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