?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 4.18.10
[linux/fpc-iii.git] / fs / btrfs / ioctl.c
blobf3d6be0c657b1a31335fb589e3a9188825bc01f1
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/buffer_head.h>
9 #include <linux/file.h>
10 #include <linux/fs.h>
11 #include <linux/fsnotify.h>
12 #include <linux/pagemap.h>
13 #include <linux/highmem.h>
14 #include <linux/time.h>
15 #include <linux/init.h>
16 #include <linux/string.h>
17 #include <linux/backing-dev.h>
18 #include <linux/mount.h>
19 #include <linux/mpage.h>
20 #include <linux/namei.h>
21 #include <linux/swap.h>
22 #include <linux/writeback.h>
23 #include <linux/compat.h>
24 #include <linux/bit_spinlock.h>
25 #include <linux/security.h>
26 #include <linux/xattr.h>
27 #include <linux/mm.h>
28 #include <linux/slab.h>
29 #include <linux/blkdev.h>
30 #include <linux/uuid.h>
31 #include <linux/btrfs.h>
32 #include <linux/uaccess.h>
33 #include <linux/iversion.h>
34 #include "ctree.h"
35 #include "disk-io.h"
36 #include "transaction.h"
37 #include "btrfs_inode.h"
38 #include "print-tree.h"
39 #include "volumes.h"
40 #include "locking.h"
41 #include "inode-map.h"
42 #include "backref.h"
43 #include "rcu-string.h"
44 #include "send.h"
45 #include "dev-replace.h"
46 #include "props.h"
47 #include "sysfs.h"
48 #include "qgroup.h"
49 #include "tree-log.h"
50 #include "compression.h"
51
52 #ifdef CONFIG_64BIT
53 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
54 * structures are incorrect, as the timespec structure from userspace
55 * is 4 bytes too small. We define these alternatives here to teach
56 * the kernel about the 32-bit struct packing.
57 */
58 struct btrfs_ioctl_timespec_32 {
59 __u64 sec;
60 __u32 nsec;
61 } __attribute__ ((__packed__));
62
63 struct btrfs_ioctl_received_subvol_args_32 {
64 char uuid[BTRFS_UUID_SIZE]; /* in */
65 __u64 stransid; /* in */
66 __u64 rtransid; /* out */
67 struct btrfs_ioctl_timespec_32 stime; /* in */
68 struct btrfs_ioctl_timespec_32 rtime; /* out */
69 __u64 flags; /* in */
70 __u64 reserved[16]; /* in */
71 } __attribute__ ((__packed__));
72
73 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
74 struct btrfs_ioctl_received_subvol_args_32)
75 #endif
76
77 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
78 struct btrfs_ioctl_send_args_32 {
79 __s64 send_fd; /* in */
80 __u64 clone_sources_count; /* in */
81 compat_uptr_t clone_sources; /* in */
82 __u64 parent_root; /* in */
83 __u64 flags; /* in */
84 __u64 reserved[4]; /* in */
85 } __attribute__ ((__packed__));
86
87 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
88 struct btrfs_ioctl_send_args_32)
89 #endif
90
91 static int btrfs_clone(struct inode *src, struct inode *inode,
92 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
93 int no_time_update);
94
95 /* Mask out flags that are inappropriate for the given type of inode. */
96 static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
97 unsigned int flags)
98 {
99 if (S_ISDIR(inode->i_mode))
100 return flags;
101 else if (S_ISREG(inode->i_mode))
102 return flags & ~FS_DIRSYNC_FL;
103 else
104 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
105 }
106
107 /*
108 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
109 * ioctl.
110 */
111 static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags)
112 {
113 unsigned int iflags = 0;
114
115 if (flags & BTRFS_INODE_SYNC)
116 iflags |= FS_SYNC_FL;
117 if (flags & BTRFS_INODE_IMMUTABLE)
118 iflags |= FS_IMMUTABLE_FL;
119 if (flags & BTRFS_INODE_APPEND)
120 iflags |= FS_APPEND_FL;
121 if (flags & BTRFS_INODE_NODUMP)
122 iflags |= FS_NODUMP_FL;
123 if (flags & BTRFS_INODE_NOATIME)
124 iflags |= FS_NOATIME_FL;
125 if (flags & BTRFS_INODE_DIRSYNC)
126 iflags |= FS_DIRSYNC_FL;
127 if (flags & BTRFS_INODE_NODATACOW)
128 iflags |= FS_NOCOW_FL;
129
130 if (flags & BTRFS_INODE_NOCOMPRESS)
131 iflags |= FS_NOCOMP_FL;
132 else if (flags & BTRFS_INODE_COMPRESS)
133 iflags |= FS_COMPR_FL;
134
135 return iflags;
136 }
137
138 /*
139 * Update inode->i_flags based on the btrfs internal flags.
140 */
141 void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
142 {
143 struct btrfs_inode *binode = BTRFS_I(inode);
144 unsigned int new_fl = 0;
145
146 if (binode->flags & BTRFS_INODE_SYNC)
147 new_fl |= S_SYNC;
148 if (binode->flags & BTRFS_INODE_IMMUTABLE)
149 new_fl |= S_IMMUTABLE;
150 if (binode->flags & BTRFS_INODE_APPEND)
151 new_fl |= S_APPEND;
152 if (binode->flags & BTRFS_INODE_NOATIME)
153 new_fl |= S_NOATIME;
154 if (binode->flags & BTRFS_INODE_DIRSYNC)
155 new_fl |= S_DIRSYNC;
156
157 set_mask_bits(&inode->i_flags,
158 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
159 new_fl);
160 }
161
162 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
163 {
164 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
165 unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags);
166
167 if (copy_to_user(arg, &flags, sizeof(flags)))
168 return -EFAULT;
169 return 0;
170 }
171
172 /* Check if @flags are a supported and valid set of FS_*_FL flags */
173 static int check_fsflags(unsigned int flags)
174 {
175 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
176 FS_NOATIME_FL | FS_NODUMP_FL | \
177 FS_SYNC_FL | FS_DIRSYNC_FL | \
178 FS_NOCOMP_FL | FS_COMPR_FL |
179 FS_NOCOW_FL))
180 return -EOPNOTSUPP;
181
182 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
183 return -EINVAL;
184
185 return 0;
186 }
187
188 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
189 {
190 struct inode *inode = file_inode(file);
191 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
192 struct btrfs_inode *binode = BTRFS_I(inode);
193 struct btrfs_root *root = binode->root;
194 struct btrfs_trans_handle *trans;
195 unsigned int fsflags, old_fsflags;
196 int ret;
197 u64 old_flags;
198 unsigned int old_i_flags;
199 umode_t mode;
200
201 if (!inode_owner_or_capable(inode))
202 return -EPERM;
203
204 if (btrfs_root_readonly(root))
205 return -EROFS;
206
207 if (copy_from_user(&fsflags, arg, sizeof(fsflags)))
208 return -EFAULT;
209
210 ret = check_fsflags(fsflags);
211 if (ret)
212 return ret;
213
214 ret = mnt_want_write_file(file);
215 if (ret)
216 return ret;
217
218 inode_lock(inode);
219
220 old_flags = binode->flags;
221 old_i_flags = inode->i_flags;
222 mode = inode->i_mode;
223
224 fsflags = btrfs_mask_fsflags_for_type(inode, fsflags);
225 old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags);
226 if ((fsflags ^ old_fsflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
227 if (!capable(CAP_LINUX_IMMUTABLE)) {
228 ret = -EPERM;
229 goto out_unlock;
230 }
231 }
232
233 if (fsflags & FS_SYNC_FL)
234 binode->flags |= BTRFS_INODE_SYNC;
235 else
236 binode->flags &= ~BTRFS_INODE_SYNC;
237 if (fsflags & FS_IMMUTABLE_FL)
238 binode->flags |= BTRFS_INODE_IMMUTABLE;
239 else
240 binode->flags &= ~BTRFS_INODE_IMMUTABLE;
241 if (fsflags & FS_APPEND_FL)
242 binode->flags |= BTRFS_INODE_APPEND;
243 else
244 binode->flags &= ~BTRFS_INODE_APPEND;
245 if (fsflags & FS_NODUMP_FL)
246 binode->flags |= BTRFS_INODE_NODUMP;
247 else
248 binode->flags &= ~BTRFS_INODE_NODUMP;
249 if (fsflags & FS_NOATIME_FL)
250 binode->flags |= BTRFS_INODE_NOATIME;
251 else
252 binode->flags &= ~BTRFS_INODE_NOATIME;
253 if (fsflags & FS_DIRSYNC_FL)
254 binode->flags |= BTRFS_INODE_DIRSYNC;
255 else
256 binode->flags &= ~BTRFS_INODE_DIRSYNC;
257 if (fsflags & FS_NOCOW_FL) {
258 if (S_ISREG(mode)) {
259 /*
260 * It's safe to turn csums off here, no extents exist.
261 * Otherwise we want the flag to reflect the real COW
262 * status of the file and will not set it.
263 */
264 if (inode->i_size == 0)
265 binode->flags |= BTRFS_INODE_NODATACOW
266 | BTRFS_INODE_NODATASUM;
267 } else {
268 binode->flags |= BTRFS_INODE_NODATACOW;
269 }
270 } else {
271 /*
272 * Revert back under same assumptions as above
273 */
274 if (S_ISREG(mode)) {
275 if (inode->i_size == 0)
276 binode->flags &= ~(BTRFS_INODE_NODATACOW
277 | BTRFS_INODE_NODATASUM);
278 } else {
279 binode->flags &= ~BTRFS_INODE_NODATACOW;
280 }
281 }
282
283 /*
284 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
285 * flag may be changed automatically if compression code won't make
286 * things smaller.
287 */
288 if (fsflags & FS_NOCOMP_FL) {
289 binode->flags &= ~BTRFS_INODE_COMPRESS;
290 binode->flags |= BTRFS_INODE_NOCOMPRESS;
291
292 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
293 if (ret && ret != -ENODATA)
294 goto out_drop;
295 } else if (fsflags & FS_COMPR_FL) {
296 const char *comp;
297
298 binode->flags |= BTRFS_INODE_COMPRESS;
299 binode->flags &= ~BTRFS_INODE_NOCOMPRESS;
300
301 comp = btrfs_compress_type2str(fs_info->compress_type);
302 if (!comp || comp[0] == 0)
303 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
304
305 ret = btrfs_set_prop(inode, "btrfs.compression",
306 comp, strlen(comp), 0);
307 if (ret)
308 goto out_drop;
309
310 } else {
311 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
312 if (ret && ret != -ENODATA)
313 goto out_drop;
314 binode->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
315 }
316
317 trans = btrfs_start_transaction(root, 1);
318 if (IS_ERR(trans)) {
319 ret = PTR_ERR(trans);
320 goto out_drop;
321 }
322
323 btrfs_sync_inode_flags_to_i_flags(inode);
324 inode_inc_iversion(inode);
325 inode->i_ctime = current_time(inode);
326 ret = btrfs_update_inode(trans, root, inode);
327
328 btrfs_end_transaction(trans);
329 out_drop:
330 if (ret) {
331 binode->flags = old_flags;
332 inode->i_flags = old_i_flags;
333 }
334
335 out_unlock:
336 inode_unlock(inode);
337 mnt_drop_write_file(file);
338 return ret;
339 }
340
341 /*
342 * Translate btrfs internal inode flags to xflags as expected by the
343 * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are
344 * silently dropped.
345 */
346 static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags)
347 {
348 unsigned int xflags = 0;
349
350 if (flags & BTRFS_INODE_APPEND)
351 xflags |= FS_XFLAG_APPEND;
352 if (flags & BTRFS_INODE_IMMUTABLE)
353 xflags |= FS_XFLAG_IMMUTABLE;
354 if (flags & BTRFS_INODE_NOATIME)
355 xflags |= FS_XFLAG_NOATIME;
356 if (flags & BTRFS_INODE_NODUMP)
357 xflags |= FS_XFLAG_NODUMP;
358 if (flags & BTRFS_INODE_SYNC)
359 xflags |= FS_XFLAG_SYNC;
360
361 return xflags;
362 }
363
364 /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */
365 static int check_xflags(unsigned int flags)
366 {
367 if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME |
368 FS_XFLAG_NODUMP | FS_XFLAG_SYNC))
369 return -EOPNOTSUPP;
370 return 0;
371 }
372
373 /*
374 * Set the xflags from the internal inode flags. The remaining items of fsxattr
375 * are zeroed.
376 */
377 static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg)
378 {
379 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
380 struct fsxattr fa;
381
382 memset(&fa, 0, sizeof(fa));
383 fa.fsx_xflags = btrfs_inode_flags_to_xflags(binode->flags);
384
385 if (copy_to_user(arg, &fa, sizeof(fa)))
386 return -EFAULT;
387
388 return 0;
389 }
390
391 static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg)
392 {
393 struct inode *inode = file_inode(file);
394 struct btrfs_inode *binode = BTRFS_I(inode);
395 struct btrfs_root *root = binode->root;
396 struct btrfs_trans_handle *trans;
397 struct fsxattr fa;
398 unsigned old_flags;
399 unsigned old_i_flags;
400 int ret = 0;
401
402 if (!inode_owner_or_capable(inode))
403 return -EPERM;
404
405 if (btrfs_root_readonly(root))
406 return -EROFS;
407
408 memset(&fa, 0, sizeof(fa));
409 if (copy_from_user(&fa, arg, sizeof(fa)))
410 return -EFAULT;
411
412 ret = check_xflags(fa.fsx_xflags);
413 if (ret)
414 return ret;
415
416 if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0)
417 return -EOPNOTSUPP;
418
419 ret = mnt_want_write_file(file);
420 if (ret)
421 return ret;
422
423 inode_lock(inode);
424
425 old_flags = binode->flags;
426 old_i_flags = inode->i_flags;
427
428 /* We need the capabilities to change append-only or immutable inode */
429 if (((old_flags & (BTRFS_INODE_APPEND | BTRFS_INODE_IMMUTABLE)) ||
430 (fa.fsx_xflags & (FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE))) &&
431 !capable(CAP_LINUX_IMMUTABLE)) {
432 ret = -EPERM;
433 goto out_unlock;
434 }
435
436 if (fa.fsx_xflags & FS_XFLAG_SYNC)
437 binode->flags |= BTRFS_INODE_SYNC;
438 else
439 binode->flags &= ~BTRFS_INODE_SYNC;
440 if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE)
441 binode->flags |= BTRFS_INODE_IMMUTABLE;
442 else
443 binode->flags &= ~BTRFS_INODE_IMMUTABLE;
444 if (fa.fsx_xflags & FS_XFLAG_APPEND)
445 binode->flags |= BTRFS_INODE_APPEND;
446 else
447 binode->flags &= ~BTRFS_INODE_APPEND;
448 if (fa.fsx_xflags & FS_XFLAG_NODUMP)
449 binode->flags |= BTRFS_INODE_NODUMP;
450 else
451 binode->flags &= ~BTRFS_INODE_NODUMP;
452 if (fa.fsx_xflags & FS_XFLAG_NOATIME)
453 binode->flags |= BTRFS_INODE_NOATIME;
454 else
455 binode->flags &= ~BTRFS_INODE_NOATIME;
456
457 /* 1 item for the inode */
458 trans = btrfs_start_transaction(root, 1);
459 if (IS_ERR(trans)) {
460 ret = PTR_ERR(trans);
461 goto out_unlock;
462 }
463
464 btrfs_sync_inode_flags_to_i_flags(inode);
465 inode_inc_iversion(inode);
466 inode->i_ctime = current_time(inode);
467 ret = btrfs_update_inode(trans, root, inode);
468
469 btrfs_end_transaction(trans);
470
471 out_unlock:
472 if (ret) {
473 binode->flags = old_flags;
474 inode->i_flags = old_i_flags;
475 }
476
477 inode_unlock(inode);
478 mnt_drop_write_file(file);
479
480 return ret;
481 }
482
483 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
484 {
485 struct inode *inode = file_inode(file);
486
487 return put_user(inode->i_generation, arg);
488 }
489
490 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
491 {
492 struct inode *inode = file_inode(file);
493 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
494 struct btrfs_device *device;
495 struct request_queue *q;
496 struct fstrim_range range;
497 u64 minlen = ULLONG_MAX;
498 u64 num_devices = 0;
499 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
500 int ret;
501
502 if (!capable(CAP_SYS_ADMIN))
503 return -EPERM;
504
505 rcu_read_lock();
506 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
507 dev_list) {
508 if (!device->bdev)
509 continue;
510 q = bdev_get_queue(device->bdev);
511 if (blk_queue_discard(q)) {
512 num_devices++;
513 minlen = min_t(u64, q->limits.discard_granularity,
514 minlen);
515 }
516 }
517 rcu_read_unlock();
518
519 if (!num_devices)
520 return -EOPNOTSUPP;
521 if (copy_from_user(&range, arg, sizeof(range)))
522 return -EFAULT;
523 if (range.start > total_bytes ||
524 range.len < fs_info->sb->s_blocksize)
525 return -EINVAL;
526
527 range.len = min(range.len, total_bytes - range.start);
528 range.minlen = max(range.minlen, minlen);
529 ret = btrfs_trim_fs(fs_info, &range);
530 if (ret < 0)
531 return ret;
532
533 if (copy_to_user(arg, &range, sizeof(range)))
534 return -EFAULT;
535
536 return 0;
537 }
538
539 int btrfs_is_empty_uuid(u8 *uuid)
540 {
541 int i;
542
543 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
544 if (uuid[i])
545 return 0;
546 }
547 return 1;
548 }
549
550 static noinline int create_subvol(struct inode *dir,
551 struct dentry *dentry,
552 const char *name, int namelen,
553 u64 *async_transid,
554 struct btrfs_qgroup_inherit *inherit)
555 {
556 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
557 struct btrfs_trans_handle *trans;
558 struct btrfs_key key;
559 struct btrfs_root_item *root_item;
560 struct btrfs_inode_item *inode_item;
561 struct extent_buffer *leaf;
562 struct btrfs_root *root = BTRFS_I(dir)->root;
563 struct btrfs_root *new_root;
564 struct btrfs_block_rsv block_rsv;
565 struct timespec64 cur_time = current_time(dir);
566 struct inode *inode;
567 int ret;
568 int err;
569 u64 objectid;
570 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
571 u64 index = 0;
572 uuid_le new_uuid;
573
574 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
575 if (!root_item)
576 return -ENOMEM;
577
578 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
579 if (ret)
580 goto fail_free;
581
582 /*
583 * Don't create subvolume whose level is not zero. Or qgroup will be
584 * screwed up since it assumes subvolume qgroup's level to be 0.
585 */
586 if (btrfs_qgroup_level(objectid)) {
587 ret = -ENOSPC;
588 goto fail_free;
589 }
590
591 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
592 /*
593 * The same as the snapshot creation, please see the comment
594 * of create_snapshot().
595 */
596 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false);
597 if (ret)
598 goto fail_free;
599
600 trans = btrfs_start_transaction(root, 0);
601 if (IS_ERR(trans)) {
602 ret = PTR_ERR(trans);
603 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
604 goto fail_free;
605 }
606 trans->block_rsv = &block_rsv;
607 trans->bytes_reserved = block_rsv.size;
608
609 ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit);
610 if (ret)
611 goto fail;
612
613 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
614 if (IS_ERR(leaf)) {
615 ret = PTR_ERR(leaf);
616 goto fail;
617 }
618
619 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
620 btrfs_set_header_bytenr(leaf, leaf->start);
621 btrfs_set_header_generation(leaf, trans->transid);
622 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
623 btrfs_set_header_owner(leaf, objectid);
624
625 write_extent_buffer_fsid(leaf, fs_info->fsid);
626 write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid);
627 btrfs_mark_buffer_dirty(leaf);
628
629 inode_item = &root_item->inode;
630 btrfs_set_stack_inode_generation(inode_item, 1);
631 btrfs_set_stack_inode_size(inode_item, 3);
632 btrfs_set_stack_inode_nlink(inode_item, 1);
633 btrfs_set_stack_inode_nbytes(inode_item,
634 fs_info->nodesize);
635 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
636
637 btrfs_set_root_flags(root_item, 0);
638 btrfs_set_root_limit(root_item, 0);
639 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
640
641 btrfs_set_root_bytenr(root_item, leaf->start);
642 btrfs_set_root_generation(root_item, trans->transid);
643 btrfs_set_root_level(root_item, 0);
644 btrfs_set_root_refs(root_item, 1);
645 btrfs_set_root_used(root_item, leaf->len);
646 btrfs_set_root_last_snapshot(root_item, 0);
647
648 btrfs_set_root_generation_v2(root_item,
649 btrfs_root_generation(root_item));
650 uuid_le_gen(&new_uuid);
651 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
652 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
653 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
654 root_item->ctime = root_item->otime;
655 btrfs_set_root_ctransid(root_item, trans->transid);
656 btrfs_set_root_otransid(root_item, trans->transid);
657
658 btrfs_tree_unlock(leaf);
659 free_extent_buffer(leaf);
660 leaf = NULL;
661
662 btrfs_set_root_dirid(root_item, new_dirid);
663
664 key.objectid = objectid;
665 key.offset = 0;
666 key.type = BTRFS_ROOT_ITEM_KEY;
667 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
668 root_item);
669 if (ret)
670 goto fail;
671
672 key.offset = (u64)-1;
673 new_root = btrfs_read_fs_root_no_name(fs_info, &key);
674 if (IS_ERR(new_root)) {
675 ret = PTR_ERR(new_root);
676 btrfs_abort_transaction(trans, ret);
677 goto fail;
678 }
679
680 btrfs_record_root_in_trans(trans, new_root);
681
682 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
683 if (ret) {
684 /* We potentially lose an unused inode item here */
685 btrfs_abort_transaction(trans, ret);
686 goto fail;
687 }
688
689 mutex_lock(&new_root->objectid_mutex);
690 new_root->highest_objectid = new_dirid;
691 mutex_unlock(&new_root->objectid_mutex);
692
693 /*
694 * insert the directory item
695 */
696 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
697 if (ret) {
698 btrfs_abort_transaction(trans, ret);
699 goto fail;
700 }
701
702 ret = btrfs_insert_dir_item(trans, root,
703 name, namelen, BTRFS_I(dir), &key,
704 BTRFS_FT_DIR, index);
705 if (ret) {
706 btrfs_abort_transaction(trans, ret);
707 goto fail;
708 }
709
710 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
711 ret = btrfs_update_inode(trans, root, dir);
712 BUG_ON(ret);
713
714 ret = btrfs_add_root_ref(trans, fs_info,
715 objectid, root->root_key.objectid,
716 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
717 BUG_ON(ret);
718
719 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
720 BTRFS_UUID_KEY_SUBVOL, objectid);
721 if (ret)
722 btrfs_abort_transaction(trans, ret);
723
724 fail:
725 kfree(root_item);
726 trans->block_rsv = NULL;
727 trans->bytes_reserved = 0;
728 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
729
730 if (async_transid) {
731 *async_transid = trans->transid;
732 err = btrfs_commit_transaction_async(trans, 1);
733 if (err)
734 err = btrfs_commit_transaction(trans);
735 } else {
736 err = btrfs_commit_transaction(trans);
737 }
738 if (err && !ret)
739 ret = err;
740
741 if (!ret) {
742 inode = btrfs_lookup_dentry(dir, dentry);
743 if (IS_ERR(inode))
744 return PTR_ERR(inode);
745 d_instantiate(dentry, inode);
746 }
747 return ret;
748
749 fail_free:
750 kfree(root_item);
751 return ret;
752 }
753
754 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
755 struct dentry *dentry,
756 u64 *async_transid, bool readonly,
757 struct btrfs_qgroup_inherit *inherit)
758 {
759 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
760 struct inode *inode;
761 struct btrfs_pending_snapshot *pending_snapshot;
762 struct btrfs_trans_handle *trans;
763 int ret;
764
765 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
766 return -EINVAL;
767
768 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
769 if (!pending_snapshot)
770 return -ENOMEM;
771
772 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
773 GFP_KERNEL);
774 pending_snapshot->path = btrfs_alloc_path();
775 if (!pending_snapshot->root_item || !pending_snapshot->path) {
776 ret = -ENOMEM;
777 goto free_pending;
778 }
779
780 atomic_inc(&root->will_be_snapshotted);
781 smp_mb__after_atomic();
782 /* wait for no snapshot writes */
783 wait_event(root->subv_writers->wait,
784 percpu_counter_sum(&root->subv_writers->counter) == 0);
785
786 ret = btrfs_start_delalloc_inodes(root);
787 if (ret)
788 goto dec_and_free;
789
790 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
791
792 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
793 BTRFS_BLOCK_RSV_TEMP);
794 /*
795 * 1 - parent dir inode
796 * 2 - dir entries
797 * 1 - root item
798 * 2 - root ref/backref
799 * 1 - root of snapshot
800 * 1 - UUID item
801 */
802 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
803 &pending_snapshot->block_rsv, 8,
804 false);
805 if (ret)
806 goto dec_and_free;
807
808 pending_snapshot->dentry = dentry;
809 pending_snapshot->root = root;
810 pending_snapshot->readonly = readonly;
811 pending_snapshot->dir = dir;
812 pending_snapshot->inherit = inherit;
813
814 trans = btrfs_start_transaction(root, 0);
815 if (IS_ERR(trans)) {
816 ret = PTR_ERR(trans);
817 goto fail;
818 }
819
820 spin_lock(&fs_info->trans_lock);
821 list_add(&pending_snapshot->list,
822 &trans->transaction->pending_snapshots);
823 spin_unlock(&fs_info->trans_lock);
824 if (async_transid) {
825 *async_transid = trans->transid;
826 ret = btrfs_commit_transaction_async(trans, 1);
827 if (ret)
828 ret = btrfs_commit_transaction(trans);
829 } else {
830 ret = btrfs_commit_transaction(trans);
831 }
832 if (ret)
833 goto fail;
834
835 ret = pending_snapshot->error;
836 if (ret)
837 goto fail;
838
839 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
840 if (ret)
841 goto fail;
842
843 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
844 if (IS_ERR(inode)) {
845 ret = PTR_ERR(inode);
846 goto fail;
847 }
848
849 d_instantiate(dentry, inode);
850 ret = 0;
851 fail:
852 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
853 dec_and_free:
854 if (atomic_dec_and_test(&root->will_be_snapshotted))
855 wake_up_var(&root->will_be_snapshotted);
856 free_pending:
857 kfree(pending_snapshot->root_item);
858 btrfs_free_path(pending_snapshot->path);
859 kfree(pending_snapshot);
860
861 return ret;
862 }
863
864 /* copy of may_delete in fs/namei.c()
865 * Check whether we can remove a link victim from directory dir, check
866 * whether the type of victim is right.
867 * 1. We can't do it if dir is read-only (done in permission())
868 * 2. We should have write and exec permissions on dir
869 * 3. We can't remove anything from append-only dir
870 * 4. We can't do anything with immutable dir (done in permission())
871 * 5. If the sticky bit on dir is set we should either
872 * a. be owner of dir, or
873 * b. be owner of victim, or
874 * c. have CAP_FOWNER capability
875 * 6. If the victim is append-only or immutable we can't do anything with
876 * links pointing to it.
877 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
878 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
879 * 9. We can't remove a root or mountpoint.
880 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
881 * nfs_async_unlink().
882 */
883
884 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
885 {
886 int error;
887
888 if (d_really_is_negative(victim))
889 return -ENOENT;
890
891 BUG_ON(d_inode(victim->d_parent) != dir);
892 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
893
894 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
895 if (error)
896 return error;
897 if (IS_APPEND(dir))
898 return -EPERM;
899 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
900 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
901 return -EPERM;
902 if (isdir) {
903 if (!d_is_dir(victim))
904 return -ENOTDIR;
905 if (IS_ROOT(victim))
906 return -EBUSY;
907 } else if (d_is_dir(victim))
908 return -EISDIR;
909 if (IS_DEADDIR(dir))
910 return -ENOENT;
911 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
912 return -EBUSY;
913 return 0;
914 }
915
916 /* copy of may_create in fs/namei.c() */
917 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
918 {
919 if (d_really_is_positive(child))
920 return -EEXIST;
921 if (IS_DEADDIR(dir))
922 return -ENOENT;
923 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
924 }
925
926 /*
927 * Create a new subvolume below @parent. This is largely modeled after
928 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
929 * inside this filesystem so it's quite a bit simpler.
930 */
931 static noinline int btrfs_mksubvol(const struct path *parent,
932 const char *name, int namelen,
933 struct btrfs_root *snap_src,
934 u64 *async_transid, bool readonly,
935 struct btrfs_qgroup_inherit *inherit)
936 {
937 struct inode *dir = d_inode(parent->dentry);
938 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
939 struct dentry *dentry;
940 int error;
941
942 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
943 if (error == -EINTR)
944 return error;
945
946 dentry = lookup_one_len(name, parent->dentry, namelen);
947 error = PTR_ERR(dentry);
948 if (IS_ERR(dentry))
949 goto out_unlock;
950
951 error = btrfs_may_create(dir, dentry);
952 if (error)
953 goto out_dput;
954
955 /*
956 * even if this name doesn't exist, we may get hash collisions.
957 * check for them now when we can safely fail
958 */
959 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
960 dir->i_ino, name,
961 namelen);
962 if (error)
963 goto out_dput;
964
965 down_read(&fs_info->subvol_sem);
966
967 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
968 goto out_up_read;
969
970 if (snap_src) {
971 error = create_snapshot(snap_src, dir, dentry,
972 async_transid, readonly, inherit);
973 } else {
974 error = create_subvol(dir, dentry, name, namelen,
975 async_transid, inherit);
976 }
977 if (!error)
978 fsnotify_mkdir(dir, dentry);
979 out_up_read:
980 up_read(&fs_info->subvol_sem);
981 out_dput:
982 dput(dentry);
983 out_unlock:
984 inode_unlock(dir);
985 return error;
986 }
987
988 /*
989 * When we're defragging a range, we don't want to kick it off again
990 * if it is really just waiting for delalloc to send it down.
991 * If we find a nice big extent or delalloc range for the bytes in the
992 * file you want to defrag, we return 0 to let you know to skip this
993 * part of the file
994 */
995 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
996 {
997 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
998 struct extent_map *em = NULL;
999 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1000 u64 end;
1001
1002 read_lock(&em_tree->lock);
1003 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
1004 read_unlock(&em_tree->lock);
1005
1006 if (em) {
1007 end = extent_map_end(em);
1008 free_extent_map(em);
1009 if (end - offset > thresh)
1010 return 0;
1011 }
1012 /* if we already have a nice delalloc here, just stop */
1013 thresh /= 2;
1014 end = count_range_bits(io_tree, &offset, offset + thresh,
1015 thresh, EXTENT_DELALLOC, 1);
1016 if (end >= thresh)
1017 return 0;
1018 return 1;
1019 }
1020
1021 /*
1022 * helper function to walk through a file and find extents
1023 * newer than a specific transid, and smaller than thresh.
1024 *
1025 * This is used by the defragging code to find new and small
1026 * extents
1027 */
1028 static int find_new_extents(struct btrfs_root *root,
1029 struct inode *inode, u64 newer_than,
1030 u64 *off, u32 thresh)
1031 {
1032 struct btrfs_path *path;
1033 struct btrfs_key min_key;
1034 struct extent_buffer *leaf;
1035 struct btrfs_file_extent_item *extent;
1036 int type;
1037 int ret;
1038 u64 ino = btrfs_ino(BTRFS_I(inode));
1039
1040 path = btrfs_alloc_path();
1041 if (!path)
1042 return -ENOMEM;
1043
1044 min_key.objectid = ino;
1045 min_key.type = BTRFS_EXTENT_DATA_KEY;
1046 min_key.offset = *off;
1047
1048 while (1) {
1049 ret = btrfs_search_forward(root, &min_key, path, newer_than);
1050 if (ret != 0)
1051 goto none;
1052 process_slot:
1053 if (min_key.objectid != ino)
1054 goto none;
1055 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1056 goto none;
1057
1058 leaf = path->nodes[0];
1059 extent = btrfs_item_ptr(leaf, path->slots[0],
1060 struct btrfs_file_extent_item);
1061
1062 type = btrfs_file_extent_type(leaf, extent);
1063 if (type == BTRFS_FILE_EXTENT_REG &&
1064 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1065 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1066 *off = min_key.offset;
1067 btrfs_free_path(path);
1068 return 0;
1069 }
1070
1071 path->slots[0]++;
1072 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1073 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1074 goto process_slot;
1075 }
1076
1077 if (min_key.offset == (u64)-1)
1078 goto none;
1079
1080 min_key.offset++;
1081 btrfs_release_path(path);
1082 }
1083 none:
1084 btrfs_free_path(path);
1085 return -ENOENT;
1086 }
1087
1088 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1089 {
1090 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1091 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1092 struct extent_map *em;
1093 u64 len = PAGE_SIZE;
1094
1095 /*
1096 * hopefully we have this extent in the tree already, try without
1097 * the full extent lock
1098 */
1099 read_lock(&em_tree->lock);
1100 em = lookup_extent_mapping(em_tree, start, len);
1101 read_unlock(&em_tree->lock);
1102
1103 if (!em) {
1104 struct extent_state *cached = NULL;
1105 u64 end = start + len - 1;
1106
1107 /* get the big lock and read metadata off disk */
1108 lock_extent_bits(io_tree, start, end, &cached);
1109 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0);
1110 unlock_extent_cached(io_tree, start, end, &cached);
1111
1112 if (IS_ERR(em))
1113 return NULL;
1114 }
1115
1116 return em;
1117 }
1118
1119 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1120 {
1121 struct extent_map *next;
1122 bool ret = true;
1123
1124 /* this is the last extent */
1125 if (em->start + em->len >= i_size_read(inode))
1126 return false;
1127
1128 next = defrag_lookup_extent(inode, em->start + em->len);
1129 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1130 ret = false;
1131 else if ((em->block_start + em->block_len == next->block_start) &&
1132 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1133 ret = false;
1134
1135 free_extent_map(next);
1136 return ret;
1137 }
1138
1139 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1140 u64 *last_len, u64 *skip, u64 *defrag_end,
1141 int compress)
1142 {
1143 struct extent_map *em;
1144 int ret = 1;
1145 bool next_mergeable = true;
1146 bool prev_mergeable = true;
1147
1148 /*
1149 * make sure that once we start defragging an extent, we keep on
1150 * defragging it
1151 */
1152 if (start < *defrag_end)
1153 return 1;
1154
1155 *skip = 0;
1156
1157 em = defrag_lookup_extent(inode, start);
1158 if (!em)
1159 return 0;
1160
1161 /* this will cover holes, and inline extents */
1162 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1163 ret = 0;
1164 goto out;
1165 }
1166
1167 if (!*defrag_end)
1168 prev_mergeable = false;
1169
1170 next_mergeable = defrag_check_next_extent(inode, em);
1171 /*
1172 * we hit a real extent, if it is big or the next extent is not a
1173 * real extent, don't bother defragging it
1174 */
1175 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1176 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1177 ret = 0;
1178 out:
1179 /*
1180 * last_len ends up being a counter of how many bytes we've defragged.
1181 * every time we choose not to defrag an extent, we reset *last_len
1182 * so that the next tiny extent will force a defrag.
1183 *
1184 * The end result of this is that tiny extents before a single big
1185 * extent will force at least part of that big extent to be defragged.
1186 */
1187 if (ret) {
1188 *defrag_end = extent_map_end(em);
1189 } else {
1190 *last_len = 0;
1191 *skip = extent_map_end(em);
1192 *defrag_end = 0;
1193 }
1194
1195 free_extent_map(em);
1196 return ret;
1197 }
1198
1199 /*
1200 * it doesn't do much good to defrag one or two pages
1201 * at a time. This pulls in a nice chunk of pages
1202 * to COW and defrag.
1203 *
1204 * It also makes sure the delalloc code has enough
1205 * dirty data to avoid making new small extents as part
1206 * of the defrag
1207 *
1208 * It's a good idea to start RA on this range
1209 * before calling this.
1210 */
1211 static int cluster_pages_for_defrag(struct inode *inode,
1212 struct page **pages,
1213 unsigned long start_index,
1214 unsigned long num_pages)
1215 {
1216 unsigned long file_end;
1217 u64 isize = i_size_read(inode);
1218 u64 page_start;
1219 u64 page_end;
1220 u64 page_cnt;
1221 int ret;
1222 int i;
1223 int i_done;
1224 struct btrfs_ordered_extent *ordered;
1225 struct extent_state *cached_state = NULL;
1226 struct extent_io_tree *tree;
1227 struct extent_changeset *data_reserved = NULL;
1228 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1229
1230 file_end = (isize - 1) >> PAGE_SHIFT;
1231 if (!isize || start_index > file_end)
1232 return 0;
1233
1234 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1235
1236 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
1237 start_index << PAGE_SHIFT,
1238 page_cnt << PAGE_SHIFT);
1239 if (ret)
1240 return ret;
1241 i_done = 0;
1242 tree = &BTRFS_I(inode)->io_tree;
1243
1244 /* step one, lock all the pages */
1245 for (i = 0; i < page_cnt; i++) {
1246 struct page *page;
1247 again:
1248 page = find_or_create_page(inode->i_mapping,
1249 start_index + i, mask);
1250 if (!page)
1251 break;
1252
1253 page_start = page_offset(page);
1254 page_end = page_start + PAGE_SIZE - 1;
1255 while (1) {
1256 lock_extent_bits(tree, page_start, page_end,
1257 &cached_state);
1258 ordered = btrfs_lookup_ordered_extent(inode,
1259 page_start);
1260 unlock_extent_cached(tree, page_start, page_end,
1261 &cached_state);
1262 if (!ordered)
1263 break;
1264
1265 unlock_page(page);
1266 btrfs_start_ordered_extent(inode, ordered, 1);
1267 btrfs_put_ordered_extent(ordered);
1268 lock_page(page);
1269 /*
1270 * we unlocked the page above, so we need check if
1271 * it was released or not.
1272 */
1273 if (page->mapping != inode->i_mapping) {
1274 unlock_page(page);
1275 put_page(page);
1276 goto again;
1277 }
1278 }
1279
1280 if (!PageUptodate(page)) {
1281 btrfs_readpage(NULL, page);
1282 lock_page(page);
1283 if (!PageUptodate(page)) {
1284 unlock_page(page);
1285 put_page(page);
1286 ret = -EIO;
1287 break;
1288 }
1289 }
1290
1291 if (page->mapping != inode->i_mapping) {
1292 unlock_page(page);
1293 put_page(page);
1294 goto again;
1295 }
1296
1297 pages[i] = page;
1298 i_done++;
1299 }
1300 if (!i_done || ret)
1301 goto out;
1302
1303 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1304 goto out;
1305
1306 /*
1307 * so now we have a nice long stream of locked
1308 * and up to date pages, lets wait on them
1309 */
1310 for (i = 0; i < i_done; i++)
1311 wait_on_page_writeback(pages[i]);
1312
1313 page_start = page_offset(pages[0]);
1314 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1315
1316 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1317 page_start, page_end - 1, &cached_state);
1318 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1319 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1320 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1321 &cached_state);
1322
1323 if (i_done != page_cnt) {
1324 spin_lock(&BTRFS_I(inode)->lock);
1325 BTRFS_I(inode)->outstanding_extents++;
1326 spin_unlock(&BTRFS_I(inode)->lock);
1327 btrfs_delalloc_release_space(inode, data_reserved,
1328 start_index << PAGE_SHIFT,
1329 (page_cnt - i_done) << PAGE_SHIFT, true);
1330 }
1331
1332
1333 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1334 &cached_state);
1335
1336 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1337 page_start, page_end - 1, &cached_state);
1338
1339 for (i = 0; i < i_done; i++) {
1340 clear_page_dirty_for_io(pages[i]);
1341 ClearPageChecked(pages[i]);
1342 set_page_extent_mapped(pages[i]);
1343 set_page_dirty(pages[i]);
1344 unlock_page(pages[i]);
1345 put_page(pages[i]);
1346 }
1347 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1348 false);
1349 extent_changeset_free(data_reserved);
1350 return i_done;
1351 out:
1352 for (i = 0; i < i_done; i++) {
1353 unlock_page(pages[i]);
1354 put_page(pages[i]);
1355 }
1356 btrfs_delalloc_release_space(inode, data_reserved,
1357 start_index << PAGE_SHIFT,
1358 page_cnt << PAGE_SHIFT, true);
1359 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1360 true);
1361 extent_changeset_free(data_reserved);
1362 return ret;
1363
1364 }
1365
1366 int btrfs_defrag_file(struct inode *inode, struct file *file,
1367 struct btrfs_ioctl_defrag_range_args *range,
1368 u64 newer_than, unsigned long max_to_defrag)
1369 {
1370 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1371 struct btrfs_root *root = BTRFS_I(inode)->root;
1372 struct file_ra_state *ra = NULL;
1373 unsigned long last_index;
1374 u64 isize = i_size_read(inode);
1375 u64 last_len = 0;
1376 u64 skip = 0;
1377 u64 defrag_end = 0;
1378 u64 newer_off = range->start;
1379 unsigned long i;
1380 unsigned long ra_index = 0;
1381 int ret;
1382 int defrag_count = 0;
1383 int compress_type = BTRFS_COMPRESS_ZLIB;
1384 u32 extent_thresh = range->extent_thresh;
1385 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1386 unsigned long cluster = max_cluster;
1387 u64 new_align = ~((u64)SZ_128K - 1);
1388 struct page **pages = NULL;
1389 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1390
1391 if (isize == 0)
1392 return 0;
1393
1394 if (range->start >= isize)
1395 return -EINVAL;
1396
1397 if (do_compress) {
1398 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1399 return -EINVAL;
1400 if (range->compress_type)
1401 compress_type = range->compress_type;
1402 }
1403
1404 if (extent_thresh == 0)
1405 extent_thresh = SZ_256K;
1406
1407 /*
1408 * If we were not given a file, allocate a readahead context. As
1409 * readahead is just an optimization, defrag will work without it so
1410 * we don't error out.
1411 */
1412 if (!file) {
1413 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1414 if (ra)
1415 file_ra_state_init(ra, inode->i_mapping);
1416 } else {
1417 ra = &file->f_ra;
1418 }
1419
1420 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1421 if (!pages) {
1422 ret = -ENOMEM;
1423 goto out_ra;
1424 }
1425
1426 /* find the last page to defrag */
1427 if (range->start + range->len > range->start) {
1428 last_index = min_t(u64, isize - 1,
1429 range->start + range->len - 1) >> PAGE_SHIFT;
1430 } else {
1431 last_index = (isize - 1) >> PAGE_SHIFT;
1432 }
1433
1434 if (newer_than) {
1435 ret = find_new_extents(root, inode, newer_than,
1436 &newer_off, SZ_64K);
1437 if (!ret) {
1438 range->start = newer_off;
1439 /*
1440 * we always align our defrag to help keep
1441 * the extents in the file evenly spaced
1442 */
1443 i = (newer_off & new_align) >> PAGE_SHIFT;
1444 } else
1445 goto out_ra;
1446 } else {
1447 i = range->start >> PAGE_SHIFT;
1448 }
1449 if (!max_to_defrag)
1450 max_to_defrag = last_index - i + 1;
1451
1452 /*
1453 * make writeback starts from i, so the defrag range can be
1454 * written sequentially.
1455 */
1456 if (i < inode->i_mapping->writeback_index)
1457 inode->i_mapping->writeback_index = i;
1458
1459 while (i <= last_index && defrag_count < max_to_defrag &&
1460 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1461 /*
1462 * make sure we stop running if someone unmounts
1463 * the FS
1464 */
1465 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1466 break;
1467
1468 if (btrfs_defrag_cancelled(fs_info)) {
1469 btrfs_debug(fs_info, "defrag_file cancelled");
1470 ret = -EAGAIN;
1471 break;
1472 }
1473
1474 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1475 extent_thresh, &last_len, &skip,
1476 &defrag_end, do_compress)){
1477 unsigned long next;
1478 /*
1479 * the should_defrag function tells us how much to skip
1480 * bump our counter by the suggested amount
1481 */
1482 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1483 i = max(i + 1, next);
1484 continue;
1485 }
1486
1487 if (!newer_than) {
1488 cluster = (PAGE_ALIGN(defrag_end) >>
1489 PAGE_SHIFT) - i;
1490 cluster = min(cluster, max_cluster);
1491 } else {
1492 cluster = max_cluster;
1493 }
1494
1495 if (i + cluster > ra_index) {
1496 ra_index = max(i, ra_index);
1497 if (ra)
1498 page_cache_sync_readahead(inode->i_mapping, ra,
1499 file, ra_index, cluster);
1500 ra_index += cluster;
1501 }
1502
1503 inode_lock(inode);
1504 if (do_compress)
1505 BTRFS_I(inode)->defrag_compress = compress_type;
1506 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1507 if (ret < 0) {
1508 inode_unlock(inode);
1509 goto out_ra;
1510 }
1511
1512 defrag_count += ret;
1513 balance_dirty_pages_ratelimited(inode->i_mapping);
1514 inode_unlock(inode);
1515
1516 if (newer_than) {
1517 if (newer_off == (u64)-1)
1518 break;
1519
1520 if (ret > 0)
1521 i += ret;
1522
1523 newer_off = max(newer_off + 1,
1524 (u64)i << PAGE_SHIFT);
1525
1526 ret = find_new_extents(root, inode, newer_than,
1527 &newer_off, SZ_64K);
1528 if (!ret) {
1529 range->start = newer_off;
1530 i = (newer_off & new_align) >> PAGE_SHIFT;
1531 } else {
1532 break;
1533 }
1534 } else {
1535 if (ret > 0) {
1536 i += ret;
1537 last_len += ret << PAGE_SHIFT;
1538 } else {
1539 i++;
1540 last_len = 0;
1541 }
1542 }
1543 }
1544
1545 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1546 filemap_flush(inode->i_mapping);
1547 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1548 &BTRFS_I(inode)->runtime_flags))
1549 filemap_flush(inode->i_mapping);
1550 }
1551
1552 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1553 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1554 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1555 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1556 }
1557
1558 ret = defrag_count;
1559
1560 out_ra:
1561 if (do_compress) {
1562 inode_lock(inode);
1563 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1564 inode_unlock(inode);
1565 }
1566 if (!file)
1567 kfree(ra);
1568 kfree(pages);
1569 return ret;
1570 }
1571
1572 static noinline int btrfs_ioctl_resize(struct file *file,
1573 void __user *arg)
1574 {
1575 struct inode *inode = file_inode(file);
1576 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1577 u64 new_size;
1578 u64 old_size;
1579 u64 devid = 1;
1580 struct btrfs_root *root = BTRFS_I(inode)->root;
1581 struct btrfs_ioctl_vol_args *vol_args;
1582 struct btrfs_trans_handle *trans;
1583 struct btrfs_device *device = NULL;
1584 char *sizestr;
1585 char *retptr;
1586 char *devstr = NULL;
1587 int ret = 0;
1588 int mod = 0;
1589
1590 if (!capable(CAP_SYS_ADMIN))
1591 return -EPERM;
1592
1593 ret = mnt_want_write_file(file);
1594 if (ret)
1595 return ret;
1596
1597 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
1598 mnt_drop_write_file(file);
1599 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1600 }
1601
1602 vol_args = memdup_user(arg, sizeof(*vol_args));
1603 if (IS_ERR(vol_args)) {
1604 ret = PTR_ERR(vol_args);
1605 goto out;
1606 }
1607
1608 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1609
1610 sizestr = vol_args->name;
1611 devstr = strchr(sizestr, ':');
1612 if (devstr) {
1613 sizestr = devstr + 1;
1614 *devstr = '\0';
1615 devstr = vol_args->name;
1616 ret = kstrtoull(devstr, 10, &devid);
1617 if (ret)
1618 goto out_free;
1619 if (!devid) {
1620 ret = -EINVAL;
1621 goto out_free;
1622 }
1623 btrfs_info(fs_info, "resizing devid %llu", devid);
1624 }
1625
1626 device = btrfs_find_device(fs_info, devid, NULL, NULL);
1627 if (!device) {
1628 btrfs_info(fs_info, "resizer unable to find device %llu",
1629 devid);
1630 ret = -ENODEV;
1631 goto out_free;
1632 }
1633
1634 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1635 btrfs_info(fs_info,
1636 "resizer unable to apply on readonly device %llu",
1637 devid);
1638 ret = -EPERM;
1639 goto out_free;
1640 }
1641
1642 if (!strcmp(sizestr, "max"))
1643 new_size = device->bdev->bd_inode->i_size;
1644 else {
1645 if (sizestr[0] == '-') {
1646 mod = -1;
1647 sizestr++;
1648 } else if (sizestr[0] == '+') {
1649 mod = 1;
1650 sizestr++;
1651 }
1652 new_size = memparse(sizestr, &retptr);
1653 if (*retptr != '\0' || new_size == 0) {
1654 ret = -EINVAL;
1655 goto out_free;
1656 }
1657 }
1658
1659 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1660 ret = -EPERM;
1661 goto out_free;
1662 }
1663
1664 old_size = btrfs_device_get_total_bytes(device);
1665
1666 if (mod < 0) {
1667 if (new_size > old_size) {
1668 ret = -EINVAL;
1669 goto out_free;
1670 }
1671 new_size = old_size - new_size;
1672 } else if (mod > 0) {
1673 if (new_size > ULLONG_MAX - old_size) {
1674 ret = -ERANGE;
1675 goto out_free;
1676 }
1677 new_size = old_size + new_size;
1678 }
1679
1680 if (new_size < SZ_256M) {
1681 ret = -EINVAL;
1682 goto out_free;
1683 }
1684 if (new_size > device->bdev->bd_inode->i_size) {
1685 ret = -EFBIG;
1686 goto out_free;
1687 }
1688
1689 new_size = round_down(new_size, fs_info->sectorsize);
1690
1691 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1692 rcu_str_deref(device->name), new_size);
1693
1694 if (new_size > old_size) {
1695 trans = btrfs_start_transaction(root, 0);
1696 if (IS_ERR(trans)) {
1697 ret = PTR_ERR(trans);
1698 goto out_free;
1699 }
1700 ret = btrfs_grow_device(trans, device, new_size);
1701 btrfs_commit_transaction(trans);
1702 } else if (new_size < old_size) {
1703 ret = btrfs_shrink_device(device, new_size);
1704 } /* equal, nothing need to do */
1705
1706 out_free:
1707 kfree(vol_args);
1708 out:
1709 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
1710 mnt_drop_write_file(file);
1711 return ret;
1712 }
1713
1714 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1715 const char *name, unsigned long fd, int subvol,
1716 u64 *transid, bool readonly,
1717 struct btrfs_qgroup_inherit *inherit)
1718 {
1719 int namelen;
1720 int ret = 0;
1721
1722 if (!S_ISDIR(file_inode(file)->i_mode))
1723 return -ENOTDIR;
1724
1725 ret = mnt_want_write_file(file);
1726 if (ret)
1727 goto out;
1728
1729 namelen = strlen(name);
1730 if (strchr(name, '/')) {
1731 ret = -EINVAL;
1732 goto out_drop_write;
1733 }
1734
1735 if (name[0] == '.' &&
1736 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1737 ret = -EEXIST;
1738 goto out_drop_write;
1739 }
1740
1741 if (subvol) {
1742 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1743 NULL, transid, readonly, inherit);
1744 } else {
1745 struct fd src = fdget(fd);
1746 struct inode *src_inode;
1747 if (!src.file) {
1748 ret = -EINVAL;
1749 goto out_drop_write;
1750 }
1751
1752 src_inode = file_inode(src.file);
1753 if (src_inode->i_sb != file_inode(file)->i_sb) {
1754 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1755 "Snapshot src from another FS");
1756 ret = -EXDEV;
1757 } else if (!inode_owner_or_capable(src_inode)) {
1758 /*
1759 * Subvolume creation is not restricted, but snapshots
1760 * are limited to own subvolumes only
1761 */
1762 ret = -EPERM;
1763 } else {
1764 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1765 BTRFS_I(src_inode)->root,
1766 transid, readonly, inherit);
1767 }
1768 fdput(src);
1769 }
1770 out_drop_write:
1771 mnt_drop_write_file(file);
1772 out:
1773 return ret;
1774 }
1775
1776 static noinline int btrfs_ioctl_snap_create(struct file *file,
1777 void __user *arg, int subvol)
1778 {
1779 struct btrfs_ioctl_vol_args *vol_args;
1780 int ret;
1781
1782 if (!S_ISDIR(file_inode(file)->i_mode))
1783 return -ENOTDIR;
1784
1785 vol_args = memdup_user(arg, sizeof(*vol_args));
1786 if (IS_ERR(vol_args))
1787 return PTR_ERR(vol_args);
1788 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1789
1790 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1791 vol_args->fd, subvol,
1792 NULL, false, NULL);
1793
1794 kfree(vol_args);
1795 return ret;
1796 }
1797
1798 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1799 void __user *arg, int subvol)
1800 {
1801 struct btrfs_ioctl_vol_args_v2 *vol_args;
1802 int ret;
1803 u64 transid = 0;
1804 u64 *ptr = NULL;
1805 bool readonly = false;
1806 struct btrfs_qgroup_inherit *inherit = NULL;
1807
1808 if (!S_ISDIR(file_inode(file)->i_mode))
1809 return -ENOTDIR;
1810
1811 vol_args = memdup_user(arg, sizeof(*vol_args));
1812 if (IS_ERR(vol_args))
1813 return PTR_ERR(vol_args);
1814 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1815
1816 if (vol_args->flags &
1817 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1818 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1819 ret = -EOPNOTSUPP;
1820 goto free_args;
1821 }
1822
1823 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1824 ptr = &transid;
1825 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1826 readonly = true;
1827 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1828 if (vol_args->size > PAGE_SIZE) {
1829 ret = -EINVAL;
1830 goto free_args;
1831 }
1832 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1833 if (IS_ERR(inherit)) {
1834 ret = PTR_ERR(inherit);
1835 goto free_args;
1836 }
1837 }
1838
1839 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1840 vol_args->fd, subvol, ptr,
1841 readonly, inherit);
1842 if (ret)
1843 goto free_inherit;
1844
1845 if (ptr && copy_to_user(arg +
1846 offsetof(struct btrfs_ioctl_vol_args_v2,
1847 transid),
1848 ptr, sizeof(*ptr)))
1849 ret = -EFAULT;
1850
1851 free_inherit:
1852 kfree(inherit);
1853 free_args:
1854 kfree(vol_args);
1855 return ret;
1856 }
1857
1858 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1859 void __user *arg)
1860 {
1861 struct inode *inode = file_inode(file);
1862 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1863 struct btrfs_root *root = BTRFS_I(inode)->root;
1864 int ret = 0;
1865 u64 flags = 0;
1866
1867 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1868 return -EINVAL;
1869
1870 down_read(&fs_info->subvol_sem);
1871 if (btrfs_root_readonly(root))
1872 flags |= BTRFS_SUBVOL_RDONLY;
1873 up_read(&fs_info->subvol_sem);
1874
1875 if (copy_to_user(arg, &flags, sizeof(flags)))
1876 ret = -EFAULT;
1877
1878 return ret;
1879 }
1880
1881 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1882 void __user *arg)
1883 {
1884 struct inode *inode = file_inode(file);
1885 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1886 struct btrfs_root *root = BTRFS_I(inode)->root;
1887 struct btrfs_trans_handle *trans;
1888 u64 root_flags;
1889 u64 flags;
1890 int ret = 0;
1891
1892 if (!inode_owner_or_capable(inode))
1893 return -EPERM;
1894
1895 ret = mnt_want_write_file(file);
1896 if (ret)
1897 goto out;
1898
1899 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1900 ret = -EINVAL;
1901 goto out_drop_write;
1902 }
1903
1904 if (copy_from_user(&flags, arg, sizeof(flags))) {
1905 ret = -EFAULT;
1906 goto out_drop_write;
1907 }
1908
1909 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1910 ret = -EINVAL;
1911 goto out_drop_write;
1912 }
1913
1914 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1915 ret = -EOPNOTSUPP;
1916 goto out_drop_write;
1917 }
1918
1919 down_write(&fs_info->subvol_sem);
1920
1921 /* nothing to do */
1922 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1923 goto out_drop_sem;
1924
1925 root_flags = btrfs_root_flags(&root->root_item);
1926 if (flags & BTRFS_SUBVOL_RDONLY) {
1927 btrfs_set_root_flags(&root->root_item,
1928 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1929 } else {
1930 /*
1931 * Block RO -> RW transition if this subvolume is involved in
1932 * send
1933 */
1934 spin_lock(&root->root_item_lock);
1935 if (root->send_in_progress == 0) {
1936 btrfs_set_root_flags(&root->root_item,
1937 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1938 spin_unlock(&root->root_item_lock);
1939 } else {
1940 spin_unlock(&root->root_item_lock);
1941 btrfs_warn(fs_info,
1942 "Attempt to set subvolume %llu read-write during send",
1943 root->root_key.objectid);
1944 ret = -EPERM;
1945 goto out_drop_sem;
1946 }
1947 }
1948
1949 trans = btrfs_start_transaction(root, 1);
1950 if (IS_ERR(trans)) {
1951 ret = PTR_ERR(trans);
1952 goto out_reset;
1953 }
1954
1955 ret = btrfs_update_root(trans, fs_info->tree_root,
1956 &root->root_key, &root->root_item);
1957 if (ret < 0) {
1958 btrfs_end_transaction(trans);
1959 goto out_reset;
1960 }
1961
1962 ret = btrfs_commit_transaction(trans);
1963
1964 out_reset:
1965 if (ret)
1966 btrfs_set_root_flags(&root->root_item, root_flags);
1967 out_drop_sem:
1968 up_write(&fs_info->subvol_sem);
1969 out_drop_write:
1970 mnt_drop_write_file(file);
1971 out:
1972 return ret;
1973 }
1974
1975 static noinline int key_in_sk(struct btrfs_key *key,
1976 struct btrfs_ioctl_search_key *sk)
1977 {
1978 struct btrfs_key test;
1979 int ret;
1980
1981 test.objectid = sk->min_objectid;
1982 test.type = sk->min_type;
1983 test.offset = sk->min_offset;
1984
1985 ret = btrfs_comp_cpu_keys(key, &test);
1986 if (ret < 0)
1987 return 0;
1988
1989 test.objectid = sk->max_objectid;
1990 test.type = sk->max_type;
1991 test.offset = sk->max_offset;
1992
1993 ret = btrfs_comp_cpu_keys(key, &test);
1994 if (ret > 0)
1995 return 0;
1996 return 1;
1997 }
1998
1999 static noinline int copy_to_sk(struct btrfs_path *path,
2000 struct btrfs_key *key,
2001 struct btrfs_ioctl_search_key *sk,
2002 size_t *buf_size,
2003 char __user *ubuf,
2004 unsigned long *sk_offset,
2005 int *num_found)
2006 {
2007 u64 found_transid;
2008 struct extent_buffer *leaf;
2009 struct btrfs_ioctl_search_header sh;
2010 struct btrfs_key test;
2011 unsigned long item_off;
2012 unsigned long item_len;
2013 int nritems;
2014 int i;
2015 int slot;
2016 int ret = 0;
2017
2018 leaf = path->nodes[0];
2019 slot = path->slots[0];
2020 nritems = btrfs_header_nritems(leaf);
2021
2022 if (btrfs_header_generation(leaf) > sk->max_transid) {
2023 i = nritems;
2024 goto advance_key;
2025 }
2026 found_transid = btrfs_header_generation(leaf);
2027
2028 for (i = slot; i < nritems; i++) {
2029 item_off = btrfs_item_ptr_offset(leaf, i);
2030 item_len = btrfs_item_size_nr(leaf, i);
2031
2032 btrfs_item_key_to_cpu(leaf, key, i);
2033 if (!key_in_sk(key, sk))
2034 continue;
2035
2036 if (sizeof(sh) + item_len > *buf_size) {
2037 if (*num_found) {
2038 ret = 1;
2039 goto out;
2040 }
2041
2042 /*
2043 * return one empty item back for v1, which does not
2044 * handle -EOVERFLOW
2045 */
2046
2047 *buf_size = sizeof(sh) + item_len;
2048 item_len = 0;
2049 ret = -EOVERFLOW;
2050 }
2051
2052 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2053 ret = 1;
2054 goto out;
2055 }
2056
2057 sh.objectid = key->objectid;
2058 sh.offset = key->offset;
2059 sh.type = key->type;
2060 sh.len = item_len;
2061 sh.transid = found_transid;
2062
2063 /* copy search result header */
2064 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2065 ret = -EFAULT;
2066 goto out;
2067 }
2068
2069 *sk_offset += sizeof(sh);
2070
2071 if (item_len) {
2072 char __user *up = ubuf + *sk_offset;
2073 /* copy the item */
2074 if (read_extent_buffer_to_user(leaf, up,
2075 item_off, item_len)) {
2076 ret = -EFAULT;
2077 goto out;
2078 }
2079
2080 *sk_offset += item_len;
2081 }
2082 (*num_found)++;
2083
2084 if (ret) /* -EOVERFLOW from above */
2085 goto out;
2086
2087 if (*num_found >= sk->nr_items) {
2088 ret = 1;
2089 goto out;
2090 }
2091 }
2092 advance_key:
2093 ret = 0;
2094 test.objectid = sk->max_objectid;
2095 test.type = sk->max_type;
2096 test.offset = sk->max_offset;
2097 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2098 ret = 1;
2099 else if (key->offset < (u64)-1)
2100 key->offset++;
2101 else if (key->type < (u8)-1) {
2102 key->offset = 0;
2103 key->type++;
2104 } else if (key->objectid < (u64)-1) {
2105 key->offset = 0;
2106 key->type = 0;
2107 key->objectid++;
2108 } else
2109 ret = 1;
2110 out:
2111 /*
2112 * 0: all items from this leaf copied, continue with next
2113 * 1: * more items can be copied, but unused buffer is too small
2114 * * all items were found
2115 * Either way, it will stops the loop which iterates to the next
2116 * leaf
2117 * -EOVERFLOW: item was to large for buffer
2118 * -EFAULT: could not copy extent buffer back to userspace
2119 */
2120 return ret;
2121 }
2122
2123 static noinline int search_ioctl(struct inode *inode,
2124 struct btrfs_ioctl_search_key *sk,
2125 size_t *buf_size,
2126 char __user *ubuf)
2127 {
2128 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2129 struct btrfs_root *root;
2130 struct btrfs_key key;
2131 struct btrfs_path *path;
2132 int ret;
2133 int num_found = 0;
2134 unsigned long sk_offset = 0;
2135
2136 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2137 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2138 return -EOVERFLOW;
2139 }
2140
2141 path = btrfs_alloc_path();
2142 if (!path)
2143 return -ENOMEM;
2144
2145 if (sk->tree_id == 0) {
2146 /* search the root of the inode that was passed */
2147 root = BTRFS_I(inode)->root;
2148 } else {
2149 key.objectid = sk->tree_id;
2150 key.type = BTRFS_ROOT_ITEM_KEY;
2151 key.offset = (u64)-1;
2152 root = btrfs_read_fs_root_no_name(info, &key);
2153 if (IS_ERR(root)) {
2154 btrfs_free_path(path);
2155 return PTR_ERR(root);
2156 }
2157 }
2158
2159 key.objectid = sk->min_objectid;
2160 key.type = sk->min_type;
2161 key.offset = sk->min_offset;
2162
2163 while (1) {
2164 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2165 if (ret != 0) {
2166 if (ret > 0)
2167 ret = 0;
2168 goto err;
2169 }
2170 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2171 &sk_offset, &num_found);
2172 btrfs_release_path(path);
2173 if (ret)
2174 break;
2175
2176 }
2177 if (ret > 0)
2178 ret = 0;
2179 err:
2180 sk->nr_items = num_found;
2181 btrfs_free_path(path);
2182 return ret;
2183 }
2184
2185 static noinline int btrfs_ioctl_tree_search(struct file *file,
2186 void __user *argp)
2187 {
2188 struct btrfs_ioctl_search_args __user *uargs;
2189 struct btrfs_ioctl_search_key sk;
2190 struct inode *inode;
2191 int ret;
2192 size_t buf_size;
2193
2194 if (!capable(CAP_SYS_ADMIN))
2195 return -EPERM;
2196
2197 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2198
2199 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2200 return -EFAULT;
2201
2202 buf_size = sizeof(uargs->buf);
2203
2204 inode = file_inode(file);
2205 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2206
2207 /*
2208 * In the origin implementation an overflow is handled by returning a
2209 * search header with a len of zero, so reset ret.
2210 */
2211 if (ret == -EOVERFLOW)
2212 ret = 0;
2213
2214 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2215 ret = -EFAULT;
2216 return ret;
2217 }
2218
2219 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2220 void __user *argp)
2221 {
2222 struct btrfs_ioctl_search_args_v2 __user *uarg;
2223 struct btrfs_ioctl_search_args_v2 args;
2224 struct inode *inode;
2225 int ret;
2226 size_t buf_size;
2227 const size_t buf_limit = SZ_16M;
2228
2229 if (!capable(CAP_SYS_ADMIN))
2230 return -EPERM;
2231
2232 /* copy search header and buffer size */
2233 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2234 if (copy_from_user(&args, uarg, sizeof(args)))
2235 return -EFAULT;
2236
2237 buf_size = args.buf_size;
2238
2239 /* limit result size to 16MB */
2240 if (buf_size > buf_limit)
2241 buf_size = buf_limit;
2242
2243 inode = file_inode(file);
2244 ret = search_ioctl(inode, &args.key, &buf_size,
2245 (char __user *)(&uarg->buf[0]));
2246 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2247 ret = -EFAULT;
2248 else if (ret == -EOVERFLOW &&
2249 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2250 ret = -EFAULT;
2251
2252 return ret;
2253 }
2254
2255 /*
2256 * Search INODE_REFs to identify path name of 'dirid' directory
2257 * in a 'tree_id' tree. and sets path name to 'name'.
2258 */
2259 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2260 u64 tree_id, u64 dirid, char *name)
2261 {
2262 struct btrfs_root *root;
2263 struct btrfs_key key;
2264 char *ptr;
2265 int ret = -1;
2266 int slot;
2267 int len;
2268 int total_len = 0;
2269 struct btrfs_inode_ref *iref;
2270 struct extent_buffer *l;
2271 struct btrfs_path *path;
2272
2273 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2274 name[0]='\0';
2275 return 0;
2276 }
2277
2278 path = btrfs_alloc_path();
2279 if (!path)
2280 return -ENOMEM;
2281
2282 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2283
2284 key.objectid = tree_id;
2285 key.type = BTRFS_ROOT_ITEM_KEY;
2286 key.offset = (u64)-1;
2287 root = btrfs_read_fs_root_no_name(info, &key);
2288 if (IS_ERR(root)) {
2289 ret = PTR_ERR(root);
2290 goto out;
2291 }
2292
2293 key.objectid = dirid;
2294 key.type = BTRFS_INODE_REF_KEY;
2295 key.offset = (u64)-1;
2296
2297 while (1) {
2298 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2299 if (ret < 0)
2300 goto out;
2301 else if (ret > 0) {
2302 ret = btrfs_previous_item(root, path, dirid,
2303 BTRFS_INODE_REF_KEY);
2304 if (ret < 0)
2305 goto out;
2306 else if (ret > 0) {
2307 ret = -ENOENT;
2308 goto out;
2309 }
2310 }
2311
2312 l = path->nodes[0];
2313 slot = path->slots[0];
2314 btrfs_item_key_to_cpu(l, &key, slot);
2315
2316 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2317 len = btrfs_inode_ref_name_len(l, iref);
2318 ptr -= len + 1;
2319 total_len += len + 1;
2320 if (ptr < name) {
2321 ret = -ENAMETOOLONG;
2322 goto out;
2323 }
2324
2325 *(ptr + len) = '/';
2326 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2327
2328 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2329 break;
2330
2331 btrfs_release_path(path);
2332 key.objectid = key.offset;
2333 key.offset = (u64)-1;
2334 dirid = key.objectid;
2335 }
2336 memmove(name, ptr, total_len);
2337 name[total_len] = '\0';
2338 ret = 0;
2339 out:
2340 btrfs_free_path(path);
2341 return ret;
2342 }
2343
2344 static int btrfs_search_path_in_tree_user(struct inode *inode,
2345 struct btrfs_ioctl_ino_lookup_user_args *args)
2346 {
2347 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2348 struct super_block *sb = inode->i_sb;
2349 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2350 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2351 u64 dirid = args->dirid;
2352 unsigned long item_off;
2353 unsigned long item_len;
2354 struct btrfs_inode_ref *iref;
2355 struct btrfs_root_ref *rref;
2356 struct btrfs_root *root;
2357 struct btrfs_path *path;
2358 struct btrfs_key key, key2;
2359 struct extent_buffer *leaf;
2360 struct inode *temp_inode;
2361 char *ptr;
2362 int slot;
2363 int len;
2364 int total_len = 0;
2365 int ret;
2366
2367 path = btrfs_alloc_path();
2368 if (!path)
2369 return -ENOMEM;
2370
2371 /*
2372 * If the bottom subvolume does not exist directly under upper_limit,
2373 * construct the path in from the bottom up.
2374 */
2375 if (dirid != upper_limit.objectid) {
2376 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2377
2378 key.objectid = treeid;
2379 key.type = BTRFS_ROOT_ITEM_KEY;
2380 key.offset = (u64)-1;
2381 root = btrfs_read_fs_root_no_name(fs_info, &key);
2382 if (IS_ERR(root)) {
2383 ret = PTR_ERR(root);
2384 goto out;
2385 }
2386
2387 key.objectid = dirid;
2388 key.type = BTRFS_INODE_REF_KEY;
2389 key.offset = (u64)-1;
2390 while (1) {
2391 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2392 if (ret < 0) {
2393 goto out;
2394 } else if (ret > 0) {
2395 ret = btrfs_previous_item(root, path, dirid,
2396 BTRFS_INODE_REF_KEY);
2397 if (ret < 0) {
2398 goto out;
2399 } else if (ret > 0) {
2400 ret = -ENOENT;
2401 goto out;
2402 }
2403 }
2404
2405 leaf = path->nodes[0];
2406 slot = path->slots[0];
2407 btrfs_item_key_to_cpu(leaf, &key, slot);
2408
2409 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2410 len = btrfs_inode_ref_name_len(leaf, iref);
2411 ptr -= len + 1;
2412 total_len += len + 1;
2413 if (ptr < args->path) {
2414 ret = -ENAMETOOLONG;
2415 goto out;
2416 }
2417
2418 *(ptr + len) = '/';
2419 read_extent_buffer(leaf, ptr,
2420 (unsigned long)(iref + 1), len);
2421
2422 /* Check the read+exec permission of this directory */
2423 ret = btrfs_previous_item(root, path, dirid,
2424 BTRFS_INODE_ITEM_KEY);
2425 if (ret < 0) {
2426 goto out;
2427 } else if (ret > 0) {
2428 ret = -ENOENT;
2429 goto out;
2430 }
2431
2432 leaf = path->nodes[0];
2433 slot = path->slots[0];
2434 btrfs_item_key_to_cpu(leaf, &key2, slot);
2435 if (key2.objectid != dirid) {
2436 ret = -ENOENT;
2437 goto out;
2438 }
2439
2440 temp_inode = btrfs_iget(sb, &key2, root, NULL);
2441 if (IS_ERR(temp_inode)) {
2442 ret = PTR_ERR(temp_inode);
2443 goto out;
2444 }
2445 ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC);
2446 iput(temp_inode);
2447 if (ret) {
2448 ret = -EACCES;
2449 goto out;
2450 }
2451
2452 if (key.offset == upper_limit.objectid)
2453 break;
2454 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2455 ret = -EACCES;
2456 goto out;
2457 }
2458
2459 btrfs_release_path(path);
2460 key.objectid = key.offset;
2461 key.offset = (u64)-1;
2462 dirid = key.objectid;
2463 }
2464
2465 memmove(args->path, ptr, total_len);
2466 args->path[total_len] = '\0';
2467 btrfs_release_path(path);
2468 }
2469
2470 /* Get the bottom subvolume's name from ROOT_REF */
2471 root = fs_info->tree_root;
2472 key.objectid = treeid;
2473 key.type = BTRFS_ROOT_REF_KEY;
2474 key.offset = args->treeid;
2475 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2476 if (ret < 0) {
2477 goto out;
2478 } else if (ret > 0) {
2479 ret = -ENOENT;
2480 goto out;
2481 }
2482
2483 leaf = path->nodes[0];
2484 slot = path->slots[0];
2485 btrfs_item_key_to_cpu(leaf, &key, slot);
2486
2487 item_off = btrfs_item_ptr_offset(leaf, slot);
2488 item_len = btrfs_item_size_nr(leaf, slot);
2489 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2490 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2491 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2492 ret = -EINVAL;
2493 goto out;
2494 }
2495
2496 /* Copy subvolume's name */
2497 item_off += sizeof(struct btrfs_root_ref);
2498 item_len -= sizeof(struct btrfs_root_ref);
2499 read_extent_buffer(leaf, args->name, item_off, item_len);
2500 args->name[item_len] = 0;
2501
2502 out:
2503 btrfs_free_path(path);
2504 return ret;
2505 }
2506
2507 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2508 void __user *argp)
2509 {
2510 struct btrfs_ioctl_ino_lookup_args *args;
2511 struct inode *inode;
2512 int ret = 0;
2513
2514 args = memdup_user(argp, sizeof(*args));
2515 if (IS_ERR(args))
2516 return PTR_ERR(args);
2517
2518 inode = file_inode(file);
2519
2520 /*
2521 * Unprivileged query to obtain the containing subvolume root id. The
2522 * path is reset so it's consistent with btrfs_search_path_in_tree.
2523 */
2524 if (args->treeid == 0)
2525 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2526
2527 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2528 args->name[0] = 0;
2529 goto out;
2530 }
2531
2532 if (!capable(CAP_SYS_ADMIN)) {
2533 ret = -EPERM;
2534 goto out;
2535 }
2536
2537 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2538 args->treeid, args->objectid,
2539 args->name);
2540
2541 out:
2542 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2543 ret = -EFAULT;
2544
2545 kfree(args);
2546 return ret;
2547 }
2548
2549 /*
2550 * Version of ino_lookup ioctl (unprivileged)
2551 *
2552 * The main differences from ino_lookup ioctl are:
2553 *
2554 * 1. Read + Exec permission will be checked using inode_permission() during
2555 * path construction. -EACCES will be returned in case of failure.
2556 * 2. Path construction will be stopped at the inode number which corresponds
2557 * to the fd with which this ioctl is called. If constructed path does not
2558 * exist under fd's inode, -EACCES will be returned.
2559 * 3. The name of bottom subvolume is also searched and filled.
2560 */
2561 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2562 {
2563 struct btrfs_ioctl_ino_lookup_user_args *args;
2564 struct inode *inode;
2565 int ret;
2566
2567 args = memdup_user(argp, sizeof(*args));
2568 if (IS_ERR(args))
2569 return PTR_ERR(args);
2570
2571 inode = file_inode(file);
2572
2573 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2574 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2575 /*
2576 * The subvolume does not exist under fd with which this is
2577 * called
2578 */
2579 kfree(args);
2580 return -EACCES;
2581 }
2582
2583 ret = btrfs_search_path_in_tree_user(inode, args);
2584
2585 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2586 ret = -EFAULT;
2587
2588 kfree(args);
2589 return ret;
2590 }
2591
2592 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2593 static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2594 {
2595 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2596 struct btrfs_fs_info *fs_info;
2597 struct btrfs_root *root;
2598 struct btrfs_path *path;
2599 struct btrfs_key key;
2600 struct btrfs_root_item *root_item;
2601 struct btrfs_root_ref *rref;
2602 struct extent_buffer *leaf;
2603 unsigned long item_off;
2604 unsigned long item_len;
2605 struct inode *inode;
2606 int slot;
2607 int ret = 0;
2608
2609 path = btrfs_alloc_path();
2610 if (!path)
2611 return -ENOMEM;
2612
2613 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2614 if (!subvol_info) {
2615 btrfs_free_path(path);
2616 return -ENOMEM;
2617 }
2618
2619 inode = file_inode(file);
2620 fs_info = BTRFS_I(inode)->root->fs_info;
2621
2622 /* Get root_item of inode's subvolume */
2623 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2624 key.type = BTRFS_ROOT_ITEM_KEY;
2625 key.offset = (u64)-1;
2626 root = btrfs_read_fs_root_no_name(fs_info, &key);
2627 if (IS_ERR(root)) {
2628 ret = PTR_ERR(root);
2629 goto out;
2630 }
2631 root_item = &root->root_item;
2632
2633 subvol_info->treeid = key.objectid;
2634
2635 subvol_info->generation = btrfs_root_generation(root_item);
2636 subvol_info->flags = btrfs_root_flags(root_item);
2637
2638 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2639 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2640 BTRFS_UUID_SIZE);
2641 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2642 BTRFS_UUID_SIZE);
2643
2644 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2645 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2646 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2647
2648 subvol_info->otransid = btrfs_root_otransid(root_item);
2649 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2650 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2651
2652 subvol_info->stransid = btrfs_root_stransid(root_item);
2653 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2654 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2655
2656 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2657 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2658 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2659
2660 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2661 /* Search root tree for ROOT_BACKREF of this subvolume */
2662 root = fs_info->tree_root;
2663
2664 key.type = BTRFS_ROOT_BACKREF_KEY;
2665 key.offset = 0;
2666 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2667 if (ret < 0) {
2668 goto out;
2669 } else if (path->slots[0] >=
2670 btrfs_header_nritems(path->nodes[0])) {
2671 ret = btrfs_next_leaf(root, path);
2672 if (ret < 0) {
2673 goto out;
2674 } else if (ret > 0) {
2675 ret = -EUCLEAN;
2676 goto out;
2677 }
2678 }
2679
2680 leaf = path->nodes[0];
2681 slot = path->slots[0];
2682 btrfs_item_key_to_cpu(leaf, &key, slot);
2683 if (key.objectid == subvol_info->treeid &&
2684 key.type == BTRFS_ROOT_BACKREF_KEY) {
2685 subvol_info->parent_id = key.offset;
2686
2687 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2688 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2689
2690 item_off = btrfs_item_ptr_offset(leaf, slot)
2691 + sizeof(struct btrfs_root_ref);
2692 item_len = btrfs_item_size_nr(leaf, slot)
2693 - sizeof(struct btrfs_root_ref);
2694 read_extent_buffer(leaf, subvol_info->name,
2695 item_off, item_len);
2696 } else {
2697 ret = -ENOENT;
2698 goto out;
2699 }
2700 }
2701
2702 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2703 ret = -EFAULT;
2704
2705 out:
2706 btrfs_free_path(path);
2707 kzfree(subvol_info);
2708 return ret;
2709 }
2710
2711 /*
2712 * Return ROOT_REF information of the subvolume containing this inode
2713 * except the subvolume name.
2714 */
2715 static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2716 {
2717 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2718 struct btrfs_root_ref *rref;
2719 struct btrfs_root *root;
2720 struct btrfs_path *path;
2721 struct btrfs_key key;
2722 struct extent_buffer *leaf;
2723 struct inode *inode;
2724 u64 objectid;
2725 int slot;
2726 int ret;
2727 u8 found;
2728
2729 path = btrfs_alloc_path();
2730 if (!path)
2731 return -ENOMEM;
2732
2733 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2734 if (IS_ERR(rootrefs)) {
2735 btrfs_free_path(path);
2736 return PTR_ERR(rootrefs);
2737 }
2738
2739 inode = file_inode(file);
2740 root = BTRFS_I(inode)->root->fs_info->tree_root;
2741 objectid = BTRFS_I(inode)->root->root_key.objectid;
2742
2743 key.objectid = objectid;
2744 key.type = BTRFS_ROOT_REF_KEY;
2745 key.offset = rootrefs->min_treeid;
2746 found = 0;
2747
2748 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2749 if (ret < 0) {
2750 goto out;
2751 } else if (path->slots[0] >=
2752 btrfs_header_nritems(path->nodes[0])) {
2753 ret = btrfs_next_leaf(root, path);
2754 if (ret < 0) {
2755 goto out;
2756 } else if (ret > 0) {
2757 ret = -EUCLEAN;
2758 goto out;
2759 }
2760 }
2761 while (1) {
2762 leaf = path->nodes[0];
2763 slot = path->slots[0];
2764
2765 btrfs_item_key_to_cpu(leaf, &key, slot);
2766 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2767 ret = 0;
2768 goto out;
2769 }
2770
2771 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2772 ret = -EOVERFLOW;
2773 goto out;
2774 }
2775
2776 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2777 rootrefs->rootref[found].treeid = key.offset;
2778 rootrefs->rootref[found].dirid =
2779 btrfs_root_ref_dirid(leaf, rref);
2780 found++;
2781
2782 ret = btrfs_next_item(root, path);
2783 if (ret < 0) {
2784 goto out;
2785 } else if (ret > 0) {
2786 ret = -EUCLEAN;
2787 goto out;
2788 }
2789 }
2790
2791 out:
2792 if (!ret || ret == -EOVERFLOW) {
2793 rootrefs->num_items = found;
2794 /* update min_treeid for next search */
2795 if (found)
2796 rootrefs->min_treeid =
2797 rootrefs->rootref[found - 1].treeid + 1;
2798 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2799 ret = -EFAULT;
2800 }
2801
2802 kfree(rootrefs);
2803 btrfs_free_path(path);
2804
2805 return ret;
2806 }
2807
2808 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2809 void __user *arg)
2810 {
2811 struct dentry *parent = file->f_path.dentry;
2812 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2813 struct dentry *dentry;
2814 struct inode *dir = d_inode(parent);
2815 struct inode *inode;
2816 struct btrfs_root *root = BTRFS_I(dir)->root;
2817 struct btrfs_root *dest = NULL;
2818 struct btrfs_ioctl_vol_args *vol_args;
2819 int namelen;
2820 int err = 0;
2821
2822 if (!S_ISDIR(dir->i_mode))
2823 return -ENOTDIR;
2824
2825 vol_args = memdup_user(arg, sizeof(*vol_args));
2826 if (IS_ERR(vol_args))
2827 return PTR_ERR(vol_args);
2828
2829 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2830 namelen = strlen(vol_args->name);
2831 if (strchr(vol_args->name, '/') ||
2832 strncmp(vol_args->name, "..", namelen) == 0) {
2833 err = -EINVAL;
2834 goto out;
2835 }
2836
2837 err = mnt_want_write_file(file);
2838 if (err)
2839 goto out;
2840
2841
2842 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2843 if (err == -EINTR)
2844 goto out_drop_write;
2845 dentry = lookup_one_len(vol_args->name, parent, namelen);
2846 if (IS_ERR(dentry)) {
2847 err = PTR_ERR(dentry);
2848 goto out_unlock_dir;
2849 }
2850
2851 if (d_really_is_negative(dentry)) {
2852 err = -ENOENT;
2853 goto out_dput;
2854 }
2855
2856 inode = d_inode(dentry);
2857 dest = BTRFS_I(inode)->root;
2858 if (!capable(CAP_SYS_ADMIN)) {
2859 /*
2860 * Regular user. Only allow this with a special mount
2861 * option, when the user has write+exec access to the
2862 * subvol root, and when rmdir(2) would have been
2863 * allowed.
2864 *
2865 * Note that this is _not_ check that the subvol is
2866 * empty or doesn't contain data that we wouldn't
2867 * otherwise be able to delete.
2868 *
2869 * Users who want to delete empty subvols should try
2870 * rmdir(2).
2871 */
2872 err = -EPERM;
2873 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2874 goto out_dput;
2875
2876 /*
2877 * Do not allow deletion if the parent dir is the same
2878 * as the dir to be deleted. That means the ioctl
2879 * must be called on the dentry referencing the root
2880 * of the subvol, not a random directory contained
2881 * within it.
2882 */
2883 err = -EINVAL;
2884 if (root == dest)
2885 goto out_dput;
2886
2887 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2888 if (err)
2889 goto out_dput;
2890 }
2891
2892 /* check if subvolume may be deleted by a user */
2893 err = btrfs_may_delete(dir, dentry, 1);
2894 if (err)
2895 goto out_dput;
2896
2897 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2898 err = -EINVAL;
2899 goto out_dput;
2900 }
2901
2902 inode_lock(inode);
2903 err = btrfs_delete_subvolume(dir, dentry);
2904 inode_unlock(inode);
2905 if (!err)
2906 d_delete(dentry);
2907
2908 out_dput:
2909 dput(dentry);
2910 out_unlock_dir:
2911 inode_unlock(dir);
2912 out_drop_write:
2913 mnt_drop_write_file(file);
2914 out:
2915 kfree(vol_args);
2916 return err;
2917 }
2918
2919 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2920 {
2921 struct inode *inode = file_inode(file);
2922 struct btrfs_root *root = BTRFS_I(inode)->root;
2923 struct btrfs_ioctl_defrag_range_args *range;
2924 int ret;
2925
2926 ret = mnt_want_write_file(file);
2927 if (ret)
2928 return ret;
2929
2930 if (btrfs_root_readonly(root)) {
2931 ret = -EROFS;
2932 goto out;
2933 }
2934
2935 switch (inode->i_mode & S_IFMT) {
2936 case S_IFDIR:
2937 if (!capable(CAP_SYS_ADMIN)) {
2938 ret = -EPERM;
2939 goto out;
2940 }
2941 ret = btrfs_defrag_root(root);
2942 break;
2943 case S_IFREG:
2944 if (!(file->f_mode & FMODE_WRITE)) {
2945 ret = -EINVAL;
2946 goto out;
2947 }
2948
2949 range = kzalloc(sizeof(*range), GFP_KERNEL);
2950 if (!range) {
2951 ret = -ENOMEM;
2952 goto out;
2953 }
2954
2955 if (argp) {
2956 if (copy_from_user(range, argp,
2957 sizeof(*range))) {
2958 ret = -EFAULT;
2959 kfree(range);
2960 goto out;
2961 }
2962 /* compression requires us to start the IO */
2963 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2964 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2965 range->extent_thresh = (u32)-1;
2966 }
2967 } else {
2968 /* the rest are all set to zero by kzalloc */
2969 range->len = (u64)-1;
2970 }
2971 ret = btrfs_defrag_file(file_inode(file), file,
2972 range, BTRFS_OLDEST_GENERATION, 0);
2973 if (ret > 0)
2974 ret = 0;
2975 kfree(range);
2976 break;
2977 default:
2978 ret = -EINVAL;
2979 }
2980 out:
2981 mnt_drop_write_file(file);
2982 return ret;
2983 }
2984
2985 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2986 {
2987 struct btrfs_ioctl_vol_args *vol_args;
2988 int ret;
2989
2990 if (!capable(CAP_SYS_ADMIN))
2991 return -EPERM;
2992
2993 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
2994 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2995
2996 vol_args = memdup_user(arg, sizeof(*vol_args));
2997 if (IS_ERR(vol_args)) {
2998 ret = PTR_ERR(vol_args);
2999 goto out;
3000 }
3001
3002 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3003 ret = btrfs_init_new_device(fs_info, vol_args->name);
3004
3005 if (!ret)
3006 btrfs_info(fs_info, "disk added %s", vol_args->name);
3007
3008 kfree(vol_args);
3009 out:
3010 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3011 return ret;
3012 }
3013
3014 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
3015 {
3016 struct inode *inode = file_inode(file);
3017 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3018 struct btrfs_ioctl_vol_args_v2 *vol_args;
3019 int ret;
3020
3021 if (!capable(CAP_SYS_ADMIN))
3022 return -EPERM;
3023
3024 ret = mnt_want_write_file(file);
3025 if (ret)
3026 return ret;
3027
3028 vol_args = memdup_user(arg, sizeof(*vol_args));
3029 if (IS_ERR(vol_args)) {
3030 ret = PTR_ERR(vol_args);
3031 goto err_drop;
3032 }
3033
3034 /* Check for compatibility reject unknown flags */
3035 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) {
3036 ret = -EOPNOTSUPP;
3037 goto out;
3038 }
3039
3040 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3041 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3042 goto out;
3043 }
3044
3045 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
3046 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
3047 } else {
3048 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3049 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3050 }
3051 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3052
3053 if (!ret) {
3054 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3055 btrfs_info(fs_info, "device deleted: id %llu",
3056 vol_args->devid);
3057 else
3058 btrfs_info(fs_info, "device deleted: %s",
3059 vol_args->name);
3060 }
3061 out:
3062 kfree(vol_args);
3063 err_drop:
3064 mnt_drop_write_file(file);
3065 return ret;
3066 }
3067
3068 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3069 {
3070 struct inode *inode = file_inode(file);
3071 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3072 struct btrfs_ioctl_vol_args *vol_args;
3073 int ret;
3074
3075 if (!capable(CAP_SYS_ADMIN))
3076 return -EPERM;
3077
3078 ret = mnt_want_write_file(file);
3079 if (ret)
3080 return ret;
3081
3082 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3083 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3084 goto out_drop_write;
3085 }
3086
3087 vol_args = memdup_user(arg, sizeof(*vol_args));
3088 if (IS_ERR(vol_args)) {
3089 ret = PTR_ERR(vol_args);
3090 goto out;
3091 }
3092
3093 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3094 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3095
3096 if (!ret)
3097 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3098 kfree(vol_args);
3099 out:
3100 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3101 out_drop_write:
3102 mnt_drop_write_file(file);
3103
3104 return ret;
3105 }
3106
3107 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3108 void __user *arg)
3109 {
3110 struct btrfs_ioctl_fs_info_args *fi_args;
3111 struct btrfs_device *device;
3112 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3113 int ret = 0;
3114
3115 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
3116 if (!fi_args)
3117 return -ENOMEM;
3118
3119 rcu_read_lock();
3120 fi_args->num_devices = fs_devices->num_devices;
3121
3122 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3123 if (device->devid > fi_args->max_id)
3124 fi_args->max_id = device->devid;
3125 }
3126 rcu_read_unlock();
3127
3128 memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid));
3129 fi_args->nodesize = fs_info->nodesize;
3130 fi_args->sectorsize = fs_info->sectorsize;
3131 fi_args->clone_alignment = fs_info->sectorsize;
3132
3133 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3134 ret = -EFAULT;
3135
3136 kfree(fi_args);
3137 return ret;
3138 }
3139
3140 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3141 void __user *arg)
3142 {
3143 struct btrfs_ioctl_dev_info_args *di_args;
3144 struct btrfs_device *dev;
3145 int ret = 0;
3146 char *s_uuid = NULL;
3147
3148 di_args = memdup_user(arg, sizeof(*di_args));
3149 if (IS_ERR(di_args))
3150 return PTR_ERR(di_args);
3151
3152 if (!btrfs_is_empty_uuid(di_args->uuid))
3153 s_uuid = di_args->uuid;
3154
3155 rcu_read_lock();
3156 dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL);
3157
3158 if (!dev) {
3159 ret = -ENODEV;
3160 goto out;
3161 }
3162
3163 di_args->devid = dev->devid;
3164 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3165 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3166 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3167 if (dev->name) {
3168 struct rcu_string *name;
3169
3170 name = rcu_dereference(dev->name);
3171 strncpy(di_args->path, name->str, sizeof(di_args->path) - 1);
3172 di_args->path[sizeof(di_args->path) - 1] = 0;
3173 } else {
3174 di_args->path[0] = '\0';
3175 }
3176
3177 out:
3178 rcu_read_unlock();
3179 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3180 ret = -EFAULT;
3181
3182 kfree(di_args);
3183 return ret;
3184 }
3185
3186 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
3187 {
3188 struct page *page;
3189
3190 page = grab_cache_page(inode->i_mapping, index);
3191 if (!page)
3192 return ERR_PTR(-ENOMEM);
3193
3194 if (!PageUptodate(page)) {
3195 int ret;
3196
3197 ret = btrfs_readpage(NULL, page);
3198 if (ret)
3199 return ERR_PTR(ret);
3200 lock_page(page);
3201 if (!PageUptodate(page)) {
3202 unlock_page(page);
3203 put_page(page);
3204 return ERR_PTR(-EIO);
3205 }
3206 if (page->mapping != inode->i_mapping) {
3207 unlock_page(page);
3208 put_page(page);
3209 return ERR_PTR(-EAGAIN);
3210 }
3211 }
3212
3213 return page;
3214 }
3215
3216 static int gather_extent_pages(struct inode *inode, struct page **pages,
3217 int num_pages, u64 off)
3218 {
3219 int i;
3220 pgoff_t index = off >> PAGE_SHIFT;
3221
3222 for (i = 0; i < num_pages; i++) {
3223 again:
3224 pages[i] = extent_same_get_page(inode, index + i);
3225 if (IS_ERR(pages[i])) {
3226 int err = PTR_ERR(pages[i]);
3227
3228 if (err == -EAGAIN)
3229 goto again;
3230 pages[i] = NULL;
3231 return err;
3232 }
3233 }
3234 return 0;
3235 }
3236
3237 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
3238 bool retry_range_locking)
3239 {
3240 /*
3241 * Do any pending delalloc/csum calculations on inode, one way or
3242 * another, and lock file content.
3243 * The locking order is:
3244 *
3245 * 1) pages
3246 * 2) range in the inode's io tree
3247 */
3248 while (1) {
3249 struct btrfs_ordered_extent *ordered;
3250 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
3251 ordered = btrfs_lookup_first_ordered_extent(inode,
3252 off + len - 1);
3253 if ((!ordered ||
3254 ordered->file_offset + ordered->len <= off ||
3255 ordered->file_offset >= off + len) &&
3256 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
3257 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
3258 if (ordered)
3259 btrfs_put_ordered_extent(ordered);
3260 break;
3261 }
3262 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
3263 if (ordered)
3264 btrfs_put_ordered_extent(ordered);
3265 if (!retry_range_locking)
3266 return -EAGAIN;
3267 btrfs_wait_ordered_range(inode, off, len);
3268 }
3269 return 0;
3270 }
3271
3272 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
3273 {
3274 inode_unlock(inode1);
3275 inode_unlock(inode2);
3276 }
3277
3278 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
3279 {
3280 if (inode1 < inode2)
3281 swap(inode1, inode2);
3282
3283 inode_lock_nested(inode1, I_MUTEX_PARENT);
3284 inode_lock_nested(inode2, I_MUTEX_CHILD);
3285 }
3286
3287 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
3288 struct inode *inode2, u64 loff2, u64 len)
3289 {
3290 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
3291 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
3292 }
3293
3294 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
3295 struct inode *inode2, u64 loff2, u64 len,
3296 bool retry_range_locking)
3297 {
3298 int ret;
3299
3300 if (inode1 < inode2) {
3301 swap(inode1, inode2);
3302 swap(loff1, loff2);
3303 }
3304 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
3305 if (ret)
3306 return ret;
3307 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
3308 if (ret)
3309 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
3310 loff1 + len - 1);
3311 return ret;
3312 }
3313
3314 struct cmp_pages {
3315 int num_pages;
3316 struct page **src_pages;
3317 struct page **dst_pages;
3318 };
3319
3320 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
3321 {
3322 int i;
3323 struct page *pg;
3324
3325 for (i = 0; i < cmp->num_pages; i++) {
3326 pg = cmp->src_pages[i];
3327 if (pg) {
3328 unlock_page(pg);
3329 put_page(pg);
3330 cmp->src_pages[i] = NULL;
3331 }
3332 pg = cmp->dst_pages[i];
3333 if (pg) {
3334 unlock_page(pg);
3335 put_page(pg);
3336 cmp->dst_pages[i] = NULL;
3337 }
3338 }
3339 }
3340
3341 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
3342 struct inode *dst, u64 dst_loff,
3343 u64 len, struct cmp_pages *cmp)
3344 {
3345 int ret;
3346 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
3347
3348 cmp->num_pages = num_pages;
3349
3350 ret = gather_extent_pages(src, cmp->src_pages, num_pages, loff);
3351 if (ret)
3352 goto out;
3353
3354 ret = gather_extent_pages(dst, cmp->dst_pages, num_pages, dst_loff);
3355
3356 out:
3357 if (ret)
3358 btrfs_cmp_data_free(cmp);
3359 return ret;
3360 }
3361
3362 static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp)
3363 {
3364 int ret = 0;
3365 int i;
3366 struct page *src_page, *dst_page;
3367 unsigned int cmp_len = PAGE_SIZE;
3368 void *addr, *dst_addr;
3369
3370 i = 0;
3371 while (len) {
3372 if (len < PAGE_SIZE)
3373 cmp_len = len;
3374
3375 BUG_ON(i >= cmp->num_pages);
3376
3377 src_page = cmp->src_pages[i];
3378 dst_page = cmp->dst_pages[i];
3379 ASSERT(PageLocked(src_page));
3380 ASSERT(PageLocked(dst_page));
3381
3382 addr = kmap_atomic(src_page);
3383 dst_addr = kmap_atomic(dst_page);
3384
3385 flush_dcache_page(src_page);
3386 flush_dcache_page(dst_page);
3387
3388 if (memcmp(addr, dst_addr, cmp_len))
3389 ret = -EBADE;
3390
3391 kunmap_atomic(addr);
3392 kunmap_atomic(dst_addr);
3393
3394 if (ret)
3395 break;
3396
3397 len -= cmp_len;
3398 i++;
3399 }
3400
3401 return ret;
3402 }
3403
3404 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3405 u64 olen)
3406 {
3407 u64 len = *plen;
3408 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3409
3410 if (off + olen > inode->i_size || off + olen < off)
3411 return -EINVAL;
3412
3413 /* if we extend to eof, continue to block boundary */
3414 if (off + len == inode->i_size)
3415 *plen = len = ALIGN(inode->i_size, bs) - off;
3416
3417 /* Check that we are block aligned - btrfs_clone() requires this */
3418 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3419 return -EINVAL;
3420
3421 return 0;
3422 }
3423
3424 static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 olen,
3425 struct inode *dst, u64 dst_loff,
3426 struct cmp_pages *cmp)
3427 {
3428 int ret;
3429 u64 len = olen;
3430 bool same_inode = (src == dst);
3431 u64 same_lock_start = 0;
3432 u64 same_lock_len = 0;
3433
3434 ret = extent_same_check_offsets(src, loff, &len, olen);
3435 if (ret)
3436 return ret;
3437
3438 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3439 if (ret)
3440 return ret;
3441
3442 if (same_inode) {
3443 /*
3444 * Single inode case wants the same checks, except we
3445 * don't want our length pushed out past i_size as
3446 * comparing that data range makes no sense.
3447 *
3448 * extent_same_check_offsets() will do this for an
3449 * unaligned length at i_size, so catch it here and
3450 * reject the request.
3451 *
3452 * This effectively means we require aligned extents
3453 * for the single-inode case, whereas the other cases
3454 * allow an unaligned length so long as it ends at
3455 * i_size.
3456 */
3457 if (len != olen)
3458 return -EINVAL;
3459
3460 /* Check for overlapping ranges */
3461 if (dst_loff + len > loff && dst_loff < loff + len)
3462 return -EINVAL;
3463
3464 same_lock_start = min_t(u64, loff, dst_loff);
3465 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3466 } else {
3467 /*
3468 * If the source and destination inodes are different, the
3469 * source's range end offset matches the source's i_size, that
3470 * i_size is not a multiple of the sector size, and the
3471 * destination range does not go past the destination's i_size,
3472 * we must round down the length to the nearest sector size
3473 * multiple. If we don't do this adjustment we end replacing
3474 * with zeroes the bytes in the range that starts at the
3475 * deduplication range's end offset and ends at the next sector
3476 * size multiple.
3477 */
3478 if (loff + olen == i_size_read(src) &&
3479 dst_loff + len < i_size_read(dst)) {
3480 const u64 sz = BTRFS_I(src)->root->fs_info->sectorsize;
3481
3482 len = round_down(i_size_read(src), sz) - loff;
3483 olen = len;
3484 }
3485 }
3486
3487 again:
3488 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, cmp);
3489 if (ret)
3490 return ret;
3491
3492 if (same_inode)
3493 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3494 false);
3495 else
3496 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3497 false);
3498 /*
3499 * If one of the inodes has dirty pages in the respective range or
3500 * ordered extents, we need to flush dellaloc and wait for all ordered
3501 * extents in the range. We must unlock the pages and the ranges in the
3502 * io trees to avoid deadlocks when flushing delalloc (requires locking
3503 * pages) and when waiting for ordered extents to complete (they require
3504 * range locking).
3505 */
3506 if (ret == -EAGAIN) {
3507 /*
3508 * Ranges in the io trees already unlocked. Now unlock all
3509 * pages before waiting for all IO to complete.
3510 */
3511 btrfs_cmp_data_free(cmp);
3512 if (same_inode) {
3513 btrfs_wait_ordered_range(src, same_lock_start,
3514 same_lock_len);
3515 } else {
3516 btrfs_wait_ordered_range(src, loff, len);
3517 btrfs_wait_ordered_range(dst, dst_loff, len);
3518 }
3519 goto again;
3520 }
3521 ASSERT(ret == 0);
3522 if (WARN_ON(ret)) {
3523 /* ranges in the io trees already unlocked */
3524 btrfs_cmp_data_free(cmp);
3525 return ret;
3526 }
3527
3528 /* pass original length for comparison so we stay within i_size */
3529 ret = btrfs_cmp_data(olen, cmp);
3530 if (ret == 0)
3531 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3532
3533 if (same_inode)
3534 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3535 same_lock_start + same_lock_len - 1);
3536 else
3537 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3538
3539 btrfs_cmp_data_free(cmp);
3540
3541 return ret;
3542 }
3543
3544 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3545
3546 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3547 struct inode *dst, u64 dst_loff)
3548 {
3549 int ret;
3550 struct cmp_pages cmp;
3551 int num_pages = PAGE_ALIGN(BTRFS_MAX_DEDUPE_LEN) >> PAGE_SHIFT;
3552 bool same_inode = (src == dst);
3553 u64 i, tail_len, chunk_count;
3554
3555 if (olen == 0)
3556 return 0;
3557
3558 if (same_inode)
3559 inode_lock(src);
3560 else
3561 btrfs_double_inode_lock(src, dst);
3562
3563 /* don't make the dst file partly checksummed */
3564 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3565 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3566 ret = -EINVAL;
3567 goto out_unlock;
3568 }
3569
3570 tail_len = olen % BTRFS_MAX_DEDUPE_LEN;
3571 chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN);
3572 if (chunk_count == 0)
3573 num_pages = PAGE_ALIGN(tail_len) >> PAGE_SHIFT;
3574
3575 /*
3576 * If deduping ranges in the same inode, locking rules make it
3577 * mandatory to always lock pages in ascending order to avoid deadlocks
3578 * with concurrent tasks (such as starting writeback/delalloc).
3579 */
3580 if (same_inode && dst_loff < loff)
3581 swap(loff, dst_loff);
3582
3583 /*
3584 * We must gather up all the pages before we initiate our extent
3585 * locking. We use an array for the page pointers. Size of the array is
3586 * bounded by len, which is in turn bounded by BTRFS_MAX_DEDUPE_LEN.
3587 */
3588 cmp.src_pages = kvmalloc_array(num_pages, sizeof(struct page *),
3589 GFP_KERNEL | __GFP_ZERO);
3590 cmp.dst_pages = kvmalloc_array(num_pages, sizeof(struct page *),
3591 GFP_KERNEL | __GFP_ZERO);
3592 if (!cmp.src_pages || !cmp.dst_pages) {
3593 ret = -ENOMEM;
3594 goto out_free;
3595 }
3596
3597 for (i = 0; i < chunk_count; i++) {
3598 ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN,
3599 dst, dst_loff, &cmp);
3600 if (ret)
3601 goto out_free;
3602
3603 loff += BTRFS_MAX_DEDUPE_LEN;
3604 dst_loff += BTRFS_MAX_DEDUPE_LEN;
3605 }
3606
3607 if (tail_len > 0)
3608 ret = btrfs_extent_same_range(src, loff, tail_len, dst,
3609 dst_loff, &cmp);
3610
3611 out_free:
3612 kvfree(cmp.src_pages);
3613 kvfree(cmp.dst_pages);
3614
3615 out_unlock:
3616 if (same_inode)
3617 inode_unlock(src);
3618 else
3619 btrfs_double_inode_unlock(src, dst);
3620
3621 return ret;
3622 }
3623
3624 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3625 struct file *dst_file, u64 dst_loff)
3626 {
3627 struct inode *src = file_inode(src_file);
3628 struct inode *dst = file_inode(dst_file);
3629 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3630 ssize_t res;
3631
3632 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
3633 /*
3634 * Btrfs does not support blocksize < page_size. As a
3635 * result, btrfs_cmp_data() won't correctly handle
3636 * this situation without an update.
3637 */
3638 return -EINVAL;
3639 }
3640
3641 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3642 if (res)
3643 return res;
3644 return olen;
3645 }
3646
3647 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3648 struct inode *inode,
3649 u64 endoff,
3650 const u64 destoff,
3651 const u64 olen,
3652 int no_time_update)
3653 {
3654 struct btrfs_root *root = BTRFS_I(inode)->root;
3655 int ret;
3656
3657 inode_inc_iversion(inode);
3658 if (!no_time_update)
3659 inode->i_mtime = inode->i_ctime = current_time(inode);
3660 /*
3661 * We round up to the block size at eof when determining which
3662 * extents to clone above, but shouldn't round up the file size.
3663 */
3664 if (endoff > destoff + olen)
3665 endoff = destoff + olen;
3666 if (endoff > inode->i_size)
3667 btrfs_i_size_write(BTRFS_I(inode), endoff);
3668
3669 ret = btrfs_update_inode(trans, root, inode);
3670 if (ret) {
3671 btrfs_abort_transaction(trans, ret);
3672 btrfs_end_transaction(trans);
3673 goto out;
3674 }
3675 ret = btrfs_end_transaction(trans);
3676 out:
3677 return ret;
3678 }
3679
3680 static void clone_update_extent_map(struct btrfs_inode *inode,
3681 const struct btrfs_trans_handle *trans,
3682 const struct btrfs_path *path,
3683 const u64 hole_offset,
3684 const u64 hole_len)
3685 {
3686 struct extent_map_tree *em_tree = &inode->extent_tree;
3687 struct extent_map *em;
3688 int ret;
3689
3690 em = alloc_extent_map();
3691 if (!em) {
3692 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3693 return;
3694 }
3695
3696 if (path) {
3697 struct btrfs_file_extent_item *fi;
3698
3699 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3700 struct btrfs_file_extent_item);
3701 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3702 em->generation = -1;
3703 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3704 BTRFS_FILE_EXTENT_INLINE)
3705 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3706 &inode->runtime_flags);
3707 } else {
3708 em->start = hole_offset;
3709 em->len = hole_len;
3710 em->ram_bytes = em->len;
3711 em->orig_start = hole_offset;
3712 em->block_start = EXTENT_MAP_HOLE;
3713 em->block_len = 0;
3714 em->orig_block_len = 0;
3715 em->compress_type = BTRFS_COMPRESS_NONE;
3716 em->generation = trans->transid;
3717 }
3718
3719 while (1) {
3720 write_lock(&em_tree->lock);
3721 ret = add_extent_mapping(em_tree, em, 1);
3722 write_unlock(&em_tree->lock);
3723 if (ret != -EEXIST) {
3724 free_extent_map(em);
3725 break;
3726 }
3727 btrfs_drop_extent_cache(inode, em->start,
3728 em->start + em->len - 1, 0);
3729 }
3730
3731 if (ret)
3732 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3733 }
3734
3735 /*
3736 * Make sure we do not end up inserting an inline extent into a file that has
3737 * already other (non-inline) extents. If a file has an inline extent it can
3738 * not have any other extents and the (single) inline extent must start at the
3739 * file offset 0. Failing to respect these rules will lead to file corruption,
3740 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3741 *
3742 * We can have extents that have been already written to disk or we can have
3743 * dirty ranges still in delalloc, in which case the extent maps and items are
3744 * created only when we run delalloc, and the delalloc ranges might fall outside
3745 * the range we are currently locking in the inode's io tree. So we check the
3746 * inode's i_size because of that (i_size updates are done while holding the
3747 * i_mutex, which we are holding here).
3748 * We also check to see if the inode has a size not greater than "datal" but has
3749 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3750 * protected against such concurrent fallocate calls by the i_mutex).
3751 *
3752 * If the file has no extents but a size greater than datal, do not allow the
3753 * copy because we would need turn the inline extent into a non-inline one (even
3754 * with NO_HOLES enabled). If we find our destination inode only has one inline
3755 * extent, just overwrite it with the source inline extent if its size is less
3756 * than the source extent's size, or we could copy the source inline extent's
3757 * data into the destination inode's inline extent if the later is greater then
3758 * the former.
3759 */
3760 static int clone_copy_inline_extent(struct inode *dst,
3761 struct btrfs_trans_handle *trans,
3762 struct btrfs_path *path,
3763 struct btrfs_key *new_key,
3764 const u64 drop_start,
3765 const u64 datal,
3766 const u64 skip,
3767 const u64 size,
3768 char *inline_data)
3769 {
3770 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
3771 struct btrfs_root *root = BTRFS_I(dst)->root;
3772 const u64 aligned_end = ALIGN(new_key->offset + datal,
3773 fs_info->sectorsize);
3774 int ret;
3775 struct btrfs_key key;
3776
3777 if (new_key->offset > 0)
3778 return -EOPNOTSUPP;
3779
3780 key.objectid = btrfs_ino(BTRFS_I(dst));
3781 key.type = BTRFS_EXTENT_DATA_KEY;
3782 key.offset = 0;
3783 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3784 if (ret < 0) {
3785 return ret;
3786 } else if (ret > 0) {
3787 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3788 ret = btrfs_next_leaf(root, path);
3789 if (ret < 0)
3790 return ret;
3791 else if (ret > 0)
3792 goto copy_inline_extent;
3793 }
3794 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3795 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3796 key.type == BTRFS_EXTENT_DATA_KEY) {
3797 ASSERT(key.offset > 0);
3798 return -EOPNOTSUPP;
3799 }
3800 } else if (i_size_read(dst) <= datal) {
3801 struct btrfs_file_extent_item *ei;
3802 u64 ext_len;
3803
3804 /*
3805 * If the file size is <= datal, make sure there are no other
3806 * extents following (can happen do to an fallocate call with
3807 * the flag FALLOC_FL_KEEP_SIZE).
3808 */
3809 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3810 struct btrfs_file_extent_item);
3811 /*
3812 * If it's an inline extent, it can not have other extents
3813 * following it.
3814 */
3815 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3816 BTRFS_FILE_EXTENT_INLINE)
3817 goto copy_inline_extent;
3818
3819 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3820 if (ext_len > aligned_end)
3821 return -EOPNOTSUPP;
3822
3823 ret = btrfs_next_item(root, path);
3824 if (ret < 0) {
3825 return ret;
3826 } else if (ret == 0) {
3827 btrfs_item_key_to_cpu(path->nodes[0], &key,
3828 path->slots[0]);
3829 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3830 key.type == BTRFS_EXTENT_DATA_KEY)
3831 return -EOPNOTSUPP;
3832 }
3833 }
3834
3835 copy_inline_extent:
3836 /*
3837 * We have no extent items, or we have an extent at offset 0 which may
3838 * or may not be inlined. All these cases are dealt the same way.
3839 */
3840 if (i_size_read(dst) > datal) {
3841 /*
3842 * If the destination inode has an inline extent...
3843 * This would require copying the data from the source inline
3844 * extent into the beginning of the destination's inline extent.
3845 * But this is really complex, both extents can be compressed
3846 * or just one of them, which would require decompressing and
3847 * re-compressing data (which could increase the new compressed
3848 * size, not allowing the compressed data to fit anymore in an
3849 * inline extent).
3850 * So just don't support this case for now (it should be rare,
3851 * we are not really saving space when cloning inline extents).
3852 */
3853 return -EOPNOTSUPP;
3854 }
3855
3856 btrfs_release_path(path);
3857 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3858 if (ret)
3859 return ret;
3860 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3861 if (ret)
3862 return ret;
3863
3864 if (skip) {
3865 const u32 start = btrfs_file_extent_calc_inline_size(0);
3866
3867 memmove(inline_data + start, inline_data + start + skip, datal);
3868 }
3869
3870 write_extent_buffer(path->nodes[0], inline_data,
3871 btrfs_item_ptr_offset(path->nodes[0],
3872 path->slots[0]),
3873 size);
3874 inode_add_bytes(dst, datal);
3875
3876 return 0;
3877 }
3878
3879 /**
3880 * btrfs_clone() - clone a range from inode file to another
3881 *
3882 * @src: Inode to clone from
3883 * @inode: Inode to clone to
3884 * @off: Offset within source to start clone from
3885 * @olen: Original length, passed by user, of range to clone
3886 * @olen_aligned: Block-aligned value of olen
3887 * @destoff: Offset within @inode to start clone
3888 * @no_time_update: Whether to update mtime/ctime on the target inode
3889 */
3890 static int btrfs_clone(struct inode *src, struct inode *inode,
3891 const u64 off, const u64 olen, const u64 olen_aligned,
3892 const u64 destoff, int no_time_update)
3893 {
3894 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3895 struct btrfs_root *root = BTRFS_I(inode)->root;
3896 struct btrfs_path *path = NULL;
3897 struct extent_buffer *leaf;
3898 struct btrfs_trans_handle *trans;
3899 char *buf = NULL;
3900 struct btrfs_key key;
3901 u32 nritems;
3902 int slot;
3903 int ret;
3904 const u64 len = olen_aligned;
3905 u64 last_dest_end = destoff;
3906
3907 ret = -ENOMEM;
3908 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3909 if (!buf)
3910 return ret;
3911
3912 path = btrfs_alloc_path();
3913 if (!path) {
3914 kvfree(buf);
3915 return ret;
3916 }
3917
3918 path->reada = READA_FORWARD;
3919 /* clone data */
3920 key.objectid = btrfs_ino(BTRFS_I(src));
3921 key.type = BTRFS_EXTENT_DATA_KEY;
3922 key.offset = off;
3923
3924 while (1) {
3925 u64 next_key_min_offset = key.offset + 1;
3926
3927 /*
3928 * note the key will change type as we walk through the
3929 * tree.
3930 */
3931 path->leave_spinning = 1;
3932 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3933 0, 0);
3934 if (ret < 0)
3935 goto out;
3936 /*
3937 * First search, if no extent item that starts at offset off was
3938 * found but the previous item is an extent item, it's possible
3939 * it might overlap our target range, therefore process it.
3940 */
3941 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3942 btrfs_item_key_to_cpu(path->nodes[0], &key,
3943 path->slots[0] - 1);
3944 if (key.type == BTRFS_EXTENT_DATA_KEY)
3945 path->slots[0]--;
3946 }
3947
3948 nritems = btrfs_header_nritems(path->nodes[0]);
3949 process_slot:
3950 if (path->slots[0] >= nritems) {
3951 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3952 if (ret < 0)
3953 goto out;
3954 if (ret > 0)
3955 break;
3956 nritems = btrfs_header_nritems(path->nodes[0]);
3957 }
3958 leaf = path->nodes[0];
3959 slot = path->slots[0];
3960
3961 btrfs_item_key_to_cpu(leaf, &key, slot);
3962 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3963 key.objectid != btrfs_ino(BTRFS_I(src)))
3964 break;
3965
3966 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3967 struct btrfs_file_extent_item *extent;
3968 int type;
3969 u32 size;
3970 struct btrfs_key new_key;
3971 u64 disko = 0, diskl = 0;
3972 u64 datao = 0, datal = 0;
3973 u8 comp;
3974 u64 drop_start;
3975
3976 extent = btrfs_item_ptr(leaf, slot,
3977 struct btrfs_file_extent_item);
3978 comp = btrfs_file_extent_compression(leaf, extent);
3979 type = btrfs_file_extent_type(leaf, extent);
3980 if (type == BTRFS_FILE_EXTENT_REG ||
3981 type == BTRFS_FILE_EXTENT_PREALLOC) {
3982 disko = btrfs_file_extent_disk_bytenr(leaf,
3983 extent);
3984 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3985 extent);
3986 datao = btrfs_file_extent_offset(leaf, extent);
3987 datal = btrfs_file_extent_num_bytes(leaf,
3988 extent);
3989 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3990 /* take upper bound, may be compressed */
3991 datal = btrfs_file_extent_ram_bytes(leaf,
3992 extent);
3993 }
3994
3995 /*
3996 * The first search might have left us at an extent
3997 * item that ends before our target range's start, can
3998 * happen if we have holes and NO_HOLES feature enabled.
3999 */
4000 if (key.offset + datal <= off) {
4001 path->slots[0]++;
4002 goto process_slot;
4003 } else if (key.offset >= off + len) {
4004 break;
4005 }
4006 next_key_min_offset = key.offset + datal;
4007 size = btrfs_item_size_nr(leaf, slot);
4008 read_extent_buffer(leaf, buf,
4009 btrfs_item_ptr_offset(leaf, slot),
4010 size);
4011
4012 btrfs_release_path(path);
4013 path->leave_spinning = 0;
4014
4015 memcpy(&new_key, &key, sizeof(new_key));
4016 new_key.objectid = btrfs_ino(BTRFS_I(inode));
4017 if (off <= key.offset)
4018 new_key.offset = key.offset + destoff - off;
4019 else
4020 new_key.offset = destoff;
4021
4022 /*
4023 * Deal with a hole that doesn't have an extent item
4024 * that represents it (NO_HOLES feature enabled).
4025 * This hole is either in the middle of the cloning
4026 * range or at the beginning (fully overlaps it or
4027 * partially overlaps it).
4028 */
4029 if (new_key.offset != last_dest_end)
4030 drop_start = last_dest_end;
4031 else
4032 drop_start = new_key.offset;
4033
4034 /*
4035 * 1 - adjusting old extent (we may have to split it)
4036 * 1 - add new extent
4037 * 1 - inode update
4038 */
4039 trans = btrfs_start_transaction(root, 3);
4040 if (IS_ERR(trans)) {
4041 ret = PTR_ERR(trans);
4042 goto out;
4043 }
4044
4045 if (type == BTRFS_FILE_EXTENT_REG ||
4046 type == BTRFS_FILE_EXTENT_PREALLOC) {
4047 /*
4048 * a | --- range to clone ---| b
4049 * | ------------- extent ------------- |
4050 */
4051
4052 /* subtract range b */
4053 if (key.offset + datal > off + len)
4054 datal = off + len - key.offset;
4055
4056 /* subtract range a */
4057 if (off > key.offset) {
4058 datao += off - key.offset;
4059 datal -= off - key.offset;
4060 }
4061
4062 ret = btrfs_drop_extents(trans, root, inode,
4063 drop_start,
4064 new_key.offset + datal,
4065 1);
4066 if (ret) {
4067 if (ret != -EOPNOTSUPP)
4068 btrfs_abort_transaction(trans,
4069 ret);
4070 btrfs_end_transaction(trans);
4071 goto out;
4072 }
4073
4074 ret = btrfs_insert_empty_item(trans, root, path,
4075 &new_key, size);
4076 if (ret) {
4077 btrfs_abort_transaction(trans, ret);
4078 btrfs_end_transaction(trans);
4079 goto out;
4080 }
4081
4082 leaf = path->nodes[0];
4083 slot = path->slots[0];
4084 write_extent_buffer(leaf, buf,
4085 btrfs_item_ptr_offset(leaf, slot),
4086 size);
4087
4088 extent = btrfs_item_ptr(leaf, slot,
4089 struct btrfs_file_extent_item);
4090
4091 /* disko == 0 means it's a hole */
4092 if (!disko)
4093 datao = 0;
4094
4095 btrfs_set_file_extent_offset(leaf, extent,
4096 datao);
4097 btrfs_set_file_extent_num_bytes(leaf, extent,
4098 datal);
4099
4100 if (disko) {
4101 inode_add_bytes(inode, datal);
4102 ret = btrfs_inc_extent_ref(trans,
4103 root,
4104 disko, diskl, 0,
4105 root->root_key.objectid,
4106 btrfs_ino(BTRFS_I(inode)),
4107 new_key.offset - datao);
4108 if (ret) {
4109 btrfs_abort_transaction(trans,
4110 ret);
4111 btrfs_end_transaction(trans);
4112 goto out;
4113
4114 }
4115 }
4116 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
4117 u64 skip = 0;
4118 u64 trim = 0;
4119
4120 if (off > key.offset) {
4121 skip = off - key.offset;
4122 new_key.offset += skip;
4123 }
4124
4125 if (key.offset + datal > off + len)
4126 trim = key.offset + datal - (off + len);
4127
4128 if (comp && (skip || trim)) {
4129 ret = -EINVAL;
4130 btrfs_end_transaction(trans);
4131 goto out;
4132 }
4133 size -= skip + trim;
4134 datal -= skip + trim;
4135
4136 ret = clone_copy_inline_extent(inode,
4137 trans, path,
4138 &new_key,
4139 drop_start,
4140 datal,
4141 skip, size, buf);
4142 if (ret) {
4143 if (ret != -EOPNOTSUPP)
4144 btrfs_abort_transaction(trans,
4145 ret);
4146 btrfs_end_transaction(trans);
4147 goto out;
4148 }
4149 leaf = path->nodes[0];
4150 slot = path->slots[0];
4151 }
4152
4153 /* If we have an implicit hole (NO_HOLES feature). */
4154 if (drop_start < new_key.offset)
4155 clone_update_extent_map(BTRFS_I(inode), trans,
4156 NULL, drop_start,
4157 new_key.offset - drop_start);
4158
4159 clone_update_extent_map(BTRFS_I(inode), trans,
4160 path, 0, 0);
4161
4162 btrfs_mark_buffer_dirty(leaf);
4163 btrfs_release_path(path);
4164
4165 last_dest_end = ALIGN(new_key.offset + datal,
4166 fs_info->sectorsize);
4167 ret = clone_finish_inode_update(trans, inode,
4168 last_dest_end,
4169 destoff, olen,
4170 no_time_update);
4171 if (ret)
4172 goto out;
4173 if (new_key.offset + datal >= destoff + len)
4174 break;
4175 }
4176 btrfs_release_path(path);
4177 key.offset = next_key_min_offset;
4178
4179 if (fatal_signal_pending(current)) {
4180 ret = -EINTR;
4181 goto out;
4182 }
4183 }
4184 ret = 0;
4185
4186 if (last_dest_end < destoff + len) {
4187 /*
4188 * We have an implicit hole (NO_HOLES feature is enabled) that
4189 * fully or partially overlaps our cloning range at its end.
4190 */
4191 btrfs_release_path(path);
4192
4193 /*
4194 * 1 - remove extent(s)
4195 * 1 - inode update
4196 */
4197 trans = btrfs_start_transaction(root, 2);
4198 if (IS_ERR(trans)) {
4199 ret = PTR_ERR(trans);
4200 goto out;
4201 }
4202 ret = btrfs_drop_extents(trans, root, inode,
4203 last_dest_end, destoff + len, 1);
4204 if (ret) {
4205 if (ret != -EOPNOTSUPP)
4206 btrfs_abort_transaction(trans, ret);
4207 btrfs_end_transaction(trans);
4208 goto out;
4209 }
4210 clone_update_extent_map(BTRFS_I(inode), trans, NULL,
4211 last_dest_end,
4212 destoff + len - last_dest_end);
4213 ret = clone_finish_inode_update(trans, inode, destoff + len,
4214 destoff, olen, no_time_update);
4215 }
4216
4217 out:
4218 btrfs_free_path(path);
4219 kvfree(buf);
4220 return ret;
4221 }
4222
4223 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
4224 u64 off, u64 olen, u64 destoff)
4225 {
4226 struct inode *inode = file_inode(file);
4227 struct inode *src = file_inode(file_src);
4228 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4229 struct btrfs_root *root = BTRFS_I(inode)->root;
4230 int ret;
4231 u64 len = olen;
4232 u64 bs = fs_info->sb->s_blocksize;
4233 int same_inode = src == inode;
4234
4235 /*
4236 * TODO:
4237 * - split compressed inline extents. annoying: we need to
4238 * decompress into destination's address_space (the file offset
4239 * may change, so source mapping won't do), then recompress (or
4240 * otherwise reinsert) a subrange.
4241 *
4242 * - split destination inode's inline extents. The inline extents can
4243 * be either compressed or non-compressed.
4244 */
4245
4246 if (btrfs_root_readonly(root))
4247 return -EROFS;
4248
4249 if (file_src->f_path.mnt != file->f_path.mnt ||
4250 src->i_sb != inode->i_sb)
4251 return -EXDEV;
4252
4253 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
4254 return -EISDIR;
4255
4256 if (!same_inode) {
4257 btrfs_double_inode_lock(src, inode);
4258 } else {
4259 inode_lock(src);
4260 }
4261
4262 /* don't make the dst file partly checksummed */
4263 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
4264 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
4265 ret = -EINVAL;
4266 goto out_unlock;
4267 }
4268
4269 /* determine range to clone */
4270 ret = -EINVAL;
4271 if (off + len > src->i_size || off + len < off)
4272 goto out_unlock;
4273 if (len == 0)
4274 olen = len = src->i_size - off;
4275 /* if we extend to eof, continue to block boundary */
4276 if (off + len == src->i_size)
4277 len = ALIGN(src->i_size, bs) - off;
4278
4279 if (len == 0) {
4280 ret = 0;
4281 goto out_unlock;
4282 }
4283
4284 /* verify the end result is block aligned */
4285 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
4286 !IS_ALIGNED(destoff, bs))
4287 goto out_unlock;
4288
4289 /* verify if ranges are overlapped within the same file */
4290 if (same_inode) {
4291 if (destoff + len > off && destoff < off + len)
4292 goto out_unlock;
4293 }
4294
4295 if (destoff > inode->i_size) {
4296 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
4297 if (ret)
4298 goto out_unlock;
4299 }
4300
4301 /*
4302 * Lock the target range too. Right after we replace the file extent
4303 * items in the fs tree (which now point to the cloned data), we might
4304 * have a worker replace them with extent items relative to a write
4305 * operation that was issued before this clone operation (i.e. confront
4306 * with inode.c:btrfs_finish_ordered_io).
4307 */
4308 if (same_inode) {
4309 u64 lock_start = min_t(u64, off, destoff);
4310 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
4311
4312 ret = lock_extent_range(src, lock_start, lock_len, true);
4313 } else {
4314 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
4315 true);
4316 }
4317 ASSERT(ret == 0);
4318 if (WARN_ON(ret)) {
4319 /* ranges in the io trees already unlocked */
4320 goto out_unlock;
4321 }
4322
4323 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
4324
4325 if (same_inode) {
4326 u64 lock_start = min_t(u64, off, destoff);
4327 u64 lock_end = max_t(u64, off, destoff) + len - 1;
4328
4329 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
4330 } else {
4331 btrfs_double_extent_unlock(src, off, inode, destoff, len);
4332 }
4333 /*
4334 * Truncate page cache pages so that future reads will see the cloned
4335 * data immediately and not the previous data.
4336 */
4337 truncate_inode_pages_range(&inode->i_data,
4338 round_down(destoff, PAGE_SIZE),
4339 round_up(destoff + len, PAGE_SIZE) - 1);
4340 out_unlock:
4341 if (!same_inode)
4342 btrfs_double_inode_unlock(src, inode);
4343 else
4344 inode_unlock(src);
4345 return ret;
4346 }
4347
4348 int btrfs_clone_file_range(struct file *src_file, loff_t off,
4349 struct file *dst_file, loff_t destoff, u64 len)
4350 {
4351 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
4352 }
4353
4354 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
4355 {
4356 struct inode *inode = file_inode(file);
4357 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4358 struct btrfs_root *root = BTRFS_I(inode)->root;
4359 struct btrfs_root *new_root;
4360 struct btrfs_dir_item *di;
4361 struct btrfs_trans_handle *trans;
4362 struct btrfs_path *path;
4363 struct btrfs_key location;
4364 struct btrfs_disk_key disk_key;
4365 u64 objectid = 0;
4366 u64 dir_id;
4367 int ret;
4368
4369 if (!capable(CAP_SYS_ADMIN))
4370 return -EPERM;
4371
4372 ret = mnt_want_write_file(file);
4373 if (ret)
4374 return ret;
4375
4376 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4377 ret = -EFAULT;
4378 goto out;
4379 }
4380
4381 if (!objectid)
4382 objectid = BTRFS_FS_TREE_OBJECTID;
4383
4384 location.objectid = objectid;
4385 location.type = BTRFS_ROOT_ITEM_KEY;
4386 location.offset = (u64)-1;
4387
4388 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
4389 if (IS_ERR(new_root)) {
4390 ret = PTR_ERR(new_root);
4391 goto out;
4392 }
4393 if (!is_fstree(new_root->objectid)) {
4394 ret = -ENOENT;
4395 goto out;
4396 }
4397
4398 path = btrfs_alloc_path();
4399 if (!path) {
4400 ret = -ENOMEM;
4401 goto out;
4402 }
4403 path->leave_spinning = 1;
4404
4405 trans = btrfs_start_transaction(root, 1);
4406 if (IS_ERR(trans)) {
4407 btrfs_free_path(path);
4408 ret = PTR_ERR(trans);
4409 goto out;
4410 }
4411
4412 dir_id = btrfs_super_root_dir(fs_info->super_copy);
4413 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
4414 dir_id, "default", 7, 1);
4415 if (IS_ERR_OR_NULL(di)) {
4416 btrfs_free_path(path);
4417 btrfs_end_transaction(trans);
4418 btrfs_err(fs_info,
4419 "Umm, you don't have the default diritem, this isn't going to work");
4420 ret = -ENOENT;
4421 goto out;
4422 }
4423
4424 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4425 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4426 btrfs_mark_buffer_dirty(path->nodes[0]);
4427 btrfs_free_path(path);
4428
4429 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
4430 btrfs_end_transaction(trans);
4431 out:
4432 mnt_drop_write_file(file);
4433 return ret;
4434 }
4435
4436 static void get_block_group_info(struct list_head *groups_list,
4437 struct btrfs_ioctl_space_info *space)
4438 {
4439 struct btrfs_block_group_cache *block_group;
4440
4441 space->total_bytes = 0;
4442 space->used_bytes = 0;
4443 space->flags = 0;
4444 list_for_each_entry(block_group, groups_list, list) {
4445 space->flags = block_group->flags;
4446 space->total_bytes += block_group->key.offset;
4447 space->used_bytes +=
4448 btrfs_block_group_used(&block_group->item);
4449 }
4450 }
4451
4452 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
4453 void __user *arg)
4454 {
4455 struct btrfs_ioctl_space_args space_args;
4456 struct btrfs_ioctl_space_info space;
4457 struct btrfs_ioctl_space_info *dest;
4458 struct btrfs_ioctl_space_info *dest_orig;
4459 struct btrfs_ioctl_space_info __user *user_dest;
4460 struct btrfs_space_info *info;
4461 static const u64 types[] = {
4462 BTRFS_BLOCK_GROUP_DATA,
4463 BTRFS_BLOCK_GROUP_SYSTEM,
4464 BTRFS_BLOCK_GROUP_METADATA,
4465 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
4466 };
4467 int num_types = 4;
4468 int alloc_size;
4469 int ret = 0;
4470 u64 slot_count = 0;
4471 int i, c;
4472
4473 if (copy_from_user(&space_args,
4474 (struct btrfs_ioctl_space_args __user *)arg,
4475 sizeof(space_args)))
4476 return -EFAULT;
4477
4478 for (i = 0; i < num_types; i++) {
4479 struct btrfs_space_info *tmp;
4480
4481 info = NULL;
4482 rcu_read_lock();
4483 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4484 list) {
4485 if (tmp->flags == types[i]) {
4486 info = tmp;
4487 break;
4488 }
4489 }
4490 rcu_read_unlock();
4491
4492 if (!info)
4493 continue;
4494
4495 down_read(&info->groups_sem);
4496 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4497 if (!list_empty(&info->block_groups[c]))
4498 slot_count++;
4499 }
4500 up_read(&info->groups_sem);
4501 }
4502
4503 /*
4504 * Global block reserve, exported as a space_info
4505 */
4506 slot_count++;
4507
4508 /* space_slots == 0 means they are asking for a count */
4509 if (space_args.space_slots == 0) {
4510 space_args.total_spaces = slot_count;
4511 goto out;
4512 }
4513
4514 slot_count = min_t(u64, space_args.space_slots, slot_count);
4515
4516 alloc_size = sizeof(*dest) * slot_count;
4517
4518 /* we generally have at most 6 or so space infos, one for each raid
4519 * level. So, a whole page should be more than enough for everyone
4520 */
4521 if (alloc_size > PAGE_SIZE)
4522 return -ENOMEM;
4523
4524 space_args.total_spaces = 0;
4525 dest = kmalloc(alloc_size, GFP_KERNEL);
4526 if (!dest)
4527 return -ENOMEM;
4528 dest_orig = dest;
4529
4530 /* now we have a buffer to copy into */
4531 for (i = 0; i < num_types; i++) {
4532 struct btrfs_space_info *tmp;
4533
4534 if (!slot_count)
4535 break;
4536
4537 info = NULL;
4538 rcu_read_lock();
4539 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4540 list) {
4541 if (tmp->flags == types[i]) {
4542 info = tmp;
4543 break;
4544 }
4545 }
4546 rcu_read_unlock();
4547
4548 if (!info)
4549 continue;
4550 down_read(&info->groups_sem);
4551 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4552 if (!list_empty(&info->block_groups[c])) {
4553 get_block_group_info(&info->block_groups[c],
4554 &space);
4555 memcpy(dest, &space, sizeof(space));
4556 dest++;
4557 space_args.total_spaces++;
4558 slot_count--;
4559 }
4560 if (!slot_count)
4561 break;
4562 }
4563 up_read(&info->groups_sem);
4564 }
4565
4566 /*
4567 * Add global block reserve
4568 */
4569 if (slot_count) {
4570 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
4571
4572 spin_lock(&block_rsv->lock);
4573 space.total_bytes = block_rsv->size;
4574 space.used_bytes = block_rsv->size - block_rsv->reserved;
4575 spin_unlock(&block_rsv->lock);
4576 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4577 memcpy(dest, &space, sizeof(space));
4578 space_args.total_spaces++;
4579 }
4580
4581 user_dest = (struct btrfs_ioctl_space_info __user *)
4582 (arg + sizeof(struct btrfs_ioctl_space_args));
4583
4584 if (copy_to_user(user_dest, dest_orig, alloc_size))
4585 ret = -EFAULT;
4586
4587 kfree(dest_orig);
4588 out:
4589 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4590 ret = -EFAULT;
4591
4592 return ret;
4593 }
4594
4595 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4596 void __user *argp)
4597 {
4598 struct btrfs_trans_handle *trans;
4599 u64 transid;
4600 int ret;
4601
4602 trans = btrfs_attach_transaction_barrier(root);
4603 if (IS_ERR(trans)) {
4604 if (PTR_ERR(trans) != -ENOENT)
4605 return PTR_ERR(trans);
4606
4607 /* No running transaction, don't bother */
4608 transid = root->fs_info->last_trans_committed;
4609 goto out;
4610 }
4611 transid = trans->transid;
4612 ret = btrfs_commit_transaction_async(trans, 0);
4613 if (ret) {
4614 btrfs_end_transaction(trans);
4615 return ret;
4616 }
4617 out:
4618 if (argp)
4619 if (copy_to_user(argp, &transid, sizeof(transid)))
4620 return -EFAULT;
4621 return 0;
4622 }
4623
4624 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
4625 void __user *argp)
4626 {
4627 u64 transid;
4628
4629 if (argp) {
4630 if (copy_from_user(&transid, argp, sizeof(transid)))
4631 return -EFAULT;
4632 } else {
4633 transid = 0; /* current trans */
4634 }
4635 return btrfs_wait_for_commit(fs_info, transid);
4636 }
4637
4638 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4639 {
4640 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4641 struct btrfs_ioctl_scrub_args *sa;
4642 int ret;
4643
4644 if (!capable(CAP_SYS_ADMIN))
4645 return -EPERM;
4646
4647 sa = memdup_user(arg, sizeof(*sa));
4648 if (IS_ERR(sa))
4649 return PTR_ERR(sa);
4650
4651 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4652 ret = mnt_want_write_file(file);
4653 if (ret)
4654 goto out;
4655 }
4656
4657 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4658 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4659 0);
4660
4661 if (copy_to_user(arg, sa, sizeof(*sa)))
4662 ret = -EFAULT;
4663
4664 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4665 mnt_drop_write_file(file);
4666 out:
4667 kfree(sa);
4668 return ret;
4669 }
4670
4671 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
4672 {
4673 if (!capable(CAP_SYS_ADMIN))
4674 return -EPERM;
4675
4676 return btrfs_scrub_cancel(fs_info);
4677 }
4678
4679 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
4680 void __user *arg)
4681 {
4682 struct btrfs_ioctl_scrub_args *sa;
4683 int ret;
4684
4685 if (!capable(CAP_SYS_ADMIN))
4686 return -EPERM;
4687
4688 sa = memdup_user(arg, sizeof(*sa));
4689 if (IS_ERR(sa))
4690 return PTR_ERR(sa);
4691
4692 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
4693
4694 if (copy_to_user(arg, sa, sizeof(*sa)))
4695 ret = -EFAULT;
4696
4697 kfree(sa);
4698 return ret;
4699 }
4700
4701 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
4702 void __user *arg)
4703 {
4704 struct btrfs_ioctl_get_dev_stats *sa;
4705 int ret;
4706
4707 sa = memdup_user(arg, sizeof(*sa));
4708 if (IS_ERR(sa))
4709 return PTR_ERR(sa);
4710
4711 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4712 kfree(sa);
4713 return -EPERM;
4714 }
4715
4716 ret = btrfs_get_dev_stats(fs_info, sa);
4717
4718 if (copy_to_user(arg, sa, sizeof(*sa)))
4719 ret = -EFAULT;
4720
4721 kfree(sa);
4722 return ret;
4723 }
4724
4725 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4726 void __user *arg)
4727 {
4728 struct btrfs_ioctl_dev_replace_args *p;
4729 int ret;
4730
4731 if (!capable(CAP_SYS_ADMIN))
4732 return -EPERM;
4733
4734 p = memdup_user(arg, sizeof(*p));
4735 if (IS_ERR(p))
4736 return PTR_ERR(p);
4737
4738 switch (p->cmd) {
4739 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4740 if (sb_rdonly(fs_info->sb)) {
4741 ret = -EROFS;
4742 goto out;
4743 }
4744 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4745 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4746 } else {
4747 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
4748 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4749 }
4750 break;
4751 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4752 btrfs_dev_replace_status(fs_info, p);
4753 ret = 0;
4754 break;
4755 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4756 p->result = btrfs_dev_replace_cancel(fs_info);
4757 ret = 0;
4758 break;
4759 default:
4760 ret = -EINVAL;
4761 break;
4762 }
4763
4764 if (copy_to_user(arg, p, sizeof(*p)))
4765 ret = -EFAULT;
4766 out:
4767 kfree(p);
4768 return ret;
4769 }
4770
4771 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4772 {
4773 int ret = 0;
4774 int i;
4775 u64 rel_ptr;
4776 int size;
4777 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4778 struct inode_fs_paths *ipath = NULL;
4779 struct btrfs_path *path;
4780
4781 if (!capable(CAP_DAC_READ_SEARCH))
4782 return -EPERM;
4783
4784 path = btrfs_alloc_path();
4785 if (!path) {
4786 ret = -ENOMEM;
4787 goto out;
4788 }
4789
4790 ipa = memdup_user(arg, sizeof(*ipa));
4791 if (IS_ERR(ipa)) {
4792 ret = PTR_ERR(ipa);
4793 ipa = NULL;
4794 goto out;
4795 }
4796
4797 size = min_t(u32, ipa->size, 4096);
4798 ipath = init_ipath(size, root, path);
4799 if (IS_ERR(ipath)) {
4800 ret = PTR_ERR(ipath);
4801 ipath = NULL;
4802 goto out;
4803 }
4804
4805 ret = paths_from_inode(ipa->inum, ipath);
4806 if (ret < 0)
4807 goto out;
4808
4809 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4810 rel_ptr = ipath->fspath->val[i] -
4811 (u64)(unsigned long)ipath->fspath->val;
4812 ipath->fspath->val[i] = rel_ptr;
4813 }
4814
4815 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
4816 ipath->fspath, size);
4817 if (ret) {
4818 ret = -EFAULT;
4819 goto out;
4820 }
4821
4822 out:
4823 btrfs_free_path(path);
4824 free_ipath(ipath);
4825 kfree(ipa);
4826
4827 return ret;
4828 }
4829
4830 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4831 {
4832 struct btrfs_data_container *inodes = ctx;
4833 const size_t c = 3 * sizeof(u64);
4834
4835 if (inodes->bytes_left >= c) {
4836 inodes->bytes_left -= c;
4837 inodes->val[inodes->elem_cnt] = inum;
4838 inodes->val[inodes->elem_cnt + 1] = offset;
4839 inodes->val[inodes->elem_cnt + 2] = root;
4840 inodes->elem_cnt += 3;
4841 } else {
4842 inodes->bytes_missing += c - inodes->bytes_left;
4843 inodes->bytes_left = 0;
4844 inodes->elem_missed += 3;
4845 }
4846
4847 return 0;
4848 }
4849
4850 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
4851 void __user *arg, int version)
4852 {
4853 int ret = 0;
4854 int size;
4855 struct btrfs_ioctl_logical_ino_args *loi;
4856 struct btrfs_data_container *inodes = NULL;
4857 struct btrfs_path *path = NULL;
4858 bool ignore_offset;
4859
4860 if (!capable(CAP_SYS_ADMIN))
4861 return -EPERM;
4862
4863 loi = memdup_user(arg, sizeof(*loi));
4864 if (IS_ERR(loi))
4865 return PTR_ERR(loi);
4866
4867 if (version == 1) {
4868 ignore_offset = false;
4869 size = min_t(u32, loi->size, SZ_64K);
4870 } else {
4871 /* All reserved bits must be 0 for now */
4872 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
4873 ret = -EINVAL;
4874 goto out_loi;
4875 }
4876 /* Only accept flags we have defined so far */
4877 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
4878 ret = -EINVAL;
4879 goto out_loi;
4880 }
4881 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
4882 size = min_t(u32, loi->size, SZ_16M);
4883 }
4884
4885 path = btrfs_alloc_path();
4886 if (!path) {
4887 ret = -ENOMEM;
4888 goto out;
4889 }
4890
4891 inodes = init_data_container(size);
4892 if (IS_ERR(inodes)) {
4893 ret = PTR_ERR(inodes);
4894 inodes = NULL;
4895 goto out;
4896 }
4897
4898 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
4899 build_ino_list, inodes, ignore_offset);
4900 if (ret == -EINVAL)
4901 ret = -ENOENT;
4902 if (ret < 0)
4903 goto out;
4904
4905 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
4906 size);
4907 if (ret)
4908 ret = -EFAULT;
4909
4910 out:
4911 btrfs_free_path(path);
4912 kvfree(inodes);
4913 out_loi:
4914 kfree(loi);
4915
4916 return ret;
4917 }
4918
4919 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
4920 struct btrfs_ioctl_balance_args *bargs)
4921 {
4922 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4923
4924 bargs->flags = bctl->flags;
4925
4926 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
4927 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4928 if (atomic_read(&fs_info->balance_pause_req))
4929 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4930 if (atomic_read(&fs_info->balance_cancel_req))
4931 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4932
4933 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4934 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4935 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4936
4937 spin_lock(&fs_info->balance_lock);
4938 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4939 spin_unlock(&fs_info->balance_lock);
4940 }
4941
4942 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4943 {
4944 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4945 struct btrfs_fs_info *fs_info = root->fs_info;
4946 struct btrfs_ioctl_balance_args *bargs;
4947 struct btrfs_balance_control *bctl;
4948 bool need_unlock; /* for mut. excl. ops lock */
4949 int ret;
4950
4951 if (!capable(CAP_SYS_ADMIN))
4952 return -EPERM;
4953
4954 ret = mnt_want_write_file(file);
4955 if (ret)
4956 return ret;
4957
4958 again:
4959 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4960 mutex_lock(&fs_info->balance_mutex);
4961 need_unlock = true;
4962 goto locked;
4963 }
4964
4965 /*
4966 * mut. excl. ops lock is locked. Three possibilities:
4967 * (1) some other op is running
4968 * (2) balance is running
4969 * (3) balance is paused -- special case (think resume)
4970 */
4971 mutex_lock(&fs_info->balance_mutex);
4972 if (fs_info->balance_ctl) {
4973 /* this is either (2) or (3) */
4974 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4975 mutex_unlock(&fs_info->balance_mutex);
4976 /*
4977 * Lock released to allow other waiters to continue,
4978 * we'll reexamine the status again.
4979 */
4980 mutex_lock(&fs_info->balance_mutex);
4981
4982 if (fs_info->balance_ctl &&
4983 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4984 /* this is (3) */
4985 need_unlock = false;
4986 goto locked;
4987 }
4988
4989 mutex_unlock(&fs_info->balance_mutex);
4990 goto again;
4991 } else {
4992 /* this is (2) */
4993 mutex_unlock(&fs_info->balance_mutex);
4994 ret = -EINPROGRESS;
4995 goto out;
4996 }
4997 } else {
4998 /* this is (1) */
4999 mutex_unlock(&fs_info->balance_mutex);
5000 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
5001 goto out;
5002 }
5003
5004 locked:
5005 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
5006
5007 if (arg) {
5008 bargs = memdup_user(arg, sizeof(*bargs));
5009 if (IS_ERR(bargs)) {
5010 ret = PTR_ERR(bargs);
5011 goto out_unlock;
5012 }
5013
5014 if (bargs->flags & BTRFS_BALANCE_RESUME) {
5015 if (!fs_info->balance_ctl) {
5016 ret = -ENOTCONN;
5017 goto out_bargs;
5018 }
5019
5020 bctl = fs_info->balance_ctl;
5021 spin_lock(&fs_info->balance_lock);
5022 bctl->flags |= BTRFS_BALANCE_RESUME;
5023 spin_unlock(&fs_info->balance_lock);
5024
5025 goto do_balance;
5026 }
5027 } else {
5028 bargs = NULL;
5029 }
5030
5031 if (fs_info->balance_ctl) {
5032 ret = -EINPROGRESS;
5033 goto out_bargs;
5034 }
5035
5036 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
5037 if (!bctl) {
5038 ret = -ENOMEM;
5039 goto out_bargs;
5040 }
5041
5042 if (arg) {
5043 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
5044 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
5045 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
5046
5047 bctl->flags = bargs->flags;
5048 } else {
5049 /* balance everything - no filters */
5050 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
5051 }
5052
5053 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
5054 ret = -EINVAL;
5055 goto out_bctl;
5056 }
5057
5058 do_balance:
5059 /*
5060 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to
5061 * btrfs_balance. bctl is freed in reset_balance_state, or, if
5062 * restriper was paused all the way until unmount, in free_fs_info.
5063 * The flag should be cleared after reset_balance_state.
5064 */
5065 need_unlock = false;
5066
5067 ret = btrfs_balance(fs_info, bctl, bargs);
5068 bctl = NULL;
5069
5070 if (arg) {
5071 if (copy_to_user(arg, bargs, sizeof(*bargs)))
5072 ret = -EFAULT;
5073 }
5074
5075 out_bctl:
5076 kfree(bctl);
5077 out_bargs:
5078 kfree(bargs);
5079 out_unlock:
5080 mutex_unlock(&fs_info->balance_mutex);
5081 if (need_unlock)
5082 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
5083 out:
5084 mnt_drop_write_file(file);
5085 return ret;
5086 }
5087
5088 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
5089 {
5090 if (!capable(CAP_SYS_ADMIN))
5091 return -EPERM;
5092
5093 switch (cmd) {
5094 case BTRFS_BALANCE_CTL_PAUSE:
5095 return btrfs_pause_balance(fs_info);
5096 case BTRFS_BALANCE_CTL_CANCEL:
5097 return btrfs_cancel_balance(fs_info);
5098 }
5099
5100 return -EINVAL;
5101 }
5102
5103 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
5104 void __user *arg)
5105 {
5106 struct btrfs_ioctl_balance_args *bargs;
5107 int ret = 0;
5108
5109 if (!capable(CAP_SYS_ADMIN))
5110 return -EPERM;
5111
5112 mutex_lock(&fs_info->balance_mutex);
5113 if (!fs_info->balance_ctl) {
5114 ret = -ENOTCONN;
5115 goto out;
5116 }
5117
5118 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
5119 if (!bargs) {
5120 ret = -ENOMEM;
5121 goto out;
5122 }
5123
5124 btrfs_update_ioctl_balance_args(fs_info, bargs);
5125
5126 if (copy_to_user(arg, bargs, sizeof(*bargs)))
5127 ret = -EFAULT;
5128
5129 kfree(bargs);
5130 out:
5131 mutex_unlock(&fs_info->balance_mutex);
5132 return ret;
5133 }
5134
5135 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5136 {
5137 struct inode *inode = file_inode(file);
5138 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5139 struct btrfs_ioctl_quota_ctl_args *sa;
5140 struct btrfs_trans_handle *trans = NULL;
5141 int ret;
5142 int err;
5143
5144 if (!capable(CAP_SYS_ADMIN))
5145 return -EPERM;
5146
5147 ret = mnt_want_write_file(file);
5148 if (ret)
5149 return ret;
5150
5151 sa = memdup_user(arg, sizeof(*sa));
5152 if (IS_ERR(sa)) {
5153 ret = PTR_ERR(sa);
5154 goto drop_write;
5155 }
5156
5157 down_write(&fs_info->subvol_sem);
5158 trans = btrfs_start_transaction(fs_info->tree_root, 2);
5159 if (IS_ERR(trans)) {
5160 ret = PTR_ERR(trans);
5161 goto out;
5162 }
5163
5164 switch (sa->cmd) {
5165 case BTRFS_QUOTA_CTL_ENABLE:
5166 ret = btrfs_quota_enable(trans, fs_info);
5167 break;
5168 case BTRFS_QUOTA_CTL_DISABLE:
5169 ret = btrfs_quota_disable(trans, fs_info);
5170 break;
5171 default:
5172 ret = -EINVAL;
5173 break;
5174 }
5175
5176 err = btrfs_commit_transaction(trans);
5177 if (err && !ret)
5178 ret = err;
5179 out:
5180 kfree(sa);
5181 up_write(&fs_info->subvol_sem);
5182 drop_write:
5183 mnt_drop_write_file(file);
5184 return ret;
5185 }
5186
5187 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5188 {
5189 struct inode *inode = file_inode(file);
5190 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5191 struct btrfs_root *root = BTRFS_I(inode)->root;
5192 struct btrfs_ioctl_qgroup_assign_args *sa;
5193 struct btrfs_trans_handle *trans;
5194 int ret;
5195 int err;
5196
5197 if (!capable(CAP_SYS_ADMIN))
5198 return -EPERM;
5199
5200 ret = mnt_want_write_file(file);
5201 if (ret)
5202 return ret;
5203
5204 sa = memdup_user(arg, sizeof(*sa));
5205 if (IS_ERR(sa)) {
5206 ret = PTR_ERR(sa);
5207 goto drop_write;
5208 }
5209
5210 trans = btrfs_join_transaction(root);
5211 if (IS_ERR(trans)) {
5212 ret = PTR_ERR(trans);
5213 goto out;
5214 }
5215
5216 if (sa->assign) {
5217 ret = btrfs_add_qgroup_relation(trans, fs_info,
5218 sa->src, sa->dst);
5219 } else {
5220 ret = btrfs_del_qgroup_relation(trans, fs_info,
5221 sa->src, sa->dst);
5222 }
5223
5224 /* update qgroup status and info */
5225 err = btrfs_run_qgroups(trans, fs_info);
5226 if (err < 0)
5227 btrfs_handle_fs_error(fs_info, err,
5228 "failed to update qgroup status and info");
5229 err = btrfs_end_transaction(trans);
5230 if (err && !ret)
5231 ret = err;
5232
5233 out:
5234 kfree(sa);
5235 drop_write:
5236 mnt_drop_write_file(file);
5237 return ret;
5238 }
5239
5240 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5241 {
5242 struct inode *inode = file_inode(file);
5243 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5244 struct btrfs_root *root = BTRFS_I(inode)->root;
5245 struct btrfs_ioctl_qgroup_create_args *sa;
5246 struct btrfs_trans_handle *trans;
5247 int ret;
5248 int err;
5249
5250 if (!capable(CAP_SYS_ADMIN))
5251 return -EPERM;
5252
5253 ret = mnt_want_write_file(file);
5254 if (ret)
5255 return ret;
5256
5257 sa = memdup_user(arg, sizeof(*sa));
5258 if (IS_ERR(sa)) {
5259 ret = PTR_ERR(sa);
5260 goto drop_write;
5261 }
5262
5263 if (!sa->qgroupid) {
5264 ret = -EINVAL;
5265 goto out;
5266 }
5267
5268 trans = btrfs_join_transaction(root);
5269 if (IS_ERR(trans)) {
5270 ret = PTR_ERR(trans);
5271 goto out;
5272 }
5273
5274 if (sa->create) {
5275 ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid);
5276 } else {
5277 ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid);
5278 }
5279
5280 err = btrfs_end_transaction(trans);
5281 if (err && !ret)
5282 ret = err;
5283
5284 out:
5285 kfree(sa);
5286 drop_write:
5287 mnt_drop_write_file(file);
5288 return ret;
5289 }
5290
5291 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5292 {
5293 struct inode *inode = file_inode(file);
5294 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5295 struct btrfs_root *root = BTRFS_I(inode)->root;
5296 struct btrfs_ioctl_qgroup_limit_args *sa;
5297 struct btrfs_trans_handle *trans;
5298 int ret;
5299 int err;
5300 u64 qgroupid;
5301
5302 if (!capable(CAP_SYS_ADMIN))
5303 return -EPERM;
5304
5305 ret = mnt_want_write_file(file);
5306 if (ret)
5307 return ret;
5308
5309 sa = memdup_user(arg, sizeof(*sa));
5310 if (IS_ERR(sa)) {
5311 ret = PTR_ERR(sa);
5312 goto drop_write;
5313 }
5314
5315 trans = btrfs_join_transaction(root);
5316 if (IS_ERR(trans)) {
5317 ret = PTR_ERR(trans);
5318 goto out;
5319 }
5320
5321 qgroupid = sa->qgroupid;
5322 if (!qgroupid) {
5323 /* take the current subvol as qgroup */
5324 qgroupid = root->root_key.objectid;
5325 }
5326
5327 ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim);
5328
5329 err = btrfs_end_transaction(trans);
5330 if (err && !ret)
5331 ret = err;
5332
5333 out:
5334 kfree(sa);
5335 drop_write:
5336 mnt_drop_write_file(file);
5337 return ret;
5338 }
5339
5340 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
5341 {
5342 struct inode *inode = file_inode(file);
5343 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5344 struct btrfs_ioctl_quota_rescan_args *qsa;
5345 int ret;
5346
5347 if (!capable(CAP_SYS_ADMIN))
5348 return -EPERM;
5349
5350 ret = mnt_want_write_file(file);
5351 if (ret)
5352 return ret;
5353
5354 qsa = memdup_user(arg, sizeof(*qsa));
5355 if (IS_ERR(qsa)) {
5356 ret = PTR_ERR(qsa);
5357 goto drop_write;
5358 }
5359
5360 if (qsa->flags) {
5361 ret = -EINVAL;
5362 goto out;
5363 }
5364
5365 ret = btrfs_qgroup_rescan(fs_info);
5366
5367 out:
5368 kfree(qsa);
5369 drop_write:
5370 mnt_drop_write_file(file);
5371 return ret;
5372 }
5373
5374 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5375 {
5376 struct inode *inode = file_inode(file);
5377 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5378 struct btrfs_ioctl_quota_rescan_args *qsa;
5379 int ret = 0;
5380
5381 if (!capable(CAP_SYS_ADMIN))
5382 return -EPERM;
5383
5384 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
5385 if (!qsa)
5386 return -ENOMEM;
5387
5388 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5389 qsa->flags = 1;
5390 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
5391 }
5392
5393 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5394 ret = -EFAULT;
5395
5396 kfree(qsa);
5397 return ret;
5398 }
5399
5400 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5401 {
5402 struct inode *inode = file_inode(file);
5403 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5404
5405 if (!capable(CAP_SYS_ADMIN))
5406 return -EPERM;
5407
5408 return btrfs_qgroup_wait_for_completion(fs_info, true);
5409 }
5410
5411 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5412 struct btrfs_ioctl_received_subvol_args *sa)
5413 {
5414 struct inode *inode = file_inode(file);
5415 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5416 struct btrfs_root *root = BTRFS_I(inode)->root;
5417 struct btrfs_root_item *root_item = &root->root_item;
5418 struct btrfs_trans_handle *trans;
5419 struct timespec64 ct = current_time(inode);
5420 int ret = 0;
5421 int received_uuid_changed;
5422
5423 if (!inode_owner_or_capable(inode))
5424 return -EPERM;
5425
5426 ret = mnt_want_write_file(file);
5427 if (ret < 0)
5428 return ret;
5429
5430 down_write(&fs_info->subvol_sem);
5431
5432 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
5433 ret = -EINVAL;
5434 goto out;
5435 }
5436
5437 if (btrfs_root_readonly(root)) {
5438 ret = -EROFS;
5439 goto out;
5440 }
5441
5442 /*
5443 * 1 - root item
5444 * 2 - uuid items (received uuid + subvol uuid)
5445 */
5446 trans = btrfs_start_transaction(root, 3);
5447 if (IS_ERR(trans)) {
5448 ret = PTR_ERR(trans);
5449 trans = NULL;
5450 goto out;
5451 }
5452
5453 sa->rtransid = trans->transid;
5454 sa->rtime.sec = ct.tv_sec;
5455 sa->rtime.nsec = ct.tv_nsec;
5456
5457 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5458 BTRFS_UUID_SIZE);
5459 if (received_uuid_changed &&
5460 !btrfs_is_empty_uuid(root_item->received_uuid)) {
5461 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
5462 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5463 root->root_key.objectid);
5464 if (ret && ret != -ENOENT) {
5465 btrfs_abort_transaction(trans, ret);
5466 btrfs_end_transaction(trans);
5467 goto out;
5468 }
5469 }
5470 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5471 btrfs_set_root_stransid(root_item, sa->stransid);
5472 btrfs_set_root_rtransid(root_item, sa->rtransid);
5473 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5474 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5475 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5476 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5477
5478 ret = btrfs_update_root(trans, fs_info->tree_root,
5479 &root->root_key, &root->root_item);
5480 if (ret < 0) {
5481 btrfs_end_transaction(trans);
5482 goto out;
5483 }
5484 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5485 ret = btrfs_uuid_tree_add(trans, sa->uuid,
5486 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5487 root->root_key.objectid);
5488 if (ret < 0 && ret != -EEXIST) {
5489 btrfs_abort_transaction(trans, ret);
5490 btrfs_end_transaction(trans);
5491 goto out;
5492 }
5493 }
5494 ret = btrfs_commit_transaction(trans);
5495 out:
5496 up_write(&fs_info->subvol_sem);
5497 mnt_drop_write_file(file);
5498 return ret;
5499 }
5500
5501 #ifdef CONFIG_64BIT
5502 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5503 void __user *arg)
5504 {
5505 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5506 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5507 int ret = 0;
5508
5509 args32 = memdup_user(arg, sizeof(*args32));
5510 if (IS_ERR(args32))
5511 return PTR_ERR(args32);
5512
5513 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5514 if (!args64) {
5515 ret = -ENOMEM;
5516 goto out;
5517 }
5518
5519 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5520 args64->stransid = args32->stransid;
5521 args64->rtransid = args32->rtransid;
5522 args64->stime.sec = args32->stime.sec;
5523 args64->stime.nsec = args32->stime.nsec;
5524 args64->rtime.sec = args32->rtime.sec;
5525 args64->rtime.nsec = args32->rtime.nsec;
5526 args64->flags = args32->flags;
5527
5528 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5529 if (ret)
5530 goto out;
5531
5532 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5533 args32->stransid = args64->stransid;
5534 args32->rtransid = args64->rtransid;
5535 args32->stime.sec = args64->stime.sec;
5536 args32->stime.nsec = args64->stime.nsec;
5537 args32->rtime.sec = args64->rtime.sec;
5538 args32->rtime.nsec = args64->rtime.nsec;
5539 args32->flags = args64->flags;
5540
5541 ret = copy_to_user(arg, args32, sizeof(*args32));
5542 if (ret)
5543 ret = -EFAULT;
5544
5545 out:
5546 kfree(args32);
5547 kfree(args64);
5548 return ret;
5549 }
5550 #endif
5551
5552 static long btrfs_ioctl_set_received_subvol(struct file *file,
5553 void __user *arg)
5554 {
5555 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5556 int ret = 0;
5557
5558 sa = memdup_user(arg, sizeof(*sa));
5559 if (IS_ERR(sa))
5560 return PTR_ERR(sa);
5561
5562 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5563
5564 if (ret)
5565 goto out;
5566
5567 ret = copy_to_user(arg, sa, sizeof(*sa));
5568 if (ret)
5569 ret = -EFAULT;
5570
5571 out:
5572 kfree(sa);
5573 return ret;
5574 }
5575
5576 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5577 {
5578 struct inode *inode = file_inode(file);
5579 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5580 size_t len;
5581 int ret;
5582 char label[BTRFS_LABEL_SIZE];
5583
5584 spin_lock(&fs_info->super_lock);
5585 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5586 spin_unlock(&fs_info->super_lock);
5587
5588 len = strnlen(label, BTRFS_LABEL_SIZE);
5589
5590 if (len == BTRFS_LABEL_SIZE) {
5591 btrfs_warn(fs_info,
5592 "label is too long, return the first %zu bytes",
5593 --len);
5594 }
5595
5596 ret = copy_to_user(arg, label, len);
5597
5598 return ret ? -EFAULT : 0;
5599 }
5600
5601 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5602 {
5603 struct inode *inode = file_inode(file);
5604 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5605 struct btrfs_root *root = BTRFS_I(inode)->root;
5606 struct btrfs_super_block *super_block = fs_info->super_copy;
5607 struct btrfs_trans_handle *trans;
5608 char label[BTRFS_LABEL_SIZE];
5609 int ret;
5610
5611 if (!capable(CAP_SYS_ADMIN))
5612 return -EPERM;
5613
5614 if (copy_from_user(label, arg, sizeof(label)))
5615 return -EFAULT;
5616
5617 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5618 btrfs_err(fs_info,
5619 "unable to set label with more than %d bytes",
5620 BTRFS_LABEL_SIZE - 1);
5621 return -EINVAL;
5622 }
5623
5624 ret = mnt_want_write_file(file);
5625 if (ret)
5626 return ret;
5627
5628 trans = btrfs_start_transaction(root, 0);
5629 if (IS_ERR(trans)) {
5630 ret = PTR_ERR(trans);
5631 goto out_unlock;
5632 }
5633
5634 spin_lock(&fs_info->super_lock);
5635 strcpy(super_block->label, label);
5636 spin_unlock(&fs_info->super_lock);
5637 ret = btrfs_commit_transaction(trans);
5638
5639 out_unlock:
5640 mnt_drop_write_file(file);
5641 return ret;
5642 }
5643
5644 #define INIT_FEATURE_FLAGS(suffix) \
5645 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5646 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5647 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5648
5649 int btrfs_ioctl_get_supported_features(void __user *arg)
5650 {
5651 static const struct btrfs_ioctl_feature_flags features[3] = {
5652 INIT_FEATURE_FLAGS(SUPP),
5653 INIT_FEATURE_FLAGS(SAFE_SET),
5654 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5655 };
5656
5657 if (copy_to_user(arg, &features, sizeof(features)))
5658 return -EFAULT;
5659
5660 return 0;
5661 }
5662
5663 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5664 {
5665 struct inode *inode = file_inode(file);
5666 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5667 struct btrfs_super_block *super_block = fs_info->super_copy;
5668 struct btrfs_ioctl_feature_flags features;
5669
5670 features.compat_flags = btrfs_super_compat_flags(super_block);
5671 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5672 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5673
5674 if (copy_to_user(arg, &features, sizeof(features)))
5675 return -EFAULT;
5676
5677 return 0;
5678 }
5679
5680 static int check_feature_bits(struct btrfs_fs_info *fs_info,
5681 enum btrfs_feature_set set,
5682 u64 change_mask, u64 flags, u64 supported_flags,
5683 u64 safe_set, u64 safe_clear)
5684 {
5685 const char *type = btrfs_feature_set_names[set];
5686 char *names;
5687 u64 disallowed, unsupported;
5688 u64 set_mask = flags & change_mask;
5689 u64 clear_mask = ~flags & change_mask;
5690
5691 unsupported = set_mask & ~supported_flags;
5692 if (unsupported) {
5693 names = btrfs_printable_features(set, unsupported);
5694 if (names) {
5695 btrfs_warn(fs_info,
5696 "this kernel does not support the %s feature bit%s",
5697 names, strchr(names, ',') ? "s" : "");
5698 kfree(names);
5699 } else
5700 btrfs_warn(fs_info,
5701 "this kernel does not support %s bits 0x%llx",
5702 type, unsupported);
5703 return -EOPNOTSUPP;
5704 }
5705
5706 disallowed = set_mask & ~safe_set;
5707 if (disallowed) {
5708 names = btrfs_printable_features(set, disallowed);
5709 if (names) {
5710 btrfs_warn(fs_info,
5711 "can't set the %s feature bit%s while mounted",
5712 names, strchr(names, ',') ? "s" : "");
5713 kfree(names);
5714 } else
5715 btrfs_warn(fs_info,
5716 "can't set %s bits 0x%llx while mounted",
5717 type, disallowed);
5718 return -EPERM;
5719 }
5720
5721 disallowed = clear_mask & ~safe_clear;
5722 if (disallowed) {
5723 names = btrfs_printable_features(set, disallowed);
5724 if (names) {
5725 btrfs_warn(fs_info,
5726 "can't clear the %s feature bit%s while mounted",
5727 names, strchr(names, ',') ? "s" : "");
5728 kfree(names);
5729 } else
5730 btrfs_warn(fs_info,
5731 "can't clear %s bits 0x%llx while mounted",
5732 type, disallowed);
5733 return -EPERM;
5734 }
5735
5736 return 0;
5737 }
5738
5739 #define check_feature(fs_info, change_mask, flags, mask_base) \
5740 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5741 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5742 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5743 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5744
5745 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5746 {
5747 struct inode *inode = file_inode(file);
5748 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5749 struct btrfs_root *root = BTRFS_I(inode)->root;
5750 struct btrfs_super_block *super_block = fs_info->super_copy;
5751 struct btrfs_ioctl_feature_flags flags[2];
5752 struct btrfs_trans_handle *trans;
5753 u64 newflags;
5754 int ret;
5755
5756 if (!capable(CAP_SYS_ADMIN))
5757 return -EPERM;
5758
5759 if (copy_from_user(flags, arg, sizeof(flags)))
5760 return -EFAULT;
5761
5762 /* Nothing to do */
5763 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5764 !flags[0].incompat_flags)
5765 return 0;
5766
5767 ret = check_feature(fs_info, flags[0].compat_flags,
5768 flags[1].compat_flags, COMPAT);
5769 if (ret)
5770 return ret;
5771
5772 ret = check_feature(fs_info, flags[0].compat_ro_flags,
5773 flags[1].compat_ro_flags, COMPAT_RO);
5774 if (ret)
5775 return ret;
5776
5777 ret = check_feature(fs_info, flags[0].incompat_flags,
5778 flags[1].incompat_flags, INCOMPAT);
5779 if (ret)
5780 return ret;
5781
5782 ret = mnt_want_write_file(file);
5783 if (ret)
5784 return ret;
5785
5786 trans = btrfs_start_transaction(root, 0);
5787 if (IS_ERR(trans)) {
5788 ret = PTR_ERR(trans);
5789 goto out_drop_write;
5790 }
5791
5792 spin_lock(&fs_info->super_lock);
5793 newflags = btrfs_super_compat_flags(super_block);
5794 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5795 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5796 btrfs_set_super_compat_flags(super_block, newflags);
5797
5798 newflags = btrfs_super_compat_ro_flags(super_block);
5799 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5800 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5801 btrfs_set_super_compat_ro_flags(super_block, newflags);
5802
5803 newflags = btrfs_super_incompat_flags(super_block);
5804 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5805 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5806 btrfs_set_super_incompat_flags(super_block, newflags);
5807 spin_unlock(&fs_info->super_lock);
5808
5809 ret = btrfs_commit_transaction(trans);
5810 out_drop_write:
5811 mnt_drop_write_file(file);
5812
5813 return ret;
5814 }
5815
5816 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
5817 {
5818 struct btrfs_ioctl_send_args *arg;
5819 int ret;
5820
5821 if (compat) {
5822 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5823 struct btrfs_ioctl_send_args_32 args32;
5824
5825 ret = copy_from_user(&args32, argp, sizeof(args32));
5826 if (ret)
5827 return -EFAULT;
5828 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
5829 if (!arg)
5830 return -ENOMEM;
5831 arg->send_fd = args32.send_fd;
5832 arg->clone_sources_count = args32.clone_sources_count;
5833 arg->clone_sources = compat_ptr(args32.clone_sources);
5834 arg->parent_root = args32.parent_root;
5835 arg->flags = args32.flags;
5836 memcpy(arg->reserved, args32.reserved,
5837 sizeof(args32.reserved));
5838 #else
5839 return -ENOTTY;
5840 #endif
5841 } else {
5842 arg = memdup_user(argp, sizeof(*arg));
5843 if (IS_ERR(arg))
5844 return PTR_ERR(arg);
5845 }
5846 ret = btrfs_ioctl_send(file, arg);
5847 kfree(arg);
5848 return ret;
5849 }
5850
5851 long btrfs_ioctl(struct file *file, unsigned int
5852 cmd, unsigned long arg)
5853 {
5854 struct inode *inode = file_inode(file);
5855 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5856 struct btrfs_root *root = BTRFS_I(inode)->root;
5857 void __user *argp = (void __user *)arg;
5858
5859 switch (cmd) {
5860 case FS_IOC_GETFLAGS:
5861 return btrfs_ioctl_getflags(file, argp);
5862 case FS_IOC_SETFLAGS:
5863 return btrfs_ioctl_setflags(file, argp);
5864 case FS_IOC_GETVERSION:
5865 return btrfs_ioctl_getversion(file, argp);
5866 case FITRIM:
5867 return btrfs_ioctl_fitrim(file, argp);
5868 case BTRFS_IOC_SNAP_CREATE:
5869 return btrfs_ioctl_snap_create(file, argp, 0);
5870 case BTRFS_IOC_SNAP_CREATE_V2:
5871 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5872 case BTRFS_IOC_SUBVOL_CREATE:
5873 return btrfs_ioctl_snap_create(file, argp, 1);
5874 case BTRFS_IOC_SUBVOL_CREATE_V2:
5875 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5876 case BTRFS_IOC_SNAP_DESTROY:
5877 return btrfs_ioctl_snap_destroy(file, argp);
5878 case BTRFS_IOC_SUBVOL_GETFLAGS:
5879 return btrfs_ioctl_subvol_getflags(file, argp);
5880 case BTRFS_IOC_SUBVOL_SETFLAGS:
5881 return btrfs_ioctl_subvol_setflags(file, argp);
5882 case BTRFS_IOC_DEFAULT_SUBVOL:
5883 return btrfs_ioctl_default_subvol(file, argp);
5884 case BTRFS_IOC_DEFRAG:
5885 return btrfs_ioctl_defrag(file, NULL);
5886 case BTRFS_IOC_DEFRAG_RANGE:
5887 return btrfs_ioctl_defrag(file, argp);
5888 case BTRFS_IOC_RESIZE:
5889 return btrfs_ioctl_resize(file, argp);
5890 case BTRFS_IOC_ADD_DEV:
5891 return btrfs_ioctl_add_dev(fs_info, argp);
5892 case BTRFS_IOC_RM_DEV:
5893 return btrfs_ioctl_rm_dev(file, argp);
5894 case BTRFS_IOC_RM_DEV_V2:
5895 return btrfs_ioctl_rm_dev_v2(file, argp);
5896 case BTRFS_IOC_FS_INFO:
5897 return btrfs_ioctl_fs_info(fs_info, argp);
5898 case BTRFS_IOC_DEV_INFO:
5899 return btrfs_ioctl_dev_info(fs_info, argp);
5900 case BTRFS_IOC_BALANCE:
5901 return btrfs_ioctl_balance(file, NULL);
5902 case BTRFS_IOC_TREE_SEARCH:
5903 return btrfs_ioctl_tree_search(file, argp);
5904 case BTRFS_IOC_TREE_SEARCH_V2:
5905 return btrfs_ioctl_tree_search_v2(file, argp);
5906 case BTRFS_IOC_INO_LOOKUP:
5907 return btrfs_ioctl_ino_lookup(file, argp);
5908 case BTRFS_IOC_INO_PATHS:
5909 return btrfs_ioctl_ino_to_path(root, argp);
5910 case BTRFS_IOC_LOGICAL_INO:
5911 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
5912 case BTRFS_IOC_LOGICAL_INO_V2:
5913 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
5914 case BTRFS_IOC_SPACE_INFO:
5915 return btrfs_ioctl_space_info(fs_info, argp);
5916 case BTRFS_IOC_SYNC: {
5917 int ret;
5918
5919 ret = btrfs_start_delalloc_roots(fs_info, -1);
5920 if (ret)
5921 return ret;
5922 ret = btrfs_sync_fs(inode->i_sb, 1);
5923 /*
5924 * The transaction thread may want to do more work,
5925 * namely it pokes the cleaner kthread that will start
5926 * processing uncleaned subvols.
5927 */
5928 wake_up_process(fs_info->transaction_kthread);
5929 return ret;
5930 }
5931 case BTRFS_IOC_START_SYNC:
5932 return btrfs_ioctl_start_sync(root, argp);
5933 case BTRFS_IOC_WAIT_SYNC:
5934 return btrfs_ioctl_wait_sync(fs_info, argp);
5935 case BTRFS_IOC_SCRUB:
5936 return btrfs_ioctl_scrub(file, argp);
5937 case BTRFS_IOC_SCRUB_CANCEL:
5938 return btrfs_ioctl_scrub_cancel(fs_info);
5939 case BTRFS_IOC_SCRUB_PROGRESS:
5940 return btrfs_ioctl_scrub_progress(fs_info, argp);
5941 case BTRFS_IOC_BALANCE_V2:
5942 return btrfs_ioctl_balance(file, argp);
5943 case BTRFS_IOC_BALANCE_CTL:
5944 return btrfs_ioctl_balance_ctl(fs_info, arg);
5945 case BTRFS_IOC_BALANCE_PROGRESS:
5946 return btrfs_ioctl_balance_progress(fs_info, argp);
5947 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5948 return btrfs_ioctl_set_received_subvol(file, argp);
5949 #ifdef CONFIG_64BIT
5950 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5951 return btrfs_ioctl_set_received_subvol_32(file, argp);
5952 #endif
5953 case BTRFS_IOC_SEND:
5954 return _btrfs_ioctl_send(file, argp, false);
5955 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5956 case BTRFS_IOC_SEND_32:
5957 return _btrfs_ioctl_send(file, argp, true);
5958 #endif
5959 case BTRFS_IOC_GET_DEV_STATS:
5960 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5961 case BTRFS_IOC_QUOTA_CTL:
5962 return btrfs_ioctl_quota_ctl(file, argp);
5963 case BTRFS_IOC_QGROUP_ASSIGN:
5964 return btrfs_ioctl_qgroup_assign(file, argp);
5965 case BTRFS_IOC_QGROUP_CREATE:
5966 return btrfs_ioctl_qgroup_create(file, argp);
5967 case BTRFS_IOC_QGROUP_LIMIT:
5968 return btrfs_ioctl_qgroup_limit(file, argp);
5969 case BTRFS_IOC_QUOTA_RESCAN:
5970 return btrfs_ioctl_quota_rescan(file, argp);
5971 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5972 return btrfs_ioctl_quota_rescan_status(file, argp);
5973 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5974 return btrfs_ioctl_quota_rescan_wait(file, argp);
5975 case BTRFS_IOC_DEV_REPLACE:
5976 return btrfs_ioctl_dev_replace(fs_info, argp);
5977 case BTRFS_IOC_GET_FSLABEL:
5978 return btrfs_ioctl_get_fslabel(file, argp);
5979 case BTRFS_IOC_SET_FSLABEL:
5980 return btrfs_ioctl_set_fslabel(file, argp);
5981 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5982 return btrfs_ioctl_get_supported_features(argp);
5983 case BTRFS_IOC_GET_FEATURES:
5984 return btrfs_ioctl_get_features(file, argp);
5985 case BTRFS_IOC_SET_FEATURES:
5986 return btrfs_ioctl_set_features(file, argp);
5987 case FS_IOC_FSGETXATTR:
5988 return btrfs_ioctl_fsgetxattr(file, argp);
5989 case FS_IOC_FSSETXATTR:
5990 return btrfs_ioctl_fssetxattr(file, argp);
5991 case BTRFS_IOC_GET_SUBVOL_INFO:
5992 return btrfs_ioctl_get_subvol_info(file, argp);
5993 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5994 return btrfs_ioctl_get_subvol_rootref(file, argp);
5995 case BTRFS_IOC_INO_LOOKUP_USER:
5996 return btrfs_ioctl_ino_lookup_user(file, argp);
5997 }
5998
5999 return -ENOTTY;
6000 }
6001
6002 #ifdef CONFIG_COMPAT
6003 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
6004 {
6005 /*
6006 * These all access 32-bit values anyway so no further
6007 * handling is necessary.
6008 */
6009 switch (cmd) {
6010 case FS_IOC32_GETFLAGS:
6011 cmd = FS_IOC_GETFLAGS;
6012 break;
6013 case FS_IOC32_SETFLAGS:
6014 cmd = FS_IOC_SETFLAGS;
6015 break;
6016 case FS_IOC32_GETVERSION:
6017 cmd = FS_IOC_GETVERSION;
6018 break;
6019 }
6020
6021 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
6022 }
6023 #endif
Linux with added FPC-III support