;
3890 if (!qgroupid) {
3891 /* take the current subvol as qgroup */
3892 qgroupid = root->root_key.objectid;
3893 }
3894
3895 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
3896
3897 err = btrfs_end_transaction(trans);
3898 if (err && !ret)
3899 ret = err;
3900
3901 out:
3902 kfree(sa);
3903 drop_write:
3904 mnt_drop_write_file(file);
3905 return ret;
3906 }
3907
btrfs_ioctl_quota_rescan(struct file * file,void __user * arg)3908 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3909 {
3910 struct inode *inode = file_inode(file);
3911 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3912 struct btrfs_ioctl_quota_rescan_args *qsa;
3913 int ret;
3914
3915 if (!capable(CAP_SYS_ADMIN))
3916 return -EPERM;
3917
3918 ret = mnt_want_write_file(file);
3919 if (ret)
3920 return ret;
3921
3922 qsa = memdup_user(arg, sizeof(*qsa));
3923 if (IS_ERR(qsa)) {
3924 ret = PTR_ERR(qsa);
3925 goto drop_write;
3926 }
3927
3928 if (qsa->flags) {
3929 ret = -EINVAL;
3930 goto out;
3931 }
3932
3933 ret = btrfs_qgroup_rescan(fs_info);
3934
3935 out:
3936 kfree(qsa);
3937 drop_write:
3938 mnt_drop_write_file(file);
3939 return ret;
3940 }
3941
btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info * fs_info,void __user * arg)3942 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
3943 void __user *arg)
3944 {
3945 struct btrfs_ioctl_quota_rescan_args qsa = {0};
3946
3947 if (!capable(CAP_SYS_ADMIN))
3948 return -EPERM;
3949
3950 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3951 qsa.flags = 1;
3952 qsa.progress = fs_info->qgroup_rescan_progress.objectid;
3953 }
3954
3955 if (copy_to_user(arg, &qsa, sizeof(qsa)))
3956 return -EFAULT;
3957
3958 return 0;
3959 }
3960
btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info * fs_info,void __user * arg)3961 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
3962 void __user *arg)
3963 {
3964 if (!capable(CAP_SYS_ADMIN))
3965 return -EPERM;
3966
3967 return btrfs_qgroup_wait_for_completion(fs_info, true);
3968 }
3969
_btrfs_ioctl_set_received_subvol(struct file * file,struct mnt_idmap * idmap,struct btrfs_ioctl_received_subvol_args * sa)3970 static long _btrfs_ioctl_set_received_subvol(struct file *file,
3971 struct mnt_idmap *idmap,
3972 struct btrfs_ioctl_received_subvol_args *sa)
3973 {
3974 struct inode *inode = file_inode(file);
3975 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3976 struct btrfs_root *root = BTRFS_I(inode)->root;
3977 struct btrfs_root_item *root_item = &root->root_item;
3978 struct btrfs_trans_handle *trans;
3979 struct timespec64 ct = current_time(inode);
3980 int ret = 0;
3981 int received_uuid_changed;
3982
3983 if (!inode_owner_or_capable(idmap, inode))
3984 return -EPERM;
3985
3986 ret = mnt_want_write_file(file);
3987 if (ret < 0)
3988 return ret;
3989
3990 down_write(&fs_info->subvol_sem);
3991
3992 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3993 ret = -EINVAL;
3994 goto out;
3995 }
3996
3997 if (btrfs_root_readonly(root)) {
3998 ret = -EROFS;
3999 goto out;
4000 }
4001
4002 /*
4003 * 1 - root item
4004 * 2 - uuid items (received uuid + subvol uuid)
4005 */
4006 trans = btrfs_start_transaction(root, 3);
4007 if (IS_ERR(trans)) {
4008 ret = PTR_ERR(trans);
4009 trans = NULL;
4010 goto out;
4011 }
4012
4013 sa->rtransid = trans->transid;
4014 sa->rtime.sec = ct.tv_sec;
4015 sa->rtime.nsec = ct.tv_nsec;
4016
4017 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4018 BTRFS_UUID_SIZE);
4019 if (received_uuid_changed &&
4020 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4021 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4022 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4023 root->root_key.objectid);
4024 if (ret && ret != -ENOENT) {
4025 btrfs_abort_transaction(trans, ret);
4026 btrfs_end_transaction(trans);
4027 goto out;
4028 }
4029 }
4030 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4031 btrfs_set_root_stransid(root_item, sa->stransid);
4032 btrfs_set_root_rtransid(root_item, sa->rtransid);
4033 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4034 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4035 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4036 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4037
4038 ret = btrfs_update_root(trans, fs_info->tree_root,
4039 &root->root_key, &root->root_item);
4040 if (ret < 0) {
4041 btrfs_end_transaction(trans);
4042 goto out;
4043 }
4044 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4045 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4046 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4047 root->root_key.objectid);
4048 if (ret < 0 && ret != -EEXIST) {
4049 btrfs_abort_transaction(trans, ret);
4050 btrfs_end_transaction(trans);
4051 goto out;
4052 }
4053 }
4054 ret = btrfs_commit_transaction(trans);
4055 out:
4056 up_write(&fs_info->subvol_sem);
4057 mnt_drop_write_file(file);
4058 return ret;
4059 }
4060
4061 #ifdef CONFIG_64BIT
btrfs_ioctl_set_received_subvol_32(struct file * file,void __user * arg)4062 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4063 void __user *arg)
4064 {
4065 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4066 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4067 int ret = 0;
4068
4069 args32 = memdup_user(arg, sizeof(*args32));
4070 if (IS_ERR(args32))
4071 return PTR_ERR(args32);
4072
4073 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4074 if (!args64) {
4075 ret = -ENOMEM;
4076 goto out;
4077 }
4078
4079 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4080 args64->stransid = args32->stransid;
4081 args64->rtransid = args32->rtransid;
4082 args64->stime.sec = args32->stime.sec;
4083 args64->stime.nsec = args32->stime.nsec;
4084 args64->rtime.sec = args32->rtime.sec;
4085 args64->rtime.nsec = args32->rtime.nsec;
4086 args64->flags = args32->flags;
4087
4088 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), args64);
4089 if (ret)
4090 goto out;
4091
4092 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4093 args32->stransid = args64->stransid;
4094 args32->rtransid = args64->rtransid;
4095 args32->stime.sec = args64->stime.sec;
4096 args32->stime.nsec = args64->stime.nsec;
4097 args32->rtime.sec = args64->rtime.sec;
4098 args32->rtime.nsec = args64->rtime.nsec;
4099 args32->flags = args64->flags;
4100
4101 ret = copy_to_user(arg, args32, sizeof(*args32));
4102 if (ret)
4103 ret = -EFAULT;
4104
4105 out:
4106 kfree(args32);
4107 kfree(args64);
4108 return ret;
4109 }
4110 #endif
4111
btrfs_ioctl_set_received_subvol(struct file * file,void __user * arg)4112 static long btrfs_ioctl_set_received_subvol(struct file *file,
4113 void __user *arg)
4114 {
4115 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4116 int ret = 0;
4117
4118 sa = memdup_user(arg, sizeof(*sa));
4119 if (IS_ERR(sa))
4120 return PTR_ERR(sa);
4121
4122 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), sa);
4123
4124 if (ret)
4125 goto out;
4126
4127 ret = copy_to_user(arg, sa, sizeof(*sa));
4128 if (ret)
4129 ret = -EFAULT;
4130
4131 out:
4132 kfree(sa);
4133 return ret;
4134 }
4135
btrfs_ioctl_get_fslabel(struct btrfs_fs_info * fs_info,void __user * arg)4136 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4137 void __user *arg)
4138 {
4139 size_t len;
4140 int ret;
4141 char label[BTRFS_LABEL_SIZE];
4142
4143 spin_lock(&fs_info->super_lock);
4144 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4145 spin_unlock(&fs_info->super_lock);
4146
4147 len = strnlen(label, BTRFS_LABEL_SIZE);
4148
4149 if (len == BTRFS_LABEL_SIZE) {
4150 btrfs_warn(fs_info,
4151 "label is too long, return the first %zu bytes",
4152 --len);
4153 }
4154
4155 ret = copy_to_user(arg, label, len);
4156
4157 return ret ? -EFAULT : 0;
4158 }
4159
btrfs_ioctl_set_fslabel(struct file * file,void __user * arg)4160 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4161 {
4162 struct inode *inode = file_inode(file);
4163 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4164 struct btrfs_root *root = BTRFS_I(inode)->root;
4165 struct btrfs_super_block *super_block = fs_info->super_copy;
4166 struct btrfs_trans_handle *trans;
4167 char label[BTRFS_LABEL_SIZE];
4168 int ret;
4169
4170 if (!capable(CAP_SYS_ADMIN))
4171 return -EPERM;
4172
4173 if (copy_from_user(label, arg, sizeof(label)))
4174 return -EFAULT;
4175
4176 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4177 btrfs_err(fs_info,
4178 "unable to set label with more than %d bytes",
4179 BTRFS_LABEL_SIZE - 1);
4180 return -EINVAL;
4181 }
4182
4183 ret = mnt_want_write_file(file);
4184 if (ret)
4185 return ret;
4186
4187 trans = btrfs_start_transaction(root, 0);
4188 if (IS_ERR(trans)) {
4189 ret = PTR_ERR(trans);
4190 goto out_unlock;
4191 }
4192
4193 spin_lock(&fs_info->super_lock);
4194 strcpy(super_block->label, label);
4195 spin_unlock(&fs_info->super_lock);
4196 ret = btrfs_commit_transaction(trans);
4197
4198 out_unlock:
4199 mnt_drop_write_file(file);
4200 return ret;
4201 }
4202
4203 #define INIT_FEATURE_FLAGS(suffix) \
4204 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4205 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4206 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4207
btrfs_ioctl_get_supported_features(void __user * arg)4208 int btrfs_ioctl_get_supported_features(void __user *arg)
4209 {
4210 static const struct btrfs_ioctl_feature_flags features[3] = {
4211 INIT_FEATURE_FLAGS(SUPP),
4212 INIT_FEATURE_FLAGS(SAFE_SET),
4213 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4214 };
4215
4216 if (copy_to_user(arg, &features, sizeof(features)))
4217 return -EFAULT;
4218
4219 return 0;
4220 }
4221
btrfs_ioctl_get_features(struct btrfs_fs_info * fs_info,void __user * arg)4222 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4223 void __user *arg)
4224 {
4225 struct btrfs_super_block *super_block = fs_info->super_copy;
4226 struct btrfs_ioctl_feature_flags features;
4227
4228 features.compat_flags = btrfs_super_compat_flags(super_block);
4229 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4230 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4231
4232 if (copy_to_user(arg, &features, sizeof(features)))
4233 return -EFAULT;
4234
4235 return 0;
4236 }
4237
check_feature_bits(struct btrfs_fs_info * fs_info,enum btrfs_feature_set set,u64 change_mask,u64 flags,u64 supported_flags,u64 safe_set,u64 safe_clear)4238 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4239 enum btrfs_feature_set set,
4240 u64 change_mask, u64 flags, u64 supported_flags,
4241 u64 safe_set, u64 safe_clear)
4242 {
4243 const char *type = btrfs_feature_set_name(set);
4244 char *names;
4245 u64 disallowed, unsupported;
4246 u64 set_mask = flags & change_mask;
4247 u64 clear_mask = ~flags & change_mask;
4248
4249 unsupported = set_mask & ~supported_flags;
4250 if (unsupported) {
4251 names = btrfs_printable_features(set, unsupported);
4252 if (names) {
4253 btrfs_warn(fs_info,
4254 "this kernel does not support the %s feature bit%s",
4255 names, strchr(names, ',') ? "s" : "");
4256 kfree(names);
4257 } else
4258 btrfs_warn(fs_info,
4259 "this kernel does not support %s bits 0x%llx",
4260 type, unsupported);
4261 return -EOPNOTSUPP;
4262 }
4263
4264 disallowed = set_mask & ~safe_set;
4265 if (disallowed) {
4266 names = btrfs_printable_features(set, disallowed);
4267 if (names) {
4268 btrfs_warn(fs_info,
4269 "can't set the %s feature bit%s while mounted",
4270 names, strchr(names, ',') ? "s" : "");
4271 kfree(names);
4272 } else
4273 btrfs_warn(fs_info,
4274 "can't set %s bits 0x%llx while mounted",
4275 type, disallowed);
4276 return -EPERM;
4277 }
4278
4279 disallowed = clear_mask & ~safe_clear;
4280 if (disallowed) {
4281 names = btrfs_printable_features(set, disallowed);
4282 if (names) {
4283 btrfs_warn(fs_info,
4284 "can't clear the %s feature bit%s while mounted",
4285 names, strchr(names, ',') ? "s" : "");
4286 kfree(names);
4287 } else
4288 btrfs_warn(fs_info,
4289 "can't clear %s bits 0x%llx while mounted",
4290 type, disallowed);
4291 return -EPERM;
4292 }
4293
4294 return 0;
4295 }
4296
4297 #define check_feature(fs_info, change_mask, flags, mask_base) \
4298 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
4299 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4300 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4301 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4302
btrfs_ioctl_set_features(struct file * file,void __user * arg)4303 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4304 {
4305 struct inode *inode = file_inode(file);
4306 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4307 struct btrfs_root *root = BTRFS_I(inode)->root;
4308 struct btrfs_super_block *super_block = fs_info->super_copy;
4309 struct btrfs_ioctl_feature_flags flags[2];
4310 struct btrfs_trans_handle *trans;
4311 u64 newflags;
4312 int ret;
4313
4314 if (!capable(CAP_SYS_ADMIN))
4315 return -EPERM;
4316
4317 if (copy_from_user(flags, arg, sizeof(flags)))
4318 return -EFAULT;
4319
4320 /* Nothing to do */
4321 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4322 !flags[0].incompat_flags)
4323 return 0;
4324
4325 ret = check_feature(fs_info, flags[0].compat_flags,
4326 flags[1].compat_flags, COMPAT);
4327 if (ret)
4328 return ret;
4329
4330 ret = check_feature(fs_info, flags[0].compat_ro_flags,
4331 flags[1].compat_ro_flags, COMPAT_RO);
4332 if (ret)
4333 return ret;
4334
4335 ret = check_feature(fs_info, flags[0].incompat_flags,
4336 flags[1].incompat_flags, INCOMPAT);
4337 if (ret)
4338 return ret;
4339
4340 ret = mnt_want_write_file(file);
4341 if (ret)
4342 return ret;
4343
4344 trans = btrfs_start_transaction(root, 0);
4345 if (IS_ERR(trans)) {
4346 ret = PTR_ERR(trans);
4347 goto out_drop_write;
4348 }
4349
4350 spin_lock(&fs_info->super_lock);
4351 newflags = btrfs_super_compat_flags(super_block);
4352 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4353 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4354 btrfs_set_super_compat_flags(super_block, newflags);
4355
4356 newflags = btrfs_super_compat_ro_flags(super_block);
4357 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4358 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4359 btrfs_set_super_compat_ro_flags(super_block, newflags);
4360
4361 newflags = btrfs_super_incompat_flags(super_block);
4362 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4363 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4364 btrfs_set_super_incompat_flags(super_block, newflags);
4365 spin_unlock(&fs_info->super_lock);
4366
4367 ret = btrfs_commit_transaction(trans);
4368 out_drop_write:
4369 mnt_drop_write_file(file);
4370
4371 return ret;
4372 }
4373
_btrfs_ioctl_send(struct inode * inode,void __user * argp,bool compat)4374 static int _btrfs_ioctl_send(struct inode *inode, void __user *argp, bool compat)
4375 {
4376 struct btrfs_ioctl_send_args *arg;
4377 int ret;
4378
4379 if (compat) {
4380 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4381 struct btrfs_ioctl_send_args_32 args32 = { 0 };
4382
4383 ret = copy_from_user(&args32, argp, sizeof(args32));
4384 if (ret)
4385 return -EFAULT;
4386 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4387 if (!arg)
4388 return -ENOMEM;
4389 arg->send_fd = args32.send_fd;
4390 arg->clone_sources_count = args32.clone_sources_count;
4391 arg->clone_sources = compat_ptr(args32.clone_sources);
4392 arg->parent_root = args32.parent_root;
4393 arg->flags = args32.flags;
4394 arg->version = args32.version;
4395 memcpy(arg->reserved, args32.reserved,
4396 sizeof(args32.reserved));
4397 #else
4398 return -ENOTTY;
4399 #endif
4400 } else {
4401 arg = memdup_user(argp, sizeof(*arg));
4402 if (IS_ERR(arg))
4403 return PTR_ERR(arg);
4404 }
4405 ret = btrfs_ioctl_send(inode, arg);
4406 kfree(arg);
4407 return ret;
4408 }
4409
btrfs_ioctl_encoded_read(struct file * file,void __user * argp,bool compat)4410 static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,
4411 bool compat)
4412 {
4413 struct btrfs_ioctl_encoded_io_args args = { 0 };
4414 size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args,
4415 flags);
4416 size_t copy_end;
4417 struct iovec iovstack[UIO_FASTIOV];
4418 struct iovec *iov = iovstack;
4419 struct iov_iter iter;
4420 loff_t pos;
4421 struct kiocb kiocb;
4422 ssize_t ret;
4423
4424 if (!capable(CAP_SYS_ADMIN)) {
4425 ret = -EPERM;
4426 goto out_acct;
4427 }
4428
4429 if (compat) {
4430 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4431 struct btrfs_ioctl_encoded_io_args_32 args32;
4432
4433 copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32,
4434 flags);
4435 if (copy_from_user(&args32, argp, copy_end)) {
4436 ret = -EFAULT;
4437 goto out_acct;
4438 }
4439 args.iov = compat_ptr(args32.iov);
4440 args.iovcnt = args32.iovcnt;
4441 args.offset = args32.offset;
4442 args.flags = args32.flags;
4443 #else
4444 return -ENOTTY;
4445 #endif
4446 } else {
4447 copy_end = copy_end_kernel;
4448 if (copy_from_user(&args, argp, copy_end)) {
4449 ret = -EFAULT;
4450 goto out_acct;
4451 }
4452 }
4453 if (args.flags != 0) {
4454 ret = -EINVAL;
4455 goto out_acct;
4456 }
4457
4458 ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4459 &iov, &iter);
4460 if (ret < 0)
4461 goto out_acct;
4462
4463 if (iov_iter_count(&iter) == 0) {
4464 ret = 0;
4465 goto out_iov;
4466 }
4467 pos = args.offset;
4468 ret = rw_verify_area(READ, file, &pos, args.len);
4469 if (ret < 0)
4470 goto out_iov;
4471
4472 init_sync_kiocb(&kiocb, file);
4473 kiocb.ki_pos = pos;
4474
4475 ret = btrfs_encoded_read(&kiocb, &iter, &args);
4476 if (ret >= 0) {
4477 fsnotify_access(file);
4478 if (copy_to_user(argp + copy_end,
4479 (char *)&args + copy_end_kernel,
4480 sizeof(args) - copy_end_kernel))
4481 ret = -EFAULT;
4482 }
4483
4484 out_iov:
4485 kfree(iov);
4486 out_acct:
4487 if (ret > 0)
4488 add_rchar(current, ret);
4489 inc_syscr(current);
4490 return ret;
4491 }
4492
btrfs_ioctl_encoded_write(struct file * file,void __user * argp,bool compat)4493 static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat)
4494 {
4495 struct btrfs_ioctl_encoded_io_args args;
4496 struct iovec iovstack[UIO_FASTIOV];
4497 struct iovec *iov = iovstack;
4498 struct iov_iter iter;
4499 loff_t pos;
4500 struct kiocb kiocb;
4501 ssize_t ret;
4502
4503 if (!capable(CAP_SYS_ADMIN)) {
4504 ret = -EPERM;
4505 goto out_acct;
4506 }
4507
4508 if (!(file->f_mode & FMODE_WRITE)) {
4509 ret = -EBADF;
4510 goto out_acct;
4511 }
4512
4513 if (compat) {
4514 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4515 struct btrfs_ioctl_encoded_io_args_32 args32;
4516
4517 if (copy_from_user(&args32, argp, sizeof(args32))) {
4518 ret = -EFAULT;
4519 goto out_acct;
4520 }
4521 args.iov = compat_ptr(args32.iov);
4522 args.iovcnt = args32.iovcnt;
4523 args.offset = args32.offset;
4524 args.flags = args32.flags;
4525 args.len = args32.len;
4526 args.unencoded_len = args32.unencoded_len;
4527 args.unencoded_offset = args32.unencoded_offset;
4528 args.compression = args32.compression;
4529 args.encryption = args32.encryption;
4530 memcpy(args.reserved, args32.reserved, sizeof(args.reserved));
4531 #else
4532 return -ENOTTY;
4533 #endif
4534 } else {
4535 if (copy_from_user(&args, argp, sizeof(args))) {
4536 ret = -EFAULT;
4537 goto out_acct;
4538 }
4539 }
4540
4541 ret = -EINVAL;
4542 if (args.flags != 0)
4543 goto out_acct;
4544 if (memchr_inv(args.reserved, 0, sizeof(args.reserved)))
4545 goto out_acct;
4546 if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
4547 args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
4548 goto out_acct;
4549 if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
4550 args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
4551 goto out_acct;
4552 if (args.unencoded_offset > args.unencoded_len)
4553 goto out_acct;
4554 if (args.len > args.unencoded_len - args.unencoded_offset)
4555 goto out_acct;
4556
4557 ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4558 &iov, &iter);
4559 if (ret < 0)
4560 goto out_acct;
4561
4562 file_start_write(file);
4563
4564 if (iov_iter_count(&iter) == 0) {
4565 ret = 0;
4566 goto out_end_write;
4567 }
4568 pos = args.offset;
4569 ret = rw_verify_area(WRITE, file, &pos, args.len);
4570 if (ret < 0)
4571 goto out_end_write;
4572
4573 init_sync_kiocb(&kiocb, file);
4574 ret = kiocb_set_rw_flags(&kiocb, 0);
4575 if (ret)
4576 goto out_end_write;
4577 kiocb.ki_pos = pos;
4578
4579 ret = btrfs_do_write_iter(&kiocb, &iter, &args);
4580 if (ret > 0)
4581 fsnotify_modify(file);
4582
4583 out_end_write:
4584 file_end_write(file);
4585 kfree(iov);
4586 out_acct:
4587 if (ret > 0)
4588 add_wchar(current, ret);
4589 inc_syscw(current);
4590 return ret;
4591 }
4592
btrfs_ioctl(struct file * file,unsigned int cmd,unsigned long arg)4593 long btrfs_ioctl(struct file *file, unsigned int
4594 cmd, unsigned long arg)
4595 {
4596 struct inode *inode = file_inode(file);
4597 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4598 struct btrfs_root *root = BTRFS_I(inode)->root;
4599 void __user *argp = (void __user *)arg;
4600
4601 switch (cmd) {
4602 case FS_IOC_GETVERSION:
4603 return btrfs_ioctl_getversion(inode, argp);
4604 case FS_IOC_GETFSLABEL:
4605 return btrfs_ioctl_get_fslabel(fs_info, argp);
4606 case FS_IOC_SETFSLABEL:
4607 return btrfs_ioctl_set_fslabel(file, argp);
4608 case FITRIM:
4609 return btrfs_ioctl_fitrim(fs_info, argp);
4610 case BTRFS_IOC_SNAP_CREATE:
4611 return btrfs_ioctl_snap_create(file, argp, 0);
4612 case BTRFS_IOC_SNAP_CREATE_V2:
4613 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4614 case BTRFS_IOC_SUBVOL_CREATE:
4615 return btrfs_ioctl_snap_create(file, argp, 1);
4616 case BTRFS_IOC_SUBVOL_CREATE_V2:
4617 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4618 case BTRFS_IOC_SNAP_DESTROY:
4619 return btrfs_ioctl_snap_destroy(file, argp, false);
4620 case BTRFS_IOC_SNAP_DESTROY_V2:
4621 return btrfs_ioctl_snap_destroy(file, argp, true);
4622 case BTRFS_IOC_SUBVOL_GETFLAGS:
4623 return btrfs_ioctl_subvol_getflags(inode, argp);
4624 case BTRFS_IOC_SUBVOL_SETFLAGS:
4625 return btrfs_ioctl_subvol_setflags(file, argp);
4626 case BTRFS_IOC_DEFAULT_SUBVOL:
4627 return btrfs_ioctl_default_subvol(file, argp);
4628 case BTRFS_IOC_DEFRAG:
4629 return btrfs_ioctl_defrag(file, NULL);
4630 case BTRFS_IOC_DEFRAG_RANGE:
4631 return btrfs_ioctl_defrag(file, argp);
4632 case BTRFS_IOC_RESIZE:
4633 return btrfs_ioctl_resize(file, argp);
4634 case BTRFS_IOC_ADD_DEV:
4635 return btrfs_ioctl_add_dev(fs_info, argp);
4636 case BTRFS_IOC_RM_DEV:
4637 return btrfs_ioctl_rm_dev(file, argp);
4638 case BTRFS_IOC_RM_DEV_V2:
4639 return btrfs_ioctl_rm_dev_v2(file, argp);
4640 case BTRFS_IOC_FS_INFO:
4641 return btrfs_ioctl_fs_info(fs_info, argp);
4642 case BTRFS_IOC_DEV_INFO:
4643 return btrfs_ioctl_dev_info(fs_info, argp);
4644 case BTRFS_IOC_TREE_SEARCH:
4645 return btrfs_ioctl_tree_search(inode, argp);
4646 case BTRFS_IOC_TREE_SEARCH_V2:
4647 return btrfs_ioctl_tree_search_v2(inode, argp);
4648 case BTRFS_IOC_INO_LOOKUP:
4649 return btrfs_ioctl_ino_lookup(root, argp);
4650 case BTRFS_IOC_INO_PATHS:
4651 return btrfs_ioctl_ino_to_path(root, argp);
4652 case BTRFS_IOC_LOGICAL_INO:
4653 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4654 case BTRFS_IOC_LOGICAL_INO_V2:
4655 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4656 case BTRFS_IOC_SPACE_INFO:
4657 return btrfs_ioctl_space_info(fs_info, argp);
4658 case BTRFS_IOC_SYNC: {
4659 int ret;
4660
4661 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4662 if (ret)
4663 return ret;
4664 ret = btrfs_sync_fs(inode->i_sb, 1);
4665 /*
4666 * The transaction thread may want to do more work,
4667 * namely it pokes the cleaner kthread that will start
4668 * processing uncleaned subvols.
4669 */
4670 wake_up_process(fs_info->transaction_kthread);
4671 return ret;
4672 }
4673 case BTRFS_IOC_START_SYNC:
4674 return btrfs_ioctl_start_sync(root, argp);
4675 case BTRFS_IOC_WAIT_SYNC:
4676 return btrfs_ioctl_wait_sync(fs_info, argp);
4677 case BTRFS_IOC_SCRUB:
4678 return btrfs_ioctl_scrub(file, argp);
4679 case BTRFS_IOC_SCRUB_CANCEL:
4680 return btrfs_ioctl_scrub_cancel(fs_info);
4681 case BTRFS_IOC_SCRUB_PROGRESS:
4682 return btrfs_ioctl_scrub_progress(fs_info, argp);
4683 case BTRFS_IOC_BALANCE_V2:
4684 return btrfs_ioctl_balance(file, argp);
4685 case BTRFS_IOC_BALANCE_CTL:
4686 return btrfs_ioctl_balance_ctl(fs_info, arg);
4687 case BTRFS_IOC_BALANCE_PROGRESS:
4688 return btrfs_ioctl_balance_progress(fs_info, argp);
4689 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4690 return btrfs_ioctl_set_received_subvol(file, argp);
4691 #ifdef CONFIG_64BIT
4692 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4693 return btrfs_ioctl_set_received_subvol_32(file, argp);
4694 #endif
4695 case BTRFS_IOC_SEND:
4696 return _btrfs_ioctl_send(inode, argp, false);
4697 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4698 case BTRFS_IOC_SEND_32:
4699 return _btrfs_ioctl_send(inode, argp, true);
4700 #endif
4701 case BTRFS_IOC_GET_DEV_STATS:
4702 return btrfs_ioctl_get_dev_stats(fs_info, argp);
4703 case BTRFS_IOC_QUOTA_CTL:
4704 return btrfs_ioctl_quota_ctl(file, argp);
4705 case BTRFS_IOC_QGROUP_ASSIGN:
4706 return btrfs_ioctl_qgroup_assign(file, argp);
4707 case BTRFS_IOC_QGROUP_CREATE:
4708 return btrfs_ioctl_qgroup_create(file, argp);
4709 case BTRFS_IOC_QGROUP_LIMIT:
4710 return btrfs_ioctl_qgroup_limit(file, argp);
4711 case BTRFS_IOC_QUOTA_RESCAN:
4712 return btrfs_ioctl_quota_rescan(file, argp);
4713 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4714 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
4715 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4716 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
4717 case BTRFS_IOC_DEV_REPLACE:
4718 return btrfs_ioctl_dev_replace(fs_info, argp);
4719 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4720 return btrfs_ioctl_get_supported_features(argp);
4721 case BTRFS_IOC_GET_FEATURES:
4722 return btrfs_ioctl_get_features(fs_info, argp);
4723 case BTRFS_IOC_SET_FEATURES:
4724 return btrfs_ioctl_set_features(file, argp);
4725 case BTRFS_IOC_GET_SUBVOL_INFO:
4726 return btrfs_ioctl_get_subvol_info(inode, argp);
4727 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
4728 return btrfs_ioctl_get_subvol_rootref(root, argp);
4729 case BTRFS_IOC_INO_LOOKUP_USER:
4730 return btrfs_ioctl_ino_lookup_user(file, argp);
4731 case FS_IOC_ENABLE_VERITY:
4732 return fsverity_ioctl_enable(file, (const void __user *)argp);
4733 case FS_IOC_MEASURE_VERITY:
4734 return fsverity_ioctl_measure(file, argp);
4735 case BTRFS_IOC_ENCODED_READ:
4736 return btrfs_ioctl_encoded_read(file, argp, false);
4737 case BTRFS_IOC_ENCODED_WRITE:
4738 return btrfs_ioctl_encoded_write(file, argp, false);
4739 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4740 case BTRFS_IOC_ENCODED_READ_32:
4741 return btrfs_ioctl_encoded_read(file, argp, true);
4742 case BTRFS_IOC_ENCODED_WRITE_32:
4743 return btrfs_ioctl_encoded_write(file, argp, true);
4744 #endif
4745 }
4746
4747 return -ENOTTY;
4748 }
4749
4750 #ifdef CONFIG_COMPAT
btrfs_compat_ioctl(struct file * file,unsigned int cmd,unsigned long arg)4751 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4752 {
4753 /*
4754 * These all access 32-bit values anyway so no further
4755 * handling is necessary.
4756 */
4757 switch (cmd) {
4758 case FS_IOC32_GETVERSION:
4759 cmd = FS_IOC_GETVERSION;
4760 break;
4761 }
4762
4763 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4764 }
4765 #endif
4766