?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
btrfs: device delete must be sysloged
[linux/fpc-iii.git] / fs / btrfs / ioctl.c
blobe5ed222267eeafa4f1846586cc44af4ca595a2dc
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
23 #include <linux/fs.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include <linux/btrfs.h>
46 #include <linux/uaccess.h>
47 #include "ctree.h"
48 #include "disk-io.h"
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
52 #include "volumes.h"
53 #include "locking.h"
54 #include "inode-map.h"
55 #include "backref.h"
56 #include "rcu-string.h"
57 #include "send.h"
58 #include "dev-replace.h"
59 #include "props.h"
60 #include "sysfs.h"
61 #include "qgroup.h"
62
63 #ifdef CONFIG_64BIT
64 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
68 */
69 struct btrfs_ioctl_timespec_32 {
70 __u64 sec;
71 __u32 nsec;
72 } __attribute__ ((__packed__));
73
74 struct btrfs_ioctl_received_subvol_args_32 {
75 char uuid[BTRFS_UUID_SIZE]; /* in */
76 __u64 stransid; /* in */
77 __u64 rtransid; /* out */
78 struct btrfs_ioctl_timespec_32 stime; /* in */
79 struct btrfs_ioctl_timespec_32 rtime; /* out */
80 __u64 flags; /* in */
81 __u64 reserved[16]; /* in */
82 } __attribute__ ((__packed__));
83
84 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
86 #endif
87
88
89 static int btrfs_clone(struct inode *src, struct inode *inode,
90 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
91
92 /* Mask out flags that are inappropriate for the given type of inode. */
93 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
94 {
95 if (S_ISDIR(mode))
96 return flags;
97 else if (S_ISREG(mode))
98 return flags & ~FS_DIRSYNC_FL;
99 else
100 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
101 }
102
103 /*
104 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
105 */
106 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
107 {
108 unsigned int iflags = 0;
109
110 if (flags & BTRFS_INODE_SYNC)
111 iflags |= FS_SYNC_FL;
112 if (flags & BTRFS_INODE_IMMUTABLE)
113 iflags |= FS_IMMUTABLE_FL;
114 if (flags & BTRFS_INODE_APPEND)
115 iflags |= FS_APPEND_FL;
116 if (flags & BTRFS_INODE_NODUMP)
117 iflags |= FS_NODUMP_FL;
118 if (flags & BTRFS_INODE_NOATIME)
119 iflags |= FS_NOATIME_FL;
120 if (flags & BTRFS_INODE_DIRSYNC)
121 iflags |= FS_DIRSYNC_FL;
122 if (flags & BTRFS_INODE_NODATACOW)
123 iflags |= FS_NOCOW_FL;
124
125 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
126 iflags |= FS_COMPR_FL;
127 else if (flags & BTRFS_INODE_NOCOMPRESS)
128 iflags |= FS_NOCOMP_FL;
129
130 return iflags;
131 }
132
133 /*
134 * Update inode->i_flags based on the btrfs internal flags.
135 */
136 void btrfs_update_iflags(struct inode *inode)
137 {
138 struct btrfs_inode *ip = BTRFS_I(inode);
139 unsigned int new_fl = 0;
140
141 if (ip->flags & BTRFS_INODE_SYNC)
142 new_fl |= S_SYNC;
143 if (ip->flags & BTRFS_INODE_IMMUTABLE)
144 new_fl |= S_IMMUTABLE;
145 if (ip->flags & BTRFS_INODE_APPEND)
146 new_fl |= S_APPEND;
147 if (ip->flags & BTRFS_INODE_NOATIME)
148 new_fl |= S_NOATIME;
149 if (ip->flags & BTRFS_INODE_DIRSYNC)
150 new_fl |= S_DIRSYNC;
151
152 set_mask_bits(&inode->i_flags,
153 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
154 new_fl);
155 }
156
157 /*
158 * Inherit flags from the parent inode.
159 *
160 * Currently only the compression flags and the cow flags are inherited.
161 */
162 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
163 {
164 unsigned int flags;
165
166 if (!dir)
167 return;
168
169 flags = BTRFS_I(dir)->flags;
170
171 if (flags & BTRFS_INODE_NOCOMPRESS) {
172 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
173 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
174 } else if (flags & BTRFS_INODE_COMPRESS) {
175 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
176 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
177 }
178
179 if (flags & BTRFS_INODE_NODATACOW) {
180 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
181 if (S_ISREG(inode->i_mode))
182 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
183 }
184
185 btrfs_update_iflags(inode);
186 }
187
188 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
189 {
190 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
191 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
192
193 if (copy_to_user(arg, &flags, sizeof(flags)))
194 return -EFAULT;
195 return 0;
196 }
197
198 static int check_flags(unsigned int flags)
199 {
200 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
201 FS_NOATIME_FL | FS_NODUMP_FL | \
202 FS_SYNC_FL | FS_DIRSYNC_FL | \
203 FS_NOCOMP_FL | FS_COMPR_FL |
204 FS_NOCOW_FL))
205 return -EOPNOTSUPP;
206
207 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
208 return -EINVAL;
209
210 return 0;
211 }
212
213 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
214 {
215 struct inode *inode = file_inode(file);
216 struct btrfs_inode *ip = BTRFS_I(inode);
217 struct btrfs_root *root = ip->root;
218 struct btrfs_trans_handle *trans;
219 unsigned int flags, oldflags;
220 int ret;
221 u64 ip_oldflags;
222 unsigned int i_oldflags;
223 umode_t mode;
224
225 if (!inode_owner_or_capable(inode))
226 return -EPERM;
227
228 if (btrfs_root_readonly(root))
229 return -EROFS;
230
231 if (copy_from_user(&flags, arg, sizeof(flags)))
232 return -EFAULT;
233
234 ret = check_flags(flags);
235 if (ret)
236 return ret;
237
238 ret = mnt_want_write_file(file);
239 if (ret)
240 return ret;
241
242 mutex_lock(&inode->i_mutex);
243
244 ip_oldflags = ip->flags;
245 i_oldflags = inode->i_flags;
246 mode = inode->i_mode;
247
248 flags = btrfs_mask_flags(inode->i_mode, flags);
249 oldflags = btrfs_flags_to_ioctl(ip->flags);
250 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
251 if (!capable(CAP_LINUX_IMMUTABLE)) {
252 ret = -EPERM;
253 goto out_unlock;
254 }
255 }
256
257 if (flags & FS_SYNC_FL)
258 ip->flags |= BTRFS_INODE_SYNC;
259 else
260 ip->flags &= ~BTRFS_INODE_SYNC;
261 if (flags & FS_IMMUTABLE_FL)
262 ip->flags |= BTRFS_INODE_IMMUTABLE;
263 else
264 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
265 if (flags & FS_APPEND_FL)
266 ip->flags |= BTRFS_INODE_APPEND;
267 else
268 ip->flags &= ~BTRFS_INODE_APPEND;
269 if (flags & FS_NODUMP_FL)
270 ip->flags |= BTRFS_INODE_NODUMP;
271 else
272 ip->flags &= ~BTRFS_INODE_NODUMP;
273 if (flags & FS_NOATIME_FL)
274 ip->flags |= BTRFS_INODE_NOATIME;
275 else
276 ip->flags &= ~BTRFS_INODE_NOATIME;
277 if (flags & FS_DIRSYNC_FL)
278 ip->flags |= BTRFS_INODE_DIRSYNC;
279 else
280 ip->flags &= ~BTRFS_INODE_DIRSYNC;
281 if (flags & FS_NOCOW_FL) {
282 if (S_ISREG(mode)) {
283 /*
284 * It's safe to turn csums off here, no extents exist.
285 * Otherwise we want the flag to reflect the real COW
286 * status of the file and will not set it.
287 */
288 if (inode->i_size == 0)
289 ip->flags |= BTRFS_INODE_NODATACOW
290 | BTRFS_INODE_NODATASUM;
291 } else {
292 ip->flags |= BTRFS_INODE_NODATACOW;
293 }
294 } else {
295 /*
296 * Revert back under same assuptions as above
297 */
298 if (S_ISREG(mode)) {
299 if (inode->i_size == 0)
300 ip->flags &= ~(BTRFS_INODE_NODATACOW
301 | BTRFS_INODE_NODATASUM);
302 } else {
303 ip->flags &= ~BTRFS_INODE_NODATACOW;
304 }
305 }
306
307 /*
308 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
309 * flag may be changed automatically if compression code won't make
310 * things smaller.
311 */
312 if (flags & FS_NOCOMP_FL) {
313 ip->flags &= ~BTRFS_INODE_COMPRESS;
314 ip->flags |= BTRFS_INODE_NOCOMPRESS;
315
316 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
317 if (ret && ret != -ENODATA)
318 goto out_drop;
319 } else if (flags & FS_COMPR_FL) {
320 const char *comp;
321
322 ip->flags |= BTRFS_INODE_COMPRESS;
323 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
324
325 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
326 comp = "lzo";
327 else
328 comp = "zlib";
329 ret = btrfs_set_prop(inode, "btrfs.compression",
330 comp, strlen(comp), 0);
331 if (ret)
332 goto out_drop;
333
334 } else {
335 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
336 }
337
338 trans = btrfs_start_transaction(root, 1);
339 if (IS_ERR(trans)) {
340 ret = PTR_ERR(trans);
341 goto out_drop;
342 }
343
344 btrfs_update_iflags(inode);
345 inode_inc_iversion(inode);
346 inode->i_ctime = CURRENT_TIME;
347 ret = btrfs_update_inode(trans, root, inode);
348
349 btrfs_end_transaction(trans, root);
350 out_drop:
351 if (ret) {
352 ip->flags = ip_oldflags;
353 inode->i_flags = i_oldflags;
354 }
355
356 out_unlock:
357 mutex_unlock(&inode->i_mutex);
358 mnt_drop_write_file(file);
359 return ret;
360 }
361
362 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
363 {
364 struct inode *inode = file_inode(file);
365
366 return put_user(inode->i_generation, arg);
367 }
368
369 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
370 {
371 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
372 struct btrfs_device *device;
373 struct request_queue *q;
374 struct fstrim_range range;
375 u64 minlen = ULLONG_MAX;
376 u64 num_devices = 0;
377 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
378 int ret;
379
380 if (!capable(CAP_SYS_ADMIN))
381 return -EPERM;
382
383 rcu_read_lock();
384 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
385 dev_list) {
386 if (!device->bdev)
387 continue;
388 q = bdev_get_queue(device->bdev);
389 if (blk_queue_discard(q)) {
390 num_devices++;
391 minlen = min((u64)q->limits.discard_granularity,
392 minlen);
393 }
394 }
395 rcu_read_unlock();
396
397 if (!num_devices)
398 return -EOPNOTSUPP;
399 if (copy_from_user(&range, arg, sizeof(range)))
400 return -EFAULT;
401 if (range.start > total_bytes ||
402 range.len < fs_info->sb->s_blocksize)
403 return -EINVAL;
404
405 range.len = min(range.len, total_bytes - range.start);
406 range.minlen = max(range.minlen, minlen);
407 ret = btrfs_trim_fs(fs_info->tree_root, &range);
408 if (ret < 0)
409 return ret;
410
411 if (copy_to_user(arg, &range, sizeof(range)))
412 return -EFAULT;
413
414 return 0;
415 }
416
417 int btrfs_is_empty_uuid(u8 *uuid)
418 {
419 int i;
420
421 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
422 if (uuid[i])
423 return 0;
424 }
425 return 1;
426 }
427
428 static noinline int create_subvol(struct inode *dir,
429 struct dentry *dentry,
430 char *name, int namelen,
431 u64 *async_transid,
432 struct btrfs_qgroup_inherit *inherit)
433 {
434 struct btrfs_trans_handle *trans;
435 struct btrfs_key key;
436 struct btrfs_root_item root_item;
437 struct btrfs_inode_item *inode_item;
438 struct extent_buffer *leaf;
439 struct btrfs_root *root = BTRFS_I(dir)->root;
440 struct btrfs_root *new_root;
441 struct btrfs_block_rsv block_rsv;
442 struct timespec cur_time = CURRENT_TIME;
443 struct inode *inode;
444 int ret;
445 int err;
446 u64 objectid;
447 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
448 u64 index = 0;
449 u64 qgroup_reserved;
450 uuid_le new_uuid;
451
452 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
453 if (ret)
454 return ret;
455
456 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
457 /*
458 * The same as the snapshot creation, please see the comment
459 * of create_snapshot().
460 */
461 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
462 8, &qgroup_reserved, false);
463 if (ret)
464 return ret;
465
466 trans = btrfs_start_transaction(root, 0);
467 if (IS_ERR(trans)) {
468 ret = PTR_ERR(trans);
469 btrfs_subvolume_release_metadata(root, &block_rsv,
470 qgroup_reserved);
471 return ret;
472 }
473 trans->block_rsv = &block_rsv;
474 trans->bytes_reserved = block_rsv.size;
475
476 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
477 if (ret)
478 goto fail;
479
480 leaf = btrfs_alloc_free_block(trans, root, root->nodesize,
481 0, objectid, NULL, 0, 0, 0);
482 if (IS_ERR(leaf)) {
483 ret = PTR_ERR(leaf);
484 goto fail;
485 }
486
487 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
488 btrfs_set_header_bytenr(leaf, leaf->start);
489 btrfs_set_header_generation(leaf, trans->transid);
490 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
491 btrfs_set_header_owner(leaf, objectid);
492
493 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
494 BTRFS_FSID_SIZE);
495 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
496 btrfs_header_chunk_tree_uuid(leaf),
497 BTRFS_UUID_SIZE);
498 btrfs_mark_buffer_dirty(leaf);
499
500 memset(&root_item, 0, sizeof(root_item));
501
502 inode_item = &root_item.inode;
503 btrfs_set_stack_inode_generation(inode_item, 1);
504 btrfs_set_stack_inode_size(inode_item, 3);
505 btrfs_set_stack_inode_nlink(inode_item, 1);
506 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
507 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
508
509 btrfs_set_root_flags(&root_item, 0);
510 btrfs_set_root_limit(&root_item, 0);
511 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
512
513 btrfs_set_root_bytenr(&root_item, leaf->start);
514 btrfs_set_root_generation(&root_item, trans->transid);
515 btrfs_set_root_level(&root_item, 0);
516 btrfs_set_root_refs(&root_item, 1);
517 btrfs_set_root_used(&root_item, leaf->len);
518 btrfs_set_root_last_snapshot(&root_item, 0);
519
520 btrfs_set_root_generation_v2(&root_item,
521 btrfs_root_generation(&root_item));
522 uuid_le_gen(&new_uuid);
523 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
524 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
525 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
526 root_item.ctime = root_item.otime;
527 btrfs_set_root_ctransid(&root_item, trans->transid);
528 btrfs_set_root_otransid(&root_item, trans->transid);
529
530 btrfs_tree_unlock(leaf);
531 free_extent_buffer(leaf);
532 leaf = NULL;
533
534 btrfs_set_root_dirid(&root_item, new_dirid);
535
536 key.objectid = objectid;
537 key.offset = 0;
538 key.type = BTRFS_ROOT_ITEM_KEY;
539 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
540 &root_item);
541 if (ret)
542 goto fail;
543
544 key.offset = (u64)-1;
545 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
546 if (IS_ERR(new_root)) {
547 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
548 ret = PTR_ERR(new_root);
549 goto fail;
550 }
551
552 btrfs_record_root_in_trans(trans, new_root);
553
554 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
555 if (ret) {
556 /* We potentially lose an unused inode item here */
557 btrfs_abort_transaction(trans, root, ret);
558 goto fail;
559 }
560
561 /*
562 * insert the directory item
563 */
564 ret = btrfs_set_inode_index(dir, &index);
565 if (ret) {
566 btrfs_abort_transaction(trans, root, ret);
567 goto fail;
568 }
569
570 ret = btrfs_insert_dir_item(trans, root,
571 name, namelen, dir, &key,
572 BTRFS_FT_DIR, index);
573 if (ret) {
574 btrfs_abort_transaction(trans, root, ret);
575 goto fail;
576 }
577
578 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
579 ret = btrfs_update_inode(trans, root, dir);
580 BUG_ON(ret);
581
582 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
583 objectid, root->root_key.objectid,
584 btrfs_ino(dir), index, name, namelen);
585 BUG_ON(ret);
586
587 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
588 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
589 objectid);
590 if (ret)
591 btrfs_abort_transaction(trans, root, ret);
592
593 fail:
594 trans->block_rsv = NULL;
595 trans->bytes_reserved = 0;
596 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
597
598 if (async_transid) {
599 *async_transid = trans->transid;
600 err = btrfs_commit_transaction_async(trans, root, 1);
601 if (err)
602 err = btrfs_commit_transaction(trans, root);
603 } else {
604 err = btrfs_commit_transaction(trans, root);
605 }
606 if (err && !ret)
607 ret = err;
608
609 if (!ret) {
610 inode = btrfs_lookup_dentry(dir, dentry);
611 if (IS_ERR(inode))
612 return PTR_ERR(inode);
613 d_instantiate(dentry, inode);
614 }
615 return ret;
616 }
617
618 static void btrfs_wait_nocow_write(struct btrfs_root *root)
619 {
620 s64 writers;
621 DEFINE_WAIT(wait);
622
623 do {
624 prepare_to_wait(&root->subv_writers->wait, &wait,
625 TASK_UNINTERRUPTIBLE);
626
627 writers = percpu_counter_sum(&root->subv_writers->counter);
628 if (writers)
629 schedule();
630
631 finish_wait(&root->subv_writers->wait, &wait);
632 } while (writers);
633 }
634
635 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
636 struct dentry *dentry, char *name, int namelen,
637 u64 *async_transid, bool readonly,
638 struct btrfs_qgroup_inherit *inherit)
639 {
640 struct inode *inode;
641 struct btrfs_pending_snapshot *pending_snapshot;
642 struct btrfs_trans_handle *trans;
643 int ret;
644
645 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
646 return -EINVAL;
647
648 atomic_inc(&root->will_be_snapshoted);
649 smp_mb__after_atomic();
650 btrfs_wait_nocow_write(root);
651
652 ret = btrfs_start_delalloc_inodes(root, 0);
653 if (ret)
654 goto out;
655
656 btrfs_wait_ordered_extents(root, -1);
657
658 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
659 if (!pending_snapshot) {
660 ret = -ENOMEM;
661 goto out;
662 }
663
664 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
665 BTRFS_BLOCK_RSV_TEMP);
666 /*
667 * 1 - parent dir inode
668 * 2 - dir entries
669 * 1 - root item
670 * 2 - root ref/backref
671 * 1 - root of snapshot
672 * 1 - UUID item
673 */
674 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
675 &pending_snapshot->block_rsv, 8,
676 &pending_snapshot->qgroup_reserved,
677 false);
678 if (ret)
679 goto free;
680
681 pending_snapshot->dentry = dentry;
682 pending_snapshot->root = root;
683 pending_snapshot->readonly = readonly;
684 pending_snapshot->dir = dir;
685 pending_snapshot->inherit = inherit;
686
687 trans = btrfs_start_transaction(root, 0);
688 if (IS_ERR(trans)) {
689 ret = PTR_ERR(trans);
690 goto fail;
691 }
692
693 spin_lock(&root->fs_info->trans_lock);
694 list_add(&pending_snapshot->list,
695 &trans->transaction->pending_snapshots);
696 spin_unlock(&root->fs_info->trans_lock);
697 if (async_transid) {
698 *async_transid = trans->transid;
699 ret = btrfs_commit_transaction_async(trans,
700 root->fs_info->extent_root, 1);
701 if (ret)
702 ret = btrfs_commit_transaction(trans, root);
703 } else {
704 ret = btrfs_commit_transaction(trans,
705 root->fs_info->extent_root);
706 }
707 if (ret)
708 goto fail;
709
710 ret = pending_snapshot->error;
711 if (ret)
712 goto fail;
713
714 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
715 if (IS_ERR(inode)) {
716 ret = PTR_ERR(inode);
717 goto fail;
718 }
719
720 d_instantiate(dentry, inode);
721 ret = 0;
722 fail:
723 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
724 &pending_snapshot->block_rsv,
725 pending_snapshot->qgroup_reserved);
726 free:
727 kfree(pending_snapshot);
728 out:
729 atomic_dec(&root->will_be_snapshoted);
730 return ret;
731 }
732
733 /* copy of check_sticky in fs/namei.c()
734 * It's inline, so penalty for filesystems that don't use sticky bit is
735 * minimal.
736 */
737 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
738 {
739 kuid_t fsuid = current_fsuid();
740
741 if (!(dir->i_mode & S_ISVTX))
742 return 0;
743 if (uid_eq(inode->i_uid, fsuid))
744 return 0;
745 if (uid_eq(dir->i_uid, fsuid))
746 return 0;
747 return !capable(CAP_FOWNER);
748 }
749
750 /* copy of may_delete in fs/namei.c()
751 * Check whether we can remove a link victim from directory dir, check
752 * whether the type of victim is right.
753 * 1. We can't do it if dir is read-only (done in permission())
754 * 2. We should have write and exec permissions on dir
755 * 3. We can't remove anything from append-only dir
756 * 4. We can't do anything with immutable dir (done in permission())
757 * 5. If the sticky bit on dir is set we should either
758 * a. be owner of dir, or
759 * b. be owner of victim, or
760 * c. have CAP_FOWNER capability
761 * 6. If the victim is append-only or immutable we can't do antyhing with
762 * links pointing to it.
763 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
764 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
765 * 9. We can't remove a root or mountpoint.
766 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
767 * nfs_async_unlink().
768 */
769
770 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
771 {
772 int error;
773
774 if (!victim->d_inode)
775 return -ENOENT;
776
777 BUG_ON(victim->d_parent->d_inode != dir);
778 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
779
780 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
781 if (error)
782 return error;
783 if (IS_APPEND(dir))
784 return -EPERM;
785 if (btrfs_check_sticky(dir, victim->d_inode)||
786 IS_APPEND(victim->d_inode)||
787 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
788 return -EPERM;
789 if (isdir) {
790 if (!S_ISDIR(victim->d_inode->i_mode))
791 return -ENOTDIR;
792 if (IS_ROOT(victim))
793 return -EBUSY;
794 } else if (S_ISDIR(victim->d_inode->i_mode))
795 return -EISDIR;
796 if (IS_DEADDIR(dir))
797 return -ENOENT;
798 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
799 return -EBUSY;
800 return 0;
801 }
802
803 /* copy of may_create in fs/namei.c() */
804 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
805 {
806 if (child->d_inode)
807 return -EEXIST;
808 if (IS_DEADDIR(dir))
809 return -ENOENT;
810 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
811 }
812
813 /*
814 * Create a new subvolume below @parent. This is largely modeled after
815 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
816 * inside this filesystem so it's quite a bit simpler.
817 */
818 static noinline int btrfs_mksubvol(struct path *parent,
819 char *name, int namelen,
820 struct btrfs_root *snap_src,
821 u64 *async_transid, bool readonly,
822 struct btrfs_qgroup_inherit *inherit)
823 {
824 struct inode *dir = parent->dentry->d_inode;
825 struct dentry *dentry;
826 int error;
827
828 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
829 if (error == -EINTR)
830 return error;
831
832 dentry = lookup_one_len(name, parent->dentry, namelen);
833 error = PTR_ERR(dentry);
834 if (IS_ERR(dentry))
835 goto out_unlock;
836
837 error = -EEXIST;
838 if (dentry->d_inode)
839 goto out_dput;
840
841 error = btrfs_may_create(dir, dentry);
842 if (error)
843 goto out_dput;
844
845 /*
846 * even if this name doesn't exist, we may get hash collisions.
847 * check for them now when we can safely fail
848 */
849 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
850 dir->i_ino, name,
851 namelen);
852 if (error)
853 goto out_dput;
854
855 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
856
857 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
858 goto out_up_read;
859
860 if (snap_src) {
861 error = create_snapshot(snap_src, dir, dentry, name, namelen,
862 async_transid, readonly, inherit);
863 } else {
864 error = create_subvol(dir, dentry, name, namelen,
865 async_transid, inherit);
866 }
867 if (!error)
868 fsnotify_mkdir(dir, dentry);
869 out_up_read:
870 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
871 out_dput:
872 dput(dentry);
873 out_unlock:
874 mutex_unlock(&dir->i_mutex);
875 return error;
876 }
877
878 /*
879 * When we're defragging a range, we don't want to kick it off again
880 * if it is really just waiting for delalloc to send it down.
881 * If we find a nice big extent or delalloc range for the bytes in the
882 * file you want to defrag, we return 0 to let you know to skip this
883 * part of the file
884 */
885 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
886 {
887 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
888 struct extent_map *em = NULL;
889 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
890 u64 end;
891
892 read_lock(&em_tree->lock);
893 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
894 read_unlock(&em_tree->lock);
895
896 if (em) {
897 end = extent_map_end(em);
898 free_extent_map(em);
899 if (end - offset > thresh)
900 return 0;
901 }
902 /* if we already have a nice delalloc here, just stop */
903 thresh /= 2;
904 end = count_range_bits(io_tree, &offset, offset + thresh,
905 thresh, EXTENT_DELALLOC, 1);
906 if (end >= thresh)
907 return 0;
908 return 1;
909 }
910
911 /*
912 * helper function to walk through a file and find extents
913 * newer than a specific transid, and smaller than thresh.
914 *
915 * This is used by the defragging code to find new and small
916 * extents
917 */
918 static int find_new_extents(struct btrfs_root *root,
919 struct inode *inode, u64 newer_than,
920 u64 *off, int thresh)
921 {
922 struct btrfs_path *path;
923 struct btrfs_key min_key;
924 struct extent_buffer *leaf;
925 struct btrfs_file_extent_item *extent;
926 int type;
927 int ret;
928 u64 ino = btrfs_ino(inode);
929
930 path = btrfs_alloc_path();
931 if (!path)
932 return -ENOMEM;
933
934 min_key.objectid = ino;
935 min_key.type = BTRFS_EXTENT_DATA_KEY;
936 min_key.offset = *off;
937
938 while (1) {
939 path->keep_locks = 1;
940 ret = btrfs_search_forward(root, &min_key, path, newer_than);
941 if (ret != 0)
942 goto none;
943 path->keep_locks = 0;
944 btrfs_unlock_up_safe(path, 1);
945 process_slot:
946 if (min_key.objectid != ino)
947 goto none;
948 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
949 goto none;
950
951 leaf = path->nodes[0];
952 extent = btrfs_item_ptr(leaf, path->slots[0],
953 struct btrfs_file_extent_item);
954
955 type = btrfs_file_extent_type(leaf, extent);
956 if (type == BTRFS_FILE_EXTENT_REG &&
957 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
958 check_defrag_in_cache(inode, min_key.offset, thresh)) {
959 *off = min_key.offset;
960 btrfs_free_path(path);
961 return 0;
962 }
963
964 path->slots[0]++;
965 if (path->slots[0] < btrfs_header_nritems(leaf)) {
966 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
967 goto process_slot;
968 }
969
970 if (min_key.offset == (u64)-1)
971 goto none;
972
973 min_key.offset++;
974 btrfs_release_path(path);
975 }
976 none:
977 btrfs_free_path(path);
978 return -ENOENT;
979 }
980
981 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
982 {
983 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
984 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
985 struct extent_map *em;
986 u64 len = PAGE_CACHE_SIZE;
987
988 /*
989 * hopefully we have this extent in the tree already, try without
990 * the full extent lock
991 */
992 read_lock(&em_tree->lock);
993 em = lookup_extent_mapping(em_tree, start, len);
994 read_unlock(&em_tree->lock);
995
996 if (!em) {
997 struct extent_state *cached = NULL;
998 u64 end = start + len - 1;
999
1000 /* get the big lock and read metadata off disk */
1001 lock_extent_bits(io_tree, start, end, 0, &cached);
1002 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1003 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1004
1005 if (IS_ERR(em))
1006 return NULL;
1007 }
1008
1009 return em;
1010 }
1011
1012 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1013 {
1014 struct extent_map *next;
1015 bool ret = true;
1016
1017 /* this is the last extent */
1018 if (em->start + em->len >= i_size_read(inode))
1019 return false;
1020
1021 next = defrag_lookup_extent(inode, em->start + em->len);
1022 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1023 ret = false;
1024 else if ((em->block_start + em->block_len == next->block_start) &&
1025 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
1026 ret = false;
1027
1028 free_extent_map(next);
1029 return ret;
1030 }
1031
1032 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
1033 u64 *last_len, u64 *skip, u64 *defrag_end,
1034 int compress)
1035 {
1036 struct extent_map *em;
1037 int ret = 1;
1038 bool next_mergeable = true;
1039
1040 /*
1041 * make sure that once we start defragging an extent, we keep on
1042 * defragging it
1043 */
1044 if (start < *defrag_end)
1045 return 1;
1046
1047 *skip = 0;
1048
1049 em = defrag_lookup_extent(inode, start);
1050 if (!em)
1051 return 0;
1052
1053 /* this will cover holes, and inline extents */
1054 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1055 ret = 0;
1056 goto out;
1057 }
1058
1059 next_mergeable = defrag_check_next_extent(inode, em);
1060 /*
1061 * we hit a real extent, if it is big or the next extent is not a
1062 * real extent, don't bother defragging it
1063 */
1064 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1065 (em->len >= thresh || !next_mergeable))
1066 ret = 0;
1067 out:
1068 /*
1069 * last_len ends up being a counter of how many bytes we've defragged.
1070 * every time we choose not to defrag an extent, we reset *last_len
1071 * so that the next tiny extent will force a defrag.
1072 *
1073 * The end result of this is that tiny extents before a single big
1074 * extent will force at least part of that big extent to be defragged.
1075 */
1076 if (ret) {
1077 *defrag_end = extent_map_end(em);
1078 } else {
1079 *last_len = 0;
1080 *skip = extent_map_end(em);
1081 *defrag_end = 0;
1082 }
1083
1084 free_extent_map(em);
1085 return ret;
1086 }
1087
1088 /*
1089 * it doesn't do much good to defrag one or two pages
1090 * at a time. This pulls in a nice chunk of pages
1091 * to COW and defrag.
1092 *
1093 * It also makes sure the delalloc code has enough
1094 * dirty data to avoid making new small extents as part
1095 * of the defrag
1096 *
1097 * It's a good idea to start RA on this range
1098 * before calling this.
1099 */
1100 static int cluster_pages_for_defrag(struct inode *inode,
1101 struct page **pages,
1102 unsigned long start_index,
1103 unsigned long num_pages)
1104 {
1105 unsigned long file_end;
1106 u64 isize = i_size_read(inode);
1107 u64 page_start;
1108 u64 page_end;
1109 u64 page_cnt;
1110 int ret;
1111 int i;
1112 int i_done;
1113 struct btrfs_ordered_extent *ordered;
1114 struct extent_state *cached_state = NULL;
1115 struct extent_io_tree *tree;
1116 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1117
1118 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1119 if (!isize || start_index > file_end)
1120 return 0;
1121
1122 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1123
1124 ret = btrfs_delalloc_reserve_space(inode,
1125 page_cnt << PAGE_CACHE_SHIFT);
1126 if (ret)
1127 return ret;
1128 i_done = 0;
1129 tree = &BTRFS_I(inode)->io_tree;
1130
1131 /* step one, lock all the pages */
1132 for (i = 0; i < page_cnt; i++) {
1133 struct page *page;
1134 again:
1135 page = find_or_create_page(inode->i_mapping,
1136 start_index + i, mask);
1137 if (!page)
1138 break;
1139
1140 page_start = page_offset(page);
1141 page_end = page_start + PAGE_CACHE_SIZE - 1;
1142 while (1) {
1143 lock_extent_bits(tree, page_start, page_end,
1144 0, &cached_state);
1145 ordered = btrfs_lookup_ordered_extent(inode,
1146 page_start);
1147 unlock_extent_cached(tree, page_start, page_end,
1148 &cached_state, GFP_NOFS);
1149 if (!ordered)
1150 break;
1151
1152 unlock_page(page);
1153 btrfs_start_ordered_extent(inode, ordered, 1);
1154 btrfs_put_ordered_extent(ordered);
1155 lock_page(page);
1156 /*
1157 * we unlocked the page above, so we need check if
1158 * it was released or not.
1159 */
1160 if (page->mapping != inode->i_mapping) {
1161 unlock_page(page);
1162 page_cache_release(page);
1163 goto again;
1164 }
1165 }
1166
1167 if (!PageUptodate(page)) {
1168 btrfs_readpage(NULL, page);
1169 lock_page(page);
1170 if (!PageUptodate(page)) {
1171 unlock_page(page);
1172 page_cache_release(page);
1173 ret = -EIO;
1174 break;
1175 }
1176 }
1177
1178 if (page->mapping != inode->i_mapping) {
1179 unlock_page(page);
1180 page_cache_release(page);
1181 goto again;
1182 }
1183
1184 pages[i] = page;
1185 i_done++;
1186 }
1187 if (!i_done || ret)
1188 goto out;
1189
1190 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1191 goto out;
1192
1193 /*
1194 * so now we have a nice long stream of locked
1195 * and up to date pages, lets wait on them
1196 */
1197 for (i = 0; i < i_done; i++)
1198 wait_on_page_writeback(pages[i]);
1199
1200 page_start = page_offset(pages[0]);
1201 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1202
1203 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1204 page_start, page_end - 1, 0, &cached_state);
1205 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1206 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1207 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1208 &cached_state, GFP_NOFS);
1209
1210 if (i_done != page_cnt) {
1211 spin_lock(&BTRFS_I(inode)->lock);
1212 BTRFS_I(inode)->outstanding_extents++;
1213 spin_unlock(&BTRFS_I(inode)->lock);
1214 btrfs_delalloc_release_space(inode,
1215 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1216 }
1217
1218
1219 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1220 &cached_state, GFP_NOFS);
1221
1222 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1223 page_start, page_end - 1, &cached_state,
1224 GFP_NOFS);
1225
1226 for (i = 0; i < i_done; i++) {
1227 clear_page_dirty_for_io(pages[i]);
1228 ClearPageChecked(pages[i]);
1229 set_page_extent_mapped(pages[i]);
1230 set_page_dirty(pages[i]);
1231 unlock_page(pages[i]);
1232 page_cache_release(pages[i]);
1233 }
1234 return i_done;
1235 out:
1236 for (i = 0; i < i_done; i++) {
1237 unlock_page(pages[i]);
1238 page_cache_release(pages[i]);
1239 }
1240 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1241 return ret;
1242
1243 }
1244
1245 int btrfs_defrag_file(struct inode *inode, struct file *file,
1246 struct btrfs_ioctl_defrag_range_args *range,
1247 u64 newer_than, unsigned long max_to_defrag)
1248 {
1249 struct btrfs_root *root = BTRFS_I(inode)->root;
1250 struct file_ra_state *ra = NULL;
1251 unsigned long last_index;
1252 u64 isize = i_size_read(inode);
1253 u64 last_len = 0;
1254 u64 skip = 0;
1255 u64 defrag_end = 0;
1256 u64 newer_off = range->start;
1257 unsigned long i;
1258 unsigned long ra_index = 0;
1259 int ret;
1260 int defrag_count = 0;
1261 int compress_type = BTRFS_COMPRESS_ZLIB;
1262 int extent_thresh = range->extent_thresh;
1263 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1264 unsigned long cluster = max_cluster;
1265 u64 new_align = ~((u64)128 * 1024 - 1);
1266 struct page **pages = NULL;
1267
1268 if (isize == 0)
1269 return 0;
1270
1271 if (range->start >= isize)
1272 return -EINVAL;
1273
1274 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1275 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1276 return -EINVAL;
1277 if (range->compress_type)
1278 compress_type = range->compress_type;
1279 }
1280
1281 if (extent_thresh == 0)
1282 extent_thresh = 256 * 1024;
1283
1284 /*
1285 * if we were not given a file, allocate a readahead
1286 * context
1287 */
1288 if (!file) {
1289 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1290 if (!ra)
1291 return -ENOMEM;
1292 file_ra_state_init(ra, inode->i_mapping);
1293 } else {
1294 ra = &file->f_ra;
1295 }
1296
1297 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1298 GFP_NOFS);
1299 if (!pages) {
1300 ret = -ENOMEM;
1301 goto out_ra;
1302 }
1303
1304 /* find the last page to defrag */
1305 if (range->start + range->len > range->start) {
1306 last_index = min_t(u64, isize - 1,
1307 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1308 } else {
1309 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1310 }
1311
1312 if (newer_than) {
1313 ret = find_new_extents(root, inode, newer_than,
1314 &newer_off, 64 * 1024);
1315 if (!ret) {
1316 range->start = newer_off;
1317 /*
1318 * we always align our defrag to help keep
1319 * the extents in the file evenly spaced
1320 */
1321 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1322 } else
1323 goto out_ra;
1324 } else {
1325 i = range->start >> PAGE_CACHE_SHIFT;
1326 }
1327 if (!max_to_defrag)
1328 max_to_defrag = last_index + 1;
1329
1330 /*
1331 * make writeback starts from i, so the defrag range can be
1332 * written sequentially.
1333 */
1334 if (i < inode->i_mapping->writeback_index)
1335 inode->i_mapping->writeback_index = i;
1336
1337 while (i <= last_index && defrag_count < max_to_defrag &&
1338 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
1339 /*
1340 * make sure we stop running if someone unmounts
1341 * the FS
1342 */
1343 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1344 break;
1345
1346 if (btrfs_defrag_cancelled(root->fs_info)) {
1347 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1348 ret = -EAGAIN;
1349 break;
1350 }
1351
1352 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1353 extent_thresh, &last_len, &skip,
1354 &defrag_end, range->flags &
1355 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1356 unsigned long next;
1357 /*
1358 * the should_defrag function tells us how much to skip
1359 * bump our counter by the suggested amount
1360 */
1361 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
1362 i = max(i + 1, next);
1363 continue;
1364 }
1365
1366 if (!newer_than) {
1367 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1368 PAGE_CACHE_SHIFT) - i;
1369 cluster = min(cluster, max_cluster);
1370 } else {
1371 cluster = max_cluster;
1372 }
1373
1374 if (i + cluster > ra_index) {
1375 ra_index = max(i, ra_index);
1376 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1377 cluster);
1378 ra_index += max_cluster;
1379 }
1380
1381 mutex_lock(&inode->i_mutex);
1382 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1383 BTRFS_I(inode)->force_compress = compress_type;
1384 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1385 if (ret < 0) {
1386 mutex_unlock(&inode->i_mutex);
1387 goto out_ra;
1388 }
1389
1390 defrag_count += ret;
1391 balance_dirty_pages_ratelimited(inode->i_mapping);
1392 mutex_unlock(&inode->i_mutex);
1393
1394 if (newer_than) {
1395 if (newer_off == (u64)-1)
1396 break;
1397
1398 if (ret > 0)
1399 i += ret;
1400
1401 newer_off = max(newer_off + 1,
1402 (u64)i << PAGE_CACHE_SHIFT);
1403
1404 ret = find_new_extents(root, inode,
1405 newer_than, &newer_off,
1406 64 * 1024);
1407 if (!ret) {
1408 range->start = newer_off;
1409 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1410 } else {
1411 break;
1412 }
1413 } else {
1414 if (ret > 0) {
1415 i += ret;
1416 last_len += ret << PAGE_CACHE_SHIFT;
1417 } else {
1418 i++;
1419 last_len = 0;
1420 }
1421 }
1422 }
1423
1424 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1425 filemap_flush(inode->i_mapping);
1426 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1427 &BTRFS_I(inode)->runtime_flags))
1428 filemap_flush(inode->i_mapping);
1429 }
1430
1431 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1432 /* the filemap_flush will queue IO into the worker threads, but
1433 * we have to make sure the IO is actually started and that
1434 * ordered extents get created before we return
1435 */
1436 atomic_inc(&root->fs_info->async_submit_draining);
1437 while (atomic_read(&root->fs_info->nr_async_submits) ||
1438 atomic_read(&root->fs_info->async_delalloc_pages)) {
1439 wait_event(root->fs_info->async_submit_wait,
1440 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1441 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1442 }
1443 atomic_dec(&root->fs_info->async_submit_draining);
1444 }
1445
1446 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1447 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1448 }
1449
1450 ret = defrag_count;
1451
1452 out_ra:
1453 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1454 mutex_lock(&inode->i_mutex);
1455 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1456 mutex_unlock(&inode->i_mutex);
1457 }
1458 if (!file)
1459 kfree(ra);
1460 kfree(pages);
1461 return ret;
1462 }
1463
1464 static noinline int btrfs_ioctl_resize(struct file *file,
1465 void __user *arg)
1466 {
1467 u64 new_size;
1468 u64 old_size;
1469 u64 devid = 1;
1470 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1471 struct btrfs_ioctl_vol_args *vol_args;
1472 struct btrfs_trans_handle *trans;
1473 struct btrfs_device *device = NULL;
1474 char *sizestr;
1475 char *retptr;
1476 char *devstr = NULL;
1477 int ret = 0;
1478 int mod = 0;
1479
1480 if (!capable(CAP_SYS_ADMIN))
1481 return -EPERM;
1482
1483 ret = mnt_want_write_file(file);
1484 if (ret)
1485 return ret;
1486
1487 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1488 1)) {
1489 mnt_drop_write_file(file);
1490 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1491 }
1492
1493 mutex_lock(&root->fs_info->volume_mutex);
1494 vol_args = memdup_user(arg, sizeof(*vol_args));
1495 if (IS_ERR(vol_args)) {
1496 ret = PTR_ERR(vol_args);
1497 goto out;
1498 }
1499
1500 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1501
1502 sizestr = vol_args->name;
1503 devstr = strchr(sizestr, ':');
1504 if (devstr) {
1505 sizestr = devstr + 1;
1506 *devstr = '\0';
1507 devstr = vol_args->name;
1508 ret = kstrtoull(devstr, 10, &devid);
1509 if (ret)
1510 goto out_free;
1511 if (!devid) {
1512 ret = -EINVAL;
1513 goto out_free;
1514 }
1515 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1516 }
1517
1518 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1519 if (!device) {
1520 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1521 devid);
1522 ret = -ENODEV;
1523 goto out_free;
1524 }
1525
1526 if (!device->writeable) {
1527 btrfs_info(root->fs_info,
1528 "resizer unable to apply on readonly device %llu",
1529 devid);
1530 ret = -EPERM;
1531 goto out_free;
1532 }
1533
1534 if (!strcmp(sizestr, "max"))
1535 new_size = device->bdev->bd_inode->i_size;
1536 else {
1537 if (sizestr[0] == '-') {
1538 mod = -1;
1539 sizestr++;
1540 } else if (sizestr[0] == '+') {
1541 mod = 1;
1542 sizestr++;
1543 }
1544 new_size = memparse(sizestr, &retptr);
1545 if (*retptr != '\0' || new_size == 0) {
1546 ret = -EINVAL;
1547 goto out_free;
1548 }
1549 }
1550
1551 if (device->is_tgtdev_for_dev_replace) {
1552 ret = -EPERM;
1553 goto out_free;
1554 }
1555
1556 old_size = device->total_bytes;
1557
1558 if (mod < 0) {
1559 if (new_size > old_size) {
1560 ret = -EINVAL;
1561 goto out_free;
1562 }
1563 new_size = old_size - new_size;
1564 } else if (mod > 0) {
1565 if (new_size > ULLONG_MAX - old_size) {
1566 ret = -ERANGE;
1567 goto out_free;
1568 }
1569 new_size = old_size + new_size;
1570 }
1571
1572 if (new_size < 256 * 1024 * 1024) {
1573 ret = -EINVAL;
1574 goto out_free;
1575 }
1576 if (new_size > device->bdev->bd_inode->i_size) {
1577 ret = -EFBIG;
1578 goto out_free;
1579 }
1580
1581 do_div(new_size, root->sectorsize);
1582 new_size *= root->sectorsize;
1583
1584 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1585 rcu_str_deref(device->name), new_size);
1586
1587 if (new_size > old_size) {
1588 trans = btrfs_start_transaction(root, 0);
1589 if (IS_ERR(trans)) {
1590 ret = PTR_ERR(trans);
1591 goto out_free;
1592 }
1593 ret = btrfs_grow_device(trans, device, new_size);
1594 btrfs_commit_transaction(trans, root);
1595 } else if (new_size < old_size) {
1596 ret = btrfs_shrink_device(device, new_size);
1597 } /* equal, nothing need to do */
1598
1599 out_free:
1600 kfree(vol_args);
1601 out:
1602 mutex_unlock(&root->fs_info->volume_mutex);
1603 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1604 mnt_drop_write_file(file);
1605 return ret;
1606 }
1607
1608 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1609 char *name, unsigned long fd, int subvol,
1610 u64 *transid, bool readonly,
1611 struct btrfs_qgroup_inherit *inherit)
1612 {
1613 int namelen;
1614 int ret = 0;
1615
1616 ret = mnt_want_write_file(file);
1617 if (ret)
1618 goto out;
1619
1620 namelen = strlen(name);
1621 if (strchr(name, '/')) {
1622 ret = -EINVAL;
1623 goto out_drop_write;
1624 }
1625
1626 if (name[0] == '.' &&
1627 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1628 ret = -EEXIST;
1629 goto out_drop_write;
1630 }
1631
1632 if (subvol) {
1633 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1634 NULL, transid, readonly, inherit);
1635 } else {
1636 struct fd src = fdget(fd);
1637 struct inode *src_inode;
1638 if (!src.file) {
1639 ret = -EINVAL;
1640 goto out_drop_write;
1641 }
1642
1643 src_inode = file_inode(src.file);
1644 if (src_inode->i_sb != file_inode(file)->i_sb) {
1645 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1646 "Snapshot src from another FS");
1647 ret = -EXDEV;
1648 } else if (!inode_owner_or_capable(src_inode)) {
1649 /*
1650 * Subvolume creation is not restricted, but snapshots
1651 * are limited to own subvolumes only
1652 */
1653 ret = -EPERM;
1654 } else {
1655 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1656 BTRFS_I(src_inode)->root,
1657 transid, readonly, inherit);
1658 }
1659 fdput(src);
1660 }
1661 out_drop_write:
1662 mnt_drop_write_file(file);
1663 out:
1664 return ret;
1665 }
1666
1667 static noinline int btrfs_ioctl_snap_create(struct file *file,
1668 void __user *arg, int subvol)
1669 {
1670 struct btrfs_ioctl_vol_args *vol_args;
1671 int ret;
1672
1673 vol_args = memdup_user(arg, sizeof(*vol_args));
1674 if (IS_ERR(vol_args))
1675 return PTR_ERR(vol_args);
1676 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1677
1678 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1679 vol_args->fd, subvol,
1680 NULL, false, NULL);
1681
1682 kfree(vol_args);
1683 return ret;
1684 }
1685
1686 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1687 void __user *arg, int subvol)
1688 {
1689 struct btrfs_ioctl_vol_args_v2 *vol_args;
1690 int ret;
1691 u64 transid = 0;
1692 u64 *ptr = NULL;
1693 bool readonly = false;
1694 struct btrfs_qgroup_inherit *inherit = NULL;
1695
1696 vol_args = memdup_user(arg, sizeof(*vol_args));
1697 if (IS_ERR(vol_args))
1698 return PTR_ERR(vol_args);
1699 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1700
1701 if (vol_args->flags &
1702 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1703 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1704 ret = -EOPNOTSUPP;
1705 goto free_args;
1706 }
1707
1708 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1709 ptr = &transid;
1710 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1711 readonly = true;
1712 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1713 if (vol_args->size > PAGE_CACHE_SIZE) {
1714 ret = -EINVAL;
1715 goto free_args;
1716 }
1717 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1718 if (IS_ERR(inherit)) {
1719 ret = PTR_ERR(inherit);
1720 goto free_args;
1721 }
1722 }
1723
1724 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1725 vol_args->fd, subvol, ptr,
1726 readonly, inherit);
1727 if (ret)
1728 goto free_inherit;
1729
1730 if (ptr && copy_to_user(arg +
1731 offsetof(struct btrfs_ioctl_vol_args_v2,
1732 transid),
1733 ptr, sizeof(*ptr)))
1734 ret = -EFAULT;
1735
1736 free_inherit:
1737 kfree(inherit);
1738 free_args:
1739 kfree(vol_args);
1740 return ret;
1741 }
1742
1743 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1744 void __user *arg)
1745 {
1746 struct inode *inode = file_inode(file);
1747 struct btrfs_root *root = BTRFS_I(inode)->root;
1748 int ret = 0;
1749 u64 flags = 0;
1750
1751 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1752 return -EINVAL;
1753
1754 down_read(&root->fs_info->subvol_sem);
1755 if (btrfs_root_readonly(root))
1756 flags |= BTRFS_SUBVOL_RDONLY;
1757 up_read(&root->fs_info->subvol_sem);
1758
1759 if (copy_to_user(arg, &flags, sizeof(flags)))
1760 ret = -EFAULT;
1761
1762 return ret;
1763 }
1764
1765 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1766 void __user *arg)
1767 {
1768 struct inode *inode = file_inode(file);
1769 struct btrfs_root *root = BTRFS_I(inode)->root;
1770 struct btrfs_trans_handle *trans;
1771 u64 root_flags;
1772 u64 flags;
1773 int ret = 0;
1774
1775 if (!inode_owner_or_capable(inode))
1776 return -EPERM;
1777
1778 ret = mnt_want_write_file(file);
1779 if (ret)
1780 goto out;
1781
1782 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1783 ret = -EINVAL;
1784 goto out_drop_write;
1785 }
1786
1787 if (copy_from_user(&flags, arg, sizeof(flags))) {
1788 ret = -EFAULT;
1789 goto out_drop_write;
1790 }
1791
1792 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1793 ret = -EINVAL;
1794 goto out_drop_write;
1795 }
1796
1797 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1798 ret = -EOPNOTSUPP;
1799 goto out_drop_write;
1800 }
1801
1802 down_write(&root->fs_info->subvol_sem);
1803
1804 /* nothing to do */
1805 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1806 goto out_drop_sem;
1807
1808 root_flags = btrfs_root_flags(&root->root_item);
1809 if (flags & BTRFS_SUBVOL_RDONLY) {
1810 btrfs_set_root_flags(&root->root_item,
1811 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1812 } else {
1813 /*
1814 * Block RO -> RW transition if this subvolume is involved in
1815 * send
1816 */
1817 spin_lock(&root->root_item_lock);
1818 if (root->send_in_progress == 0) {
1819 btrfs_set_root_flags(&root->root_item,
1820 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1821 spin_unlock(&root->root_item_lock);
1822 } else {
1823 spin_unlock(&root->root_item_lock);
1824 btrfs_warn(root->fs_info,
1825 "Attempt to set subvolume %llu read-write during send",
1826 root->root_key.objectid);
1827 ret = -EPERM;
1828 goto out_drop_sem;
1829 }
1830 }
1831
1832 trans = btrfs_start_transaction(root, 1);
1833 if (IS_ERR(trans)) {
1834 ret = PTR_ERR(trans);
1835 goto out_reset;
1836 }
1837
1838 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1839 &root->root_key, &root->root_item);
1840
1841 btrfs_commit_transaction(trans, root);
1842 out_reset:
1843 if (ret)
1844 btrfs_set_root_flags(&root->root_item, root_flags);
1845 out_drop_sem:
1846 up_write(&root->fs_info->subvol_sem);
1847 out_drop_write:
1848 mnt_drop_write_file(file);
1849 out:
1850 return ret;
1851 }
1852
1853 /*
1854 * helper to check if the subvolume references other subvolumes
1855 */
1856 static noinline int may_destroy_subvol(struct btrfs_root *root)
1857 {
1858 struct btrfs_path *path;
1859 struct btrfs_dir_item *di;
1860 struct btrfs_key key;
1861 u64 dir_id;
1862 int ret;
1863
1864 path = btrfs_alloc_path();
1865 if (!path)
1866 return -ENOMEM;
1867
1868 /* Make sure this root isn't set as the default subvol */
1869 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1870 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1871 dir_id, "default", 7, 0);
1872 if (di && !IS_ERR(di)) {
1873 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1874 if (key.objectid == root->root_key.objectid) {
1875 ret = -EPERM;
1876 btrfs_err(root->fs_info, "deleting default subvolume "
1877 "%llu is not allowed", key.objectid);
1878 goto out;
1879 }
1880 btrfs_release_path(path);
1881 }
1882
1883 key.objectid = root->root_key.objectid;
1884 key.type = BTRFS_ROOT_REF_KEY;
1885 key.offset = (u64)-1;
1886
1887 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1888 &key, path, 0, 0);
1889 if (ret < 0)
1890 goto out;
1891 BUG_ON(ret == 0);
1892
1893 ret = 0;
1894 if (path->slots[0] > 0) {
1895 path->slots[0]--;
1896 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1897 if (key.objectid == root->root_key.objectid &&
1898 key.type == BTRFS_ROOT_REF_KEY)
1899 ret = -ENOTEMPTY;
1900 }
1901 out:
1902 btrfs_free_path(path);
1903 return ret;
1904 }
1905
1906 static noinline int key_in_sk(struct btrfs_key *key,
1907 struct btrfs_ioctl_search_key *sk)
1908 {
1909 struct btrfs_key test;
1910 int ret;
1911
1912 test.objectid = sk->min_objectid;
1913 test.type = sk->min_type;
1914 test.offset = sk->min_offset;
1915
1916 ret = btrfs_comp_cpu_keys(key, &test);
1917 if (ret < 0)
1918 return 0;
1919
1920 test.objectid = sk->max_objectid;
1921 test.type = sk->max_type;
1922 test.offset = sk->max_offset;
1923
1924 ret = btrfs_comp_cpu_keys(key, &test);
1925 if (ret > 0)
1926 return 0;
1927 return 1;
1928 }
1929
1930 static noinline int copy_to_sk(struct btrfs_root *root,
1931 struct btrfs_path *path,
1932 struct btrfs_key *key,
1933 struct btrfs_ioctl_search_key *sk,
1934 size_t *buf_size,
1935 char __user *ubuf,
1936 unsigned long *sk_offset,
1937 int *num_found)
1938 {
1939 u64 found_transid;
1940 struct extent_buffer *leaf;
1941 struct btrfs_ioctl_search_header sh;
1942 unsigned long item_off;
1943 unsigned long item_len;
1944 int nritems;
1945 int i;
1946 int slot;
1947 int ret = 0;
1948
1949 leaf = path->nodes[0];
1950 slot = path->slots[0];
1951 nritems = btrfs_header_nritems(leaf);
1952
1953 if (btrfs_header_generation(leaf) > sk->max_transid) {
1954 i = nritems;
1955 goto advance_key;
1956 }
1957 found_transid = btrfs_header_generation(leaf);
1958
1959 for (i = slot; i < nritems; i++) {
1960 item_off = btrfs_item_ptr_offset(leaf, i);
1961 item_len = btrfs_item_size_nr(leaf, i);
1962
1963 btrfs_item_key_to_cpu(leaf, key, i);
1964 if (!key_in_sk(key, sk))
1965 continue;
1966
1967 if (sizeof(sh) + item_len > *buf_size) {
1968 if (*num_found) {
1969 ret = 1;
1970 goto out;
1971 }
1972
1973 /*
1974 * return one empty item back for v1, which does not
1975 * handle -EOVERFLOW
1976 */
1977
1978 *buf_size = sizeof(sh) + item_len;
1979 item_len = 0;
1980 ret = -EOVERFLOW;
1981 }
1982
1983 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1984 ret = 1;
1985 goto out;
1986 }
1987
1988 sh.objectid = key->objectid;
1989 sh.offset = key->offset;
1990 sh.type = key->type;
1991 sh.len = item_len;
1992 sh.transid = found_transid;
1993
1994 /* copy search result header */
1995 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1996 ret = -EFAULT;
1997 goto out;
1998 }
1999
2000 *sk_offset += sizeof(sh);
2001
2002 if (item_len) {
2003 char __user *up = ubuf + *sk_offset;
2004 /* copy the item */
2005 if (read_extent_buffer_to_user(leaf, up,
2006 item_off, item_len)) {
2007 ret = -EFAULT;
2008 goto out;
2009 }
2010
2011 *sk_offset += item_len;
2012 }
2013 (*num_found)++;
2014
2015 if (ret) /* -EOVERFLOW from above */
2016 goto out;
2017
2018 if (*num_found >= sk->nr_items) {
2019 ret = 1;
2020 goto out;
2021 }
2022 }
2023 advance_key:
2024 ret = 0;
2025 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
2026 key->offset++;
2027 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2028 key->offset = 0;
2029 key->type++;
2030 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2031 key->offset = 0;
2032 key->type = 0;
2033 key->objectid++;
2034 } else
2035 ret = 1;
2036 out:
2037 /*
2038 * 0: all items from this leaf copied, continue with next
2039 * 1: * more items can be copied, but unused buffer is too small
2040 * * all items were found
2041 * Either way, it will stops the loop which iterates to the next
2042 * leaf
2043 * -EOVERFLOW: item was to large for buffer
2044 * -EFAULT: could not copy extent buffer back to userspace
2045 */
2046 return ret;
2047 }
2048
2049 static noinline int search_ioctl(struct inode *inode,
2050 struct btrfs_ioctl_search_key *sk,
2051 size_t *buf_size,
2052 char __user *ubuf)
2053 {
2054 struct btrfs_root *root;
2055 struct btrfs_key key;
2056 struct btrfs_path *path;
2057 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2058 int ret;
2059 int num_found = 0;
2060 unsigned long sk_offset = 0;
2061
2062 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2063 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2064 return -EOVERFLOW;
2065 }
2066
2067 path = btrfs_alloc_path();
2068 if (!path)
2069 return -ENOMEM;
2070
2071 if (sk->tree_id == 0) {
2072 /* search the root of the inode that was passed */
2073 root = BTRFS_I(inode)->root;
2074 } else {
2075 key.objectid = sk->tree_id;
2076 key.type = BTRFS_ROOT_ITEM_KEY;
2077 key.offset = (u64)-1;
2078 root = btrfs_read_fs_root_no_name(info, &key);
2079 if (IS_ERR(root)) {
2080 printk(KERN_ERR "BTRFS: could not find root %llu\n",
2081 sk->tree_id);
2082 btrfs_free_path(path);
2083 return -ENOENT;
2084 }
2085 }
2086
2087 key.objectid = sk->min_objectid;
2088 key.type = sk->min_type;
2089 key.offset = sk->min_offset;
2090
2091 path->keep_locks = 1;
2092
2093 while (1) {
2094 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2095 if (ret != 0) {
2096 if (ret > 0)
2097 ret = 0;
2098 goto err;
2099 }
2100 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2101 &sk_offset, &num_found);
2102 btrfs_release_path(path);
2103 if (ret)
2104 break;
2105
2106 }
2107 if (ret > 0)
2108 ret = 0;
2109 err:
2110 sk->nr_items = num_found;
2111 btrfs_free_path(path);
2112 return ret;
2113 }
2114
2115 static noinline int btrfs_ioctl_tree_search(struct file *file,
2116 void __user *argp)
2117 {
2118 struct btrfs_ioctl_search_args __user *uargs;
2119 struct btrfs_ioctl_search_key sk;
2120 struct inode *inode;
2121 int ret;
2122 size_t buf_size;
2123
2124 if (!capable(CAP_SYS_ADMIN))
2125 return -EPERM;
2126
2127 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2128
2129 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2130 return -EFAULT;
2131
2132 buf_size = sizeof(uargs->buf);
2133
2134 inode = file_inode(file);
2135 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2136
2137 /*
2138 * In the origin implementation an overflow is handled by returning a
2139 * search header with a len of zero, so reset ret.
2140 */
2141 if (ret == -EOVERFLOW)
2142 ret = 0;
2143
2144 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2145 ret = -EFAULT;
2146 return ret;
2147 }
2148
2149 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2150 void __user *argp)
2151 {
2152 struct btrfs_ioctl_search_args_v2 __user *uarg;
2153 struct btrfs_ioctl_search_args_v2 args;
2154 struct inode *inode;
2155 int ret;
2156 size_t buf_size;
2157 const size_t buf_limit = 16 * 1024 * 1024;
2158
2159 if (!capable(CAP_SYS_ADMIN))
2160 return -EPERM;
2161
2162 /* copy search header and buffer size */
2163 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2164 if (copy_from_user(&args, uarg, sizeof(args)))
2165 return -EFAULT;
2166
2167 buf_size = args.buf_size;
2168
2169 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2170 return -EOVERFLOW;
2171
2172 /* limit result size to 16MB */
2173 if (buf_size > buf_limit)
2174 buf_size = buf_limit;
2175
2176 inode = file_inode(file);
2177 ret = search_ioctl(inode, &args.key, &buf_size,
2178 (char *)(&uarg->buf[0]));
2179 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2180 ret = -EFAULT;
2181 else if (ret == -EOVERFLOW &&
2182 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2183 ret = -EFAULT;
2184
2185 return ret;
2186 }
2187
2188 /*
2189 * Search INODE_REFs to identify path name of 'dirid' directory
2190 * in a 'tree_id' tree. and sets path name to 'name'.
2191 */
2192 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2193 u64 tree_id, u64 dirid, char *name)
2194 {
2195 struct btrfs_root *root;
2196 struct btrfs_key key;
2197 char *ptr;
2198 int ret = -1;
2199 int slot;
2200 int len;
2201 int total_len = 0;
2202 struct btrfs_inode_ref *iref;
2203 struct extent_buffer *l;
2204 struct btrfs_path *path;
2205
2206 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2207 name[0]='\0';
2208 return 0;
2209 }
2210
2211 path = btrfs_alloc_path();
2212 if (!path)
2213 return -ENOMEM;
2214
2215 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2216
2217 key.objectid = tree_id;
2218 key.type = BTRFS_ROOT_ITEM_KEY;
2219 key.offset = (u64)-1;
2220 root = btrfs_read_fs_root_no_name(info, &key);
2221 if (IS_ERR(root)) {
2222 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2223 ret = -ENOENT;
2224 goto out;
2225 }
2226
2227 key.objectid = dirid;
2228 key.type = BTRFS_INODE_REF_KEY;
2229 key.offset = (u64)-1;
2230
2231 while (1) {
2232 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2233 if (ret < 0)
2234 goto out;
2235 else if (ret > 0) {
2236 ret = btrfs_previous_item(root, path, dirid,
2237 BTRFS_INODE_REF_KEY);
2238 if (ret < 0)
2239 goto out;
2240 else if (ret > 0) {
2241 ret = -ENOENT;
2242 goto out;
2243 }
2244 }
2245
2246 l = path->nodes[0];
2247 slot = path->slots[0];
2248 btrfs_item_key_to_cpu(l, &key, slot);
2249
2250 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2251 len = btrfs_inode_ref_name_len(l, iref);
2252 ptr -= len + 1;
2253 total_len += len + 1;
2254 if (ptr < name) {
2255 ret = -ENAMETOOLONG;
2256 goto out;
2257 }
2258
2259 *(ptr + len) = '/';
2260 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2261
2262 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2263 break;
2264
2265 btrfs_release_path(path);
2266 key.objectid = key.offset;
2267 key.offset = (u64)-1;
2268 dirid = key.objectid;
2269 }
2270 memmove(name, ptr, total_len);
2271 name[total_len] = '\0';
2272 ret = 0;
2273 out:
2274 btrfs_free_path(path);
2275 return ret;
2276 }
2277
2278 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2279 void __user *argp)
2280 {
2281 struct btrfs_ioctl_ino_lookup_args *args;
2282 struct inode *inode;
2283 int ret;
2284
2285 if (!capable(CAP_SYS_ADMIN))
2286 return -EPERM;
2287
2288 args = memdup_user(argp, sizeof(*args));
2289 if (IS_ERR(args))
2290 return PTR_ERR(args);
2291
2292 inode = file_inode(file);
2293
2294 if (args->treeid == 0)
2295 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2296
2297 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2298 args->treeid, args->objectid,
2299 args->name);
2300
2301 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2302 ret = -EFAULT;
2303
2304 kfree(args);
2305 return ret;
2306 }
2307
2308 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2309 void __user *arg)
2310 {
2311 struct dentry *parent = file->f_path.dentry;
2312 struct dentry *dentry;
2313 struct inode *dir = parent->d_inode;
2314 struct inode *inode;
2315 struct btrfs_root *root = BTRFS_I(dir)->root;
2316 struct btrfs_root *dest = NULL;
2317 struct btrfs_ioctl_vol_args *vol_args;
2318 struct btrfs_trans_handle *trans;
2319 struct btrfs_block_rsv block_rsv;
2320 u64 root_flags;
2321 u64 qgroup_reserved;
2322 int namelen;
2323 int ret;
2324 int err = 0;
2325
2326 vol_args = memdup_user(arg, sizeof(*vol_args));
2327 if (IS_ERR(vol_args))
2328 return PTR_ERR(vol_args);
2329
2330 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2331 namelen = strlen(vol_args->name);
2332 if (strchr(vol_args->name, '/') ||
2333 strncmp(vol_args->name, "..", namelen) == 0) {
2334 err = -EINVAL;
2335 goto out;
2336 }
2337
2338 err = mnt_want_write_file(file);
2339 if (err)
2340 goto out;
2341
2342
2343 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2344 if (err == -EINTR)
2345 goto out_drop_write;
2346 dentry = lookup_one_len(vol_args->name, parent, namelen);
2347 if (IS_ERR(dentry)) {
2348 err = PTR_ERR(dentry);
2349 goto out_unlock_dir;
2350 }
2351
2352 if (!dentry->d_inode) {
2353 err = -ENOENT;
2354 goto out_dput;
2355 }
2356
2357 inode = dentry->d_inode;
2358 dest = BTRFS_I(inode)->root;
2359 if (!capable(CAP_SYS_ADMIN)) {
2360 /*
2361 * Regular user. Only allow this with a special mount
2362 * option, when the user has write+exec access to the
2363 * subvol root, and when rmdir(2) would have been
2364 * allowed.
2365 *
2366 * Note that this is _not_ check that the subvol is
2367 * empty or doesn't contain data that we wouldn't
2368 * otherwise be able to delete.
2369 *
2370 * Users who want to delete empty subvols should try
2371 * rmdir(2).
2372 */
2373 err = -EPERM;
2374 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2375 goto out_dput;
2376
2377 /*
2378 * Do not allow deletion if the parent dir is the same
2379 * as the dir to be deleted. That means the ioctl
2380 * must be called on the dentry referencing the root
2381 * of the subvol, not a random directory contained
2382 * within it.
2383 */
2384 err = -EINVAL;
2385 if (root == dest)
2386 goto out_dput;
2387
2388 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2389 if (err)
2390 goto out_dput;
2391 }
2392
2393 /* check if subvolume may be deleted by a user */
2394 err = btrfs_may_delete(dir, dentry, 1);
2395 if (err)
2396 goto out_dput;
2397
2398 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2399 err = -EINVAL;
2400 goto out_dput;
2401 }
2402
2403 mutex_lock(&inode->i_mutex);
2404
2405 /*
2406 * Don't allow to delete a subvolume with send in progress. This is
2407 * inside the i_mutex so the error handling that has to drop the bit
2408 * again is not run concurrently.
2409 */
2410 spin_lock(&dest->root_item_lock);
2411 root_flags = btrfs_root_flags(&dest->root_item);
2412 if (dest->send_in_progress == 0) {
2413 btrfs_set_root_flags(&dest->root_item,
2414 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2415 spin_unlock(&dest->root_item_lock);
2416 } else {
2417 spin_unlock(&dest->root_item_lock);
2418 btrfs_warn(root->fs_info,
2419 "Attempt to delete subvolume %llu during send",
2420 dest->root_key.objectid);
2421 err = -EPERM;
2422 goto out_dput;
2423 }
2424
2425 err = d_invalidate(dentry);
2426 if (err)
2427 goto out_unlock;
2428
2429 down_write(&root->fs_info->subvol_sem);
2430
2431 err = may_destroy_subvol(dest);
2432 if (err)
2433 goto out_up_write;
2434
2435 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2436 /*
2437 * One for dir inode, two for dir entries, two for root
2438 * ref/backref.
2439 */
2440 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2441 5, &qgroup_reserved, true);
2442 if (err)
2443 goto out_up_write;
2444
2445 trans = btrfs_start_transaction(root, 0);
2446 if (IS_ERR(trans)) {
2447 err = PTR_ERR(trans);
2448 goto out_release;
2449 }
2450 trans->block_rsv = &block_rsv;
2451 trans->bytes_reserved = block_rsv.size;
2452
2453 ret = btrfs_unlink_subvol(trans, root, dir,
2454 dest->root_key.objectid,
2455 dentry->d_name.name,
2456 dentry->d_name.len);
2457 if (ret) {
2458 err = ret;
2459 btrfs_abort_transaction(trans, root, ret);
2460 goto out_end_trans;
2461 }
2462
2463 btrfs_record_root_in_trans(trans, dest);
2464
2465 memset(&dest->root_item.drop_progress, 0,
2466 sizeof(dest->root_item.drop_progress));
2467 dest->root_item.drop_level = 0;
2468 btrfs_set_root_refs(&dest->root_item, 0);
2469
2470 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2471 ret = btrfs_insert_orphan_item(trans,
2472 root->fs_info->tree_root,
2473 dest->root_key.objectid);
2474 if (ret) {
2475 btrfs_abort_transaction(trans, root, ret);
2476 err = ret;
2477 goto out_end_trans;
2478 }
2479 }
2480
2481 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2482 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2483 dest->root_key.objectid);
2484 if (ret && ret != -ENOENT) {
2485 btrfs_abort_transaction(trans, root, ret);
2486 err = ret;
2487 goto out_end_trans;
2488 }
2489 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2490 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2491 dest->root_item.received_uuid,
2492 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2493 dest->root_key.objectid);
2494 if (ret && ret != -ENOENT) {
2495 btrfs_abort_transaction(trans, root, ret);
2496 err = ret;
2497 goto out_end_trans;
2498 }
2499 }
2500
2501 out_end_trans:
2502 trans->block_rsv = NULL;
2503 trans->bytes_reserved = 0;
2504 ret = btrfs_end_transaction(trans, root);
2505 if (ret && !err)
2506 err = ret;
2507 inode->i_flags |= S_DEAD;
2508 out_release:
2509 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2510 out_up_write:
2511 up_write(&root->fs_info->subvol_sem);
2512 out_unlock:
2513 if (err) {
2514 spin_lock(&dest->root_item_lock);
2515 root_flags = btrfs_root_flags(&dest->root_item);
2516 btrfs_set_root_flags(&dest->root_item,
2517 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2518 spin_unlock(&dest->root_item_lock);
2519 }
2520 mutex_unlock(&inode->i_mutex);
2521 if (!err) {
2522 shrink_dcache_sb(root->fs_info->sb);
2523 btrfs_invalidate_inodes(dest);
2524 d_delete(dentry);
2525 ASSERT(dest->send_in_progress == 0);
2526
2527 /* the last ref */
2528 if (dest->ino_cache_inode) {
2529 iput(dest->ino_cache_inode);
2530 dest->ino_cache_inode = NULL;
2531 }
2532 }
2533 out_dput:
2534 dput(dentry);
2535 out_unlock_dir:
2536 mutex_unlock(&dir->i_mutex);
2537 out_drop_write:
2538 mnt_drop_write_file(file);
2539 out:
2540 kfree(vol_args);
2541 return err;
2542 }
2543
2544 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2545 {
2546 struct inode *inode = file_inode(file);
2547 struct btrfs_root *root = BTRFS_I(inode)->root;
2548 struct btrfs_ioctl_defrag_range_args *range;
2549 int ret;
2550
2551 ret = mnt_want_write_file(file);
2552 if (ret)
2553 return ret;
2554
2555 if (btrfs_root_readonly(root)) {
2556 ret = -EROFS;
2557 goto out;
2558 }
2559
2560 switch (inode->i_mode & S_IFMT) {
2561 case S_IFDIR:
2562 if (!capable(CAP_SYS_ADMIN)) {
2563 ret = -EPERM;
2564 goto out;
2565 }
2566 ret = btrfs_defrag_root(root);
2567 if (ret)
2568 goto out;
2569 ret = btrfs_defrag_root(root->fs_info->extent_root);
2570 break;
2571 case S_IFREG:
2572 if (!(file->f_mode & FMODE_WRITE)) {
2573 ret = -EINVAL;
2574 goto out;
2575 }
2576
2577 range = kzalloc(sizeof(*range), GFP_KERNEL);
2578 if (!range) {
2579 ret = -ENOMEM;
2580 goto out;
2581 }
2582
2583 if (argp) {
2584 if (copy_from_user(range, argp,
2585 sizeof(*range))) {
2586 ret = -EFAULT;
2587 kfree(range);
2588 goto out;
2589 }
2590 /* compression requires us to start the IO */
2591 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2592 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2593 range->extent_thresh = (u32)-1;
2594 }
2595 } else {
2596 /* the rest are all set to zero by kzalloc */
2597 range->len = (u64)-1;
2598 }
2599 ret = btrfs_defrag_file(file_inode(file), file,
2600 range, 0, 0);
2601 if (ret > 0)
2602 ret = 0;
2603 kfree(range);
2604 break;
2605 default:
2606 ret = -EINVAL;
2607 }
2608 out:
2609 mnt_drop_write_file(file);
2610 return ret;
2611 }
2612
2613 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2614 {
2615 struct btrfs_ioctl_vol_args *vol_args;
2616 int ret;
2617
2618 if (!capable(CAP_SYS_ADMIN))
2619 return -EPERM;
2620
2621 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2622 1)) {
2623 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2624 }
2625
2626 mutex_lock(&root->fs_info->volume_mutex);
2627 vol_args = memdup_user(arg, sizeof(*vol_args));
2628 if (IS_ERR(vol_args)) {
2629 ret = PTR_ERR(vol_args);
2630 goto out;
2631 }
2632
2633 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2634 ret = btrfs_init_new_device(root, vol_args->name);
2635
2636 if (!ret)
2637 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2638
2639 kfree(vol_args);
2640 out:
2641 mutex_unlock(&root->fs_info->volume_mutex);
2642 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2643 return ret;
2644 }
2645
2646 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2647 {
2648 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2649 struct btrfs_ioctl_vol_args *vol_args;
2650 int ret;
2651
2652 if (!capable(CAP_SYS_ADMIN))
2653 return -EPERM;
2654
2655 ret = mnt_want_write_file(file);
2656 if (ret)
2657 return ret;
2658
2659 vol_args = memdup_user(arg, sizeof(*vol_args));
2660 if (IS_ERR(vol_args)) {
2661 ret = PTR_ERR(vol_args);
2662 goto err_drop;
2663 }
2664
2665 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2666
2667 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2668 1)) {
2669 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2670 goto out;
2671 }
2672
2673 mutex_lock(&root->fs_info->volume_mutex);
2674 ret = btrfs_rm_device(root, vol_args->name);
2675 mutex_unlock(&root->fs_info->volume_mutex);
2676 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2677
2678 if (!ret)
2679 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2680
2681 out:
2682 kfree(vol_args);
2683 err_drop:
2684 mnt_drop_write_file(file);
2685 return ret;
2686 }
2687
2688 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2689 {
2690 struct btrfs_ioctl_fs_info_args *fi_args;
2691 struct btrfs_device *device;
2692 struct btrfs_device *next;
2693 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2694 int ret = 0;
2695
2696 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2697 if (!fi_args)
2698 return -ENOMEM;
2699
2700 mutex_lock(&fs_devices->device_list_mutex);
2701 fi_args->num_devices = fs_devices->num_devices;
2702 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2703
2704 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2705 if (device->devid > fi_args->max_id)
2706 fi_args->max_id = device->devid;
2707 }
2708 mutex_unlock(&fs_devices->device_list_mutex);
2709
2710 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2711 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2712 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2713
2714 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2715 ret = -EFAULT;
2716
2717 kfree(fi_args);
2718 return ret;
2719 }
2720
2721 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2722 {
2723 struct btrfs_ioctl_dev_info_args *di_args;
2724 struct btrfs_device *dev;
2725 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2726 int ret = 0;
2727 char *s_uuid = NULL;
2728
2729 di_args = memdup_user(arg, sizeof(*di_args));
2730 if (IS_ERR(di_args))
2731 return PTR_ERR(di_args);
2732
2733 if (!btrfs_is_empty_uuid(di_args->uuid))
2734 s_uuid = di_args->uuid;
2735
2736 mutex_lock(&fs_devices->device_list_mutex);
2737 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2738
2739 if (!dev) {
2740 ret = -ENODEV;
2741 goto out;
2742 }
2743
2744 di_args->devid = dev->devid;
2745 di_args->bytes_used = dev->bytes_used;
2746 di_args->total_bytes = dev->total_bytes;
2747 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2748 if (dev->name) {
2749 struct rcu_string *name;
2750
2751 rcu_read_lock();
2752 name = rcu_dereference(dev->name);
2753 strncpy(di_args->path, name->str, sizeof(di_args->path));
2754 rcu_read_unlock();
2755 di_args->path[sizeof(di_args->path) - 1] = 0;
2756 } else {
2757 di_args->path[0] = '\0';
2758 }
2759
2760 out:
2761 mutex_unlock(&fs_devices->device_list_mutex);
2762 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2763 ret = -EFAULT;
2764
2765 kfree(di_args);
2766 return ret;
2767 }
2768
2769 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2770 {
2771 struct page *page;
2772 pgoff_t index;
2773 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2774
2775 index = off >> PAGE_CACHE_SHIFT;
2776
2777 page = grab_cache_page(inode->i_mapping, index);
2778 if (!page)
2779 return NULL;
2780
2781 if (!PageUptodate(page)) {
2782 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2783 0))
2784 return NULL;
2785 lock_page(page);
2786 if (!PageUptodate(page)) {
2787 unlock_page(page);
2788 page_cache_release(page);
2789 return NULL;
2790 }
2791 }
2792 unlock_page(page);
2793
2794 return page;
2795 }
2796
2797 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2798 {
2799 /* do any pending delalloc/csum calc on src, one way or
2800 another, and lock file content */
2801 while (1) {
2802 struct btrfs_ordered_extent *ordered;
2803 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2804 ordered = btrfs_lookup_first_ordered_extent(inode,
2805 off + len - 1);
2806 if ((!ordered ||
2807 ordered->file_offset + ordered->len <= off ||
2808 ordered->file_offset >= off + len) &&
2809 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2810 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2811 if (ordered)
2812 btrfs_put_ordered_extent(ordered);
2813 break;
2814 }
2815 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2816 if (ordered)
2817 btrfs_put_ordered_extent(ordered);
2818 btrfs_wait_ordered_range(inode, off, len);
2819 }
2820 }
2821
2822 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2823 struct inode *inode2, u64 loff2, u64 len)
2824 {
2825 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2826 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2827
2828 mutex_unlock(&inode1->i_mutex);
2829 mutex_unlock(&inode2->i_mutex);
2830 }
2831
2832 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2833 struct inode *inode2, u64 loff2, u64 len)
2834 {
2835 if (inode1 < inode2) {
2836 swap(inode1, inode2);
2837 swap(loff1, loff2);
2838 }
2839
2840 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2841 lock_extent_range(inode1, loff1, len);
2842 if (inode1 != inode2) {
2843 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2844 lock_extent_range(inode2, loff2, len);
2845 }
2846 }
2847
2848 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2849 u64 dst_loff, u64 len)
2850 {
2851 int ret = 0;
2852 struct page *src_page, *dst_page;
2853 unsigned int cmp_len = PAGE_CACHE_SIZE;
2854 void *addr, *dst_addr;
2855
2856 while (len) {
2857 if (len < PAGE_CACHE_SIZE)
2858 cmp_len = len;
2859
2860 src_page = extent_same_get_page(src, loff);
2861 if (!src_page)
2862 return -EINVAL;
2863 dst_page = extent_same_get_page(dst, dst_loff);
2864 if (!dst_page) {
2865 page_cache_release(src_page);
2866 return -EINVAL;
2867 }
2868 addr = kmap_atomic(src_page);
2869 dst_addr = kmap_atomic(dst_page);
2870
2871 flush_dcache_page(src_page);
2872 flush_dcache_page(dst_page);
2873
2874 if (memcmp(addr, dst_addr, cmp_len))
2875 ret = BTRFS_SAME_DATA_DIFFERS;
2876
2877 kunmap_atomic(addr);
2878 kunmap_atomic(dst_addr);
2879 page_cache_release(src_page);
2880 page_cache_release(dst_page);
2881
2882 if (ret)
2883 break;
2884
2885 loff += cmp_len;
2886 dst_loff += cmp_len;
2887 len -= cmp_len;
2888 }
2889
2890 return ret;
2891 }
2892
2893 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2894 {
2895 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2896
2897 if (off + len > inode->i_size || off + len < off)
2898 return -EINVAL;
2899 /* Check that we are block aligned - btrfs_clone() requires this */
2900 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2901 return -EINVAL;
2902
2903 return 0;
2904 }
2905
2906 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2907 struct inode *dst, u64 dst_loff)
2908 {
2909 int ret;
2910
2911 /*
2912 * btrfs_clone() can't handle extents in the same file
2913 * yet. Once that works, we can drop this check and replace it
2914 * with a check for the same inode, but overlapping extents.
2915 */
2916 if (src == dst)
2917 return -EINVAL;
2918
2919 btrfs_double_lock(src, loff, dst, dst_loff, len);
2920
2921 ret = extent_same_check_offsets(src, loff, len);
2922 if (ret)
2923 goto out_unlock;
2924
2925 ret = extent_same_check_offsets(dst, dst_loff, len);
2926 if (ret)
2927 goto out_unlock;
2928
2929 /* don't make the dst file partly checksummed */
2930 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2931 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2932 ret = -EINVAL;
2933 goto out_unlock;
2934 }
2935
2936 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2937 if (ret == 0)
2938 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2939
2940 out_unlock:
2941 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2942
2943 return ret;
2944 }
2945
2946 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2947
2948 static long btrfs_ioctl_file_extent_same(struct file *file,
2949 struct btrfs_ioctl_same_args __user *argp)
2950 {
2951 struct btrfs_ioctl_same_args *same;
2952 struct btrfs_ioctl_same_extent_info *info;
2953 struct inode *src = file_inode(file);
2954 u64 off;
2955 u64 len;
2956 int i;
2957 int ret;
2958 unsigned long size;
2959 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2960 bool is_admin = capable(CAP_SYS_ADMIN);
2961 u16 count;
2962
2963 if (!(file->f_mode & FMODE_READ))
2964 return -EINVAL;
2965
2966 ret = mnt_want_write_file(file);
2967 if (ret)
2968 return ret;
2969
2970 if (get_user(count, &argp->dest_count)) {
2971 ret = -EFAULT;
2972 goto out;
2973 }
2974
2975 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
2976
2977 same = memdup_user(argp, size);
2978
2979 if (IS_ERR(same)) {
2980 ret = PTR_ERR(same);
2981 goto out;
2982 }
2983
2984 off = same->logical_offset;
2985 len = same->length;
2986
2987 /*
2988 * Limit the total length we will dedupe for each operation.
2989 * This is intended to bound the total time spent in this
2990 * ioctl to something sane.
2991 */
2992 if (len > BTRFS_MAX_DEDUPE_LEN)
2993 len = BTRFS_MAX_DEDUPE_LEN;
2994
2995 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2996 /*
2997 * Btrfs does not support blocksize < page_size. As a
2998 * result, btrfs_cmp_data() won't correctly handle
2999 * this situation without an update.
3000 */
3001 ret = -EINVAL;
3002 goto out;
3003 }
3004
3005 ret = -EISDIR;
3006 if (S_ISDIR(src->i_mode))
3007 goto out;
3008
3009 ret = -EACCES;
3010 if (!S_ISREG(src->i_mode))
3011 goto out;
3012
3013 /* pre-format output fields to sane values */
3014 for (i = 0; i < count; i++) {
3015 same->info[i].bytes_deduped = 0ULL;
3016 same->info[i].status = 0;
3017 }
3018
3019 for (i = 0, info = same->info; i < count; i++, info++) {
3020 struct inode *dst;
3021 struct fd dst_file = fdget(info->fd);
3022 if (!dst_file.file) {
3023 info->status = -EBADF;
3024 continue;
3025 }
3026 dst = file_inode(dst_file.file);
3027
3028 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
3029 info->status = -EINVAL;
3030 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3031 info->status = -EXDEV;
3032 } else if (S_ISDIR(dst->i_mode)) {
3033 info->status = -EISDIR;
3034 } else if (!S_ISREG(dst->i_mode)) {
3035 info->status = -EACCES;
3036 } else {
3037 info->status = btrfs_extent_same(src, off, len, dst,
3038 info->logical_offset);
3039 if (info->status == 0)
3040 info->bytes_deduped += len;
3041 }
3042 fdput(dst_file);
3043 }
3044
3045 ret = copy_to_user(argp, same, size);
3046 if (ret)
3047 ret = -EFAULT;
3048
3049 out:
3050 mnt_drop_write_file(file);
3051 return ret;
3052 }
3053
3054 /* Helper to check and see if this root currently has a ref on the given disk
3055 * bytenr. If it does then we need to update the quota for this root. This
3056 * doesn't do anything if quotas aren't enabled.
3057 */
3058 static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3059 u64 disko)
3060 {
3061 struct seq_list tree_mod_seq_elem = {};
3062 struct ulist *roots;
3063 struct ulist_iterator uiter;
3064 struct ulist_node *root_node = NULL;
3065 int ret;
3066
3067 if (!root->fs_info->quota_enabled)
3068 return 1;
3069
3070 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3071 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3072 tree_mod_seq_elem.seq, &roots);
3073 if (ret < 0)
3074 goto out;
3075 ret = 0;
3076 ULIST_ITER_INIT(&uiter);
3077 while ((root_node = ulist_next(roots, &uiter))) {
3078 if (root_node->val == root->objectid) {
3079 ret = 1;
3080 break;
3081 }
3082 }
3083 ulist_free(roots);
3084 out:
3085 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3086 return ret;
3087 }
3088
3089 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3090 struct inode *inode,
3091 u64 endoff,
3092 const u64 destoff,
3093 const u64 olen)
3094 {
3095 struct btrfs_root *root = BTRFS_I(inode)->root;
3096 int ret;
3097
3098 inode_inc_iversion(inode);
3099 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3100 /*
3101 * We round up to the block size at eof when determining which
3102 * extents to clone above, but shouldn't round up the file size.
3103 */
3104 if (endoff > destoff + olen)
3105 endoff = destoff + olen;
3106 if (endoff > inode->i_size)
3107 btrfs_i_size_write(inode, endoff);
3108
3109 ret = btrfs_update_inode(trans, root, inode);
3110 if (ret) {
3111 btrfs_abort_transaction(trans, root, ret);
3112 btrfs_end_transaction(trans, root);
3113 goto out;
3114 }
3115 ret = btrfs_end_transaction(trans, root);
3116 out:
3117 return ret;
3118 }
3119
3120 static void clone_update_extent_map(struct inode *inode,
3121 const struct btrfs_trans_handle *trans,
3122 const struct btrfs_path *path,
3123 const u64 hole_offset,
3124 const u64 hole_len)
3125 {
3126 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3127 struct extent_map *em;
3128 int ret;
3129
3130 em = alloc_extent_map();
3131 if (!em) {
3132 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3133 &BTRFS_I(inode)->runtime_flags);
3134 return;
3135 }
3136
3137 if (path) {
3138 struct btrfs_file_extent_item *fi;
3139
3140 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3141 struct btrfs_file_extent_item);
3142 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3143 em->generation = -1;
3144 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3145 BTRFS_FILE_EXTENT_INLINE)
3146 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3147 &BTRFS_I(inode)->runtime_flags);
3148 } else {
3149 em->start = hole_offset;
3150 em->len = hole_len;
3151 em->ram_bytes = em->len;
3152 em->orig_start = hole_offset;
3153 em->block_start = EXTENT_MAP_HOLE;
3154 em->block_len = 0;
3155 em->orig_block_len = 0;
3156 em->compress_type = BTRFS_COMPRESS_NONE;
3157 em->generation = trans->transid;
3158 }
3159
3160 while (1) {
3161 write_lock(&em_tree->lock);
3162 ret = add_extent_mapping(em_tree, em, 1);
3163 write_unlock(&em_tree->lock);
3164 if (ret != -EEXIST) {
3165 free_extent_map(em);
3166 break;
3167 }
3168 btrfs_drop_extent_cache(inode, em->start,
3169 em->start + em->len - 1, 0);
3170 }
3171
3172 if (unlikely(ret))
3173 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3174 &BTRFS_I(inode)->runtime_flags);
3175 }
3176
3177 /**
3178 * btrfs_clone() - clone a range from inode file to another
3179 *
3180 * @src: Inode to clone from
3181 * @inode: Inode to clone to
3182 * @off: Offset within source to start clone from
3183 * @olen: Original length, passed by user, of range to clone
3184 * @olen_aligned: Block-aligned value of olen, extent_same uses
3185 * identical values here
3186 * @destoff: Offset within @inode to start clone
3187 */
3188 static int btrfs_clone(struct inode *src, struct inode *inode,
3189 const u64 off, const u64 olen, const u64 olen_aligned,
3190 const u64 destoff)
3191 {
3192 struct btrfs_root *root = BTRFS_I(inode)->root;
3193 struct btrfs_path *path = NULL;
3194 struct extent_buffer *leaf;
3195 struct btrfs_trans_handle *trans;
3196 char *buf = NULL;
3197 struct btrfs_key key;
3198 u32 nritems;
3199 int slot;
3200 int ret;
3201 int no_quota;
3202 const u64 len = olen_aligned;
3203 u64 last_disko = 0;
3204 u64 last_dest_end = destoff;
3205
3206 ret = -ENOMEM;
3207 buf = vmalloc(root->nodesize);
3208 if (!buf)
3209 return ret;
3210
3211 path = btrfs_alloc_path();
3212 if (!path) {
3213 vfree(buf);
3214 return ret;
3215 }
3216
3217 path->reada = 2;
3218 /* clone data */
3219 key.objectid = btrfs_ino(src);
3220 key.type = BTRFS_EXTENT_DATA_KEY;
3221 key.offset = off;
3222
3223 while (1) {
3224 /*
3225 * note the key will change type as we walk through the
3226 * tree.
3227 */
3228 path->leave_spinning = 1;
3229 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3230 0, 0);
3231 if (ret < 0)
3232 goto out;
3233 /*
3234 * First search, if no extent item that starts at offset off was
3235 * found but the previous item is an extent item, it's possible
3236 * it might overlap our target range, therefore process it.
3237 */
3238 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3239 btrfs_item_key_to_cpu(path->nodes[0], &key,
3240 path->slots[0] - 1);
3241 if (key.type == BTRFS_EXTENT_DATA_KEY)
3242 path->slots[0]--;
3243 }
3244
3245 nritems = btrfs_header_nritems(path->nodes[0]);
3246 process_slot:
3247 no_quota = 1;
3248 if (path->slots[0] >= nritems) {
3249 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3250 if (ret < 0)
3251 goto out;
3252 if (ret > 0)
3253 break;
3254 nritems = btrfs_header_nritems(path->nodes[0]);
3255 }
3256 leaf = path->nodes[0];
3257 slot = path->slots[0];
3258
3259 btrfs_item_key_to_cpu(leaf, &key, slot);
3260 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3261 key.objectid != btrfs_ino(src))
3262 break;
3263
3264 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3265 struct btrfs_file_extent_item *extent;
3266 int type;
3267 u32 size;
3268 struct btrfs_key new_key;
3269 u64 disko = 0, diskl = 0;
3270 u64 datao = 0, datal = 0;
3271 u8 comp;
3272 u64 drop_start;
3273
3274 extent = btrfs_item_ptr(leaf, slot,
3275 struct btrfs_file_extent_item);
3276 comp = btrfs_file_extent_compression(leaf, extent);
3277 type = btrfs_file_extent_type(leaf, extent);
3278 if (type == BTRFS_FILE_EXTENT_REG ||
3279 type == BTRFS_FILE_EXTENT_PREALLOC) {
3280 disko = btrfs_file_extent_disk_bytenr(leaf,
3281 extent);
3282 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3283 extent);
3284 datao = btrfs_file_extent_offset(leaf, extent);
3285 datal = btrfs_file_extent_num_bytes(leaf,
3286 extent);
3287 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3288 /* take upper bound, may be compressed */
3289 datal = btrfs_file_extent_ram_bytes(leaf,
3290 extent);
3291 }
3292
3293 /*
3294 * The first search might have left us at an extent
3295 * item that ends before our target range's start, can
3296 * happen if we have holes and NO_HOLES feature enabled.
3297 */
3298 if (key.offset + datal <= off) {
3299 path->slots[0]++;
3300 goto process_slot;
3301 } else if (key.offset >= off + len) {
3302 break;
3303 }
3304
3305 size = btrfs_item_size_nr(leaf, slot);
3306 read_extent_buffer(leaf, buf,
3307 btrfs_item_ptr_offset(leaf, slot),
3308 size);
3309
3310 btrfs_release_path(path);
3311 path->leave_spinning = 0;
3312
3313 memcpy(&new_key, &key, sizeof(new_key));
3314 new_key.objectid = btrfs_ino(inode);
3315 if (off <= key.offset)
3316 new_key.offset = key.offset + destoff - off;
3317 else
3318 new_key.offset = destoff;
3319
3320 /*
3321 * Deal with a hole that doesn't have an extent item
3322 * that represents it (NO_HOLES feature enabled).
3323 * This hole is either in the middle of the cloning
3324 * range or at the beginning (fully overlaps it or
3325 * partially overlaps it).
3326 */
3327 if (new_key.offset != last_dest_end)
3328 drop_start = last_dest_end;
3329 else
3330 drop_start = new_key.offset;
3331
3332 /*
3333 * 1 - adjusting old extent (we may have to split it)
3334 * 1 - add new extent
3335 * 1 - inode update
3336 */
3337 trans = btrfs_start_transaction(root, 3);
3338 if (IS_ERR(trans)) {
3339 ret = PTR_ERR(trans);
3340 goto out;
3341 }
3342
3343 if (type == BTRFS_FILE_EXTENT_REG ||
3344 type == BTRFS_FILE_EXTENT_PREALLOC) {
3345 /*
3346 * a | --- range to clone ---| b
3347 * | ------------- extent ------------- |
3348 */
3349
3350 /* subtract range b */
3351 if (key.offset + datal > off + len)
3352 datal = off + len - key.offset;
3353
3354 /* subtract range a */
3355 if (off > key.offset) {
3356 datao += off - key.offset;
3357 datal -= off - key.offset;
3358 }
3359
3360 ret = btrfs_drop_extents(trans, root, inode,
3361 drop_start,
3362 new_key.offset + datal,
3363 1);
3364 if (ret) {
3365 if (ret != -EOPNOTSUPP)
3366 btrfs_abort_transaction(trans,
3367 root, ret);
3368 btrfs_end_transaction(trans, root);
3369 goto out;
3370 }
3371
3372 ret = btrfs_insert_empty_item(trans, root, path,
3373 &new_key, size);
3374 if (ret) {
3375 btrfs_abort_transaction(trans, root,
3376 ret);
3377 btrfs_end_transaction(trans, root);
3378 goto out;
3379 }
3380
3381 leaf = path->nodes[0];
3382 slot = path->slots[0];
3383 write_extent_buffer(leaf, buf,
3384 btrfs_item_ptr_offset(leaf, slot),
3385 size);
3386
3387 extent = btrfs_item_ptr(leaf, slot,
3388 struct btrfs_file_extent_item);
3389
3390 /* disko == 0 means it's a hole */
3391 if (!disko)
3392 datao = 0;
3393
3394 btrfs_set_file_extent_offset(leaf, extent,
3395 datao);
3396 btrfs_set_file_extent_num_bytes(leaf, extent,
3397 datal);
3398
3399 /*
3400 * We need to look up the roots that point at
3401 * this bytenr and see if the new root does. If
3402 * it does not we need to make sure we update
3403 * quotas appropriately.
3404 */
3405 if (disko && root != BTRFS_I(src)->root &&
3406 disko != last_disko) {
3407 no_quota = check_ref(trans, root,
3408 disko);
3409 if (no_quota < 0) {
3410 btrfs_abort_transaction(trans,
3411 root,
3412 ret);
3413 btrfs_end_transaction(trans,
3414 root);
3415 ret = no_quota;
3416 goto out;
3417 }
3418 }
3419
3420 if (disko) {
3421 inode_add_bytes(inode, datal);
3422 ret = btrfs_inc_extent_ref(trans, root,
3423 disko, diskl, 0,
3424 root->root_key.objectid,
3425 btrfs_ino(inode),
3426 new_key.offset - datao,
3427 no_quota);
3428 if (ret) {
3429 btrfs_abort_transaction(trans,
3430 root,
3431 ret);
3432 btrfs_end_transaction(trans,
3433 root);
3434 goto out;
3435
3436 }
3437 }
3438 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3439 u64 skip = 0;
3440 u64 trim = 0;
3441 u64 aligned_end = 0;
3442
3443 if (off > key.offset) {
3444 skip = off - key.offset;
3445 new_key.offset += skip;
3446 }
3447
3448 if (key.offset + datal > off + len)
3449 trim = key.offset + datal - (off + len);
3450
3451 if (comp && (skip || trim)) {
3452 ret = -EINVAL;
3453 btrfs_end_transaction(trans, root);
3454 goto out;
3455 }
3456 size -= skip + trim;
3457 datal -= skip + trim;
3458
3459 aligned_end = ALIGN(new_key.offset + datal,
3460 root->sectorsize);
3461 ret = btrfs_drop_extents(trans, root, inode,
3462 drop_start,
3463 aligned_end,
3464 1);
3465 if (ret) {
3466 if (ret != -EOPNOTSUPP)
3467 btrfs_abort_transaction(trans,
3468 root, ret);
3469 btrfs_end_transaction(trans, root);
3470 goto out;
3471 }
3472
3473 ret = btrfs_insert_empty_item(trans, root, path,
3474 &new_key, size);
3475 if (ret) {
3476 btrfs_abort_transaction(trans, root,
3477 ret);
3478 btrfs_end_transaction(trans, root);
3479 goto out;
3480 }
3481
3482 if (skip) {
3483 u32 start =
3484 btrfs_file_extent_calc_inline_size(0);
3485 memmove(buf+start, buf+start+skip,
3486 datal);
3487 }
3488
3489 leaf = path->nodes[0];
3490 slot = path->slots[0];
3491 write_extent_buffer(leaf, buf,
3492 btrfs_item_ptr_offset(leaf, slot),
3493 size);
3494 inode_add_bytes(inode, datal);
3495 }
3496
3497 /* If we have an implicit hole (NO_HOLES feature). */
3498 if (drop_start < new_key.offset)
3499 clone_update_extent_map(inode, trans,
3500 NULL, drop_start,
3501 new_key.offset - drop_start);
3502
3503 clone_update_extent_map(inode, trans, path, 0, 0);
3504
3505 btrfs_mark_buffer_dirty(leaf);
3506 btrfs_release_path(path);
3507
3508 last_dest_end = ALIGN(new_key.offset + datal,
3509 root->sectorsize);
3510 ret = clone_finish_inode_update(trans, inode,
3511 last_dest_end,
3512 destoff, olen);
3513 if (ret)
3514 goto out;
3515 if (new_key.offset + datal >= destoff + len)
3516 break;
3517 }
3518 btrfs_release_path(path);
3519 key.offset++;
3520 }
3521 ret = 0;
3522
3523 if (last_dest_end < destoff + len) {
3524 /*
3525 * We have an implicit hole (NO_HOLES feature is enabled) that
3526 * fully or partially overlaps our cloning range at its end.
3527 */
3528 btrfs_release_path(path);
3529
3530 /*
3531 * 1 - remove extent(s)
3532 * 1 - inode update
3533 */
3534 trans = btrfs_start_transaction(root, 2);
3535 if (IS_ERR(trans)) {
3536 ret = PTR_ERR(trans);
3537 goto out;
3538 }
3539 ret = btrfs_drop_extents(trans, root, inode,
3540 last_dest_end, destoff + len, 1);
3541 if (ret) {
3542 if (ret != -EOPNOTSUPP)
3543 btrfs_abort_transaction(trans, root, ret);
3544 btrfs_end_transaction(trans, root);
3545 goto out;
3546 }
3547 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3548 destoff + len - last_dest_end);
3549 ret = clone_finish_inode_update(trans, inode, destoff + len,
3550 destoff, olen);
3551 }
3552
3553 out:
3554 btrfs_free_path(path);
3555 vfree(buf);
3556 return ret;
3557 }
3558
3559 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3560 u64 off, u64 olen, u64 destoff)
3561 {
3562 struct inode *inode = file_inode(file);
3563 struct btrfs_root *root = BTRFS_I(inode)->root;
3564 struct fd src_file;
3565 struct inode *src;
3566 int ret;
3567 u64 len = olen;
3568 u64 bs = root->fs_info->sb->s_blocksize;
3569 int same_inode = 0;
3570
3571 /*
3572 * TODO:
3573 * - split compressed inline extents. annoying: we need to
3574 * decompress into destination's address_space (the file offset
3575 * may change, so source mapping won't do), then recompress (or
3576 * otherwise reinsert) a subrange.
3577 *
3578 * - split destination inode's inline extents. The inline extents can
3579 * be either compressed or non-compressed.
3580 */
3581
3582 /* the destination must be opened for writing */
3583 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3584 return -EINVAL;
3585
3586 if (btrfs_root_readonly(root))
3587 return -EROFS;
3588
3589 ret = mnt_want_write_file(file);
3590 if (ret)
3591 return ret;
3592
3593 src_file = fdget(srcfd);
3594 if (!src_file.file) {
3595 ret = -EBADF;
3596 goto out_drop_write;
3597 }
3598
3599 ret = -EXDEV;
3600 if (src_file.file->f_path.mnt != file->f_path.mnt)
3601 goto out_fput;
3602
3603 src = file_inode(src_file.file);
3604
3605 ret = -EINVAL;
3606 if (src == inode)
3607 same_inode = 1;
3608
3609 /* the src must be open for reading */
3610 if (!(src_file.file->f_mode & FMODE_READ))
3611 goto out_fput;
3612
3613 /* don't make the dst file partly checksummed */
3614 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3615 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3616 goto out_fput;
3617
3618 ret = -EISDIR;
3619 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3620 goto out_fput;
3621
3622 ret = -EXDEV;
3623 if (src->i_sb != inode->i_sb)
3624 goto out_fput;
3625
3626 if (!same_inode) {
3627 if (inode < src) {
3628 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3629 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3630 } else {
3631 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3632 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3633 }
3634 } else {
3635 mutex_lock(&src->i_mutex);
3636 }
3637
3638 /* determine range to clone */
3639 ret = -EINVAL;
3640 if (off + len > src->i_size || off + len < off)
3641 goto out_unlock;
3642 if (len == 0)
3643 olen = len = src->i_size - off;
3644 /* if we extend to eof, continue to block boundary */
3645 if (off + len == src->i_size)
3646 len = ALIGN(src->i_size, bs) - off;
3647
3648 /* verify the end result is block aligned */
3649 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3650 !IS_ALIGNED(destoff, bs))
3651 goto out_unlock;
3652
3653 /* verify if ranges are overlapped within the same file */
3654 if (same_inode) {
3655 if (destoff + len > off && destoff < off + len)
3656 goto out_unlock;
3657 }
3658
3659 if (destoff > inode->i_size) {
3660 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3661 if (ret)
3662 goto out_unlock;
3663 }
3664
3665 /*
3666 * Lock the target range too. Right after we replace the file extent
3667 * items in the fs tree (which now point to the cloned data), we might
3668 * have a worker replace them with extent items relative to a write
3669 * operation that was issued before this clone operation (i.e. confront
3670 * with inode.c:btrfs_finish_ordered_io).
3671 */
3672 if (same_inode) {
3673 u64 lock_start = min_t(u64, off, destoff);
3674 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3675
3676 lock_extent_range(src, lock_start, lock_len);
3677 } else {
3678 lock_extent_range(src, off, len);
3679 lock_extent_range(inode, destoff, len);
3680 }
3681
3682 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3683
3684 if (same_inode) {
3685 u64 lock_start = min_t(u64, off, destoff);
3686 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3687
3688 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3689 } else {
3690 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3691 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3692 destoff + len - 1);
3693 }
3694 /*
3695 * Truncate page cache pages so that future reads will see the cloned
3696 * data immediately and not the previous data.
3697 */
3698 truncate_inode_pages_range(&inode->i_data, destoff,
3699 PAGE_CACHE_ALIGN(destoff + len) - 1);
3700 out_unlock:
3701 if (!same_inode) {
3702 if (inode < src) {
3703 mutex_unlock(&src->i_mutex);
3704 mutex_unlock(&inode->i_mutex);
3705 } else {
3706 mutex_unlock(&inode->i_mutex);
3707 mutex_unlock(&src->i_mutex);
3708 }
3709 } else {
3710 mutex_unlock(&src->i_mutex);
3711 }
3712 out_fput:
3713 fdput(src_file);
3714 out_drop_write:
3715 mnt_drop_write_file(file);
3716 return ret;
3717 }
3718
3719 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3720 {
3721 struct btrfs_ioctl_clone_range_args args;
3722
3723 if (copy_from_user(&args, argp, sizeof(args)))
3724 return -EFAULT;
3725 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3726 args.src_length, args.dest_offset);
3727 }
3728
3729 /*
3730 * there are many ways the trans_start and trans_end ioctls can lead
3731 * to deadlocks. They should only be used by applications that
3732 * basically own the machine, and have a very in depth understanding
3733 * of all the possible deadlocks and enospc problems.
3734 */
3735 static long btrfs_ioctl_trans_start(struct file *file)
3736 {
3737 struct inode *inode = file_inode(file);
3738 struct btrfs_root *root = BTRFS_I(inode)->root;
3739 struct btrfs_trans_handle *trans;
3740 int ret;
3741
3742 ret = -EPERM;
3743 if (!capable(CAP_SYS_ADMIN))
3744 goto out;
3745
3746 ret = -EINPROGRESS;
3747 if (file->private_data)
3748 goto out;
3749
3750 ret = -EROFS;
3751 if (btrfs_root_readonly(root))
3752 goto out;
3753
3754 ret = mnt_want_write_file(file);
3755 if (ret)
3756 goto out;
3757
3758 atomic_inc(&root->fs_info->open_ioctl_trans);
3759
3760 ret = -ENOMEM;
3761 trans = btrfs_start_ioctl_transaction(root);
3762 if (IS_ERR(trans))
3763 goto out_drop;
3764
3765 file->private_data = trans;
3766 return 0;
3767
3768 out_drop:
3769 atomic_dec(&root->fs_info->open_ioctl_trans);
3770 mnt_drop_write_file(file);
3771 out:
3772 return ret;
3773 }
3774
3775 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3776 {
3777 struct inode *inode = file_inode(file);
3778 struct btrfs_root *root = BTRFS_I(inode)->root;
3779 struct btrfs_root *new_root;
3780 struct btrfs_dir_item *di;
3781 struct btrfs_trans_handle *trans;
3782 struct btrfs_path *path;
3783 struct btrfs_key location;
3784 struct btrfs_disk_key disk_key;
3785 u64 objectid = 0;
3786 u64 dir_id;
3787 int ret;
3788
3789 if (!capable(CAP_SYS_ADMIN))
3790 return -EPERM;
3791
3792 ret = mnt_want_write_file(file);
3793 if (ret)
3794 return ret;
3795
3796 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3797 ret = -EFAULT;
3798 goto out;
3799 }
3800
3801 if (!objectid)
3802 objectid = BTRFS_FS_TREE_OBJECTID;
3803
3804 location.objectid = objectid;
3805 location.type = BTRFS_ROOT_ITEM_KEY;
3806 location.offset = (u64)-1;
3807
3808 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3809 if (IS_ERR(new_root)) {
3810 ret = PTR_ERR(new_root);
3811 goto out;
3812 }
3813
3814 path = btrfs_alloc_path();
3815 if (!path) {
3816 ret = -ENOMEM;
3817 goto out;
3818 }
3819 path->leave_spinning = 1;
3820
3821 trans = btrfs_start_transaction(root, 1);
3822 if (IS_ERR(trans)) {
3823 btrfs_free_path(path);
3824 ret = PTR_ERR(trans);
3825 goto out;
3826 }
3827
3828 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3829 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3830 dir_id, "default", 7, 1);
3831 if (IS_ERR_OR_NULL(di)) {
3832 btrfs_free_path(path);
3833 btrfs_end_transaction(trans, root);
3834 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3835 "item, this isn't going to work");
3836 ret = -ENOENT;
3837 goto out;
3838 }
3839
3840 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3841 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3842 btrfs_mark_buffer_dirty(path->nodes[0]);
3843 btrfs_free_path(path);
3844
3845 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3846 btrfs_end_transaction(trans, root);
3847 out:
3848 mnt_drop_write_file(file);
3849 return ret;
3850 }
3851
3852 void btrfs_get_block_group_info(struct list_head *groups_list,
3853 struct btrfs_ioctl_space_info *space)
3854 {
3855 struct btrfs_block_group_cache *block_group;
3856
3857 space->total_bytes = 0;
3858 space->used_bytes = 0;
3859 space->flags = 0;
3860 list_for_each_entry(block_group, groups_list, list) {
3861 space->flags = block_group->flags;
3862 space->total_bytes += block_group->key.offset;
3863 space->used_bytes +=
3864 btrfs_block_group_used(&block_group->item);
3865 }
3866 }
3867
3868 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3869 {
3870 struct btrfs_ioctl_space_args space_args;
3871 struct btrfs_ioctl_space_info space;
3872 struct btrfs_ioctl_space_info *dest;
3873 struct btrfs_ioctl_space_info *dest_orig;
3874 struct btrfs_ioctl_space_info __user *user_dest;
3875 struct btrfs_space_info *info;
3876 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3877 BTRFS_BLOCK_GROUP_SYSTEM,
3878 BTRFS_BLOCK_GROUP_METADATA,
3879 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3880 int num_types = 4;
3881 int alloc_size;
3882 int ret = 0;
3883 u64 slot_count = 0;
3884 int i, c;
3885
3886 if (copy_from_user(&space_args,
3887 (struct btrfs_ioctl_space_args __user *)arg,
3888 sizeof(space_args)))
3889 return -EFAULT;
3890
3891 for (i = 0; i < num_types; i++) {
3892 struct btrfs_space_info *tmp;
3893
3894 info = NULL;
3895 rcu_read_lock();
3896 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3897 list) {
3898 if (tmp->flags == types[i]) {
3899 info = tmp;
3900 break;
3901 }
3902 }
3903 rcu_read_unlock();
3904
3905 if (!info)
3906 continue;
3907
3908 down_read(&info->groups_sem);
3909 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3910 if (!list_empty(&info->block_groups[c]))
3911 slot_count++;
3912 }
3913 up_read(&info->groups_sem);
3914 }
3915
3916 /*
3917 * Global block reserve, exported as a space_info
3918 */
3919 slot_count++;
3920
3921 /* space_slots == 0 means they are asking for a count */
3922 if (space_args.space_slots == 0) {
3923 space_args.total_spaces = slot_count;
3924 goto out;
3925 }
3926
3927 slot_count = min_t(u64, space_args.space_slots, slot_count);
3928
3929 alloc_size = sizeof(*dest) * slot_count;
3930
3931 /* we generally have at most 6 or so space infos, one for each raid
3932 * level. So, a whole page should be more than enough for everyone
3933 */
3934 if (alloc_size > PAGE_CACHE_SIZE)
3935 return -ENOMEM;
3936
3937 space_args.total_spaces = 0;
3938 dest = kmalloc(alloc_size, GFP_NOFS);
3939 if (!dest)
3940 return -ENOMEM;
3941 dest_orig = dest;
3942
3943 /* now we have a buffer to copy into */
3944 for (i = 0; i < num_types; i++) {
3945 struct btrfs_space_info *tmp;
3946
3947 if (!slot_count)
3948 break;
3949
3950 info = NULL;
3951 rcu_read_lock();
3952 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3953 list) {
3954 if (tmp->flags == types[i]) {
3955 info = tmp;
3956 break;
3957 }
3958 }
3959 rcu_read_unlock();
3960
3961 if (!info)
3962 continue;
3963 down_read(&info->groups_sem);
3964 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3965 if (!list_empty(&info->block_groups[c])) {
3966 btrfs_get_block_group_info(
3967 &info->block_groups[c], &space);
3968 memcpy(dest, &space, sizeof(space));
3969 dest++;
3970 space_args.total_spaces++;
3971 slot_count--;
3972 }
3973 if (!slot_count)
3974 break;
3975 }
3976 up_read(&info->groups_sem);
3977 }
3978
3979 /*
3980 * Add global block reserve
3981 */
3982 if (slot_count) {
3983 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3984
3985 spin_lock(&block_rsv->lock);
3986 space.total_bytes = block_rsv->size;
3987 space.used_bytes = block_rsv->size - block_rsv->reserved;
3988 spin_unlock(&block_rsv->lock);
3989 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3990 memcpy(dest, &space, sizeof(space));
3991 space_args.total_spaces++;
3992 }
3993
3994 user_dest = (struct btrfs_ioctl_space_info __user *)
3995 (arg + sizeof(struct btrfs_ioctl_space_args));
3996
3997 if (copy_to_user(user_dest, dest_orig, alloc_size))
3998 ret = -EFAULT;
3999
4000 kfree(dest_orig);
4001 out:
4002 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4003 ret = -EFAULT;
4004
4005 return ret;
4006 }
4007
4008 /*
4009 * there are many ways the trans_start and trans_end ioctls can lead
4010 * to deadlocks. They should only be used by applications that
4011 * basically own the machine, and have a very in depth understanding
4012 * of all the possible deadlocks and enospc problems.
4013 */
4014 long btrfs_ioctl_trans_end(struct file *file)
4015 {
4016 struct inode *inode = file_inode(file);
4017 struct btrfs_root *root = BTRFS_I(inode)->root;
4018 struct btrfs_trans_handle *trans;
4019
4020 trans = file->private_data;
4021 if (!trans)
4022 return -EINVAL;
4023 file->private_data = NULL;
4024
4025 btrfs_end_transaction(trans, root);
4026
4027 atomic_dec(&root->fs_info->open_ioctl_trans);
4028
4029 mnt_drop_write_file(file);
4030 return 0;
4031 }
4032
4033 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4034 void __user *argp)
4035 {
4036 struct btrfs_trans_handle *trans;
4037 u64 transid;
4038 int ret;
4039
4040 trans = btrfs_attach_transaction_barrier(root);
4041 if (IS_ERR(trans)) {
4042 if (PTR_ERR(trans) != -ENOENT)
4043 return PTR_ERR(trans);
4044
4045 /* No running transaction, don't bother */
4046 transid = root->fs_info->last_trans_committed;
4047 goto out;
4048 }
4049 transid = trans->transid;
4050 ret = btrfs_commit_transaction_async(trans, root, 0);
4051 if (ret) {
4052 btrfs_end_transaction(trans, root);
4053 return ret;
4054 }
4055 out:
4056 if (argp)
4057 if (copy_to_user(argp, &transid, sizeof(transid)))
4058 return -EFAULT;
4059 return 0;
4060 }
4061
4062 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4063 void __user *argp)
4064 {
4065 u64 transid;
4066
4067 if (argp) {
4068 if (copy_from_user(&transid, argp, sizeof(transid)))
4069 return -EFAULT;
4070 } else {
4071 transid = 0; /* current trans */
4072 }
4073 return btrfs_wait_for_commit(root, transid);
4074 }
4075
4076 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4077 {
4078 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4079 struct btrfs_ioctl_scrub_args *sa;
4080 int ret;
4081
4082 if (!capable(CAP_SYS_ADMIN))
4083 return -EPERM;
4084
4085 sa = memdup_user(arg, sizeof(*sa));
4086 if (IS_ERR(sa))
4087 return PTR_ERR(sa);
4088
4089 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4090 ret = mnt_want_write_file(file);
4091 if (ret)
4092 goto out;
4093 }
4094
4095 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4096 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4097 0);
4098
4099 if (copy_to_user(arg, sa, sizeof(*sa)))
4100 ret = -EFAULT;
4101
4102 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4103 mnt_drop_write_file(file);
4104 out:
4105 kfree(sa);
4106 return ret;
4107 }
4108
4109 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4110 {
4111 if (!capable(CAP_SYS_ADMIN))
4112 return -EPERM;
4113
4114 return btrfs_scrub_cancel(root->fs_info);
4115 }
4116
4117 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4118 void __user *arg)
4119 {
4120 struct btrfs_ioctl_scrub_args *sa;
4121 int ret;
4122
4123 if (!capable(CAP_SYS_ADMIN))
4124 return -EPERM;
4125
4126 sa = memdup_user(arg, sizeof(*sa));
4127 if (IS_ERR(sa))
4128 return PTR_ERR(sa);
4129
4130 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4131
4132 if (copy_to_user(arg, sa, sizeof(*sa)))
4133 ret = -EFAULT;
4134
4135 kfree(sa);
4136 return ret;
4137 }
4138
4139 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4140 void __user *arg)
4141 {
4142 struct btrfs_ioctl_get_dev_stats *sa;
4143 int ret;
4144
4145 sa = memdup_user(arg, sizeof(*sa));
4146 if (IS_ERR(sa))
4147 return PTR_ERR(sa);
4148
4149 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4150 kfree(sa);
4151 return -EPERM;
4152 }
4153
4154 ret = btrfs_get_dev_stats(root, sa);
4155
4156 if (copy_to_user(arg, sa, sizeof(*sa)))
4157 ret = -EFAULT;
4158
4159 kfree(sa);
4160 return ret;
4161 }
4162
4163 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4164 {
4165 struct btrfs_ioctl_dev_replace_args *p;
4166 int ret;
4167
4168 if (!capable(CAP_SYS_ADMIN))
4169 return -EPERM;
4170
4171 p = memdup_user(arg, sizeof(*p));
4172 if (IS_ERR(p))
4173 return PTR_ERR(p);
4174
4175 switch (p->cmd) {
4176 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4177 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4178 ret = -EROFS;
4179 goto out;
4180 }
4181 if (atomic_xchg(
4182 &root->fs_info->mutually_exclusive_operation_running,
4183 1)) {
4184 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4185 } else {
4186 ret = btrfs_dev_replace_start(root, p);
4187 atomic_set(
4188 &root->fs_info->mutually_exclusive_operation_running,
4189 0);
4190 }
4191 break;
4192 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4193 btrfs_dev_replace_status(root->fs_info, p);
4194 ret = 0;
4195 break;
4196 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4197 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4198 break;
4199 default:
4200 ret = -EINVAL;
4201 break;
4202 }
4203
4204 if (copy_to_user(arg, p, sizeof(*p)))
4205 ret = -EFAULT;
4206 out:
4207 kfree(p);
4208 return ret;
4209 }
4210
4211 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4212 {
4213 int ret = 0;
4214 int i;
4215 u64 rel_ptr;
4216 int size;
4217 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4218 struct inode_fs_paths *ipath = NULL;
4219 struct btrfs_path *path;
4220
4221 if (!capable(CAP_DAC_READ_SEARCH))
4222 return -EPERM;
4223
4224 path = btrfs_alloc_path();
4225 if (!path) {
4226 ret = -ENOMEM;
4227 goto out;
4228 }
4229
4230 ipa = memdup_user(arg, sizeof(*ipa));
4231 if (IS_ERR(ipa)) {
4232 ret = PTR_ERR(ipa);
4233 ipa = NULL;
4234 goto out;
4235 }
4236
4237 size = min_t(u32, ipa->size, 4096);
4238 ipath = init_ipath(size, root, path);
4239 if (IS_ERR(ipath)) {
4240 ret = PTR_ERR(ipath);
4241 ipath = NULL;
4242 goto out;
4243 }
4244
4245 ret = paths_from_inode(ipa->inum, ipath);
4246 if (ret < 0)
4247 goto out;
4248
4249 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4250 rel_ptr = ipath->fspath->val[i] -
4251 (u64)(unsigned long)ipath->fspath->val;
4252 ipath->fspath->val[i] = rel_ptr;
4253 }
4254
4255 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4256 (void *)(unsigned long)ipath->fspath, size);
4257 if (ret) {
4258 ret = -EFAULT;
4259 goto out;
4260 }
4261
4262 out:
4263 btrfs_free_path(path);
4264 free_ipath(ipath);
4265 kfree(ipa);
4266
4267 return ret;
4268 }
4269
4270 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4271 {
4272 struct btrfs_data_container *inodes = ctx;
4273 const size_t c = 3 * sizeof(u64);
4274
4275 if (inodes->bytes_left >= c) {
4276 inodes->bytes_left -= c;
4277 inodes->val[inodes->elem_cnt] = inum;
4278 inodes->val[inodes->elem_cnt + 1] = offset;
4279 inodes->val[inodes->elem_cnt + 2] = root;
4280 inodes->elem_cnt += 3;
4281 } else {
4282 inodes->bytes_missing += c - inodes->bytes_left;
4283 inodes->bytes_left = 0;
4284 inodes->elem_missed += 3;
4285 }
4286
4287 return 0;
4288 }
4289
4290 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4291 void __user *arg)
4292 {
4293 int ret = 0;
4294 int size;
4295 struct btrfs_ioctl_logical_ino_args *loi;
4296 struct btrfs_data_container *inodes = NULL;
4297 struct btrfs_path *path = NULL;
4298
4299 if (!capable(CAP_SYS_ADMIN))
4300 return -EPERM;
4301
4302 loi = memdup_user(arg, sizeof(*loi));
4303 if (IS_ERR(loi)) {
4304 ret = PTR_ERR(loi);
4305 loi = NULL;
4306 goto out;
4307 }
4308
4309 path = btrfs_alloc_path();
4310 if (!path) {
4311 ret = -ENOMEM;
4312 goto out;
4313 }
4314
4315 size = min_t(u32, loi->size, 64 * 1024);
4316 inodes = init_data_container(size);
4317 if (IS_ERR(inodes)) {
4318 ret = PTR_ERR(inodes);
4319 inodes = NULL;
4320 goto out;
4321 }
4322
4323 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4324 build_ino_list, inodes);
4325 if (ret == -EINVAL)
4326 ret = -ENOENT;
4327 if (ret < 0)
4328 goto out;
4329
4330 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4331 (void *)(unsigned long)inodes, size);
4332 if (ret)
4333 ret = -EFAULT;
4334
4335 out:
4336 btrfs_free_path(path);
4337 vfree(inodes);
4338 kfree(loi);
4339
4340 return ret;
4341 }
4342
4343 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4344 struct btrfs_ioctl_balance_args *bargs)
4345 {
4346 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4347
4348 bargs->flags = bctl->flags;
4349
4350 if (atomic_read(&fs_info->balance_running))
4351 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4352 if (atomic_read(&fs_info->balance_pause_req))
4353 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4354 if (atomic_read(&fs_info->balance_cancel_req))
4355 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4356
4357 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4358 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4359 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4360
4361 if (lock) {
4362 spin_lock(&fs_info->balance_lock);
4363 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4364 spin_unlock(&fs_info->balance_lock);
4365 } else {
4366 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4367 }
4368 }
4369
4370 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4371 {
4372 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4373 struct btrfs_fs_info *fs_info = root->fs_info;
4374 struct btrfs_ioctl_balance_args *bargs;
4375 struct btrfs_balance_control *bctl;
4376 bool need_unlock; /* for mut. excl. ops lock */
4377 int ret;
4378
4379 if (!capable(CAP_SYS_ADMIN))
4380 return -EPERM;
4381
4382 ret = mnt_want_write_file(file);
4383 if (ret)
4384 return ret;
4385
4386 again:
4387 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4388 mutex_lock(&fs_info->volume_mutex);
4389 mutex_lock(&fs_info->balance_mutex);
4390 need_unlock = true;
4391 goto locked;
4392 }
4393
4394 /*
4395 * mut. excl. ops lock is locked. Three possibilites:
4396 * (1) some other op is running
4397 * (2) balance is running
4398 * (3) balance is paused -- special case (think resume)
4399 */
4400 mutex_lock(&fs_info->balance_mutex);
4401 if (fs_info->balance_ctl) {
4402 /* this is either (2) or (3) */
4403 if (!atomic_read(&fs_info->balance_running)) {
4404 mutex_unlock(&fs_info->balance_mutex);
4405 if (!mutex_trylock(&fs_info->volume_mutex))
4406 goto again;
4407 mutex_lock(&fs_info->balance_mutex);
4408
4409 if (fs_info->balance_ctl &&
4410 !atomic_read(&fs_info->balance_running)) {
4411 /* this is (3) */
4412 need_unlock = false;
4413 goto locked;
4414 }
4415
4416 mutex_unlock(&fs_info->balance_mutex);
4417 mutex_unlock(&fs_info->volume_mutex);
4418 goto again;
4419 } else {
4420 /* this is (2) */
4421 mutex_unlock(&fs_info->balance_mutex);
4422 ret = -EINPROGRESS;
4423 goto out;
4424 }
4425 } else {
4426 /* this is (1) */
4427 mutex_unlock(&fs_info->balance_mutex);
4428 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4429 goto out;
4430 }
4431
4432 locked:
4433 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4434
4435 if (arg) {
4436 bargs = memdup_user(arg, sizeof(*bargs));
4437 if (IS_ERR(bargs)) {
4438 ret = PTR_ERR(bargs);
4439 goto out_unlock;
4440 }
4441
4442 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4443 if (!fs_info->balance_ctl) {
4444 ret = -ENOTCONN;
4445 goto out_bargs;
4446 }
4447
4448 bctl = fs_info->balance_ctl;
4449 spin_lock(&fs_info->balance_lock);
4450 bctl->flags |= BTRFS_BALANCE_RESUME;
4451 spin_unlock(&fs_info->balance_lock);
4452
4453 goto do_balance;
4454 }
4455 } else {
4456 bargs = NULL;
4457 }
4458
4459 if (fs_info->balance_ctl) {
4460 ret = -EINPROGRESS;
4461 goto out_bargs;
4462 }
4463
4464 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4465 if (!bctl) {
4466 ret = -ENOMEM;
4467 goto out_bargs;
4468 }
4469
4470 bctl->fs_info = fs_info;
4471 if (arg) {
4472 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4473 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4474 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4475
4476 bctl->flags = bargs->flags;
4477 } else {
4478 /* balance everything - no filters */
4479 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4480 }
4481
4482 do_balance:
4483 /*
4484 * Ownership of bctl and mutually_exclusive_operation_running
4485 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4486 * or, if restriper was paused all the way until unmount, in
4487 * free_fs_info. mutually_exclusive_operation_running is
4488 * cleared in __cancel_balance.
4489 */
4490 need_unlock = false;
4491
4492 ret = btrfs_balance(bctl, bargs);
4493
4494 if (arg) {
4495 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4496 ret = -EFAULT;
4497 }
4498
4499 out_bargs:
4500 kfree(bargs);
4501 out_unlock:
4502 mutex_unlock(&fs_info->balance_mutex);
4503 mutex_unlock(&fs_info->volume_mutex);
4504 if (need_unlock)
4505 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4506 out:
4507 mnt_drop_write_file(file);
4508 return ret;
4509 }
4510
4511 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4512 {
4513 if (!capable(CAP_SYS_ADMIN))
4514 return -EPERM;
4515
4516 switch (cmd) {
4517 case BTRFS_BALANCE_CTL_PAUSE:
4518 return btrfs_pause_balance(root->fs_info);
4519 case BTRFS_BALANCE_CTL_CANCEL:
4520 return btrfs_cancel_balance(root->fs_info);
4521 }
4522
4523 return -EINVAL;
4524 }
4525
4526 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4527 void __user *arg)
4528 {
4529 struct btrfs_fs_info *fs_info = root->fs_info;
4530 struct btrfs_ioctl_balance_args *bargs;
4531 int ret = 0;
4532
4533 if (!capable(CAP_SYS_ADMIN))
4534 return -EPERM;
4535
4536 mutex_lock(&fs_info->balance_mutex);
4537 if (!fs_info->balance_ctl) {
4538 ret = -ENOTCONN;
4539 goto out;
4540 }
4541
4542 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4543 if (!bargs) {
4544 ret = -ENOMEM;
4545 goto out;
4546 }
4547
4548 update_ioctl_balance_args(fs_info, 1, bargs);
4549
4550 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4551 ret = -EFAULT;
4552
4553 kfree(bargs);
4554 out:
4555 mutex_unlock(&fs_info->balance_mutex);
4556 return ret;
4557 }
4558
4559 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4560 {
4561 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4562 struct btrfs_ioctl_quota_ctl_args *sa;
4563 struct btrfs_trans_handle *trans = NULL;
4564 int ret;
4565 int err;
4566
4567 if (!capable(CAP_SYS_ADMIN))
4568 return -EPERM;
4569
4570 ret = mnt_want_write_file(file);
4571 if (ret)
4572 return ret;
4573
4574 sa = memdup_user(arg, sizeof(*sa));
4575 if (IS_ERR(sa)) {
4576 ret = PTR_ERR(sa);
4577 goto drop_write;
4578 }
4579
4580 down_write(&root->fs_info->subvol_sem);
4581 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4582 if (IS_ERR(trans)) {
4583 ret = PTR_ERR(trans);
4584 goto out;
4585 }
4586
4587 switch (sa->cmd) {
4588 case BTRFS_QUOTA_CTL_ENABLE:
4589 ret = btrfs_quota_enable(trans, root->fs_info);
4590 break;
4591 case BTRFS_QUOTA_CTL_DISABLE:
4592 ret = btrfs_quota_disable(trans, root->fs_info);
4593 break;
4594 default:
4595 ret = -EINVAL;
4596 break;
4597 }
4598
4599 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4600 if (err && !ret)
4601 ret = err;
4602 out:
4603 kfree(sa);
4604 up_write(&root->fs_info->subvol_sem);
4605 drop_write:
4606 mnt_drop_write_file(file);
4607 return ret;
4608 }
4609
4610 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4611 {
4612 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4613 struct btrfs_ioctl_qgroup_assign_args *sa;
4614 struct btrfs_trans_handle *trans;
4615 int ret;
4616 int err;
4617
4618 if (!capable(CAP_SYS_ADMIN))
4619 return -EPERM;
4620
4621 ret = mnt_want_write_file(file);
4622 if (ret)
4623 return ret;
4624
4625 sa = memdup_user(arg, sizeof(*sa));
4626 if (IS_ERR(sa)) {
4627 ret = PTR_ERR(sa);
4628 goto drop_write;
4629 }
4630
4631 trans = btrfs_join_transaction(root);
4632 if (IS_ERR(trans)) {
4633 ret = PTR_ERR(trans);
4634 goto out;
4635 }
4636
4637 /* FIXME: check if the IDs really exist */
4638 if (sa->assign) {
4639 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4640 sa->src, sa->dst);
4641 } else {
4642 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4643 sa->src, sa->dst);
4644 }
4645
4646 err = btrfs_end_transaction(trans, root);
4647 if (err && !ret)
4648 ret = err;
4649
4650 out:
4651 kfree(sa);
4652 drop_write:
4653 mnt_drop_write_file(file);
4654 return ret;
4655 }
4656
4657 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4658 {
4659 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4660 struct btrfs_ioctl_qgroup_create_args *sa;
4661 struct btrfs_trans_handle *trans;
4662 int ret;
4663 int err;
4664
4665 if (!capable(CAP_SYS_ADMIN))
4666 return -EPERM;
4667
4668 ret = mnt_want_write_file(file);
4669 if (ret)
4670 return ret;
4671
4672 sa = memdup_user(arg, sizeof(*sa));
4673 if (IS_ERR(sa)) {
4674 ret = PTR_ERR(sa);
4675 goto drop_write;
4676 }
4677
4678 if (!sa->qgroupid) {
4679 ret = -EINVAL;
4680 goto out;
4681 }
4682
4683 trans = btrfs_join_transaction(root);
4684 if (IS_ERR(trans)) {
4685 ret = PTR_ERR(trans);
4686 goto out;
4687 }
4688
4689 /* FIXME: check if the IDs really exist */
4690 if (sa->create) {
4691 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4692 NULL);
4693 } else {
4694 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4695 }
4696
4697 err = btrfs_end_transaction(trans, root);
4698 if (err && !ret)
4699 ret = err;
4700
4701 out:
4702 kfree(sa);
4703 drop_write:
4704 mnt_drop_write_file(file);
4705 return ret;
4706 }
4707
4708 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4709 {
4710 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4711 struct btrfs_ioctl_qgroup_limit_args *sa;
4712 struct btrfs_trans_handle *trans;
4713 int ret;
4714 int err;
4715 u64 qgroupid;
4716
4717 if (!capable(CAP_SYS_ADMIN))
4718 return -EPERM;
4719
4720 ret = mnt_want_write_file(file);
4721 if (ret)
4722 return ret;
4723
4724 sa = memdup_user(arg, sizeof(*sa));
4725 if (IS_ERR(sa)) {
4726 ret = PTR_ERR(sa);
4727 goto drop_write;
4728 }
4729
4730 trans = btrfs_join_transaction(root);
4731 if (IS_ERR(trans)) {
4732 ret = PTR_ERR(trans);
4733 goto out;
4734 }
4735
4736 qgroupid = sa->qgroupid;
4737 if (!qgroupid) {
4738 /* take the current subvol as qgroup */
4739 qgroupid = root->root_key.objectid;
4740 }
4741
4742 /* FIXME: check if the IDs really exist */
4743 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4744
4745 err = btrfs_end_transaction(trans, root);
4746 if (err && !ret)
4747 ret = err;
4748
4749 out:
4750 kfree(sa);
4751 drop_write:
4752 mnt_drop_write_file(file);
4753 return ret;
4754 }
4755
4756 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4757 {
4758 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4759 struct btrfs_ioctl_quota_rescan_args *qsa;
4760 int ret;
4761
4762 if (!capable(CAP_SYS_ADMIN))
4763 return -EPERM;
4764
4765 ret = mnt_want_write_file(file);
4766 if (ret)
4767 return ret;
4768
4769 qsa = memdup_user(arg, sizeof(*qsa));
4770 if (IS_ERR(qsa)) {
4771 ret = PTR_ERR(qsa);
4772 goto drop_write;
4773 }
4774
4775 if (qsa->flags) {
4776 ret = -EINVAL;
4777 goto out;
4778 }
4779
4780 ret = btrfs_qgroup_rescan(root->fs_info);
4781
4782 out:
4783 kfree(qsa);
4784 drop_write:
4785 mnt_drop_write_file(file);
4786 return ret;
4787 }
4788
4789 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4790 {
4791 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4792 struct btrfs_ioctl_quota_rescan_args *qsa;
4793 int ret = 0;
4794
4795 if (!capable(CAP_SYS_ADMIN))
4796 return -EPERM;
4797
4798 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4799 if (!qsa)
4800 return -ENOMEM;
4801
4802 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4803 qsa->flags = 1;
4804 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4805 }
4806
4807 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4808 ret = -EFAULT;
4809
4810 kfree(qsa);
4811 return ret;
4812 }
4813
4814 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4815 {
4816 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4817
4818 if (!capable(CAP_SYS_ADMIN))
4819 return -EPERM;
4820
4821 return btrfs_qgroup_wait_for_completion(root->fs_info);
4822 }
4823
4824 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4825 struct btrfs_ioctl_received_subvol_args *sa)
4826 {
4827 struct inode *inode = file_inode(file);
4828 struct btrfs_root *root = BTRFS_I(inode)->root;
4829 struct btrfs_root_item *root_item = &root->root_item;
4830 struct btrfs_trans_handle *trans;
4831 struct timespec ct = CURRENT_TIME;
4832 int ret = 0;
4833 int received_uuid_changed;
4834
4835 if (!inode_owner_or_capable(inode))
4836 return -EPERM;
4837
4838 ret = mnt_want_write_file(file);
4839 if (ret < 0)
4840 return ret;
4841
4842 down_write(&root->fs_info->subvol_sem);
4843
4844 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4845 ret = -EINVAL;
4846 goto out;
4847 }
4848
4849 if (btrfs_root_readonly(root)) {
4850 ret = -EROFS;
4851 goto out;
4852 }
4853
4854 /*
4855 * 1 - root item
4856 * 2 - uuid items (received uuid + subvol uuid)
4857 */
4858 trans = btrfs_start_transaction(root, 3);
4859 if (IS_ERR(trans)) {
4860 ret = PTR_ERR(trans);
4861 trans = NULL;
4862 goto out;
4863 }
4864
4865 sa->rtransid = trans->transid;
4866 sa->rtime.sec = ct.tv_sec;
4867 sa->rtime.nsec = ct.tv_nsec;
4868
4869 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4870 BTRFS_UUID_SIZE);
4871 if (received_uuid_changed &&
4872 !btrfs_is_empty_uuid(root_item->received_uuid))
4873 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4874 root_item->received_uuid,
4875 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4876 root->root_key.objectid);
4877 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4878 btrfs_set_root_stransid(root_item, sa->stransid);
4879 btrfs_set_root_rtransid(root_item, sa->rtransid);
4880 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4881 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4882 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4883 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4884
4885 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4886 &root->root_key, &root->root_item);
4887 if (ret < 0) {
4888 btrfs_end_transaction(trans, root);
4889 goto out;
4890 }
4891 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4892 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4893 sa->uuid,
4894 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4895 root->root_key.objectid);
4896 if (ret < 0 && ret != -EEXIST) {
4897 btrfs_abort_transaction(trans, root, ret);
4898 goto out;
4899 }
4900 }
4901 ret = btrfs_commit_transaction(trans, root);
4902 if (ret < 0) {
4903 btrfs_abort_transaction(trans, root, ret);
4904 goto out;
4905 }
4906
4907 out:
4908 up_write(&root->fs_info->subvol_sem);
4909 mnt_drop_write_file(file);
4910 return ret;
4911 }
4912
4913 #ifdef CONFIG_64BIT
4914 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4915 void __user *arg)
4916 {
4917 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4918 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4919 int ret = 0;
4920
4921 args32 = memdup_user(arg, sizeof(*args32));
4922 if (IS_ERR(args32)) {
4923 ret = PTR_ERR(args32);
4924 args32 = NULL;
4925 goto out;
4926 }
4927
4928 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4929 if (!args64) {
4930 ret = -ENOMEM;
4931 goto out;
4932 }
4933
4934 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4935 args64->stransid = args32->stransid;
4936 args64->rtransid = args32->rtransid;
4937 args64->stime.sec = args32->stime.sec;
4938 args64->stime.nsec = args32->stime.nsec;
4939 args64->rtime.sec = args32->rtime.sec;
4940 args64->rtime.nsec = args32->rtime.nsec;
4941 args64->flags = args32->flags;
4942
4943 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4944 if (ret)
4945 goto out;
4946
4947 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4948 args32->stransid = args64->stransid;
4949 args32->rtransid = args64->rtransid;
4950 args32->stime.sec = args64->stime.sec;
4951 args32->stime.nsec = args64->stime.nsec;
4952 args32->rtime.sec = args64->rtime.sec;
4953 args32->rtime.nsec = args64->rtime.nsec;
4954 args32->flags = args64->flags;
4955
4956 ret = copy_to_user(arg, args32, sizeof(*args32));
4957 if (ret)
4958 ret = -EFAULT;
4959
4960 out:
4961 kfree(args32);
4962 kfree(args64);
4963 return ret;
4964 }
4965 #endif
4966
4967 static long btrfs_ioctl_set_received_subvol(struct file *file,
4968 void __user *arg)
4969 {
4970 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4971 int ret = 0;
4972
4973 sa = memdup_user(arg, sizeof(*sa));
4974 if (IS_ERR(sa)) {
4975 ret = PTR_ERR(sa);
4976 sa = NULL;
4977 goto out;
4978 }
4979
4980 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4981
4982 if (ret)
4983 goto out;
4984
4985 ret = copy_to_user(arg, sa, sizeof(*sa));
4986 if (ret)
4987 ret = -EFAULT;
4988
4989 out:
4990 kfree(sa);
4991 return ret;
4992 }
4993
4994 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4995 {
4996 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4997 size_t len;
4998 int ret;
4999 char label[BTRFS_LABEL_SIZE];
5000
5001 spin_lock(&root->fs_info->super_lock);
5002 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5003 spin_unlock(&root->fs_info->super_lock);
5004
5005 len = strnlen(label, BTRFS_LABEL_SIZE);
5006
5007 if (len == BTRFS_LABEL_SIZE) {
5008 btrfs_warn(root->fs_info,
5009 "label is too long, return the first %zu bytes", --len);
5010 }
5011
5012 ret = copy_to_user(arg, label, len);
5013
5014 return ret ? -EFAULT : 0;
5015 }
5016
5017 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5018 {
5019 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5020 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5021 struct btrfs_trans_handle *trans;
5022 char label[BTRFS_LABEL_SIZE];
5023 int ret;
5024
5025 if (!capable(CAP_SYS_ADMIN))
5026 return -EPERM;
5027
5028 if (copy_from_user(label, arg, sizeof(label)))
5029 return -EFAULT;
5030
5031 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5032 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5033 BTRFS_LABEL_SIZE - 1);
5034 return -EINVAL;
5035 }
5036
5037 ret = mnt_want_write_file(file);
5038 if (ret)
5039 return ret;
5040
5041 trans = btrfs_start_transaction(root, 0);
5042 if (IS_ERR(trans)) {
5043 ret = PTR_ERR(trans);
5044 goto out_unlock;
5045 }
5046
5047 spin_lock(&root->fs_info->super_lock);
5048 strcpy(super_block->label, label);
5049 spin_unlock(&root->fs_info->super_lock);
5050 ret = btrfs_commit_transaction(trans, root);
5051
5052 out_unlock:
5053 mnt_drop_write_file(file);
5054 return ret;
5055 }
5056
5057 #define INIT_FEATURE_FLAGS(suffix) \
5058 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5059 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5060 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5061
5062 static int btrfs_ioctl_get_supported_features(struct file *file,
5063 void __user *arg)
5064 {
5065 static struct btrfs_ioctl_feature_flags features[3] = {
5066 INIT_FEATURE_FLAGS(SUPP),
5067 INIT_FEATURE_FLAGS(SAFE_SET),
5068 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5069 };
5070
5071 if (copy_to_user(arg, &features, sizeof(features)))
5072 return -EFAULT;
5073
5074 return 0;
5075 }
5076
5077 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5078 {
5079 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5080 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5081 struct btrfs_ioctl_feature_flags features;
5082
5083 features.compat_flags = btrfs_super_compat_flags(super_block);
5084 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5085 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5086
5087 if (copy_to_user(arg, &features, sizeof(features)))
5088 return -EFAULT;
5089
5090 return 0;
5091 }
5092
5093 static int check_feature_bits(struct btrfs_root *root,
5094 enum btrfs_feature_set set,
5095 u64 change_mask, u64 flags, u64 supported_flags,
5096 u64 safe_set, u64 safe_clear)
5097 {
5098 const char *type = btrfs_feature_set_names[set];
5099 char *names;
5100 u64 disallowed, unsupported;
5101 u64 set_mask = flags & change_mask;
5102 u64 clear_mask = ~flags & change_mask;
5103
5104 unsupported = set_mask & ~supported_flags;
5105 if (unsupported) {
5106 names = btrfs_printable_features(set, unsupported);
5107 if (names) {
5108 btrfs_warn(root->fs_info,
5109 "this kernel does not support the %s feature bit%s",
5110 names, strchr(names, ',') ? "s" : "");
5111 kfree(names);
5112 } else
5113 btrfs_warn(root->fs_info,
5114 "this kernel does not support %s bits 0x%llx",
5115 type, unsupported);
5116 return -EOPNOTSUPP;
5117 }
5118
5119 disallowed = set_mask & ~safe_set;
5120 if (disallowed) {
5121 names = btrfs_printable_features(set, disallowed);
5122 if (names) {
5123 btrfs_warn(root->fs_info,
5124 "can't set the %s feature bit%s while mounted",
5125 names, strchr(names, ',') ? "s" : "");
5126 kfree(names);
5127 } else
5128 btrfs_warn(root->fs_info,
5129 "can't set %s bits 0x%llx while mounted",
5130 type, disallowed);
5131 return -EPERM;
5132 }
5133
5134 disallowed = clear_mask & ~safe_clear;
5135 if (disallowed) {
5136 names = btrfs_printable_features(set, disallowed);
5137 if (names) {
5138 btrfs_warn(root->fs_info,
5139 "can't clear the %s feature bit%s while mounted",
5140 names, strchr(names, ',') ? "s" : "");
5141 kfree(names);
5142 } else
5143 btrfs_warn(root->fs_info,
5144 "can't clear %s bits 0x%llx while mounted",
5145 type, disallowed);
5146 return -EPERM;
5147 }
5148
5149 return 0;
5150 }
5151
5152 #define check_feature(root, change_mask, flags, mask_base) \
5153 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5154 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5155 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5156 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5157
5158 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5159 {
5160 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5161 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5162 struct btrfs_ioctl_feature_flags flags[2];
5163 struct btrfs_trans_handle *trans;
5164 u64 newflags;
5165 int ret;
5166
5167 if (!capable(CAP_SYS_ADMIN))
5168 return -EPERM;
5169
5170 if (copy_from_user(flags, arg, sizeof(flags)))
5171 return -EFAULT;
5172
5173 /* Nothing to do */
5174 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5175 !flags[0].incompat_flags)
5176 return 0;
5177
5178 ret = check_feature(root, flags[0].compat_flags,
5179 flags[1].compat_flags, COMPAT);
5180 if (ret)
5181 return ret;
5182
5183 ret = check_feature(root, flags[0].compat_ro_flags,
5184 flags[1].compat_ro_flags, COMPAT_RO);
5185 if (ret)
5186 return ret;
5187
5188 ret = check_feature(root, flags[0].incompat_flags,
5189 flags[1].incompat_flags, INCOMPAT);
5190 if (ret)
5191 return ret;
5192
5193 trans = btrfs_start_transaction(root, 0);
5194 if (IS_ERR(trans))
5195 return PTR_ERR(trans);
5196
5197 spin_lock(&root->fs_info->super_lock);
5198 newflags = btrfs_super_compat_flags(super_block);
5199 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5200 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5201 btrfs_set_super_compat_flags(super_block, newflags);
5202
5203 newflags = btrfs_super_compat_ro_flags(super_block);
5204 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5205 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5206 btrfs_set_super_compat_ro_flags(super_block, newflags);
5207
5208 newflags = btrfs_super_incompat_flags(super_block);
5209 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5210 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5211 btrfs_set_super_incompat_flags(super_block, newflags);
5212 spin_unlock(&root->fs_info->super_lock);
5213
5214 return btrfs_commit_transaction(trans, root);
5215 }
5216
5217 long btrfs_ioctl(struct file *file, unsigned int
5218 cmd, unsigned long arg)
5219 {
5220 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5221 void __user *argp = (void __user *)arg;
5222
5223 switch (cmd) {
5224 case FS_IOC_GETFLAGS:
5225 return btrfs_ioctl_getflags(file, argp);
5226 case FS_IOC_SETFLAGS:
5227 return btrfs_ioctl_setflags(file, argp);
5228 case FS_IOC_GETVERSION:
5229 return btrfs_ioctl_getversion(file, argp);
5230 case FITRIM:
5231 return btrfs_ioctl_fitrim(file, argp);
5232 case BTRFS_IOC_SNAP_CREATE:
5233 return btrfs_ioctl_snap_create(file, argp, 0);
5234 case BTRFS_IOC_SNAP_CREATE_V2:
5235 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5236 case BTRFS_IOC_SUBVOL_CREATE:
5237 return btrfs_ioctl_snap_create(file, argp, 1);
5238 case BTRFS_IOC_SUBVOL_CREATE_V2:
5239 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5240 case BTRFS_IOC_SNAP_DESTROY:
5241 return btrfs_ioctl_snap_destroy(file, argp);
5242 case BTRFS_IOC_SUBVOL_GETFLAGS:
5243 return btrfs_ioctl_subvol_getflags(file, argp);
5244 case BTRFS_IOC_SUBVOL_SETFLAGS:
5245 return btrfs_ioctl_subvol_setflags(file, argp);
5246 case BTRFS_IOC_DEFAULT_SUBVOL:
5247 return btrfs_ioctl_default_subvol(file, argp);
5248 case BTRFS_IOC_DEFRAG:
5249 return btrfs_ioctl_defrag(file, NULL);
5250 case BTRFS_IOC_DEFRAG_RANGE:
5251 return btrfs_ioctl_defrag(file, argp);
5252 case BTRFS_IOC_RESIZE:
5253 return btrfs_ioctl_resize(file, argp);
5254 case BTRFS_IOC_ADD_DEV:
5255 return btrfs_ioctl_add_dev(root, argp);
5256 case BTRFS_IOC_RM_DEV:
5257 return btrfs_ioctl_rm_dev(file, argp);
5258 case BTRFS_IOC_FS_INFO:
5259 return btrfs_ioctl_fs_info(root, argp);
5260 case BTRFS_IOC_DEV_INFO:
5261 return btrfs_ioctl_dev_info(root, argp);
5262 case BTRFS_IOC_BALANCE:
5263 return btrfs_ioctl_balance(file, NULL);
5264 case BTRFS_IOC_CLONE:
5265 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5266 case BTRFS_IOC_CLONE_RANGE:
5267 return btrfs_ioctl_clone_range(file, argp);
5268 case BTRFS_IOC_TRANS_START:
5269 return btrfs_ioctl_trans_start(file);
5270 case BTRFS_IOC_TRANS_END:
5271 return btrfs_ioctl_trans_end(file);
5272 case BTRFS_IOC_TREE_SEARCH:
5273 return btrfs_ioctl_tree_search(file, argp);
5274 case BTRFS_IOC_TREE_SEARCH_V2:
5275 return btrfs_ioctl_tree_search_v2(file, argp);
5276 case BTRFS_IOC_INO_LOOKUP:
5277 return btrfs_ioctl_ino_lookup(file, argp);
5278 case BTRFS_IOC_INO_PATHS:
5279 return btrfs_ioctl_ino_to_path(root, argp);
5280 case BTRFS_IOC_LOGICAL_INO:
5281 return btrfs_ioctl_logical_to_ino(root, argp);
5282 case BTRFS_IOC_SPACE_INFO:
5283 return btrfs_ioctl_space_info(root, argp);
5284 case BTRFS_IOC_SYNC: {
5285 int ret;
5286
5287 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5288 if (ret)
5289 return ret;
5290 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
5291 return ret;
5292 }
5293 case BTRFS_IOC_START_SYNC:
5294 return btrfs_ioctl_start_sync(root, argp);
5295 case BTRFS_IOC_WAIT_SYNC:
5296 return btrfs_ioctl_wait_sync(root, argp);
5297 case BTRFS_IOC_SCRUB:
5298 return btrfs_ioctl_scrub(file, argp);
5299 case BTRFS_IOC_SCRUB_CANCEL:
5300 return btrfs_ioctl_scrub_cancel(root, argp);
5301 case BTRFS_IOC_SCRUB_PROGRESS:
5302 return btrfs_ioctl_scrub_progress(root, argp);
5303 case BTRFS_IOC_BALANCE_V2:
5304 return btrfs_ioctl_balance(file, argp);
5305 case BTRFS_IOC_BALANCE_CTL:
5306 return btrfs_ioctl_balance_ctl(root, arg);
5307 case BTRFS_IOC_BALANCE_PROGRESS:
5308 return btrfs_ioctl_balance_progress(root, argp);
5309 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5310 return btrfs_ioctl_set_received_subvol(file, argp);
5311 #ifdef CONFIG_64BIT
5312 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5313 return btrfs_ioctl_set_received_subvol_32(file, argp);
5314 #endif
5315 case BTRFS_IOC_SEND:
5316 return btrfs_ioctl_send(file, argp);
5317 case BTRFS_IOC_GET_DEV_STATS:
5318 return btrfs_ioctl_get_dev_stats(root, argp);
5319 case BTRFS_IOC_QUOTA_CTL:
5320 return btrfs_ioctl_quota_ctl(file, argp);
5321 case BTRFS_IOC_QGROUP_ASSIGN:
5322 return btrfs_ioctl_qgroup_assign(file, argp);
5323 case BTRFS_IOC_QGROUP_CREATE:
5324 return btrfs_ioctl_qgroup_create(file, argp);
5325 case BTRFS_IOC_QGROUP_LIMIT:
5326 return btrfs_ioctl_qgroup_limit(file, argp);
5327 case BTRFS_IOC_QUOTA_RESCAN:
5328 return btrfs_ioctl_quota_rescan(file, argp);
5329 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5330 return btrfs_ioctl_quota_rescan_status(file, argp);
5331 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5332 return btrfs_ioctl_quota_rescan_wait(file, argp);
5333 case BTRFS_IOC_DEV_REPLACE:
5334 return btrfs_ioctl_dev_replace(root, argp);
5335 case BTRFS_IOC_GET_FSLABEL:
5336 return btrfs_ioctl_get_fslabel(file, argp);
5337 case BTRFS_IOC_SET_FSLABEL:
5338 return btrfs_ioctl_set_fslabel(file, argp);
5339 case BTRFS_IOC_FILE_EXTENT_SAME:
5340 return btrfs_ioctl_file_extent_same(file, argp);
5341 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5342 return btrfs_ioctl_get_supported_features(file, argp);
5343 case BTRFS_IOC_GET_FEATURES:
5344 return btrfs_ioctl_get_features(file, argp);
5345 case BTRFS_IOC_SET_FEATURES:
5346 return btrfs_ioctl_set_features(file, argp);
5347 }
5348
5349 return -ENOTTY;
5350 }
Linux with added FPC-III support