]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - fs/btrfs/ioctl.c
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
btrfs: root->fs_info cleanup, add fs_info convenience variables
[mirror_ubuntu-bionic-kernel.git] / fs / btrfs / ioctl.c
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
23 #include <linux/fs.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/compat.h>
37 #include <linux/bit_spinlock.h>
38 #include <linux/security.h>
39 #include <linux/xattr.h>
40 #include <linux/vmalloc.h>
41 #include <linux/slab.h>
42 #include <linux/blkdev.h>
43 #include <linux/uuid.h>
44 #include <linux/btrfs.h>
45 #include <linux/uaccess.h>
46 #include "ctree.h"
47 #include "disk-io.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "print-tree.h"
51 #include "volumes.h"
52 #include "locking.h"
53 #include "inode-map.h"
54 #include "backref.h"
55 #include "rcu-string.h"
56 #include "send.h"
57 #include "dev-replace.h"
58 #include "props.h"
59 #include "sysfs.h"
60 #include "qgroup.h"
61 #include "tree-log.h"
62 #include "compression.h"
63
64 #ifdef CONFIG_64BIT
65 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
66 * structures are incorrect, as the timespec structure from userspace
67 * is 4 bytes too small. We define these alternatives here to teach
68 * the kernel about the 32-bit struct packing.
69 */
70 struct btrfs_ioctl_timespec_32 {
71 __u64 sec;
72 __u32 nsec;
73 } __attribute__ ((__packed__));
74
75 struct btrfs_ioctl_received_subvol_args_32 {
76 char uuid[BTRFS_UUID_SIZE]; /* in */
77 __u64 stransid; /* in */
78 __u64 rtransid; /* out */
79 struct btrfs_ioctl_timespec_32 stime; /* in */
80 struct btrfs_ioctl_timespec_32 rtime; /* out */
81 __u64 flags; /* in */
82 __u64 reserved[16]; /* in */
83 } __attribute__ ((__packed__));
84
85 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
86 struct btrfs_ioctl_received_subvol_args_32)
87 #endif
88
89
90 static int btrfs_clone(struct inode *src, struct inode *inode,
91 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
92 int no_time_update);
93
94 /* Mask out flags that are inappropriate for the given type of inode. */
95 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
96 {
97 if (S_ISDIR(mode))
98 return flags;
99 else if (S_ISREG(mode))
100 return flags & ~FS_DIRSYNC_FL;
101 else
102 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
103 }
104
105 /*
106 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
107 */
108 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
109 {
110 unsigned int iflags = 0;
111
112 if (flags & BTRFS_INODE_SYNC)
113 iflags |= FS_SYNC_FL;
114 if (flags & BTRFS_INODE_IMMUTABLE)
115 iflags |= FS_IMMUTABLE_FL;
116 if (flags & BTRFS_INODE_APPEND)
117 iflags |= FS_APPEND_FL;
118 if (flags & BTRFS_INODE_NODUMP)
119 iflags |= FS_NODUMP_FL;
120 if (flags & BTRFS_INODE_NOATIME)
121 iflags |= FS_NOATIME_FL;
122 if (flags & BTRFS_INODE_DIRSYNC)
123 iflags |= FS_DIRSYNC_FL;
124 if (flags & BTRFS_INODE_NODATACOW)
125 iflags |= FS_NOCOW_FL;
126
127 if (flags & BTRFS_INODE_NOCOMPRESS)
128 iflags |= FS_NOCOMP_FL;
129 else if (flags & BTRFS_INODE_COMPRESS)
130 iflags |= FS_COMPR_FL;
131
132 return iflags;
133 }
134
135 /*
136 * Update inode->i_flags based on the btrfs internal flags.
137 */
138 void btrfs_update_iflags(struct inode *inode)
139 {
140 struct btrfs_inode *ip = BTRFS_I(inode);
141 unsigned int new_fl = 0;
142
143 if (ip->flags & BTRFS_INODE_SYNC)
144 new_fl |= S_SYNC;
145 if (ip->flags & BTRFS_INODE_IMMUTABLE)
146 new_fl |= S_IMMUTABLE;
147 if (ip->flags & BTRFS_INODE_APPEND)
148 new_fl |= S_APPEND;
149 if (ip->flags & BTRFS_INODE_NOATIME)
150 new_fl |= S_NOATIME;
151 if (ip->flags & BTRFS_INODE_DIRSYNC)
152 new_fl |= S_DIRSYNC;
153
154 set_mask_bits(&inode->i_flags,
155 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
156 new_fl);
157 }
158
159 /*
160 * Inherit flags from the parent inode.
161 *
162 * Currently only the compression flags and the cow flags are inherited.
163 */
164 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
165 {
166 unsigned int flags;
167
168 if (!dir)
169 return;
170
171 flags = BTRFS_I(dir)->flags;
172
173 if (flags & BTRFS_INODE_NOCOMPRESS) {
174 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
175 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
176 } else if (flags & BTRFS_INODE_COMPRESS) {
177 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
178 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
179 }
180
181 if (flags & BTRFS_INODE_NODATACOW) {
182 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
183 if (S_ISREG(inode->i_mode))
184 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
185 }
186
187 btrfs_update_iflags(inode);
188 }
189
190 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
191 {
192 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
193 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
194
195 if (copy_to_user(arg, &flags, sizeof(flags)))
196 return -EFAULT;
197 return 0;
198 }
199
200 static int check_flags(unsigned int flags)
201 {
202 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
203 FS_NOATIME_FL | FS_NODUMP_FL | \
204 FS_SYNC_FL | FS_DIRSYNC_FL | \
205 FS_NOCOMP_FL | FS_COMPR_FL |
206 FS_NOCOW_FL))
207 return -EOPNOTSUPP;
208
209 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
210 return -EINVAL;
211
212 return 0;
213 }
214
215 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
216 {
217 struct inode *inode = file_inode(file);
218 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
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 (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 inode *inode = file_inode(file);
378 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
379 struct btrfs_device *device;
380 struct request_queue *q;
381 struct fstrim_range range;
382 u64 minlen = ULLONG_MAX;
383 u64 num_devices = 0;
384 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
385 int ret;
386
387 if (!capable(CAP_SYS_ADMIN))
388 return -EPERM;
389
390 rcu_read_lock();
391 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
392 dev_list) {
393 if (!device->bdev)
394 continue;
395 q = bdev_get_queue(device->bdev);
396 if (blk_queue_discard(q)) {
397 num_devices++;
398 minlen = min((u64)q->limits.discard_granularity,
399 minlen);
400 }
401 }
402 rcu_read_unlock();
403
404 if (!num_devices)
405 return -EOPNOTSUPP;
406 if (copy_from_user(&range, arg, sizeof(range)))
407 return -EFAULT;
408 if (range.start > total_bytes ||
409 range.len < fs_info->sb->s_blocksize)
410 return -EINVAL;
411
412 range.len = min(range.len, total_bytes - range.start);
413 range.minlen = max(range.minlen, minlen);
414 ret = btrfs_trim_fs(fs_info->tree_root, &range);
415 if (ret < 0)
416 return ret;
417
418 if (copy_to_user(arg, &range, sizeof(range)))
419 return -EFAULT;
420
421 return 0;
422 }
423
424 int btrfs_is_empty_uuid(u8 *uuid)
425 {
426 int i;
427
428 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
429 if (uuid[i])
430 return 0;
431 }
432 return 1;
433 }
434
435 static noinline int create_subvol(struct inode *dir,
436 struct dentry *dentry,
437 char *name, int namelen,
438 u64 *async_transid,
439 struct btrfs_qgroup_inherit *inherit)
440 {
441 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
442 struct btrfs_trans_handle *trans;
443 struct btrfs_key key;
444 struct btrfs_root_item *root_item;
445 struct btrfs_inode_item *inode_item;
446 struct extent_buffer *leaf;
447 struct btrfs_root *root = BTRFS_I(dir)->root;
448 struct btrfs_root *new_root;
449 struct btrfs_block_rsv block_rsv;
450 struct timespec cur_time = current_time(dir);
451 struct inode *inode;
452 int ret;
453 int err;
454 u64 objectid;
455 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
456 u64 index = 0;
457 u64 qgroup_reserved;
458 uuid_le new_uuid;
459
460 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
461 if (!root_item)
462 return -ENOMEM;
463
464 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
465 if (ret)
466 goto fail_free;
467
468 /*
469 * Don't create subvolume whose level is not zero. Or qgroup will be
470 * screwed up since it assumes subvolume qgroup's level to be 0.
471 */
472 if (btrfs_qgroup_level(objectid)) {
473 ret = -ENOSPC;
474 goto fail_free;
475 }
476
477 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
478 /*
479 * The same as the snapshot creation, please see the comment
480 * of create_snapshot().
481 */
482 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
483 8, &qgroup_reserved, false);
484 if (ret)
485 goto fail_free;
486
487 trans = btrfs_start_transaction(root, 0);
488 if (IS_ERR(trans)) {
489 ret = PTR_ERR(trans);
490 btrfs_subvolume_release_metadata(root, &block_rsv,
491 qgroup_reserved);
492 goto fail_free;
493 }
494 trans->block_rsv = &block_rsv;
495 trans->bytes_reserved = block_rsv.size;
496
497 ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit);
498 if (ret)
499 goto fail;
500
501 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
502 if (IS_ERR(leaf)) {
503 ret = PTR_ERR(leaf);
504 goto fail;
505 }
506
507 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
508 btrfs_set_header_bytenr(leaf, leaf->start);
509 btrfs_set_header_generation(leaf, trans->transid);
510 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
511 btrfs_set_header_owner(leaf, objectid);
512
513 write_extent_buffer_fsid(leaf, fs_info->fsid);
514 write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid);
515 btrfs_mark_buffer_dirty(leaf);
516
517 inode_item = &root_item->inode;
518 btrfs_set_stack_inode_generation(inode_item, 1);
519 btrfs_set_stack_inode_size(inode_item, 3);
520 btrfs_set_stack_inode_nlink(inode_item, 1);
521 btrfs_set_stack_inode_nbytes(inode_item,
522 fs_info->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, 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(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, fs_info,
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, fs_info, root_item->uuid,
608 BTRFS_UUID_KEY_SUBVOL, objectid);
609 if (ret)
610 btrfs_abort_transaction(trans, ret);
611
612 fail:
613 kfree(root_item);
614 trans->block_rsv = NULL;
615 trans->bytes_reserved = 0;
616 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
617
618 if (async_transid) {
619 *async_transid = trans->transid;
620 err = btrfs_commit_transaction_async(trans, root, 1);
621 if (err)
622 err = btrfs_commit_transaction(trans, root);
623 } else {
624 err = btrfs_commit_transaction(trans, root);
625 }
626 if (err && !ret)
627 ret = err;
628
629 if (!ret) {
630 inode = btrfs_lookup_dentry(dir, dentry);
631 if (IS_ERR(inode))
632 return PTR_ERR(inode);
633 d_instantiate(dentry, inode);
634 }
635 return ret;
636
637 fail_free:
638 kfree(root_item);
639 return ret;
640 }
641
642 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
643 {
644 s64 writers;
645 DEFINE_WAIT(wait);
646
647 do {
648 prepare_to_wait(&root->subv_writers->wait, &wait,
649 TASK_UNINTERRUPTIBLE);
650
651 writers = percpu_counter_sum(&root->subv_writers->counter);
652 if (writers)
653 schedule();
654
655 finish_wait(&root->subv_writers->wait, &wait);
656 } while (writers);
657 }
658
659 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
660 struct dentry *dentry, char *name, int namelen,
661 u64 *async_transid, bool readonly,
662 struct btrfs_qgroup_inherit *inherit)
663 {
664 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
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(&fs_info->trans_lock);
725 list_add(&pending_snapshot->list,
726 &trans->transaction->pending_snapshots);
727 spin_unlock(&fs_info->trans_lock);
728 if (async_transid) {
729 *async_transid = trans->transid;
730 ret = btrfs_commit_transaction_async(trans,
731 fs_info->extent_root, 1);
732 if (ret)
733 ret = btrfs_commit_transaction(trans, root);
734 } else {
735 ret = btrfs_commit_transaction(trans,
736 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 btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
847 struct dentry *dentry;
848 int error;
849
850 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
851 if (error == -EINTR)
852 return error;
853
854 dentry = lookup_one_len(name, parent->dentry, namelen);
855 error = PTR_ERR(dentry);
856 if (IS_ERR(dentry))
857 goto out_unlock;
858
859 error = btrfs_may_create(dir, dentry);
860 if (error)
861 goto out_dput;
862
863 /*
864 * even if this name doesn't exist, we may get hash collisions.
865 * check for them now when we can safely fail
866 */
867 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
868 dir->i_ino, name,
869 namelen);
870 if (error)
871 goto out_dput;
872
873 down_read(&fs_info->subvol_sem);
874
875 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
876 goto out_up_read;
877
878 if (snap_src) {
879 error = create_snapshot(snap_src, dir, dentry, name, namelen,
880 async_transid, readonly, inherit);
881 } else {
882 error = create_subvol(dir, dentry, name, namelen,
883 async_transid, inherit);
884 }
885 if (!error)
886 fsnotify_mkdir(dir, dentry);
887 out_up_read:
888 up_read(&fs_info->subvol_sem);
889 out_dput:
890 dput(dentry);
891 out_unlock:
892 inode_unlock(dir);
893 return error;
894 }
895
896 /*
897 * When we're defragging a range, we don't want to kick it off again
898 * if it is really just waiting for delalloc to send it down.
899 * If we find a nice big extent or delalloc range for the bytes in the
900 * file you want to defrag, we return 0 to let you know to skip this
901 * part of the file
902 */
903 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
904 {
905 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
906 struct extent_map *em = NULL;
907 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
908 u64 end;
909
910 read_lock(&em_tree->lock);
911 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
912 read_unlock(&em_tree->lock);
913
914 if (em) {
915 end = extent_map_end(em);
916 free_extent_map(em);
917 if (end - offset > thresh)
918 return 0;
919 }
920 /* if we already have a nice delalloc here, just stop */
921 thresh /= 2;
922 end = count_range_bits(io_tree, &offset, offset + thresh,
923 thresh, EXTENT_DELALLOC, 1);
924 if (end >= thresh)
925 return 0;
926 return 1;
927 }
928
929 /*
930 * helper function to walk through a file and find extents
931 * newer than a specific transid, and smaller than thresh.
932 *
933 * This is used by the defragging code to find new and small
934 * extents
935 */
936 static int find_new_extents(struct btrfs_root *root,
937 struct inode *inode, u64 newer_than,
938 u64 *off, u32 thresh)
939 {
940 struct btrfs_path *path;
941 struct btrfs_key min_key;
942 struct extent_buffer *leaf;
943 struct btrfs_file_extent_item *extent;
944 int type;
945 int ret;
946 u64 ino = btrfs_ino(inode);
947
948 path = btrfs_alloc_path();
949 if (!path)
950 return -ENOMEM;
951
952 min_key.objectid = ino;
953 min_key.type = BTRFS_EXTENT_DATA_KEY;
954 min_key.offset = *off;
955
956 while (1) {
957 ret = btrfs_search_forward(root, &min_key, path, newer_than);
958 if (ret != 0)
959 goto none;
960 process_slot:
961 if (min_key.objectid != ino)
962 goto none;
963 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
964 goto none;
965
966 leaf = path->nodes[0];
967 extent = btrfs_item_ptr(leaf, path->slots[0],
968 struct btrfs_file_extent_item);
969
970 type = btrfs_file_extent_type(leaf, extent);
971 if (type == BTRFS_FILE_EXTENT_REG &&
972 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
973 check_defrag_in_cache(inode, min_key.offset, thresh)) {
974 *off = min_key.offset;
975 btrfs_free_path(path);
976 return 0;
977 }
978
979 path->slots[0]++;
980 if (path->slots[0] < btrfs_header_nritems(leaf)) {
981 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
982 goto process_slot;
983 }
984
985 if (min_key.offset == (u64)-1)
986 goto none;
987
988 min_key.offset++;
989 btrfs_release_path(path);
990 }
991 none:
992 btrfs_free_path(path);
993 return -ENOENT;
994 }
995
996 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
997 {
998 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
999 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1000 struct extent_map *em;
1001 u64 len = PAGE_SIZE;
1002
1003 /*
1004 * hopefully we have this extent in the tree already, try without
1005 * the full extent lock
1006 */
1007 read_lock(&em_tree->lock);
1008 em = lookup_extent_mapping(em_tree, start, len);
1009 read_unlock(&em_tree->lock);
1010
1011 if (!em) {
1012 struct extent_state *cached = NULL;
1013 u64 end = start + len - 1;
1014
1015 /* get the big lock and read metadata off disk */
1016 lock_extent_bits(io_tree, start, end, &cached);
1017 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1018 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1019
1020 if (IS_ERR(em))
1021 return NULL;
1022 }
1023
1024 return em;
1025 }
1026
1027 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1028 {
1029 struct extent_map *next;
1030 bool ret = true;
1031
1032 /* this is the last extent */
1033 if (em->start + em->len >= i_size_read(inode))
1034 return false;
1035
1036 next = defrag_lookup_extent(inode, em->start + em->len);
1037 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1038 ret = false;
1039 else if ((em->block_start + em->block_len == next->block_start) &&
1040 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1041 ret = false;
1042
1043 free_extent_map(next);
1044 return ret;
1045 }
1046
1047 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1048 u64 *last_len, u64 *skip, u64 *defrag_end,
1049 int compress)
1050 {
1051 struct extent_map *em;
1052 int ret = 1;
1053 bool next_mergeable = true;
1054 bool prev_mergeable = true;
1055
1056 /*
1057 * make sure that once we start defragging an extent, we keep on
1058 * defragging it
1059 */
1060 if (start < *defrag_end)
1061 return 1;
1062
1063 *skip = 0;
1064
1065 em = defrag_lookup_extent(inode, start);
1066 if (!em)
1067 return 0;
1068
1069 /* this will cover holes, and inline extents */
1070 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1071 ret = 0;
1072 goto out;
1073 }
1074
1075 if (!*defrag_end)
1076 prev_mergeable = false;
1077
1078 next_mergeable = defrag_check_next_extent(inode, em);
1079 /*
1080 * we hit a real extent, if it is big or the next extent is not a
1081 * real extent, don't bother defragging it
1082 */
1083 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1084 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1085 ret = 0;
1086 out:
1087 /*
1088 * last_len ends up being a counter of how many bytes we've defragged.
1089 * every time we choose not to defrag an extent, we reset *last_len
1090 * so that the next tiny extent will force a defrag.
1091 *
1092 * The end result of this is that tiny extents before a single big
1093 * extent will force at least part of that big extent to be defragged.
1094 */
1095 if (ret) {
1096 *defrag_end = extent_map_end(em);
1097 } else {
1098 *last_len = 0;
1099 *skip = extent_map_end(em);
1100 *defrag_end = 0;
1101 }
1102
1103 free_extent_map(em);
1104 return ret;
1105 }
1106
1107 /*
1108 * it doesn't do much good to defrag one or two pages
1109 * at a time. This pulls in a nice chunk of pages
1110 * to COW and defrag.
1111 *
1112 * It also makes sure the delalloc code has enough
1113 * dirty data to avoid making new small extents as part
1114 * of the defrag
1115 *
1116 * It's a good idea to start RA on this range
1117 * before calling this.
1118 */
1119 static int cluster_pages_for_defrag(struct inode *inode,
1120 struct page **pages,
1121 unsigned long start_index,
1122 unsigned long num_pages)
1123 {
1124 unsigned long file_end;
1125 u64 isize = i_size_read(inode);
1126 u64 page_start;
1127 u64 page_end;
1128 u64 page_cnt;
1129 int ret;
1130 int i;
1131 int i_done;
1132 struct btrfs_ordered_extent *ordered;
1133 struct extent_state *cached_state = NULL;
1134 struct extent_io_tree *tree;
1135 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1136
1137 file_end = (isize - 1) >> PAGE_SHIFT;
1138 if (!isize || start_index > file_end)
1139 return 0;
1140
1141 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1142
1143 ret = btrfs_delalloc_reserve_space(inode,
1144 start_index << PAGE_SHIFT,
1145 page_cnt << PAGE_SHIFT);
1146 if (ret)
1147 return ret;
1148 i_done = 0;
1149 tree = &BTRFS_I(inode)->io_tree;
1150
1151 /* step one, lock all the pages */
1152 for (i = 0; i < page_cnt; i++) {
1153 struct page *page;
1154 again:
1155 page = find_or_create_page(inode->i_mapping,
1156 start_index + i, mask);
1157 if (!page)
1158 break;
1159
1160 page_start = page_offset(page);
1161 page_end = page_start + PAGE_SIZE - 1;
1162 while (1) {
1163 lock_extent_bits(tree, page_start, page_end,
1164 &cached_state);
1165 ordered = btrfs_lookup_ordered_extent(inode,
1166 page_start);
1167 unlock_extent_cached(tree, page_start, page_end,
1168 &cached_state, GFP_NOFS);
1169 if (!ordered)
1170 break;
1171
1172 unlock_page(page);
1173 btrfs_start_ordered_extent(inode, ordered, 1);
1174 btrfs_put_ordered_extent(ordered);
1175 lock_page(page);
1176 /*
1177 * we unlocked the page above, so we need check if
1178 * it was released or not.
1179 */
1180 if (page->mapping != inode->i_mapping) {
1181 unlock_page(page);
1182 put_page(page);
1183 goto again;
1184 }
1185 }
1186
1187 if (!PageUptodate(page)) {
1188 btrfs_readpage(NULL, page);
1189 lock_page(page);
1190 if (!PageUptodate(page)) {
1191 unlock_page(page);
1192 put_page(page);
1193 ret = -EIO;
1194 break;
1195 }
1196 }
1197
1198 if (page->mapping != inode->i_mapping) {
1199 unlock_page(page);
1200 put_page(page);
1201 goto again;
1202 }
1203
1204 pages[i] = page;
1205 i_done++;
1206 }
1207 if (!i_done || ret)
1208 goto out;
1209
1210 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1211 goto out;
1212
1213 /*
1214 * so now we have a nice long stream of locked
1215 * and up to date pages, lets wait on them
1216 */
1217 for (i = 0; i < i_done; i++)
1218 wait_on_page_writeback(pages[i]);
1219
1220 page_start = page_offset(pages[0]);
1221 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1222
1223 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1224 page_start, page_end - 1, &cached_state);
1225 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1226 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1227 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1228 &cached_state, GFP_NOFS);
1229
1230 if (i_done != page_cnt) {
1231 spin_lock(&BTRFS_I(inode)->lock);
1232 BTRFS_I(inode)->outstanding_extents++;
1233 spin_unlock(&BTRFS_I(inode)->lock);
1234 btrfs_delalloc_release_space(inode,
1235 start_index << PAGE_SHIFT,
1236 (page_cnt - i_done) << PAGE_SHIFT);
1237 }
1238
1239
1240 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1241 &cached_state);
1242
1243 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1244 page_start, page_end - 1, &cached_state,
1245 GFP_NOFS);
1246
1247 for (i = 0; i < i_done; i++) {
1248 clear_page_dirty_for_io(pages[i]);
1249 ClearPageChecked(pages[i]);
1250 set_page_extent_mapped(pages[i]);
1251 set_page_dirty(pages[i]);
1252 unlock_page(pages[i]);
1253 put_page(pages[i]);
1254 }
1255 return i_done;
1256 out:
1257 for (i = 0; i < i_done; i++) {
1258 unlock_page(pages[i]);
1259 put_page(pages[i]);
1260 }
1261 btrfs_delalloc_release_space(inode,
1262 start_index << PAGE_SHIFT,
1263 page_cnt << PAGE_SHIFT);
1264 return ret;
1265
1266 }
1267
1268 int btrfs_defrag_file(struct inode *inode, struct file *file,
1269 struct btrfs_ioctl_defrag_range_args *range,
1270 u64 newer_than, unsigned long max_to_defrag)
1271 {
1272 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1273 struct btrfs_root *root = BTRFS_I(inode)->root;
1274 struct file_ra_state *ra = NULL;
1275 unsigned long last_index;
1276 u64 isize = i_size_read(inode);
1277 u64 last_len = 0;
1278 u64 skip = 0;
1279 u64 defrag_end = 0;
1280 u64 newer_off = range->start;
1281 unsigned long i;
1282 unsigned long ra_index = 0;
1283 int ret;
1284 int defrag_count = 0;
1285 int compress_type = BTRFS_COMPRESS_ZLIB;
1286 u32 extent_thresh = range->extent_thresh;
1287 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1288 unsigned long cluster = max_cluster;
1289 u64 new_align = ~((u64)SZ_128K - 1);
1290 struct page **pages = NULL;
1291
1292 if (isize == 0)
1293 return 0;
1294
1295 if (range->start >= isize)
1296 return -EINVAL;
1297
1298 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1299 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1300 return -EINVAL;
1301 if (range->compress_type)
1302 compress_type = range->compress_type;
1303 }
1304
1305 if (extent_thresh == 0)
1306 extent_thresh = SZ_256K;
1307
1308 /*
1309 * if we were not given a file, allocate a readahead
1310 * context
1311 */
1312 if (!file) {
1313 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1314 if (!ra)
1315 return -ENOMEM;
1316 file_ra_state_init(ra, inode->i_mapping);
1317 } else {
1318 ra = &file->f_ra;
1319 }
1320
1321 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1322 GFP_NOFS);
1323 if (!pages) {
1324 ret = -ENOMEM;
1325 goto out_ra;
1326 }
1327
1328 /* find the last page to defrag */
1329 if (range->start + range->len > range->start) {
1330 last_index = min_t(u64, isize - 1,
1331 range->start + range->len - 1) >> PAGE_SHIFT;
1332 } else {
1333 last_index = (isize - 1) >> PAGE_SHIFT;
1334 }
1335
1336 if (newer_than) {
1337 ret = find_new_extents(root, inode, newer_than,
1338 &newer_off, SZ_64K);
1339 if (!ret) {
1340 range->start = newer_off;
1341 /*
1342 * we always align our defrag to help keep
1343 * the extents in the file evenly spaced
1344 */
1345 i = (newer_off & new_align) >> PAGE_SHIFT;
1346 } else
1347 goto out_ra;
1348 } else {
1349 i = range->start >> PAGE_SHIFT;
1350 }
1351 if (!max_to_defrag)
1352 max_to_defrag = last_index - i + 1;
1353
1354 /*
1355 * make writeback starts from i, so the defrag range can be
1356 * written sequentially.
1357 */
1358 if (i < inode->i_mapping->writeback_index)
1359 inode->i_mapping->writeback_index = i;
1360
1361 while (i <= last_index && defrag_count < max_to_defrag &&
1362 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1363 /*
1364 * make sure we stop running if someone unmounts
1365 * the FS
1366 */
1367 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1368 break;
1369
1370 if (btrfs_defrag_cancelled(fs_info)) {
1371 btrfs_debug(fs_info, "defrag_file cancelled");
1372 ret = -EAGAIN;
1373 break;
1374 }
1375
1376 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1377 extent_thresh, &last_len, &skip,
1378 &defrag_end, range->flags &
1379 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1380 unsigned long next;
1381 /*
1382 * the should_defrag function tells us how much to skip
1383 * bump our counter by the suggested amount
1384 */
1385 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1386 i = max(i + 1, next);
1387 continue;
1388 }
1389
1390 if (!newer_than) {
1391 cluster = (PAGE_ALIGN(defrag_end) >>
1392 PAGE_SHIFT) - i;
1393 cluster = min(cluster, max_cluster);
1394 } else {
1395 cluster = max_cluster;
1396 }
1397
1398 if (i + cluster > ra_index) {
1399 ra_index = max(i, ra_index);
1400 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1401 cluster);
1402 ra_index += cluster;
1403 }
1404
1405 inode_lock(inode);
1406 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1407 BTRFS_I(inode)->force_compress = compress_type;
1408 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1409 if (ret < 0) {
1410 inode_unlock(inode);
1411 goto out_ra;
1412 }
1413
1414 defrag_count += ret;
1415 balance_dirty_pages_ratelimited(inode->i_mapping);
1416 inode_unlock(inode);
1417
1418 if (newer_than) {
1419 if (newer_off == (u64)-1)
1420 break;
1421
1422 if (ret > 0)
1423 i += ret;
1424
1425 newer_off = max(newer_off + 1,
1426 (u64)i << PAGE_SHIFT);
1427
1428 ret = find_new_extents(root, inode, newer_than,
1429 &newer_off, SZ_64K);
1430 if (!ret) {
1431 range->start = newer_off;
1432 i = (newer_off & new_align) >> PAGE_SHIFT;
1433 } else {
1434 break;
1435 }
1436 } else {
1437 if (ret > 0) {
1438 i += ret;
1439 last_len += ret << PAGE_SHIFT;
1440 } else {
1441 i++;
1442 last_len = 0;
1443 }
1444 }
1445 }
1446
1447 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1448 filemap_flush(inode->i_mapping);
1449 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1450 &BTRFS_I(inode)->runtime_flags))
1451 filemap_flush(inode->i_mapping);
1452 }
1453
1454 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1455 /* the filemap_flush will queue IO into the worker threads, but
1456 * we have to make sure the IO is actually started and that
1457 * ordered extents get created before we return
1458 */
1459 atomic_inc(&fs_info->async_submit_draining);
1460 while (atomic_read(&fs_info->nr_async_submits) ||
1461 atomic_read(&fs_info->async_delalloc_pages)) {
1462 wait_event(fs_info->async_submit_wait,
1463 (atomic_read(&fs_info->nr_async_submits) == 0 &&
1464 atomic_read(&fs_info->async_delalloc_pages) == 0));
1465 }
1466 atomic_dec(&fs_info->async_submit_draining);
1467 }
1468
1469 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1470 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1471 }
1472
1473 ret = defrag_count;
1474
1475 out_ra:
1476 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1477 inode_lock(inode);
1478 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1479 inode_unlock(inode);
1480 }
1481 if (!file)
1482 kfree(ra);
1483 kfree(pages);
1484 return ret;
1485 }
1486
1487 static noinline int btrfs_ioctl_resize(struct file *file,
1488 void __user *arg)
1489 {
1490 struct inode *inode = file_inode(file);
1491 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1492 u64 new_size;
1493 u64 old_size;
1494 u64 devid = 1;
1495 struct btrfs_root *root = BTRFS_I(inode)->root;
1496 struct btrfs_ioctl_vol_args *vol_args;
1497 struct btrfs_trans_handle *trans;
1498 struct btrfs_device *device = NULL;
1499 char *sizestr;
1500 char *retptr;
1501 char *devstr = NULL;
1502 int ret = 0;
1503 int mod = 0;
1504
1505 if (!capable(CAP_SYS_ADMIN))
1506 return -EPERM;
1507
1508 ret = mnt_want_write_file(file);
1509 if (ret)
1510 return ret;
1511
1512 if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
1513 mnt_drop_write_file(file);
1514 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1515 }
1516
1517 mutex_lock(&fs_info->volume_mutex);
1518 vol_args = memdup_user(arg, sizeof(*vol_args));
1519 if (IS_ERR(vol_args)) {
1520 ret = PTR_ERR(vol_args);
1521 goto out;
1522 }
1523
1524 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1525
1526 sizestr = vol_args->name;
1527 devstr = strchr(sizestr, ':');
1528 if (devstr) {
1529 sizestr = devstr + 1;
1530 *devstr = '\0';
1531 devstr = vol_args->name;
1532 ret = kstrtoull(devstr, 10, &devid);
1533 if (ret)
1534 goto out_free;
1535 if (!devid) {
1536 ret = -EINVAL;
1537 goto out_free;
1538 }
1539 btrfs_info(fs_info, "resizing devid %llu", devid);
1540 }
1541
1542 device = btrfs_find_device(fs_info, devid, NULL, NULL);
1543 if (!device) {
1544 btrfs_info(fs_info, "resizer unable to find device %llu",
1545 devid);
1546 ret = -ENODEV;
1547 goto out_free;
1548 }
1549
1550 if (!device->writeable) {
1551 btrfs_info(fs_info,
1552 "resizer unable to apply on readonly device %llu",
1553 devid);
1554 ret = -EPERM;
1555 goto out_free;
1556 }
1557
1558 if (!strcmp(sizestr, "max"))
1559 new_size = device->bdev->bd_inode->i_size;
1560 else {
1561 if (sizestr[0] == '-') {
1562 mod = -1;
1563 sizestr++;
1564 } else if (sizestr[0] == '+') {
1565 mod = 1;
1566 sizestr++;
1567 }
1568 new_size = memparse(sizestr, &retptr);
1569 if (*retptr != '\0' || new_size == 0) {
1570 ret = -EINVAL;
1571 goto out_free;
1572 }
1573 }
1574
1575 if (device->is_tgtdev_for_dev_replace) {
1576 ret = -EPERM;
1577 goto out_free;
1578 }
1579
1580 old_size = btrfs_device_get_total_bytes(device);
1581
1582 if (mod < 0) {
1583 if (new_size > old_size) {
1584 ret = -EINVAL;
1585 goto out_free;
1586 }
1587 new_size = old_size - new_size;
1588 } else if (mod > 0) {
1589 if (new_size > ULLONG_MAX - old_size) {
1590 ret = -ERANGE;
1591 goto out_free;
1592 }
1593 new_size = old_size + new_size;
1594 }
1595
1596 if (new_size < SZ_256M) {
1597 ret = -EINVAL;
1598 goto out_free;
1599 }
1600 if (new_size > device->bdev->bd_inode->i_size) {
1601 ret = -EFBIG;
1602 goto out_free;
1603 }
1604
1605 new_size = div_u64(new_size, fs_info->sectorsize);
1606 new_size *= fs_info->sectorsize;
1607
1608 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1609 rcu_str_deref(device->name), new_size);
1610
1611 if (new_size > old_size) {
1612 trans = btrfs_start_transaction(root, 0);
1613 if (IS_ERR(trans)) {
1614 ret = PTR_ERR(trans);
1615 goto out_free;
1616 }
1617 ret = btrfs_grow_device(trans, device, new_size);
1618 btrfs_commit_transaction(trans, root);
1619 } else if (new_size < old_size) {
1620 ret = btrfs_shrink_device(device, new_size);
1621 } /* equal, nothing need to do */
1622
1623 out_free:
1624 kfree(vol_args);
1625 out:
1626 mutex_unlock(&fs_info->volume_mutex);
1627 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
1628 mnt_drop_write_file(file);
1629 return ret;
1630 }
1631
1632 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1633 char *name, unsigned long fd, int subvol,
1634 u64 *transid, bool readonly,
1635 struct btrfs_qgroup_inherit *inherit)
1636 {
1637 int namelen;
1638 int ret = 0;
1639
1640 if (!S_ISDIR(file_inode(file)->i_mode))
1641 return -ENOTDIR;
1642
1643 ret = mnt_want_write_file(file);
1644 if (ret)
1645 goto out;
1646
1647 namelen = strlen(name);
1648 if (strchr(name, '/')) {
1649 ret = -EINVAL;
1650 goto out_drop_write;
1651 }
1652
1653 if (name[0] == '.' &&
1654 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1655 ret = -EEXIST;
1656 goto out_drop_write;
1657 }
1658
1659 if (subvol) {
1660 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1661 NULL, transid, readonly, inherit);
1662 } else {
1663 struct fd src = fdget(fd);
1664 struct inode *src_inode;
1665 if (!src.file) {
1666 ret = -EINVAL;
1667 goto out_drop_write;
1668 }
1669
1670 src_inode = file_inode(src.file);
1671 if (src_inode->i_sb != file_inode(file)->i_sb) {
1672 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1673 "Snapshot src from another FS");
1674 ret = -EXDEV;
1675 } else if (!inode_owner_or_capable(src_inode)) {
1676 /*
1677 * Subvolume creation is not restricted, but snapshots
1678 * are limited to own subvolumes only
1679 */
1680 ret = -EPERM;
1681 } else {
1682 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1683 BTRFS_I(src_inode)->root,
1684 transid, readonly, inherit);
1685 }
1686 fdput(src);
1687 }
1688 out_drop_write:
1689 mnt_drop_write_file(file);
1690 out:
1691 return ret;
1692 }
1693
1694 static noinline int btrfs_ioctl_snap_create(struct file *file,
1695 void __user *arg, int subvol)
1696 {
1697 struct btrfs_ioctl_vol_args *vol_args;
1698 int ret;
1699
1700 if (!S_ISDIR(file_inode(file)->i_mode))
1701 return -ENOTDIR;
1702
1703 vol_args = memdup_user(arg, sizeof(*vol_args));
1704 if (IS_ERR(vol_args))
1705 return PTR_ERR(vol_args);
1706 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1707
1708 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1709 vol_args->fd, subvol,
1710 NULL, false, NULL);
1711
1712 kfree(vol_args);
1713 return ret;
1714 }
1715
1716 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1717 void __user *arg, int subvol)
1718 {
1719 struct btrfs_ioctl_vol_args_v2 *vol_args;
1720 int ret;
1721 u64 transid = 0;
1722 u64 *ptr = NULL;
1723 bool readonly = false;
1724 struct btrfs_qgroup_inherit *inherit = NULL;
1725
1726 if (!S_ISDIR(file_inode(file)->i_mode))
1727 return -ENOTDIR;
1728
1729 vol_args = memdup_user(arg, sizeof(*vol_args));
1730 if (IS_ERR(vol_args))
1731 return PTR_ERR(vol_args);
1732 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1733
1734 if (vol_args->flags &
1735 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1736 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1737 ret = -EOPNOTSUPP;
1738 goto free_args;
1739 }
1740
1741 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1742 ptr = &transid;
1743 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1744 readonly = true;
1745 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1746 if (vol_args->size > PAGE_SIZE) {
1747 ret = -EINVAL;
1748 goto free_args;
1749 }
1750 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1751 if (IS_ERR(inherit)) {
1752 ret = PTR_ERR(inherit);
1753 goto free_args;
1754 }
1755 }
1756
1757 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1758 vol_args->fd, subvol, ptr,
1759 readonly, inherit);
1760 if (ret)
1761 goto free_inherit;
1762
1763 if (ptr && copy_to_user(arg +
1764 offsetof(struct btrfs_ioctl_vol_args_v2,
1765 transid),
1766 ptr, sizeof(*ptr)))
1767 ret = -EFAULT;
1768
1769 free_inherit:
1770 kfree(inherit);
1771 free_args:
1772 kfree(vol_args);
1773 return ret;
1774 }
1775
1776 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1777 void __user *arg)
1778 {
1779 struct inode *inode = file_inode(file);
1780 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1781 struct btrfs_root *root = BTRFS_I(inode)->root;
1782 int ret = 0;
1783 u64 flags = 0;
1784
1785 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1786 return -EINVAL;
1787
1788 down_read(&fs_info->subvol_sem);
1789 if (btrfs_root_readonly(root))
1790 flags |= BTRFS_SUBVOL_RDONLY;
1791 up_read(&fs_info->subvol_sem);
1792
1793 if (copy_to_user(arg, &flags, sizeof(flags)))
1794 ret = -EFAULT;
1795
1796 return ret;
1797 }
1798
1799 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1800 void __user *arg)
1801 {
1802 struct inode *inode = file_inode(file);
1803 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1804 struct btrfs_root *root = BTRFS_I(inode)->root;
1805 struct btrfs_trans_handle *trans;
1806 u64 root_flags;
1807 u64 flags;
1808 int ret = 0;
1809
1810 if (!inode_owner_or_capable(inode))
1811 return -EPERM;
1812
1813 ret = mnt_want_write_file(file);
1814 if (ret)
1815 goto out;
1816
1817 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1818 ret = -EINVAL;
1819 goto out_drop_write;
1820 }
1821
1822 if (copy_from_user(&flags, arg, sizeof(flags))) {
1823 ret = -EFAULT;
1824 goto out_drop_write;
1825 }
1826
1827 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1828 ret = -EINVAL;
1829 goto out_drop_write;
1830 }
1831
1832 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1833 ret = -EOPNOTSUPP;
1834 goto out_drop_write;
1835 }
1836
1837 down_write(&fs_info->subvol_sem);
1838
1839 /* nothing to do */
1840 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1841 goto out_drop_sem;
1842
1843 root_flags = btrfs_root_flags(&root->root_item);
1844 if (flags & BTRFS_SUBVOL_RDONLY) {
1845 btrfs_set_root_flags(&root->root_item,
1846 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1847 } else {
1848 /*
1849 * Block RO -> RW transition if this subvolume is involved in
1850 * send
1851 */
1852 spin_lock(&root->root_item_lock);
1853 if (root->send_in_progress == 0) {
1854 btrfs_set_root_flags(&root->root_item,
1855 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1856 spin_unlock(&root->root_item_lock);
1857 } else {
1858 spin_unlock(&root->root_item_lock);
1859 btrfs_warn(fs_info,
1860 "Attempt to set subvolume %llu read-write during send",
1861 root->root_key.objectid);
1862 ret = -EPERM;
1863 goto out_drop_sem;
1864 }
1865 }
1866
1867 trans = btrfs_start_transaction(root, 1);
1868 if (IS_ERR(trans)) {
1869 ret = PTR_ERR(trans);
1870 goto out_reset;
1871 }
1872
1873 ret = btrfs_update_root(trans, fs_info->tree_root,
1874 &root->root_key, &root->root_item);
1875
1876 btrfs_commit_transaction(trans, root);
1877 out_reset:
1878 if (ret)
1879 btrfs_set_root_flags(&root->root_item, root_flags);
1880 out_drop_sem:
1881 up_write(&fs_info->subvol_sem);
1882 out_drop_write:
1883 mnt_drop_write_file(file);
1884 out:
1885 return ret;
1886 }
1887
1888 /*
1889 * helper to check if the subvolume references other subvolumes
1890 */
1891 static noinline int may_destroy_subvol(struct btrfs_root *root)
1892 {
1893 struct btrfs_fs_info *fs_info = root->fs_info;
1894 struct btrfs_path *path;
1895 struct btrfs_dir_item *di;
1896 struct btrfs_key key;
1897 u64 dir_id;
1898 int ret;
1899
1900 path = btrfs_alloc_path();
1901 if (!path)
1902 return -ENOMEM;
1903
1904 /* Make sure this root isn't set as the default subvol */
1905 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1906 di = btrfs_lookup_dir_item(NULL, fs_info->tree_root, path,
1907 dir_id, "default", 7, 0);
1908 if (di && !IS_ERR(di)) {
1909 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1910 if (key.objectid == root->root_key.objectid) {
1911 ret = -EPERM;
1912 btrfs_err(fs_info,
1913 "deleting default subvolume %llu is not allowed",
1914 key.objectid);
1915 goto out;
1916 }
1917 btrfs_release_path(path);
1918 }
1919
1920 key.objectid = root->root_key.objectid;
1921 key.type = BTRFS_ROOT_REF_KEY;
1922 key.offset = (u64)-1;
1923
1924 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
1925 if (ret < 0)
1926 goto out;
1927 BUG_ON(ret == 0);
1928
1929 ret = 0;
1930 if (path->slots[0] > 0) {
1931 path->slots[0]--;
1932 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1933 if (key.objectid == root->root_key.objectid &&
1934 key.type == BTRFS_ROOT_REF_KEY)
1935 ret = -ENOTEMPTY;
1936 }
1937 out:
1938 btrfs_free_path(path);
1939 return ret;
1940 }
1941
1942 static noinline int key_in_sk(struct btrfs_key *key,
1943 struct btrfs_ioctl_search_key *sk)
1944 {
1945 struct btrfs_key test;
1946 int ret;
1947
1948 test.objectid = sk->min_objectid;
1949 test.type = sk->min_type;
1950 test.offset = sk->min_offset;
1951
1952 ret = btrfs_comp_cpu_keys(key, &test);
1953 if (ret < 0)
1954 return 0;
1955
1956 test.objectid = sk->max_objectid;
1957 test.type = sk->max_type;
1958 test.offset = sk->max_offset;
1959
1960 ret = btrfs_comp_cpu_keys(key, &test);
1961 if (ret > 0)
1962 return 0;
1963 return 1;
1964 }
1965
1966 static noinline int copy_to_sk(struct btrfs_path *path,
1967 struct btrfs_key *key,
1968 struct btrfs_ioctl_search_key *sk,
1969 size_t *buf_size,
1970 char __user *ubuf,
1971 unsigned long *sk_offset,
1972 int *num_found)
1973 {
1974 u64 found_transid;
1975 struct extent_buffer *leaf;
1976 struct btrfs_ioctl_search_header sh;
1977 struct btrfs_key test;
1978 unsigned long item_off;
1979 unsigned long item_len;
1980 int nritems;
1981 int i;
1982 int slot;
1983 int ret = 0;
1984
1985 leaf = path->nodes[0];
1986 slot = path->slots[0];
1987 nritems = btrfs_header_nritems(leaf);
1988
1989 if (btrfs_header_generation(leaf) > sk->max_transid) {
1990 i = nritems;
1991 goto advance_key;
1992 }
1993 found_transid = btrfs_header_generation(leaf);
1994
1995 for (i = slot; i < nritems; i++) {
1996 item_off = btrfs_item_ptr_offset(leaf, i);
1997 item_len = btrfs_item_size_nr(leaf, i);
1998
1999 btrfs_item_key_to_cpu(leaf, key, i);
2000 if (!key_in_sk(key, sk))
2001 continue;
2002
2003 if (sizeof(sh) + item_len > *buf_size) {
2004 if (*num_found) {
2005 ret = 1;
2006 goto out;
2007 }
2008
2009 /*
2010 * return one empty item back for v1, which does not
2011 * handle -EOVERFLOW
2012 */
2013
2014 *buf_size = sizeof(sh) + item_len;
2015 item_len = 0;
2016 ret = -EOVERFLOW;
2017 }
2018
2019 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2020 ret = 1;
2021 goto out;
2022 }
2023
2024 sh.objectid = key->objectid;
2025 sh.offset = key->offset;
2026 sh.type = key->type;
2027 sh.len = item_len;
2028 sh.transid = found_transid;
2029
2030 /* copy search result header */
2031 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2032 ret = -EFAULT;
2033 goto out;
2034 }
2035
2036 *sk_offset += sizeof(sh);
2037
2038 if (item_len) {
2039 char __user *up = ubuf + *sk_offset;
2040 /* copy the item */
2041 if (read_extent_buffer_to_user(leaf, up,
2042 item_off, item_len)) {
2043 ret = -EFAULT;
2044 goto out;
2045 }
2046
2047 *sk_offset += item_len;
2048 }
2049 (*num_found)++;
2050
2051 if (ret) /* -EOVERFLOW from above */
2052 goto out;
2053
2054 if (*num_found >= sk->nr_items) {
2055 ret = 1;
2056 goto out;
2057 }
2058 }
2059 advance_key:
2060 ret = 0;
2061 test.objectid = sk->max_objectid;
2062 test.type = sk->max_type;
2063 test.offset = sk->max_offset;
2064 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2065 ret = 1;
2066 else if (key->offset < (u64)-1)
2067 key->offset++;
2068 else if (key->type < (u8)-1) {
2069 key->offset = 0;
2070 key->type++;
2071 } else if (key->objectid < (u64)-1) {
2072 key->offset = 0;
2073 key->type = 0;
2074 key->objectid++;
2075 } else
2076 ret = 1;
2077 out:
2078 /*
2079 * 0: all items from this leaf copied, continue with next
2080 * 1: * more items can be copied, but unused buffer is too small
2081 * * all items were found
2082 * Either way, it will stops the loop which iterates to the next
2083 * leaf
2084 * -EOVERFLOW: item was to large for buffer
2085 * -EFAULT: could not copy extent buffer back to userspace
2086 */
2087 return ret;
2088 }
2089
2090 static noinline int search_ioctl(struct inode *inode,
2091 struct btrfs_ioctl_search_key *sk,
2092 size_t *buf_size,
2093 char __user *ubuf)
2094 {
2095 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2096 struct btrfs_root *root;
2097 struct btrfs_key key;
2098 struct btrfs_path *path;
2099 int ret;
2100 int num_found = 0;
2101 unsigned long sk_offset = 0;
2102
2103 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2104 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2105 return -EOVERFLOW;
2106 }
2107
2108 path = btrfs_alloc_path();
2109 if (!path)
2110 return -ENOMEM;
2111
2112 if (sk->tree_id == 0) {
2113 /* search the root of the inode that was passed */
2114 root = BTRFS_I(inode)->root;
2115 } else {
2116 key.objectid = sk->tree_id;
2117 key.type = BTRFS_ROOT_ITEM_KEY;
2118 key.offset = (u64)-1;
2119 root = btrfs_read_fs_root_no_name(info, &key);
2120 if (IS_ERR(root)) {
2121 btrfs_free_path(path);
2122 return -ENOENT;
2123 }
2124 }
2125
2126 key.objectid = sk->min_objectid;
2127 key.type = sk->min_type;
2128 key.offset = sk->min_offset;
2129
2130 while (1) {
2131 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2132 if (ret != 0) {
2133 if (ret > 0)
2134 ret = 0;
2135 goto err;
2136 }
2137 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2138 &sk_offset, &num_found);
2139 btrfs_release_path(path);
2140 if (ret)
2141 break;
2142
2143 }
2144 if (ret > 0)
2145 ret = 0;
2146 err:
2147 sk->nr_items = num_found;
2148 btrfs_free_path(path);
2149 return ret;
2150 }
2151
2152 static noinline int btrfs_ioctl_tree_search(struct file *file,
2153 void __user *argp)
2154 {
2155 struct btrfs_ioctl_search_args __user *uargs;
2156 struct btrfs_ioctl_search_key sk;
2157 struct inode *inode;
2158 int ret;
2159 size_t buf_size;
2160
2161 if (!capable(CAP_SYS_ADMIN))
2162 return -EPERM;
2163
2164 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2165
2166 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2167 return -EFAULT;
2168
2169 buf_size = sizeof(uargs->buf);
2170
2171 inode = file_inode(file);
2172 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2173
2174 /*
2175 * In the origin implementation an overflow is handled by returning a
2176 * search header with a len of zero, so reset ret.
2177 */
2178 if (ret == -EOVERFLOW)
2179 ret = 0;
2180
2181 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2182 ret = -EFAULT;
2183 return ret;
2184 }
2185
2186 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2187 void __user *argp)
2188 {
2189 struct btrfs_ioctl_search_args_v2 __user *uarg;
2190 struct btrfs_ioctl_search_args_v2 args;
2191 struct inode *inode;
2192 int ret;
2193 size_t buf_size;
2194 const size_t buf_limit = SZ_16M;
2195
2196 if (!capable(CAP_SYS_ADMIN))
2197 return -EPERM;
2198
2199 /* copy search header and buffer size */
2200 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2201 if (copy_from_user(&args, uarg, sizeof(args)))
2202 return -EFAULT;
2203
2204 buf_size = args.buf_size;
2205
2206 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2207 return -EOVERFLOW;
2208
2209 /* limit result size to 16MB */
2210 if (buf_size > buf_limit)
2211 buf_size = buf_limit;
2212
2213 inode = file_inode(file);
2214 ret = search_ioctl(inode, &args.key, &buf_size,
2215 (char *)(&uarg->buf[0]));
2216 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2217 ret = -EFAULT;
2218 else if (ret == -EOVERFLOW &&
2219 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2220 ret = -EFAULT;
2221
2222 return ret;
2223 }
2224
2225 /*
2226 * Search INODE_REFs to identify path name of 'dirid' directory
2227 * in a 'tree_id' tree. and sets path name to 'name'.
2228 */
2229 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2230 u64 tree_id, u64 dirid, char *name)
2231 {
2232 struct btrfs_root *root;
2233 struct btrfs_key key;
2234 char *ptr;
2235 int ret = -1;
2236 int slot;
2237 int len;
2238 int total_len = 0;
2239 struct btrfs_inode_ref *iref;
2240 struct extent_buffer *l;
2241 struct btrfs_path *path;
2242
2243 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2244 name[0]='\0';
2245 return 0;
2246 }
2247
2248 path = btrfs_alloc_path();
2249 if (!path)
2250 return -ENOMEM;
2251
2252 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2253
2254 key.objectid = tree_id;
2255 key.type = BTRFS_ROOT_ITEM_KEY;
2256 key.offset = (u64)-1;
2257 root = btrfs_read_fs_root_no_name(info, &key);
2258 if (IS_ERR(root)) {
2259 btrfs_err(info, "could not find root %llu", tree_id);
2260 ret = -ENOENT;
2261 goto out;
2262 }
2263
2264 key.objectid = dirid;
2265 key.type = BTRFS_INODE_REF_KEY;
2266 key.offset = (u64)-1;
2267
2268 while (1) {
2269 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2270 if (ret < 0)
2271 goto out;
2272 else if (ret > 0) {
2273 ret = btrfs_previous_item(root, path, dirid,
2274 BTRFS_INODE_REF_KEY);
2275 if (ret < 0)
2276 goto out;
2277 else if (ret > 0) {
2278 ret = -ENOENT;
2279 goto out;
2280 }
2281 }
2282
2283 l = path->nodes[0];
2284 slot = path->slots[0];
2285 btrfs_item_key_to_cpu(l, &key, slot);
2286
2287 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2288 len = btrfs_inode_ref_name_len(l, iref);
2289 ptr -= len + 1;
2290 total_len += len + 1;
2291 if (ptr < name) {
2292 ret = -ENAMETOOLONG;
2293 goto out;
2294 }
2295
2296 *(ptr + len) = '/';
2297 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2298
2299 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2300 break;
2301
2302 btrfs_release_path(path);
2303 key.objectid = key.offset;
2304 key.offset = (u64)-1;
2305 dirid = key.objectid;
2306 }
2307 memmove(name, ptr, total_len);
2308 name[total_len] = '\0';
2309 ret = 0;
2310 out:
2311 btrfs_free_path(path);
2312 return ret;
2313 }
2314
2315 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2316 void __user *argp)
2317 {
2318 struct btrfs_ioctl_ino_lookup_args *args;
2319 struct inode *inode;
2320 int ret = 0;
2321
2322 args = memdup_user(argp, sizeof(*args));
2323 if (IS_ERR(args))
2324 return PTR_ERR(args);
2325
2326 inode = file_inode(file);
2327
2328 /*
2329 * Unprivileged query to obtain the containing subvolume root id. The
2330 * path is reset so it's consistent with btrfs_search_path_in_tree.
2331 */
2332 if (args->treeid == 0)
2333 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2334
2335 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2336 args->name[0] = 0;
2337 goto out;
2338 }
2339
2340 if (!capable(CAP_SYS_ADMIN)) {
2341 ret = -EPERM;
2342 goto out;
2343 }
2344
2345 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2346 args->treeid, args->objectid,
2347 args->name);
2348
2349 out:
2350 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2351 ret = -EFAULT;
2352
2353 kfree(args);
2354 return ret;
2355 }
2356
2357 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2358 void __user *arg)
2359 {
2360 struct dentry *parent = file->f_path.dentry;
2361 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2362 struct dentry *dentry;
2363 struct inode *dir = d_inode(parent);
2364 struct inode *inode;
2365 struct btrfs_root *root = BTRFS_I(dir)->root;
2366 struct btrfs_root *dest = NULL;
2367 struct btrfs_ioctl_vol_args *vol_args;
2368 struct btrfs_trans_handle *trans;
2369 struct btrfs_block_rsv block_rsv;
2370 u64 root_flags;
2371 u64 qgroup_reserved;
2372 int namelen;
2373 int ret;
2374 int err = 0;
2375
2376 if (!S_ISDIR(dir->i_mode))
2377 return -ENOTDIR;
2378
2379 vol_args = memdup_user(arg, sizeof(*vol_args));
2380 if (IS_ERR(vol_args))
2381 return PTR_ERR(vol_args);
2382
2383 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2384 namelen = strlen(vol_args->name);
2385 if (strchr(vol_args->name, '/') ||
2386 strncmp(vol_args->name, "..", namelen) == 0) {
2387 err = -EINVAL;
2388 goto out;
2389 }
2390
2391 err = mnt_want_write_file(file);
2392 if (err)
2393 goto out;
2394
2395
2396 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2397 if (err == -EINTR)
2398 goto out_drop_write;
2399 dentry = lookup_one_len(vol_args->name, parent, namelen);
2400 if (IS_ERR(dentry)) {
2401 err = PTR_ERR(dentry);
2402 goto out_unlock_dir;
2403 }
2404
2405 if (d_really_is_negative(dentry)) {
2406 err = -ENOENT;
2407 goto out_dput;
2408 }
2409
2410 inode = d_inode(dentry);
2411 dest = BTRFS_I(inode)->root;
2412 if (!capable(CAP_SYS_ADMIN)) {
2413 /*
2414 * Regular user. Only allow this with a special mount
2415 * option, when the user has write+exec access to the
2416 * subvol root, and when rmdir(2) would have been
2417 * allowed.
2418 *
2419 * Note that this is _not_ check that the subvol is
2420 * empty or doesn't contain data that we wouldn't
2421 * otherwise be able to delete.
2422 *
2423 * Users who want to delete empty subvols should try
2424 * rmdir(2).
2425 */
2426 err = -EPERM;
2427 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2428 goto out_dput;
2429
2430 /*
2431 * Do not allow deletion if the parent dir is the same
2432 * as the dir to be deleted. That means the ioctl
2433 * must be called on the dentry referencing the root
2434 * of the subvol, not a random directory contained
2435 * within it.
2436 */
2437 err = -EINVAL;
2438 if (root == dest)
2439 goto out_dput;
2440
2441 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2442 if (err)
2443 goto out_dput;
2444 }
2445
2446 /* check if subvolume may be deleted by a user */
2447 err = btrfs_may_delete(dir, dentry, 1);
2448 if (err)
2449 goto out_dput;
2450
2451 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2452 err = -EINVAL;
2453 goto out_dput;
2454 }
2455
2456 inode_lock(inode);
2457
2458 /*
2459 * Don't allow to delete a subvolume with send in progress. This is
2460 * inside the i_mutex so the error handling that has to drop the bit
2461 * again is not run concurrently.
2462 */
2463 spin_lock(&dest->root_item_lock);
2464 root_flags = btrfs_root_flags(&dest->root_item);
2465 if (dest->send_in_progress == 0) {
2466 btrfs_set_root_flags(&dest->root_item,
2467 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2468 spin_unlock(&dest->root_item_lock);
2469 } else {
2470 spin_unlock(&dest->root_item_lock);
2471 btrfs_warn(fs_info,
2472 "Attempt to delete subvolume %llu during send",
2473 dest->root_key.objectid);
2474 err = -EPERM;
2475 goto out_unlock_inode;
2476 }
2477
2478 down_write(&fs_info->subvol_sem);
2479
2480 err = may_destroy_subvol(dest);
2481 if (err)
2482 goto out_up_write;
2483
2484 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2485 /*
2486 * One for dir inode, two for dir entries, two for root
2487 * ref/backref.
2488 */
2489 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2490 5, &qgroup_reserved, true);
2491 if (err)
2492 goto out_up_write;
2493
2494 trans = btrfs_start_transaction(root, 0);
2495 if (IS_ERR(trans)) {
2496 err = PTR_ERR(trans);
2497 goto out_release;
2498 }
2499 trans->block_rsv = &block_rsv;
2500 trans->bytes_reserved = block_rsv.size;
2501
2502 btrfs_record_snapshot_destroy(trans, dir);
2503
2504 ret = btrfs_unlink_subvol(trans, root, dir,
2505 dest->root_key.objectid,
2506 dentry->d_name.name,
2507 dentry->d_name.len);
2508 if (ret) {
2509 err = ret;
2510 btrfs_abort_transaction(trans, ret);
2511 goto out_end_trans;
2512 }
2513
2514 btrfs_record_root_in_trans(trans, dest);
2515
2516 memset(&dest->root_item.drop_progress, 0,
2517 sizeof(dest->root_item.drop_progress));
2518 dest->root_item.drop_level = 0;
2519 btrfs_set_root_refs(&dest->root_item, 0);
2520
2521 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2522 ret = btrfs_insert_orphan_item(trans,
2523 fs_info->tree_root,
2524 dest->root_key.objectid);
2525 if (ret) {
2526 btrfs_abort_transaction(trans, ret);
2527 err = ret;
2528 goto out_end_trans;
2529 }
2530 }
2531
2532 ret = btrfs_uuid_tree_rem(trans, fs_info, dest->root_item.uuid,
2533 BTRFS_UUID_KEY_SUBVOL,
2534 dest->root_key.objectid);
2535 if (ret && ret != -ENOENT) {
2536 btrfs_abort_transaction(trans, ret);
2537 err = ret;
2538 goto out_end_trans;
2539 }
2540 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2541 ret = btrfs_uuid_tree_rem(trans, fs_info,
2542 dest->root_item.received_uuid,
2543 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2544 dest->root_key.objectid);
2545 if (ret && ret != -ENOENT) {
2546 btrfs_abort_transaction(trans, ret);
2547 err = ret;
2548 goto out_end_trans;
2549 }
2550 }
2551
2552 out_end_trans:
2553 trans->block_rsv = NULL;
2554 trans->bytes_reserved = 0;
2555 ret = btrfs_end_transaction(trans, root);
2556 if (ret && !err)
2557 err = ret;
2558 inode->i_flags |= S_DEAD;
2559 out_release:
2560 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2561 out_up_write:
2562 up_write(&fs_info->subvol_sem);
2563 if (err) {
2564 spin_lock(&dest->root_item_lock);
2565 root_flags = btrfs_root_flags(&dest->root_item);
2566 btrfs_set_root_flags(&dest->root_item,
2567 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2568 spin_unlock(&dest->root_item_lock);
2569 }
2570 out_unlock_inode:
2571 inode_unlock(inode);
2572 if (!err) {
2573 d_invalidate(dentry);
2574 btrfs_invalidate_inodes(dest);
2575 d_delete(dentry);
2576 ASSERT(dest->send_in_progress == 0);
2577
2578 /* the last ref */
2579 if (dest->ino_cache_inode) {
2580 iput(dest->ino_cache_inode);
2581 dest->ino_cache_inode = NULL;
2582 }
2583 }
2584 out_dput:
2585 dput(dentry);
2586 out_unlock_dir:
2587 inode_unlock(dir);
2588 out_drop_write:
2589 mnt_drop_write_file(file);
2590 out:
2591 kfree(vol_args);
2592 return err;
2593 }
2594
2595 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2596 {
2597 struct inode *inode = file_inode(file);
2598 struct btrfs_root *root = BTRFS_I(inode)->root;
2599 struct btrfs_ioctl_defrag_range_args *range;
2600 int ret;
2601
2602 ret = mnt_want_write_file(file);
2603 if (ret)
2604 return ret;
2605
2606 if (btrfs_root_readonly(root)) {
2607 ret = -EROFS;
2608 goto out;
2609 }
2610
2611 switch (inode->i_mode & S_IFMT) {
2612 case S_IFDIR:
2613 if (!capable(CAP_SYS_ADMIN)) {
2614 ret = -EPERM;
2615 goto out;
2616 }
2617 ret = btrfs_defrag_root(root);
2618 if (ret)
2619 goto out;
2620 ret = btrfs_defrag_root(root->fs_info->extent_root);
2621 break;
2622 case S_IFREG:
2623 if (!(file->f_mode & FMODE_WRITE)) {
2624 ret = -EINVAL;
2625 goto out;
2626 }
2627
2628 range = kzalloc(sizeof(*range), GFP_KERNEL);
2629 if (!range) {
2630 ret = -ENOMEM;
2631 goto out;
2632 }
2633
2634 if (argp) {
2635 if (copy_from_user(range, argp,
2636 sizeof(*range))) {
2637 ret = -EFAULT;
2638 kfree(range);
2639 goto out;
2640 }
2641 /* compression requires us to start the IO */
2642 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2643 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2644 range->extent_thresh = (u32)-1;
2645 }
2646 } else {
2647 /* the rest are all set to zero by kzalloc */
2648 range->len = (u64)-1;
2649 }
2650 ret = btrfs_defrag_file(file_inode(file), file,
2651 range, 0, 0);
2652 if (ret > 0)
2653 ret = 0;
2654 kfree(range);
2655 break;
2656 default:
2657 ret = -EINVAL;
2658 }
2659 out:
2660 mnt_drop_write_file(file);
2661 return ret;
2662 }
2663
2664 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2665 {
2666 struct btrfs_fs_info *fs_info = root->fs_info;
2667 struct btrfs_ioctl_vol_args *vol_args;
2668 int ret;
2669
2670 if (!capable(CAP_SYS_ADMIN))
2671 return -EPERM;
2672
2673 if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1))
2674 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2675
2676 mutex_lock(&fs_info->volume_mutex);
2677 vol_args = memdup_user(arg, sizeof(*vol_args));
2678 if (IS_ERR(vol_args)) {
2679 ret = PTR_ERR(vol_args);
2680 goto out;
2681 }
2682
2683 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2684 ret = btrfs_init_new_device(root->fs_info, vol_args->name);
2685
2686 if (!ret)
2687 btrfs_info(fs_info, "disk added %s", vol_args->name);
2688
2689 kfree(vol_args);
2690 out:
2691 mutex_unlock(&fs_info->volume_mutex);
2692 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
2693 return ret;
2694 }
2695
2696 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2697 {
2698 struct inode *inode = file_inode(file);
2699 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2700 struct btrfs_root *root = BTRFS_I(inode)->root;
2701 struct btrfs_ioctl_vol_args_v2 *vol_args;
2702 int ret;
2703
2704 if (!capable(CAP_SYS_ADMIN))
2705 return -EPERM;
2706
2707 ret = mnt_want_write_file(file);
2708 if (ret)
2709 return ret;
2710
2711 vol_args = memdup_user(arg, sizeof(*vol_args));
2712 if (IS_ERR(vol_args)) {
2713 ret = PTR_ERR(vol_args);
2714 goto err_drop;
2715 }
2716
2717 /* Check for compatibility reject unknown flags */
2718 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED)
2719 return -EOPNOTSUPP;
2720
2721 if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
2722 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2723 goto out;
2724 }
2725
2726 mutex_lock(&fs_info->volume_mutex);
2727 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2728 ret = btrfs_rm_device(root, NULL, vol_args->devid);
2729 } else {
2730 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2731 ret = btrfs_rm_device(root, vol_args->name, 0);
2732 }
2733 mutex_unlock(&fs_info->volume_mutex);
2734 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
2735
2736 if (!ret) {
2737 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2738 btrfs_info(fs_info, "device deleted: id %llu",
2739 vol_args->devid);
2740 else
2741 btrfs_info(fs_info, "device deleted: %s",
2742 vol_args->name);
2743 }
2744 out:
2745 kfree(vol_args);
2746 err_drop:
2747 mnt_drop_write_file(file);
2748 return ret;
2749 }
2750
2751 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2752 {
2753 struct inode *inode = file_inode(file);
2754 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2755 struct btrfs_root *root = BTRFS_I(inode)->root;
2756 struct btrfs_ioctl_vol_args *vol_args;
2757 int ret;
2758
2759 if (!capable(CAP_SYS_ADMIN))
2760 return -EPERM;
2761
2762 ret = mnt_want_write_file(file);
2763 if (ret)
2764 return ret;
2765
2766 if (atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
2767 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2768 goto out_drop_write;
2769 }
2770
2771 vol_args = memdup_user(arg, sizeof(*vol_args));
2772 if (IS_ERR(vol_args)) {
2773 ret = PTR_ERR(vol_args);
2774 goto out;
2775 }
2776
2777 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2778 mutex_lock(&fs_info->volume_mutex);
2779 ret = btrfs_rm_device(root, vol_args->name, 0);
2780 mutex_unlock(&fs_info->volume_mutex);
2781
2782 if (!ret)
2783 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2784 kfree(vol_args);
2785 out:
2786 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
2787 out_drop_write:
2788 mnt_drop_write_file(file);
2789
2790 return ret;
2791 }
2792
2793 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2794 {
2795 struct btrfs_fs_info *fs_info = root->fs_info;
2796 struct btrfs_ioctl_fs_info_args *fi_args;
2797 struct btrfs_device *device;
2798 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2799 int ret = 0;
2800
2801 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2802 if (!fi_args)
2803 return -ENOMEM;
2804
2805 mutex_lock(&fs_devices->device_list_mutex);
2806 fi_args->num_devices = fs_devices->num_devices;
2807 memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid));
2808
2809 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2810 if (device->devid > fi_args->max_id)
2811 fi_args->max_id = device->devid;
2812 }
2813 mutex_unlock(&fs_devices->device_list_mutex);
2814
2815 fi_args->nodesize = fs_info->super_copy->nodesize;
2816 fi_args->sectorsize = fs_info->super_copy->sectorsize;
2817 fi_args->clone_alignment = fs_info->super_copy->sectorsize;
2818
2819 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2820 ret = -EFAULT;
2821
2822 kfree(fi_args);
2823 return ret;
2824 }
2825
2826 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2827 {
2828 struct btrfs_fs_info *fs_info = root->fs_info;
2829 struct btrfs_ioctl_dev_info_args *di_args;
2830 struct btrfs_device *dev;
2831 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2832 int ret = 0;
2833 char *s_uuid = NULL;
2834
2835 di_args = memdup_user(arg, sizeof(*di_args));
2836 if (IS_ERR(di_args))
2837 return PTR_ERR(di_args);
2838
2839 if (!btrfs_is_empty_uuid(di_args->uuid))
2840 s_uuid = di_args->uuid;
2841
2842 mutex_lock(&fs_devices->device_list_mutex);
2843 dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL);
2844
2845 if (!dev) {
2846 ret = -ENODEV;
2847 goto out;
2848 }
2849
2850 di_args->devid = dev->devid;
2851 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2852 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2853 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2854 if (dev->name) {
2855 struct rcu_string *name;
2856
2857 rcu_read_lock();
2858 name = rcu_dereference(dev->name);
2859 strncpy(di_args->path, name->str, sizeof(di_args->path));
2860 rcu_read_unlock();
2861 di_args->path[sizeof(di_args->path) - 1] = 0;
2862 } else {
2863 di_args->path[0] = '\0';
2864 }
2865
2866 out:
2867 mutex_unlock(&fs_devices->device_list_mutex);
2868 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2869 ret = -EFAULT;
2870
2871 kfree(di_args);
2872 return ret;
2873 }
2874
2875 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2876 {
2877 struct page *page;
2878
2879 page = grab_cache_page(inode->i_mapping, index);
2880 if (!page)
2881 return ERR_PTR(-ENOMEM);
2882
2883 if (!PageUptodate(page)) {
2884 int ret;
2885
2886 ret = btrfs_readpage(NULL, page);
2887 if (ret)
2888 return ERR_PTR(ret);
2889 lock_page(page);
2890 if (!PageUptodate(page)) {
2891 unlock_page(page);
2892 put_page(page);
2893 return ERR_PTR(-EIO);
2894 }
2895 if (page->mapping != inode->i_mapping) {
2896 unlock_page(page);
2897 put_page(page);
2898 return ERR_PTR(-EAGAIN);
2899 }
2900 }
2901
2902 return page;
2903 }
2904
2905 static int gather_extent_pages(struct inode *inode, struct page **pages,
2906 int num_pages, u64 off)
2907 {
2908 int i;
2909 pgoff_t index = off >> PAGE_SHIFT;
2910
2911 for (i = 0; i < num_pages; i++) {
2912 again:
2913 pages[i] = extent_same_get_page(inode, index + i);
2914 if (IS_ERR(pages[i])) {
2915 int err = PTR_ERR(pages[i]);
2916
2917 if (err == -EAGAIN)
2918 goto again;
2919 pages[i] = NULL;
2920 return err;
2921 }
2922 }
2923 return 0;
2924 }
2925
2926 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2927 bool retry_range_locking)
2928 {
2929 /*
2930 * Do any pending delalloc/csum calculations on inode, one way or
2931 * another, and lock file content.
2932 * The locking order is:
2933 *
2934 * 1) pages
2935 * 2) range in the inode's io tree
2936 */
2937 while (1) {
2938 struct btrfs_ordered_extent *ordered;
2939 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2940 ordered = btrfs_lookup_first_ordered_extent(inode,
2941 off + len - 1);
2942 if ((!ordered ||
2943 ordered->file_offset + ordered->len <= off ||
2944 ordered->file_offset >= off + len) &&
2945 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2946 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2947 if (ordered)
2948 btrfs_put_ordered_extent(ordered);
2949 break;
2950 }
2951 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2952 if (ordered)
2953 btrfs_put_ordered_extent(ordered);
2954 if (!retry_range_locking)
2955 return -EAGAIN;
2956 btrfs_wait_ordered_range(inode, off, len);
2957 }
2958 return 0;
2959 }
2960
2961 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2962 {
2963 inode_unlock(inode1);
2964 inode_unlock(inode2);
2965 }
2966
2967 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2968 {
2969 if (inode1 < inode2)
2970 swap(inode1, inode2);
2971
2972 inode_lock_nested(inode1, I_MUTEX_PARENT);
2973 inode_lock_nested(inode2, I_MUTEX_CHILD);
2974 }
2975
2976 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2977 struct inode *inode2, u64 loff2, u64 len)
2978 {
2979 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2980 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2981 }
2982
2983 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2984 struct inode *inode2, u64 loff2, u64 len,
2985 bool retry_range_locking)
2986 {
2987 int ret;
2988
2989 if (inode1 < inode2) {
2990 swap(inode1, inode2);
2991 swap(loff1, loff2);
2992 }
2993 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2994 if (ret)
2995 return ret;
2996 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2997 if (ret)
2998 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2999 loff1 + len - 1);
3000 return ret;
3001 }
3002
3003 struct cmp_pages {
3004 int num_pages;
3005 struct page **src_pages;
3006 struct page **dst_pages;
3007 };
3008
3009 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
3010 {
3011 int i;
3012 struct page *pg;
3013
3014 for (i = 0; i < cmp->num_pages; i++) {
3015 pg = cmp->src_pages[i];
3016 if (pg) {
3017 unlock_page(pg);
3018 put_page(pg);
3019 }
3020 pg = cmp->dst_pages[i];
3021 if (pg) {
3022 unlock_page(pg);
3023 put_page(pg);
3024 }
3025 }
3026 kfree(cmp->src_pages);
3027 kfree(cmp->dst_pages);
3028 }
3029
3030 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
3031 struct inode *dst, u64 dst_loff,
3032 u64 len, struct cmp_pages *cmp)
3033 {
3034 int ret;
3035 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
3036 struct page **src_pgarr, **dst_pgarr;
3037
3038 /*
3039 * We must gather up all the pages before we initiate our
3040 * extent locking. We use an array for the page pointers. Size
3041 * of the array is bounded by len, which is in turn bounded by
3042 * BTRFS_MAX_DEDUPE_LEN.
3043 */
3044 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
3045 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
3046 if (!src_pgarr || !dst_pgarr) {
3047 kfree(src_pgarr);
3048 kfree(dst_pgarr);
3049 return -ENOMEM;
3050 }
3051 cmp->num_pages = num_pages;
3052 cmp->src_pages = src_pgarr;
3053 cmp->dst_pages = dst_pgarr;
3054
3055 ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff);
3056 if (ret)
3057 goto out;
3058
3059 ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff);
3060
3061 out:
3062 if (ret)
3063 btrfs_cmp_data_free(cmp);
3064 return 0;
3065 }
3066
3067 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
3068 u64 dst_loff, u64 len, struct cmp_pages *cmp)
3069 {
3070 int ret = 0;
3071 int i;
3072 struct page *src_page, *dst_page;
3073 unsigned int cmp_len = PAGE_SIZE;
3074 void *addr, *dst_addr;
3075
3076 i = 0;
3077 while (len) {
3078 if (len < PAGE_SIZE)
3079 cmp_len = len;
3080
3081 BUG_ON(i >= cmp->num_pages);
3082
3083 src_page = cmp->src_pages[i];
3084 dst_page = cmp->dst_pages[i];
3085 ASSERT(PageLocked(src_page));
3086 ASSERT(PageLocked(dst_page));
3087
3088 addr = kmap_atomic(src_page);
3089 dst_addr = kmap_atomic(dst_page);
3090
3091 flush_dcache_page(src_page);
3092 flush_dcache_page(dst_page);
3093
3094 if (memcmp(addr, dst_addr, cmp_len))
3095 ret = -EBADE;
3096
3097 kunmap_atomic(addr);
3098 kunmap_atomic(dst_addr);
3099
3100 if (ret)
3101 break;
3102
3103 len -= cmp_len;
3104 i++;
3105 }
3106
3107 return ret;
3108 }
3109
3110 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3111 u64 olen)
3112 {
3113 u64 len = *plen;
3114 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3115
3116 if (off + olen > inode->i_size || off + olen < off)
3117 return -EINVAL;
3118
3119 /* if we extend to eof, continue to block boundary */
3120 if (off + len == inode->i_size)
3121 *plen = len = ALIGN(inode->i_size, bs) - off;
3122
3123 /* Check that we are block aligned - btrfs_clone() requires this */
3124 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3125 return -EINVAL;
3126
3127 return 0;
3128 }
3129
3130 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3131 struct inode *dst, u64 dst_loff)
3132 {
3133 int ret;
3134 u64 len = olen;
3135 struct cmp_pages cmp;
3136 int same_inode = 0;
3137 u64 same_lock_start = 0;
3138 u64 same_lock_len = 0;
3139
3140 if (src == dst)
3141 same_inode = 1;
3142
3143 if (len == 0)
3144 return 0;
3145
3146 if (same_inode) {
3147 inode_lock(src);
3148
3149 ret = extent_same_check_offsets(src, loff, &len, olen);
3150 if (ret)
3151 goto out_unlock;
3152 ret = extent_same_check_offsets(src, dst_loff, &len, olen);
3153 if (ret)
3154 goto out_unlock;
3155
3156 /*
3157 * Single inode case wants the same checks, except we
3158 * don't want our length pushed out past i_size as
3159 * comparing that data range makes no sense.
3160 *
3161 * extent_same_check_offsets() will do this for an
3162 * unaligned length at i_size, so catch it here and
3163 * reject the request.
3164 *
3165 * This effectively means we require aligned extents
3166 * for the single-inode case, whereas the other cases
3167 * allow an unaligned length so long as it ends at
3168 * i_size.
3169 */
3170 if (len != olen) {
3171 ret = -EINVAL;
3172 goto out_unlock;
3173 }
3174
3175 /* Check for overlapping ranges */
3176 if (dst_loff + len > loff && dst_loff < loff + len) {
3177 ret = -EINVAL;
3178 goto out_unlock;
3179 }
3180
3181 same_lock_start = min_t(u64, loff, dst_loff);
3182 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3183 } else {
3184 btrfs_double_inode_lock(src, dst);
3185
3186 ret = extent_same_check_offsets(src, loff, &len, olen);
3187 if (ret)
3188 goto out_unlock;
3189
3190 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3191 if (ret)
3192 goto out_unlock;
3193 }
3194
3195 /* don't make the dst file partly checksummed */
3196 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3197 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3198 ret = -EINVAL;
3199 goto out_unlock;
3200 }
3201
3202 again:
3203 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3204 if (ret)
3205 goto out_unlock;
3206
3207 if (same_inode)
3208 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3209 false);
3210 else
3211 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3212 false);
3213 /*
3214 * If one of the inodes has dirty pages in the respective range or
3215 * ordered extents, we need to flush dellaloc and wait for all ordered
3216 * extents in the range. We must unlock the pages and the ranges in the
3217 * io trees to avoid deadlocks when flushing delalloc (requires locking
3218 * pages) and when waiting for ordered extents to complete (they require
3219 * range locking).
3220 */
3221 if (ret == -EAGAIN) {
3222 /*
3223 * Ranges in the io trees already unlocked. Now unlock all
3224 * pages before waiting for all IO to complete.
3225 */
3226 btrfs_cmp_data_free(&cmp);
3227 if (same_inode) {
3228 btrfs_wait_ordered_range(src, same_lock_start,
3229 same_lock_len);
3230 } else {
3231 btrfs_wait_ordered_range(src, loff, len);
3232 btrfs_wait_ordered_range(dst, dst_loff, len);
3233 }
3234 goto again;
3235 }
3236 ASSERT(ret == 0);
3237 if (WARN_ON(ret)) {
3238 /* ranges in the io trees already unlocked */
3239 btrfs_cmp_data_free(&cmp);
3240 return ret;
3241 }
3242
3243 /* pass original length for comparison so we stay within i_size */
3244 ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp);
3245 if (ret == 0)
3246 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3247
3248 if (same_inode)
3249 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3250 same_lock_start + same_lock_len - 1);
3251 else
3252 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3253
3254 btrfs_cmp_data_free(&cmp);
3255 out_unlock:
3256 if (same_inode)
3257 inode_unlock(src);
3258 else
3259 btrfs_double_inode_unlock(src, dst);
3260
3261 return ret;
3262 }
3263
3264 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3265
3266 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3267 struct file *dst_file, u64 dst_loff)
3268 {
3269 struct inode *src = file_inode(src_file);
3270 struct inode *dst = file_inode(dst_file);
3271 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3272 ssize_t res;
3273
3274 if (olen > BTRFS_MAX_DEDUPE_LEN)
3275 olen = BTRFS_MAX_DEDUPE_LEN;
3276
3277 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
3278 /*
3279 * Btrfs does not support blocksize < page_size. As a
3280 * result, btrfs_cmp_data() won't correctly handle
3281 * this situation without an update.
3282 */
3283 return -EINVAL;
3284 }
3285
3286 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3287 if (res)
3288 return res;
3289 return olen;
3290 }
3291
3292 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3293 struct inode *inode,
3294 u64 endoff,
3295 const u64 destoff,
3296 const u64 olen,
3297 int no_time_update)
3298 {
3299 struct btrfs_root *root = BTRFS_I(inode)->root;
3300 int ret;
3301
3302 inode_inc_iversion(inode);
3303 if (!no_time_update)
3304 inode->i_mtime = inode->i_ctime = current_time(inode);
3305 /*
3306 * We round up to the block size at eof when determining which
3307 * extents to clone above, but shouldn't round up the file size.
3308 */
3309 if (endoff > destoff + olen)
3310 endoff = destoff + olen;
3311 if (endoff > inode->i_size)
3312 btrfs_i_size_write(inode, endoff);
3313
3314 ret = btrfs_update_inode(trans, root, inode);
3315 if (ret) {
3316 btrfs_abort_transaction(trans, ret);
3317 btrfs_end_transaction(trans, root);
3318 goto out;
3319 }
3320 ret = btrfs_end_transaction(trans, root);
3321 out:
3322 return ret;
3323 }
3324
3325 static void clone_update_extent_map(struct inode *inode,
3326 const struct btrfs_trans_handle *trans,
3327 const struct btrfs_path *path,
3328 const u64 hole_offset,
3329 const u64 hole_len)
3330 {
3331 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3332 struct extent_map *em;
3333 int ret;
3334
3335 em = alloc_extent_map();
3336 if (!em) {
3337 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3338 &BTRFS_I(inode)->runtime_flags);
3339 return;
3340 }
3341
3342 if (path) {
3343 struct btrfs_file_extent_item *fi;
3344
3345 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3346 struct btrfs_file_extent_item);
3347 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3348 em->generation = -1;
3349 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3350 BTRFS_FILE_EXTENT_INLINE)
3351 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3352 &BTRFS_I(inode)->runtime_flags);
3353 } else {
3354 em->start = hole_offset;
3355 em->len = hole_len;
3356 em->ram_bytes = em->len;
3357 em->orig_start = hole_offset;
3358 em->block_start = EXTENT_MAP_HOLE;
3359 em->block_len = 0;
3360 em->orig_block_len = 0;
3361 em->compress_type = BTRFS_COMPRESS_NONE;
3362 em->generation = trans->transid;
3363 }
3364
3365 while (1) {
3366 write_lock(&em_tree->lock);
3367 ret = add_extent_mapping(em_tree, em, 1);
3368 write_unlock(&em_tree->lock);
3369 if (ret != -EEXIST) {
3370 free_extent_map(em);
3371 break;
3372 }
3373 btrfs_drop_extent_cache(inode, em->start,
3374 em->start + em->len - 1, 0);
3375 }
3376
3377 if (ret)
3378 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3379 &BTRFS_I(inode)->runtime_flags);
3380 }
3381
3382 /*
3383 * Make sure we do not end up inserting an inline extent into a file that has
3384 * already other (non-inline) extents. If a file has an inline extent it can
3385 * not have any other extents and the (single) inline extent must start at the
3386 * file offset 0. Failing to respect these rules will lead to file corruption,
3387 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3388 *
3389 * We can have extents that have been already written to disk or we can have
3390 * dirty ranges still in delalloc, in which case the extent maps and items are
3391 * created only when we run delalloc, and the delalloc ranges might fall outside
3392 * the range we are currently locking in the inode's io tree. So we check the
3393 * inode's i_size because of that (i_size updates are done while holding the
3394 * i_mutex, which we are holding here).
3395 * We also check to see if the inode has a size not greater than "datal" but has
3396 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3397 * protected against such concurrent fallocate calls by the i_mutex).
3398 *
3399 * If the file has no extents but a size greater than datal, do not allow the
3400 * copy because we would need turn the inline extent into a non-inline one (even
3401 * with NO_HOLES enabled). If we find our destination inode only has one inline
3402 * extent, just overwrite it with the source inline extent if its size is less
3403 * than the source extent's size, or we could copy the source inline extent's
3404 * data into the destination inode's inline extent if the later is greater then
3405 * the former.
3406 */
3407 static int clone_copy_inline_extent(struct inode *src,
3408 struct inode *dst,
3409 struct btrfs_trans_handle *trans,
3410 struct btrfs_path *path,
3411 struct btrfs_key *new_key,
3412 const u64 drop_start,
3413 const u64 datal,
3414 const u64 skip,
3415 const u64 size,
3416 char *inline_data)
3417 {
3418 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
3419 struct btrfs_root *root = BTRFS_I(dst)->root;
3420 const u64 aligned_end = ALIGN(new_key->offset + datal,
3421 fs_info->sectorsize);
3422 int ret;
3423 struct btrfs_key key;
3424
3425 if (new_key->offset > 0)
3426 return -EOPNOTSUPP;
3427
3428 key.objectid = btrfs_ino(dst);
3429 key.type = BTRFS_EXTENT_DATA_KEY;
3430 key.offset = 0;
3431 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3432 if (ret < 0) {
3433 return ret;
3434 } else if (ret > 0) {
3435 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3436 ret = btrfs_next_leaf(root, path);
3437 if (ret < 0)
3438 return ret;
3439 else if (ret > 0)
3440 goto copy_inline_extent;
3441 }
3442 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3443 if (key.objectid == btrfs_ino(dst) &&
3444 key.type == BTRFS_EXTENT_DATA_KEY) {
3445 ASSERT(key.offset > 0);
3446 return -EOPNOTSUPP;
3447 }
3448 } else if (i_size_read(dst) <= datal) {
3449 struct btrfs_file_extent_item *ei;
3450 u64 ext_len;
3451
3452 /*
3453 * If the file size is <= datal, make sure there are no other
3454 * extents following (can happen do to an fallocate call with
3455 * the flag FALLOC_FL_KEEP_SIZE).
3456 */
3457 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3458 struct btrfs_file_extent_item);
3459 /*
3460 * If it's an inline extent, it can not have other extents
3461 * following it.
3462 */
3463 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3464 BTRFS_FILE_EXTENT_INLINE)
3465 goto copy_inline_extent;
3466
3467 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3468 if (ext_len > aligned_end)
3469 return -EOPNOTSUPP;
3470
3471 ret = btrfs_next_item(root, path);
3472 if (ret < 0) {
3473 return ret;
3474 } else if (ret == 0) {
3475 btrfs_item_key_to_cpu(path->nodes[0], &key,
3476 path->slots[0]);
3477 if (key.objectid == btrfs_ino(dst) &&
3478 key.type == BTRFS_EXTENT_DATA_KEY)
3479 return -EOPNOTSUPP;
3480 }
3481 }
3482
3483 copy_inline_extent:
3484 /*
3485 * We have no extent items, or we have an extent at offset 0 which may
3486 * or may not be inlined. All these cases are dealt the same way.
3487 */
3488 if (i_size_read(dst) > datal) {
3489 /*
3490 * If the destination inode has an inline extent...
3491 * This would require copying the data from the source inline
3492 * extent into the beginning of the destination's inline extent.
3493 * But this is really complex, both extents can be compressed
3494 * or just one of them, which would require decompressing and
3495 * re-compressing data (which could increase the new compressed
3496 * size, not allowing the compressed data to fit anymore in an
3497 * inline extent).
3498 * So just don't support this case for now (it should be rare,
3499 * we are not really saving space when cloning inline extents).
3500 */
3501 return -EOPNOTSUPP;
3502 }
3503
3504 btrfs_release_path(path);
3505 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3506 if (ret)
3507 return ret;
3508 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3509 if (ret)
3510 return ret;
3511
3512 if (skip) {
3513 const u32 start = btrfs_file_extent_calc_inline_size(0);
3514
3515 memmove(inline_data + start, inline_data + start + skip, datal);
3516 }
3517
3518 write_extent_buffer(path->nodes[0], inline_data,
3519 btrfs_item_ptr_offset(path->nodes[0],
3520 path->slots[0]),
3521 size);
3522 inode_add_bytes(dst, datal);
3523
3524 return 0;
3525 }
3526
3527 /**
3528 * btrfs_clone() - clone a range from inode file to another
3529 *
3530 * @src: Inode to clone from
3531 * @inode: Inode to clone to
3532 * @off: Offset within source to start clone from
3533 * @olen: Original length, passed by user, of range to clone
3534 * @olen_aligned: Block-aligned value of olen
3535 * @destoff: Offset within @inode to start clone
3536 * @no_time_update: Whether to update mtime/ctime on the target inode
3537 */
3538 static int btrfs_clone(struct inode *src, struct inode *inode,
3539 const u64 off, const u64 olen, const u64 olen_aligned,
3540 const u64 destoff, int no_time_update)
3541 {
3542 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3543 struct btrfs_root *root = BTRFS_I(inode)->root;
3544 struct btrfs_path *path = NULL;
3545 struct extent_buffer *leaf;
3546 struct btrfs_trans_handle *trans;
3547 char *buf = NULL;
3548 struct btrfs_key key;
3549 u32 nritems;
3550 int slot;
3551 int ret;
3552 const u64 len = olen_aligned;
3553 u64 last_dest_end = destoff;
3554
3555 ret = -ENOMEM;
3556 buf = kmalloc(fs_info->nodesize, GFP_KERNEL | __GFP_NOWARN);
3557 if (!buf) {
3558 buf = vmalloc(fs_info->nodesize);
3559 if (!buf)
3560 return ret;
3561 }
3562
3563 path = btrfs_alloc_path();
3564 if (!path) {
3565 kvfree(buf);
3566 return ret;
3567 }
3568
3569 path->reada = READA_FORWARD;
3570 /* clone data */
3571 key.objectid = btrfs_ino(src);
3572 key.type = BTRFS_EXTENT_DATA_KEY;
3573 key.offset = off;
3574
3575 while (1) {
3576 u64 next_key_min_offset = key.offset + 1;
3577
3578 /*
3579 * note the key will change type as we walk through the
3580 * tree.
3581 */
3582 path->leave_spinning = 1;
3583 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3584 0, 0);
3585 if (ret < 0)
3586 goto out;
3587 /*
3588 * First search, if no extent item that starts at offset off was
3589 * found but the previous item is an extent item, it's possible
3590 * it might overlap our target range, therefore process it.
3591 */
3592 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3593 btrfs_item_key_to_cpu(path->nodes[0], &key,
3594 path->slots[0] - 1);
3595 if (key.type == BTRFS_EXTENT_DATA_KEY)
3596 path->slots[0]--;
3597 }
3598
3599 nritems = btrfs_header_nritems(path->nodes[0]);
3600 process_slot:
3601 if (path->slots[0] >= nritems) {
3602 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3603 if (ret < 0)
3604 goto out;
3605 if (ret > 0)
3606 break;
3607 nritems = btrfs_header_nritems(path->nodes[0]);
3608 }
3609 leaf = path->nodes[0];
3610 slot = path->slots[0];
3611
3612 btrfs_item_key_to_cpu(leaf, &key, slot);
3613 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3614 key.objectid != btrfs_ino(src))
3615 break;
3616
3617 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3618 struct btrfs_file_extent_item *extent;
3619 int type;
3620 u32 size;
3621 struct btrfs_key new_key;
3622 u64 disko = 0, diskl = 0;
3623 u64 datao = 0, datal = 0;
3624 u8 comp;
3625 u64 drop_start;
3626
3627 extent = btrfs_item_ptr(leaf, slot,
3628 struct btrfs_file_extent_item);
3629 comp = btrfs_file_extent_compression(leaf, extent);
3630 type = btrfs_file_extent_type(leaf, extent);
3631 if (type == BTRFS_FILE_EXTENT_REG ||
3632 type == BTRFS_FILE_EXTENT_PREALLOC) {
3633 disko = btrfs_file_extent_disk_bytenr(leaf,
3634 extent);
3635 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3636 extent);
3637 datao = btrfs_file_extent_offset(leaf, extent);
3638 datal = btrfs_file_extent_num_bytes(leaf,
3639 extent);
3640 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3641 /* take upper bound, may be compressed */
3642 datal = btrfs_file_extent_ram_bytes(leaf,
3643 extent);
3644 }
3645
3646 /*
3647 * The first search might have left us at an extent
3648 * item that ends before our target range's start, can
3649 * happen if we have holes and NO_HOLES feature enabled.
3650 */
3651 if (key.offset + datal <= off) {
3652 path->slots[0]++;
3653 goto process_slot;
3654 } else if (key.offset >= off + len) {
3655 break;
3656 }
3657 next_key_min_offset = key.offset + datal;
3658 size = btrfs_item_size_nr(leaf, slot);
3659 read_extent_buffer(leaf, buf,
3660 btrfs_item_ptr_offset(leaf, slot),
3661 size);
3662
3663 btrfs_release_path(path);
3664 path->leave_spinning = 0;
3665
3666 memcpy(&new_key, &key, sizeof(new_key));
3667 new_key.objectid = btrfs_ino(inode);
3668 if (off <= key.offset)
3669 new_key.offset = key.offset + destoff - off;
3670 else
3671 new_key.offset = destoff;
3672
3673 /*
3674 * Deal with a hole that doesn't have an extent item
3675 * that represents it (NO_HOLES feature enabled).
3676 * This hole is either in the middle of the cloning
3677 * range or at the beginning (fully overlaps it or
3678 * partially overlaps it).
3679 */
3680 if (new_key.offset != last_dest_end)
3681 drop_start = last_dest_end;
3682 else
3683 drop_start = new_key.offset;
3684
3685 /*
3686 * 1 - adjusting old extent (we may have to split it)
3687 * 1 - add new extent
3688 * 1 - inode update
3689 */
3690 trans = btrfs_start_transaction(root, 3);
3691 if (IS_ERR(trans)) {
3692 ret = PTR_ERR(trans);
3693 goto out;
3694 }
3695
3696 if (type == BTRFS_FILE_EXTENT_REG ||
3697 type == BTRFS_FILE_EXTENT_PREALLOC) {
3698 /*
3699 * a | --- range to clone ---| b
3700 * | ------------- extent ------------- |
3701 */
3702
3703 /* subtract range b */
3704 if (key.offset + datal > off + len)
3705 datal = off + len - key.offset;
3706
3707 /* subtract range a */
3708 if (off > key.offset) {
3709 datao += off - key.offset;
3710 datal -= off - key.offset;
3711 }
3712
3713 ret = btrfs_drop_extents(trans, root, inode,
3714 drop_start,
3715 new_key.offset + datal,
3716 1);
3717 if (ret) {
3718 if (ret != -EOPNOTSUPP)
3719 btrfs_abort_transaction(trans,
3720 ret);
3721 btrfs_end_transaction(trans, root);
3722 goto out;
3723 }
3724
3725 ret = btrfs_insert_empty_item(trans, root, path,
3726 &new_key, size);
3727 if (ret) {
3728 btrfs_abort_transaction(trans, ret);
3729 btrfs_end_transaction(trans, root);
3730 goto out;
3731 }
3732
3733 leaf = path->nodes[0];
3734 slot = path->slots[0];
3735 write_extent_buffer(leaf, buf,
3736 btrfs_item_ptr_offset(leaf, slot),
3737 size);
3738
3739 extent = btrfs_item_ptr(leaf, slot,
3740 struct btrfs_file_extent_item);
3741
3742 /* disko == 0 means it's a hole */
3743 if (!disko)
3744 datao = 0;
3745
3746 btrfs_set_file_extent_offset(leaf, extent,
3747 datao);
3748 btrfs_set_file_extent_num_bytes(leaf, extent,
3749 datal);
3750
3751 if (disko) {
3752 inode_add_bytes(inode, datal);
3753 ret = btrfs_inc_extent_ref(trans, root,
3754 disko, diskl, 0,
3755 root->root_key.objectid,
3756 btrfs_ino(inode),
3757 new_key.offset - datao);
3758 if (ret) {
3759 btrfs_abort_transaction(trans,
3760 ret);
3761 btrfs_end_transaction(trans,
3762 root);
3763 goto out;
3764
3765 }
3766 }
3767 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3768 u64 skip = 0;
3769 u64 trim = 0;
3770
3771 if (off > key.offset) {
3772 skip = off - key.offset;
3773 new_key.offset += skip;
3774 }
3775
3776 if (key.offset + datal > off + len)
3777 trim = key.offset + datal - (off + len);
3778
3779 if (comp && (skip || trim)) {
3780 ret = -EINVAL;
3781 btrfs_end_transaction(trans, root);
3782 goto out;
3783 }
3784 size -= skip + trim;
3785 datal -= skip + trim;
3786
3787 ret = clone_copy_inline_extent(src, inode,
3788 trans, path,
3789 &new_key,
3790 drop_start,
3791 datal,
3792 skip, size, buf);
3793 if (ret) {
3794 if (ret != -EOPNOTSUPP)
3795 btrfs_abort_transaction(trans,
3796 ret);
3797 btrfs_end_transaction(trans, root);
3798 goto out;
3799 }
3800 leaf = path->nodes[0];
3801 slot = path->slots[0];
3802 }
3803
3804 /* If we have an implicit hole (NO_HOLES feature). */
3805 if (drop_start < new_key.offset)
3806 clone_update_extent_map(inode, trans,
3807 NULL, drop_start,
3808 new_key.offset - drop_start);
3809
3810 clone_update_extent_map(inode, trans, path, 0, 0);
3811
3812 btrfs_mark_buffer_dirty(leaf);
3813 btrfs_release_path(path);
3814
3815 last_dest_end = ALIGN(new_key.offset + datal,
3816 fs_info->sectorsize);
3817 ret = clone_finish_inode_update(trans, inode,
3818 last_dest_end,
3819 destoff, olen,
3820 no_time_update);
3821 if (ret)
3822 goto out;
3823 if (new_key.offset + datal >= destoff + len)
3824 break;
3825 }
3826 btrfs_release_path(path);
3827 key.offset = next_key_min_offset;
3828
3829 if (fatal_signal_pending(current)) {
3830 ret = -EINTR;
3831 goto out;
3832 }
3833 }
3834 ret = 0;
3835
3836 if (last_dest_end < destoff + len) {
3837 /*
3838 * We have an implicit hole (NO_HOLES feature is enabled) that
3839 * fully or partially overlaps our cloning range at its end.
3840 */
3841 btrfs_release_path(path);
3842
3843 /*
3844 * 1 - remove extent(s)
3845 * 1 - inode update
3846 */
3847 trans = btrfs_start_transaction(root, 2);
3848 if (IS_ERR(trans)) {
3849 ret = PTR_ERR(trans);
3850 goto out;
3851 }
3852 ret = btrfs_drop_extents(trans, root, inode,
3853 last_dest_end, destoff + len, 1);
3854 if (ret) {
3855 if (ret != -EOPNOTSUPP)
3856 btrfs_abort_transaction(trans, ret);
3857 btrfs_end_transaction(trans, root);
3858 goto out;
3859 }
3860 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3861 destoff + len - last_dest_end);
3862 ret = clone_finish_inode_update(trans, inode, destoff + len,
3863 destoff, olen, no_time_update);
3864 }
3865
3866 out:
3867 btrfs_free_path(path);
3868 kvfree(buf);
3869 return ret;
3870 }
3871
3872 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3873 u64 off, u64 olen, u64 destoff)
3874 {
3875 struct inode *inode = file_inode(file);
3876 struct inode *src = file_inode(file_src);
3877 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3878 struct btrfs_root *root = BTRFS_I(inode)->root;
3879 int ret;
3880 u64 len = olen;
3881 u64 bs = fs_info->sb->s_blocksize;
3882 int same_inode = src == inode;
3883
3884 /*
3885 * TODO:
3886 * - split compressed inline extents. annoying: we need to
3887 * decompress into destination's address_space (the file offset
3888 * may change, so source mapping won't do), then recompress (or
3889 * otherwise reinsert) a subrange.
3890 *
3891 * - split destination inode's inline extents. The inline extents can
3892 * be either compressed or non-compressed.
3893 */
3894
3895 if (btrfs_root_readonly(root))
3896 return -EROFS;
3897
3898 if (file_src->f_path.mnt != file->f_path.mnt ||
3899 src->i_sb != inode->i_sb)
3900 return -EXDEV;
3901
3902 /* don't make the dst file partly checksummed */
3903 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3904 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3905 return -EINVAL;
3906
3907 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3908 return -EISDIR;
3909
3910 if (!same_inode) {
3911 btrfs_double_inode_lock(src, inode);
3912 } else {
3913 inode_lock(src);
3914 }
3915
3916 /* determine range to clone */
3917 ret = -EINVAL;
3918 if (off + len > src->i_size || off + len < off)
3919 goto out_unlock;
3920 if (len == 0)
3921 olen = len = src->i_size - off;
3922 /* if we extend to eof, continue to block boundary */
3923 if (off + len == src->i_size)
3924 len = ALIGN(src->i_size, bs) - off;
3925
3926 if (len == 0) {
3927 ret = 0;
3928 goto out_unlock;
3929 }
3930
3931 /* verify the end result is block aligned */
3932 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3933 !IS_ALIGNED(destoff, bs))
3934 goto out_unlock;
3935
3936 /* verify if ranges are overlapped within the same file */
3937 if (same_inode) {
3938 if (destoff + len > off && destoff < off + len)
3939 goto out_unlock;
3940 }
3941
3942 if (destoff > inode->i_size) {
3943 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3944 if (ret)
3945 goto out_unlock;
3946 }
3947
3948 /*
3949 * Lock the target range too. Right after we replace the file extent
3950 * items in the fs tree (which now point to the cloned data), we might
3951 * have a worker replace them with extent items relative to a write
3952 * operation that was issued before this clone operation (i.e. confront
3953 * with inode.c:btrfs_finish_ordered_io).
3954 */
3955 if (same_inode) {
3956 u64 lock_start = min_t(u64, off, destoff);
3957 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3958
3959 ret = lock_extent_range(src, lock_start, lock_len, true);
3960 } else {
3961 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3962 true);
3963 }
3964 ASSERT(ret == 0);
3965 if (WARN_ON(ret)) {
3966 /* ranges in the io trees already unlocked */
3967 goto out_unlock;
3968 }
3969
3970 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3971
3972 if (same_inode) {
3973 u64 lock_start = min_t(u64, off, destoff);
3974 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3975
3976 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3977 } else {
3978 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3979 }
3980 /*
3981 * Truncate page cache pages so that future reads will see the cloned
3982 * data immediately and not the previous data.
3983 */
3984 truncate_inode_pages_range(&inode->i_data,
3985 round_down(destoff, PAGE_SIZE),
3986 round_up(destoff + len, PAGE_SIZE) - 1);
3987 out_unlock:
3988 if (!same_inode)
3989 btrfs_double_inode_unlock(src, inode);
3990 else
3991 inode_unlock(src);
3992 return ret;
3993 }
3994
3995 ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in,
3996 struct file *file_out, loff_t pos_out,
3997 size_t len, unsigned int flags)
3998 {
3999 ssize_t ret;
4000
4001 ret = btrfs_clone_files(file_out, file_in, pos_in, len, pos_out);
4002 if (ret == 0)
4003 ret = len;
4004 return ret;
4005 }
4006
4007 int btrfs_clone_file_range(struct file *src_file, loff_t off,
4008 struct file *dst_file, loff_t destoff, u64 len)
4009 {
4010 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
4011 }
4012
4013 /*
4014 * there are many ways the trans_start and trans_end ioctls can lead
4015 * to deadlocks. They should only be used by applications that
4016 * basically own the machine, and have a very in depth understanding
4017 * of all the possible deadlocks and enospc problems.
4018 */
4019 static long btrfs_ioctl_trans_start(struct file *file)
4020 {
4021 struct inode *inode = file_inode(file);
4022 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4023 struct btrfs_root *root = BTRFS_I(inode)->root;
4024 struct btrfs_trans_handle *trans;
4025 int ret;
4026
4027 ret = -EPERM;
4028 if (!capable(CAP_SYS_ADMIN))
4029 goto out;
4030
4031 ret = -EINPROGRESS;
4032 if (file->private_data)
4033 goto out;
4034
4035 ret = -EROFS;
4036 if (btrfs_root_readonly(root))
4037 goto out;
4038
4039 ret = mnt_want_write_file(file);
4040 if (ret)
4041 goto out;
4042
4043 atomic_inc(&fs_info->open_ioctl_trans);
4044
4045 ret = -ENOMEM;
4046 trans = btrfs_start_ioctl_transaction(root);
4047 if (IS_ERR(trans))
4048 goto out_drop;
4049
4050 file->private_data = trans;
4051 return 0;
4052
4053 out_drop:
4054 atomic_dec(&fs_info->open_ioctl_trans);
4055 mnt_drop_write_file(file);
4056 out:
4057 return ret;
4058 }
4059
4060 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
4061 {
4062 struct inode *inode = file_inode(file);
4063 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4064 struct btrfs_root *root = BTRFS_I(inode)->root;
4065 struct btrfs_root *new_root;
4066 struct btrfs_dir_item *di;
4067 struct btrfs_trans_handle *trans;
4068 struct btrfs_path *path;
4069 struct btrfs_key location;
4070 struct btrfs_disk_key disk_key;
4071 u64 objectid = 0;
4072 u64 dir_id;
4073 int ret;
4074
4075 if (!capable(CAP_SYS_ADMIN))
4076 return -EPERM;
4077
4078 ret = mnt_want_write_file(file);
4079 if (ret)
4080 return ret;
4081
4082 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4083 ret = -EFAULT;
4084 goto out;
4085 }
4086
4087 if (!objectid)
4088 objectid = BTRFS_FS_TREE_OBJECTID;
4089
4090 location.objectid = objectid;
4091 location.type = BTRFS_ROOT_ITEM_KEY;
4092 location.offset = (u64)-1;
4093
4094 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
4095 if (IS_ERR(new_root)) {
4096 ret = PTR_ERR(new_root);
4097 goto out;
4098 }
4099
4100 path = btrfs_alloc_path();
4101 if (!path) {
4102 ret = -ENOMEM;
4103 goto out;
4104 }
4105 path->leave_spinning = 1;
4106
4107 trans = btrfs_start_transaction(root, 1);
4108 if (IS_ERR(trans)) {
4109 btrfs_free_path(path);
4110 ret = PTR_ERR(trans);
4111 goto out;
4112 }
4113
4114 dir_id = btrfs_super_root_dir(fs_info->super_copy);
4115 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
4116 dir_id, "default", 7, 1);
4117 if (IS_ERR_OR_NULL(di)) {
4118 btrfs_free_path(path);
4119 btrfs_end_transaction(trans, root);
4120 btrfs_err(fs_info,
4121 "Umm, you don't have the default diritem, this isn't going to work");
4122 ret = -ENOENT;
4123 goto out;
4124 }
4125
4126 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4127 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4128 btrfs_mark_buffer_dirty(path->nodes[0]);
4129 btrfs_free_path(path);
4130
4131 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
4132 btrfs_end_transaction(trans, root);
4133 out:
4134 mnt_drop_write_file(file);
4135 return ret;
4136 }
4137
4138 void btrfs_get_block_group_info(struct list_head *groups_list,
4139 struct btrfs_ioctl_space_info *space)
4140 {
4141 struct btrfs_block_group_cache *block_group;
4142
4143 space->total_bytes = 0;
4144 space->used_bytes = 0;
4145 space->flags = 0;
4146 list_for_each_entry(block_group, groups_list, list) {
4147 space->flags = block_group->flags;
4148 space->total_bytes += block_group->key.offset;
4149 space->used_bytes +=
4150 btrfs_block_group_used(&block_group->item);
4151 }
4152 }
4153
4154 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
4155 {
4156 struct btrfs_fs_info *fs_info = root->fs_info;
4157 struct btrfs_ioctl_space_args space_args;
4158 struct btrfs_ioctl_space_info space;
4159 struct btrfs_ioctl_space_info *dest;
4160 struct btrfs_ioctl_space_info *dest_orig;
4161 struct btrfs_ioctl_space_info __user *user_dest;
4162 struct btrfs_space_info *info;
4163 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4164 BTRFS_BLOCK_GROUP_SYSTEM,
4165 BTRFS_BLOCK_GROUP_METADATA,
4166 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4167 int num_types = 4;
4168 int alloc_size;
4169 int ret = 0;
4170 u64 slot_count = 0;
4171 int i, c;
4172
4173 if (copy_from_user(&space_args,
4174 (struct btrfs_ioctl_space_args __user *)arg,
4175 sizeof(space_args)))
4176 return -EFAULT;
4177
4178 for (i = 0; i < num_types; i++) {
4179 struct btrfs_space_info *tmp;
4180
4181 info = NULL;
4182 rcu_read_lock();
4183 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4184 list) {
4185 if (tmp->flags == types[i]) {
4186 info = tmp;
4187 break;
4188 }
4189 }
4190 rcu_read_unlock();
4191
4192 if (!info)
4193 continue;
4194
4195 down_read(&info->groups_sem);
4196 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4197 if (!list_empty(&info->block_groups[c]))
4198 slot_count++;
4199 }
4200 up_read(&info->groups_sem);
4201 }
4202
4203 /*
4204 * Global block reserve, exported as a space_info
4205 */
4206 slot_count++;
4207
4208 /* space_slots == 0 means they are asking for a count */
4209 if (space_args.space_slots == 0) {
4210 space_args.total_spaces = slot_count;
4211 goto out;
4212 }
4213
4214 slot_count = min_t(u64, space_args.space_slots, slot_count);
4215
4216 alloc_size = sizeof(*dest) * slot_count;
4217
4218 /* we generally have at most 6 or so space infos, one for each raid
4219 * level. So, a whole page should be more than enough for everyone
4220 */
4221 if (alloc_size > PAGE_SIZE)
4222 return -ENOMEM;
4223
4224 space_args.total_spaces = 0;
4225 dest = kmalloc(alloc_size, GFP_KERNEL);
4226 if (!dest)
4227 return -ENOMEM;
4228 dest_orig = dest;
4229
4230 /* now we have a buffer to copy into */
4231 for (i = 0; i < num_types; i++) {
4232 struct btrfs_space_info *tmp;
4233
4234 if (!slot_count)
4235 break;
4236
4237 info = NULL;
4238 rcu_read_lock();
4239 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4240 list) {
4241 if (tmp->flags == types[i]) {
4242 info = tmp;
4243 break;
4244 }
4245 }
4246 rcu_read_unlock();
4247
4248 if (!info)
4249 continue;
4250 down_read(&info->groups_sem);
4251 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4252 if (!list_empty(&info->block_groups[c])) {
4253 btrfs_get_block_group_info(
4254 &info->block_groups[c], &space);
4255 memcpy(dest, &space, sizeof(space));
4256 dest++;
4257 space_args.total_spaces++;
4258 slot_count--;
4259 }
4260 if (!slot_count)
4261 break;
4262 }
4263 up_read(&info->groups_sem);
4264 }
4265
4266 /*
4267 * Add global block reserve
4268 */
4269 if (slot_count) {
4270 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
4271
4272 spin_lock(&block_rsv->lock);
4273 space.total_bytes = block_rsv->size;
4274 space.used_bytes = block_rsv->size - block_rsv->reserved;
4275 spin_unlock(&block_rsv->lock);
4276 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4277 memcpy(dest, &space, sizeof(space));
4278 space_args.total_spaces++;
4279 }
4280
4281 user_dest = (struct btrfs_ioctl_space_info __user *)
4282 (arg + sizeof(struct btrfs_ioctl_space_args));
4283
4284 if (copy_to_user(user_dest, dest_orig, alloc_size))
4285 ret = -EFAULT;
4286
4287 kfree(dest_orig);
4288 out:
4289 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4290 ret = -EFAULT;
4291
4292 return ret;
4293 }
4294
4295 /*
4296 * there are many ways the trans_start and trans_end ioctls can lead
4297 * to deadlocks. They should only be used by applications that
4298 * basically own the machine, and have a very in depth understanding
4299 * of all the possible deadlocks and enospc problems.
4300 */
4301 long btrfs_ioctl_trans_end(struct file *file)
4302 {
4303 struct inode *inode = file_inode(file);
4304 struct btrfs_root *root = BTRFS_I(inode)->root;
4305 struct btrfs_trans_handle *trans;
4306
4307 trans = file->private_data;
4308 if (!trans)
4309 return -EINVAL;
4310 file->private_data = NULL;
4311
4312 btrfs_end_transaction(trans, root);
4313
4314 atomic_dec(&root->fs_info->open_ioctl_trans);
4315
4316 mnt_drop_write_file(file);
4317 return 0;
4318 }
4319
4320 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4321 void __user *argp)
4322 {
4323 struct btrfs_trans_handle *trans;
4324 u64 transid;
4325 int ret;
4326
4327 trans = btrfs_attach_transaction_barrier(root);
4328 if (IS_ERR(trans)) {
4329 if (PTR_ERR(trans) != -ENOENT)
4330 return PTR_ERR(trans);
4331
4332 /* No running transaction, don't bother */
4333 transid = root->fs_info->last_trans_committed;
4334 goto out;
4335 }
4336 transid = trans->transid;
4337 ret = btrfs_commit_transaction_async(trans, root, 0);
4338 if (ret) {
4339 btrfs_end_transaction(trans, root);
4340 return ret;
4341 }
4342 out:
4343 if (argp)
4344 if (copy_to_user(argp, &transid, sizeof(transid)))
4345 return -EFAULT;
4346 return 0;
4347 }
4348
4349 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4350 void __user *argp)
4351 {
4352 u64 transid;
4353
4354 if (argp) {
4355 if (copy_from_user(&transid, argp, sizeof(transid)))
4356 return -EFAULT;
4357 } else {
4358 transid = 0; /* current trans */
4359 }
4360 return btrfs_wait_for_commit(root, transid);
4361 }
4362
4363 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4364 {
4365 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4366 struct btrfs_ioctl_scrub_args *sa;
4367 int ret;
4368
4369 if (!capable(CAP_SYS_ADMIN))
4370 return -EPERM;
4371
4372 sa = memdup_user(arg, sizeof(*sa));
4373 if (IS_ERR(sa))
4374 return PTR_ERR(sa);
4375
4376 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4377 ret = mnt_want_write_file(file);
4378 if (ret)
4379 goto out;
4380 }
4381
4382 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4383 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4384 0);
4385
4386 if (copy_to_user(arg, sa, sizeof(*sa)))
4387 ret = -EFAULT;
4388
4389 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4390 mnt_drop_write_file(file);
4391 out:
4392 kfree(sa);
4393 return ret;
4394 }
4395
4396 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4397 {
4398 if (!capable(CAP_SYS_ADMIN))
4399 return -EPERM;
4400
4401 return btrfs_scrub_cancel(root->fs_info);
4402 }
4403
4404 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4405 void __user *arg)
4406 {
4407 struct btrfs_ioctl_scrub_args *sa;
4408 int ret;
4409
4410 if (!capable(CAP_SYS_ADMIN))
4411 return -EPERM;
4412
4413 sa = memdup_user(arg, sizeof(*sa));
4414 if (IS_ERR(sa))
4415 return PTR_ERR(sa);
4416
4417 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4418
4419 if (copy_to_user(arg, sa, sizeof(*sa)))
4420 ret = -EFAULT;
4421
4422 kfree(sa);
4423 return ret;
4424 }
4425
4426 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4427 void __user *arg)
4428 {
4429 struct btrfs_ioctl_get_dev_stats *sa;
4430 int ret;
4431
4432 sa = memdup_user(arg, sizeof(*sa));
4433 if (IS_ERR(sa))
4434 return PTR_ERR(sa);
4435
4436 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4437 kfree(sa);
4438 return -EPERM;
4439 }
4440
4441 ret = btrfs_get_dev_stats(root, sa);
4442
4443 if (copy_to_user(arg, sa, sizeof(*sa)))
4444 ret = -EFAULT;
4445
4446 kfree(sa);
4447 return ret;
4448 }
4449
4450 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4451 {
4452 struct btrfs_fs_info *fs_info = root->fs_info;
4453 struct btrfs_ioctl_dev_replace_args *p;
4454 int ret;
4455
4456 if (!capable(CAP_SYS_ADMIN))
4457 return -EPERM;
4458
4459 p = memdup_user(arg, sizeof(*p));
4460 if (IS_ERR(p))
4461 return PTR_ERR(p);
4462
4463 switch (p->cmd) {
4464 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4465 if (fs_info->sb->s_flags & MS_RDONLY) {
4466 ret = -EROFS;
4467 goto out;
4468 }
4469 if (atomic_xchg(
4470 &fs_info->mutually_exclusive_operation_running, 1)) {
4471 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4472 } else {
4473 ret = btrfs_dev_replace_by_ioctl(root, p);
4474 atomic_set(
4475 &fs_info->mutually_exclusive_operation_running, 0);
4476 }
4477 break;
4478 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4479 btrfs_dev_replace_status(fs_info, p);
4480 ret = 0;
4481 break;
4482 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4483 ret = btrfs_dev_replace_cancel(fs_info, p);
4484 break;
4485 default:
4486 ret = -EINVAL;
4487 break;
4488 }
4489
4490 if (copy_to_user(arg, p, sizeof(*p)))
4491 ret = -EFAULT;
4492 out:
4493 kfree(p);
4494 return ret;
4495 }
4496
4497 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4498 {
4499 int ret = 0;
4500 int i;
4501 u64 rel_ptr;
4502 int size;
4503 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4504 struct inode_fs_paths *ipath = NULL;
4505 struct btrfs_path *path;
4506
4507 if (!capable(CAP_DAC_READ_SEARCH))
4508 return -EPERM;
4509
4510 path = btrfs_alloc_path();
4511 if (!path) {
4512 ret = -ENOMEM;
4513 goto out;
4514 }
4515
4516 ipa = memdup_user(arg, sizeof(*ipa));
4517 if (IS_ERR(ipa)) {
4518 ret = PTR_ERR(ipa);
4519 ipa = NULL;
4520 goto out;
4521 }
4522
4523 size = min_t(u32, ipa->size, 4096);
4524 ipath = init_ipath(size, root, path);
4525 if (IS_ERR(ipath)) {
4526 ret = PTR_ERR(ipath);
4527 ipath = NULL;
4528 goto out;
4529 }
4530
4531 ret = paths_from_inode(ipa->inum, ipath);
4532 if (ret < 0)
4533 goto out;
4534
4535 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4536 rel_ptr = ipath->fspath->val[i] -
4537 (u64)(unsigned long)ipath->fspath->val;
4538 ipath->fspath->val[i] = rel_ptr;
4539 }
4540
4541 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4542 (void *)(unsigned long)ipath->fspath, size);
4543 if (ret) {
4544 ret = -EFAULT;
4545 goto out;
4546 }
4547
4548 out:
4549 btrfs_free_path(path);
4550 free_ipath(ipath);
4551 kfree(ipa);
4552
4553 return ret;
4554 }
4555
4556 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4557 {
4558 struct btrfs_data_container *inodes = ctx;
4559 const size_t c = 3 * sizeof(u64);
4560
4561 if (inodes->bytes_left >= c) {
4562 inodes->bytes_left -= c;
4563 inodes->val[inodes->elem_cnt] = inum;
4564 inodes->val[inodes->elem_cnt + 1] = offset;
4565 inodes->val[inodes->elem_cnt + 2] = root;
4566 inodes->elem_cnt += 3;
4567 } else {
4568 inodes->bytes_missing += c - inodes->bytes_left;
4569 inodes->bytes_left = 0;
4570 inodes->elem_missed += 3;
4571 }
4572
4573 return 0;
4574 }
4575
4576 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4577 void __user *arg)
4578 {
4579 int ret = 0;
4580 int size;
4581 struct btrfs_ioctl_logical_ino_args *loi;
4582 struct btrfs_data_container *inodes = NULL;
4583 struct btrfs_path *path = NULL;
4584
4585 if (!capable(CAP_SYS_ADMIN))
4586 return -EPERM;
4587
4588 loi = memdup_user(arg, sizeof(*loi));
4589 if (IS_ERR(loi))
4590 return PTR_ERR(loi);
4591
4592 path = btrfs_alloc_path();
4593 if (!path) {
4594 ret = -ENOMEM;
4595 goto out;
4596 }
4597
4598 size = min_t(u32, loi->size, SZ_64K);
4599 inodes = init_data_container(size);
4600 if (IS_ERR(inodes)) {
4601 ret = PTR_ERR(inodes);
4602 inodes = NULL;
4603 goto out;
4604 }
4605
4606 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4607 build_ino_list, inodes);
4608 if (ret == -EINVAL)
4609 ret = -ENOENT;
4610 if (ret < 0)
4611 goto out;
4612
4613 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4614 (void *)(unsigned long)inodes, size);
4615 if (ret)
4616 ret = -EFAULT;
4617
4618 out:
4619 btrfs_free_path(path);
4620 vfree(inodes);
4621 kfree(loi);
4622
4623 return ret;
4624 }
4625
4626 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4627 struct btrfs_ioctl_balance_args *bargs)
4628 {
4629 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4630
4631 bargs->flags = bctl->flags;
4632
4633 if (atomic_read(&fs_info->balance_running))
4634 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4635 if (atomic_read(&fs_info->balance_pause_req))
4636 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4637 if (atomic_read(&fs_info->balance_cancel_req))
4638 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4639
4640 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4641 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4642 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4643
4644 if (lock) {
4645 spin_lock(&fs_info->balance_lock);
4646 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4647 spin_unlock(&fs_info->balance_lock);
4648 } else {
4649 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4650 }
4651 }
4652
4653 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4654 {
4655 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4656 struct btrfs_fs_info *fs_info = root->fs_info;
4657 struct btrfs_ioctl_balance_args *bargs;
4658 struct btrfs_balance_control *bctl;
4659 bool need_unlock; /* for mut. excl. ops lock */
4660 int ret;
4661
4662 if (!capable(CAP_SYS_ADMIN))
4663 return -EPERM;
4664
4665 ret = mnt_want_write_file(file);
4666 if (ret)
4667 return ret;
4668
4669 again:
4670 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4671 mutex_lock(&fs_info->volume_mutex);
4672 mutex_lock(&fs_info->balance_mutex);
4673 need_unlock = true;
4674 goto locked;
4675 }
4676
4677 /*
4678 * mut. excl. ops lock is locked. Three possibilities:
4679 * (1) some other op is running
4680 * (2) balance is running
4681 * (3) balance is paused -- special case (think resume)
4682 */
4683 mutex_lock(&fs_info->balance_mutex);
4684 if (fs_info->balance_ctl) {
4685 /* this is either (2) or (3) */
4686 if (!atomic_read(&fs_info->balance_running)) {
4687 mutex_unlock(&fs_info->balance_mutex);
4688 if (!mutex_trylock(&fs_info->volume_mutex))
4689 goto again;
4690 mutex_lock(&fs_info->balance_mutex);
4691
4692 if (fs_info->balance_ctl &&
4693 !atomic_read(&fs_info->balance_running)) {
4694 /* this is (3) */
4695 need_unlock = false;
4696 goto locked;
4697 }
4698
4699 mutex_unlock(&fs_info->balance_mutex);
4700 mutex_unlock(&fs_info->volume_mutex);
4701 goto again;
4702 } else {
4703 /* this is (2) */
4704 mutex_unlock(&fs_info->balance_mutex);
4705 ret = -EINPROGRESS;
4706 goto out;
4707 }
4708 } else {
4709 /* this is (1) */
4710 mutex_unlock(&fs_info->balance_mutex);
4711 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4712 goto out;
4713 }
4714
4715 locked:
4716 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4717
4718 if (arg) {
4719 bargs = memdup_user(arg, sizeof(*bargs));
4720 if (IS_ERR(bargs)) {
4721 ret = PTR_ERR(bargs);
4722 goto out_unlock;
4723 }
4724
4725 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4726 if (!fs_info->balance_ctl) {
4727 ret = -ENOTCONN;
4728 goto out_bargs;
4729 }
4730
4731 bctl = fs_info->balance_ctl;
4732 spin_lock(&fs_info->balance_lock);
4733 bctl->flags |= BTRFS_BALANCE_RESUME;
4734 spin_unlock(&fs_info->balance_lock);
4735
4736 goto do_balance;
4737 }
4738 } else {
4739 bargs = NULL;
4740 }
4741
4742 if (fs_info->balance_ctl) {
4743 ret = -EINPROGRESS;
4744 goto out_bargs;
4745 }
4746
4747 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4748 if (!bctl) {
4749 ret = -ENOMEM;
4750 goto out_bargs;
4751 }
4752
4753 bctl->fs_info = fs_info;
4754 if (arg) {
4755 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4756 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4757 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4758
4759 bctl->flags = bargs->flags;
4760 } else {
4761 /* balance everything - no filters */
4762 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4763 }
4764
4765 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4766 ret = -EINVAL;
4767 goto out_bctl;
4768 }
4769
4770 do_balance:
4771 /*
4772 * Ownership of bctl and mutually_exclusive_operation_running
4773 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4774 * or, if restriper was paused all the way until unmount, in
4775 * free_fs_info. mutually_exclusive_operation_running is
4776 * cleared in __cancel_balance.
4777 */
4778 need_unlock = false;
4779
4780 ret = btrfs_balance(bctl, bargs);
4781 bctl = NULL;
4782
4783 if (arg) {
4784 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4785 ret = -EFAULT;
4786 }
4787
4788 out_bctl:
4789 kfree(bctl);
4790 out_bargs:
4791 kfree(bargs);
4792 out_unlock:
4793 mutex_unlock(&fs_info->balance_mutex);
4794 mutex_unlock(&fs_info->volume_mutex);
4795 if (need_unlock)
4796 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4797 out:
4798 mnt_drop_write_file(file);
4799 return ret;
4800 }
4801
4802 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4803 {
4804 struct btrfs_fs_info *fs_info = root->fs_info;
4805
4806 if (!capable(CAP_SYS_ADMIN))
4807 return -EPERM;
4808
4809 switch (cmd) {
4810 case BTRFS_BALANCE_CTL_PAUSE:
4811 return btrfs_pause_balance(fs_info);
4812 case BTRFS_BALANCE_CTL_CANCEL:
4813 return btrfs_cancel_balance(fs_info);
4814 }
4815
4816 return -EINVAL;
4817 }
4818
4819 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4820 void __user *arg)
4821 {
4822 struct btrfs_fs_info *fs_info = root->fs_info;
4823 struct btrfs_ioctl_balance_args *bargs;
4824 int ret = 0;
4825
4826 if (!capable(CAP_SYS_ADMIN))
4827 return -EPERM;
4828
4829 mutex_lock(&fs_info->balance_mutex);
4830 if (!fs_info->balance_ctl) {
4831 ret = -ENOTCONN;
4832 goto out;
4833 }
4834
4835 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4836 if (!bargs) {
4837 ret = -ENOMEM;
4838 goto out;
4839 }
4840
4841 update_ioctl_balance_args(fs_info, 1, bargs);
4842
4843 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4844 ret = -EFAULT;
4845
4846 kfree(bargs);
4847 out:
4848 mutex_unlock(&fs_info->balance_mutex);
4849 return ret;
4850 }
4851
4852 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4853 {
4854 struct inode *inode = file_inode(file);
4855 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4856 struct btrfs_ioctl_quota_ctl_args *sa;
4857 struct btrfs_trans_handle *trans = NULL;
4858 int ret;
4859 int err;
4860
4861 if (!capable(CAP_SYS_ADMIN))
4862 return -EPERM;
4863
4864 ret = mnt_want_write_file(file);
4865 if (ret)
4866 return ret;
4867
4868 sa = memdup_user(arg, sizeof(*sa));
4869 if (IS_ERR(sa)) {
4870 ret = PTR_ERR(sa);
4871 goto drop_write;
4872 }
4873
4874 down_write(&fs_info->subvol_sem);
4875 trans = btrfs_start_transaction(fs_info->tree_root, 2);
4876 if (IS_ERR(trans)) {
4877 ret = PTR_ERR(trans);
4878 goto out;
4879 }
4880
4881 switch (sa->cmd) {
4882 case BTRFS_QUOTA_CTL_ENABLE:
4883 ret = btrfs_quota_enable(trans, fs_info);
4884 break;
4885 case BTRFS_QUOTA_CTL_DISABLE:
4886 ret = btrfs_quota_disable(trans, fs_info);
4887 break;
4888 default:
4889 ret = -EINVAL;
4890 break;
4891 }
4892
4893 err = btrfs_commit_transaction(trans, fs_info->tree_root);
4894 if (err && !ret)
4895 ret = err;
4896 out:
4897 kfree(sa);
4898 up_write(&fs_info->subvol_sem);
4899 drop_write:
4900 mnt_drop_write_file(file);
4901 return ret;
4902 }
4903
4904 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4905 {
4906 struct inode *inode = file_inode(file);
4907 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4908 struct btrfs_root *root = BTRFS_I(inode)->root;
4909 struct btrfs_ioctl_qgroup_assign_args *sa;
4910 struct btrfs_trans_handle *trans;
4911 int ret;
4912 int err;
4913
4914 if (!capable(CAP_SYS_ADMIN))
4915 return -EPERM;
4916
4917 ret = mnt_want_write_file(file);
4918 if (ret)
4919 return ret;
4920
4921 sa = memdup_user(arg, sizeof(*sa));
4922 if (IS_ERR(sa)) {
4923 ret = PTR_ERR(sa);
4924 goto drop_write;
4925 }
4926
4927 trans = btrfs_join_transaction(root);
4928 if (IS_ERR(trans)) {
4929 ret = PTR_ERR(trans);
4930 goto out;
4931 }
4932
4933 /* FIXME: check if the IDs really exist */
4934 if (sa->assign) {
4935 ret = btrfs_add_qgroup_relation(trans, fs_info,
4936 sa->src, sa->dst);
4937 } else {
4938 ret = btrfs_del_qgroup_relation(trans, fs_info,
4939 sa->src, sa->dst);
4940 }
4941
4942 /* update qgroup status and info */
4943 err = btrfs_run_qgroups(trans, fs_info);
4944 if (err < 0)
4945 btrfs_handle_fs_error(fs_info, err,
4946 "failed to update qgroup status and info");
4947 err = btrfs_end_transaction(trans, root);
4948 if (err && !ret)
4949 ret = err;
4950
4951 out:
4952 kfree(sa);
4953 drop_write:
4954 mnt_drop_write_file(file);
4955 return ret;
4956 }
4957
4958 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4959 {
4960 struct inode *inode = file_inode(file);
4961 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4962 struct btrfs_root *root = BTRFS_I(inode)->root;
4963 struct btrfs_ioctl_qgroup_create_args *sa;
4964 struct btrfs_trans_handle *trans;
4965 int ret;
4966 int err;
4967
4968 if (!capable(CAP_SYS_ADMIN))
4969 return -EPERM;
4970
4971 ret = mnt_want_write_file(file);
4972 if (ret)
4973 return ret;
4974
4975 sa = memdup_user(arg, sizeof(*sa));
4976 if (IS_ERR(sa)) {
4977 ret = PTR_ERR(sa);
4978 goto drop_write;
4979 }
4980
4981 if (!sa->qgroupid) {
4982 ret = -EINVAL;
4983 goto out;
4984 }
4985
4986 trans = btrfs_join_transaction(root);
4987 if (IS_ERR(trans)) {
4988 ret = PTR_ERR(trans);
4989 goto out;
4990 }
4991
4992 /* FIXME: check if the IDs really exist */
4993 if (sa->create) {
4994 ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid);
4995 } else {
4996 ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid);
4997 }
4998
4999 err = btrfs_end_transaction(trans, root);
5000 if (err && !ret)
5001 ret = err;
5002
5003 out:
5004 kfree(sa);
5005 drop_write:
5006 mnt_drop_write_file(file);
5007 return ret;
5008 }
5009
5010 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5011 {
5012 struct inode *inode = file_inode(file);
5013 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5014 struct btrfs_root *root = BTRFS_I(inode)->root;
5015 struct btrfs_ioctl_qgroup_limit_args *sa;
5016 struct btrfs_trans_handle *trans;
5017 int ret;
5018 int err;
5019 u64 qgroupid;
5020
5021 if (!capable(CAP_SYS_ADMIN))
5022 return -EPERM;
5023
5024 ret = mnt_want_write_file(file);
5025 if (ret)
5026 return ret;
5027
5028 sa = memdup_user(arg, sizeof(*sa));
5029 if (IS_ERR(sa)) {
5030 ret = PTR_ERR(sa);
5031 goto drop_write;
5032 }
5033
5034 trans = btrfs_join_transaction(root);
5035 if (IS_ERR(trans)) {
5036 ret = PTR_ERR(trans);
5037 goto out;
5038 }
5039
5040 qgroupid = sa->qgroupid;
5041 if (!qgroupid) {
5042 /* take the current subvol as qgroup */
5043 qgroupid = root->root_key.objectid;
5044 }
5045
5046 /* FIXME: check if the IDs really exist */
5047 ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim);
5048
5049 err = btrfs_end_transaction(trans, root);
5050 if (err && !ret)
5051 ret = err;
5052
5053 out:
5054 kfree(sa);
5055 drop_write:
5056 mnt_drop_write_file(file);
5057 return ret;
5058 }
5059
5060 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
5061 {
5062 struct inode *inode = file_inode(file);
5063 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5064 struct btrfs_ioctl_quota_rescan_args *qsa;
5065 int ret;
5066
5067 if (!capable(CAP_SYS_ADMIN))
5068 return -EPERM;
5069
5070 ret = mnt_want_write_file(file);
5071 if (ret)
5072 return ret;
5073
5074 qsa = memdup_user(arg, sizeof(*qsa));
5075 if (IS_ERR(qsa)) {
5076 ret = PTR_ERR(qsa);
5077 goto drop_write;
5078 }
5079
5080 if (qsa->flags) {
5081 ret = -EINVAL;
5082 goto out;
5083 }
5084
5085 ret = btrfs_qgroup_rescan(fs_info);
5086
5087 out:
5088 kfree(qsa);
5089 drop_write:
5090 mnt_drop_write_file(file);
5091 return ret;
5092 }
5093
5094 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5095 {
5096 struct inode *inode = file_inode(file);
5097 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5098 struct btrfs_ioctl_quota_rescan_args *qsa;
5099 int ret = 0;
5100
5101 if (!capable(CAP_SYS_ADMIN))
5102 return -EPERM;
5103
5104 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
5105 if (!qsa)
5106 return -ENOMEM;
5107
5108 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5109 qsa->flags = 1;
5110 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
5111 }
5112
5113 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5114 ret = -EFAULT;
5115
5116 kfree(qsa);
5117 return ret;
5118 }
5119
5120 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5121 {
5122 struct inode *inode = file_inode(file);
5123 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5124
5125 if (!capable(CAP_SYS_ADMIN))
5126 return -EPERM;
5127
5128 return btrfs_qgroup_wait_for_completion(fs_info, true);
5129 }
5130
5131 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5132 struct btrfs_ioctl_received_subvol_args *sa)
5133 {
5134 struct inode *inode = file_inode(file);
5135 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5136 struct btrfs_root *root = BTRFS_I(inode)->root;
5137 struct btrfs_root_item *root_item = &root->root_item;
5138 struct btrfs_trans_handle *trans;
5139 struct timespec ct = current_time(inode);
5140 int ret = 0;
5141 int received_uuid_changed;
5142
5143 if (!inode_owner_or_capable(inode))
5144 return -EPERM;
5145
5146 ret = mnt_want_write_file(file);
5147 if (ret < 0)
5148 return ret;
5149
5150 down_write(&fs_info->subvol_sem);
5151
5152 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
5153 ret = -EINVAL;
5154 goto out;
5155 }
5156
5157 if (btrfs_root_readonly(root)) {
5158 ret = -EROFS;
5159 goto out;
5160 }
5161
5162 /*
5163 * 1 - root item
5164 * 2 - uuid items (received uuid + subvol uuid)
5165 */
5166 trans = btrfs_start_transaction(root, 3);
5167 if (IS_ERR(trans)) {
5168 ret = PTR_ERR(trans);
5169 trans = NULL;
5170 goto out;
5171 }
5172
5173 sa->rtransid = trans->transid;
5174 sa->rtime.sec = ct.tv_sec;
5175 sa->rtime.nsec = ct.tv_nsec;
5176
5177 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5178 BTRFS_UUID_SIZE);
5179 if (received_uuid_changed &&
5180 !btrfs_is_empty_uuid(root_item->received_uuid))
5181 btrfs_uuid_tree_rem(trans, fs_info, root_item->received_uuid,
5182 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5183 root->root_key.objectid);
5184 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5185 btrfs_set_root_stransid(root_item, sa->stransid);
5186 btrfs_set_root_rtransid(root_item, sa->rtransid);
5187 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5188 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5189 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5190 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5191
5192 ret = btrfs_update_root(trans, fs_info->tree_root,
5193 &root->root_key, &root->root_item);
5194 if (ret < 0) {
5195 btrfs_end_transaction(trans, root);
5196 goto out;
5197 }
5198 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5199 ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid,
5200 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5201 root->root_key.objectid);
5202 if (ret < 0 && ret != -EEXIST) {
5203 btrfs_abort_transaction(trans, ret);
5204 goto out;
5205 }
5206 }
5207 ret = btrfs_commit_transaction(trans, root);
5208 if (ret < 0) {
5209 btrfs_abort_transaction(trans, ret);
5210 goto out;
5211 }
5212
5213 out:
5214 up_write(&fs_info->subvol_sem);
5215 mnt_drop_write_file(file);
5216 return ret;
5217 }
5218
5219 #ifdef CONFIG_64BIT
5220 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5221 void __user *arg)
5222 {
5223 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5224 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5225 int ret = 0;
5226
5227 args32 = memdup_user(arg, sizeof(*args32));
5228 if (IS_ERR(args32))
5229 return PTR_ERR(args32);
5230
5231 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5232 if (!args64) {
5233 ret = -ENOMEM;
5234 goto out;
5235 }
5236
5237 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5238 args64->stransid = args32->stransid;
5239 args64->rtransid = args32->rtransid;
5240 args64->stime.sec = args32->stime.sec;
5241 args64->stime.nsec = args32->stime.nsec;
5242 args64->rtime.sec = args32->rtime.sec;
5243 args64->rtime.nsec = args32->rtime.nsec;
5244 args64->flags = args32->flags;
5245
5246 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5247 if (ret)
5248 goto out;
5249
5250 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5251 args32->stransid = args64->stransid;
5252 args32->rtransid = args64->rtransid;
5253 args32->stime.sec = args64->stime.sec;
5254 args32->stime.nsec = args64->stime.nsec;
5255 args32->rtime.sec = args64->rtime.sec;
5256 args32->rtime.nsec = args64->rtime.nsec;
5257 args32->flags = args64->flags;
5258
5259 ret = copy_to_user(arg, args32, sizeof(*args32));
5260 if (ret)
5261 ret = -EFAULT;
5262
5263 out:
5264 kfree(args32);
5265 kfree(args64);
5266 return ret;
5267 }
5268 #endif
5269
5270 static long btrfs_ioctl_set_received_subvol(struct file *file,
5271 void __user *arg)
5272 {
5273 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5274 int ret = 0;
5275
5276 sa = memdup_user(arg, sizeof(*sa));
5277 if (IS_ERR(sa))
5278 return PTR_ERR(sa);
5279
5280 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5281
5282 if (ret)
5283 goto out;
5284
5285 ret = copy_to_user(arg, sa, sizeof(*sa));
5286 if (ret)
5287 ret = -EFAULT;
5288
5289 out:
5290 kfree(sa);
5291 return ret;
5292 }
5293
5294 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5295 {
5296 struct inode *inode = file_inode(file);
5297 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5298 size_t len;
5299 int ret;
5300 char label[BTRFS_LABEL_SIZE];
5301
5302 spin_lock(&fs_info->super_lock);
5303 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5304 spin_unlock(&fs_info->super_lock);
5305
5306 len = strnlen(label, BTRFS_LABEL_SIZE);
5307
5308 if (len == BTRFS_LABEL_SIZE) {
5309 btrfs_warn(fs_info,
5310 "label is too long, return the first %zu bytes",
5311 --len);
5312 }
5313
5314 ret = copy_to_user(arg, label, len);
5315
5316 return ret ? -EFAULT : 0;
5317 }
5318
5319 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5320 {
5321 struct inode *inode = file_inode(file);
5322 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5323 struct btrfs_root *root = BTRFS_I(inode)->root;
5324 struct btrfs_super_block *super_block = fs_info->super_copy;
5325 struct btrfs_trans_handle *trans;
5326 char label[BTRFS_LABEL_SIZE];
5327 int ret;
5328
5329 if (!capable(CAP_SYS_ADMIN))
5330 return -EPERM;
5331
5332 if (copy_from_user(label, arg, sizeof(label)))
5333 return -EFAULT;
5334
5335 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5336 btrfs_err(fs_info,
5337 "unable to set label with more than %d bytes",
5338 BTRFS_LABEL_SIZE - 1);
5339 return -EINVAL;
5340 }
5341
5342 ret = mnt_want_write_file(file);
5343 if (ret)
5344 return ret;
5345
5346 trans = btrfs_start_transaction(root, 0);
5347 if (IS_ERR(trans)) {
5348 ret = PTR_ERR(trans);
5349 goto out_unlock;
5350 }
5351
5352 spin_lock(&fs_info->super_lock);
5353 strcpy(super_block->label, label);
5354 spin_unlock(&fs_info->super_lock);
5355 ret = btrfs_commit_transaction(trans, root);
5356
5357 out_unlock:
5358 mnt_drop_write_file(file);
5359 return ret;
5360 }
5361
5362 #define INIT_FEATURE_FLAGS(suffix) \
5363 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5364 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5365 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5366
5367 int btrfs_ioctl_get_supported_features(void __user *arg)
5368 {
5369 static const struct btrfs_ioctl_feature_flags features[3] = {
5370 INIT_FEATURE_FLAGS(SUPP),
5371 INIT_FEATURE_FLAGS(SAFE_SET),
5372 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5373 };
5374
5375 if (copy_to_user(arg, &features, sizeof(features)))
5376 return -EFAULT;
5377
5378 return 0;
5379 }
5380
5381 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5382 {
5383 struct inode *inode = file_inode(file);
5384 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5385 struct btrfs_super_block *super_block = fs_info->super_copy;
5386 struct btrfs_ioctl_feature_flags features;
5387
5388 features.compat_flags = btrfs_super_compat_flags(super_block);
5389 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5390 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5391
5392 if (copy_to_user(arg, &features, sizeof(features)))
5393 return -EFAULT;
5394
5395 return 0;
5396 }
5397
5398 static int check_feature_bits(struct btrfs_root *root,
5399 enum btrfs_feature_set set,
5400 u64 change_mask, u64 flags, u64 supported_flags,
5401 u64 safe_set, u64 safe_clear)
5402 {
5403 struct btrfs_fs_info *fs_info = root->fs_info;
5404 const char *type = btrfs_feature_set_names[set];
5405 char *names;
5406 u64 disallowed, unsupported;
5407 u64 set_mask = flags & change_mask;
5408 u64 clear_mask = ~flags & change_mask;
5409
5410 unsupported = set_mask & ~supported_flags;
5411 if (unsupported) {
5412 names = btrfs_printable_features(set, unsupported);
5413 if (names) {
5414 btrfs_warn(fs_info,
5415 "this kernel does not support the %s feature bit%s",
5416 names, strchr(names, ',') ? "s" : "");
5417 kfree(names);
5418 } else
5419 btrfs_warn(fs_info,
5420 "this kernel does not support %s bits 0x%llx",
5421 type, unsupported);
5422 return -EOPNOTSUPP;
5423 }
5424
5425 disallowed = set_mask & ~safe_set;
5426 if (disallowed) {
5427 names = btrfs_printable_features(set, disallowed);
5428 if (names) {
5429 btrfs_warn(fs_info,
5430 "can't set the %s feature bit%s while mounted",
5431 names, strchr(names, ',') ? "s" : "");
5432 kfree(names);
5433 } else
5434 btrfs_warn(fs_info,
5435 "can't set %s bits 0x%llx while mounted",
5436 type, disallowed);
5437 return -EPERM;
5438 }
5439
5440 disallowed = clear_mask & ~safe_clear;
5441 if (disallowed) {
5442 names = btrfs_printable_features(set, disallowed);
5443 if (names) {
5444 btrfs_warn(fs_info,
5445 "can't clear the %s feature bit%s while mounted",
5446 names, strchr(names, ',') ? "s" : "");
5447 kfree(names);
5448 } else
5449 btrfs_warn(fs_info,
5450 "can't clear %s bits 0x%llx while mounted",
5451 type, disallowed);
5452 return -EPERM;
5453 }
5454
5455 return 0;
5456 }
5457
5458 #define check_feature(root, change_mask, flags, mask_base) \
5459 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5460 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5461 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5462 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5463
5464 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5465 {
5466 struct inode *inode = file_inode(file);
5467 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5468 struct btrfs_root *root = BTRFS_I(inode)->root;
5469 struct btrfs_super_block *super_block = fs_info->super_copy;
5470 struct btrfs_ioctl_feature_flags flags[2];
5471 struct btrfs_trans_handle *trans;
5472 u64 newflags;
5473 int ret;
5474
5475 if (!capable(CAP_SYS_ADMIN))
5476 return -EPERM;
5477
5478 if (copy_from_user(flags, arg, sizeof(flags)))
5479 return -EFAULT;
5480
5481 /* Nothing to do */
5482 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5483 !flags[0].incompat_flags)
5484 return 0;
5485
5486 ret = check_feature(root, flags[0].compat_flags,
5487 flags[1].compat_flags, COMPAT);
5488 if (ret)
5489 return ret;
5490
5491 ret = check_feature(root, flags[0].compat_ro_flags,
5492 flags[1].compat_ro_flags, COMPAT_RO);
5493 if (ret)
5494 return ret;
5495
5496 ret = check_feature(root, flags[0].incompat_flags,
5497 flags[1].incompat_flags, INCOMPAT);
5498 if (ret)
5499 return ret;
5500
5501 ret = mnt_want_write_file(file);
5502 if (ret)
5503 return ret;
5504
5505 trans = btrfs_start_transaction(root, 0);
5506 if (IS_ERR(trans)) {
5507 ret = PTR_ERR(trans);
5508 goto out_drop_write;
5509 }
5510
5511 spin_lock(&fs_info->super_lock);
5512 newflags = btrfs_super_compat_flags(super_block);
5513 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5514 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5515 btrfs_set_super_compat_flags(super_block, newflags);
5516
5517 newflags = btrfs_super_compat_ro_flags(super_block);
5518 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5519 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5520 btrfs_set_super_compat_ro_flags(super_block, newflags);
5521
5522 newflags = btrfs_super_incompat_flags(super_block);
5523 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5524 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5525 btrfs_set_super_incompat_flags(super_block, newflags);
5526 spin_unlock(&fs_info->super_lock);
5527
5528 ret = btrfs_commit_transaction(trans, root);
5529 out_drop_write:
5530 mnt_drop_write_file(file);
5531
5532 return ret;
5533 }
5534
5535 long btrfs_ioctl(struct file *file, unsigned int
5536 cmd, unsigned long arg)
5537 {
5538 struct inode *inode = file_inode(file);
5539 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5540 struct btrfs_root *root = BTRFS_I(inode)->root;
5541 void __user *argp = (void __user *)arg;
5542
5543 switch (cmd) {
5544 case FS_IOC_GETFLAGS:
5545 return btrfs_ioctl_getflags(file, argp);
5546 case FS_IOC_SETFLAGS:
5547 return btrfs_ioctl_setflags(file, argp);
5548 case FS_IOC_GETVERSION:
5549 return btrfs_ioctl_getversion(file, argp);
5550 case FITRIM:
5551 return btrfs_ioctl_fitrim(file, argp);
5552 case BTRFS_IOC_SNAP_CREATE:
5553 return btrfs_ioctl_snap_create(file, argp, 0);
5554 case BTRFS_IOC_SNAP_CREATE_V2:
5555 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5556 case BTRFS_IOC_SUBVOL_CREATE:
5557 return btrfs_ioctl_snap_create(file, argp, 1);
5558 case BTRFS_IOC_SUBVOL_CREATE_V2:
5559 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5560 case BTRFS_IOC_SNAP_DESTROY:
5561 return btrfs_ioctl_snap_destroy(file, argp);
5562 case BTRFS_IOC_SUBVOL_GETFLAGS:
5563 return btrfs_ioctl_subvol_getflags(file, argp);
5564 case BTRFS_IOC_SUBVOL_SETFLAGS:
5565 return btrfs_ioctl_subvol_setflags(file, argp);
5566 case BTRFS_IOC_DEFAULT_SUBVOL:
5567 return btrfs_ioctl_default_subvol(file, argp);
5568 case BTRFS_IOC_DEFRAG:
5569 return btrfs_ioctl_defrag(file, NULL);
5570 case BTRFS_IOC_DEFRAG_RANGE:
5571 return btrfs_ioctl_defrag(file, argp);
5572 case BTRFS_IOC_RESIZE:
5573 return btrfs_ioctl_resize(file, argp);
5574 case BTRFS_IOC_ADD_DEV:
5575 return btrfs_ioctl_add_dev(root, argp);
5576 case BTRFS_IOC_RM_DEV:
5577 return btrfs_ioctl_rm_dev(file, argp);
5578 case BTRFS_IOC_RM_DEV_V2:
5579 return btrfs_ioctl_rm_dev_v2(file, argp);
5580 case BTRFS_IOC_FS_INFO:
5581 return btrfs_ioctl_fs_info(root, argp);
5582 case BTRFS_IOC_DEV_INFO:
5583 return btrfs_ioctl_dev_info(root, argp);
5584 case BTRFS_IOC_BALANCE:
5585 return btrfs_ioctl_balance(file, NULL);
5586 case BTRFS_IOC_TRANS_START:
5587 return btrfs_ioctl_trans_start(file);
5588 case BTRFS_IOC_TRANS_END:
5589 return btrfs_ioctl_trans_end(file);
5590 case BTRFS_IOC_TREE_SEARCH:
5591 return btrfs_ioctl_tree_search(file, argp);
5592 case BTRFS_IOC_TREE_SEARCH_V2:
5593 return btrfs_ioctl_tree_search_v2(file, argp);
5594 case BTRFS_IOC_INO_LOOKUP:
5595 return btrfs_ioctl_ino_lookup(file, argp);
5596 case BTRFS_IOC_INO_PATHS:
5597 return btrfs_ioctl_ino_to_path(root, argp);
5598 case BTRFS_IOC_LOGICAL_INO:
5599 return btrfs_ioctl_logical_to_ino(root, argp);
5600 case BTRFS_IOC_SPACE_INFO:
5601 return btrfs_ioctl_space_info(root, argp);
5602 case BTRFS_IOC_SYNC: {
5603 int ret;
5604
5605 ret = btrfs_start_delalloc_roots(fs_info, 0, -1);
5606 if (ret)
5607 return ret;
5608 ret = btrfs_sync_fs(inode->i_sb, 1);
5609 /*
5610 * The transaction thread may want to do more work,
5611 * namely it pokes the cleaner kthread that will start
5612 * processing uncleaned subvols.
5613 */
5614 wake_up_process(fs_info->transaction_kthread);
5615 return ret;
5616 }
5617 case BTRFS_IOC_START_SYNC:
5618 return btrfs_ioctl_start_sync(root, argp);
5619 case BTRFS_IOC_WAIT_SYNC:
5620 return btrfs_ioctl_wait_sync(root, argp);
5621 case BTRFS_IOC_SCRUB:
5622 return btrfs_ioctl_scrub(file, argp);
5623 case BTRFS_IOC_SCRUB_CANCEL:
5624 return btrfs_ioctl_scrub_cancel(root, argp);
5625 case BTRFS_IOC_SCRUB_PROGRESS:
5626 return btrfs_ioctl_scrub_progress(root, argp);
5627 case BTRFS_IOC_BALANCE_V2:
5628 return btrfs_ioctl_balance(file, argp);
5629 case BTRFS_IOC_BALANCE_CTL:
5630 return btrfs_ioctl_balance_ctl(root, arg);
5631 case BTRFS_IOC_BALANCE_PROGRESS:
5632 return btrfs_ioctl_balance_progress(root, argp);
5633 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5634 return btrfs_ioctl_set_received_subvol(file, argp);
5635 #ifdef CONFIG_64BIT
5636 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5637 return btrfs_ioctl_set_received_subvol_32(file, argp);
5638 #endif
5639 case BTRFS_IOC_SEND:
5640 return btrfs_ioctl_send(file, argp);
5641 case BTRFS_IOC_GET_DEV_STATS:
5642 return btrfs_ioctl_get_dev_stats(root, argp);
5643 case BTRFS_IOC_QUOTA_CTL:
5644 return btrfs_ioctl_quota_ctl(file, argp);
5645 case BTRFS_IOC_QGROUP_ASSIGN:
5646 return btrfs_ioctl_qgroup_assign(file, argp);
5647 case BTRFS_IOC_QGROUP_CREATE:
5648 return btrfs_ioctl_qgroup_create(file, argp);
5649 case BTRFS_IOC_QGROUP_LIMIT:
5650 return btrfs_ioctl_qgroup_limit(file, argp);
5651 case BTRFS_IOC_QUOTA_RESCAN:
5652 return btrfs_ioctl_quota_rescan(file, argp);
5653 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5654 return btrfs_ioctl_quota_rescan_status(file, argp);
5655 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5656 return btrfs_ioctl_quota_rescan_wait(file, argp);
5657 case BTRFS_IOC_DEV_REPLACE:
5658 return btrfs_ioctl_dev_replace(root, argp);
5659 case BTRFS_IOC_GET_FSLABEL:
5660 return btrfs_ioctl_get_fslabel(file, argp);
5661 case BTRFS_IOC_SET_FSLABEL:
5662 return btrfs_ioctl_set_fslabel(file, argp);
5663 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5664 return btrfs_ioctl_get_supported_features(argp);
5665 case BTRFS_IOC_GET_FEATURES:
5666 return btrfs_ioctl_get_features(file, argp);
5667 case BTRFS_IOC_SET_FEATURES:
5668 return btrfs_ioctl_set_features(file, argp);
5669 }
5670
5671 return -ENOTTY;
5672 }
5673
5674 #ifdef CONFIG_COMPAT
5675 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5676 {
5677 switch (cmd) {
5678 case FS_IOC32_GETFLAGS:
5679 cmd = FS_IOC_GETFLAGS;
5680 break;
5681 case FS_IOC32_SETFLAGS:
5682 cmd = FS_IOC_SETFLAGS;
5683 break;
5684 case FS_IOC32_GETVERSION:
5685 cmd = FS_IOC_GETVERSION;
5686 break;
5687 default:
5688 return -ENOIOCTLCMD;
5689 }
5690
5691 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5692 }
5693 #endif
ubuntu-bionic-kernel mirror