projects / GitHub / moto-9609 / android_kernel_motorola_exynos9610.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'for-4.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave...
[GitHub/moto-9609/android_kernel_motorola_exynos9610.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 #include "tree-log.h"
63 #include "compression.h"
64
65 #ifdef CONFIG_64BIT
66 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
67 * structures are incorrect, as the timespec structure from userspace
68 * is 4 bytes too small. We define these alternatives here to teach
69 * the kernel about the 32-bit struct packing.
70 */
71 struct btrfs_ioctl_timespec_32 {
72 __u64 sec;
73 __u32 nsec;
74 } __attribute__ ((__packed__));
75
76 struct btrfs_ioctl_received_subvol_args_32 {
77 char uuid[BTRFS_UUID_SIZE]; /* in */
78 __u64 stransid; /* in */
79 __u64 rtransid; /* out */
80 struct btrfs_ioctl_timespec_32 stime; /* in */
81 struct btrfs_ioctl_timespec_32 rtime; /* out */
82 __u64 flags; /* in */
83 __u64 reserved[16]; /* in */
84 } __attribute__ ((__packed__));
85
86 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
87 struct btrfs_ioctl_received_subvol_args_32)
88 #endif
89
90
91 static int btrfs_clone(struct inode *src, struct inode *inode,
92 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
93 int no_time_update);
94
95 /* Mask out flags that are inappropriate for the given type of inode. */
96 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
97 {
98 if (S_ISDIR(mode))
99 return flags;
100 else if (S_ISREG(mode))
101 return flags & ~FS_DIRSYNC_FL;
102 else
103 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
104 }
105
106 /*
107 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
108 */
109 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
110 {
111 unsigned int iflags = 0;
112
113 if (flags & BTRFS_INODE_SYNC)
114 iflags |= FS_SYNC_FL;
115 if (flags & BTRFS_INODE_IMMUTABLE)
116 iflags |= FS_IMMUTABLE_FL;
117 if (flags & BTRFS_INODE_APPEND)
118 iflags |= FS_APPEND_FL;
119 if (flags & BTRFS_INODE_NODUMP)
120 iflags |= FS_NODUMP_FL;
121 if (flags & BTRFS_INODE_NOATIME)
122 iflags |= FS_NOATIME_FL;
123 if (flags & BTRFS_INODE_DIRSYNC)
124 iflags |= FS_DIRSYNC_FL;
125 if (flags & BTRFS_INODE_NODATACOW)
126 iflags |= FS_NOCOW_FL;
127
128 if (flags & BTRFS_INODE_NOCOMPRESS)
129 iflags |= FS_NOCOMP_FL;
130 else if (flags & BTRFS_INODE_COMPRESS)
131 iflags |= FS_COMPR_FL;
132
133 return iflags;
134 }
135
136 /*
137 * Update inode->i_flags based on the btrfs internal flags.
138 */
139 void btrfs_update_iflags(struct inode *inode)
140 {
141 struct btrfs_inode *ip = BTRFS_I(inode);
142 unsigned int new_fl = 0;
143
144 if (ip->flags & BTRFS_INODE_SYNC)
145 new_fl |= S_SYNC;
146 if (ip->flags & BTRFS_INODE_IMMUTABLE)
147 new_fl |= S_IMMUTABLE;
148 if (ip->flags & BTRFS_INODE_APPEND)
149 new_fl |= S_APPEND;
150 if (ip->flags & BTRFS_INODE_NOATIME)
151 new_fl |= S_NOATIME;
152 if (ip->flags & BTRFS_INODE_DIRSYNC)
153 new_fl |= S_DIRSYNC;
154
155 set_mask_bits(&inode->i_flags,
156 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
157 new_fl);
158 }
159
160 /*
161 * Inherit flags from the parent inode.
162 *
163 * Currently only the compression flags and the cow flags are inherited.
164 */
165 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
166 {
167 unsigned int flags;
168
169 if (!dir)
170 return;
171
172 flags = BTRFS_I(dir)->flags;
173
174 if (flags & BTRFS_INODE_NOCOMPRESS) {
175 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
176 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
177 } else if (flags & BTRFS_INODE_COMPRESS) {
178 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
179 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
180 }
181
182 if (flags & BTRFS_INODE_NODATACOW) {
183 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
184 if (S_ISREG(inode->i_mode))
185 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
186 }
187
188 btrfs_update_iflags(inode);
189 }
190
191 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
192 {
193 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
194 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
195
196 if (copy_to_user(arg, &flags, sizeof(flags)))
197 return -EFAULT;
198 return 0;
199 }
200
201 static int check_flags(unsigned int flags)
202 {
203 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
204 FS_NOATIME_FL | FS_NODUMP_FL | \
205 FS_SYNC_FL | FS_DIRSYNC_FL | \
206 FS_NOCOMP_FL | FS_COMPR_FL |
207 FS_NOCOW_FL))
208 return -EOPNOTSUPP;
209
210 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
211 return -EINVAL;
212
213 return 0;
214 }
215
216 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
217 {
218 struct inode *inode = file_inode(file);
219 struct btrfs_inode *ip = BTRFS_I(inode);
220 struct btrfs_root *root = ip->root;
221 struct btrfs_trans_handle *trans;
222 unsigned int flags, oldflags;
223 int ret;
224 u64 ip_oldflags;
225 unsigned int i_oldflags;
226 umode_t mode;
227
228 if (!inode_owner_or_capable(inode))
229 return -EPERM;
230
231 if (btrfs_root_readonly(root))
232 return -EROFS;
233
234 if (copy_from_user(&flags, arg, sizeof(flags)))
235 return -EFAULT;
236
237 ret = check_flags(flags);
238 if (ret)
239 return ret;
240
241 ret = mnt_want_write_file(file);
242 if (ret)
243 return ret;
244
245 inode_lock(inode);
246
247 ip_oldflags = ip->flags;
248 i_oldflags = inode->i_flags;
249 mode = inode->i_mode;
250
251 flags = btrfs_mask_flags(inode->i_mode, flags);
252 oldflags = btrfs_flags_to_ioctl(ip->flags);
253 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
254 if (!capable(CAP_LINUX_IMMUTABLE)) {
255 ret = -EPERM;
256 goto out_unlock;
257 }
258 }
259
260 if (flags & FS_SYNC_FL)
261 ip->flags |= BTRFS_INODE_SYNC;
262 else
263 ip->flags &= ~BTRFS_INODE_SYNC;
264 if (flags & FS_IMMUTABLE_FL)
265 ip->flags |= BTRFS_INODE_IMMUTABLE;
266 else
267 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
268 if (flags & FS_APPEND_FL)
269 ip->flags |= BTRFS_INODE_APPEND;
270 else
271 ip->flags &= ~BTRFS_INODE_APPEND;
272 if (flags & FS_NODUMP_FL)
273 ip->flags |= BTRFS_INODE_NODUMP;
274 else
275 ip->flags &= ~BTRFS_INODE_NODUMP;
276 if (flags & FS_NOATIME_FL)
277 ip->flags |= BTRFS_INODE_NOATIME;
278 else
279 ip->flags &= ~BTRFS_INODE_NOATIME;
280 if (flags & FS_DIRSYNC_FL)
281 ip->flags |= BTRFS_INODE_DIRSYNC;
282 else
283 ip->flags &= ~BTRFS_INODE_DIRSYNC;
284 if (flags & FS_NOCOW_FL) {
285 if (S_ISREG(mode)) {
286 /*
287 * It's safe to turn csums off here, no extents exist.
288 * Otherwise we want the flag to reflect the real COW
289 * status of the file and will not set it.
290 */
291 if (inode->i_size == 0)
292 ip->flags |= BTRFS_INODE_NODATACOW
293 | BTRFS_INODE_NODATASUM;
294 } else {
295 ip->flags |= BTRFS_INODE_NODATACOW;
296 }
297 } else {
298 /*
299 * Revert back under same assumptions as above
300 */
301 if (S_ISREG(mode)) {
302 if (inode->i_size == 0)
303 ip->flags &= ~(BTRFS_INODE_NODATACOW
304 | BTRFS_INODE_NODATASUM);
305 } else {
306 ip->flags &= ~BTRFS_INODE_NODATACOW;
307 }
308 }
309
310 /*
311 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
312 * flag may be changed automatically if compression code won't make
313 * things smaller.
314 */
315 if (flags & FS_NOCOMP_FL) {
316 ip->flags &= ~BTRFS_INODE_COMPRESS;
317 ip->flags |= BTRFS_INODE_NOCOMPRESS;
318
319 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
320 if (ret && ret != -ENODATA)
321 goto out_drop;
322 } else if (flags & FS_COMPR_FL) {
323 const char *comp;
324
325 ip->flags |= BTRFS_INODE_COMPRESS;
326 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
327
328 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
329 comp = "lzo";
330 else
331 comp = "zlib";
332 ret = btrfs_set_prop(inode, "btrfs.compression",
333 comp, strlen(comp), 0);
334 if (ret)
335 goto out_drop;
336
337 } else {
338 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
339 if (ret && ret != -ENODATA)
340 goto out_drop;
341 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
342 }
343
344 trans = btrfs_start_transaction(root, 1);
345 if (IS_ERR(trans)) {
346 ret = PTR_ERR(trans);
347 goto out_drop;
348 }
349
350 btrfs_update_iflags(inode);
351 inode_inc_iversion(inode);
352 inode->i_ctime = current_time(inode);
353 ret = btrfs_update_inode(trans, root, inode);
354
355 btrfs_end_transaction(trans, root);
356 out_drop:
357 if (ret) {
358 ip->flags = ip_oldflags;
359 inode->i_flags = i_oldflags;
360 }
361
362 out_unlock:
363 inode_unlock(inode);
364 mnt_drop_write_file(file);
365 return ret;
366 }
367
368 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
369 {
370 struct inode *inode = file_inode(file);
371
372 return put_user(inode->i_generation, arg);
373 }
374
375 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
376 {
377 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
378 struct btrfs_device *device;
379 struct request_queue *q;
380 struct fstrim_range range;
381 u64 minlen = ULLONG_MAX;
382 u64 num_devices = 0;
383 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
384 int ret;
385
386 if (!capable(CAP_SYS_ADMIN))
387 return -EPERM;
388
389 rcu_read_lock();
390 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
391 dev_list) {
392 if (!device->bdev)
393 continue;
394 q = bdev_get_queue(device->bdev);
395 if (blk_queue_discard(q)) {
396 num_devices++;
397 minlen = min((u64)q->limits.discard_granularity,
398 minlen);
399 }
400 }
401 rcu_read_unlock();
402
403 if (!num_devices)
404 return -EOPNOTSUPP;
405 if (copy_from_user(&range, arg, sizeof(range)))
406 return -EFAULT;
407 if (range.start > total_bytes ||
408 range.len < fs_info->sb->s_blocksize)
409 return -EINVAL;
410
411 range.len = min(range.len, total_bytes - range.start);
412 range.minlen = max(range.minlen, minlen);
413 ret = btrfs_trim_fs(fs_info->tree_root, &range);
414 if (ret < 0)
415 return ret;
416
417 if (copy_to_user(arg, &range, sizeof(range)))
418 return -EFAULT;
419
420 return 0;
421 }
422
423 int btrfs_is_empty_uuid(u8 *uuid)
424 {
425 int i;
426
427 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
428 if (uuid[i])
429 return 0;
430 }
431 return 1;
432 }
433
434 static noinline int create_subvol(struct inode *dir,
435 struct dentry *dentry,
436 char *name, int namelen,
437 u64 *async_transid,
438 struct btrfs_qgroup_inherit *inherit)
439 {
440 struct btrfs_trans_handle *trans;
441 struct btrfs_key key;
442 struct btrfs_root_item *root_item;
443 struct btrfs_inode_item *inode_item;
444 struct extent_buffer *leaf;
445 struct btrfs_root *root = BTRFS_I(dir)->root;
446 struct btrfs_root *new_root;
447 struct btrfs_block_rsv block_rsv;
448 struct timespec cur_time = current_time(dir);
449 struct inode *inode;
450 int ret;
451 int err;
452 u64 objectid;
453 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
454 u64 index = 0;
455 u64 qgroup_reserved;
456 uuid_le new_uuid;
457
458 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
459 if (!root_item)
460 return -ENOMEM;
461
462 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
463 if (ret)
464 goto fail_free;
465
466 /*
467 * Don't create subvolume whose level is not zero. Or qgroup will be
468 * screwed up since it assumes subvolume qgroup's level to be 0.
469 */
470 if (btrfs_qgroup_level(objectid)) {
471 ret = -ENOSPC;
472 goto fail_free;
473 }
474
475 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
476 /*
477 * The same as the snapshot creation, please see the comment
478 * of create_snapshot().
479 */
480 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
481 8, &qgroup_reserved, false);
482 if (ret)
483 goto fail_free;
484
485 trans = btrfs_start_transaction(root, 0);
486 if (IS_ERR(trans)) {
487 ret = PTR_ERR(trans);
488 btrfs_subvolume_release_metadata(root, &block_rsv,
489 qgroup_reserved);
490 goto fail_free;
491 }
492 trans->block_rsv = &block_rsv;
493 trans->bytes_reserved = block_rsv.size;
494
495 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
496 if (ret)
497 goto fail;
498
499 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
500 if (IS_ERR(leaf)) {
501 ret = PTR_ERR(leaf);
502 goto fail;
503 }
504
505 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
506 btrfs_set_header_bytenr(leaf, leaf->start);
507 btrfs_set_header_generation(leaf, trans->transid);
508 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
509 btrfs_set_header_owner(leaf, objectid);
510
511 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
512 BTRFS_FSID_SIZE);
513 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
514 btrfs_header_chunk_tree_uuid(leaf),
515 BTRFS_UUID_SIZE);
516 btrfs_mark_buffer_dirty(leaf);
517
518 inode_item = &root_item->inode;
519 btrfs_set_stack_inode_generation(inode_item, 1);
520 btrfs_set_stack_inode_size(inode_item, 3);
521 btrfs_set_stack_inode_nlink(inode_item, 1);
522 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
523 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
524
525 btrfs_set_root_flags(root_item, 0);
526 btrfs_set_root_limit(root_item, 0);
527 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
528
529 btrfs_set_root_bytenr(root_item, leaf->start);
530 btrfs_set_root_generation(root_item, trans->transid);
531 btrfs_set_root_level(root_item, 0);
532 btrfs_set_root_refs(root_item, 1);
533 btrfs_set_root_used(root_item, leaf->len);
534 btrfs_set_root_last_snapshot(root_item, 0);
535
536 btrfs_set_root_generation_v2(root_item,
537 btrfs_root_generation(root_item));
538 uuid_le_gen(&new_uuid);
539 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
540 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
541 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
542 root_item->ctime = root_item->otime;
543 btrfs_set_root_ctransid(root_item, trans->transid);
544 btrfs_set_root_otransid(root_item, trans->transid);
545
546 btrfs_tree_unlock(leaf);
547 free_extent_buffer(leaf);
548 leaf = NULL;
549
550 btrfs_set_root_dirid(root_item, new_dirid);
551
552 key.objectid = objectid;
553 key.offset = 0;
554 key.type = BTRFS_ROOT_ITEM_KEY;
555 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
556 root_item);
557 if (ret)
558 goto fail;
559
560 key.offset = (u64)-1;
561 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
562 if (IS_ERR(new_root)) {
563 ret = PTR_ERR(new_root);
564 btrfs_abort_transaction(trans, ret);
565 goto fail;
566 }
567
568 btrfs_record_root_in_trans(trans, new_root);
569
570 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
571 if (ret) {
572 /* We potentially lose an unused inode item here */
573 btrfs_abort_transaction(trans, ret);
574 goto fail;
575 }
576
577 mutex_lock(&new_root->objectid_mutex);
578 new_root->highest_objectid = new_dirid;
579 mutex_unlock(&new_root->objectid_mutex);
580
581 /*
582 * insert the directory item
583 */
584 ret = btrfs_set_inode_index(dir, &index);
585 if (ret) {
586 btrfs_abort_transaction(trans, ret);
587 goto fail;
588 }
589
590 ret = btrfs_insert_dir_item(trans, root,
591 name, namelen, dir, &key,
592 BTRFS_FT_DIR, index);
593 if (ret) {
594 btrfs_abort_transaction(trans, ret);
595 goto fail;
596 }
597
598 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
599 ret = btrfs_update_inode(trans, root, dir);
600 BUG_ON(ret);
601
602 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
603 objectid, root->root_key.objectid,
604 btrfs_ino(dir), index, name, namelen);
605 BUG_ON(ret);
606
607 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
608 root_item->uuid, BTRFS_UUID_KEY_SUBVOL,
609 objectid);
610 if (ret)
611 btrfs_abort_transaction(trans, ret);
612
613 fail:
614 kfree(root_item);
615 trans->block_rsv = NULL;
616 trans->bytes_reserved = 0;
617 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
618
619 if (async_transid) {
620 *async_transid = trans->transid;
621 err = btrfs_commit_transaction_async(trans, root, 1);
622 if (err)
623 err = btrfs_commit_transaction(trans, root);
624 } else {
625 err = btrfs_commit_transaction(trans, root);
626 }
627 if (err && !ret)
628 ret = err;
629
630 if (!ret) {
631 inode = btrfs_lookup_dentry(dir, dentry);
632 if (IS_ERR(inode))
633 return PTR_ERR(inode);
634 d_instantiate(dentry, inode);
635 }
636 return ret;
637
638 fail_free:
639 kfree(root_item);
640 return ret;
641 }
642
643 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
644 {
645 s64 writers;
646 DEFINE_WAIT(wait);
647
648 do {
649 prepare_to_wait(&root->subv_writers->wait, &wait,
650 TASK_UNINTERRUPTIBLE);
651
652 writers = percpu_counter_sum(&root->subv_writers->counter);
653 if (writers)
654 schedule();
655
656 finish_wait(&root->subv_writers->wait, &wait);
657 } while (writers);
658 }
659
660 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
661 struct dentry *dentry, char *name, int namelen,
662 u64 *async_transid, bool readonly,
663 struct btrfs_qgroup_inherit *inherit)
664 {
665 struct inode *inode;
666 struct btrfs_pending_snapshot *pending_snapshot;
667 struct btrfs_trans_handle *trans;
668 int ret;
669
670 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
671 return -EINVAL;
672
673 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
674 if (!pending_snapshot)
675 return -ENOMEM;
676
677 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
678 GFP_NOFS);
679 pending_snapshot->path = btrfs_alloc_path();
680 if (!pending_snapshot->root_item || !pending_snapshot->path) {
681 ret = -ENOMEM;
682 goto free_pending;
683 }
684
685 atomic_inc(&root->will_be_snapshoted);
686 smp_mb__after_atomic();
687 btrfs_wait_for_no_snapshoting_writes(root);
688
689 ret = btrfs_start_delalloc_inodes(root, 0);
690 if (ret)
691 goto dec_and_free;
692
693 btrfs_wait_ordered_extents(root, -1, 0, (u64)-1);
694
695 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
696 BTRFS_BLOCK_RSV_TEMP);
697 /*
698 * 1 - parent dir inode
699 * 2 - dir entries
700 * 1 - root item
701 * 2 - root ref/backref
702 * 1 - root of snapshot
703 * 1 - UUID item
704 */
705 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
706 &pending_snapshot->block_rsv, 8,
707 &pending_snapshot->qgroup_reserved,
708 false);
709 if (ret)
710 goto dec_and_free;
711
712 pending_snapshot->dentry = dentry;
713 pending_snapshot->root = root;
714 pending_snapshot->readonly = readonly;
715 pending_snapshot->dir = dir;
716 pending_snapshot->inherit = inherit;
717
718 trans = btrfs_start_transaction(root, 0);
719 if (IS_ERR(trans)) {
720 ret = PTR_ERR(trans);
721 goto fail;
722 }
723
724 spin_lock(&root->fs_info->trans_lock);
725 list_add(&pending_snapshot->list,
726 &trans->transaction->pending_snapshots);
727 spin_unlock(&root->fs_info->trans_lock);
728 if (async_transid) {
729 *async_transid = trans->transid;
730 ret = btrfs_commit_transaction_async(trans,
731 root->fs_info->extent_root, 1);
732 if (ret)
733 ret = btrfs_commit_transaction(trans, root);
734 } else {
735 ret = btrfs_commit_transaction(trans,
736 root->fs_info->extent_root);
737 }
738 if (ret)
739 goto fail;
740
741 ret = pending_snapshot->error;
742 if (ret)
743 goto fail;
744
745 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
746 if (ret)
747 goto fail;
748
749 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
750 if (IS_ERR(inode)) {
751 ret = PTR_ERR(inode);
752 goto fail;
753 }
754
755 d_instantiate(dentry, inode);
756 ret = 0;
757 fail:
758 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
759 &pending_snapshot->block_rsv,
760 pending_snapshot->qgroup_reserved);
761 dec_and_free:
762 if (atomic_dec_and_test(&root->will_be_snapshoted))
763 wake_up_atomic_t(&root->will_be_snapshoted);
764 free_pending:
765 kfree(pending_snapshot->root_item);
766 btrfs_free_path(pending_snapshot->path);
767 kfree(pending_snapshot);
768
769 return ret;
770 }
771
772 /* copy of may_delete in fs/namei.c()
773 * Check whether we can remove a link victim from directory dir, check
774 * whether the type of victim is right.
775 * 1. We can't do it if dir is read-only (done in permission())
776 * 2. We should have write and exec permissions on dir
777 * 3. We can't remove anything from append-only dir
778 * 4. We can't do anything with immutable dir (done in permission())
779 * 5. If the sticky bit on dir is set we should either
780 * a. be owner of dir, or
781 * b. be owner of victim, or
782 * c. have CAP_FOWNER capability
783 * 6. If the victim is append-only or immutable we can't do anything with
784 * links pointing to it.
785 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
786 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
787 * 9. We can't remove a root or mountpoint.
788 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
789 * nfs_async_unlink().
790 */
791
792 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
793 {
794 int error;
795
796 if (d_really_is_negative(victim))
797 return -ENOENT;
798
799 BUG_ON(d_inode(victim->d_parent) != dir);
800 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
801
802 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
803 if (error)
804 return error;
805 if (IS_APPEND(dir))
806 return -EPERM;
807 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
808 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
809 return -EPERM;
810 if (isdir) {
811 if (!d_is_dir(victim))
812 return -ENOTDIR;
813 if (IS_ROOT(victim))
814 return -EBUSY;
815 } else if (d_is_dir(victim))
816 return -EISDIR;
817 if (IS_DEADDIR(dir))
818 return -ENOENT;
819 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
820 return -EBUSY;
821 return 0;
822 }
823
824 /* copy of may_create in fs/namei.c() */
825 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
826 {
827 if (d_really_is_positive(child))
828 return -EEXIST;
829 if (IS_DEADDIR(dir))
830 return -ENOENT;
831 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
832 }
833
834 /*
835 * Create a new subvolume below @parent. This is largely modeled after
836 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
837 * inside this filesystem so it's quite a bit simpler.
838 */
839 static noinline int btrfs_mksubvol(struct path *parent,
840 char *name, int namelen,
841 struct btrfs_root *snap_src,
842 u64 *async_transid, bool readonly,
843 struct btrfs_qgroup_inherit *inherit)
844 {
845 struct inode *dir = d_inode(parent->dentry);
846 struct dentry *dentry;
847 int error;
848
849 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
850 if (error == -EINTR)
851 return error;
852
853 dentry = lookup_one_len(name, parent->dentry, namelen);
854 error = PTR_ERR(dentry);
855 if (IS_ERR(dentry))
856 goto out_unlock;
857
858 error = btrfs_may_create(dir, dentry);
859 if (error)
860 goto out_dput;
861
862 /*
863 * even if this name doesn't exist, we may get hash collisions.
864 * check for them now when we can safely fail
865 */
866 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
867 dir->i_ino, name,
868 namelen);
869 if (error)
870 goto out_dput;
871
872 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
873
874 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
875 goto out_up_read;
876
877 if (snap_src) {
878 error = create_snapshot(snap_src, dir, dentry, name, namelen,
879 async_transid, readonly, inherit);
880 } else {
881 error = create_subvol(dir, dentry, name, namelen,
882 async_transid, inherit);
883 }
884 if (!error)
885 fsnotify_mkdir(dir, dentry);
886 out_up_read:
887 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
888 out_dput:
889 dput(dentry);
890 out_unlock:
891 inode_unlock(dir);
892 return error;
893 }
894
895 /*
896 * When we're defragging a range, we don't want to kick it off again
897 * if it is really just waiting for delalloc to send it down.
898 * If we find a nice big extent or delalloc range for the bytes in the
899 * file you want to defrag, we return 0 to let you know to skip this
900 * part of the file
901 */
902 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
903 {
904 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
905 struct extent_map *em = NULL;
906 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
907 u64 end;
908
909 read_lock(&em_tree->lock);
910 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
911 read_unlock(&em_tree->lock);
912
913 if (em) {
914 end = extent_map_end(em);
915 free_extent_map(em);
916 if (end - offset > thresh)
917 return 0;
918 }
919 /* if we already have a nice delalloc here, just stop */
920 thresh /= 2;
921 end = count_range_bits(io_tree, &offset, offset + thresh,
922 thresh, EXTENT_DELALLOC, 1);
923 if (end >= thresh)
924 return 0;
925 return 1;
926 }
927
928 /*
929 * helper function to walk through a file and find extents
930 * newer than a specific transid, and smaller than thresh.
931 *
932 * This is used by the defragging code to find new and small
933 * extents
934 */
935 static int find_new_extents(struct btrfs_root *root,
936 struct inode *inode, u64 newer_than,
937 u64 *off, u32 thresh)
938 {
939 struct btrfs_path *path;
940 struct btrfs_key min_key;
941 struct extent_buffer *leaf;
942 struct btrfs_file_extent_item *extent;
943 int type;
944 int ret;
945 u64 ino = btrfs_ino(inode);
946
947 path = btrfs_alloc_path();
948 if (!path)
949 return -ENOMEM;
950
951 min_key.objectid = ino;
952 min_key.type = BTRFS_EXTENT_DATA_KEY;
953 min_key.offset = *off;
954
955 while (1) {
956 ret = btrfs_search_forward(root, &min_key, path, newer_than);
957 if (ret != 0)
958 goto none;
959 process_slot:
960 if (min_key.objectid != ino)
961 goto none;
962 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
963 goto none;
964
965 leaf = path->nodes[0];
966 extent = btrfs_item_ptr(leaf, path->slots[0],
967 struct btrfs_file_extent_item);
968
969 type = btrfs_file_extent_type(leaf, extent);
970 if (type == BTRFS_FILE_EXTENT_REG &&
971 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
972 check_defrag_in_cache(inode, min_key.offset, thresh)) {
973 *off = min_key.offset;
974 btrfs_free_path(path);
975 return 0;
976 }
977
978 path->slots[0]++;
979 if (path->slots[0] < btrfs_header_nritems(leaf)) {
980 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
981 goto process_slot;
982 }
983
984 if (min_key.offset == (u64)-1)
985 goto none;
986
987 min_key.offset++;
988 btrfs_release_path(path);
989 }
990 none:
991 btrfs_free_path(path);
992 return -ENOENT;
993 }
994
995 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
996 {
997 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
998 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
999 struct extent_map *em;
1000 u64 len = PAGE_SIZE;
1001
1002 /*
1003 * hopefully we have this extent in the tree already, try without
1004 * the full extent lock
1005 */
1006 read_lock(&em_tree->lock);
1007 em = lookup_extent_mapping(em_tree, start, len);
1008 read_unlock(&em_tree->lock);
1009
1010 if (!em) {
1011 struct extent_state *cached = NULL;
1012 u64 end = start + len - 1;
1013
1014 /* get the big lock and read metadata off disk */
1015 lock_extent_bits(io_tree, start, end, &cached);
1016 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1017 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1018
1019 if (IS_ERR(em))
1020 return NULL;
1021 }
1022
1023 return em;
1024 }
1025
1026 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1027 {
1028 struct extent_map *next;
1029 bool ret = true;
1030
1031 /* this is the last extent */
1032 if (em->start + em->len >= i_size_read(inode))
1033 return false;
1034
1035 next = defrag_lookup_extent(inode, em->start + em->len);
1036 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1037 ret = false;
1038 else if ((em->block_start + em->block_len == next->block_start) &&
1039 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1040 ret = false;
1041
1042 free_extent_map(next);
1043 return ret;
1044 }
1045
1046 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1047 u64 *last_len, u64 *skip, u64 *defrag_end,
1048 int compress)
1049 {
1050 struct extent_map *em;
1051 int ret = 1;
1052 bool next_mergeable = true;
1053 bool prev_mergeable = true;
1054
1055 /*
1056 * make sure that once we start defragging an extent, we keep on
1057 * defragging it
1058 */
1059 if (start < *defrag_end)
1060 return 1;
1061
1062 *skip = 0;
1063
1064 em = defrag_lookup_extent(inode, start);
1065 if (!em)
1066 return 0;
1067
1068 /* this will cover holes, and inline extents */
1069 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1070 ret = 0;
1071 goto out;
1072 }
1073
1074 if (!*defrag_end)
1075 prev_mergeable = false;
1076
1077 next_mergeable = defrag_check_next_extent(inode, em);
1078 /*
1079 * we hit a real extent, if it is big or the next extent is not a
1080 * real extent, don't bother defragging it
1081 */
1082 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1083 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1084 ret = 0;
1085 out:
1086 /*
1087 * last_len ends up being a counter of how many bytes we've defragged.
1088 * every time we choose not to defrag an extent, we reset *last_len
1089 * so that the next tiny extent will force a defrag.
1090 *
1091 * The end result of this is that tiny extents before a single big
1092 * extent will force at least part of that big extent to be defragged.
1093 */
1094 if (ret) {
1095 *defrag_end = extent_map_end(em);
1096 } else {
1097 *last_len = 0;
1098 *skip = extent_map_end(em);
1099 *defrag_end = 0;
1100 }
1101
1102 free_extent_map(em);
1103 return ret;
1104 }
1105
1106 /*
1107 * it doesn't do much good to defrag one or two pages
1108 * at a time. This pulls in a nice chunk of pages
1109 * to COW and defrag.
1110 *
1111 * It also makes sure the delalloc code has enough
1112 * dirty data to avoid making new small extents as part
1113 * of the defrag
1114 *
1115 * It's a good idea to start RA on this range
1116 * before calling this.
1117 */
1118 static int cluster_pages_for_defrag(struct inode *inode,
1119 struct page **pages,
1120 unsigned long start_index,
1121 unsigned long num_pages)
1122 {
1123 unsigned long file_end;
1124 u64 isize = i_size_read(inode);
1125 u64 page_start;
1126 u64 page_end;
1127 u64 page_cnt;
1128 int ret;
1129 int i;
1130 int i_done;
1131 struct btrfs_ordered_extent *ordered;
1132 struct extent_state *cached_state = NULL;
1133 struct extent_io_tree *tree;
1134 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1135
1136 file_end = (isize - 1) >> PAGE_SHIFT;
1137 if (!isize || start_index > file_end)
1138 return 0;
1139
1140 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1141
1142 ret = btrfs_delalloc_reserve_space(inode,
1143 start_index << PAGE_SHIFT,
1144 page_cnt << PAGE_SHIFT);
1145 if (ret)
1146 return ret;
1147 i_done = 0;
1148 tree = &BTRFS_I(inode)->io_tree;
1149
1150 /* step one, lock all the pages */
1151 for (i = 0; i < page_cnt; i++) {
1152 struct page *page;
1153 again:
1154 page = find_or_create_page(inode->i_mapping,
1155 start_index + i, mask);
1156 if (!page)
1157 break;
1158
1159 page_start = page_offset(page);
1160 page_end = page_start + PAGE_SIZE - 1;
1161 while (1) {
1162 lock_extent_bits(tree, page_start, page_end,
1163 &cached_state);
1164 ordered = btrfs_lookup_ordered_extent(inode,
1165 page_start);
1166 unlock_extent_cached(tree, page_start, page_end,
1167 &cached_state, GFP_NOFS);
1168 if (!ordered)
1169 break;
1170
1171 unlock_page(page);
1172 btrfs_start_ordered_extent(inode, ordered, 1);
1173 btrfs_put_ordered_extent(ordered);
1174 lock_page(page);
1175 /*
1176 * we unlocked the page above, so we need check if
1177 * it was released or not.
1178 */
1179 if (page->mapping != inode->i_mapping) {
1180 unlock_page(page);
1181 put_page(page);
1182 goto again;
1183 }
1184 }
1185
1186 if (!PageUptodate(page)) {
1187 btrfs_readpage(NULL, page);
1188 lock_page(page);
1189 if (!PageUptodate(page)) {
1190 unlock_page(page);
1191 put_page(page);
1192 ret = -EIO;
1193 break;
1194 }
1195 }
1196
1197 if (page->mapping != inode->i_mapping) {
1198 unlock_page(page);
1199 put_page(page);
1200 goto again;
1201 }
1202
1203 pages[i] = page;
1204 i_done++;
1205 }
1206 if (!i_done || ret)
1207 goto out;
1208
1209 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1210 goto out;
1211
1212 /*
1213 * so now we have a nice long stream of locked
1214 * and up to date pages, lets wait on them
1215 */
1216 for (i = 0; i < i_done; i++)
1217 wait_on_page_writeback(pages[i]);
1218
1219 page_start = page_offset(pages[0]);
1220 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1221
1222 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1223 page_start, page_end - 1, &cached_state);
1224 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1225 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1226 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1227 &cached_state, GFP_NOFS);
1228
1229 if (i_done != page_cnt) {
1230 spin_lock(&BTRFS_I(inode)->lock);
1231 BTRFS_I(inode)->outstanding_extents++;
1232 spin_unlock(&BTRFS_I(inode)->lock);
1233 btrfs_delalloc_release_space(inode,
1234 start_index << PAGE_SHIFT,
1235 (page_cnt - i_done) << PAGE_SHIFT);
1236 }
1237
1238
1239 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1240 &cached_state);
1241
1242 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1243 page_start, page_end - 1, &cached_state,
1244 GFP_NOFS);
1245
1246 for (i = 0; i < i_done; i++) {
1247 clear_page_dirty_for_io(pages[i]);
1248 ClearPageChecked(pages[i]);
1249 set_page_extent_mapped(pages[i]);
1250 set_page_dirty(pages[i]);
1251 unlock_page(pages[i]);
1252 put_page(pages[i]);
1253 }
1254 return i_done;
1255 out:
1256 for (i = 0; i < i_done; i++) {
1257 unlock_page(pages[i]);
1258 put_page(pages[i]);
1259 }
1260 btrfs_delalloc_release_space(inode,
1261 start_index << PAGE_SHIFT,
1262 page_cnt << PAGE_SHIFT);
1263 return ret;
1264
1265 }
1266
1267 int btrfs_defrag_file(struct inode *inode, struct file *file,
1268 struct btrfs_ioctl_defrag_range_args *range,
1269 u64 newer_than, unsigned long max_to_defrag)
1270 {
1271 struct btrfs_root *root = BTRFS_I(inode)->root;
1272 struct file_ra_state *ra = NULL;
1273 unsigned long last_index;
1274 u64 isize = i_size_read(inode);
1275 u64 last_len = 0;
1276 u64 skip = 0;
1277 u64 defrag_end = 0;
1278 u64 newer_off = range->start;
1279 unsigned long i;
1280 unsigned long ra_index = 0;
1281 int ret;
1282 int defrag_count = 0;
1283 int compress_type = BTRFS_COMPRESS_ZLIB;
1284 u32 extent_thresh = range->extent_thresh;
1285 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1286 unsigned long cluster = max_cluster;
1287 u64 new_align = ~((u64)SZ_128K - 1);
1288 struct page **pages = NULL;
1289
1290 if (isize == 0)
1291 return 0;
1292
1293 if (range->start >= isize)
1294 return -EINVAL;
1295
1296 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1297 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1298 return -EINVAL;
1299 if (range->compress_type)
1300 compress_type = range->compress_type;
1301 }
1302
1303 if (extent_thresh == 0)
1304 extent_thresh = SZ_256K;
1305
1306 /*
1307 * if we were not given a file, allocate a readahead
1308 * context
1309 */
1310 if (!file) {
1311 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1312 if (!ra)
1313 return -ENOMEM;
1314 file_ra_state_init(ra, inode->i_mapping);
1315 } else {
1316 ra = &file->f_ra;
1317 }
1318
1319 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1320 GFP_NOFS);
1321 if (!pages) {
1322 ret = -ENOMEM;
1323 goto out_ra;
1324 }
1325
1326 /* find the last page to defrag */
1327 if (range->start + range->len > range->start) {
1328 last_index = min_t(u64, isize - 1,
1329 range->start + range->len - 1) >> PAGE_SHIFT;
1330 } else {
1331 last_index = (isize - 1) >> PAGE_SHIFT;
1332 }
1333
1334 if (newer_than) {
1335 ret = find_new_extents(root, inode, newer_than,
1336 &newer_off, SZ_64K);
1337 if (!ret) {
1338 range->start = newer_off;
1339 /*
1340 * we always align our defrag to help keep
1341 * the extents in the file evenly spaced
1342 */
1343 i = (newer_off & new_align) >> PAGE_SHIFT;
1344 } else
1345 goto out_ra;
1346 } else {
1347 i = range->start >> PAGE_SHIFT;
1348 }
1349 if (!max_to_defrag)
1350 max_to_defrag = last_index - i + 1;
1351
1352 /*
1353 * make writeback starts from i, so the defrag range can be
1354 * written sequentially.
1355 */
1356 if (i < inode->i_mapping->writeback_index)
1357 inode->i_mapping->writeback_index = i;
1358
1359 while (i <= last_index && defrag_count < max_to_defrag &&
1360 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1361 /*
1362 * make sure we stop running if someone unmounts
1363 * the FS
1364 */
1365 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1366 break;
1367
1368 if (btrfs_defrag_cancelled(root->fs_info)) {
1369 btrfs_debug(root->fs_info, "defrag_file cancelled");
1370 ret = -EAGAIN;
1371 break;
1372 }
1373
1374 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1375 extent_thresh, &last_len, &skip,
1376 &defrag_end, range->flags &
1377 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1378 unsigned long next;
1379 /*
1380 * the should_defrag function tells us how much to skip
1381 * bump our counter by the suggested amount
1382 */
1383 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1384 i = max(i + 1, next);
1385 continue;
1386 }
1387
1388 if (!newer_than) {
1389 cluster = (PAGE_ALIGN(defrag_end) >>
1390 PAGE_SHIFT) - i;
1391 cluster = min(cluster, max_cluster);
1392 } else {
1393 cluster = max_cluster;
1394 }
1395
1396 if (i + cluster > ra_index) {
1397 ra_index = max(i, ra_index);
1398 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1399 cluster);
1400 ra_index += cluster;
1401 }
1402
1403 inode_lock(inode);
1404 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1405 BTRFS_I(inode)->force_compress = compress_type;
1406 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1407 if (ret < 0) {
1408 inode_unlock(inode);
1409 goto out_ra;
1410 }
1411
1412 defrag_count += ret;
1413 balance_dirty_pages_ratelimited(inode->i_mapping);
1414 inode_unlock(inode);
1415
1416 if (newer_than) {
1417 if (newer_off == (u64)-1)
1418 break;
1419
1420 if (ret > 0)
1421 i += ret;
1422
1423 newer_off = max(newer_off + 1,
1424 (u64)i << PAGE_SHIFT);
1425
1426 ret = find_new_extents(root, inode, newer_than,
1427 &newer_off, SZ_64K);
1428 if (!ret) {
1429 range->start = newer_off;
1430 i = (newer_off & new_align) >> PAGE_SHIFT;
1431 } else {
1432 break;
1433 }
1434 } else {
1435 if (ret > 0) {
1436 i += ret;
1437 last_len += ret << PAGE_SHIFT;
1438 } else {
1439 i++;
1440 last_len = 0;
1441 }
1442 }
1443 }
1444
1445 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1446 filemap_flush(inode->i_mapping);
1447 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1448 &BTRFS_I(inode)->runtime_flags))
1449 filemap_flush(inode->i_mapping);
1450 }
1451
1452 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1453 /* the filemap_flush will queue IO into the worker threads, but
1454 * we have to make sure the IO is actually started and that
1455 * ordered extents get created before we return
1456 */
1457 atomic_inc(&root->fs_info->async_submit_draining);
1458 while (atomic_read(&root->fs_info->nr_async_submits) ||
1459 atomic_read(&root->fs_info->async_delalloc_pages)) {
1460 wait_event(root->fs_info->async_submit_wait,
1461 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1462 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1463 }
1464 atomic_dec(&root->fs_info->async_submit_draining);
1465 }
1466
1467 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1468 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1469 }
1470
1471 ret = defrag_count;
1472
1473 out_ra:
1474 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1475 inode_lock(inode);
1476 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1477 inode_unlock(inode);
1478 }
1479 if (!file)
1480 kfree(ra);
1481 kfree(pages);
1482 return ret;
1483 }
1484
1485 static noinline int btrfs_ioctl_resize(struct file *file,
1486 void __user *arg)
1487 {
1488 u64 new_size;
1489 u64 old_size;
1490 u64 devid = 1;
1491 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1492 struct btrfs_ioctl_vol_args *vol_args;
1493 struct btrfs_trans_handle *trans;
1494 struct btrfs_device *device = NULL;
1495 char *sizestr;
1496 char *retptr;
1497 char *devstr = NULL;
1498 int ret = 0;
1499 int mod = 0;
1500
1501 if (!capable(CAP_SYS_ADMIN))
1502 return -EPERM;
1503
1504 ret = mnt_want_write_file(file);
1505 if (ret)
1506 return ret;
1507
1508 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1509 1)) {
1510 mnt_drop_write_file(file);
1511 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1512 }
1513
1514 mutex_lock(&root->fs_info->volume_mutex);
1515 vol_args = memdup_user(arg, sizeof(*vol_args));
1516 if (IS_ERR(vol_args)) {
1517 ret = PTR_ERR(vol_args);
1518 goto out;
1519 }
1520
1521 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1522
1523 sizestr = vol_args->name;
1524 devstr = strchr(sizestr, ':');
1525 if (devstr) {
1526 sizestr = devstr + 1;
1527 *devstr = '\0';
1528 devstr = vol_args->name;
1529 ret = kstrtoull(devstr, 10, &devid);
1530 if (ret)
1531 goto out_free;
1532 if (!devid) {
1533 ret = -EINVAL;
1534 goto out_free;
1535 }
1536 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1537 }
1538
1539 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1540 if (!device) {
1541 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1542 devid);
1543 ret = -ENODEV;
1544 goto out_free;
1545 }
1546
1547 if (!device->writeable) {
1548 btrfs_info(root->fs_info,
1549 "resizer unable to apply on readonly device %llu",
1550 devid);
1551 ret = -EPERM;
1552 goto out_free;
1553 }
1554
1555 if (!strcmp(sizestr, "max"))
1556 new_size = device->bdev->bd_inode->i_size;
1557 else {
1558 if (sizestr[0] == '-') {
1559 mod = -1;
1560 sizestr++;
1561 } else if (sizestr[0] == '+') {
1562 mod = 1;
1563 sizestr++;
1564 }
1565 new_size = memparse(sizestr, &retptr);
1566 if (*retptr != '\0' || new_size == 0) {
1567 ret = -EINVAL;
1568 goto out_free;
1569 }
1570 }
1571
1572 if (device->is_tgtdev_for_dev_replace) {
1573 ret = -EPERM;
1574 goto out_free;
1575 }
1576
1577 old_size = btrfs_device_get_total_bytes(device);
1578
1579 if (mod < 0) {
1580 if (new_size > old_size) {
1581 ret = -EINVAL;
1582 goto out_free;
1583 }
1584 new_size = old_size - new_size;
1585 } else if (mod > 0) {
1586 if (new_size > ULLONG_MAX - old_size) {
1587 ret = -ERANGE;
1588 goto out_free;
1589 }
1590 new_size = old_size + new_size;
1591 }
1592
1593 if (new_size < SZ_256M) {
1594 ret = -EINVAL;
1595 goto out_free;
1596 }
1597 if (new_size > device->bdev->bd_inode->i_size) {
1598 ret = -EFBIG;
1599 goto out_free;
1600 }
1601
1602 new_size = div_u64(new_size, root->sectorsize);
1603 new_size *= root->sectorsize;
1604
1605 btrfs_info_in_rcu(root->fs_info, "new size for %s is %llu",
1606 rcu_str_deref(device->name), new_size);
1607
1608 if (new_size > old_size) {
1609 trans = btrfs_start_transaction(root, 0);
1610 if (IS_ERR(trans)) {
1611 ret = PTR_ERR(trans);
1612 goto out_free;
1613 }
1614 ret = btrfs_grow_device(trans, device, new_size);
1615 btrfs_commit_transaction(trans, root);
1616 } else if (new_size < old_size) {
1617 ret = btrfs_shrink_device(device, new_size);
1618 } /* equal, nothing need to do */
1619
1620 out_free:
1621 kfree(vol_args);
1622 out:
1623 mutex_unlock(&root->fs_info->volume_mutex);
1624 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1625 mnt_drop_write_file(file);
1626 return ret;
1627 }
1628
1629 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1630 char *name, unsigned long fd, int subvol,
1631 u64 *transid, bool readonly,
1632 struct btrfs_qgroup_inherit *inherit)
1633 {
1634 int namelen;
1635 int ret = 0;
1636
1637 if (!S_ISDIR(file_inode(file)->i_mode))
1638 return -ENOTDIR;
1639
1640 ret = mnt_want_write_file(file);
1641 if (ret)
1642 goto out;
1643
1644 namelen = strlen(name);
1645 if (strchr(name, '/')) {
1646 ret = -EINVAL;
1647 goto out_drop_write;
1648 }
1649
1650 if (name[0] == '.' &&
1651 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1652 ret = -EEXIST;
1653 goto out_drop_write;
1654 }
1655
1656 if (subvol) {
1657 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1658 NULL, transid, readonly, inherit);
1659 } else {
1660 struct fd src = fdget(fd);
1661 struct inode *src_inode;
1662 if (!src.file) {
1663 ret = -EINVAL;
1664 goto out_drop_write;
1665 }
1666
1667 src_inode = file_inode(src.file);
1668 if (src_inode->i_sb != file_inode(file)->i_sb) {
1669 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1670 "Snapshot src from another FS");
1671 ret = -EXDEV;
1672 } else if (!inode_owner_or_capable(src_inode)) {
1673 /*
1674 * Subvolume creation is not restricted, but snapshots
1675 * are limited to own subvolumes only
1676 */
1677 ret = -EPERM;
1678 } else {
1679 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1680 BTRFS_I(src_inode)->root,
1681 transid, readonly, inherit);
1682 }
1683 fdput(src);
1684 }
1685 out_drop_write:
1686 mnt_drop_write_file(file);
1687 out:
1688 return ret;
1689 }
1690
1691 static noinline int btrfs_ioctl_snap_create(struct file *file,
1692 void __user *arg, int subvol)
1693 {
1694 struct btrfs_ioctl_vol_args *vol_args;
1695 int ret;
1696
1697 if (!S_ISDIR(file_inode(file)->i_mode))
1698 return -ENOTDIR;
1699
1700 vol_args = memdup_user(arg, sizeof(*vol_args));
1701 if (IS_ERR(vol_args))
1702 return PTR_ERR(vol_args);
1703 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1704
1705 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1706 vol_args->fd, subvol,
1707 NULL, false, NULL);
1708
1709 kfree(vol_args);
1710 return ret;
1711 }
1712
1713 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1714 void __user *arg, int subvol)
1715 {
1716 struct btrfs_ioctl_vol_args_v2 *vol_args;
1717 int ret;
1718 u64 transid = 0;
1719 u64 *ptr = NULL;
1720 bool readonly = false;
1721 struct btrfs_qgroup_inherit *inherit = NULL;
1722
1723 if (!S_ISDIR(file_inode(file)->i_mode))
1724 return -ENOTDIR;
1725
1726 vol_args = memdup_user(arg, sizeof(*vol_args));
1727 if (IS_ERR(vol_args))
1728 return PTR_ERR(vol_args);
1729 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1730
1731 if (vol_args->flags &
1732 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1733 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1734 ret = -EOPNOTSUPP;
1735 goto free_args;
1736 }
1737
1738 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1739 ptr = &transid;
1740 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1741 readonly = true;
1742 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1743 if (vol_args->size > PAGE_SIZE) {
1744 ret = -EINVAL;
1745 goto free_args;
1746 }
1747 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1748 if (IS_ERR(inherit)) {
1749 ret = PTR_ERR(inherit);
1750 goto free_args;
1751 }
1752 }
1753
1754 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1755 vol_args->fd, subvol, ptr,
1756 readonly, inherit);
1757 if (ret)
1758 goto free_inherit;
1759
1760 if (ptr && copy_to_user(arg +
1761 offsetof(struct btrfs_ioctl_vol_args_v2,
1762 transid),
1763 ptr, sizeof(*ptr)))
1764 ret = -EFAULT;
1765
1766 free_inherit:
1767 kfree(inherit);
1768 free_args:
1769 kfree(vol_args);
1770 return ret;
1771 }
1772
1773 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1774 void __user *arg)
1775 {
1776 struct inode *inode = file_inode(file);
1777 struct btrfs_root *root = BTRFS_I(inode)->root;
1778 int ret = 0;
1779 u64 flags = 0;
1780
1781 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1782 return -EINVAL;
1783
1784 down_read(&root->fs_info->subvol_sem);
1785 if (btrfs_root_readonly(root))
1786 flags |= BTRFS_SUBVOL_RDONLY;
1787 up_read(&root->fs_info->subvol_sem);
1788
1789 if (copy_to_user(arg, &flags, sizeof(flags)))
1790 ret = -EFAULT;
1791
1792 return ret;
1793 }
1794
1795 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1796 void __user *arg)
1797 {
1798 struct inode *inode = file_inode(file);
1799 struct btrfs_root *root = BTRFS_I(inode)->root;
1800 struct btrfs_trans_handle *trans;
1801 u64 root_flags;
1802 u64 flags;
1803 int ret = 0;
1804
1805 if (!inode_owner_or_capable(inode))
1806 return -EPERM;
1807
1808 ret = mnt_want_write_file(file);
1809 if (ret)
1810 goto out;
1811
1812 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1813 ret = -EINVAL;
1814 goto out_drop_write;
1815 }
1816
1817 if (copy_from_user(&flags, arg, sizeof(flags))) {
1818 ret = -EFAULT;
1819 goto out_drop_write;
1820 }
1821
1822 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1823 ret = -EINVAL;
1824 goto out_drop_write;
1825 }
1826
1827 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1828 ret = -EOPNOTSUPP;
1829 goto out_drop_write;
1830 }
1831
1832 down_write(&root->fs_info->subvol_sem);
1833
1834 /* nothing to do */
1835 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1836 goto out_drop_sem;
1837
1838 root_flags = btrfs_root_flags(&root->root_item);
1839 if (flags & BTRFS_SUBVOL_RDONLY) {
1840 btrfs_set_root_flags(&root->root_item,
1841 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1842 } else {
1843 /*
1844 * Block RO -> RW transition if this subvolume is involved in
1845 * send
1846 */
1847 spin_lock(&root->root_item_lock);
1848 if (root->send_in_progress == 0) {
1849 btrfs_set_root_flags(&root->root_item,
1850 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1851 spin_unlock(&root->root_item_lock);
1852 } else {
1853 spin_unlock(&root->root_item_lock);
1854 btrfs_warn(root->fs_info,
1855 "Attempt to set subvolume %llu read-write during send",
1856 root->root_key.objectid);
1857 ret = -EPERM;
1858 goto out_drop_sem;
1859 }
1860 }
1861
1862 trans = btrfs_start_transaction(root, 1);
1863 if (IS_ERR(trans)) {
1864 ret = PTR_ERR(trans);
1865 goto out_reset;
1866 }
1867
1868 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1869 &root->root_key, &root->root_item);
1870
1871 btrfs_commit_transaction(trans, root);
1872 out_reset:
1873 if (ret)
1874 btrfs_set_root_flags(&root->root_item, root_flags);
1875 out_drop_sem:
1876 up_write(&root->fs_info->subvol_sem);
1877 out_drop_write:
1878 mnt_drop_write_file(file);
1879 out:
1880 return ret;
1881 }
1882
1883 /*
1884 * helper to check if the subvolume references other subvolumes
1885 */
1886 static noinline int may_destroy_subvol(struct btrfs_root *root)
1887 {
1888 struct btrfs_path *path;
1889 struct btrfs_dir_item *di;
1890 struct btrfs_key key;
1891 u64 dir_id;
1892 int ret;
1893
1894 path = btrfs_alloc_path();
1895 if (!path)
1896 return -ENOMEM;
1897
1898 /* Make sure this root isn't set as the default subvol */
1899 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1900 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1901 dir_id, "default", 7, 0);
1902 if (di && !IS_ERR(di)) {
1903 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1904 if (key.objectid == root->root_key.objectid) {
1905 ret = -EPERM;
1906 btrfs_err(root->fs_info,
1907 "deleting default subvolume %llu is not allowed",
1908 key.objectid);
1909 goto out;
1910 }
1911 btrfs_release_path(path);
1912 }
1913
1914 key.objectid = root->root_key.objectid;
1915 key.type = BTRFS_ROOT_REF_KEY;
1916 key.offset = (u64)-1;
1917
1918 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1919 &key, path, 0, 0);
1920 if (ret < 0)
1921 goto out;
1922 BUG_ON(ret == 0);
1923
1924 ret = 0;
1925 if (path->slots[0] > 0) {
1926 path->slots[0]--;
1927 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1928 if (key.objectid == root->root_key.objectid &&
1929 key.type == BTRFS_ROOT_REF_KEY)
1930 ret = -ENOTEMPTY;
1931 }
1932 out:
1933 btrfs_free_path(path);
1934 return ret;
1935 }
1936
1937 static noinline int key_in_sk(struct btrfs_key *key,
1938 struct btrfs_ioctl_search_key *sk)
1939 {
1940 struct btrfs_key test;
1941 int ret;
1942
1943 test.objectid = sk->min_objectid;
1944 test.type = sk->min_type;
1945 test.offset = sk->min_offset;
1946
1947 ret = btrfs_comp_cpu_keys(key, &test);
1948 if (ret < 0)
1949 return 0;
1950
1951 test.objectid = sk->max_objectid;
1952 test.type = sk->max_type;
1953 test.offset = sk->max_offset;
1954
1955 ret = btrfs_comp_cpu_keys(key, &test);
1956 if (ret > 0)
1957 return 0;
1958 return 1;
1959 }
1960
1961 static noinline int copy_to_sk(struct btrfs_path *path,
1962 struct btrfs_key *key,
1963 struct btrfs_ioctl_search_key *sk,
1964 size_t *buf_size,
1965 char __user *ubuf,
1966 unsigned long *sk_offset,
1967 int *num_found)
1968 {
1969 u64 found_transid;
1970 struct extent_buffer *leaf;
1971 struct btrfs_ioctl_search_header sh;
1972 struct btrfs_key test;
1973 unsigned long item_off;
1974 unsigned long item_len;
1975 int nritems;
1976 int i;
1977 int slot;
1978 int ret = 0;
1979
1980 leaf = path->nodes[0];
1981 slot = path->slots[0];
1982 nritems = btrfs_header_nritems(leaf);
1983
1984 if (btrfs_header_generation(leaf) > sk->max_transid) {
1985 i = nritems;
1986 goto advance_key;
1987 }
1988 found_transid = btrfs_header_generation(leaf);
1989
1990 for (i = slot; i < nritems; i++) {
1991 item_off = btrfs_item_ptr_offset(leaf, i);
1992 item_len = btrfs_item_size_nr(leaf, i);
1993
1994 btrfs_item_key_to_cpu(leaf, key, i);
1995 if (!key_in_sk(key, sk))
1996 continue;
1997
1998 if (sizeof(sh) + item_len > *buf_size) {
1999 if (*num_found) {
2000 ret = 1;
2001 goto out;
2002 }
2003
2004 /*
2005 * return one empty item back for v1, which does not
2006 * handle -EOVERFLOW
2007 */
2008
2009 *buf_size = sizeof(sh) + item_len;
2010 item_len = 0;
2011 ret = -EOVERFLOW;
2012 }
2013
2014 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2015 ret = 1;
2016 goto out;
2017 }
2018
2019 sh.objectid = key->objectid;
2020 sh.offset = key->offset;
2021 sh.type = key->type;
2022 sh.len = item_len;
2023 sh.transid = found_transid;
2024
2025 /* copy search result header */
2026 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2027 ret = -EFAULT;
2028 goto out;
2029 }
2030
2031 *sk_offset += sizeof(sh);
2032
2033 if (item_len) {
2034 char __user *up = ubuf + *sk_offset;
2035 /* copy the item */
2036 if (read_extent_buffer_to_user(leaf, up,
2037 item_off, item_len)) {
2038 ret = -EFAULT;
2039 goto out;
2040 }
2041
2042 *sk_offset += item_len;
2043 }
2044 (*num_found)++;
2045
2046 if (ret) /* -EOVERFLOW from above */
2047 goto out;
2048
2049 if (*num_found >= sk->nr_items) {
2050 ret = 1;
2051 goto out;
2052 }
2053 }
2054 advance_key:
2055 ret = 0;
2056 test.objectid = sk->max_objectid;
2057 test.type = sk->max_type;
2058 test.offset = sk->max_offset;
2059 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2060 ret = 1;
2061 else if (key->offset < (u64)-1)
2062 key->offset++;
2063 else if (key->type < (u8)-1) {
2064 key->offset = 0;
2065 key->type++;
2066 } else if (key->objectid < (u64)-1) {
2067 key->offset = 0;
2068 key->type = 0;
2069 key->objectid++;
2070 } else
2071 ret = 1;
2072 out:
2073 /*
2074 * 0: all items from this leaf copied, continue with next
2075 * 1: * more items can be copied, but unused buffer is too small
2076 * * all items were found
2077 * Either way, it will stops the loop which iterates to the next
2078 * leaf
2079 * -EOVERFLOW: item was to large for buffer
2080 * -EFAULT: could not copy extent buffer back to userspace
2081 */
2082 return ret;
2083 }
2084
2085 static noinline int search_ioctl(struct inode *inode,
2086 struct btrfs_ioctl_search_key *sk,
2087 size_t *buf_size,
2088 char __user *ubuf)
2089 {
2090 struct btrfs_root *root;
2091 struct btrfs_key key;
2092 struct btrfs_path *path;
2093 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2094 int ret;
2095 int num_found = 0;
2096 unsigned long sk_offset = 0;
2097
2098 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2099 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2100 return -EOVERFLOW;
2101 }
2102
2103 path = btrfs_alloc_path();
2104 if (!path)
2105 return -ENOMEM;
2106
2107 if (sk->tree_id == 0) {
2108 /* search the root of the inode that was passed */
2109 root = BTRFS_I(inode)->root;
2110 } else {
2111 key.objectid = sk->tree_id;
2112 key.type = BTRFS_ROOT_ITEM_KEY;
2113 key.offset = (u64)-1;
2114 root = btrfs_read_fs_root_no_name(info, &key);
2115 if (IS_ERR(root)) {
2116 btrfs_free_path(path);
2117 return -ENOENT;
2118 }
2119 }
2120
2121 key.objectid = sk->min_objectid;
2122 key.type = sk->min_type;
2123 key.offset = sk->min_offset;
2124
2125 while (1) {
2126 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2127 if (ret != 0) {
2128 if (ret > 0)
2129 ret = 0;
2130 goto err;
2131 }
2132 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2133 &sk_offset, &num_found);
2134 btrfs_release_path(path);
2135 if (ret)
2136 break;
2137
2138 }
2139 if (ret > 0)
2140 ret = 0;
2141 err:
2142 sk->nr_items = num_found;
2143 btrfs_free_path(path);
2144 return ret;
2145 }
2146
2147 static noinline int btrfs_ioctl_tree_search(struct file *file,
2148 void __user *argp)
2149 {
2150 struct btrfs_ioctl_search_args __user *uargs;
2151 struct btrfs_ioctl_search_key sk;
2152 struct inode *inode;
2153 int ret;
2154 size_t buf_size;
2155
2156 if (!capable(CAP_SYS_ADMIN))
2157 return -EPERM;
2158
2159 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2160
2161 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2162 return -EFAULT;
2163
2164 buf_size = sizeof(uargs->buf);
2165
2166 inode = file_inode(file);
2167 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2168
2169 /*
2170 * In the origin implementation an overflow is handled by returning a
2171 * search header with a len of zero, so reset ret.
2172 */
2173 if (ret == -EOVERFLOW)
2174 ret = 0;
2175
2176 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2177 ret = -EFAULT;
2178 return ret;
2179 }
2180
2181 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2182 void __user *argp)
2183 {
2184 struct btrfs_ioctl_search_args_v2 __user *uarg;
2185 struct btrfs_ioctl_search_args_v2 args;
2186 struct inode *inode;
2187 int ret;
2188 size_t buf_size;
2189 const size_t buf_limit = SZ_16M;
2190
2191 if (!capable(CAP_SYS_ADMIN))
2192 return -EPERM;
2193
2194 /* copy search header and buffer size */
2195 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2196 if (copy_from_user(&args, uarg, sizeof(args)))
2197 return -EFAULT;
2198
2199 buf_size = args.buf_size;
2200
2201 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2202 return -EOVERFLOW;
2203
2204 /* limit result size to 16MB */
2205 if (buf_size > buf_limit)
2206 buf_size = buf_limit;
2207
2208 inode = file_inode(file);
2209 ret = search_ioctl(inode, &args.key, &buf_size,
2210 (char *)(&uarg->buf[0]));
2211 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2212 ret = -EFAULT;
2213 else if (ret == -EOVERFLOW &&
2214 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2215 ret = -EFAULT;
2216
2217 return ret;
2218 }
2219
2220 /*
2221 * Search INODE_REFs to identify path name of 'dirid' directory
2222 * in a 'tree_id' tree. and sets path name to 'name'.
2223 */
2224 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2225 u64 tree_id, u64 dirid, char *name)
2226 {
2227 struct btrfs_root *root;
2228 struct btrfs_key key;
2229 char *ptr;
2230 int ret = -1;
2231 int slot;
2232 int len;
2233 int total_len = 0;
2234 struct btrfs_inode_ref *iref;
2235 struct extent_buffer *l;
2236 struct btrfs_path *path;
2237
2238 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2239 name[0]='\0';
2240 return 0;
2241 }
2242
2243 path = btrfs_alloc_path();
2244 if (!path)
2245 return -ENOMEM;
2246
2247 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2248
2249 key.objectid = tree_id;
2250 key.type = BTRFS_ROOT_ITEM_KEY;
2251 key.offset = (u64)-1;
2252 root = btrfs_read_fs_root_no_name(info, &key);
2253 if (IS_ERR(root)) {
2254 btrfs_err(info, "could not find root %llu", tree_id);
2255 ret = -ENOENT;
2256 goto out;
2257 }
2258
2259 key.objectid = dirid;
2260 key.type = BTRFS_INODE_REF_KEY;
2261 key.offset = (u64)-1;
2262
2263 while (1) {
2264 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2265 if (ret < 0)
2266 goto out;
2267 else if (ret > 0) {
2268 ret = btrfs_previous_item(root, path, dirid,
2269 BTRFS_INODE_REF_KEY);
2270 if (ret < 0)
2271 goto out;
2272 else if (ret > 0) {
2273 ret = -ENOENT;
2274 goto out;
2275 }
2276 }
2277
2278 l = path->nodes[0];
2279 slot = path->slots[0];
2280 btrfs_item_key_to_cpu(l, &key, slot);
2281
2282 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2283 len = btrfs_inode_ref_name_len(l, iref);
2284 ptr -= len + 1;
2285 total_len += len + 1;
2286 if (ptr < name) {
2287 ret = -ENAMETOOLONG;
2288 goto out;
2289 }
2290
2291 *(ptr + len) = '/';
2292 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2293
2294 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2295 break;
2296
2297 btrfs_release_path(path);
2298 key.objectid = key.offset;
2299 key.offset = (u64)-1;
2300 dirid = key.objectid;
2301 }
2302 memmove(name, ptr, total_len);
2303 name[total_len] = '\0';
2304 ret = 0;
2305 out:
2306 btrfs_free_path(path);
2307 return ret;
2308 }
2309
2310 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2311 void __user *argp)
2312 {
2313 struct btrfs_ioctl_ino_lookup_args *args;
2314 struct inode *inode;
2315 int ret = 0;
2316
2317 args = memdup_user(argp, sizeof(*args));
2318 if (IS_ERR(args))
2319 return PTR_ERR(args);
2320
2321 inode = file_inode(file);
2322
2323 /*
2324 * Unprivileged query to obtain the containing subvolume root id. The
2325 * path is reset so it's consistent with btrfs_search_path_in_tree.
2326 */
2327 if (args->treeid == 0)
2328 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2329
2330 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2331 args->name[0] = 0;
2332 goto out;
2333 }
2334
2335 if (!capable(CAP_SYS_ADMIN)) {
2336 ret = -EPERM;
2337 goto out;
2338 }
2339
2340 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2341 args->treeid, args->objectid,
2342 args->name);
2343
2344 out:
2345 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2346 ret = -EFAULT;
2347
2348 kfree(args);
2349 return ret;
2350 }
2351
2352 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2353 void __user *arg)
2354 {
2355 struct dentry *parent = file->f_path.dentry;
2356 struct dentry *dentry;
2357 struct inode *dir = d_inode(parent);
2358 struct inode *inode;
2359 struct btrfs_root *root = BTRFS_I(dir)->root;
2360 struct btrfs_root *dest = NULL;
2361 struct btrfs_ioctl_vol_args *vol_args;
2362 struct btrfs_trans_handle *trans;
2363 struct btrfs_block_rsv block_rsv;
2364 u64 root_flags;
2365 u64 qgroup_reserved;
2366 int namelen;
2367 int ret;
2368 int err = 0;
2369
2370 if (!S_ISDIR(dir->i_mode))
2371 return -ENOTDIR;
2372
2373 vol_args = memdup_user(arg, sizeof(*vol_args));
2374 if (IS_ERR(vol_args))
2375 return PTR_ERR(vol_args);
2376
2377 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2378 namelen = strlen(vol_args->name);
2379 if (strchr(vol_args->name, '/') ||
2380 strncmp(vol_args->name, "..", namelen) == 0) {
2381 err = -EINVAL;
2382 goto out;
2383 }
2384
2385 err = mnt_want_write_file(file);
2386 if (err)
2387 goto out;
2388
2389
2390 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2391 if (err == -EINTR)
2392 goto out_drop_write;
2393 dentry = lookup_one_len(vol_args->name, parent, namelen);
2394 if (IS_ERR(dentry)) {
2395 err = PTR_ERR(dentry);
2396 goto out_unlock_dir;
2397 }
2398
2399 if (d_really_is_negative(dentry)) {
2400 err = -ENOENT;
2401 goto out_dput;
2402 }
2403
2404 inode = d_inode(dentry);
2405 dest = BTRFS_I(inode)->root;
2406 if (!capable(CAP_SYS_ADMIN)) {
2407 /*
2408 * Regular user. Only allow this with a special mount
2409 * option, when the user has write+exec access to the
2410 * subvol root, and when rmdir(2) would have been
2411 * allowed.
2412 *
2413 * Note that this is _not_ check that the subvol is
2414 * empty or doesn't contain data that we wouldn't
2415 * otherwise be able to delete.
2416 *
2417 * Users who want to delete empty subvols should try
2418 * rmdir(2).
2419 */
2420 err = -EPERM;
2421 if (!btrfs_test_opt(root->fs_info, USER_SUBVOL_RM_ALLOWED))
2422 goto out_dput;
2423
2424 /*
2425 * Do not allow deletion if the parent dir is the same
2426 * as the dir to be deleted. That means the ioctl
2427 * must be called on the dentry referencing the root
2428 * of the subvol, not a random directory contained
2429 * within it.
2430 */
2431 err = -EINVAL;
2432 if (root == dest)
2433 goto out_dput;
2434
2435 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2436 if (err)
2437 goto out_dput;
2438 }
2439
2440 /* check if subvolume may be deleted by a user */
2441 err = btrfs_may_delete(dir, dentry, 1);
2442 if (err)
2443 goto out_dput;
2444
2445 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2446 err = -EINVAL;
2447 goto out_dput;
2448 }
2449
2450 inode_lock(inode);
2451
2452 /*
2453 * Don't allow to delete a subvolume with send in progress. This is
2454 * inside the i_mutex so the error handling that has to drop the bit
2455 * again is not run concurrently.
2456 */
2457 spin_lock(&dest->root_item_lock);
2458 root_flags = btrfs_root_flags(&dest->root_item);
2459 if (dest->send_in_progress == 0) {
2460 btrfs_set_root_flags(&dest->root_item,
2461 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2462 spin_unlock(&dest->root_item_lock);
2463 } else {
2464 spin_unlock(&dest->root_item_lock);
2465 btrfs_warn(root->fs_info,
2466 "Attempt to delete subvolume %llu during send",
2467 dest->root_key.objectid);
2468 err = -EPERM;
2469 goto out_unlock_inode;
2470 }
2471
2472 down_write(&root->fs_info->subvol_sem);
2473
2474 err = may_destroy_subvol(dest);
2475 if (err)
2476 goto out_up_write;
2477
2478 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2479 /*
2480 * One for dir inode, two for dir entries, two for root
2481 * ref/backref.
2482 */
2483 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2484 5, &qgroup_reserved, true);
2485 if (err)
2486 goto out_up_write;
2487
2488 trans = btrfs_start_transaction(root, 0);
2489 if (IS_ERR(trans)) {
2490 err = PTR_ERR(trans);
2491 goto out_release;
2492 }
2493 trans->block_rsv = &block_rsv;
2494 trans->bytes_reserved = block_rsv.size;
2495
2496 btrfs_record_snapshot_destroy(trans, dir);
2497
2498 ret = btrfs_unlink_subvol(trans, root, dir,
2499 dest->root_key.objectid,
2500 dentry->d_name.name,
2501 dentry->d_name.len);
2502 if (ret) {
2503 err = ret;
2504 btrfs_abort_transaction(trans, ret);
2505 goto out_end_trans;
2506 }
2507
2508 btrfs_record_root_in_trans(trans, dest);
2509
2510 memset(&dest->root_item.drop_progress, 0,
2511 sizeof(dest->root_item.drop_progress));
2512 dest->root_item.drop_level = 0;
2513 btrfs_set_root_refs(&dest->root_item, 0);
2514
2515 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2516 ret = btrfs_insert_orphan_item(trans,
2517 root->fs_info->tree_root,
2518 dest->root_key.objectid);
2519 if (ret) {
2520 btrfs_abort_transaction(trans, ret);
2521 err = ret;
2522 goto out_end_trans;
2523 }
2524 }
2525
2526 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2527 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2528 dest->root_key.objectid);
2529 if (ret && ret != -ENOENT) {
2530 btrfs_abort_transaction(trans, ret);
2531 err = ret;
2532 goto out_end_trans;
2533 }
2534 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2535 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2536 dest->root_item.received_uuid,
2537 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2538 dest->root_key.objectid);
2539 if (ret && ret != -ENOENT) {
2540 btrfs_abort_transaction(trans, ret);
2541 err = ret;
2542 goto out_end_trans;
2543 }
2544 }
2545
2546 out_end_trans:
2547 trans->block_rsv = NULL;
2548 trans->bytes_reserved = 0;
2549 ret = btrfs_end_transaction(trans, root);
2550 if (ret && !err)
2551 err = ret;
2552 inode->i_flags |= S_DEAD;
2553 out_release:
2554 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2555 out_up_write:
2556 up_write(&root->fs_info->subvol_sem);
2557 if (err) {
2558 spin_lock(&dest->root_item_lock);
2559 root_flags = btrfs_root_flags(&dest->root_item);
2560 btrfs_set_root_flags(&dest->root_item,
2561 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2562 spin_unlock(&dest->root_item_lock);
2563 }
2564 out_unlock_inode:
2565 inode_unlock(inode);
2566 if (!err) {
2567 d_invalidate(dentry);
2568 btrfs_invalidate_inodes(dest);
2569 d_delete(dentry);
2570 ASSERT(dest->send_in_progress == 0);
2571
2572 /* the last ref */
2573 if (dest->ino_cache_inode) {
2574 iput(dest->ino_cache_inode);
2575 dest->ino_cache_inode = NULL;
2576 }
2577 }
2578 out_dput:
2579 dput(dentry);
2580 out_unlock_dir:
2581 inode_unlock(dir);
2582 out_drop_write:
2583 mnt_drop_write_file(file);
2584 out:
2585 kfree(vol_args);
2586 return err;
2587 }
2588
2589 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2590 {
2591 struct inode *inode = file_inode(file);
2592 struct btrfs_root *root = BTRFS_I(inode)->root;
2593 struct btrfs_ioctl_defrag_range_args *range;
2594 int ret;
2595
2596 ret = mnt_want_write_file(file);
2597 if (ret)
2598 return ret;
2599
2600 if (btrfs_root_readonly(root)) {
2601 ret = -EROFS;
2602 goto out;
2603 }
2604
2605 switch (inode->i_mode & S_IFMT) {
2606 case S_IFDIR:
2607 if (!capable(CAP_SYS_ADMIN)) {
2608 ret = -EPERM;
2609 goto out;
2610 }
2611 ret = btrfs_defrag_root(root);
2612 if (ret)
2613 goto out;
2614 ret = btrfs_defrag_root(root->fs_info->extent_root);
2615 break;
2616 case S_IFREG:
2617 if (!(file->f_mode & FMODE_WRITE)) {
2618 ret = -EINVAL;
2619 goto out;
2620 }
2621
2622 range = kzalloc(sizeof(*range), GFP_KERNEL);
2623 if (!range) {
2624 ret = -ENOMEM;
2625 goto out;
2626 }
2627
2628 if (argp) {
2629 if (copy_from_user(range, argp,
2630 sizeof(*range))) {
2631 ret = -EFAULT;
2632 kfree(range);
2633 goto out;
2634 }
2635 /* compression requires us to start the IO */
2636 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2637 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2638 range->extent_thresh = (u32)-1;
2639 }
2640 } else {
2641 /* the rest are all set to zero by kzalloc */
2642 range->len = (u64)-1;
2643 }
2644 ret = btrfs_defrag_file(file_inode(file), file,
2645 range, 0, 0);
2646 if (ret > 0)
2647 ret = 0;
2648 kfree(range);
2649 break;
2650 default:
2651 ret = -EINVAL;
2652 }
2653 out:
2654 mnt_drop_write_file(file);
2655 return ret;
2656 }
2657
2658 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2659 {
2660 struct btrfs_ioctl_vol_args *vol_args;
2661 int ret;
2662
2663 if (!capable(CAP_SYS_ADMIN))
2664 return -EPERM;
2665
2666 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2667 1)) {
2668 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2669 }
2670
2671 mutex_lock(&root->fs_info->volume_mutex);
2672 vol_args = memdup_user(arg, sizeof(*vol_args));
2673 if (IS_ERR(vol_args)) {
2674 ret = PTR_ERR(vol_args);
2675 goto out;
2676 }
2677
2678 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2679 ret = btrfs_init_new_device(root, vol_args->name);
2680
2681 if (!ret)
2682 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2683
2684 kfree(vol_args);
2685 out:
2686 mutex_unlock(&root->fs_info->volume_mutex);
2687 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2688 return ret;
2689 }
2690
2691 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2692 {
2693 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2694 struct btrfs_ioctl_vol_args_v2 *vol_args;
2695 int ret;
2696
2697 if (!capable(CAP_SYS_ADMIN))
2698 return -EPERM;
2699
2700 ret = mnt_want_write_file(file);
2701 if (ret)
2702 return ret;
2703
2704 vol_args = memdup_user(arg, sizeof(*vol_args));
2705 if (IS_ERR(vol_args)) {
2706 ret = PTR_ERR(vol_args);
2707 goto err_drop;
2708 }
2709
2710 /* Check for compatibility reject unknown flags */
2711 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED)
2712 return -EOPNOTSUPP;
2713
2714 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2715 1)) {
2716 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2717 goto out;
2718 }
2719
2720 mutex_lock(&root->fs_info->volume_mutex);
2721 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2722 ret = btrfs_rm_device(root, NULL, vol_args->devid);
2723 } else {
2724 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2725 ret = btrfs_rm_device(root, vol_args->name, 0);
2726 }
2727 mutex_unlock(&root->fs_info->volume_mutex);
2728 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2729
2730 if (!ret) {
2731 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2732 btrfs_info(root->fs_info, "device deleted: id %llu",
2733 vol_args->devid);
2734 else
2735 btrfs_info(root->fs_info, "device deleted: %s",
2736 vol_args->name);
2737 }
2738 out:
2739 kfree(vol_args);
2740 err_drop:
2741 mnt_drop_write_file(file);
2742 return ret;
2743 }
2744
2745 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2746 {
2747 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2748 struct btrfs_ioctl_vol_args *vol_args;
2749 int ret;
2750
2751 if (!capable(CAP_SYS_ADMIN))
2752 return -EPERM;
2753
2754 ret = mnt_want_write_file(file);
2755 if (ret)
2756 return ret;
2757
2758 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2759 1)) {
2760 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2761 goto out_drop_write;
2762 }
2763
2764 vol_args = memdup_user(arg, sizeof(*vol_args));
2765 if (IS_ERR(vol_args)) {
2766 ret = PTR_ERR(vol_args);
2767 goto out;
2768 }
2769
2770 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2771 mutex_lock(&root->fs_info->volume_mutex);
2772 ret = btrfs_rm_device(root, vol_args->name, 0);
2773 mutex_unlock(&root->fs_info->volume_mutex);
2774
2775 if (!ret)
2776 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2777 kfree(vol_args);
2778 out:
2779 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2780 out_drop_write:
2781 mnt_drop_write_file(file);
2782
2783 return ret;
2784 }
2785
2786 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2787 {
2788 struct btrfs_ioctl_fs_info_args *fi_args;
2789 struct btrfs_device *device;
2790 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2791 int ret = 0;
2792
2793 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2794 if (!fi_args)
2795 return -ENOMEM;
2796
2797 mutex_lock(&fs_devices->device_list_mutex);
2798 fi_args->num_devices = fs_devices->num_devices;
2799 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2800
2801 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2802 if (device->devid > fi_args->max_id)
2803 fi_args->max_id = device->devid;
2804 }
2805 mutex_unlock(&fs_devices->device_list_mutex);
2806
2807 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2808 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2809 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2810
2811 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2812 ret = -EFAULT;
2813
2814 kfree(fi_args);
2815 return ret;
2816 }
2817
2818 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2819 {
2820 struct btrfs_ioctl_dev_info_args *di_args;
2821 struct btrfs_device *dev;
2822 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2823 int ret = 0;
2824 char *s_uuid = NULL;
2825
2826 di_args = memdup_user(arg, sizeof(*di_args));
2827 if (IS_ERR(di_args))
2828 return PTR_ERR(di_args);
2829
2830 if (!btrfs_is_empty_uuid(di_args->uuid))
2831 s_uuid = di_args->uuid;
2832
2833 mutex_lock(&fs_devices->device_list_mutex);
2834 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2835
2836 if (!dev) {
2837 ret = -ENODEV;
2838 goto out;
2839 }
2840
2841 di_args->devid = dev->devid;
2842 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2843 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2844 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2845 if (dev->name) {
2846 struct rcu_string *name;
2847
2848 rcu_read_lock();
2849 name = rcu_dereference(dev->name);
2850 strncpy(di_args->path, name->str, sizeof(di_args->path));
2851 rcu_read_unlock();
2852 di_args->path[sizeof(di_args->path) - 1] = 0;
2853 } else {
2854 di_args->path[0] = '\0';
2855 }
2856
2857 out:
2858 mutex_unlock(&fs_devices->device_list_mutex);
2859 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2860 ret = -EFAULT;
2861
2862 kfree(di_args);
2863 return ret;
2864 }
2865
2866 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2867 {
2868 struct page *page;
2869
2870 page = grab_cache_page(inode->i_mapping, index);
2871 if (!page)
2872 return ERR_PTR(-ENOMEM);
2873
2874 if (!PageUptodate(page)) {
2875 int ret;
2876
2877 ret = btrfs_readpage(NULL, page);
2878 if (ret)
2879 return ERR_PTR(ret);
2880 lock_page(page);
2881 if (!PageUptodate(page)) {
2882 unlock_page(page);
2883 put_page(page);
2884 return ERR_PTR(-EIO);
2885 }
2886 if (page->mapping != inode->i_mapping) {
2887 unlock_page(page);
2888 put_page(page);
2889 return ERR_PTR(-EAGAIN);
2890 }
2891 }
2892
2893 return page;
2894 }
2895
2896 static int gather_extent_pages(struct inode *inode, struct page **pages,
2897 int num_pages, u64 off)
2898 {
2899 int i;
2900 pgoff_t index = off >> PAGE_SHIFT;
2901
2902 for (i = 0; i < num_pages; i++) {
2903 again:
2904 pages[i] = extent_same_get_page(inode, index + i);
2905 if (IS_ERR(pages[i])) {
2906 int err = PTR_ERR(pages[i]);
2907
2908 if (err == -EAGAIN)
2909 goto again;
2910 pages[i] = NULL;
2911 return err;
2912 }
2913 }
2914 return 0;
2915 }
2916
2917 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2918 bool retry_range_locking)
2919 {
2920 /*
2921 * Do any pending delalloc/csum calculations on inode, one way or
2922 * another, and lock file content.
2923 * The locking order is:
2924 *
2925 * 1) pages
2926 * 2) range in the inode's io tree
2927 */
2928 while (1) {
2929 struct btrfs_ordered_extent *ordered;
2930 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2931 ordered = btrfs_lookup_first_ordered_extent(inode,
2932 off + len - 1);
2933 if ((!ordered ||
2934 ordered->file_offset + ordered->len <= off ||
2935 ordered->file_offset >= off + len) &&
2936 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2937 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2938 if (ordered)
2939 btrfs_put_ordered_extent(ordered);
2940 break;
2941 }
2942 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2943 if (ordered)
2944 btrfs_put_ordered_extent(ordered);
2945 if (!retry_range_locking)
2946 return -EAGAIN;
2947 btrfs_wait_ordered_range(inode, off, len);
2948 }
2949 return 0;
2950 }
2951
2952 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2953 {
2954 inode_unlock(inode1);
2955 inode_unlock(inode2);
2956 }
2957
2958 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2959 {
2960 if (inode1 < inode2)
2961 swap(inode1, inode2);
2962
2963 inode_lock_nested(inode1, I_MUTEX_PARENT);
2964 inode_lock_nested(inode2, I_MUTEX_CHILD);
2965 }
2966
2967 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2968 struct inode *inode2, u64 loff2, u64 len)
2969 {
2970 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2971 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2972 }
2973
2974 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2975 struct inode *inode2, u64 loff2, u64 len,
2976 bool retry_range_locking)
2977 {
2978 int ret;
2979
2980 if (inode1 < inode2) {
2981 swap(inode1, inode2);
2982 swap(loff1, loff2);
2983 }
2984 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2985 if (ret)
2986 return ret;
2987 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2988 if (ret)
2989 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2990 loff1 + len - 1);
2991 return ret;
2992 }
2993
2994 struct cmp_pages {
2995 int num_pages;
2996 struct page **src_pages;
2997 struct page **dst_pages;
2998 };
2999
3000 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
3001 {
3002 int i;
3003 struct page *pg;
3004
3005 for (i = 0; i < cmp->num_pages; i++) {
3006 pg = cmp->src_pages[i];
3007 if (pg) {
3008 unlock_page(pg);
3009 put_page(pg);
3010 }
3011 pg = cmp->dst_pages[i];
3012 if (pg) {
3013 unlock_page(pg);
3014 put_page(pg);
3015 }
3016 }
3017 kfree(cmp->src_pages);
3018 kfree(cmp->dst_pages);
3019 }
3020
3021 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
3022 struct inode *dst, u64 dst_loff,
3023 u64 len, struct cmp_pages *cmp)
3024 {
3025 int ret;
3026 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
3027 struct page **src_pgarr, **dst_pgarr;
3028
3029 /*
3030 * We must gather up all the pages before we initiate our
3031 * extent locking. We use an array for the page pointers. Size
3032 * of the array is bounded by len, which is in turn bounded by
3033 * BTRFS_MAX_DEDUPE_LEN.
3034 */
3035 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
3036 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
3037 if (!src_pgarr || !dst_pgarr) {
3038 kfree(src_pgarr);
3039 kfree(dst_pgarr);
3040 return -ENOMEM;
3041 }
3042 cmp->num_pages = num_pages;
3043 cmp->src_pages = src_pgarr;
3044 cmp->dst_pages = dst_pgarr;
3045
3046 ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff);
3047 if (ret)
3048 goto out;
3049
3050 ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff);
3051
3052 out:
3053 if (ret)
3054 btrfs_cmp_data_free(cmp);
3055 return 0;
3056 }
3057
3058 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
3059 u64 dst_loff, u64 len, struct cmp_pages *cmp)
3060 {
3061 int ret = 0;
3062 int i;
3063 struct page *src_page, *dst_page;
3064 unsigned int cmp_len = PAGE_SIZE;
3065 void *addr, *dst_addr;
3066
3067 i = 0;
3068 while (len) {
3069 if (len < PAGE_SIZE)
3070 cmp_len = len;
3071
3072 BUG_ON(i >= cmp->num_pages);
3073
3074 src_page = cmp->src_pages[i];
3075 dst_page = cmp->dst_pages[i];
3076 ASSERT(PageLocked(src_page));
3077 ASSERT(PageLocked(dst_page));
3078
3079 addr = kmap_atomic(src_page);
3080 dst_addr = kmap_atomic(dst_page);
3081
3082 flush_dcache_page(src_page);
3083 flush_dcache_page(dst_page);
3084
3085 if (memcmp(addr, dst_addr, cmp_len))
3086 ret = -EBADE;
3087
3088 kunmap_atomic(addr);
3089 kunmap_atomic(dst_addr);
3090
3091 if (ret)
3092 break;
3093
3094 len -= cmp_len;
3095 i++;
3096 }
3097
3098 return ret;
3099 }
3100
3101 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3102 u64 olen)
3103 {
3104 u64 len = *plen;
3105 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3106
3107 if (off + olen > inode->i_size || off + olen < off)
3108 return -EINVAL;
3109
3110 /* if we extend to eof, continue to block boundary */
3111 if (off + len == inode->i_size)
3112 *plen = len = ALIGN(inode->i_size, bs) - off;
3113
3114 /* Check that we are block aligned - btrfs_clone() requires this */
3115 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3116 return -EINVAL;
3117
3118 return 0;
3119 }
3120
3121 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3122 struct inode *dst, u64 dst_loff)
3123 {
3124 int ret;
3125 u64 len = olen;
3126 struct cmp_pages cmp;
3127 int same_inode = 0;
3128 u64 same_lock_start = 0;
3129 u64 same_lock_len = 0;
3130
3131 if (src == dst)
3132 same_inode = 1;
3133
3134 if (len == 0)
3135 return 0;
3136
3137 if (same_inode) {
3138 inode_lock(src);
3139
3140 ret = extent_same_check_offsets(src, loff, &len, olen);
3141 if (ret)
3142 goto out_unlock;
3143 ret = extent_same_check_offsets(src, dst_loff, &len, olen);
3144 if (ret)
3145 goto out_unlock;
3146
3147 /*
3148 * Single inode case wants the same checks, except we
3149 * don't want our length pushed out past i_size as
3150 * comparing that data range makes no sense.
3151 *
3152 * extent_same_check_offsets() will do this for an
3153 * unaligned length at i_size, so catch it here and
3154 * reject the request.
3155 *
3156 * This effectively means we require aligned extents
3157 * for the single-inode case, whereas the other cases
3158 * allow an unaligned length so long as it ends at
3159 * i_size.
3160 */
3161 if (len != olen) {
3162 ret = -EINVAL;
3163 goto out_unlock;
3164 }
3165
3166 /* Check for overlapping ranges */
3167 if (dst_loff + len > loff && dst_loff < loff + len) {
3168 ret = -EINVAL;
3169 goto out_unlock;
3170 }
3171
3172 same_lock_start = min_t(u64, loff, dst_loff);
3173 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3174 } else {
3175 btrfs_double_inode_lock(src, dst);
3176
3177 ret = extent_same_check_offsets(src, loff, &len, olen);
3178 if (ret)
3179 goto out_unlock;
3180
3181 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3182 if (ret)
3183 goto out_unlock;
3184 }
3185
3186 /* don't make the dst file partly checksummed */
3187 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3188 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3189 ret = -EINVAL;
3190 goto out_unlock;
3191 }
3192
3193 again:
3194 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3195 if (ret)
3196 goto out_unlock;
3197
3198 if (same_inode)
3199 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3200 false);
3201 else
3202 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3203 false);
3204 /*
3205 * If one of the inodes has dirty pages in the respective range or
3206 * ordered extents, we need to flush dellaloc and wait for all ordered
3207 * extents in the range. We must unlock the pages and the ranges in the
3208 * io trees to avoid deadlocks when flushing delalloc (requires locking
3209 * pages) and when waiting for ordered extents to complete (they require
3210 * range locking).
3211 */
3212 if (ret == -EAGAIN) {
3213 /*
3214 * Ranges in the io trees already unlocked. Now unlock all
3215 * pages before waiting for all IO to complete.
3216 */
3217 btrfs_cmp_data_free(&cmp);
3218 if (same_inode) {
3219 btrfs_wait_ordered_range(src, same_lock_start,
3220 same_lock_len);
3221 } else {
3222 btrfs_wait_ordered_range(src, loff, len);
3223 btrfs_wait_ordered_range(dst, dst_loff, len);
3224 }
3225 goto again;
3226 }
3227 ASSERT(ret == 0);
3228 if (WARN_ON(ret)) {
3229 /* ranges in the io trees already unlocked */
3230 btrfs_cmp_data_free(&cmp);
3231 return ret;
3232 }
3233
3234 /* pass original length for comparison so we stay within i_size */
3235 ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp);
3236 if (ret == 0)
3237 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3238
3239 if (same_inode)
3240 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3241 same_lock_start + same_lock_len - 1);
3242 else
3243 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3244
3245 btrfs_cmp_data_free(&cmp);
3246 out_unlock:
3247 if (same_inode)
3248 inode_unlock(src);
3249 else
3250 btrfs_double_inode_unlock(src, dst);
3251
3252 return ret;
3253 }
3254
3255 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3256
3257 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3258 struct file *dst_file, u64 dst_loff)
3259 {
3260 struct inode *src = file_inode(src_file);
3261 struct inode *dst = file_inode(dst_file);
3262 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3263 ssize_t res;
3264
3265 if (olen > BTRFS_MAX_DEDUPE_LEN)
3266 olen = BTRFS_MAX_DEDUPE_LEN;
3267
3268 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
3269 /*
3270 * Btrfs does not support blocksize < page_size. As a
3271 * result, btrfs_cmp_data() won't correctly handle
3272 * this situation without an update.
3273 */
3274 return -EINVAL;
3275 }
3276
3277 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3278 if (res)
3279 return res;
3280 return olen;
3281 }
3282
3283 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3284 struct inode *inode,
3285 u64 endoff,
3286 const u64 destoff,
3287 const u64 olen,
3288 int no_time_update)
3289 {
3290 struct btrfs_root *root = BTRFS_I(inode)->root;
3291 int ret;
3292
3293 inode_inc_iversion(inode);
3294 if (!no_time_update)
3295 inode->i_mtime = inode->i_ctime = current_time(inode);
3296 /*
3297 * We round up to the block size at eof when determining which
3298 * extents to clone above, but shouldn't round up the file size.
3299 */
3300 if (endoff > destoff + olen)
3301 endoff = destoff + olen;
3302 if (endoff > inode->i_size)
3303 btrfs_i_size_write(inode, endoff);
3304
3305 ret = btrfs_update_inode(trans, root, inode);
3306 if (ret) {
3307 btrfs_abort_transaction(trans, ret);
3308 btrfs_end_transaction(trans, root);
3309 goto out;
3310 }
3311 ret = btrfs_end_transaction(trans, root);
3312 out:
3313 return ret;
3314 }
3315
3316 static void clone_update_extent_map(struct inode *inode,
3317 const struct btrfs_trans_handle *trans,
3318 const struct btrfs_path *path,
3319 const u64 hole_offset,
3320 const u64 hole_len)
3321 {
3322 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3323 struct extent_map *em;
3324 int ret;
3325
3326 em = alloc_extent_map();
3327 if (!em) {
3328 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3329 &BTRFS_I(inode)->runtime_flags);
3330 return;
3331 }
3332
3333 if (path) {
3334 struct btrfs_file_extent_item *fi;
3335
3336 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3337 struct btrfs_file_extent_item);
3338 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3339 em->generation = -1;
3340 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3341 BTRFS_FILE_EXTENT_INLINE)
3342 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3343 &BTRFS_I(inode)->runtime_flags);
3344 } else {
3345 em->start = hole_offset;
3346 em->len = hole_len;
3347 em->ram_bytes = em->len;
3348 em->orig_start = hole_offset;
3349 em->block_start = EXTENT_MAP_HOLE;
3350 em->block_len = 0;
3351 em->orig_block_len = 0;
3352 em->compress_type = BTRFS_COMPRESS_NONE;
3353 em->generation = trans->transid;
3354 }
3355
3356 while (1) {
3357 write_lock(&em_tree->lock);
3358 ret = add_extent_mapping(em_tree, em, 1);
3359 write_unlock(&em_tree->lock);
3360 if (ret != -EEXIST) {
3361 free_extent_map(em);
3362 break;
3363 }
3364 btrfs_drop_extent_cache(inode, em->start,
3365 em->start + em->len - 1, 0);
3366 }
3367
3368 if (ret)
3369 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3370 &BTRFS_I(inode)->runtime_flags);
3371 }
3372
3373 /*
3374 * Make sure we do not end up inserting an inline extent into a file that has
3375 * already other (non-inline) extents. If a file has an inline extent it can
3376 * not have any other extents and the (single) inline extent must start at the
3377 * file offset 0. Failing to respect these rules will lead to file corruption,
3378 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3379 *
3380 * We can have extents that have been already written to disk or we can have
3381 * dirty ranges still in delalloc, in which case the extent maps and items are
3382 * created only when we run delalloc, and the delalloc ranges might fall outside
3383 * the range we are currently locking in the inode's io tree. So we check the
3384 * inode's i_size because of that (i_size updates are done while holding the
3385 * i_mutex, which we are holding here).
3386 * We also check to see if the inode has a size not greater than "datal" but has
3387 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3388 * protected against such concurrent fallocate calls by the i_mutex).
3389 *
3390 * If the file has no extents but a size greater than datal, do not allow the
3391 * copy because we would need turn the inline extent into a non-inline one (even
3392 * with NO_HOLES enabled). If we find our destination inode only has one inline
3393 * extent, just overwrite it with the source inline extent if its size is less
3394 * than the source extent's size, or we could copy the source inline extent's
3395 * data into the destination inode's inline extent if the later is greater then
3396 * the former.
3397 */
3398 static int clone_copy_inline_extent(struct inode *src,
3399 struct inode *dst,
3400 struct btrfs_trans_handle *trans,
3401 struct btrfs_path *path,
3402 struct btrfs_key *new_key,
3403 const u64 drop_start,
3404 const u64 datal,
3405 const u64 skip,
3406 const u64 size,
3407 char *inline_data)
3408 {
3409 struct btrfs_root *root = BTRFS_I(dst)->root;
3410 const u64 aligned_end = ALIGN(new_key->offset + datal,
3411 root->sectorsize);
3412 int ret;
3413 struct btrfs_key key;
3414
3415 if (new_key->offset > 0)
3416 return -EOPNOTSUPP;
3417
3418 key.objectid = btrfs_ino(dst);
3419 key.type = BTRFS_EXTENT_DATA_KEY;
3420 key.offset = 0;
3421 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3422 if (ret < 0) {
3423 return ret;
3424 } else if (ret > 0) {
3425 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3426 ret = btrfs_next_leaf(root, path);
3427 if (ret < 0)
3428 return ret;
3429 else if (ret > 0)
3430 goto copy_inline_extent;
3431 }
3432 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3433 if (key.objectid == btrfs_ino(dst) &&
3434 key.type == BTRFS_EXTENT_DATA_KEY) {
3435 ASSERT(key.offset > 0);
3436 return -EOPNOTSUPP;
3437 }
3438 } else if (i_size_read(dst) <= datal) {
3439 struct btrfs_file_extent_item *ei;
3440 u64 ext_len;
3441
3442 /*
3443 * If the file size is <= datal, make sure there are no other
3444 * extents following (can happen do to an fallocate call with
3445 * the flag FALLOC_FL_KEEP_SIZE).
3446 */
3447 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3448 struct btrfs_file_extent_item);
3449 /*
3450 * If it's an inline extent, it can not have other extents
3451 * following it.
3452 */
3453 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3454 BTRFS_FILE_EXTENT_INLINE)
3455 goto copy_inline_extent;
3456
3457 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3458 if (ext_len > aligned_end)
3459 return -EOPNOTSUPP;
3460
3461 ret = btrfs_next_item(root, path);
3462 if (ret < 0) {
3463 return ret;
3464 } else if (ret == 0) {
3465 btrfs_item_key_to_cpu(path->nodes[0], &key,
3466 path->slots[0]);
3467 if (key.objectid == btrfs_ino(dst) &&
3468 key.type == BTRFS_EXTENT_DATA_KEY)
3469 return -EOPNOTSUPP;
3470 }
3471 }
3472
3473 copy_inline_extent:
3474 /*
3475 * We have no extent items, or we have an extent at offset 0 which may
3476 * or may not be inlined. All these cases are dealt the same way.
3477 */
3478 if (i_size_read(dst) > datal) {
3479 /*
3480 * If the destination inode has an inline extent...
3481 * This would require copying the data from the source inline
3482 * extent into the beginning of the destination's inline extent.
3483 * But this is really complex, both extents can be compressed
3484 * or just one of them, which would require decompressing and
3485 * re-compressing data (which could increase the new compressed
3486 * size, not allowing the compressed data to fit anymore in an
3487 * inline extent).
3488 * So just don't support this case for now (it should be rare,
3489 * we are not really saving space when cloning inline extents).
3490 */
3491 return -EOPNOTSUPP;
3492 }
3493
3494 btrfs_release_path(path);
3495 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3496 if (ret)
3497 return ret;
3498 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3499 if (ret)
3500 return ret;
3501
3502 if (skip) {
3503 const u32 start = btrfs_file_extent_calc_inline_size(0);
3504
3505 memmove(inline_data + start, inline_data + start + skip, datal);
3506 }
3507
3508 write_extent_buffer(path->nodes[0], inline_data,
3509 btrfs_item_ptr_offset(path->nodes[0],
3510 path->slots[0]),
3511 size);
3512 inode_add_bytes(dst, datal);
3513
3514 return 0;
3515 }
3516
3517 /**
3518 * btrfs_clone() - clone a range from inode file to another
3519 *
3520 * @src: Inode to clone from
3521 * @inode: Inode to clone to
3522 * @off: Offset within source to start clone from
3523 * @olen: Original length, passed by user, of range to clone
3524 * @olen_aligned: Block-aligned value of olen
3525 * @destoff: Offset within @inode to start clone
3526 * @no_time_update: Whether to update mtime/ctime on the target inode
3527 */
3528 static int btrfs_clone(struct inode *src, struct inode *inode,
3529 const u64 off, const u64 olen, const u64 olen_aligned,
3530 const u64 destoff, int no_time_update)
3531 {
3532 struct btrfs_root *root = BTRFS_I(inode)->root;
3533 struct btrfs_path *path = NULL;
3534 struct extent_buffer *leaf;
3535 struct btrfs_trans_handle *trans;
3536 char *buf = NULL;
3537 struct btrfs_key key;
3538 u32 nritems;
3539 int slot;
3540 int ret;
3541 const u64 len = olen_aligned;
3542 u64 last_dest_end = destoff;
3543
3544 ret = -ENOMEM;
3545 buf = kmalloc(root->nodesize, GFP_KERNEL | __GFP_NOWARN);
3546 if (!buf) {
3547 buf = vmalloc(root->nodesize);
3548 if (!buf)
3549 return ret;
3550 }
3551
3552 path = btrfs_alloc_path();
3553 if (!path) {
3554 kvfree(buf);
3555 return ret;
3556 }
3557
3558 path->reada = READA_FORWARD;
3559 /* clone data */
3560 key.objectid = btrfs_ino(src);
3561 key.type = BTRFS_EXTENT_DATA_KEY;
3562 key.offset = off;
3563
3564 while (1) {
3565 u64 next_key_min_offset = key.offset + 1;
3566
3567 /*
3568 * note the key will change type as we walk through the
3569 * tree.
3570 */
3571 path->leave_spinning = 1;
3572 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3573 0, 0);
3574 if (ret < 0)
3575 goto out;
3576 /*
3577 * First search, if no extent item that starts at offset off was
3578 * found but the previous item is an extent item, it's possible
3579 * it might overlap our target range, therefore process it.
3580 */
3581 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3582 btrfs_item_key_to_cpu(path->nodes[0], &key,
3583 path->slots[0] - 1);
3584 if (key.type == BTRFS_EXTENT_DATA_KEY)
3585 path->slots[0]--;
3586 }
3587
3588 nritems = btrfs_header_nritems(path->nodes[0]);
3589 process_slot:
3590 if (path->slots[0] >= nritems) {
3591 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3592 if (ret < 0)
3593 goto out;
3594 if (ret > 0)
3595 break;
3596 nritems = btrfs_header_nritems(path->nodes[0]);
3597 }
3598 leaf = path->nodes[0];
3599 slot = path->slots[0];
3600
3601 btrfs_item_key_to_cpu(leaf, &key, slot);
3602 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3603 key.objectid != btrfs_ino(src))
3604 break;
3605
3606 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3607 struct btrfs_file_extent_item *extent;
3608 int type;
3609 u32 size;
3610 struct btrfs_key new_key;
3611 u64 disko = 0, diskl = 0;
3612 u64 datao = 0, datal = 0;
3613 u8 comp;
3614 u64 drop_start;
3615
3616 extent = btrfs_item_ptr(leaf, slot,
3617 struct btrfs_file_extent_item);
3618 comp = btrfs_file_extent_compression(leaf, extent);
3619 type = btrfs_file_extent_type(leaf, extent);
3620 if (type == BTRFS_FILE_EXTENT_REG ||
3621 type == BTRFS_FILE_EXTENT_PREALLOC) {
3622 disko = btrfs_file_extent_disk_bytenr(leaf,
3623 extent);
3624 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3625 extent);
3626 datao = btrfs_file_extent_offset(leaf, extent);
3627 datal = btrfs_file_extent_num_bytes(leaf,
3628 extent);
3629 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3630 /* take upper bound, may be compressed */
3631 datal = btrfs_file_extent_ram_bytes(leaf,
3632 extent);
3633 }
3634
3635 /*
3636 * The first search might have left us at an extent
3637 * item that ends before our target range's start, can
3638 * happen if we have holes and NO_HOLES feature enabled.
3639 */
3640 if (key.offset + datal <= off) {
3641 path->slots[0]++;
3642 goto process_slot;
3643 } else if (key.offset >= off + len) {
3644 break;
3645 }
3646 next_key_min_offset = key.offset + datal;
3647 size = btrfs_item_size_nr(leaf, slot);
3648 read_extent_buffer(leaf, buf,
3649 btrfs_item_ptr_offset(leaf, slot),
3650 size);
3651
3652 btrfs_release_path(path);
3653 path->leave_spinning = 0;
3654
3655 memcpy(&new_key, &key, sizeof(new_key));
3656 new_key.objectid = btrfs_ino(inode);
3657 if (off <= key.offset)
3658 new_key.offset = key.offset + destoff - off;
3659 else
3660 new_key.offset = destoff;
3661
3662 /*
3663 * Deal with a hole that doesn't have an extent item
3664 * that represents it (NO_HOLES feature enabled).
3665 * This hole is either in the middle of the cloning
3666 * range or at the beginning (fully overlaps it or
3667 * partially overlaps it).
3668 */
3669 if (new_key.offset != last_dest_end)
3670 drop_start = last_dest_end;
3671 else
3672 drop_start = new_key.offset;
3673
3674 /*
3675 * 1 - adjusting old extent (we may have to split it)
3676 * 1 - add new extent
3677 * 1 - inode update
3678 */
3679 trans = btrfs_start_transaction(root, 3);
3680 if (IS_ERR(trans)) {
3681 ret = PTR_ERR(trans);
3682 goto out;
3683 }
3684
3685 if (type == BTRFS_FILE_EXTENT_REG ||
3686 type == BTRFS_FILE_EXTENT_PREALLOC) {
3687 /*
3688 * a | --- range to clone ---| b
3689 * | ------------- extent ------------- |
3690 */
3691
3692 /* subtract range b */
3693 if (key.offset + datal > off + len)
3694 datal = off + len - key.offset;
3695
3696 /* subtract range a */
3697 if (off > key.offset) {
3698 datao += off - key.offset;
3699 datal -= off - key.offset;
3700 }
3701
3702 ret = btrfs_drop_extents(trans, root, inode,
3703 drop_start,
3704 new_key.offset + datal,
3705 1);
3706 if (ret) {
3707 if (ret != -EOPNOTSUPP)
3708 btrfs_abort_transaction(trans,
3709 ret);
3710 btrfs_end_transaction(trans, root);
3711 goto out;
3712 }
3713
3714 ret = btrfs_insert_empty_item(trans, root, path,
3715 &new_key, size);
3716 if (ret) {
3717 btrfs_abort_transaction(trans, ret);
3718 btrfs_end_transaction(trans, root);
3719 goto out;
3720 }
3721
3722 leaf = path->nodes[0];
3723 slot = path->slots[0];
3724 write_extent_buffer(leaf, buf,
3725 btrfs_item_ptr_offset(leaf, slot),
3726 size);
3727
3728 extent = btrfs_item_ptr(leaf, slot,
3729 struct btrfs_file_extent_item);
3730
3731 /* disko == 0 means it's a hole */
3732 if (!disko)
3733 datao = 0;
3734
3735 btrfs_set_file_extent_offset(leaf, extent,
3736 datao);
3737 btrfs_set_file_extent_num_bytes(leaf, extent,
3738 datal);
3739
3740 if (disko) {
3741 inode_add_bytes(inode, datal);
3742 ret = btrfs_inc_extent_ref(trans, root,
3743 disko, diskl, 0,
3744 root->root_key.objectid,
3745 btrfs_ino(inode),
3746 new_key.offset - datao);
3747 if (ret) {
3748 btrfs_abort_transaction(trans,
3749 ret);
3750 btrfs_end_transaction(trans,
3751 root);
3752 goto out;
3753
3754 }
3755 }
3756 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3757 u64 skip = 0;
3758 u64 trim = 0;
3759
3760 if (off > key.offset) {
3761 skip = off - key.offset;
3762 new_key.offset += skip;
3763 }
3764
3765 if (key.offset + datal > off + len)
3766 trim = key.offset + datal - (off + len);
3767
3768 if (comp && (skip || trim)) {
3769 ret = -EINVAL;
3770 btrfs_end_transaction(trans, root);
3771 goto out;
3772 }
3773 size -= skip + trim;
3774 datal -= skip + trim;
3775
3776 ret = clone_copy_inline_extent(src, inode,
3777 trans, path,
3778 &new_key,
3779 drop_start,
3780 datal,
3781 skip, size, buf);
3782 if (ret) {
3783 if (ret != -EOPNOTSUPP)
3784 btrfs_abort_transaction(trans,
3785 ret);
3786 btrfs_end_transaction(trans, root);
3787 goto out;
3788 }
3789 leaf = path->nodes[0];
3790 slot = path->slots[0];
3791 }
3792
3793 /* If we have an implicit hole (NO_HOLES feature). */
3794 if (drop_start < new_key.offset)
3795 clone_update_extent_map(inode, trans,
3796 NULL, drop_start,
3797 new_key.offset - drop_start);
3798
3799 clone_update_extent_map(inode, trans, path, 0, 0);
3800
3801 btrfs_mark_buffer_dirty(leaf);
3802 btrfs_release_path(path);
3803
3804 last_dest_end = ALIGN(new_key.offset + datal,
3805 root->sectorsize);
3806 ret = clone_finish_inode_update(trans, inode,
3807 last_dest_end,
3808 destoff, olen,
3809 no_time_update);
3810 if (ret)
3811 goto out;
3812 if (new_key.offset + datal >= destoff + len)
3813 break;
3814 }
3815 btrfs_release_path(path);
3816 key.offset = next_key_min_offset;
3817
3818 if (fatal_signal_pending(current)) {
3819 ret = -EINTR;
3820 goto out;
3821 }
3822 }
3823 ret = 0;
3824
3825 if (last_dest_end < destoff + len) {
3826 /*
3827 * We have an implicit hole (NO_HOLES feature is enabled) that
3828 * fully or partially overlaps our cloning range at its end.
3829 */
3830 btrfs_release_path(path);
3831
3832 /*
3833 * 1 - remove extent(s)
3834 * 1 - inode update
3835 */
3836 trans = btrfs_start_transaction(root, 2);
3837 if (IS_ERR(trans)) {
3838 ret = PTR_ERR(trans);
3839 goto out;
3840 }
3841 ret = btrfs_drop_extents(trans, root, inode,
3842 last_dest_end, destoff + len, 1);
3843 if (ret) {
3844 if (ret != -EOPNOTSUPP)
3845 btrfs_abort_transaction(trans, ret);
3846 btrfs_end_transaction(trans, root);
3847 goto out;
3848 }
3849 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3850 destoff + len - last_dest_end);
3851 ret = clone_finish_inode_update(trans, inode, destoff + len,
3852 destoff, olen, no_time_update);
3853 }
3854
3855 out:
3856 btrfs_free_path(path);
3857 kvfree(buf);
3858 return ret;
3859 }
3860
3861 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3862 u64 off, u64 olen, u64 destoff)
3863 {
3864 struct inode *inode = file_inode(file);
3865 struct inode *src = file_inode(file_src);
3866 struct btrfs_root *root = BTRFS_I(inode)->root;
3867 int ret;
3868 u64 len = olen;
3869 u64 bs = root->fs_info->sb->s_blocksize;
3870 int same_inode = src == inode;
3871
3872 /*
3873 * TODO:
3874 * - split compressed inline extents. annoying: we need to
3875 * decompress into destination's address_space (the file offset
3876 * may change, so source mapping won't do), then recompress (or
3877 * otherwise reinsert) a subrange.
3878 *
3879 * - split destination inode's inline extents. The inline extents can
3880 * be either compressed or non-compressed.
3881 */
3882
3883 if (btrfs_root_readonly(root))
3884 return -EROFS;
3885
3886 if (file_src->f_path.mnt != file->f_path.mnt ||
3887 src->i_sb != inode->i_sb)
3888 return -EXDEV;
3889
3890 /* don't make the dst file partly checksummed */
3891 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3892 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3893 return -EINVAL;
3894
3895 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3896 return -EISDIR;
3897
3898 if (!same_inode) {
3899 btrfs_double_inode_lock(src, inode);
3900 } else {
3901 inode_lock(src);
3902 }
3903
3904 /* determine range to clone */
3905 ret = -EINVAL;
3906 if (off + len > src->i_size || off + len < off)
3907 goto out_unlock;
3908 if (len == 0)
3909 olen = len = src->i_size - off;
3910 /* if we extend to eof, continue to block boundary */
3911 if (off + len == src->i_size)
3912 len = ALIGN(src->i_size, bs) - off;
3913
3914 if (len == 0) {
3915 ret = 0;
3916 goto out_unlock;
3917 }
3918
3919 /* verify the end result is block aligned */
3920 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3921 !IS_ALIGNED(destoff, bs))
3922 goto out_unlock;
3923
3924 /* verify if ranges are overlapped within the same file */
3925 if (same_inode) {
3926 if (destoff + len > off && destoff < off + len)
3927 goto out_unlock;
3928 }
3929
3930 if (destoff > inode->i_size) {
3931 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3932 if (ret)
3933 goto out_unlock;
3934 }
3935
3936 /*
3937 * Lock the target range too. Right after we replace the file extent
3938 * items in the fs tree (which now point to the cloned data), we might
3939 * have a worker replace them with extent items relative to a write
3940 * operation that was issued before this clone operation (i.e. confront
3941 * with inode.c:btrfs_finish_ordered_io).
3942 */
3943 if (same_inode) {
3944 u64 lock_start = min_t(u64, off, destoff);
3945 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3946
3947 ret = lock_extent_range(src, lock_start, lock_len, true);
3948 } else {
3949 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3950 true);
3951 }
3952 ASSERT(ret == 0);
3953 if (WARN_ON(ret)) {
3954 /* ranges in the io trees already unlocked */
3955 goto out_unlock;
3956 }
3957
3958 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3959
3960 if (same_inode) {
3961 u64 lock_start = min_t(u64, off, destoff);
3962 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3963
3964 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3965 } else {
3966 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3967 }
3968 /*
3969 * Truncate page cache pages so that future reads will see the cloned
3970 * data immediately and not the previous data.
3971 */
3972 truncate_inode_pages_range(&inode->i_data,
3973 round_down(destoff, PAGE_SIZE),
3974 round_up(destoff + len, PAGE_SIZE) - 1);
3975 out_unlock:
3976 if (!same_inode)
3977 btrfs_double_inode_unlock(src, inode);
3978 else
3979 inode_unlock(src);
3980 return ret;
3981 }
3982
3983 ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in,
3984 struct file *file_out, loff_t pos_out,
3985 size_t len, unsigned int flags)
3986 {
3987 ssize_t ret;
3988
3989 ret = btrfs_clone_files(file_out, file_in, pos_in, len, pos_out);
3990 if (ret == 0)
3991 ret = len;
3992 return ret;
3993 }
3994
3995 int btrfs_clone_file_range(struct file *src_file, loff_t off,
3996 struct file *dst_file, loff_t destoff, u64 len)
3997 {
3998 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
3999 }
4000
4001 /*
4002 * there are many ways the trans_start and trans_end ioctls can lead
4003 * to deadlocks. They should only be used by applications that
4004 * basically own the machine, and have a very in depth understanding
4005 * of all the possible deadlocks and enospc problems.
4006 */
4007 static long btrfs_ioctl_trans_start(struct file *file)
4008 {
4009 struct inode *inode = file_inode(file);
4010 struct btrfs_root *root = BTRFS_I(inode)->root;
4011 struct btrfs_trans_handle *trans;
4012 int ret;
4013
4014 ret = -EPERM;
4015 if (!capable(CAP_SYS_ADMIN))
4016 goto out;
4017
4018 ret = -EINPROGRESS;
4019 if (file->private_data)
4020 goto out;
4021
4022 ret = -EROFS;
4023 if (btrfs_root_readonly(root))
4024 goto out;
4025
4026 ret = mnt_want_write_file(file);
4027 if (ret)
4028 goto out;
4029
4030 atomic_inc(&root->fs_info->open_ioctl_trans);
4031
4032 ret = -ENOMEM;
4033 trans = btrfs_start_ioctl_transaction(root);
4034 if (IS_ERR(trans))
4035 goto out_drop;
4036
4037 file->private_data = trans;
4038 return 0;
4039
4040 out_drop:
4041 atomic_dec(&root->fs_info->open_ioctl_trans);
4042 mnt_drop_write_file(file);
4043 out:
4044 return ret;
4045 }
4046
4047 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
4048 {
4049 struct inode *inode = file_inode(file);
4050 struct btrfs_root *root = BTRFS_I(inode)->root;
4051 struct btrfs_root *new_root;
4052 struct btrfs_dir_item *di;
4053 struct btrfs_trans_handle *trans;
4054 struct btrfs_path *path;
4055 struct btrfs_key location;
4056 struct btrfs_disk_key disk_key;
4057 u64 objectid = 0;
4058 u64 dir_id;
4059 int ret;
4060
4061 if (!capable(CAP_SYS_ADMIN))
4062 return -EPERM;
4063
4064 ret = mnt_want_write_file(file);
4065 if (ret)
4066 return ret;
4067
4068 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4069 ret = -EFAULT;
4070 goto out;
4071 }
4072
4073 if (!objectid)
4074 objectid = BTRFS_FS_TREE_OBJECTID;
4075
4076 location.objectid = objectid;
4077 location.type = BTRFS_ROOT_ITEM_KEY;
4078 location.offset = (u64)-1;
4079
4080 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
4081 if (IS_ERR(new_root)) {
4082 ret = PTR_ERR(new_root);
4083 goto out;
4084 }
4085
4086 path = btrfs_alloc_path();
4087 if (!path) {
4088 ret = -ENOMEM;
4089 goto out;
4090 }
4091 path->leave_spinning = 1;
4092
4093 trans = btrfs_start_transaction(root, 1);
4094 if (IS_ERR(trans)) {
4095 btrfs_free_path(path);
4096 ret = PTR_ERR(trans);
4097 goto out;
4098 }
4099
4100 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
4101 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
4102 dir_id, "default", 7, 1);
4103 if (IS_ERR_OR_NULL(di)) {
4104 btrfs_free_path(path);
4105 btrfs_end_transaction(trans, root);
4106 btrfs_err(new_root->fs_info,
4107 "Umm, you don't have the default diritem, this isn't going to work");
4108 ret = -ENOENT;
4109 goto out;
4110 }
4111
4112 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4113 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4114 btrfs_mark_buffer_dirty(path->nodes[0]);
4115 btrfs_free_path(path);
4116
4117 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
4118 btrfs_end_transaction(trans, root);
4119 out:
4120 mnt_drop_write_file(file);
4121 return ret;
4122 }
4123
4124 void btrfs_get_block_group_info(struct list_head *groups_list,
4125 struct btrfs_ioctl_space_info *space)
4126 {
4127 struct btrfs_block_group_cache *block_group;
4128
4129 space->total_bytes = 0;
4130 space->used_bytes = 0;
4131 space->flags = 0;
4132 list_for_each_entry(block_group, groups_list, list) {
4133 space->flags = block_group->flags;
4134 space->total_bytes += block_group->key.offset;
4135 space->used_bytes +=
4136 btrfs_block_group_used(&block_group->item);
4137 }
4138 }
4139
4140 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
4141 {
4142 struct btrfs_ioctl_space_args space_args;
4143 struct btrfs_ioctl_space_info space;
4144 struct btrfs_ioctl_space_info *dest;
4145 struct btrfs_ioctl_space_info *dest_orig;
4146 struct btrfs_ioctl_space_info __user *user_dest;
4147 struct btrfs_space_info *info;
4148 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4149 BTRFS_BLOCK_GROUP_SYSTEM,
4150 BTRFS_BLOCK_GROUP_METADATA,
4151 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4152 int num_types = 4;
4153 int alloc_size;
4154 int ret = 0;
4155 u64 slot_count = 0;
4156 int i, c;
4157
4158 if (copy_from_user(&space_args,
4159 (struct btrfs_ioctl_space_args __user *)arg,
4160 sizeof(space_args)))
4161 return -EFAULT;
4162
4163 for (i = 0; i < num_types; i++) {
4164 struct btrfs_space_info *tmp;
4165
4166 info = NULL;
4167 rcu_read_lock();
4168 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4169 list) {
4170 if (tmp->flags == types[i]) {
4171 info = tmp;
4172 break;
4173 }
4174 }
4175 rcu_read_unlock();
4176
4177 if (!info)
4178 continue;
4179
4180 down_read(&info->groups_sem);
4181 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4182 if (!list_empty(&info->block_groups[c]))
4183 slot_count++;
4184 }
4185 up_read(&info->groups_sem);
4186 }
4187
4188 /*
4189 * Global block reserve, exported as a space_info
4190 */
4191 slot_count++;
4192
4193 /* space_slots == 0 means they are asking for a count */
4194 if (space_args.space_slots == 0) {
4195 space_args.total_spaces = slot_count;
4196 goto out;
4197 }
4198
4199 slot_count = min_t(u64, space_args.space_slots, slot_count);
4200
4201 alloc_size = sizeof(*dest) * slot_count;
4202
4203 /* we generally have at most 6 or so space infos, one for each raid
4204 * level. So, a whole page should be more than enough for everyone
4205 */
4206 if (alloc_size > PAGE_SIZE)
4207 return -ENOMEM;
4208
4209 space_args.total_spaces = 0;
4210 dest = kmalloc(alloc_size, GFP_KERNEL);
4211 if (!dest)
4212 return -ENOMEM;
4213 dest_orig = dest;
4214
4215 /* now we have a buffer to copy into */
4216 for (i = 0; i < num_types; i++) {
4217 struct btrfs_space_info *tmp;
4218
4219 if (!slot_count)
4220 break;
4221
4222 info = NULL;
4223 rcu_read_lock();
4224 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4225 list) {
4226 if (tmp->flags == types[i]) {
4227 info = tmp;
4228 break;
4229 }
4230 }
4231 rcu_read_unlock();
4232
4233 if (!info)
4234 continue;
4235 down_read(&info->groups_sem);
4236 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4237 if (!list_empty(&info->block_groups[c])) {
4238 btrfs_get_block_group_info(
4239 &info->block_groups[c], &space);
4240 memcpy(dest, &space, sizeof(space));
4241 dest++;
4242 space_args.total_spaces++;
4243 slot_count--;
4244 }
4245 if (!slot_count)
4246 break;
4247 }
4248 up_read(&info->groups_sem);
4249 }
4250
4251 /*
4252 * Add global block reserve
4253 */
4254 if (slot_count) {
4255 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4256
4257 spin_lock(&block_rsv->lock);
4258 space.total_bytes = block_rsv->size;
4259 space.used_bytes = block_rsv->size - block_rsv->reserved;
4260 spin_unlock(&block_rsv->lock);
4261 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4262 memcpy(dest, &space, sizeof(space));
4263 space_args.total_spaces++;
4264 }
4265
4266 user_dest = (struct btrfs_ioctl_space_info __user *)
4267 (arg + sizeof(struct btrfs_ioctl_space_args));
4268
4269 if (copy_to_user(user_dest, dest_orig, alloc_size))
4270 ret = -EFAULT;
4271
4272 kfree(dest_orig);
4273 out:
4274 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4275 ret = -EFAULT;
4276
4277 return ret;
4278 }
4279
4280 /*
4281 * there are many ways the trans_start and trans_end ioctls can lead
4282 * to deadlocks. They should only be used by applications that
4283 * basically own the machine, and have a very in depth understanding
4284 * of all the possible deadlocks and enospc problems.
4285 */
4286 long btrfs_ioctl_trans_end(struct file *file)
4287 {
4288 struct inode *inode = file_inode(file);
4289 struct btrfs_root *root = BTRFS_I(inode)->root;
4290 struct btrfs_trans_handle *trans;
4291
4292 trans = file->private_data;
4293 if (!trans)
4294 return -EINVAL;
4295 file->private_data = NULL;
4296
4297 btrfs_end_transaction(trans, root);
4298
4299 atomic_dec(&root->fs_info->open_ioctl_trans);
4300
4301 mnt_drop_write_file(file);
4302 return 0;
4303 }
4304
4305 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4306 void __user *argp)
4307 {
4308 struct btrfs_trans_handle *trans;
4309 u64 transid;
4310 int ret;
4311
4312 trans = btrfs_attach_transaction_barrier(root);
4313 if (IS_ERR(trans)) {
4314 if (PTR_ERR(trans) != -ENOENT)
4315 return PTR_ERR(trans);
4316
4317 /* No running transaction, don't bother */
4318 transid = root->fs_info->last_trans_committed;
4319 goto out;
4320 }
4321 transid = trans->transid;
4322 ret = btrfs_commit_transaction_async(trans, root, 0);
4323 if (ret) {
4324 btrfs_end_transaction(trans, root);
4325 return ret;
4326 }
4327 out:
4328 if (argp)
4329 if (copy_to_user(argp, &transid, sizeof(transid)))
4330 return -EFAULT;
4331 return 0;
4332 }
4333
4334 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4335 void __user *argp)
4336 {
4337 u64 transid;
4338
4339 if (argp) {
4340 if (copy_from_user(&transid, argp, sizeof(transid)))
4341 return -EFAULT;
4342 } else {
4343 transid = 0; /* current trans */
4344 }
4345 return btrfs_wait_for_commit(root, transid);
4346 }
4347
4348 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4349 {
4350 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4351 struct btrfs_ioctl_scrub_args *sa;
4352 int ret;
4353
4354 if (!capable(CAP_SYS_ADMIN))
4355 return -EPERM;
4356
4357 sa = memdup_user(arg, sizeof(*sa));
4358 if (IS_ERR(sa))
4359 return PTR_ERR(sa);
4360
4361 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4362 ret = mnt_want_write_file(file);
4363 if (ret)
4364 goto out;
4365 }
4366
4367 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4368 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4369 0);
4370
4371 if (copy_to_user(arg, sa, sizeof(*sa)))
4372 ret = -EFAULT;
4373
4374 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4375 mnt_drop_write_file(file);
4376 out:
4377 kfree(sa);
4378 return ret;
4379 }
4380
4381 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4382 {
4383 if (!capable(CAP_SYS_ADMIN))
4384 return -EPERM;
4385
4386 return btrfs_scrub_cancel(root->fs_info);
4387 }
4388
4389 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4390 void __user *arg)
4391 {
4392 struct btrfs_ioctl_scrub_args *sa;
4393 int ret;
4394
4395 if (!capable(CAP_SYS_ADMIN))
4396 return -EPERM;
4397
4398 sa = memdup_user(arg, sizeof(*sa));
4399 if (IS_ERR(sa))
4400 return PTR_ERR(sa);
4401
4402 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4403
4404 if (copy_to_user(arg, sa, sizeof(*sa)))
4405 ret = -EFAULT;
4406
4407 kfree(sa);
4408 return ret;
4409 }
4410
4411 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4412 void __user *arg)
4413 {
4414 struct btrfs_ioctl_get_dev_stats *sa;
4415 int ret;
4416
4417 sa = memdup_user(arg, sizeof(*sa));
4418 if (IS_ERR(sa))
4419 return PTR_ERR(sa);
4420
4421 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4422 kfree(sa);
4423 return -EPERM;
4424 }
4425
4426 ret = btrfs_get_dev_stats(root, sa);
4427
4428 if (copy_to_user(arg, sa, sizeof(*sa)))
4429 ret = -EFAULT;
4430
4431 kfree(sa);
4432 return ret;
4433 }
4434
4435 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4436 {
4437 struct btrfs_ioctl_dev_replace_args *p;
4438 int ret;
4439
4440 if (!capable(CAP_SYS_ADMIN))
4441 return -EPERM;
4442
4443 p = memdup_user(arg, sizeof(*p));
4444 if (IS_ERR(p))
4445 return PTR_ERR(p);
4446
4447 switch (p->cmd) {
4448 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4449 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4450 ret = -EROFS;
4451 goto out;
4452 }
4453 if (atomic_xchg(
4454 &root->fs_info->mutually_exclusive_operation_running,
4455 1)) {
4456 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4457 } else {
4458 ret = btrfs_dev_replace_by_ioctl(root, p);
4459 atomic_set(
4460 &root->fs_info->mutually_exclusive_operation_running,
4461 0);
4462 }
4463 break;
4464 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4465 btrfs_dev_replace_status(root->fs_info, p);
4466 ret = 0;
4467 break;
4468 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4469 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4470 break;
4471 default:
4472 ret = -EINVAL;
4473 break;
4474 }
4475
4476 if (copy_to_user(arg, p, sizeof(*p)))
4477 ret = -EFAULT;
4478 out:
4479 kfree(p);
4480 return ret;
4481 }
4482
4483 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4484 {
4485 int ret = 0;
4486 int i;
4487 u64 rel_ptr;
4488 int size;
4489 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4490 struct inode_fs_paths *ipath = NULL;
4491 struct btrfs_path *path;
4492
4493 if (!capable(CAP_DAC_READ_SEARCH))
4494 return -EPERM;
4495
4496 path = btrfs_alloc_path();
4497 if (!path) {
4498 ret = -ENOMEM;
4499 goto out;
4500 }
4501
4502 ipa = memdup_user(arg, sizeof(*ipa));
4503 if (IS_ERR(ipa)) {
4504 ret = PTR_ERR(ipa);
4505 ipa = NULL;
4506 goto out;
4507 }
4508
4509 size = min_t(u32, ipa->size, 4096);
4510 ipath = init_ipath(size, root, path);
4511 if (IS_ERR(ipath)) {
4512 ret = PTR_ERR(ipath);
4513 ipath = NULL;
4514 goto out;
4515 }
4516
4517 ret = paths_from_inode(ipa->inum, ipath);
4518 if (ret < 0)
4519 goto out;
4520
4521 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4522 rel_ptr = ipath->fspath->val[i] -
4523 (u64)(unsigned long)ipath->fspath->val;
4524 ipath->fspath->val[i] = rel_ptr;
4525 }
4526
4527 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4528 (void *)(unsigned long)ipath->fspath, size);
4529 if (ret) {
4530 ret = -EFAULT;
4531 goto out;
4532 }
4533
4534 out:
4535 btrfs_free_path(path);
4536 free_ipath(ipath);
4537 kfree(ipa);
4538
4539 return ret;
4540 }
4541
4542 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4543 {
4544 struct btrfs_data_container *inodes = ctx;
4545 const size_t c = 3 * sizeof(u64);
4546
4547 if (inodes->bytes_left >= c) {
4548 inodes->bytes_left -= c;
4549 inodes->val[inodes->elem_cnt] = inum;
4550 inodes->val[inodes->elem_cnt + 1] = offset;
4551 inodes->val[inodes->elem_cnt + 2] = root;
4552 inodes->elem_cnt += 3;
4553 } else {
4554 inodes->bytes_missing += c - inodes->bytes_left;
4555 inodes->bytes_left = 0;
4556 inodes->elem_missed += 3;
4557 }
4558
4559 return 0;
4560 }
4561
4562 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4563 void __user *arg)
4564 {
4565 int ret = 0;
4566 int size;
4567 struct btrfs_ioctl_logical_ino_args *loi;
4568 struct btrfs_data_container *inodes = NULL;
4569 struct btrfs_path *path = NULL;
4570
4571 if (!capable(CAP_SYS_ADMIN))
4572 return -EPERM;
4573
4574 loi = memdup_user(arg, sizeof(*loi));
4575 if (IS_ERR(loi)) {
4576 ret = PTR_ERR(loi);
4577 loi = NULL;
4578 goto out;
4579 }
4580
4581 path = btrfs_alloc_path();
4582 if (!path) {
4583 ret = -ENOMEM;
4584 goto out;
4585 }
4586
4587 size = min_t(u32, loi->size, SZ_64K);
4588 inodes = init_data_container(size);
4589 if (IS_ERR(inodes)) {
4590 ret = PTR_ERR(inodes);
4591 inodes = NULL;
4592 goto out;
4593 }
4594
4595 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4596 build_ino_list, inodes);
4597 if (ret == -EINVAL)
4598 ret = -ENOENT;
4599 if (ret < 0)
4600 goto out;
4601
4602 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4603 (void *)(unsigned long)inodes, size);
4604 if (ret)
4605 ret = -EFAULT;
4606
4607 out:
4608 btrfs_free_path(path);
4609 vfree(inodes);
4610 kfree(loi);
4611
4612 return ret;
4613 }
4614
4615 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4616 struct btrfs_ioctl_balance_args *bargs)
4617 {
4618 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4619
4620 bargs->flags = bctl->flags;
4621
4622 if (atomic_read(&fs_info->balance_running))
4623 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4624 if (atomic_read(&fs_info->balance_pause_req))
4625 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4626 if (atomic_read(&fs_info->balance_cancel_req))
4627 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4628
4629 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4630 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4631 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4632
4633 if (lock) {
4634 spin_lock(&fs_info->balance_lock);
4635 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4636 spin_unlock(&fs_info->balance_lock);
4637 } else {
4638 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4639 }
4640 }
4641
4642 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4643 {
4644 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4645 struct btrfs_fs_info *fs_info = root->fs_info;
4646 struct btrfs_ioctl_balance_args *bargs;
4647 struct btrfs_balance_control *bctl;
4648 bool need_unlock; /* for mut. excl. ops lock */
4649 int ret;
4650
4651 if (!capable(CAP_SYS_ADMIN))
4652 return -EPERM;
4653
4654 ret = mnt_want_write_file(file);
4655 if (ret)
4656 return ret;
4657
4658 again:
4659 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4660 mutex_lock(&fs_info->volume_mutex);
4661 mutex_lock(&fs_info->balance_mutex);
4662 need_unlock = true;
4663 goto locked;
4664 }
4665
4666 /*
4667 * mut. excl. ops lock is locked. Three possibilities:
4668 * (1) some other op is running
4669 * (2) balance is running
4670 * (3) balance is paused -- special case (think resume)
4671 */
4672 mutex_lock(&fs_info->balance_mutex);
4673 if (fs_info->balance_ctl) {
4674 /* this is either (2) or (3) */
4675 if (!atomic_read(&fs_info->balance_running)) {
4676 mutex_unlock(&fs_info->balance_mutex);
4677 if (!mutex_trylock(&fs_info->volume_mutex))
4678 goto again;
4679 mutex_lock(&fs_info->balance_mutex);
4680
4681 if (fs_info->balance_ctl &&
4682 !atomic_read(&fs_info->balance_running)) {
4683 /* this is (3) */
4684 need_unlock = false;
4685 goto locked;
4686 }
4687
4688 mutex_unlock(&fs_info->balance_mutex);
4689 mutex_unlock(&fs_info->volume_mutex);
4690 goto again;
4691 } else {
4692 /* this is (2) */
4693 mutex_unlock(&fs_info->balance_mutex);
4694 ret = -EINPROGRESS;
4695 goto out;
4696 }
4697 } else {
4698 /* this is (1) */
4699 mutex_unlock(&fs_info->balance_mutex);
4700 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4701 goto out;
4702 }
4703
4704 locked:
4705 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4706
4707 if (arg) {
4708 bargs = memdup_user(arg, sizeof(*bargs));
4709 if (IS_ERR(bargs)) {
4710 ret = PTR_ERR(bargs);
4711 goto out_unlock;
4712 }
4713
4714 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4715 if (!fs_info->balance_ctl) {
4716 ret = -ENOTCONN;
4717 goto out_bargs;
4718 }
4719
4720 bctl = fs_info->balance_ctl;
4721 spin_lock(&fs_info->balance_lock);
4722 bctl->flags |= BTRFS_BALANCE_RESUME;
4723 spin_unlock(&fs_info->balance_lock);
4724
4725 goto do_balance;
4726 }
4727 } else {
4728 bargs = NULL;
4729 }
4730
4731 if (fs_info->balance_ctl) {
4732 ret = -EINPROGRESS;
4733 goto out_bargs;
4734 }
4735
4736 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4737 if (!bctl) {
4738 ret = -ENOMEM;
4739 goto out_bargs;
4740 }
4741
4742 bctl->fs_info = fs_info;
4743 if (arg) {
4744 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4745 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4746 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4747
4748 bctl->flags = bargs->flags;
4749 } else {
4750 /* balance everything - no filters */
4751 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4752 }
4753
4754 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4755 ret = -EINVAL;
4756 goto out_bctl;
4757 }
4758
4759 do_balance:
4760 /*
4761 * Ownership of bctl and mutually_exclusive_operation_running
4762 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4763 * or, if restriper was paused all the way until unmount, in
4764 * free_fs_info. mutually_exclusive_operation_running is
4765 * cleared in __cancel_balance.
4766 */
4767 need_unlock = false;
4768
4769 ret = btrfs_balance(bctl, bargs);
4770 bctl = NULL;
4771
4772 if (arg) {
4773 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4774 ret = -EFAULT;
4775 }
4776
4777 out_bctl:
4778 kfree(bctl);
4779 out_bargs:
4780 kfree(bargs);
4781 out_unlock:
4782 mutex_unlock(&fs_info->balance_mutex);
4783 mutex_unlock(&fs_info->volume_mutex);
4784 if (need_unlock)
4785 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4786 out:
4787 mnt_drop_write_file(file);
4788 return ret;
4789 }
4790
4791 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4792 {
4793 if (!capable(CAP_SYS_ADMIN))
4794 return -EPERM;
4795
4796 switch (cmd) {
4797 case BTRFS_BALANCE_CTL_PAUSE:
4798 return btrfs_pause_balance(root->fs_info);
4799 case BTRFS_BALANCE_CTL_CANCEL:
4800 return btrfs_cancel_balance(root->fs_info);
4801 }
4802
4803 return -EINVAL;
4804 }
4805
4806 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4807 void __user *arg)
4808 {
4809 struct btrfs_fs_info *fs_info = root->fs_info;
4810 struct btrfs_ioctl_balance_args *bargs;
4811 int ret = 0;
4812
4813 if (!capable(CAP_SYS_ADMIN))
4814 return -EPERM;
4815
4816 mutex_lock(&fs_info->balance_mutex);
4817 if (!fs_info->balance_ctl) {
4818 ret = -ENOTCONN;
4819 goto out;
4820 }
4821
4822 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4823 if (!bargs) {
4824 ret = -ENOMEM;
4825 goto out;
4826 }
4827
4828 update_ioctl_balance_args(fs_info, 1, bargs);
4829
4830 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4831 ret = -EFAULT;
4832
4833 kfree(bargs);
4834 out:
4835 mutex_unlock(&fs_info->balance_mutex);
4836 return ret;
4837 }
4838
4839 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4840 {
4841 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4842 struct btrfs_ioctl_quota_ctl_args *sa;
4843 struct btrfs_trans_handle *trans = NULL;
4844 int ret;
4845 int err;
4846
4847 if (!capable(CAP_SYS_ADMIN))
4848 return -EPERM;
4849
4850 ret = mnt_want_write_file(file);
4851 if (ret)
4852 return ret;
4853
4854 sa = memdup_user(arg, sizeof(*sa));
4855 if (IS_ERR(sa)) {
4856 ret = PTR_ERR(sa);
4857 goto drop_write;
4858 }
4859
4860 down_write(&root->fs_info->subvol_sem);
4861 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4862 if (IS_ERR(trans)) {
4863 ret = PTR_ERR(trans);
4864 goto out;
4865 }
4866
4867 switch (sa->cmd) {
4868 case BTRFS_QUOTA_CTL_ENABLE:
4869 ret = btrfs_quota_enable(trans, root->fs_info);
4870 break;
4871 case BTRFS_QUOTA_CTL_DISABLE:
4872 ret = btrfs_quota_disable(trans, root->fs_info);
4873 break;
4874 default:
4875 ret = -EINVAL;
4876 break;
4877 }
4878
4879 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4880 if (err && !ret)
4881 ret = err;
4882 out:
4883 kfree(sa);
4884 up_write(&root->fs_info->subvol_sem);
4885 drop_write:
4886 mnt_drop_write_file(file);
4887 return ret;
4888 }
4889
4890 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4891 {
4892 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4893 struct btrfs_ioctl_qgroup_assign_args *sa;
4894 struct btrfs_trans_handle *trans;
4895 int ret;
4896 int err;
4897
4898 if (!capable(CAP_SYS_ADMIN))
4899 return -EPERM;
4900
4901 ret = mnt_want_write_file(file);
4902 if (ret)
4903 return ret;
4904
4905 sa = memdup_user(arg, sizeof(*sa));
4906 if (IS_ERR(sa)) {
4907 ret = PTR_ERR(sa);
4908 goto drop_write;
4909 }
4910
4911 trans = btrfs_join_transaction(root);
4912 if (IS_ERR(trans)) {
4913 ret = PTR_ERR(trans);
4914 goto out;
4915 }
4916
4917 /* FIXME: check if the IDs really exist */
4918 if (sa->assign) {
4919 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4920 sa->src, sa->dst);
4921 } else {
4922 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4923 sa->src, sa->dst);
4924 }
4925
4926 /* update qgroup status and info */
4927 err = btrfs_run_qgroups(trans, root->fs_info);
4928 if (err < 0)
4929 btrfs_handle_fs_error(root->fs_info, err,
4930 "failed to update qgroup status and info");
4931 err = btrfs_end_transaction(trans, root);
4932 if (err && !ret)
4933 ret = err;
4934
4935 out:
4936 kfree(sa);
4937 drop_write:
4938 mnt_drop_write_file(file);
4939 return ret;
4940 }
4941
4942 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4943 {
4944 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4945 struct btrfs_ioctl_qgroup_create_args *sa;
4946 struct btrfs_trans_handle *trans;
4947 int ret;
4948 int err;
4949
4950 if (!capable(CAP_SYS_ADMIN))
4951 return -EPERM;
4952
4953 ret = mnt_want_write_file(file);
4954 if (ret)
4955 return ret;
4956
4957 sa = memdup_user(arg, sizeof(*sa));
4958 if (IS_ERR(sa)) {
4959 ret = PTR_ERR(sa);
4960 goto drop_write;
4961 }
4962
4963 if (!sa->qgroupid) {
4964 ret = -EINVAL;
4965 goto out;
4966 }
4967
4968 trans = btrfs_join_transaction(root);
4969 if (IS_ERR(trans)) {
4970 ret = PTR_ERR(trans);
4971 goto out;
4972 }
4973
4974 /* FIXME: check if the IDs really exist */
4975 if (sa->create) {
4976 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
4977 } else {
4978 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4979 }
4980
4981 err = btrfs_end_transaction(trans, root);
4982 if (err && !ret)
4983 ret = err;
4984
4985 out:
4986 kfree(sa);
4987 drop_write:
4988 mnt_drop_write_file(file);
4989 return ret;
4990 }
4991
4992 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4993 {
4994 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4995 struct btrfs_ioctl_qgroup_limit_args *sa;
4996 struct btrfs_trans_handle *trans;
4997 int ret;
4998 int err;
4999 u64 qgroupid;
5000
5001 if (!capable(CAP_SYS_ADMIN))
5002 return -EPERM;
5003
5004 ret = mnt_want_write_file(file);
5005 if (ret)
5006 return ret;
5007
5008 sa = memdup_user(arg, sizeof(*sa));
5009 if (IS_ERR(sa)) {
5010 ret = PTR_ERR(sa);
5011 goto drop_write;
5012 }
5013
5014 trans = btrfs_join_transaction(root);
5015 if (IS_ERR(trans)) {
5016 ret = PTR_ERR(trans);
5017 goto out;
5018 }
5019
5020 qgroupid = sa->qgroupid;
5021 if (!qgroupid) {
5022 /* take the current subvol as qgroup */
5023 qgroupid = root->root_key.objectid;
5024 }
5025
5026 /* FIXME: check if the IDs really exist */
5027 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
5028
5029 err = btrfs_end_transaction(trans, root);
5030 if (err && !ret)
5031 ret = err;
5032
5033 out:
5034 kfree(sa);
5035 drop_write:
5036 mnt_drop_write_file(file);
5037 return ret;
5038 }
5039
5040 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
5041 {
5042 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5043 struct btrfs_ioctl_quota_rescan_args *qsa;
5044 int ret;
5045
5046 if (!capable(CAP_SYS_ADMIN))
5047 return -EPERM;
5048
5049 ret = mnt_want_write_file(file);
5050 if (ret)
5051 return ret;
5052
5053 qsa = memdup_user(arg, sizeof(*qsa));
5054 if (IS_ERR(qsa)) {
5055 ret = PTR_ERR(qsa);
5056 goto drop_write;
5057 }
5058
5059 if (qsa->flags) {
5060 ret = -EINVAL;
5061 goto out;
5062 }
5063
5064 ret = btrfs_qgroup_rescan(root->fs_info);
5065
5066 out:
5067 kfree(qsa);
5068 drop_write:
5069 mnt_drop_write_file(file);
5070 return ret;
5071 }
5072
5073 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5074 {
5075 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5076 struct btrfs_ioctl_quota_rescan_args *qsa;
5077 int ret = 0;
5078
5079 if (!capable(CAP_SYS_ADMIN))
5080 return -EPERM;
5081
5082 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
5083 if (!qsa)
5084 return -ENOMEM;
5085
5086 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5087 qsa->flags = 1;
5088 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
5089 }
5090
5091 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5092 ret = -EFAULT;
5093
5094 kfree(qsa);
5095 return ret;
5096 }
5097
5098 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5099 {
5100 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5101
5102 if (!capable(CAP_SYS_ADMIN))
5103 return -EPERM;
5104
5105 return btrfs_qgroup_wait_for_completion(root->fs_info, true);
5106 }
5107
5108 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5109 struct btrfs_ioctl_received_subvol_args *sa)
5110 {
5111 struct inode *inode = file_inode(file);
5112 struct btrfs_root *root = BTRFS_I(inode)->root;
5113 struct btrfs_root_item *root_item = &root->root_item;
5114 struct btrfs_trans_handle *trans;
5115 struct timespec ct = current_time(inode);
5116 int ret = 0;
5117 int received_uuid_changed;
5118
5119 if (!inode_owner_or_capable(inode))
5120 return -EPERM;
5121
5122 ret = mnt_want_write_file(file);
5123 if (ret < 0)
5124 return ret;
5125
5126 down_write(&root->fs_info->subvol_sem);
5127
5128 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
5129 ret = -EINVAL;
5130 goto out;
5131 }
5132
5133 if (btrfs_root_readonly(root)) {
5134 ret = -EROFS;
5135 goto out;
5136 }
5137
5138 /*
5139 * 1 - root item
5140 * 2 - uuid items (received uuid + subvol uuid)
5141 */
5142 trans = btrfs_start_transaction(root, 3);
5143 if (IS_ERR(trans)) {
5144 ret = PTR_ERR(trans);
5145 trans = NULL;
5146 goto out;
5147 }
5148
5149 sa->rtransid = trans->transid;
5150 sa->rtime.sec = ct.tv_sec;
5151 sa->rtime.nsec = ct.tv_nsec;
5152
5153 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5154 BTRFS_UUID_SIZE);
5155 if (received_uuid_changed &&
5156 !btrfs_is_empty_uuid(root_item->received_uuid))
5157 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
5158 root_item->received_uuid,
5159 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5160 root->root_key.objectid);
5161 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5162 btrfs_set_root_stransid(root_item, sa->stransid);
5163 btrfs_set_root_rtransid(root_item, sa->rtransid);
5164 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5165 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5166 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5167 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5168
5169 ret = btrfs_update_root(trans, root->fs_info->tree_root,
5170 &root->root_key, &root->root_item);
5171 if (ret < 0) {
5172 btrfs_end_transaction(trans, root);
5173 goto out;
5174 }
5175 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5176 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
5177 sa->uuid,
5178 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5179 root->root_key.objectid);
5180 if (ret < 0 && ret != -EEXIST) {
5181 btrfs_abort_transaction(trans, ret);
5182 goto out;
5183 }
5184 }
5185 ret = btrfs_commit_transaction(trans, root);
5186 if (ret < 0) {
5187 btrfs_abort_transaction(trans, ret);
5188 goto out;
5189 }
5190
5191 out:
5192 up_write(&root->fs_info->subvol_sem);
5193 mnt_drop_write_file(file);
5194 return ret;
5195 }
5196
5197 #ifdef CONFIG_64BIT
5198 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5199 void __user *arg)
5200 {
5201 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5202 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5203 int ret = 0;
5204
5205 args32 = memdup_user(arg, sizeof(*args32));
5206 if (IS_ERR(args32)) {
5207 ret = PTR_ERR(args32);
5208 args32 = NULL;
5209 goto out;
5210 }
5211
5212 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5213 if (!args64) {
5214 ret = -ENOMEM;
5215 goto out;
5216 }
5217
5218 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5219 args64->stransid = args32->stransid;
5220 args64->rtransid = args32->rtransid;
5221 args64->stime.sec = args32->stime.sec;
5222 args64->stime.nsec = args32->stime.nsec;
5223 args64->rtime.sec = args32->rtime.sec;
5224 args64->rtime.nsec = args32->rtime.nsec;
5225 args64->flags = args32->flags;
5226
5227 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5228 if (ret)
5229 goto out;
5230
5231 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5232 args32->stransid = args64->stransid;
5233 args32->rtransid = args64->rtransid;
5234 args32->stime.sec = args64->stime.sec;
5235 args32->stime.nsec = args64->stime.nsec;
5236 args32->rtime.sec = args64->rtime.sec;
5237 args32->rtime.nsec = args64->rtime.nsec;
5238 args32->flags = args64->flags;
5239
5240 ret = copy_to_user(arg, args32, sizeof(*args32));
5241 if (ret)
5242 ret = -EFAULT;
5243
5244 out:
5245 kfree(args32);
5246 kfree(args64);
5247 return ret;
5248 }
5249 #endif
5250
5251 static long btrfs_ioctl_set_received_subvol(struct file *file,
5252 void __user *arg)
5253 {
5254 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5255 int ret = 0;
5256
5257 sa = memdup_user(arg, sizeof(*sa));
5258 if (IS_ERR(sa)) {
5259 ret = PTR_ERR(sa);
5260 sa = NULL;
5261 goto out;
5262 }
5263
5264 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5265
5266 if (ret)
5267 goto out;
5268
5269 ret = copy_to_user(arg, sa, sizeof(*sa));
5270 if (ret)
5271 ret = -EFAULT;
5272
5273 out:
5274 kfree(sa);
5275 return ret;
5276 }
5277
5278 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5279 {
5280 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5281 size_t len;
5282 int ret;
5283 char label[BTRFS_LABEL_SIZE];
5284
5285 spin_lock(&root->fs_info->super_lock);
5286 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5287 spin_unlock(&root->fs_info->super_lock);
5288
5289 len = strnlen(label, BTRFS_LABEL_SIZE);
5290
5291 if (len == BTRFS_LABEL_SIZE) {
5292 btrfs_warn(root->fs_info,
5293 "label is too long, return the first %zu bytes", --len);
5294 }
5295
5296 ret = copy_to_user(arg, label, len);
5297
5298 return ret ? -EFAULT : 0;
5299 }
5300
5301 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5302 {
5303 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5304 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5305 struct btrfs_trans_handle *trans;
5306 char label[BTRFS_LABEL_SIZE];
5307 int ret;
5308
5309 if (!capable(CAP_SYS_ADMIN))
5310 return -EPERM;
5311
5312 if (copy_from_user(label, arg, sizeof(label)))
5313 return -EFAULT;
5314
5315 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5316 btrfs_err(root->fs_info,
5317 "unable to set label with more than %d bytes",
5318 BTRFS_LABEL_SIZE - 1);
5319 return -EINVAL;
5320 }
5321
5322 ret = mnt_want_write_file(file);
5323 if (ret)
5324 return ret;
5325
5326 trans = btrfs_start_transaction(root, 0);
5327 if (IS_ERR(trans)) {
5328 ret = PTR_ERR(trans);
5329 goto out_unlock;
5330 }
5331
5332 spin_lock(&root->fs_info->super_lock);
5333 strcpy(super_block->label, label);
5334 spin_unlock(&root->fs_info->super_lock);
5335 ret = btrfs_commit_transaction(trans, root);
5336
5337 out_unlock:
5338 mnt_drop_write_file(file);
5339 return ret;
5340 }
5341
5342 #define INIT_FEATURE_FLAGS(suffix) \
5343 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5344 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5345 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5346
5347 int btrfs_ioctl_get_supported_features(void __user *arg)
5348 {
5349 static const struct btrfs_ioctl_feature_flags features[3] = {
5350 INIT_FEATURE_FLAGS(SUPP),
5351 INIT_FEATURE_FLAGS(SAFE_SET),
5352 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5353 };
5354
5355 if (copy_to_user(arg, &features, sizeof(features)))
5356 return -EFAULT;
5357
5358 return 0;
5359 }
5360
5361 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5362 {
5363 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5364 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5365 struct btrfs_ioctl_feature_flags features;
5366
5367 features.compat_flags = btrfs_super_compat_flags(super_block);
5368 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5369 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5370
5371 if (copy_to_user(arg, &features, sizeof(features)))
5372 return -EFAULT;
5373
5374 return 0;
5375 }
5376
5377 static int check_feature_bits(struct btrfs_root *root,
5378 enum btrfs_feature_set set,
5379 u64 change_mask, u64 flags, u64 supported_flags,
5380 u64 safe_set, u64 safe_clear)
5381 {
5382 const char *type = btrfs_feature_set_names[set];
5383 char *names;
5384 u64 disallowed, unsupported;
5385 u64 set_mask = flags & change_mask;
5386 u64 clear_mask = ~flags & change_mask;
5387
5388 unsupported = set_mask & ~supported_flags;
5389 if (unsupported) {
5390 names = btrfs_printable_features(set, unsupported);
5391 if (names) {
5392 btrfs_warn(root->fs_info,
5393 "this kernel does not support the %s feature bit%s",
5394 names, strchr(names, ',') ? "s" : "");
5395 kfree(names);
5396 } else
5397 btrfs_warn(root->fs_info,
5398 "this kernel does not support %s bits 0x%llx",
5399 type, unsupported);
5400 return -EOPNOTSUPP;
5401 }
5402
5403 disallowed = set_mask & ~safe_set;
5404 if (disallowed) {
5405 names = btrfs_printable_features(set, disallowed);
5406 if (names) {
5407 btrfs_warn(root->fs_info,
5408 "can't set the %s feature bit%s while mounted",
5409 names, strchr(names, ',') ? "s" : "");
5410 kfree(names);
5411 } else
5412 btrfs_warn(root->fs_info,
5413 "can't set %s bits 0x%llx while mounted",
5414 type, disallowed);
5415 return -EPERM;
5416 }
5417
5418 disallowed = clear_mask & ~safe_clear;
5419 if (disallowed) {
5420 names = btrfs_printable_features(set, disallowed);
5421 if (names) {
5422 btrfs_warn(root->fs_info,
5423 "can't clear the %s feature bit%s while mounted",
5424 names, strchr(names, ',') ? "s" : "");
5425 kfree(names);
5426 } else
5427 btrfs_warn(root->fs_info,
5428 "can't clear %s bits 0x%llx while mounted",
5429 type, disallowed);
5430 return -EPERM;
5431 }
5432
5433 return 0;
5434 }
5435
5436 #define check_feature(root, change_mask, flags, mask_base) \
5437 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5438 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5439 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5440 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5441
5442 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5443 {
5444 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5445 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5446 struct btrfs_ioctl_feature_flags flags[2];
5447 struct btrfs_trans_handle *trans;
5448 u64 newflags;
5449 int ret;
5450
5451 if (!capable(CAP_SYS_ADMIN))
5452 return -EPERM;
5453
5454 if (copy_from_user(flags, arg, sizeof(flags)))
5455 return -EFAULT;
5456
5457 /* Nothing to do */
5458 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5459 !flags[0].incompat_flags)
5460 return 0;
5461
5462 ret = check_feature(root, flags[0].compat_flags,
5463 flags[1].compat_flags, COMPAT);
5464 if (ret)
5465 return ret;
5466
5467 ret = check_feature(root, flags[0].compat_ro_flags,
5468 flags[1].compat_ro_flags, COMPAT_RO);
5469 if (ret)
5470 return ret;
5471
5472 ret = check_feature(root, flags[0].incompat_flags,
5473 flags[1].incompat_flags, INCOMPAT);
5474 if (ret)
5475 return ret;
5476
5477 ret = mnt_want_write_file(file);
5478 if (ret)
5479 return ret;
5480
5481 trans = btrfs_start_transaction(root, 0);
5482 if (IS_ERR(trans)) {
5483 ret = PTR_ERR(trans);
5484 goto out_drop_write;
5485 }
5486
5487 spin_lock(&root->fs_info->super_lock);
5488 newflags = btrfs_super_compat_flags(super_block);
5489 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5490 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5491 btrfs_set_super_compat_flags(super_block, newflags);
5492
5493 newflags = btrfs_super_compat_ro_flags(super_block);
5494 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5495 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5496 btrfs_set_super_compat_ro_flags(super_block, newflags);
5497
5498 newflags = btrfs_super_incompat_flags(super_block);
5499 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5500 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5501 btrfs_set_super_incompat_flags(super_block, newflags);
5502 spin_unlock(&root->fs_info->super_lock);
5503
5504 ret = btrfs_commit_transaction(trans, root);
5505 out_drop_write:
5506 mnt_drop_write_file(file);
5507
5508 return ret;
5509 }
5510
5511 long btrfs_ioctl(struct file *file, unsigned int
5512 cmd, unsigned long arg)
5513 {
5514 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5515 void __user *argp = (void __user *)arg;
5516
5517 switch (cmd) {
5518 case FS_IOC_GETFLAGS:
5519 return btrfs_ioctl_getflags(file, argp);
5520 case FS_IOC_SETFLAGS:
5521 return btrfs_ioctl_setflags(file, argp);
5522 case FS_IOC_GETVERSION:
5523 return btrfs_ioctl_getversion(file, argp);
5524 case FITRIM:
5525 return btrfs_ioctl_fitrim(file, argp);
5526 case BTRFS_IOC_SNAP_CREATE:
5527 return btrfs_ioctl_snap_create(file, argp, 0);
5528 case BTRFS_IOC_SNAP_CREATE_V2:
5529 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5530 case BTRFS_IOC_SUBVOL_CREATE:
5531 return btrfs_ioctl_snap_create(file, argp, 1);
5532 case BTRFS_IOC_SUBVOL_CREATE_V2:
5533 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5534 case BTRFS_IOC_SNAP_DESTROY:
5535 return btrfs_ioctl_snap_destroy(file, argp);
5536 case BTRFS_IOC_SUBVOL_GETFLAGS:
5537 return btrfs_ioctl_subvol_getflags(file, argp);
5538 case BTRFS_IOC_SUBVOL_SETFLAGS:
5539 return btrfs_ioctl_subvol_setflags(file, argp);
5540 case BTRFS_IOC_DEFAULT_SUBVOL:
5541 return btrfs_ioctl_default_subvol(file, argp);
5542 case BTRFS_IOC_DEFRAG:
5543 return btrfs_ioctl_defrag(file, NULL);
5544 case BTRFS_IOC_DEFRAG_RANGE:
5545 return btrfs_ioctl_defrag(file, argp);
5546 case BTRFS_IOC_RESIZE:
5547 return btrfs_ioctl_resize(file, argp);
5548 case BTRFS_IOC_ADD_DEV:
5549 return btrfs_ioctl_add_dev(root, argp);
5550 case BTRFS_IOC_RM_DEV:
5551 return btrfs_ioctl_rm_dev(file, argp);
5552 case BTRFS_IOC_RM_DEV_V2:
5553 return btrfs_ioctl_rm_dev_v2(file, argp);
5554 case BTRFS_IOC_FS_INFO:
5555 return btrfs_ioctl_fs_info(root, argp);
5556 case BTRFS_IOC_DEV_INFO:
5557 return btrfs_ioctl_dev_info(root, argp);
5558 case BTRFS_IOC_BALANCE:
5559 return btrfs_ioctl_balance(file, NULL);
5560 case BTRFS_IOC_TRANS_START:
5561 return btrfs_ioctl_trans_start(file);
5562 case BTRFS_IOC_TRANS_END:
5563 return btrfs_ioctl_trans_end(file);
5564 case BTRFS_IOC_TREE_SEARCH:
5565 return btrfs_ioctl_tree_search(file, argp);
5566 case BTRFS_IOC_TREE_SEARCH_V2:
5567 return btrfs_ioctl_tree_search_v2(file, argp);
5568 case BTRFS_IOC_INO_LOOKUP:
5569 return btrfs_ioctl_ino_lookup(file, argp);
5570 case BTRFS_IOC_INO_PATHS:
5571 return btrfs_ioctl_ino_to_path(root, argp);
5572 case BTRFS_IOC_LOGICAL_INO:
5573 return btrfs_ioctl_logical_to_ino(root, argp);
5574 case BTRFS_IOC_SPACE_INFO:
5575 return btrfs_ioctl_space_info(root, argp);
5576 case BTRFS_IOC_SYNC: {
5577 int ret;
5578
5579 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5580 if (ret)
5581 return ret;
5582 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
5583 /*
5584 * The transaction thread may want to do more work,
5585 * namely it pokes the cleaner kthread that will start
5586 * processing uncleaned subvols.
5587 */
5588 wake_up_process(root->fs_info->transaction_kthread);
5589 return ret;
5590 }
5591 case BTRFS_IOC_START_SYNC:
5592 return btrfs_ioctl_start_sync(root, argp);
5593 case BTRFS_IOC_WAIT_SYNC:
5594 return btrfs_ioctl_wait_sync(root, argp);
5595 case BTRFS_IOC_SCRUB:
5596 return btrfs_ioctl_scrub(file, argp);
5597 case BTRFS_IOC_SCRUB_CANCEL:
5598 return btrfs_ioctl_scrub_cancel(root, argp);
5599 case BTRFS_IOC_SCRUB_PROGRESS:
5600 return btrfs_ioctl_scrub_progress(root, argp);
5601 case BTRFS_IOC_BALANCE_V2:
5602 return btrfs_ioctl_balance(file, argp);
5603 case BTRFS_IOC_BALANCE_CTL:
5604 return btrfs_ioctl_balance_ctl(root, arg);
5605 case BTRFS_IOC_BALANCE_PROGRESS:
5606 return btrfs_ioctl_balance_progress(root, argp);
5607 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5608 return btrfs_ioctl_set_received_subvol(file, argp);
5609 #ifdef CONFIG_64BIT
5610 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5611 return btrfs_ioctl_set_received_subvol_32(file, argp);
5612 #endif
5613 case BTRFS_IOC_SEND:
5614 return btrfs_ioctl_send(file, argp);
5615 case BTRFS_IOC_GET_DEV_STATS:
5616 return btrfs_ioctl_get_dev_stats(root, argp);
5617 case BTRFS_IOC_QUOTA_CTL:
5618 return btrfs_ioctl_quota_ctl(file, argp);
5619 case BTRFS_IOC_QGROUP_ASSIGN:
5620 return btrfs_ioctl_qgroup_assign(file, argp);
5621 case BTRFS_IOC_QGROUP_CREATE:
5622 return btrfs_ioctl_qgroup_create(file, argp);
5623 case BTRFS_IOC_QGROUP_LIMIT:
5624 return btrfs_ioctl_qgroup_limit(file, argp);
5625 case BTRFS_IOC_QUOTA_RESCAN:
5626 return btrfs_ioctl_quota_rescan(file, argp);
5627 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5628 return btrfs_ioctl_quota_rescan_status(file, argp);
5629 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5630 return btrfs_ioctl_quota_rescan_wait(file, argp);
5631 case BTRFS_IOC_DEV_REPLACE:
5632 return btrfs_ioctl_dev_replace(root, argp);
5633 case BTRFS_IOC_GET_FSLABEL:
5634 return btrfs_ioctl_get_fslabel(file, argp);
5635 case BTRFS_IOC_SET_FSLABEL:
5636 return btrfs_ioctl_set_fslabel(file, argp);
5637 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5638 return btrfs_ioctl_get_supported_features(argp);
5639 case BTRFS_IOC_GET_FEATURES:
5640 return btrfs_ioctl_get_features(file, argp);
5641 case BTRFS_IOC_SET_FEATURES:
5642 return btrfs_ioctl_set_features(file, argp);
5643 }
5644
5645 return -ENOTTY;
5646 }
5647
5648 #ifdef CONFIG_COMPAT
5649 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5650 {
5651 switch (cmd) {
5652 case FS_IOC32_GETFLAGS:
5653 cmd = FS_IOC_GETFLAGS;
5654 break;
5655 case FS_IOC32_SETFLAGS:
5656 cmd = FS_IOC_SETFLAGS;
5657 break;
5658 case FS_IOC32_GETVERSION:
5659 cmd = FS_IOC_GETVERSION;
5660 break;
5661 default:
5662 return -ENOIOCTLCMD;
5663 }
5664
5665 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5666 }
5667 #endif