projects / linux-2.6-block.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
Merge branches 'acpi-resources', 'acpi-battery', 'acpi-doc' and 'acpi-pnp'
[linux-2.6-block.git] / fs / btrfs / ioctl.c
... / ...
CommitLineData
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 */
69struct btrfs_ioctl_timespec_32 {
70 __u64 sec;
71 __u32 nsec;
72} __attribute__ ((__packed__));
73
74struct 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
89static 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. */
93static 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 */
106static 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 */
136void 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 */
162void 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
188static 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
198static 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
213static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
214{
215 struct inode *inode = file_inode(file);
216 struct btrfs_inode *ip = BTRFS_I(inode);
217 struct btrfs_root *root = ip->root;
218 struct btrfs_trans_handle *trans;
219 unsigned int flags, oldflags;
220 int ret;
221 u64 ip_oldflags;
222 unsigned int i_oldflags;
223 umode_t mode;
224
225 if (!inode_owner_or_capable(inode))
226 return -EPERM;
227
228 if (btrfs_root_readonly(root))
229 return -EROFS;
230
231 if (copy_from_user(&flags, arg, sizeof(flags)))
232 return -EFAULT;
233
234 ret = check_flags(flags);
235 if (ret)
236 return ret;
237
238 ret = mnt_want_write_file(file);
239 if (ret)
240 return ret;
241
242 mutex_lock(&inode->i_mutex);
243
244 ip_oldflags = ip->flags;
245 i_oldflags = inode->i_flags;
246 mode = inode->i_mode;
247
248 flags = btrfs_mask_flags(inode->i_mode, flags);
249 oldflags = btrfs_flags_to_ioctl(ip->flags);
250 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
251 if (!capable(CAP_LINUX_IMMUTABLE)) {
252 ret = -EPERM;
253 goto out_unlock;
254 }
255 }
256
257 if (flags & FS_SYNC_FL)
258 ip->flags |= BTRFS_INODE_SYNC;
259 else
260 ip->flags &= ~BTRFS_INODE_SYNC;
261 if (flags & FS_IMMUTABLE_FL)
262 ip->flags |= BTRFS_INODE_IMMUTABLE;
263 else
264 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
265 if (flags & FS_APPEND_FL)
266 ip->flags |= BTRFS_INODE_APPEND;
267 else
268 ip->flags &= ~BTRFS_INODE_APPEND;
269 if (flags & FS_NODUMP_FL)
270 ip->flags |= BTRFS_INODE_NODUMP;
271 else
272 ip->flags &= ~BTRFS_INODE_NODUMP;
273 if (flags & FS_NOATIME_FL)
274 ip->flags |= BTRFS_INODE_NOATIME;
275 else
276 ip->flags &= ~BTRFS_INODE_NOATIME;
277 if (flags & FS_DIRSYNC_FL)
278 ip->flags |= BTRFS_INODE_DIRSYNC;
279 else
280 ip->flags &= ~BTRFS_INODE_DIRSYNC;
281 if (flags & FS_NOCOW_FL) {
282 if (S_ISREG(mode)) {
283 /*
284 * It's safe to turn csums off here, no extents exist.
285 * Otherwise we want the flag to reflect the real COW
286 * status of the file and will not set it.
287 */
288 if (inode->i_size == 0)
289 ip->flags |= BTRFS_INODE_NODATACOW
290 | BTRFS_INODE_NODATASUM;
291 } else {
292 ip->flags |= BTRFS_INODE_NODATACOW;
293 }
294 } else {
295 /*
296 * Revert back under same assuptions as above
297 */
298 if (S_ISREG(mode)) {
299 if (inode->i_size == 0)
300 ip->flags &= ~(BTRFS_INODE_NODATACOW
301 | BTRFS_INODE_NODATASUM);
302 } else {
303 ip->flags &= ~BTRFS_INODE_NODATACOW;
304 }
305 }
306
307 /*
308 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
309 * flag may be changed automatically if compression code won't make
310 * things smaller.
311 */
312 if (flags & FS_NOCOMP_FL) {
313 ip->flags &= ~BTRFS_INODE_COMPRESS;
314 ip->flags |= BTRFS_INODE_NOCOMPRESS;
315
316 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
317 if (ret && ret != -ENODATA)
318 goto out_drop;
319 } else if (flags & FS_COMPR_FL) {
320 const char *comp;
321
322 ip->flags |= BTRFS_INODE_COMPRESS;
323 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
324
325 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
326 comp = "lzo";
327 else
328 comp = "zlib";
329 ret = btrfs_set_prop(inode, "btrfs.compression",
330 comp, strlen(comp), 0);
331 if (ret)
332 goto out_drop;
333
334 } else {
335 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
336 if (ret && ret != -ENODATA)
337 goto out_drop;
338 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
339 }
340
341 trans = btrfs_start_transaction(root, 1);
342 if (IS_ERR(trans)) {
343 ret = PTR_ERR(trans);
344 goto out_drop;
345 }
346
347 btrfs_update_iflags(inode);
348 inode_inc_iversion(inode);
349 inode->i_ctime = CURRENT_TIME;
350 ret = btrfs_update_inode(trans, root, inode);
351
352 btrfs_end_transaction(trans, root);
353 out_drop:
354 if (ret) {
355 ip->flags = ip_oldflags;
356 inode->i_flags = i_oldflags;
357 }
358
359 out_unlock:
360 mutex_unlock(&inode->i_mutex);
361 mnt_drop_write_file(file);
362 return ret;
363}
364
365static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
366{
367 struct inode *inode = file_inode(file);
368
369 return put_user(inode->i_generation, arg);
370}
371
372static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
373{
374 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
375 struct btrfs_device *device;
376 struct request_queue *q;
377 struct fstrim_range range;
378 u64 minlen = ULLONG_MAX;
379 u64 num_devices = 0;
380 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
381 int ret;
382
383 if (!capable(CAP_SYS_ADMIN))
384 return -EPERM;
385
386 rcu_read_lock();
387 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
388 dev_list) {
389 if (!device->bdev)
390 continue;
391 q = bdev_get_queue(device->bdev);
392 if (blk_queue_discard(q)) {
393 num_devices++;
394 minlen = min((u64)q->limits.discard_granularity,
395 minlen);
396 }
397 }
398 rcu_read_unlock();
399
400 if (!num_devices)
401 return -EOPNOTSUPP;
402 if (copy_from_user(&range, arg, sizeof(range)))
403 return -EFAULT;
404 if (range.start > total_bytes ||
405 range.len < fs_info->sb->s_blocksize)
406 return -EINVAL;
407
408 range.len = min(range.len, total_bytes - range.start);
409 range.minlen = max(range.minlen, minlen);
410 ret = btrfs_trim_fs(fs_info->tree_root, &range);
411 if (ret < 0)
412 return ret;
413
414 if (copy_to_user(arg, &range, sizeof(range)))
415 return -EFAULT;
416
417 return 0;
418}
419
420int btrfs_is_empty_uuid(u8 *uuid)
421{
422 int i;
423
424 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
425 if (uuid[i])
426 return 0;
427 }
428 return 1;
429}
430
431static noinline int create_subvol(struct inode *dir,
432 struct dentry *dentry,
433 char *name, int namelen,
434 u64 *async_transid,
435 struct btrfs_qgroup_inherit *inherit)
436{
437 struct btrfs_trans_handle *trans;
438 struct btrfs_key key;
439 struct btrfs_root_item root_item;
440 struct btrfs_inode_item *inode_item;
441 struct extent_buffer *leaf;
442 struct btrfs_root *root = BTRFS_I(dir)->root;
443 struct btrfs_root *new_root;
444 struct btrfs_block_rsv block_rsv;
445 struct timespec cur_time = CURRENT_TIME;
446 struct inode *inode;
447 int ret;
448 int err;
449 u64 objectid;
450 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
451 u64 index = 0;
452 u64 qgroup_reserved;
453 uuid_le new_uuid;
454
455 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
456 if (ret)
457 return ret;
458
459 /*
460 * Don't create subvolume whose level is not zero. Or qgroup will be
461 * screwed up since it assume subvolme qgroup's level to be 0.
462 */
463 if (btrfs_qgroup_level(objectid))
464 return -ENOSPC;
465
466 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
467 /*
468 * The same as the snapshot creation, please see the comment
469 * of create_snapshot().
470 */
471 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
472 8, &qgroup_reserved, false);
473 if (ret)
474 return ret;
475
476 trans = btrfs_start_transaction(root, 0);
477 if (IS_ERR(trans)) {
478 ret = PTR_ERR(trans);
479 btrfs_subvolume_release_metadata(root, &block_rsv,
480 qgroup_reserved);
481 return ret;
482 }
483 trans->block_rsv = &block_rsv;
484 trans->bytes_reserved = block_rsv.size;
485
486 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
487 if (ret)
488 goto fail;
489
490 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
491 if (IS_ERR(leaf)) {
492 ret = PTR_ERR(leaf);
493 goto fail;
494 }
495
496 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
497 btrfs_set_header_bytenr(leaf, leaf->start);
498 btrfs_set_header_generation(leaf, trans->transid);
499 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
500 btrfs_set_header_owner(leaf, objectid);
501
502 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
503 BTRFS_FSID_SIZE);
504 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
505 btrfs_header_chunk_tree_uuid(leaf),
506 BTRFS_UUID_SIZE);
507 btrfs_mark_buffer_dirty(leaf);
508
509 memset(&root_item, 0, sizeof(root_item));
510
511 inode_item = &root_item.inode;
512 btrfs_set_stack_inode_generation(inode_item, 1);
513 btrfs_set_stack_inode_size(inode_item, 3);
514 btrfs_set_stack_inode_nlink(inode_item, 1);
515 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
516 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
517
518 btrfs_set_root_flags(&root_item, 0);
519 btrfs_set_root_limit(&root_item, 0);
520 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
521
522 btrfs_set_root_bytenr(&root_item, leaf->start);
523 btrfs_set_root_generation(&root_item, trans->transid);
524 btrfs_set_root_level(&root_item, 0);
525 btrfs_set_root_refs(&root_item, 1);
526 btrfs_set_root_used(&root_item, leaf->len);
527 btrfs_set_root_last_snapshot(&root_item, 0);
528
529 btrfs_set_root_generation_v2(&root_item,
530 btrfs_root_generation(&root_item));
531 uuid_le_gen(&new_uuid);
532 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
533 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
534 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
535 root_item.ctime = root_item.otime;
536 btrfs_set_root_ctransid(&root_item, trans->transid);
537 btrfs_set_root_otransid(&root_item, trans->transid);
538
539 btrfs_tree_unlock(leaf);
540 free_extent_buffer(leaf);
541 leaf = NULL;
542
543 btrfs_set_root_dirid(&root_item, new_dirid);
544
545 key.objectid = objectid;
546 key.offset = 0;
547 key.type = BTRFS_ROOT_ITEM_KEY;
548 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
549 &root_item);
550 if (ret)
551 goto fail;
552
553 key.offset = (u64)-1;
554 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
555 if (IS_ERR(new_root)) {
556 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
557 ret = PTR_ERR(new_root);
558 goto fail;
559 }
560
561 btrfs_record_root_in_trans(trans, new_root);
562
563 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
564 if (ret) {
565 /* We potentially lose an unused inode item here */
566 btrfs_abort_transaction(trans, root, ret);
567 goto fail;
568 }
569
570 /*
571 * insert the directory item
572 */
573 ret = btrfs_set_inode_index(dir, &index);
574 if (ret) {
575 btrfs_abort_transaction(trans, root, ret);
576 goto fail;
577 }
578
579 ret = btrfs_insert_dir_item(trans, root,
580 name, namelen, dir, &key,
581 BTRFS_FT_DIR, index);
582 if (ret) {
583 btrfs_abort_transaction(trans, root, ret);
584 goto fail;
585 }
586
587 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
588 ret = btrfs_update_inode(trans, root, dir);
589 BUG_ON(ret);
590
591 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
592 objectid, root->root_key.objectid,
593 btrfs_ino(dir), index, name, namelen);
594 BUG_ON(ret);
595
596 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
597 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
598 objectid);
599 if (ret)
600 btrfs_abort_transaction(trans, root, ret);
601
602fail:
603 trans->block_rsv = NULL;
604 trans->bytes_reserved = 0;
605 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
606
607 if (async_transid) {
608 *async_transid = trans->transid;
609 err = btrfs_commit_transaction_async(trans, root, 1);
610 if (err)
611 err = btrfs_commit_transaction(trans, root);
612 } else {
613 err = btrfs_commit_transaction(trans, root);
614 }
615 if (err && !ret)
616 ret = err;
617
618 if (!ret) {
619 inode = btrfs_lookup_dentry(dir, dentry);
620 if (IS_ERR(inode))
621 return PTR_ERR(inode);
622 d_instantiate(dentry, inode);
623 }
624 return ret;
625}
626
627static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
628{
629 s64 writers;
630 DEFINE_WAIT(wait);
631
632 do {
633 prepare_to_wait(&root->subv_writers->wait, &wait,
634 TASK_UNINTERRUPTIBLE);
635
636 writers = percpu_counter_sum(&root->subv_writers->counter);
637 if (writers)
638 schedule();
639
640 finish_wait(&root->subv_writers->wait, &wait);
641 } while (writers);
642}
643
644static int create_snapshot(struct btrfs_root *root, struct inode *dir,
645 struct dentry *dentry, char *name, int namelen,
646 u64 *async_transid, bool readonly,
647 struct btrfs_qgroup_inherit *inherit)
648{
649 struct inode *inode;
650 struct btrfs_pending_snapshot *pending_snapshot;
651 struct btrfs_trans_handle *trans;
652 int ret;
653
654 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
655 return -EINVAL;
656
657 atomic_inc(&root->will_be_snapshoted);
658 smp_mb__after_atomic();
659 btrfs_wait_for_no_snapshoting_writes(root);
660
661 ret = btrfs_start_delalloc_inodes(root, 0);
662 if (ret)
663 goto out;
664
665 btrfs_wait_ordered_extents(root, -1);
666
667 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
668 if (!pending_snapshot) {
669 ret = -ENOMEM;
670 goto out;
671 }
672
673 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
674 BTRFS_BLOCK_RSV_TEMP);
675 /*
676 * 1 - parent dir inode
677 * 2 - dir entries
678 * 1 - root item
679 * 2 - root ref/backref
680 * 1 - root of snapshot
681 * 1 - UUID item
682 */
683 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
684 &pending_snapshot->block_rsv, 8,
685 &pending_snapshot->qgroup_reserved,
686 false);
687 if (ret)
688 goto free;
689
690 pending_snapshot->dentry = dentry;
691 pending_snapshot->root = root;
692 pending_snapshot->readonly = readonly;
693 pending_snapshot->dir = dir;
694 pending_snapshot->inherit = inherit;
695
696 trans = btrfs_start_transaction(root, 0);
697 if (IS_ERR(trans)) {
698 ret = PTR_ERR(trans);
699 goto fail;
700 }
701
702 spin_lock(&root->fs_info->trans_lock);
703 list_add(&pending_snapshot->list,
704 &trans->transaction->pending_snapshots);
705 spin_unlock(&root->fs_info->trans_lock);
706 if (async_transid) {
707 *async_transid = trans->transid;
708 ret = btrfs_commit_transaction_async(trans,
709 root->fs_info->extent_root, 1);
710 if (ret)
711 ret = btrfs_commit_transaction(trans, root);
712 } else {
713 ret = btrfs_commit_transaction(trans,
714 root->fs_info->extent_root);
715 }
716 if (ret)
717 goto fail;
718
719 ret = pending_snapshot->error;
720 if (ret)
721 goto fail;
722
723 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
724 if (ret)
725 goto fail;
726
727 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
728 if (IS_ERR(inode)) {
729 ret = PTR_ERR(inode);
730 goto fail;
731 }
732
733 d_instantiate(dentry, inode);
734 ret = 0;
735fail:
736 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
737 &pending_snapshot->block_rsv,
738 pending_snapshot->qgroup_reserved);
739free:
740 kfree(pending_snapshot);
741out:
742 if (atomic_dec_and_test(&root->will_be_snapshoted))
743 wake_up_atomic_t(&root->will_be_snapshoted);
744 return ret;
745}
746
747/* copy of may_delete in fs/namei.c()
748 * Check whether we can remove a link victim from directory dir, check
749 * whether the type of victim is right.
750 * 1. We can't do it if dir is read-only (done in permission())
751 * 2. We should have write and exec permissions on dir
752 * 3. We can't remove anything from append-only dir
753 * 4. We can't do anything with immutable dir (done in permission())
754 * 5. If the sticky bit on dir is set we should either
755 * a. be owner of dir, or
756 * b. be owner of victim, or
757 * c. have CAP_FOWNER capability
758 * 6. If the victim is append-only or immutable we can't do antyhing with
759 * links pointing to it.
760 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
761 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
762 * 9. We can't remove a root or mountpoint.
763 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
764 * nfs_async_unlink().
765 */
766
767static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
768{
769 int error;
770
771 if (d_really_is_negative(victim))
772 return -ENOENT;
773
774 BUG_ON(d_inode(victim->d_parent) != dir);
775 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
776
777 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
778 if (error)
779 return error;
780 if (IS_APPEND(dir))
781 return -EPERM;
782 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
783 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
784 return -EPERM;
785 if (isdir) {
786 if (!d_is_dir(victim))
787 return -ENOTDIR;
788 if (IS_ROOT(victim))
789 return -EBUSY;
790 } else if (d_is_dir(victim))
791 return -EISDIR;
792 if (IS_DEADDIR(dir))
793 return -ENOENT;
794 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
795 return -EBUSY;
796 return 0;
797}
798
799/* copy of may_create in fs/namei.c() */
800static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
801{
802 if (d_really_is_positive(child))
803 return -EEXIST;
804 if (IS_DEADDIR(dir))
805 return -ENOENT;
806 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
807}
808
809/*
810 * Create a new subvolume below @parent. This is largely modeled after
811 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
812 * inside this filesystem so it's quite a bit simpler.
813 */
814static noinline int btrfs_mksubvol(struct path *parent,
815 char *name, int namelen,
816 struct btrfs_root *snap_src,
817 u64 *async_transid, bool readonly,
818 struct btrfs_qgroup_inherit *inherit)
819{
820 struct inode *dir = d_inode(parent->dentry);
821 struct dentry *dentry;
822 int error;
823
824 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
825 if (error == -EINTR)
826 return error;
827
828 dentry = lookup_one_len(name, parent->dentry, namelen);
829 error = PTR_ERR(dentry);
830 if (IS_ERR(dentry))
831 goto out_unlock;
832
833 error = -EEXIST;
834 if (d_really_is_positive(dentry))
835 goto out_dput;
836
837 error = btrfs_may_create(dir, dentry);
838 if (error)
839 goto out_dput;
840
841 /*
842 * even if this name doesn't exist, we may get hash collisions.
843 * check for them now when we can safely fail
844 */
845 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
846 dir->i_ino, name,
847 namelen);
848 if (error)
849 goto out_dput;
850
851 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
852
853 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
854 goto out_up_read;
855
856 if (snap_src) {
857 error = create_snapshot(snap_src, dir, dentry, name, namelen,
858 async_transid, readonly, inherit);
859 } else {
860 error = create_subvol(dir, dentry, name, namelen,
861 async_transid, inherit);
862 }
863 if (!error)
864 fsnotify_mkdir(dir, dentry);
865out_up_read:
866 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
867out_dput:
868 dput(dentry);
869out_unlock:
870 mutex_unlock(&dir->i_mutex);
871 return error;
872}
873
874/*
875 * When we're defragging a range, we don't want to kick it off again
876 * if it is really just waiting for delalloc to send it down.
877 * If we find a nice big extent or delalloc range for the bytes in the
878 * file you want to defrag, we return 0 to let you know to skip this
879 * part of the file
880 */
881static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
882{
883 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
884 struct extent_map *em = NULL;
885 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
886 u64 end;
887
888 read_lock(&em_tree->lock);
889 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
890 read_unlock(&em_tree->lock);
891
892 if (em) {
893 end = extent_map_end(em);
894 free_extent_map(em);
895 if (end - offset > thresh)
896 return 0;
897 }
898 /* if we already have a nice delalloc here, just stop */
899 thresh /= 2;
900 end = count_range_bits(io_tree, &offset, offset + thresh,
901 thresh, EXTENT_DELALLOC, 1);
902 if (end >= thresh)
903 return 0;
904 return 1;
905}
906
907/*
908 * helper function to walk through a file and find extents
909 * newer than a specific transid, and smaller than thresh.
910 *
911 * This is used by the defragging code to find new and small
912 * extents
913 */
914static int find_new_extents(struct btrfs_root *root,
915 struct inode *inode, u64 newer_than,
916 u64 *off, u32 thresh)
917{
918 struct btrfs_path *path;
919 struct btrfs_key min_key;
920 struct extent_buffer *leaf;
921 struct btrfs_file_extent_item *extent;
922 int type;
923 int ret;
924 u64 ino = btrfs_ino(inode);
925
926 path = btrfs_alloc_path();
927 if (!path)
928 return -ENOMEM;
929
930 min_key.objectid = ino;
931 min_key.type = BTRFS_EXTENT_DATA_KEY;
932 min_key.offset = *off;
933
934 while (1) {
935 ret = btrfs_search_forward(root, &min_key, path, newer_than);
936 if (ret != 0)
937 goto none;
938process_slot:
939 if (min_key.objectid != ino)
940 goto none;
941 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
942 goto none;
943
944 leaf = path->nodes[0];
945 extent = btrfs_item_ptr(leaf, path->slots[0],
946 struct btrfs_file_extent_item);
947
948 type = btrfs_file_extent_type(leaf, extent);
949 if (type == BTRFS_FILE_EXTENT_REG &&
950 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
951 check_defrag_in_cache(inode, min_key.offset, thresh)) {
952 *off = min_key.offset;
953 btrfs_free_path(path);
954 return 0;
955 }
956
957 path->slots[0]++;
958 if (path->slots[0] < btrfs_header_nritems(leaf)) {
959 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
960 goto process_slot;
961 }
962
963 if (min_key.offset == (u64)-1)
964 goto none;
965
966 min_key.offset++;
967 btrfs_release_path(path);
968 }
969none:
970 btrfs_free_path(path);
971 return -ENOENT;
972}
973
974static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
975{
976 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
977 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
978 struct extent_map *em;
979 u64 len = PAGE_CACHE_SIZE;
980
981 /*
982 * hopefully we have this extent in the tree already, try without
983 * the full extent lock
984 */
985 read_lock(&em_tree->lock);
986 em = lookup_extent_mapping(em_tree, start, len);
987 read_unlock(&em_tree->lock);
988
989 if (!em) {
990 struct extent_state *cached = NULL;
991 u64 end = start + len - 1;
992
993 /* get the big lock and read metadata off disk */
994 lock_extent_bits(io_tree, start, end, 0, &cached);
995 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
996 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
997
998 if (IS_ERR(em))
999 return NULL;
1000 }
1001
1002 return em;
1003}
1004
1005static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1006{
1007 struct extent_map *next;
1008 bool ret = true;
1009
1010 /* this is the last extent */
1011 if (em->start + em->len >= i_size_read(inode))
1012 return false;
1013
1014 next = defrag_lookup_extent(inode, em->start + em->len);
1015 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1016 ret = false;
1017 else if ((em->block_start + em->block_len == next->block_start) &&
1018 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
1019 ret = false;
1020
1021 free_extent_map(next);
1022 return ret;
1023}
1024
1025static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1026 u64 *last_len, u64 *skip, u64 *defrag_end,
1027 int compress)
1028{
1029 struct extent_map *em;
1030 int ret = 1;
1031 bool next_mergeable = true;
1032
1033 /*
1034 * make sure that once we start defragging an extent, we keep on
1035 * defragging it
1036 */
1037 if (start < *defrag_end)
1038 return 1;
1039
1040 *skip = 0;
1041
1042 em = defrag_lookup_extent(inode, start);
1043 if (!em)
1044 return 0;
1045
1046 /* this will cover holes, and inline extents */
1047 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1048 ret = 0;
1049 goto out;
1050 }
1051
1052 next_mergeable = defrag_check_next_extent(inode, em);
1053 /*
1054 * we hit a real extent, if it is big or the next extent is not a
1055 * real extent, don't bother defragging it
1056 */
1057 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1058 (em->len >= thresh || !next_mergeable))
1059 ret = 0;
1060out:
1061 /*
1062 * last_len ends up being a counter of how many bytes we've defragged.
1063 * every time we choose not to defrag an extent, we reset *last_len
1064 * so that the next tiny extent will force a defrag.
1065 *
1066 * The end result of this is that tiny extents before a single big
1067 * extent will force at least part of that big extent to be defragged.
1068 */
1069 if (ret) {
1070 *defrag_end = extent_map_end(em);
1071 } else {
1072 *last_len = 0;
1073 *skip = extent_map_end(em);
1074 *defrag_end = 0;
1075 }
1076
1077 free_extent_map(em);
1078 return ret;
1079}
1080
1081/*
1082 * it doesn't do much good to defrag one or two pages
1083 * at a time. This pulls in a nice chunk of pages
1084 * to COW and defrag.
1085 *
1086 * It also makes sure the delalloc code has enough
1087 * dirty data to avoid making new small extents as part
1088 * of the defrag
1089 *
1090 * It's a good idea to start RA on this range
1091 * before calling this.
1092 */
1093static int cluster_pages_for_defrag(struct inode *inode,
1094 struct page **pages,
1095 unsigned long start_index,
1096 unsigned long num_pages)
1097{
1098 unsigned long file_end;
1099 u64 isize = i_size_read(inode);
1100 u64 page_start;
1101 u64 page_end;
1102 u64 page_cnt;
1103 int ret;
1104 int i;
1105 int i_done;
1106 struct btrfs_ordered_extent *ordered;
1107 struct extent_state *cached_state = NULL;
1108 struct extent_io_tree *tree;
1109 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1110
1111 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1112 if (!isize || start_index > file_end)
1113 return 0;
1114
1115 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1116
1117 ret = btrfs_delalloc_reserve_space(inode,
1118 page_cnt << PAGE_CACHE_SHIFT);
1119 if (ret)
1120 return ret;
1121 i_done = 0;
1122 tree = &BTRFS_I(inode)->io_tree;
1123
1124 /* step one, lock all the pages */
1125 for (i = 0; i < page_cnt; i++) {
1126 struct page *page;
1127again:
1128 page = find_or_create_page(inode->i_mapping,
1129 start_index + i, mask);
1130 if (!page)
1131 break;
1132
1133 page_start = page_offset(page);
1134 page_end = page_start + PAGE_CACHE_SIZE - 1;
1135 while (1) {
1136 lock_extent_bits(tree, page_start, page_end,
1137 0, &cached_state);
1138 ordered = btrfs_lookup_ordered_extent(inode,
1139 page_start);
1140 unlock_extent_cached(tree, page_start, page_end,
1141 &cached_state, GFP_NOFS);
1142 if (!ordered)
1143 break;
1144
1145 unlock_page(page);
1146 btrfs_start_ordered_extent(inode, ordered, 1);
1147 btrfs_put_ordered_extent(ordered);
1148 lock_page(page);
1149 /*
1150 * we unlocked the page above, so we need check if
1151 * it was released or not.
1152 */
1153 if (page->mapping != inode->i_mapping) {
1154 unlock_page(page);
1155 page_cache_release(page);
1156 goto again;
1157 }
1158 }
1159
1160 if (!PageUptodate(page)) {
1161 btrfs_readpage(NULL, page);
1162 lock_page(page);
1163 if (!PageUptodate(page)) {
1164 unlock_page(page);
1165 page_cache_release(page);
1166 ret = -EIO;
1167 break;
1168 }
1169 }
1170
1171 if (page->mapping != inode->i_mapping) {
1172 unlock_page(page);
1173 page_cache_release(page);
1174 goto again;
1175 }
1176
1177 pages[i] = page;
1178 i_done++;
1179 }
1180 if (!i_done || ret)
1181 goto out;
1182
1183 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1184 goto out;
1185
1186 /*
1187 * so now we have a nice long stream of locked
1188 * and up to date pages, lets wait on them
1189 */
1190 for (i = 0; i < i_done; i++)
1191 wait_on_page_writeback(pages[i]);
1192
1193 page_start = page_offset(pages[0]);
1194 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1195
1196 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1197 page_start, page_end - 1, 0, &cached_state);
1198 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1199 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1200 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1201 &cached_state, GFP_NOFS);
1202
1203 if (i_done != page_cnt) {
1204 spin_lock(&BTRFS_I(inode)->lock);
1205 BTRFS_I(inode)->outstanding_extents++;
1206 spin_unlock(&BTRFS_I(inode)->lock);
1207 btrfs_delalloc_release_space(inode,
1208 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1209 }
1210
1211
1212 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1213 &cached_state, GFP_NOFS);
1214
1215 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1216 page_start, page_end - 1, &cached_state,
1217 GFP_NOFS);
1218
1219 for (i = 0; i < i_done; i++) {
1220 clear_page_dirty_for_io(pages[i]);
1221 ClearPageChecked(pages[i]);
1222 set_page_extent_mapped(pages[i]);
1223 set_page_dirty(pages[i]);
1224 unlock_page(pages[i]);
1225 page_cache_release(pages[i]);
1226 }
1227 return i_done;
1228out:
1229 for (i = 0; i < i_done; i++) {
1230 unlock_page(pages[i]);
1231 page_cache_release(pages[i]);
1232 }
1233 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1234 return ret;
1235
1236}
1237
1238int btrfs_defrag_file(struct inode *inode, struct file *file,
1239 struct btrfs_ioctl_defrag_range_args *range,
1240 u64 newer_than, unsigned long max_to_defrag)
1241{
1242 struct btrfs_root *root = BTRFS_I(inode)->root;
1243 struct file_ra_state *ra = NULL;
1244 unsigned long last_index;
1245 u64 isize = i_size_read(inode);
1246 u64 last_len = 0;
1247 u64 skip = 0;
1248 u64 defrag_end = 0;
1249 u64 newer_off = range->start;
1250 unsigned long i;
1251 unsigned long ra_index = 0;
1252 int ret;
1253 int defrag_count = 0;
1254 int compress_type = BTRFS_COMPRESS_ZLIB;
1255 u32 extent_thresh = range->extent_thresh;
1256 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1257 unsigned long cluster = max_cluster;
1258 u64 new_align = ~((u64)128 * 1024 - 1);
1259 struct page **pages = NULL;
1260
1261 if (isize == 0)
1262 return 0;
1263
1264 if (range->start >= isize)
1265 return -EINVAL;
1266
1267 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1268 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1269 return -EINVAL;
1270 if (range->compress_type)
1271 compress_type = range->compress_type;
1272 }
1273
1274 if (extent_thresh == 0)
1275 extent_thresh = 256 * 1024;
1276
1277 /*
1278 * if we were not given a file, allocate a readahead
1279 * context
1280 */
1281 if (!file) {
1282 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1283 if (!ra)
1284 return -ENOMEM;
1285 file_ra_state_init(ra, inode->i_mapping);
1286 } else {
1287 ra = &file->f_ra;
1288 }
1289
1290 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1291 GFP_NOFS);
1292 if (!pages) {
1293 ret = -ENOMEM;
1294 goto out_ra;
1295 }
1296
1297 /* find the last page to defrag */
1298 if (range->start + range->len > range->start) {
1299 last_index = min_t(u64, isize - 1,
1300 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1301 } else {
1302 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1303 }
1304
1305 if (newer_than) {
1306 ret = find_new_extents(root, inode, newer_than,
1307 &newer_off, 64 * 1024);
1308 if (!ret) {
1309 range->start = newer_off;
1310 /*
1311 * we always align our defrag to help keep
1312 * the extents in the file evenly spaced
1313 */
1314 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1315 } else
1316 goto out_ra;
1317 } else {
1318 i = range->start >> PAGE_CACHE_SHIFT;
1319 }
1320 if (!max_to_defrag)
1321 max_to_defrag = last_index + 1;
1322
1323 /*
1324 * make writeback starts from i, so the defrag range can be
1325 * written sequentially.
1326 */
1327 if (i < inode->i_mapping->writeback_index)
1328 inode->i_mapping->writeback_index = i;
1329
1330 while (i <= last_index && defrag_count < max_to_defrag &&
1331 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
1332 /*
1333 * make sure we stop running if someone unmounts
1334 * the FS
1335 */
1336 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1337 break;
1338
1339 if (btrfs_defrag_cancelled(root->fs_info)) {
1340 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1341 ret = -EAGAIN;
1342 break;
1343 }
1344
1345 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1346 extent_thresh, &last_len, &skip,
1347 &defrag_end, range->flags &
1348 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1349 unsigned long next;
1350 /*
1351 * the should_defrag function tells us how much to skip
1352 * bump our counter by the suggested amount
1353 */
1354 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
1355 i = max(i + 1, next);
1356 continue;
1357 }
1358
1359 if (!newer_than) {
1360 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1361 PAGE_CACHE_SHIFT) - i;
1362 cluster = min(cluster, max_cluster);
1363 } else {
1364 cluster = max_cluster;
1365 }
1366
1367 if (i + cluster > ra_index) {
1368 ra_index = max(i, ra_index);
1369 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1370 cluster);
1371 ra_index += max_cluster;
1372 }
1373
1374 mutex_lock(&inode->i_mutex);
1375 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1376 BTRFS_I(inode)->force_compress = compress_type;
1377 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1378 if (ret < 0) {
1379 mutex_unlock(&inode->i_mutex);
1380 goto out_ra;
1381 }
1382
1383 defrag_count += ret;
1384 balance_dirty_pages_ratelimited(inode->i_mapping);
1385 mutex_unlock(&inode->i_mutex);
1386
1387 if (newer_than) {
1388 if (newer_off == (u64)-1)
1389 break;
1390
1391 if (ret > 0)
1392 i += ret;
1393
1394 newer_off = max(newer_off + 1,
1395 (u64)i << PAGE_CACHE_SHIFT);
1396
1397 ret = find_new_extents(root, inode,
1398 newer_than, &newer_off,
1399 64 * 1024);
1400 if (!ret) {
1401 range->start = newer_off;
1402 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1403 } else {
1404 break;
1405 }
1406 } else {
1407 if (ret > 0) {
1408 i += ret;
1409 last_len += ret << PAGE_CACHE_SHIFT;
1410 } else {
1411 i++;
1412 last_len = 0;
1413 }
1414 }
1415 }
1416
1417 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1418 filemap_flush(inode->i_mapping);
1419 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1420 &BTRFS_I(inode)->runtime_flags))
1421 filemap_flush(inode->i_mapping);
1422 }
1423
1424 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1425 /* the filemap_flush will queue IO into the worker threads, but
1426 * we have to make sure the IO is actually started and that
1427 * ordered extents get created before we return
1428 */
1429 atomic_inc(&root->fs_info->async_submit_draining);
1430 while (atomic_read(&root->fs_info->nr_async_submits) ||
1431 atomic_read(&root->fs_info->async_delalloc_pages)) {
1432 wait_event(root->fs_info->async_submit_wait,
1433 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1434 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1435 }
1436 atomic_dec(&root->fs_info->async_submit_draining);
1437 }
1438
1439 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1440 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1441 }
1442
1443 ret = defrag_count;
1444
1445out_ra:
1446 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1447 mutex_lock(&inode->i_mutex);
1448 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1449 mutex_unlock(&inode->i_mutex);
1450 }
1451 if (!file)
1452 kfree(ra);
1453 kfree(pages);
1454 return ret;
1455}
1456
1457static noinline int btrfs_ioctl_resize(struct file *file,
1458 void __user *arg)
1459{
1460 u64 new_size;
1461 u64 old_size;
1462 u64 devid = 1;
1463 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1464 struct btrfs_ioctl_vol_args *vol_args;
1465 struct btrfs_trans_handle *trans;
1466 struct btrfs_device *device = NULL;
1467 char *sizestr;
1468 char *retptr;
1469 char *devstr = NULL;
1470 int ret = 0;
1471 int mod = 0;
1472
1473 if (!capable(CAP_SYS_ADMIN))
1474 return -EPERM;
1475
1476 ret = mnt_want_write_file(file);
1477 if (ret)
1478 return ret;
1479
1480 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1481 1)) {
1482 mnt_drop_write_file(file);
1483 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1484 }
1485
1486 mutex_lock(&root->fs_info->volume_mutex);
1487 vol_args = memdup_user(arg, sizeof(*vol_args));
1488 if (IS_ERR(vol_args)) {
1489 ret = PTR_ERR(vol_args);
1490 goto out;
1491 }
1492
1493 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1494
1495 sizestr = vol_args->name;
1496 devstr = strchr(sizestr, ':');
1497 if (devstr) {
1498 sizestr = devstr + 1;
1499 *devstr = '\0';
1500 devstr = vol_args->name;
1501 ret = kstrtoull(devstr, 10, &devid);
1502 if (ret)
1503 goto out_free;
1504 if (!devid) {
1505 ret = -EINVAL;
1506 goto out_free;
1507 }
1508 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1509 }
1510
1511 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1512 if (!device) {
1513 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1514 devid);
1515 ret = -ENODEV;
1516 goto out_free;
1517 }
1518
1519 if (!device->writeable) {
1520 btrfs_info(root->fs_info,
1521 "resizer unable to apply on readonly device %llu",
1522 devid);
1523 ret = -EPERM;
1524 goto out_free;
1525 }
1526
1527 if (!strcmp(sizestr, "max"))
1528 new_size = device->bdev->bd_inode->i_size;
1529 else {
1530 if (sizestr[0] == '-') {
1531 mod = -1;
1532 sizestr++;
1533 } else if (sizestr[0] == '+') {
1534 mod = 1;
1535 sizestr++;
1536 }
1537 new_size = memparse(sizestr, &retptr);
1538 if (*retptr != '\0' || new_size == 0) {
1539 ret = -EINVAL;
1540 goto out_free;
1541 }
1542 }
1543
1544 if (device->is_tgtdev_for_dev_replace) {
1545 ret = -EPERM;
1546 goto out_free;
1547 }
1548
1549 old_size = btrfs_device_get_total_bytes(device);
1550
1551 if (mod < 0) {
1552 if (new_size > old_size) {
1553 ret = -EINVAL;
1554 goto out_free;
1555 }
1556 new_size = old_size - new_size;
1557 } else if (mod > 0) {
1558 if (new_size > ULLONG_MAX - old_size) {
1559 ret = -ERANGE;
1560 goto out_free;
1561 }
1562 new_size = old_size + new_size;
1563 }
1564
1565 if (new_size < 256 * 1024 * 1024) {
1566 ret = -EINVAL;
1567 goto out_free;
1568 }
1569 if (new_size > device->bdev->bd_inode->i_size) {
1570 ret = -EFBIG;
1571 goto out_free;
1572 }
1573
1574 new_size = div_u64(new_size, root->sectorsize);
1575 new_size *= root->sectorsize;
1576
1577 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1578 rcu_str_deref(device->name), new_size);
1579
1580 if (new_size > old_size) {
1581 trans = btrfs_start_transaction(root, 0);
1582 if (IS_ERR(trans)) {
1583 ret = PTR_ERR(trans);
1584 goto out_free;
1585 }
1586 ret = btrfs_grow_device(trans, device, new_size);
1587 btrfs_commit_transaction(trans, root);
1588 } else if (new_size < old_size) {
1589 ret = btrfs_shrink_device(device, new_size);
1590 } /* equal, nothing need to do */
1591
1592out_free:
1593 kfree(vol_args);
1594out:
1595 mutex_unlock(&root->fs_info->volume_mutex);
1596 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1597 mnt_drop_write_file(file);
1598 return ret;
1599}
1600
1601static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1602 char *name, unsigned long fd, int subvol,
1603 u64 *transid, bool readonly,
1604 struct btrfs_qgroup_inherit *inherit)
1605{
1606 int namelen;
1607 int ret = 0;
1608
1609 ret = mnt_want_write_file(file);
1610 if (ret)
1611 goto out;
1612
1613 namelen = strlen(name);
1614 if (strchr(name, '/')) {
1615 ret = -EINVAL;
1616 goto out_drop_write;
1617 }
1618
1619 if (name[0] == '.' &&
1620 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1621 ret = -EEXIST;
1622 goto out_drop_write;
1623 }
1624
1625 if (subvol) {
1626 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1627 NULL, transid, readonly, inherit);
1628 } else {
1629 struct fd src = fdget(fd);
1630 struct inode *src_inode;
1631 if (!src.file) {
1632 ret = -EINVAL;
1633 goto out_drop_write;
1634 }
1635
1636 src_inode = file_inode(src.file);
1637 if (src_inode->i_sb != file_inode(file)->i_sb) {
1638 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1639 "Snapshot src from another FS");
1640 ret = -EXDEV;
1641 } else if (!inode_owner_or_capable(src_inode)) {
1642 /*
1643 * Subvolume creation is not restricted, but snapshots
1644 * are limited to own subvolumes only
1645 */
1646 ret = -EPERM;
1647 } else {
1648 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1649 BTRFS_I(src_inode)->root,
1650 transid, readonly, inherit);
1651 }
1652 fdput(src);
1653 }
1654out_drop_write:
1655 mnt_drop_write_file(file);
1656out:
1657 return ret;
1658}
1659
1660static noinline int btrfs_ioctl_snap_create(struct file *file,
1661 void __user *arg, int subvol)
1662{
1663 struct btrfs_ioctl_vol_args *vol_args;
1664 int ret;
1665
1666 vol_args = memdup_user(arg, sizeof(*vol_args));
1667 if (IS_ERR(vol_args))
1668 return PTR_ERR(vol_args);
1669 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1670
1671 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1672 vol_args->fd, subvol,
1673 NULL, false, NULL);
1674
1675 kfree(vol_args);
1676 return ret;
1677}
1678
1679static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1680 void __user *arg, int subvol)
1681{
1682 struct btrfs_ioctl_vol_args_v2 *vol_args;
1683 int ret;
1684 u64 transid = 0;
1685 u64 *ptr = NULL;
1686 bool readonly = false;
1687 struct btrfs_qgroup_inherit *inherit = NULL;
1688
1689 vol_args = memdup_user(arg, sizeof(*vol_args));
1690 if (IS_ERR(vol_args))
1691 return PTR_ERR(vol_args);
1692 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1693
1694 if (vol_args->flags &
1695 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1696 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1697 ret = -EOPNOTSUPP;
1698 goto free_args;
1699 }
1700
1701 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1702 ptr = &transid;
1703 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1704 readonly = true;
1705 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1706 if (vol_args->size > PAGE_CACHE_SIZE) {
1707 ret = -EINVAL;
1708 goto free_args;
1709 }
1710 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1711 if (IS_ERR(inherit)) {
1712 ret = PTR_ERR(inherit);
1713 goto free_args;
1714 }
1715 }
1716
1717 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1718 vol_args->fd, subvol, ptr,
1719 readonly, inherit);
1720 if (ret)
1721 goto free_inherit;
1722
1723 if (ptr && copy_to_user(arg +
1724 offsetof(struct btrfs_ioctl_vol_args_v2,
1725 transid),
1726 ptr, sizeof(*ptr)))
1727 ret = -EFAULT;
1728
1729free_inherit:
1730 kfree(inherit);
1731free_args:
1732 kfree(vol_args);
1733 return ret;
1734}
1735
1736static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1737 void __user *arg)
1738{
1739 struct inode *inode = file_inode(file);
1740 struct btrfs_root *root = BTRFS_I(inode)->root;
1741 int ret = 0;
1742 u64 flags = 0;
1743
1744 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1745 return -EINVAL;
1746
1747 down_read(&root->fs_info->subvol_sem);
1748 if (btrfs_root_readonly(root))
1749 flags |= BTRFS_SUBVOL_RDONLY;
1750 up_read(&root->fs_info->subvol_sem);
1751
1752 if (copy_to_user(arg, &flags, sizeof(flags)))
1753 ret = -EFAULT;
1754
1755 return ret;
1756}
1757
1758static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1759 void __user *arg)
1760{
1761 struct inode *inode = file_inode(file);
1762 struct btrfs_root *root = BTRFS_I(inode)->root;
1763 struct btrfs_trans_handle *trans;
1764 u64 root_flags;
1765 u64 flags;
1766 int ret = 0;
1767
1768 if (!inode_owner_or_capable(inode))
1769 return -EPERM;
1770
1771 ret = mnt_want_write_file(file);
1772 if (ret)
1773 goto out;
1774
1775 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1776 ret = -EINVAL;
1777 goto out_drop_write;
1778 }
1779
1780 if (copy_from_user(&flags, arg, sizeof(flags))) {
1781 ret = -EFAULT;
1782 goto out_drop_write;
1783 }
1784
1785 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1786 ret = -EINVAL;
1787 goto out_drop_write;
1788 }
1789
1790 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1791 ret = -EOPNOTSUPP;
1792 goto out_drop_write;
1793 }
1794
1795 down_write(&root->fs_info->subvol_sem);
1796
1797 /* nothing to do */
1798 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1799 goto out_drop_sem;
1800
1801 root_flags = btrfs_root_flags(&root->root_item);
1802 if (flags & BTRFS_SUBVOL_RDONLY) {
1803 btrfs_set_root_flags(&root->root_item,
1804 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1805 } else {
1806 /*
1807 * Block RO -> RW transition if this subvolume is involved in
1808 * send
1809 */
1810 spin_lock(&root->root_item_lock);
1811 if (root->send_in_progress == 0) {
1812 btrfs_set_root_flags(&root->root_item,
1813 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1814 spin_unlock(&root->root_item_lock);
1815 } else {
1816 spin_unlock(&root->root_item_lock);
1817 btrfs_warn(root->fs_info,
1818 "Attempt to set subvolume %llu read-write during send",
1819 root->root_key.objectid);
1820 ret = -EPERM;
1821 goto out_drop_sem;
1822 }
1823 }
1824
1825 trans = btrfs_start_transaction(root, 1);
1826 if (IS_ERR(trans)) {
1827 ret = PTR_ERR(trans);
1828 goto out_reset;
1829 }
1830
1831 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1832 &root->root_key, &root->root_item);
1833
1834 btrfs_commit_transaction(trans, root);
1835out_reset:
1836 if (ret)
1837 btrfs_set_root_flags(&root->root_item, root_flags);
1838out_drop_sem:
1839 up_write(&root->fs_info->subvol_sem);
1840out_drop_write:
1841 mnt_drop_write_file(file);
1842out:
1843 return ret;
1844}
1845
1846/*
1847 * helper to check if the subvolume references other subvolumes
1848 */
1849static noinline int may_destroy_subvol(struct btrfs_root *root)
1850{
1851 struct btrfs_path *path;
1852 struct btrfs_dir_item *di;
1853 struct btrfs_key key;
1854 u64 dir_id;
1855 int ret;
1856
1857 path = btrfs_alloc_path();
1858 if (!path)
1859 return -ENOMEM;
1860
1861 /* Make sure this root isn't set as the default subvol */
1862 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1863 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1864 dir_id, "default", 7, 0);
1865 if (di && !IS_ERR(di)) {
1866 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1867 if (key.objectid == root->root_key.objectid) {
1868 ret = -EPERM;
1869 btrfs_err(root->fs_info, "deleting default subvolume "
1870 "%llu is not allowed", key.objectid);
1871 goto out;
1872 }
1873 btrfs_release_path(path);
1874 }
1875
1876 key.objectid = root->root_key.objectid;
1877 key.type = BTRFS_ROOT_REF_KEY;
1878 key.offset = (u64)-1;
1879
1880 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1881 &key, path, 0, 0);
1882 if (ret < 0)
1883 goto out;
1884 BUG_ON(ret == 0);
1885
1886 ret = 0;
1887 if (path->slots[0] > 0) {
1888 path->slots[0]--;
1889 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1890 if (key.objectid == root->root_key.objectid &&
1891 key.type == BTRFS_ROOT_REF_KEY)
1892 ret = -ENOTEMPTY;
1893 }
1894out:
1895 btrfs_free_path(path);
1896 return ret;
1897}
1898
1899static noinline int key_in_sk(struct btrfs_key *key,
1900 struct btrfs_ioctl_search_key *sk)
1901{
1902 struct btrfs_key test;
1903 int ret;
1904
1905 test.objectid = sk->min_objectid;
1906 test.type = sk->min_type;
1907 test.offset = sk->min_offset;
1908
1909 ret = btrfs_comp_cpu_keys(key, &test);
1910 if (ret < 0)
1911 return 0;
1912
1913 test.objectid = sk->max_objectid;
1914 test.type = sk->max_type;
1915 test.offset = sk->max_offset;
1916
1917 ret = btrfs_comp_cpu_keys(key, &test);
1918 if (ret > 0)
1919 return 0;
1920 return 1;
1921}
1922
1923static noinline int copy_to_sk(struct btrfs_root *root,
1924 struct btrfs_path *path,
1925 struct btrfs_key *key,
1926 struct btrfs_ioctl_search_key *sk,
1927 size_t *buf_size,
1928 char __user *ubuf,
1929 unsigned long *sk_offset,
1930 int *num_found)
1931{
1932 u64 found_transid;
1933 struct extent_buffer *leaf;
1934 struct btrfs_ioctl_search_header sh;
1935 unsigned long item_off;
1936 unsigned long item_len;
1937 int nritems;
1938 int i;
1939 int slot;
1940 int ret = 0;
1941
1942 leaf = path->nodes[0];
1943 slot = path->slots[0];
1944 nritems = btrfs_header_nritems(leaf);
1945
1946 if (btrfs_header_generation(leaf) > sk->max_transid) {
1947 i = nritems;
1948 goto advance_key;
1949 }
1950 found_transid = btrfs_header_generation(leaf);
1951
1952 for (i = slot; i < nritems; i++) {
1953 item_off = btrfs_item_ptr_offset(leaf, i);
1954 item_len = btrfs_item_size_nr(leaf, i);
1955
1956 btrfs_item_key_to_cpu(leaf, key, i);
1957 if (!key_in_sk(key, sk))
1958 continue;
1959
1960 if (sizeof(sh) + item_len > *buf_size) {
1961 if (*num_found) {
1962 ret = 1;
1963 goto out;
1964 }
1965
1966 /*
1967 * return one empty item back for v1, which does not
1968 * handle -EOVERFLOW
1969 */
1970
1971 *buf_size = sizeof(sh) + item_len;
1972 item_len = 0;
1973 ret = -EOVERFLOW;
1974 }
1975
1976 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1977 ret = 1;
1978 goto out;
1979 }
1980
1981 sh.objectid = key->objectid;
1982 sh.offset = key->offset;
1983 sh.type = key->type;
1984 sh.len = item_len;
1985 sh.transid = found_transid;
1986
1987 /* copy search result header */
1988 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1989 ret = -EFAULT;
1990 goto out;
1991 }
1992
1993 *sk_offset += sizeof(sh);
1994
1995 if (item_len) {
1996 char __user *up = ubuf + *sk_offset;
1997 /* copy the item */
1998 if (read_extent_buffer_to_user(leaf, up,
1999 item_off, item_len)) {
2000 ret = -EFAULT;
2001 goto out;
2002 }
2003
2004 *sk_offset += item_len;
2005 }
2006 (*num_found)++;
2007
2008 if (ret) /* -EOVERFLOW from above */
2009 goto out;
2010
2011 if (*num_found >= sk->nr_items) {
2012 ret = 1;
2013 goto out;
2014 }
2015 }
2016advance_key:
2017 ret = 0;
2018 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
2019 key->offset++;
2020 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2021 key->offset = 0;
2022 key->type++;
2023 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2024 key->offset = 0;
2025 key->type = 0;
2026 key->objectid++;
2027 } else
2028 ret = 1;
2029out:
2030 /*
2031 * 0: all items from this leaf copied, continue with next
2032 * 1: * more items can be copied, but unused buffer is too small
2033 * * all items were found
2034 * Either way, it will stops the loop which iterates to the next
2035 * leaf
2036 * -EOVERFLOW: item was to large for buffer
2037 * -EFAULT: could not copy extent buffer back to userspace
2038 */
2039 return ret;
2040}
2041
2042static noinline int search_ioctl(struct inode *inode,
2043 struct btrfs_ioctl_search_key *sk,
2044 size_t *buf_size,
2045 char __user *ubuf)
2046{
2047 struct btrfs_root *root;
2048 struct btrfs_key key;
2049 struct btrfs_path *path;
2050 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2051 int ret;
2052 int num_found = 0;
2053 unsigned long sk_offset = 0;
2054
2055 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2056 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2057 return -EOVERFLOW;
2058 }
2059
2060 path = btrfs_alloc_path();
2061 if (!path)
2062 return -ENOMEM;
2063
2064 if (sk->tree_id == 0) {
2065 /* search the root of the inode that was passed */
2066 root = BTRFS_I(inode)->root;
2067 } else {
2068 key.objectid = sk->tree_id;
2069 key.type = BTRFS_ROOT_ITEM_KEY;
2070 key.offset = (u64)-1;
2071 root = btrfs_read_fs_root_no_name(info, &key);
2072 if (IS_ERR(root)) {
2073 printk(KERN_ERR "BTRFS: could not find root %llu\n",
2074 sk->tree_id);
2075 btrfs_free_path(path);
2076 return -ENOENT;
2077 }
2078 }
2079
2080 key.objectid = sk->min_objectid;
2081 key.type = sk->min_type;
2082 key.offset = sk->min_offset;
2083
2084 while (1) {
2085 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2086 if (ret != 0) {
2087 if (ret > 0)
2088 ret = 0;
2089 goto err;
2090 }
2091 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2092 &sk_offset, &num_found);
2093 btrfs_release_path(path);
2094 if (ret)
2095 break;
2096
2097 }
2098 if (ret > 0)
2099 ret = 0;
2100err:
2101 sk->nr_items = num_found;
2102 btrfs_free_path(path);
2103 return ret;
2104}
2105
2106static noinline int btrfs_ioctl_tree_search(struct file *file,
2107 void __user *argp)
2108{
2109 struct btrfs_ioctl_search_args __user *uargs;
2110 struct btrfs_ioctl_search_key sk;
2111 struct inode *inode;
2112 int ret;
2113 size_t buf_size;
2114
2115 if (!capable(CAP_SYS_ADMIN))
2116 return -EPERM;
2117
2118 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2119
2120 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2121 return -EFAULT;
2122
2123 buf_size = sizeof(uargs->buf);
2124
2125 inode = file_inode(file);
2126 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2127
2128 /*
2129 * In the origin implementation an overflow is handled by returning a
2130 * search header with a len of zero, so reset ret.
2131 */
2132 if (ret == -EOVERFLOW)
2133 ret = 0;
2134
2135 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2136 ret = -EFAULT;
2137 return ret;
2138}
2139
2140static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2141 void __user *argp)
2142{
2143 struct btrfs_ioctl_search_args_v2 __user *uarg;
2144 struct btrfs_ioctl_search_args_v2 args;
2145 struct inode *inode;
2146 int ret;
2147 size_t buf_size;
2148 const size_t buf_limit = 16 * 1024 * 1024;
2149
2150 if (!capable(CAP_SYS_ADMIN))
2151 return -EPERM;
2152
2153 /* copy search header and buffer size */
2154 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2155 if (copy_from_user(&args, uarg, sizeof(args)))
2156 return -EFAULT;
2157
2158 buf_size = args.buf_size;
2159
2160 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2161 return -EOVERFLOW;
2162
2163 /* limit result size to 16MB */
2164 if (buf_size > buf_limit)
2165 buf_size = buf_limit;
2166
2167 inode = file_inode(file);
2168 ret = search_ioctl(inode, &args.key, &buf_size,
2169 (char *)(&uarg->buf[0]));
2170 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2171 ret = -EFAULT;
2172 else if (ret == -EOVERFLOW &&
2173 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2174 ret = -EFAULT;
2175
2176 return ret;
2177}
2178
2179/*
2180 * Search INODE_REFs to identify path name of 'dirid' directory
2181 * in a 'tree_id' tree. and sets path name to 'name'.
2182 */
2183static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2184 u64 tree_id, u64 dirid, char *name)
2185{
2186 struct btrfs_root *root;
2187 struct btrfs_key key;
2188 char *ptr;
2189 int ret = -1;
2190 int slot;
2191 int len;
2192 int total_len = 0;
2193 struct btrfs_inode_ref *iref;
2194 struct extent_buffer *l;
2195 struct btrfs_path *path;
2196
2197 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2198 name[0]='\0';
2199 return 0;
2200 }
2201
2202 path = btrfs_alloc_path();
2203 if (!path)
2204 return -ENOMEM;
2205
2206 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2207
2208 key.objectid = tree_id;
2209 key.type = BTRFS_ROOT_ITEM_KEY;
2210 key.offset = (u64)-1;
2211 root = btrfs_read_fs_root_no_name(info, &key);
2212 if (IS_ERR(root)) {
2213 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2214 ret = -ENOENT;
2215 goto out;
2216 }
2217
2218 key.objectid = dirid;
2219 key.type = BTRFS_INODE_REF_KEY;
2220 key.offset = (u64)-1;
2221
2222 while (1) {
2223 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2224 if (ret < 0)
2225 goto out;
2226 else if (ret > 0) {
2227 ret = btrfs_previous_item(root, path, dirid,
2228 BTRFS_INODE_REF_KEY);
2229 if (ret < 0)
2230 goto out;
2231 else if (ret > 0) {
2232 ret = -ENOENT;
2233 goto out;
2234 }
2235 }
2236
2237 l = path->nodes[0];
2238 slot = path->slots[0];
2239 btrfs_item_key_to_cpu(l, &key, slot);
2240
2241 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2242 len = btrfs_inode_ref_name_len(l, iref);
2243 ptr -= len + 1;
2244 total_len += len + 1;
2245 if (ptr < name) {
2246 ret = -ENAMETOOLONG;
2247 goto out;
2248 }
2249
2250 *(ptr + len) = '/';
2251 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2252
2253 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2254 break;
2255
2256 btrfs_release_path(path);
2257 key.objectid = key.offset;
2258 key.offset = (u64)-1;
2259 dirid = key.objectid;
2260 }
2261 memmove(name, ptr, total_len);
2262 name[total_len] = '\0';
2263 ret = 0;
2264out:
2265 btrfs_free_path(path);
2266 return ret;
2267}
2268
2269static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2270 void __user *argp)
2271{
2272 struct btrfs_ioctl_ino_lookup_args *args;
2273 struct inode *inode;
2274 int ret;
2275
2276 if (!capable(CAP_SYS_ADMIN))
2277 return -EPERM;
2278
2279 args = memdup_user(argp, sizeof(*args));
2280 if (IS_ERR(args))
2281 return PTR_ERR(args);
2282
2283 inode = file_inode(file);
2284
2285 if (args->treeid == 0)
2286 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2287
2288 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2289 args->treeid, args->objectid,
2290 args->name);
2291
2292 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2293 ret = -EFAULT;
2294
2295 kfree(args);
2296 return ret;
2297}
2298
2299static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2300 void __user *arg)
2301{
2302 struct dentry *parent = file->f_path.dentry;
2303 struct dentry *dentry;
2304 struct inode *dir = d_inode(parent);
2305 struct inode *inode;
2306 struct btrfs_root *root = BTRFS_I(dir)->root;
2307 struct btrfs_root *dest = NULL;
2308 struct btrfs_ioctl_vol_args *vol_args;
2309 struct btrfs_trans_handle *trans;
2310 struct btrfs_block_rsv block_rsv;
2311 u64 root_flags;
2312 u64 qgroup_reserved;
2313 int namelen;
2314 int ret;
2315 int err = 0;
2316
2317 vol_args = memdup_user(arg, sizeof(*vol_args));
2318 if (IS_ERR(vol_args))
2319 return PTR_ERR(vol_args);
2320
2321 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2322 namelen = strlen(vol_args->name);
2323 if (strchr(vol_args->name, '/') ||
2324 strncmp(vol_args->name, "..", namelen) == 0) {
2325 err = -EINVAL;
2326 goto out;
2327 }
2328
2329 err = mnt_want_write_file(file);
2330 if (err)
2331 goto out;
2332
2333
2334 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2335 if (err == -EINTR)
2336 goto out_drop_write;
2337 dentry = lookup_one_len(vol_args->name, parent, namelen);
2338 if (IS_ERR(dentry)) {
2339 err = PTR_ERR(dentry);
2340 goto out_unlock_dir;
2341 }
2342
2343 if (d_really_is_negative(dentry)) {
2344 err = -ENOENT;
2345 goto out_dput;
2346 }
2347
2348 inode = d_inode(dentry);
2349 dest = BTRFS_I(inode)->root;
2350 if (!capable(CAP_SYS_ADMIN)) {
2351 /*
2352 * Regular user. Only allow this with a special mount
2353 * option, when the user has write+exec access to the
2354 * subvol root, and when rmdir(2) would have been
2355 * allowed.
2356 *
2357 * Note that this is _not_ check that the subvol is
2358 * empty or doesn't contain data that we wouldn't
2359 * otherwise be able to delete.
2360 *
2361 * Users who want to delete empty subvols should try
2362 * rmdir(2).
2363 */
2364 err = -EPERM;
2365 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2366 goto out_dput;
2367
2368 /*
2369 * Do not allow deletion if the parent dir is the same
2370 * as the dir to be deleted. That means the ioctl
2371 * must be called on the dentry referencing the root
2372 * of the subvol, not a random directory contained
2373 * within it.
2374 */
2375 err = -EINVAL;
2376 if (root == dest)
2377 goto out_dput;
2378
2379 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2380 if (err)
2381 goto out_dput;
2382 }
2383
2384 /* check if subvolume may be deleted by a user */
2385 err = btrfs_may_delete(dir, dentry, 1);
2386 if (err)
2387 goto out_dput;
2388
2389 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2390 err = -EINVAL;
2391 goto out_dput;
2392 }
2393
2394 mutex_lock(&inode->i_mutex);
2395
2396 /*
2397 * Don't allow to delete a subvolume with send in progress. This is
2398 * inside the i_mutex so the error handling that has to drop the bit
2399 * again is not run concurrently.
2400 */
2401 spin_lock(&dest->root_item_lock);
2402 root_flags = btrfs_root_flags(&dest->root_item);
2403 if (dest->send_in_progress == 0) {
2404 btrfs_set_root_flags(&dest->root_item,
2405 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2406 spin_unlock(&dest->root_item_lock);
2407 } else {
2408 spin_unlock(&dest->root_item_lock);
2409 btrfs_warn(root->fs_info,
2410 "Attempt to delete subvolume %llu during send",
2411 dest->root_key.objectid);
2412 err = -EPERM;
2413 goto out_unlock_inode;
2414 }
2415
2416 d_invalidate(dentry);
2417
2418 down_write(&root->fs_info->subvol_sem);
2419
2420 err = may_destroy_subvol(dest);
2421 if (err)
2422 goto out_up_write;
2423
2424 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2425 /*
2426 * One for dir inode, two for dir entries, two for root
2427 * ref/backref.
2428 */
2429 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2430 5, &qgroup_reserved, true);
2431 if (err)
2432 goto out_up_write;
2433
2434 trans = btrfs_start_transaction(root, 0);
2435 if (IS_ERR(trans)) {
2436 err = PTR_ERR(trans);
2437 goto out_release;
2438 }
2439 trans->block_rsv = &block_rsv;
2440 trans->bytes_reserved = block_rsv.size;
2441
2442 ret = btrfs_unlink_subvol(trans, root, dir,
2443 dest->root_key.objectid,
2444 dentry->d_name.name,
2445 dentry->d_name.len);
2446 if (ret) {
2447 err = ret;
2448 btrfs_abort_transaction(trans, root, ret);
2449 goto out_end_trans;
2450 }
2451
2452 btrfs_record_root_in_trans(trans, dest);
2453
2454 memset(&dest->root_item.drop_progress, 0,
2455 sizeof(dest->root_item.drop_progress));
2456 dest->root_item.drop_level = 0;
2457 btrfs_set_root_refs(&dest->root_item, 0);
2458
2459 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2460 ret = btrfs_insert_orphan_item(trans,
2461 root->fs_info->tree_root,
2462 dest->root_key.objectid);
2463 if (ret) {
2464 btrfs_abort_transaction(trans, root, ret);
2465 err = ret;
2466 goto out_end_trans;
2467 }
2468 }
2469
2470 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2471 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2472 dest->root_key.objectid);
2473 if (ret && ret != -ENOENT) {
2474 btrfs_abort_transaction(trans, root, ret);
2475 err = ret;
2476 goto out_end_trans;
2477 }
2478 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2479 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2480 dest->root_item.received_uuid,
2481 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2482 dest->root_key.objectid);
2483 if (ret && ret != -ENOENT) {
2484 btrfs_abort_transaction(trans, root, ret);
2485 err = ret;
2486 goto out_end_trans;
2487 }
2488 }
2489
2490out_end_trans:
2491 trans->block_rsv = NULL;
2492 trans->bytes_reserved = 0;
2493 ret = btrfs_end_transaction(trans, root);
2494 if (ret && !err)
2495 err = ret;
2496 inode->i_flags |= S_DEAD;
2497out_release:
2498 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2499out_up_write:
2500 up_write(&root->fs_info->subvol_sem);
2501 if (err) {
2502 spin_lock(&dest->root_item_lock);
2503 root_flags = btrfs_root_flags(&dest->root_item);
2504 btrfs_set_root_flags(&dest->root_item,
2505 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2506 spin_unlock(&dest->root_item_lock);
2507 }
2508out_unlock_inode:
2509 mutex_unlock(&inode->i_mutex);
2510 if (!err) {
2511 shrink_dcache_sb(root->fs_info->sb);
2512 btrfs_invalidate_inodes(dest);
2513 d_delete(dentry);
2514 ASSERT(dest->send_in_progress == 0);
2515
2516 /* the last ref */
2517 if (dest->ino_cache_inode) {
2518 iput(dest->ino_cache_inode);
2519 dest->ino_cache_inode = NULL;
2520 }
2521 }
2522out_dput:
2523 dput(dentry);
2524out_unlock_dir:
2525 mutex_unlock(&dir->i_mutex);
2526out_drop_write:
2527 mnt_drop_write_file(file);
2528out:
2529 kfree(vol_args);
2530 return err;
2531}
2532
2533static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2534{
2535 struct inode *inode = file_inode(file);
2536 struct btrfs_root *root = BTRFS_I(inode)->root;
2537 struct btrfs_ioctl_defrag_range_args *range;
2538 int ret;
2539
2540 ret = mnt_want_write_file(file);
2541 if (ret)
2542 return ret;
2543
2544 if (btrfs_root_readonly(root)) {
2545 ret = -EROFS;
2546 goto out;
2547 }
2548
2549 switch (inode->i_mode & S_IFMT) {
2550 case S_IFDIR:
2551 if (!capable(CAP_SYS_ADMIN)) {
2552 ret = -EPERM;
2553 goto out;
2554 }
2555 ret = btrfs_defrag_root(root);
2556 if (ret)
2557 goto out;
2558 ret = btrfs_defrag_root(root->fs_info->extent_root);
2559 break;
2560 case S_IFREG:
2561 if (!(file->f_mode & FMODE_WRITE)) {
2562 ret = -EINVAL;
2563 goto out;
2564 }
2565
2566 range = kzalloc(sizeof(*range), GFP_KERNEL);
2567 if (!range) {
2568 ret = -ENOMEM;
2569 goto out;
2570 }
2571
2572 if (argp) {
2573 if (copy_from_user(range, argp,
2574 sizeof(*range))) {
2575 ret = -EFAULT;
2576 kfree(range);
2577 goto out;
2578 }
2579 /* compression requires us to start the IO */
2580 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2581 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2582 range->extent_thresh = (u32)-1;
2583 }
2584 } else {
2585 /* the rest are all set to zero by kzalloc */
2586 range->len = (u64)-1;
2587 }
2588 ret = btrfs_defrag_file(file_inode(file), file,
2589 range, 0, 0);
2590 if (ret > 0)
2591 ret = 0;
2592 kfree(range);
2593 break;
2594 default:
2595 ret = -EINVAL;
2596 }
2597out:
2598 mnt_drop_write_file(file);
2599 return ret;
2600}
2601
2602static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2603{
2604 struct btrfs_ioctl_vol_args *vol_args;
2605 int ret;
2606
2607 if (!capable(CAP_SYS_ADMIN))
2608 return -EPERM;
2609
2610 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2611 1)) {
2612 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2613 }
2614
2615 mutex_lock(&root->fs_info->volume_mutex);
2616 vol_args = memdup_user(arg, sizeof(*vol_args));
2617 if (IS_ERR(vol_args)) {
2618 ret = PTR_ERR(vol_args);
2619 goto out;
2620 }
2621
2622 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2623 ret = btrfs_init_new_device(root, vol_args->name);
2624
2625 if (!ret)
2626 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2627
2628 kfree(vol_args);
2629out:
2630 mutex_unlock(&root->fs_info->volume_mutex);
2631 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2632 return ret;
2633}
2634
2635static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2636{
2637 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2638 struct btrfs_ioctl_vol_args *vol_args;
2639 int ret;
2640
2641 if (!capable(CAP_SYS_ADMIN))
2642 return -EPERM;
2643
2644 ret = mnt_want_write_file(file);
2645 if (ret)
2646 return ret;
2647
2648 vol_args = memdup_user(arg, sizeof(*vol_args));
2649 if (IS_ERR(vol_args)) {
2650 ret = PTR_ERR(vol_args);
2651 goto err_drop;
2652 }
2653
2654 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2655
2656 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2657 1)) {
2658 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2659 goto out;
2660 }
2661
2662 mutex_lock(&root->fs_info->volume_mutex);
2663 ret = btrfs_rm_device(root, vol_args->name);
2664 mutex_unlock(&root->fs_info->volume_mutex);
2665 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2666
2667 if (!ret)
2668 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2669
2670out:
2671 kfree(vol_args);
2672err_drop:
2673 mnt_drop_write_file(file);
2674 return ret;
2675}
2676
2677static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2678{
2679 struct btrfs_ioctl_fs_info_args *fi_args;
2680 struct btrfs_device *device;
2681 struct btrfs_device *next;
2682 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2683 int ret = 0;
2684
2685 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2686 if (!fi_args)
2687 return -ENOMEM;
2688
2689 mutex_lock(&fs_devices->device_list_mutex);
2690 fi_args->num_devices = fs_devices->num_devices;
2691 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2692
2693 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2694 if (device->devid > fi_args->max_id)
2695 fi_args->max_id = device->devid;
2696 }
2697 mutex_unlock(&fs_devices->device_list_mutex);
2698
2699 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2700 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2701 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2702
2703 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2704 ret = -EFAULT;
2705
2706 kfree(fi_args);
2707 return ret;
2708}
2709
2710static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2711{
2712 struct btrfs_ioctl_dev_info_args *di_args;
2713 struct btrfs_device *dev;
2714 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2715 int ret = 0;
2716 char *s_uuid = NULL;
2717
2718 di_args = memdup_user(arg, sizeof(*di_args));
2719 if (IS_ERR(di_args))
2720 return PTR_ERR(di_args);
2721
2722 if (!btrfs_is_empty_uuid(di_args->uuid))
2723 s_uuid = di_args->uuid;
2724
2725 mutex_lock(&fs_devices->device_list_mutex);
2726 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2727
2728 if (!dev) {
2729 ret = -ENODEV;
2730 goto out;
2731 }
2732
2733 di_args->devid = dev->devid;
2734 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2735 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2736 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2737 if (dev->name) {
2738 struct rcu_string *name;
2739
2740 rcu_read_lock();
2741 name = rcu_dereference(dev->name);
2742 strncpy(di_args->path, name->str, sizeof(di_args->path));
2743 rcu_read_unlock();
2744 di_args->path[sizeof(di_args->path) - 1] = 0;
2745 } else {
2746 di_args->path[0] = '\0';
2747 }
2748
2749out:
2750 mutex_unlock(&fs_devices->device_list_mutex);
2751 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2752 ret = -EFAULT;
2753
2754 kfree(di_args);
2755 return ret;
2756}
2757
2758static struct page *extent_same_get_page(struct inode *inode, u64 off)
2759{
2760 struct page *page;
2761 pgoff_t index;
2762 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2763
2764 index = off >> PAGE_CACHE_SHIFT;
2765
2766 page = grab_cache_page(inode->i_mapping, index);
2767 if (!page)
2768 return NULL;
2769
2770 if (!PageUptodate(page)) {
2771 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2772 0))
2773 return NULL;
2774 lock_page(page);
2775 if (!PageUptodate(page)) {
2776 unlock_page(page);
2777 page_cache_release(page);
2778 return NULL;
2779 }
2780 }
2781 unlock_page(page);
2782
2783 return page;
2784}
2785
2786static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2787{
2788 /* do any pending delalloc/csum calc on src, one way or
2789 another, and lock file content */
2790 while (1) {
2791 struct btrfs_ordered_extent *ordered;
2792 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2793 ordered = btrfs_lookup_first_ordered_extent(inode,
2794 off + len - 1);
2795 if ((!ordered ||
2796 ordered->file_offset + ordered->len <= off ||
2797 ordered->file_offset >= off + len) &&
2798 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2799 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2800 if (ordered)
2801 btrfs_put_ordered_extent(ordered);
2802 break;
2803 }
2804 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2805 if (ordered)
2806 btrfs_put_ordered_extent(ordered);
2807 btrfs_wait_ordered_range(inode, off, len);
2808 }
2809}
2810
2811static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2812 struct inode *inode2, u64 loff2, u64 len)
2813{
2814 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2815 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2816
2817 mutex_unlock(&inode1->i_mutex);
2818 mutex_unlock(&inode2->i_mutex);
2819}
2820
2821static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2822 struct inode *inode2, u64 loff2, u64 len)
2823{
2824 if (inode1 < inode2) {
2825 swap(inode1, inode2);
2826 swap(loff1, loff2);
2827 }
2828
2829 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2830 lock_extent_range(inode1, loff1, len);
2831 if (inode1 != inode2) {
2832 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2833 lock_extent_range(inode2, loff2, len);
2834 }
2835}
2836
2837static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2838 u64 dst_loff, u64 len)
2839{
2840 int ret = 0;
2841 struct page *src_page, *dst_page;
2842 unsigned int cmp_len = PAGE_CACHE_SIZE;
2843 void *addr, *dst_addr;
2844
2845 while (len) {
2846 if (len < PAGE_CACHE_SIZE)
2847 cmp_len = len;
2848
2849 src_page = extent_same_get_page(src, loff);
2850 if (!src_page)
2851 return -EINVAL;
2852 dst_page = extent_same_get_page(dst, dst_loff);
2853 if (!dst_page) {
2854 page_cache_release(src_page);
2855 return -EINVAL;
2856 }
2857 addr = kmap_atomic(src_page);
2858 dst_addr = kmap_atomic(dst_page);
2859
2860 flush_dcache_page(src_page);
2861 flush_dcache_page(dst_page);
2862
2863 if (memcmp(addr, dst_addr, cmp_len))
2864 ret = BTRFS_SAME_DATA_DIFFERS;
2865
2866 kunmap_atomic(addr);
2867 kunmap_atomic(dst_addr);
2868 page_cache_release(src_page);
2869 page_cache_release(dst_page);
2870
2871 if (ret)
2872 break;
2873
2874 loff += cmp_len;
2875 dst_loff += cmp_len;
2876 len -= cmp_len;
2877 }
2878
2879 return ret;
2880}
2881
2882static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2883{
2884 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2885
2886 if (off + len > inode->i_size || off + len < off)
2887 return -EINVAL;
2888 /* Check that we are block aligned - btrfs_clone() requires this */
2889 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2890 return -EINVAL;
2891
2892 return 0;
2893}
2894
2895static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2896 struct inode *dst, u64 dst_loff)
2897{
2898 int ret;
2899
2900 /*
2901 * btrfs_clone() can't handle extents in the same file
2902 * yet. Once that works, we can drop this check and replace it
2903 * with a check for the same inode, but overlapping extents.
2904 */
2905 if (src == dst)
2906 return -EINVAL;
2907
2908 if (len == 0)
2909 return 0;
2910
2911 btrfs_double_lock(src, loff, dst, dst_loff, len);
2912
2913 ret = extent_same_check_offsets(src, loff, len);
2914 if (ret)
2915 goto out_unlock;
2916
2917 ret = extent_same_check_offsets(dst, dst_loff, len);
2918 if (ret)
2919 goto out_unlock;
2920
2921 /* don't make the dst file partly checksummed */
2922 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2923 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2924 ret = -EINVAL;
2925 goto out_unlock;
2926 }
2927
2928 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2929 if (ret == 0)
2930 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2931
2932out_unlock:
2933 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2934
2935 return ret;
2936}
2937
2938#define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2939
2940static long btrfs_ioctl_file_extent_same(struct file *file,
2941 struct btrfs_ioctl_same_args __user *argp)
2942{
2943 struct btrfs_ioctl_same_args *same;
2944 struct btrfs_ioctl_same_extent_info *info;
2945 struct inode *src = file_inode(file);
2946 u64 off;
2947 u64 len;
2948 int i;
2949 int ret;
2950 unsigned long size;
2951 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2952 bool is_admin = capable(CAP_SYS_ADMIN);
2953 u16 count;
2954
2955 if (!(file->f_mode & FMODE_READ))
2956 return -EINVAL;
2957
2958 ret = mnt_want_write_file(file);
2959 if (ret)
2960 return ret;
2961
2962 if (get_user(count, &argp->dest_count)) {
2963 ret = -EFAULT;
2964 goto out;
2965 }
2966
2967 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
2968
2969 same = memdup_user(argp, size);
2970
2971 if (IS_ERR(same)) {
2972 ret = PTR_ERR(same);
2973 goto out;
2974 }
2975
2976 off = same->logical_offset;
2977 len = same->length;
2978
2979 /*
2980 * Limit the total length we will dedupe for each operation.
2981 * This is intended to bound the total time spent in this
2982 * ioctl to something sane.
2983 */
2984 if (len > BTRFS_MAX_DEDUPE_LEN)
2985 len = BTRFS_MAX_DEDUPE_LEN;
2986
2987 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2988 /*
2989 * Btrfs does not support blocksize < page_size. As a
2990 * result, btrfs_cmp_data() won't correctly handle
2991 * this situation without an update.
2992 */
2993 ret = -EINVAL;
2994 goto out;
2995 }
2996
2997 ret = -EISDIR;
2998 if (S_ISDIR(src->i_mode))
2999 goto out;
3000
3001 ret = -EACCES;
3002 if (!S_ISREG(src->i_mode))
3003 goto out;
3004
3005 /* pre-format output fields to sane values */
3006 for (i = 0; i < count; i++) {
3007 same->info[i].bytes_deduped = 0ULL;
3008 same->info[i].status = 0;
3009 }
3010
3011 for (i = 0, info = same->info; i < count; i++, info++) {
3012 struct inode *dst;
3013 struct fd dst_file = fdget(info->fd);
3014 if (!dst_file.file) {
3015 info->status = -EBADF;
3016 continue;
3017 }
3018 dst = file_inode(dst_file.file);
3019
3020 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
3021 info->status = -EINVAL;
3022 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3023 info->status = -EXDEV;
3024 } else if (S_ISDIR(dst->i_mode)) {
3025 info->status = -EISDIR;
3026 } else if (!S_ISREG(dst->i_mode)) {
3027 info->status = -EACCES;
3028 } else {
3029 info->status = btrfs_extent_same(src, off, len, dst,
3030 info->logical_offset);
3031 if (info->status == 0)
3032 info->bytes_deduped += len;
3033 }
3034 fdput(dst_file);
3035 }
3036
3037 ret = copy_to_user(argp, same, size);
3038 if (ret)
3039 ret = -EFAULT;
3040
3041out:
3042 mnt_drop_write_file(file);
3043 return ret;
3044}
3045
3046/* Helper to check and see if this root currently has a ref on the given disk
3047 * bytenr. If it does then we need to update the quota for this root. This
3048 * doesn't do anything if quotas aren't enabled.
3049 */
3050static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3051 u64 disko)
3052{
3053 struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem);
3054 struct ulist *roots;
3055 struct ulist_iterator uiter;
3056 struct ulist_node *root_node = NULL;
3057 int ret;
3058
3059 if (!root->fs_info->quota_enabled)
3060 return 1;
3061
3062 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3063 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3064 tree_mod_seq_elem.seq, &roots);
3065 if (ret < 0)
3066 goto out;
3067 ret = 0;
3068 ULIST_ITER_INIT(&uiter);
3069 while ((root_node = ulist_next(roots, &uiter))) {
3070 if (root_node->val == root->objectid) {
3071 ret = 1;
3072 break;
3073 }
3074 }
3075 ulist_free(roots);
3076out:
3077 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3078 return ret;
3079}
3080
3081static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3082 struct inode *inode,
3083 u64 endoff,
3084 const u64 destoff,
3085 const u64 olen)
3086{
3087 struct btrfs_root *root = BTRFS_I(inode)->root;
3088 int ret;
3089
3090 inode_inc_iversion(inode);
3091 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3092 /*
3093 * We round up to the block size at eof when determining which
3094 * extents to clone above, but shouldn't round up the file size.
3095 */
3096 if (endoff > destoff + olen)
3097 endoff = destoff + olen;
3098 if (endoff > inode->i_size)
3099 btrfs_i_size_write(inode, endoff);
3100
3101 ret = btrfs_update_inode(trans, root, inode);
3102 if (ret) {
3103 btrfs_abort_transaction(trans, root, ret);
3104 btrfs_end_transaction(trans, root);
3105 goto out;
3106 }
3107 ret = btrfs_end_transaction(trans, root);
3108out:
3109 return ret;
3110}
3111
3112static void clone_update_extent_map(struct inode *inode,
3113 const struct btrfs_trans_handle *trans,
3114 const struct btrfs_path *path,
3115 const u64 hole_offset,
3116 const u64 hole_len)
3117{
3118 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3119 struct extent_map *em;
3120 int ret;
3121
3122 em = alloc_extent_map();
3123 if (!em) {
3124 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3125 &BTRFS_I(inode)->runtime_flags);
3126 return;
3127 }
3128
3129 if (path) {
3130 struct btrfs_file_extent_item *fi;
3131
3132 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3133 struct btrfs_file_extent_item);
3134 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3135 em->generation = -1;
3136 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3137 BTRFS_FILE_EXTENT_INLINE)
3138 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3139 &BTRFS_I(inode)->runtime_flags);
3140 } else {
3141 em->start = hole_offset;
3142 em->len = hole_len;
3143 em->ram_bytes = em->len;
3144 em->orig_start = hole_offset;
3145 em->block_start = EXTENT_MAP_HOLE;
3146 em->block_len = 0;
3147 em->orig_block_len = 0;
3148 em->compress_type = BTRFS_COMPRESS_NONE;
3149 em->generation = trans->transid;
3150 }
3151
3152 while (1) {
3153 write_lock(&em_tree->lock);
3154 ret = add_extent_mapping(em_tree, em, 1);
3155 write_unlock(&em_tree->lock);
3156 if (ret != -EEXIST) {
3157 free_extent_map(em);
3158 break;
3159 }
3160 btrfs_drop_extent_cache(inode, em->start,
3161 em->start + em->len - 1, 0);
3162 }
3163
3164 if (ret)
3165 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3166 &BTRFS_I(inode)->runtime_flags);
3167}
3168
3169/**
3170 * btrfs_clone() - clone a range from inode file to another
3171 *
3172 * @src: Inode to clone from
3173 * @inode: Inode to clone to
3174 * @off: Offset within source to start clone from
3175 * @olen: Original length, passed by user, of range to clone
3176 * @olen_aligned: Block-aligned value of olen, extent_same uses
3177 * identical values here
3178 * @destoff: Offset within @inode to start clone
3179 */
3180static int btrfs_clone(struct inode *src, struct inode *inode,
3181 const u64 off, const u64 olen, const u64 olen_aligned,
3182 const u64 destoff)
3183{
3184 struct btrfs_root *root = BTRFS_I(inode)->root;
3185 struct btrfs_path *path = NULL;
3186 struct extent_buffer *leaf;
3187 struct btrfs_trans_handle *trans;
3188 char *buf = NULL;
3189 struct btrfs_key key;
3190 u32 nritems;
3191 int slot;
3192 int ret;
3193 int no_quota;
3194 const u64 len = olen_aligned;
3195 u64 last_disko = 0;
3196 u64 last_dest_end = destoff;
3197
3198 ret = -ENOMEM;
3199 buf = vmalloc(root->nodesize);
3200 if (!buf)
3201 return ret;
3202
3203 path = btrfs_alloc_path();
3204 if (!path) {
3205 vfree(buf);
3206 return ret;
3207 }
3208
3209 path->reada = 2;
3210 /* clone data */
3211 key.objectid = btrfs_ino(src);
3212 key.type = BTRFS_EXTENT_DATA_KEY;
3213 key.offset = off;
3214
3215 while (1) {
3216 u64 next_key_min_offset = key.offset + 1;
3217
3218 /*
3219 * note the key will change type as we walk through the
3220 * tree.
3221 */
3222 path->leave_spinning = 1;
3223 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3224 0, 0);
3225 if (ret < 0)
3226 goto out;
3227 /*
3228 * First search, if no extent item that starts at offset off was
3229 * found but the previous item is an extent item, it's possible
3230 * it might overlap our target range, therefore process it.
3231 */
3232 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3233 btrfs_item_key_to_cpu(path->nodes[0], &key,
3234 path->slots[0] - 1);
3235 if (key.type == BTRFS_EXTENT_DATA_KEY)
3236 path->slots[0]--;
3237 }
3238
3239 nritems = btrfs_header_nritems(path->nodes[0]);
3240process_slot:
3241 no_quota = 1;
3242 if (path->slots[0] >= nritems) {
3243 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3244 if (ret < 0)
3245 goto out;
3246 if (ret > 0)
3247 break;
3248 nritems = btrfs_header_nritems(path->nodes[0]);
3249 }
3250 leaf = path->nodes[0];
3251 slot = path->slots[0];
3252
3253 btrfs_item_key_to_cpu(leaf, &key, slot);
3254 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3255 key.objectid != btrfs_ino(src))
3256 break;
3257
3258 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3259 struct btrfs_file_extent_item *extent;
3260 int type;
3261 u32 size;
3262 struct btrfs_key new_key;
3263 u64 disko = 0, diskl = 0;
3264 u64 datao = 0, datal = 0;
3265 u8 comp;
3266 u64 drop_start;
3267
3268 extent = btrfs_item_ptr(leaf, slot,
3269 struct btrfs_file_extent_item);
3270 comp = btrfs_file_extent_compression(leaf, extent);
3271 type = btrfs_file_extent_type(leaf, extent);
3272 if (type == BTRFS_FILE_EXTENT_REG ||
3273 type == BTRFS_FILE_EXTENT_PREALLOC) {
3274 disko = btrfs_file_extent_disk_bytenr(leaf,
3275 extent);
3276 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3277 extent);
3278 datao = btrfs_file_extent_offset(leaf, extent);
3279 datal = btrfs_file_extent_num_bytes(leaf,
3280 extent);
3281 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3282 /* take upper bound, may be compressed */
3283 datal = btrfs_file_extent_ram_bytes(leaf,
3284 extent);
3285 }
3286
3287 /*
3288 * The first search might have left us at an extent
3289 * item that ends before our target range's start, can
3290 * happen if we have holes and NO_HOLES feature enabled.
3291 */
3292 if (key.offset + datal <= off) {
3293 path->slots[0]++;
3294 goto process_slot;
3295 } else if (key.offset >= off + len) {
3296 break;
3297 }
3298 next_key_min_offset = key.offset + datal;
3299 size = btrfs_item_size_nr(leaf, slot);
3300 read_extent_buffer(leaf, buf,
3301 btrfs_item_ptr_offset(leaf, slot),
3302 size);
3303
3304 btrfs_release_path(path);
3305 path->leave_spinning = 0;
3306
3307 memcpy(&new_key, &key, sizeof(new_key));
3308 new_key.objectid = btrfs_ino(inode);
3309 if (off <= key.offset)
3310 new_key.offset = key.offset + destoff - off;
3311 else
3312 new_key.offset = destoff;
3313
3314 /*
3315 * Deal with a hole that doesn't have an extent item
3316 * that represents it (NO_HOLES feature enabled).
3317 * This hole is either in the middle of the cloning
3318 * range or at the beginning (fully overlaps it or
3319 * partially overlaps it).
3320 */
3321 if (new_key.offset != last_dest_end)
3322 drop_start = last_dest_end;
3323 else
3324 drop_start = new_key.offset;
3325
3326 /*
3327 * 1 - adjusting old extent (we may have to split it)
3328 * 1 - add new extent
3329 * 1 - inode update
3330 */
3331 trans = btrfs_start_transaction(root, 3);
3332 if (IS_ERR(trans)) {
3333 ret = PTR_ERR(trans);
3334 goto out;
3335 }
3336
3337 if (type == BTRFS_FILE_EXTENT_REG ||
3338 type == BTRFS_FILE_EXTENT_PREALLOC) {
3339 /*
3340 * a | --- range to clone ---| b
3341 * | ------------- extent ------------- |
3342 */
3343
3344 /* subtract range b */
3345 if (key.offset + datal > off + len)
3346 datal = off + len - key.offset;
3347
3348 /* subtract range a */
3349 if (off > key.offset) {
3350 datao += off - key.offset;
3351 datal -= off - key.offset;
3352 }
3353
3354 ret = btrfs_drop_extents(trans, root, inode,
3355 drop_start,
3356 new_key.offset + datal,
3357 1);
3358 if (ret) {
3359 if (ret != -EOPNOTSUPP)
3360 btrfs_abort_transaction(trans,
3361 root, ret);
3362 btrfs_end_transaction(trans, root);
3363 goto out;
3364 }
3365
3366 ret = btrfs_insert_empty_item(trans, root, path,
3367 &new_key, size);
3368 if (ret) {
3369 btrfs_abort_transaction(trans, root,
3370 ret);
3371 btrfs_end_transaction(trans, root);
3372 goto out;
3373 }
3374
3375 leaf = path->nodes[0];
3376 slot = path->slots[0];
3377 write_extent_buffer(leaf, buf,
3378 btrfs_item_ptr_offset(leaf, slot),
3379 size);
3380
3381 extent = btrfs_item_ptr(leaf, slot,
3382 struct btrfs_file_extent_item);
3383
3384 /* disko == 0 means it's a hole */
3385 if (!disko)
3386 datao = 0;
3387
3388 btrfs_set_file_extent_offset(leaf, extent,
3389 datao);
3390 btrfs_set_file_extent_num_bytes(leaf, extent,
3391 datal);
3392
3393 /*
3394 * We need to look up the roots that point at
3395 * this bytenr and see if the new root does. If
3396 * it does not we need to make sure we update
3397 * quotas appropriately.
3398 */
3399 if (disko && root != BTRFS_I(src)->root &&
3400 disko != last_disko) {
3401 no_quota = check_ref(trans, root,
3402 disko);
3403 if (no_quota < 0) {
3404 btrfs_abort_transaction(trans,
3405 root,
3406 ret);
3407 btrfs_end_transaction(trans,
3408 root);
3409 ret = no_quota;
3410 goto out;
3411 }
3412 }
3413
3414 if (disko) {
3415 inode_add_bytes(inode, datal);
3416 ret = btrfs_inc_extent_ref(trans, root,
3417 disko, diskl, 0,
3418 root->root_key.objectid,
3419 btrfs_ino(inode),
3420 new_key.offset - datao,
3421 no_quota);
3422 if (ret) {
3423 btrfs_abort_transaction(trans,
3424 root,
3425 ret);
3426 btrfs_end_transaction(trans,
3427 root);
3428 goto out;
3429
3430 }
3431 }
3432 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3433 u64 skip = 0;
3434 u64 trim = 0;
3435 u64 aligned_end = 0;
3436
3437 if (off > key.offset) {
3438 skip = off - key.offset;
3439 new_key.offset += skip;
3440 }
3441
3442 if (key.offset + datal > off + len)
3443 trim = key.offset + datal - (off + len);
3444
3445 if (comp && (skip || trim)) {
3446 ret = -EINVAL;
3447 btrfs_end_transaction(trans, root);
3448 goto out;
3449 }
3450 size -= skip + trim;
3451 datal -= skip + trim;
3452
3453 aligned_end = ALIGN(new_key.offset + datal,
3454 root->sectorsize);
3455 ret = btrfs_drop_extents(trans, root, inode,
3456 drop_start,
3457 aligned_end,
3458 1);
3459 if (ret) {
3460 if (ret != -EOPNOTSUPP)
3461 btrfs_abort_transaction(trans,
3462 root, ret);
3463 btrfs_end_transaction(trans, root);
3464 goto out;
3465 }
3466
3467 ret = btrfs_insert_empty_item(trans, root, path,
3468 &new_key, size);
3469 if (ret) {
3470 btrfs_abort_transaction(trans, root,
3471 ret);
3472 btrfs_end_transaction(trans, root);
3473 goto out;
3474 }
3475
3476 if (skip) {
3477 u32 start =
3478 btrfs_file_extent_calc_inline_size(0);
3479 memmove(buf+start, buf+start+skip,
3480 datal);
3481 }
3482
3483 leaf = path->nodes[0];
3484 slot = path->slots[0];
3485 write_extent_buffer(leaf, buf,
3486 btrfs_item_ptr_offset(leaf, slot),
3487 size);
3488 inode_add_bytes(inode, datal);
3489 }
3490
3491 /* If we have an implicit hole (NO_HOLES feature). */
3492 if (drop_start < new_key.offset)
3493 clone_update_extent_map(inode, trans,
3494 NULL, drop_start,
3495 new_key.offset - drop_start);
3496
3497 clone_update_extent_map(inode, trans, path, 0, 0);
3498
3499 btrfs_mark_buffer_dirty(leaf);
3500 btrfs_release_path(path);
3501
3502 last_dest_end = ALIGN(new_key.offset + datal,
3503 root->sectorsize);
3504 ret = clone_finish_inode_update(trans, inode,
3505 last_dest_end,
3506 destoff, olen);
3507 if (ret)
3508 goto out;
3509 if (new_key.offset + datal >= destoff + len)
3510 break;
3511 }
3512 btrfs_release_path(path);
3513 key.offset = next_key_min_offset;
3514 }
3515 ret = 0;
3516
3517 if (last_dest_end < destoff + len) {
3518 /*
3519 * We have an implicit hole (NO_HOLES feature is enabled) that
3520 * fully or partially overlaps our cloning range at its end.
3521 */
3522 btrfs_release_path(path);
3523
3524 /*
3525 * 1 - remove extent(s)
3526 * 1 - inode update
3527 */
3528 trans = btrfs_start_transaction(root, 2);
3529 if (IS_ERR(trans)) {
3530 ret = PTR_ERR(trans);
3531 goto out;
3532 }
3533 ret = btrfs_drop_extents(trans, root, inode,
3534 last_dest_end, destoff + len, 1);
3535 if (ret) {
3536 if (ret != -EOPNOTSUPP)
3537 btrfs_abort_transaction(trans, root, ret);
3538 btrfs_end_transaction(trans, root);
3539 goto out;
3540 }
3541 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3542 destoff + len - last_dest_end);
3543 ret = clone_finish_inode_update(trans, inode, destoff + len,
3544 destoff, olen);
3545 }
3546
3547out:
3548 btrfs_free_path(path);
3549 vfree(buf);
3550 return ret;
3551}
3552
3553static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3554 u64 off, u64 olen, u64 destoff)
3555{
3556 struct inode *inode = file_inode(file);
3557 struct btrfs_root *root = BTRFS_I(inode)->root;
3558 struct fd src_file;
3559 struct inode *src;
3560 int ret;
3561 u64 len = olen;
3562 u64 bs = root->fs_info->sb->s_blocksize;
3563 int same_inode = 0;
3564
3565 /*
3566 * TODO:
3567 * - split compressed inline extents. annoying: we need to
3568 * decompress into destination's address_space (the file offset
3569 * may change, so source mapping won't do), then recompress (or
3570 * otherwise reinsert) a subrange.
3571 *
3572 * - split destination inode's inline extents. The inline extents can
3573 * be either compressed or non-compressed.
3574 */
3575
3576 /* the destination must be opened for writing */
3577 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3578 return -EINVAL;
3579
3580 if (btrfs_root_readonly(root))
3581 return -EROFS;
3582
3583 ret = mnt_want_write_file(file);
3584 if (ret)
3585 return ret;
3586
3587 src_file = fdget(srcfd);
3588 if (!src_file.file) {
3589 ret = -EBADF;
3590 goto out_drop_write;
3591 }
3592
3593 ret = -EXDEV;
3594 if (src_file.file->f_path.mnt != file->f_path.mnt)
3595 goto out_fput;
3596
3597 src = file_inode(src_file.file);
3598
3599 ret = -EINVAL;
3600 if (src == inode)
3601 same_inode = 1;
3602
3603 /* the src must be open for reading */
3604 if (!(src_file.file->f_mode & FMODE_READ))
3605 goto out_fput;
3606
3607 /* don't make the dst file partly checksummed */
3608 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3609 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3610 goto out_fput;
3611
3612 ret = -EISDIR;
3613 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3614 goto out_fput;
3615
3616 ret = -EXDEV;
3617 if (src->i_sb != inode->i_sb)
3618 goto out_fput;
3619
3620 if (!same_inode) {
3621 if (inode < src) {
3622 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3623 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3624 } else {
3625 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3626 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3627 }
3628 } else {
3629 mutex_lock(&src->i_mutex);
3630 }
3631
3632 /* determine range to clone */
3633 ret = -EINVAL;
3634 if (off + len > src->i_size || off + len < off)
3635 goto out_unlock;
3636 if (len == 0)
3637 olen = len = src->i_size - off;
3638 /* if we extend to eof, continue to block boundary */
3639 if (off + len == src->i_size)
3640 len = ALIGN(src->i_size, bs) - off;
3641
3642 if (len == 0) {
3643 ret = 0;
3644 goto out_unlock;
3645 }
3646
3647 /* verify the end result is block aligned */
3648 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3649 !IS_ALIGNED(destoff, bs))
3650 goto out_unlock;
3651
3652 /* verify if ranges are overlapped within the same file */
3653 if (same_inode) {
3654 if (destoff + len > off && destoff < off + len)
3655 goto out_unlock;
3656 }
3657
3658 if (destoff > inode->i_size) {
3659 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3660 if (ret)
3661 goto out_unlock;
3662 }
3663
3664 /*
3665 * Lock the target range too. Right after we replace the file extent
3666 * items in the fs tree (which now point to the cloned data), we might
3667 * have a worker replace them with extent items relative to a write
3668 * operation that was issued before this clone operation (i.e. confront
3669 * with inode.c:btrfs_finish_ordered_io).
3670 */
3671 if (same_inode) {
3672 u64 lock_start = min_t(u64, off, destoff);
3673 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3674
3675 lock_extent_range(src, lock_start, lock_len);
3676 } else {
3677 lock_extent_range(src, off, len);
3678 lock_extent_range(inode, destoff, len);
3679 }
3680
3681 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3682
3683 if (same_inode) {
3684 u64 lock_start = min_t(u64, off, destoff);
3685 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3686
3687 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3688 } else {
3689 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3690 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3691 destoff + len - 1);
3692 }
3693 /*
3694 * Truncate page cache pages so that future reads will see the cloned
3695 * data immediately and not the previous data.
3696 */
3697 truncate_inode_pages_range(&inode->i_data, destoff,
3698 PAGE_CACHE_ALIGN(destoff + len) - 1);
3699out_unlock:
3700 if (!same_inode) {
3701 if (inode < src) {
3702 mutex_unlock(&src->i_mutex);
3703 mutex_unlock(&inode->i_mutex);
3704 } else {
3705 mutex_unlock(&inode->i_mutex);
3706 mutex_unlock(&src->i_mutex);
3707 }
3708 } else {
3709 mutex_unlock(&src->i_mutex);
3710 }
3711out_fput:
3712 fdput(src_file);
3713out_drop_write:
3714 mnt_drop_write_file(file);
3715 return ret;
3716}
3717
3718static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3719{
3720 struct btrfs_ioctl_clone_range_args args;
3721
3722 if (copy_from_user(&args, argp, sizeof(args)))
3723 return -EFAULT;
3724 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3725 args.src_length, args.dest_offset);
3726}
3727
3728/*
3729 * there are many ways the trans_start and trans_end ioctls can lead
3730 * to deadlocks. They should only be used by applications that
3731 * basically own the machine, and have a very in depth understanding
3732 * of all the possible deadlocks and enospc problems.
3733 */
3734static long btrfs_ioctl_trans_start(struct file *file)
3735{
3736 struct inode *inode = file_inode(file);
3737 struct btrfs_root *root = BTRFS_I(inode)->root;
3738 struct btrfs_trans_handle *trans;
3739 int ret;
3740
3741 ret = -EPERM;
3742 if (!capable(CAP_SYS_ADMIN))
3743 goto out;
3744
3745 ret = -EINPROGRESS;
3746 if (file->private_data)
3747 goto out;
3748
3749 ret = -EROFS;
3750 if (btrfs_root_readonly(root))
3751 goto out;
3752
3753 ret = mnt_want_write_file(file);
3754 if (ret)
3755 goto out;
3756
3757 atomic_inc(&root->fs_info->open_ioctl_trans);
3758
3759 ret = -ENOMEM;
3760 trans = btrfs_start_ioctl_transaction(root);
3761 if (IS_ERR(trans))
3762 goto out_drop;
3763
3764 file->private_data = trans;
3765 return 0;
3766
3767out_drop:
3768 atomic_dec(&root->fs_info->open_ioctl_trans);
3769 mnt_drop_write_file(file);
3770out:
3771 return ret;
3772}
3773
3774static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3775{
3776 struct inode *inode = file_inode(file);
3777 struct btrfs_root *root = BTRFS_I(inode)->root;
3778 struct btrfs_root *new_root;
3779 struct btrfs_dir_item *di;
3780 struct btrfs_trans_handle *trans;
3781 struct btrfs_path *path;
3782 struct btrfs_key location;
3783 struct btrfs_disk_key disk_key;
3784 u64 objectid = 0;
3785 u64 dir_id;
3786 int ret;
3787
3788 if (!capable(CAP_SYS_ADMIN))
3789 return -EPERM;
3790
3791 ret = mnt_want_write_file(file);
3792 if (ret)
3793 return ret;
3794
3795 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3796 ret = -EFAULT;
3797 goto out;
3798 }
3799
3800 if (!objectid)
3801 objectid = BTRFS_FS_TREE_OBJECTID;
3802
3803 location.objectid = objectid;
3804 location.type = BTRFS_ROOT_ITEM_KEY;
3805 location.offset = (u64)-1;
3806
3807 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3808 if (IS_ERR(new_root)) {
3809 ret = PTR_ERR(new_root);
3810 goto out;
3811 }
3812
3813 path = btrfs_alloc_path();
3814 if (!path) {
3815 ret = -ENOMEM;
3816 goto out;
3817 }
3818 path->leave_spinning = 1;
3819
3820 trans = btrfs_start_transaction(root, 1);
3821 if (IS_ERR(trans)) {
3822 btrfs_free_path(path);
3823 ret = PTR_ERR(trans);
3824 goto out;
3825 }
3826
3827 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3828 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3829 dir_id, "default", 7, 1);
3830 if (IS_ERR_OR_NULL(di)) {
3831 btrfs_free_path(path);
3832 btrfs_end_transaction(trans, root);
3833 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3834 "item, this isn't going to work");
3835 ret = -ENOENT;
3836 goto out;
3837 }
3838
3839 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3840 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3841 btrfs_mark_buffer_dirty(path->nodes[0]);
3842 btrfs_free_path(path);
3843
3844 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3845 btrfs_end_transaction(trans, root);
3846out:
3847 mnt_drop_write_file(file);
3848 return ret;
3849}
3850
3851void btrfs_get_block_group_info(struct list_head *groups_list,
3852 struct btrfs_ioctl_space_info *space)
3853{
3854 struct btrfs_block_group_cache *block_group;
3855
3856 space->total_bytes = 0;
3857 space->used_bytes = 0;
3858 space->flags = 0;
3859 list_for_each_entry(block_group, groups_list, list) {
3860 space->flags = block_group->flags;
3861 space->total_bytes += block_group->key.offset;
3862 space->used_bytes +=
3863 btrfs_block_group_used(&block_group->item);
3864 }
3865}
3866
3867static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3868{
3869 struct btrfs_ioctl_space_args space_args;
3870 struct btrfs_ioctl_space_info space;
3871 struct btrfs_ioctl_space_info *dest;
3872 struct btrfs_ioctl_space_info *dest_orig;
3873 struct btrfs_ioctl_space_info __user *user_dest;
3874 struct btrfs_space_info *info;
3875 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3876 BTRFS_BLOCK_GROUP_SYSTEM,
3877 BTRFS_BLOCK_GROUP_METADATA,
3878 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3879 int num_types = 4;
3880 int alloc_size;
3881 int ret = 0;
3882 u64 slot_count = 0;
3883 int i, c;
3884
3885 if (copy_from_user(&space_args,
3886 (struct btrfs_ioctl_space_args __user *)arg,
3887 sizeof(space_args)))
3888 return -EFAULT;
3889
3890 for (i = 0; i < num_types; i++) {
3891 struct btrfs_space_info *tmp;
3892
3893 info = NULL;
3894 rcu_read_lock();
3895 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3896 list) {
3897 if (tmp->flags == types[i]) {
3898 info = tmp;
3899 break;
3900 }
3901 }
3902 rcu_read_unlock();
3903
3904 if (!info)
3905 continue;
3906
3907 down_read(&info->groups_sem);
3908 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3909 if (!list_empty(&info->block_groups[c]))
3910 slot_count++;
3911 }
3912 up_read(&info->groups_sem);
3913 }
3914
3915 /*
3916 * Global block reserve, exported as a space_info
3917 */
3918 slot_count++;
3919
3920 /* space_slots == 0 means they are asking for a count */
3921 if (space_args.space_slots == 0) {
3922 space_args.total_spaces = slot_count;
3923 goto out;
3924 }
3925
3926 slot_count = min_t(u64, space_args.space_slots, slot_count);
3927
3928 alloc_size = sizeof(*dest) * slot_count;
3929
3930 /* we generally have at most 6 or so space infos, one for each raid
3931 * level. So, a whole page should be more than enough for everyone
3932 */
3933 if (alloc_size > PAGE_CACHE_SIZE)
3934 return -ENOMEM;
3935
3936 space_args.total_spaces = 0;
3937 dest = kmalloc(alloc_size, GFP_NOFS);
3938 if (!dest)
3939 return -ENOMEM;
3940 dest_orig = dest;
3941
3942 /* now we have a buffer to copy into */
3943 for (i = 0; i < num_types; i++) {
3944 struct btrfs_space_info *tmp;
3945
3946 if (!slot_count)
3947 break;
3948
3949 info = NULL;
3950 rcu_read_lock();
3951 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3952 list) {
3953 if (tmp->flags == types[i]) {
3954 info = tmp;
3955 break;
3956 }
3957 }
3958 rcu_read_unlock();
3959
3960 if (!info)
3961 continue;
3962 down_read(&info->groups_sem);
3963 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3964 if (!list_empty(&info->block_groups[c])) {
3965 btrfs_get_block_group_info(
3966 &info->block_groups[c], &space);
3967 memcpy(dest, &space, sizeof(space));
3968 dest++;
3969 space_args.total_spaces++;
3970 slot_count--;
3971 }
3972 if (!slot_count)
3973 break;
3974 }
3975 up_read(&info->groups_sem);
3976 }
3977
3978 /*
3979 * Add global block reserve
3980 */
3981 if (slot_count) {
3982 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3983
3984 spin_lock(&block_rsv->lock);
3985 space.total_bytes = block_rsv->size;
3986 space.used_bytes = block_rsv->size - block_rsv->reserved;
3987 spin_unlock(&block_rsv->lock);
3988 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3989 memcpy(dest, &space, sizeof(space));
3990 space_args.total_spaces++;
3991 }
3992
3993 user_dest = (struct btrfs_ioctl_space_info __user *)
3994 (arg + sizeof(struct btrfs_ioctl_space_args));
3995
3996 if (copy_to_user(user_dest, dest_orig, alloc_size))
3997 ret = -EFAULT;
3998
3999 kfree(dest_orig);
4000out:
4001 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4002 ret = -EFAULT;
4003
4004 return ret;
4005}
4006
4007/*
4008 * there are many ways the trans_start and trans_end ioctls can lead
4009 * to deadlocks. They should only be used by applications that
4010 * basically own the machine, and have a very in depth understanding
4011 * of all the possible deadlocks and enospc problems.
4012 */
4013long btrfs_ioctl_trans_end(struct file *file)
4014{
4015 struct inode *inode = file_inode(file);
4016 struct btrfs_root *root = BTRFS_I(inode)->root;
4017 struct btrfs_trans_handle *trans;
4018
4019 trans = file->private_data;
4020 if (!trans)
4021 return -EINVAL;
4022 file->private_data = NULL;
4023
4024 btrfs_end_transaction(trans, root);
4025
4026 atomic_dec(&root->fs_info->open_ioctl_trans);
4027
4028 mnt_drop_write_file(file);
4029 return 0;
4030}
4031
4032static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4033 void __user *argp)
4034{
4035 struct btrfs_trans_handle *trans;
4036 u64 transid;
4037 int ret;
4038
4039 trans = btrfs_attach_transaction_barrier(root);
4040 if (IS_ERR(trans)) {
4041 if (PTR_ERR(trans) != -ENOENT)
4042 return PTR_ERR(trans);
4043
4044 /* No running transaction, don't bother */
4045 transid = root->fs_info->last_trans_committed;
4046 goto out;
4047 }
4048 transid = trans->transid;
4049 ret = btrfs_commit_transaction_async(trans, root, 0);
4050 if (ret) {
4051 btrfs_end_transaction(trans, root);
4052 return ret;
4053 }
4054out:
4055 if (argp)
4056 if (copy_to_user(argp, &transid, sizeof(transid)))
4057 return -EFAULT;
4058 return 0;
4059}
4060
4061static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4062 void __user *argp)
4063{
4064 u64 transid;
4065
4066 if (argp) {
4067 if (copy_from_user(&transid, argp, sizeof(transid)))
4068 return -EFAULT;
4069 } else {
4070 transid = 0; /* current trans */
4071 }
4072 return btrfs_wait_for_commit(root, transid);
4073}
4074
4075static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4076{
4077 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4078 struct btrfs_ioctl_scrub_args *sa;
4079 int ret;
4080
4081 if (!capable(CAP_SYS_ADMIN))
4082 return -EPERM;
4083
4084 sa = memdup_user(arg, sizeof(*sa));
4085 if (IS_ERR(sa))
4086 return PTR_ERR(sa);
4087
4088 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4089 ret = mnt_want_write_file(file);
4090 if (ret)
4091 goto out;
4092 }
4093
4094 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4095 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4096 0);
4097
4098 if (copy_to_user(arg, sa, sizeof(*sa)))
4099 ret = -EFAULT;
4100
4101 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4102 mnt_drop_write_file(file);
4103out:
4104 kfree(sa);
4105 return ret;
4106}
4107
4108static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4109{
4110 if (!capable(CAP_SYS_ADMIN))
4111 return -EPERM;
4112
4113 return btrfs_scrub_cancel(root->fs_info);
4114}
4115
4116static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4117 void __user *arg)
4118{
4119 struct btrfs_ioctl_scrub_args *sa;
4120 int ret;
4121
4122 if (!capable(CAP_SYS_ADMIN))
4123 return -EPERM;
4124
4125 sa = memdup_user(arg, sizeof(*sa));
4126 if (IS_ERR(sa))
4127 return PTR_ERR(sa);
4128
4129 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4130
4131 if (copy_to_user(arg, sa, sizeof(*sa)))
4132 ret = -EFAULT;
4133
4134 kfree(sa);
4135 return ret;
4136}
4137
4138static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4139 void __user *arg)
4140{
4141 struct btrfs_ioctl_get_dev_stats *sa;
4142 int ret;
4143
4144 sa = memdup_user(arg, sizeof(*sa));
4145 if (IS_ERR(sa))
4146 return PTR_ERR(sa);
4147
4148 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4149 kfree(sa);
4150 return -EPERM;
4151 }
4152
4153 ret = btrfs_get_dev_stats(root, sa);
4154
4155 if (copy_to_user(arg, sa, sizeof(*sa)))
4156 ret = -EFAULT;
4157
4158 kfree(sa);
4159 return ret;
4160}
4161
4162static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4163{
4164 struct btrfs_ioctl_dev_replace_args *p;
4165 int ret;
4166
4167 if (!capable(CAP_SYS_ADMIN))
4168 return -EPERM;
4169
4170 p = memdup_user(arg, sizeof(*p));
4171 if (IS_ERR(p))
4172 return PTR_ERR(p);
4173
4174 switch (p->cmd) {
4175 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4176 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4177 ret = -EROFS;
4178 goto out;
4179 }
4180 if (atomic_xchg(
4181 &root->fs_info->mutually_exclusive_operation_running,
4182 1)) {
4183 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4184 } else {
4185 ret = btrfs_dev_replace_start(root, p);
4186 atomic_set(
4187 &root->fs_info->mutually_exclusive_operation_running,
4188 0);
4189 }
4190 break;
4191 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4192 btrfs_dev_replace_status(root->fs_info, p);
4193 ret = 0;
4194 break;
4195 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4196 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4197 break;
4198 default:
4199 ret = -EINVAL;
4200 break;
4201 }
4202
4203 if (copy_to_user(arg, p, sizeof(*p)))
4204 ret = -EFAULT;
4205out:
4206 kfree(p);
4207 return ret;
4208}
4209
4210static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4211{
4212 int ret = 0;
4213 int i;
4214 u64 rel_ptr;
4215 int size;
4216 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4217 struct inode_fs_paths *ipath = NULL;
4218 struct btrfs_path *path;
4219
4220 if (!capable(CAP_DAC_READ_SEARCH))
4221 return -EPERM;
4222
4223 path = btrfs_alloc_path();
4224 if (!path) {
4225 ret = -ENOMEM;
4226 goto out;
4227 }
4228
4229 ipa = memdup_user(arg, sizeof(*ipa));
4230 if (IS_ERR(ipa)) {
4231 ret = PTR_ERR(ipa);
4232 ipa = NULL;
4233 goto out;
4234 }
4235
4236 size = min_t(u32, ipa->size, 4096);
4237 ipath = init_ipath(size, root, path);
4238 if (IS_ERR(ipath)) {
4239 ret = PTR_ERR(ipath);
4240 ipath = NULL;
4241 goto out;
4242 }
4243
4244 ret = paths_from_inode(ipa->inum, ipath);
4245 if (ret < 0)
4246 goto out;
4247
4248 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4249 rel_ptr = ipath->fspath->val[i] -
4250 (u64)(unsigned long)ipath->fspath->val;
4251 ipath->fspath->val[i] = rel_ptr;
4252 }
4253
4254 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4255 (void *)(unsigned long)ipath->fspath, size);
4256 if (ret) {
4257 ret = -EFAULT;
4258 goto out;
4259 }
4260
4261out:
4262 btrfs_free_path(path);
4263 free_ipath(ipath);
4264 kfree(ipa);
4265
4266 return ret;
4267}
4268
4269static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4270{
4271 struct btrfs_data_container *inodes = ctx;
4272 const size_t c = 3 * sizeof(u64);
4273
4274 if (inodes->bytes_left >= c) {
4275 inodes->bytes_left -= c;
4276 inodes->val[inodes->elem_cnt] = inum;
4277 inodes->val[inodes->elem_cnt + 1] = offset;
4278 inodes->val[inodes->elem_cnt + 2] = root;
4279 inodes->elem_cnt += 3;
4280 } else {
4281 inodes->bytes_missing += c - inodes->bytes_left;
4282 inodes->bytes_left = 0;
4283 inodes->elem_missed += 3;
4284 }
4285
4286 return 0;
4287}
4288
4289static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4290 void __user *arg)
4291{
4292 int ret = 0;
4293 int size;
4294 struct btrfs_ioctl_logical_ino_args *loi;
4295 struct btrfs_data_container *inodes = NULL;
4296 struct btrfs_path *path = NULL;
4297
4298 if (!capable(CAP_SYS_ADMIN))
4299 return -EPERM;
4300
4301 loi = memdup_user(arg, sizeof(*loi));
4302 if (IS_ERR(loi)) {
4303 ret = PTR_ERR(loi);
4304 loi = NULL;
4305 goto out;
4306 }
4307
4308 path = btrfs_alloc_path();
4309 if (!path) {
4310 ret = -ENOMEM;
4311 goto out;
4312 }
4313
4314 size = min_t(u32, loi->size, 64 * 1024);
4315 inodes = init_data_container(size);
4316 if (IS_ERR(inodes)) {
4317 ret = PTR_ERR(inodes);
4318 inodes = NULL;
4319 goto out;
4320 }
4321
4322 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4323 build_ino_list, inodes);
4324 if (ret == -EINVAL)
4325 ret = -ENOENT;
4326 if (ret < 0)
4327 goto out;
4328
4329 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4330 (void *)(unsigned long)inodes, size);
4331 if (ret)
4332 ret = -EFAULT;
4333
4334out:
4335 btrfs_free_path(path);
4336 vfree(inodes);
4337 kfree(loi);
4338
4339 return ret;
4340}
4341
4342void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4343 struct btrfs_ioctl_balance_args *bargs)
4344{
4345 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4346
4347 bargs->flags = bctl->flags;
4348
4349 if (atomic_read(&fs_info->balance_running))
4350 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4351 if (atomic_read(&fs_info->balance_pause_req))
4352 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4353 if (atomic_read(&fs_info->balance_cancel_req))
4354 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4355
4356 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4357 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4358 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4359
4360 if (lock) {
4361 spin_lock(&fs_info->balance_lock);
4362 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4363 spin_unlock(&fs_info->balance_lock);
4364 } else {
4365 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4366 }
4367}
4368
4369static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4370{
4371 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4372 struct btrfs_fs_info *fs_info = root->fs_info;
4373 struct btrfs_ioctl_balance_args *bargs;
4374 struct btrfs_balance_control *bctl;
4375 bool need_unlock; /* for mut. excl. ops lock */
4376 int ret;
4377
4378 if (!capable(CAP_SYS_ADMIN))
4379 return -EPERM;
4380
4381 ret = mnt_want_write_file(file);
4382 if (ret)
4383 return ret;
4384
4385again:
4386 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4387 mutex_lock(&fs_info->volume_mutex);
4388 mutex_lock(&fs_info->balance_mutex);
4389 need_unlock = true;
4390 goto locked;
4391 }
4392
4393 /*
4394 * mut. excl. ops lock is locked. Three possibilites:
4395 * (1) some other op is running
4396 * (2) balance is running
4397 * (3) balance is paused -- special case (think resume)
4398 */
4399 mutex_lock(&fs_info->balance_mutex);
4400 if (fs_info->balance_ctl) {
4401 /* this is either (2) or (3) */
4402 if (!atomic_read(&fs_info->balance_running)) {
4403 mutex_unlock(&fs_info->balance_mutex);
4404 if (!mutex_trylock(&fs_info->volume_mutex))
4405 goto again;
4406 mutex_lock(&fs_info->balance_mutex);
4407
4408 if (fs_info->balance_ctl &&
4409 !atomic_read(&fs_info->balance_running)) {
4410 /* this is (3) */
4411 need_unlock = false;
4412 goto locked;
4413 }
4414
4415 mutex_unlock(&fs_info->balance_mutex);
4416 mutex_unlock(&fs_info->volume_mutex);
4417 goto again;
4418 } else {
4419 /* this is (2) */
4420 mutex_unlock(&fs_info->balance_mutex);
4421 ret = -EINPROGRESS;
4422 goto out;
4423 }
4424 } else {
4425 /* this is (1) */
4426 mutex_unlock(&fs_info->balance_mutex);
4427 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4428 goto out;
4429 }
4430
4431locked:
4432 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4433
4434 if (arg) {
4435 bargs = memdup_user(arg, sizeof(*bargs));
4436 if (IS_ERR(bargs)) {
4437 ret = PTR_ERR(bargs);
4438 goto out_unlock;
4439 }
4440
4441 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4442 if (!fs_info->balance_ctl) {
4443 ret = -ENOTCONN;
4444 goto out_bargs;
4445 }
4446
4447 bctl = fs_info->balance_ctl;
4448 spin_lock(&fs_info->balance_lock);
4449 bctl->flags |= BTRFS_BALANCE_RESUME;
4450 spin_unlock(&fs_info->balance_lock);
4451
4452 goto do_balance;
4453 }
4454 } else {
4455 bargs = NULL;
4456 }
4457
4458 if (fs_info->balance_ctl) {
4459 ret = -EINPROGRESS;
4460 goto out_bargs;
4461 }
4462
4463 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4464 if (!bctl) {
4465 ret = -ENOMEM;
4466 goto out_bargs;
4467 }
4468
4469 bctl->fs_info = fs_info;
4470 if (arg) {
4471 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4472 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4473 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4474
4475 bctl->flags = bargs->flags;
4476 } else {
4477 /* balance everything - no filters */
4478 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4479 }
4480
4481do_balance:
4482 /*
4483 * Ownership of bctl and mutually_exclusive_operation_running
4484 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4485 * or, if restriper was paused all the way until unmount, in
4486 * free_fs_info. mutually_exclusive_operation_running is
4487 * cleared in __cancel_balance.
4488 */
4489 need_unlock = false;
4490
4491 ret = btrfs_balance(bctl, bargs);
4492
4493 if (arg) {
4494 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4495 ret = -EFAULT;
4496 }
4497
4498out_bargs:
4499 kfree(bargs);
4500out_unlock:
4501 mutex_unlock(&fs_info->balance_mutex);
4502 mutex_unlock(&fs_info->volume_mutex);
4503 if (need_unlock)
4504 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4505out:
4506 mnt_drop_write_file(file);
4507 return ret;
4508}
4509
4510static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4511{
4512 if (!capable(CAP_SYS_ADMIN))
4513 return -EPERM;
4514
4515 switch (cmd) {
4516 case BTRFS_BALANCE_CTL_PAUSE:
4517 return btrfs_pause_balance(root->fs_info);
4518 case BTRFS_BALANCE_CTL_CANCEL:
4519 return btrfs_cancel_balance(root->fs_info);
4520 }
4521
4522 return -EINVAL;
4523}
4524
4525static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4526 void __user *arg)
4527{
4528 struct btrfs_fs_info *fs_info = root->fs_info;
4529 struct btrfs_ioctl_balance_args *bargs;
4530 int ret = 0;
4531
4532 if (!capable(CAP_SYS_ADMIN))
4533 return -EPERM;
4534
4535 mutex_lock(&fs_info->balance_mutex);
4536 if (!fs_info->balance_ctl) {
4537 ret = -ENOTCONN;
4538 goto out;
4539 }
4540
4541 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4542 if (!bargs) {
4543 ret = -ENOMEM;
4544 goto out;
4545 }
4546
4547 update_ioctl_balance_args(fs_info, 1, bargs);
4548
4549 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4550 ret = -EFAULT;
4551
4552 kfree(bargs);
4553out:
4554 mutex_unlock(&fs_info->balance_mutex);
4555 return ret;
4556}
4557
4558static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4559{
4560 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4561 struct btrfs_ioctl_quota_ctl_args *sa;
4562 struct btrfs_trans_handle *trans = NULL;
4563 int ret;
4564 int err;
4565
4566 if (!capable(CAP_SYS_ADMIN))
4567 return -EPERM;
4568
4569 ret = mnt_want_write_file(file);
4570 if (ret)
4571 return ret;
4572
4573 sa = memdup_user(arg, sizeof(*sa));
4574 if (IS_ERR(sa)) {
4575 ret = PTR_ERR(sa);
4576 goto drop_write;
4577 }
4578
4579 down_write(&root->fs_info->subvol_sem);
4580 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4581 if (IS_ERR(trans)) {
4582 ret = PTR_ERR(trans);
4583 goto out;
4584 }
4585
4586 switch (sa->cmd) {
4587 case BTRFS_QUOTA_CTL_ENABLE:
4588 ret = btrfs_quota_enable(trans, root->fs_info);
4589 break;
4590 case BTRFS_QUOTA_CTL_DISABLE:
4591 ret = btrfs_quota_disable(trans, root->fs_info);
4592 break;
4593 default:
4594 ret = -EINVAL;
4595 break;
4596 }
4597
4598 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4599 if (err && !ret)
4600 ret = err;
4601out:
4602 kfree(sa);
4603 up_write(&root->fs_info->subvol_sem);
4604drop_write:
4605 mnt_drop_write_file(file);
4606 return ret;
4607}
4608
4609static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4610{
4611 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4612 struct btrfs_ioctl_qgroup_assign_args *sa;
4613 struct btrfs_trans_handle *trans;
4614 int ret;
4615 int err;
4616
4617 if (!capable(CAP_SYS_ADMIN))
4618 return -EPERM;
4619
4620 ret = mnt_want_write_file(file);
4621 if (ret)
4622 return ret;
4623
4624 sa = memdup_user(arg, sizeof(*sa));
4625 if (IS_ERR(sa)) {
4626 ret = PTR_ERR(sa);
4627 goto drop_write;
4628 }
4629
4630 trans = btrfs_join_transaction(root);
4631 if (IS_ERR(trans)) {
4632 ret = PTR_ERR(trans);
4633 goto out;
4634 }
4635
4636 /* FIXME: check if the IDs really exist */
4637 if (sa->assign) {
4638 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4639 sa->src, sa->dst);
4640 } else {
4641 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4642 sa->src, sa->dst);
4643 }
4644
4645 /* update qgroup status and info */
4646 err = btrfs_run_qgroups(trans, root->fs_info);
4647 if (err < 0)
4648 btrfs_error(root->fs_info, ret,
4649 "failed to update qgroup status and info\n");
4650 err = btrfs_end_transaction(trans, root);
4651 if (err && !ret)
4652 ret = err;
4653
4654out:
4655 kfree(sa);
4656drop_write:
4657 mnt_drop_write_file(file);
4658 return ret;
4659}
4660
4661static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4662{
4663 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4664 struct btrfs_ioctl_qgroup_create_args *sa;
4665 struct btrfs_trans_handle *trans;
4666 int ret;
4667 int err;
4668
4669 if (!capable(CAP_SYS_ADMIN))
4670 return -EPERM;
4671
4672 ret = mnt_want_write_file(file);
4673 if (ret)
4674 return ret;
4675
4676 sa = memdup_user(arg, sizeof(*sa));
4677 if (IS_ERR(sa)) {
4678 ret = PTR_ERR(sa);
4679 goto drop_write;
4680 }
4681
4682 if (!sa->qgroupid) {
4683 ret = -EINVAL;
4684 goto out;
4685 }
4686
4687 trans = btrfs_join_transaction(root);
4688 if (IS_ERR(trans)) {
4689 ret = PTR_ERR(trans);
4690 goto out;
4691 }
4692
4693 /* FIXME: check if the IDs really exist */
4694 if (sa->create) {
4695 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
4696 } else {
4697 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4698 }
4699
4700 err = btrfs_end_transaction(trans, root);
4701 if (err && !ret)
4702 ret = err;
4703
4704out:
4705 kfree(sa);
4706drop_write:
4707 mnt_drop_write_file(file);
4708 return ret;
4709}
4710
4711static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4712{
4713 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4714 struct btrfs_ioctl_qgroup_limit_args *sa;
4715 struct btrfs_trans_handle *trans;
4716 int ret;
4717 int err;
4718 u64 qgroupid;
4719
4720 if (!capable(CAP_SYS_ADMIN))
4721 return -EPERM;
4722
4723 ret = mnt_want_write_file(file);
4724 if (ret)
4725 return ret;
4726
4727 sa = memdup_user(arg, sizeof(*sa));
4728 if (IS_ERR(sa)) {
4729 ret = PTR_ERR(sa);
4730 goto drop_write;
4731 }
4732
4733 trans = btrfs_join_transaction(root);
4734 if (IS_ERR(trans)) {
4735 ret = PTR_ERR(trans);
4736 goto out;
4737 }
4738
4739 qgroupid = sa->qgroupid;
4740 if (!qgroupid) {
4741 /* take the current subvol as qgroup */
4742 qgroupid = root->root_key.objectid;
4743 }
4744
4745 /* FIXME: check if the IDs really exist */
4746 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4747
4748 err = btrfs_end_transaction(trans, root);
4749 if (err && !ret)
4750 ret = err;
4751
4752out:
4753 kfree(sa);
4754drop_write:
4755 mnt_drop_write_file(file);
4756 return ret;
4757}
4758
4759static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4760{
4761 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4762 struct btrfs_ioctl_quota_rescan_args *qsa;
4763 int ret;
4764
4765 if (!capable(CAP_SYS_ADMIN))
4766 return -EPERM;
4767
4768 ret = mnt_want_write_file(file);
4769 if (ret)
4770 return ret;
4771
4772 qsa = memdup_user(arg, sizeof(*qsa));
4773 if (IS_ERR(qsa)) {
4774 ret = PTR_ERR(qsa);
4775 goto drop_write;
4776 }
4777
4778 if (qsa->flags) {
4779 ret = -EINVAL;
4780 goto out;
4781 }
4782
4783 ret = btrfs_qgroup_rescan(root->fs_info);
4784
4785out:
4786 kfree(qsa);
4787drop_write:
4788 mnt_drop_write_file(file);
4789 return ret;
4790}
4791
4792static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4793{
4794 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4795 struct btrfs_ioctl_quota_rescan_args *qsa;
4796 int ret = 0;
4797
4798 if (!capable(CAP_SYS_ADMIN))
4799 return -EPERM;
4800
4801 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4802 if (!qsa)
4803 return -ENOMEM;
4804
4805 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4806 qsa->flags = 1;
4807 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4808 }
4809
4810 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4811 ret = -EFAULT;
4812
4813 kfree(qsa);
4814 return ret;
4815}
4816
4817static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4818{
4819 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4820
4821 if (!capable(CAP_SYS_ADMIN))
4822 return -EPERM;
4823
4824 return btrfs_qgroup_wait_for_completion(root->fs_info);
4825}
4826
4827static long _btrfs_ioctl_set_received_subvol(struct file *file,
4828 struct btrfs_ioctl_received_subvol_args *sa)
4829{
4830 struct inode *inode = file_inode(file);
4831 struct btrfs_root *root = BTRFS_I(inode)->root;
4832 struct btrfs_root_item *root_item = &root->root_item;
4833 struct btrfs_trans_handle *trans;
4834 struct timespec ct = CURRENT_TIME;
4835 int ret = 0;
4836 int received_uuid_changed;
4837
4838 if (!inode_owner_or_capable(inode))
4839 return -EPERM;
4840
4841 ret = mnt_want_write_file(file);
4842 if (ret < 0)
4843 return ret;
4844
4845 down_write(&root->fs_info->subvol_sem);
4846
4847 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4848 ret = -EINVAL;
4849 goto out;
4850 }
4851
4852 if (btrfs_root_readonly(root)) {
4853 ret = -EROFS;
4854 goto out;
4855 }
4856
4857 /*
4858 * 1 - root item
4859 * 2 - uuid items (received uuid + subvol uuid)
4860 */
4861 trans = btrfs_start_transaction(root, 3);
4862 if (IS_ERR(trans)) {
4863 ret = PTR_ERR(trans);
4864 trans = NULL;
4865 goto out;
4866 }
4867
4868 sa->rtransid = trans->transid;
4869 sa->rtime.sec = ct.tv_sec;
4870 sa->rtime.nsec = ct.tv_nsec;
4871
4872 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4873 BTRFS_UUID_SIZE);
4874 if (received_uuid_changed &&
4875 !btrfs_is_empty_uuid(root_item->received_uuid))
4876 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4877 root_item->received_uuid,
4878 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4879 root->root_key.objectid);
4880 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4881 btrfs_set_root_stransid(root_item, sa->stransid);
4882 btrfs_set_root_rtransid(root_item, sa->rtransid);
4883 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4884 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4885 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4886 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4887
4888 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4889 &root->root_key, &root->root_item);
4890 if (ret < 0) {
4891 btrfs_end_transaction(trans, root);
4892 goto out;
4893 }
4894 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4895 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4896 sa->uuid,
4897 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4898 root->root_key.objectid);
4899 if (ret < 0 && ret != -EEXIST) {
4900 btrfs_abort_transaction(trans, root, ret);
4901 goto out;
4902 }
4903 }
4904 ret = btrfs_commit_transaction(trans, root);
4905 if (ret < 0) {
4906 btrfs_abort_transaction(trans, root, ret);
4907 goto out;
4908 }
4909
4910out:
4911 up_write(&root->fs_info->subvol_sem);
4912 mnt_drop_write_file(file);
4913 return ret;
4914}
4915
4916#ifdef CONFIG_64BIT
4917static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4918 void __user *arg)
4919{
4920 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4921 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4922 int ret = 0;
4923
4924 args32 = memdup_user(arg, sizeof(*args32));
4925 if (IS_ERR(args32)) {
4926 ret = PTR_ERR(args32);
4927 args32 = NULL;
4928 goto out;
4929 }
4930
4931 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4932 if (!args64) {
4933 ret = -ENOMEM;
4934 goto out;
4935 }
4936
4937 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4938 args64->stransid = args32->stransid;
4939 args64->rtransid = args32->rtransid;
4940 args64->stime.sec = args32->stime.sec;
4941 args64->stime.nsec = args32->stime.nsec;
4942 args64->rtime.sec = args32->rtime.sec;
4943 args64->rtime.nsec = args32->rtime.nsec;
4944 args64->flags = args32->flags;
4945
4946 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4947 if (ret)
4948 goto out;
4949
4950 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4951 args32->stransid = args64->stransid;
4952 args32->rtransid = args64->rtransid;
4953 args32->stime.sec = args64->stime.sec;
4954 args32->stime.nsec = args64->stime.nsec;
4955 args32->rtime.sec = args64->rtime.sec;
4956 args32->rtime.nsec = args64->rtime.nsec;
4957 args32->flags = args64->flags;
4958
4959 ret = copy_to_user(arg, args32, sizeof(*args32));
4960 if (ret)
4961 ret = -EFAULT;
4962
4963out:
4964 kfree(args32);
4965 kfree(args64);
4966 return ret;
4967}
4968#endif
4969
4970static long btrfs_ioctl_set_received_subvol(struct file *file,
4971 void __user *arg)
4972{
4973 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4974 int ret = 0;
4975
4976 sa = memdup_user(arg, sizeof(*sa));
4977 if (IS_ERR(sa)) {
4978 ret = PTR_ERR(sa);
4979 sa = NULL;
4980 goto out;
4981 }
4982
4983 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4984
4985 if (ret)
4986 goto out;
4987
4988 ret = copy_to_user(arg, sa, sizeof(*sa));
4989 if (ret)
4990 ret = -EFAULT;
4991
4992out:
4993 kfree(sa);
4994 return ret;
4995}
4996
4997static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4998{
4999 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5000 size_t len;
5001 int ret;
5002 char label[BTRFS_LABEL_SIZE];
5003
5004 spin_lock(&root->fs_info->super_lock);
5005 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5006 spin_unlock(&root->fs_info->super_lock);
5007
5008 len = strnlen(label, BTRFS_LABEL_SIZE);
5009
5010 if (len == BTRFS_LABEL_SIZE) {
5011 btrfs_warn(root->fs_info,
5012 "label is too long, return the first %zu bytes", --len);
5013 }
5014
5015 ret = copy_to_user(arg, label, len);
5016
5017 return ret ? -EFAULT : 0;
5018}
5019
5020static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5021{
5022 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5023 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5024 struct btrfs_trans_handle *trans;
5025 char label[BTRFS_LABEL_SIZE];
5026 int ret;
5027
5028 if (!capable(CAP_SYS_ADMIN))
5029 return -EPERM;
5030
5031 if (copy_from_user(label, arg, sizeof(label)))
5032 return -EFAULT;
5033
5034 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5035 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5036 BTRFS_LABEL_SIZE - 1);
5037 return -EINVAL;
5038 }
5039
5040 ret = mnt_want_write_file(file);
5041 if (ret)
5042 return ret;
5043
5044 trans = btrfs_start_transaction(root, 0);
5045 if (IS_ERR(trans)) {
5046 ret = PTR_ERR(trans);
5047 goto out_unlock;
5048 }
5049
5050 spin_lock(&root->fs_info->super_lock);
5051 strcpy(super_block->label, label);
5052 spin_unlock(&root->fs_info->super_lock);
5053 ret = btrfs_commit_transaction(trans, root);
5054
5055out_unlock:
5056 mnt_drop_write_file(file);
5057 return ret;
5058}
5059
5060#define INIT_FEATURE_FLAGS(suffix) \
5061 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5062 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5063 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5064
5065static int btrfs_ioctl_get_supported_features(struct file *file,
5066 void __user *arg)
5067{
5068 static struct btrfs_ioctl_feature_flags features[3] = {
5069 INIT_FEATURE_FLAGS(SUPP),
5070 INIT_FEATURE_FLAGS(SAFE_SET),
5071 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5072 };
5073
5074 if (copy_to_user(arg, &features, sizeof(features)))
5075 return -EFAULT;
5076
5077 return 0;
5078}
5079
5080static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5081{
5082 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5083 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5084 struct btrfs_ioctl_feature_flags features;
5085
5086 features.compat_flags = btrfs_super_compat_flags(super_block);
5087 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5088 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5089
5090 if (copy_to_user(arg, &features, sizeof(features)))
5091 return -EFAULT;
5092
5093 return 0;
5094}
5095
5096static int check_feature_bits(struct btrfs_root *root,
5097 enum btrfs_feature_set set,
5098 u64 change_mask, u64 flags, u64 supported_flags,
5099 u64 safe_set, u64 safe_clear)
5100{
5101 const char *type = btrfs_feature_set_names[set];
5102 char *names;
5103 u64 disallowed, unsupported;
5104 u64 set_mask = flags & change_mask;
5105 u64 clear_mask = ~flags & change_mask;
5106
5107 unsupported = set_mask & ~supported_flags;
5108 if (unsupported) {
5109 names = btrfs_printable_features(set, unsupported);
5110 if (names) {
5111 btrfs_warn(root->fs_info,
5112 "this kernel does not support the %s feature bit%s",
5113 names, strchr(names, ',') ? "s" : "");
5114 kfree(names);
5115 } else
5116 btrfs_warn(root->fs_info,
5117 "this kernel does not support %s bits 0x%llx",
5118 type, unsupported);
5119 return -EOPNOTSUPP;
5120 }
5121
5122 disallowed = set_mask & ~safe_set;
5123 if (disallowed) {
5124 names = btrfs_printable_features(set, disallowed);
5125 if (names) {
5126 btrfs_warn(root->fs_info,
5127 "can't set the %s feature bit%s while mounted",
5128 names, strchr(names, ',') ? "s" : "");
5129 kfree(names);
5130 } else
5131 btrfs_warn(root->fs_info,
5132 "can't set %s bits 0x%llx while mounted",
5133 type, disallowed);
5134 return -EPERM;
5135 }
5136
5137 disallowed = clear_mask & ~safe_clear;
5138 if (disallowed) {
5139 names = btrfs_printable_features(set, disallowed);
5140 if (names) {
5141 btrfs_warn(root->fs_info,
5142 "can't clear the %s feature bit%s while mounted",
5143 names, strchr(names, ',') ? "s" : "");
5144 kfree(names);
5145 } else
5146 btrfs_warn(root->fs_info,
5147 "can't clear %s bits 0x%llx while mounted",
5148 type, disallowed);
5149 return -EPERM;
5150 }
5151
5152 return 0;
5153}
5154
5155#define check_feature(root, change_mask, flags, mask_base) \
5156check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5157 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5158 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5159 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5160
5161static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5162{
5163 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5164 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5165 struct btrfs_ioctl_feature_flags flags[2];
5166 struct btrfs_trans_handle *trans;
5167 u64 newflags;
5168 int ret;
5169
5170 if (!capable(CAP_SYS_ADMIN))
5171 return -EPERM;
5172
5173 if (copy_from_user(flags, arg, sizeof(flags)))
5174 return -EFAULT;
5175
5176 /* Nothing to do */
5177 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5178 !flags[0].incompat_flags)
5179 return 0;
5180
5181 ret = check_feature(root, flags[0].compat_flags,
5182 flags[1].compat_flags, COMPAT);
5183 if (ret)
5184 return ret;
5185
5186 ret = check_feature(root, flags[0].compat_ro_flags,
5187 flags[1].compat_ro_flags, COMPAT_RO);
5188 if (ret)
5189 return ret;
5190
5191 ret = check_feature(root, flags[0].incompat_flags,
5192 flags[1].incompat_flags, INCOMPAT);
5193 if (ret)
5194 return ret;
5195
5196 trans = btrfs_start_transaction(root, 0);
5197 if (IS_ERR(trans))
5198 return PTR_ERR(trans);
5199
5200 spin_lock(&root->fs_info->super_lock);
5201 newflags = btrfs_super_compat_flags(super_block);
5202 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5203 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5204 btrfs_set_super_compat_flags(super_block, newflags);
5205
5206 newflags = btrfs_super_compat_ro_flags(super_block);
5207 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5208 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5209 btrfs_set_super_compat_ro_flags(super_block, newflags);
5210
5211 newflags = btrfs_super_incompat_flags(super_block);
5212 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5213 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5214 btrfs_set_super_incompat_flags(super_block, newflags);
5215 spin_unlock(&root->fs_info->super_lock);
5216
5217 return btrfs_commit_transaction(trans, root);
5218}
5219
5220long btrfs_ioctl(struct file *file, unsigned int
5221 cmd, unsigned long arg)
5222{
5223 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5224 void __user *argp = (void __user *)arg;
5225
5226 switch (cmd) {
5227 case FS_IOC_GETFLAGS:
5228 return btrfs_ioctl_getflags(file, argp);
5229 case FS_IOC_SETFLAGS:
5230 return btrfs_ioctl_setflags(file, argp);
5231 case FS_IOC_GETVERSION:
5232 return btrfs_ioctl_getversion(file, argp);
5233 case FITRIM:
5234 return btrfs_ioctl_fitrim(file, argp);
5235 case BTRFS_IOC_SNAP_CREATE:
5236 return btrfs_ioctl_snap_create(file, argp, 0);
5237 case BTRFS_IOC_SNAP_CREATE_V2:
5238 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5239 case BTRFS_IOC_SUBVOL_CREATE:
5240 return btrfs_ioctl_snap_create(file, argp, 1);
5241 case BTRFS_IOC_SUBVOL_CREATE_V2:
5242 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5243 case BTRFS_IOC_SNAP_DESTROY:
5244 return btrfs_ioctl_snap_destroy(file, argp);
5245 case BTRFS_IOC_SUBVOL_GETFLAGS:
5246 return btrfs_ioctl_subvol_getflags(file, argp);
5247 case BTRFS_IOC_SUBVOL_SETFLAGS:
5248 return btrfs_ioctl_subvol_setflags(file, argp);
5249 case BTRFS_IOC_DEFAULT_SUBVOL:
5250 return btrfs_ioctl_default_subvol(file, argp);
5251 case BTRFS_IOC_DEFRAG:
5252 return btrfs_ioctl_defrag(file, NULL);
5253 case BTRFS_IOC_DEFRAG_RANGE:
5254 return btrfs_ioctl_defrag(file, argp);
5255 case BTRFS_IOC_RESIZE:
5256 return btrfs_ioctl_resize(file, argp);
5257 case BTRFS_IOC_ADD_DEV:
5258 return btrfs_ioctl_add_dev(root, argp);
5259 case BTRFS_IOC_RM_DEV:
5260 return btrfs_ioctl_rm_dev(file, argp);
5261 case BTRFS_IOC_FS_INFO:
5262 return btrfs_ioctl_fs_info(root, argp);
5263 case BTRFS_IOC_DEV_INFO:
5264 return btrfs_ioctl_dev_info(root, argp);
5265 case BTRFS_IOC_BALANCE:
5266 return btrfs_ioctl_balance(file, NULL);
5267 case BTRFS_IOC_CLONE:
5268 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5269 case BTRFS_IOC_CLONE_RANGE:
5270 return btrfs_ioctl_clone_range(file, argp);
5271 case BTRFS_IOC_TRANS_START:
5272 return btrfs_ioctl_trans_start(file);
5273 case BTRFS_IOC_TRANS_END:
5274 return btrfs_ioctl_trans_end(file);
5275 case BTRFS_IOC_TREE_SEARCH:
5276 return btrfs_ioctl_tree_search(file, argp);
5277 case BTRFS_IOC_TREE_SEARCH_V2:
5278 return btrfs_ioctl_tree_search_v2(file, argp);
5279 case BTRFS_IOC_INO_LOOKUP:
5280 return btrfs_ioctl_ino_lookup(file, argp);
5281 case BTRFS_IOC_INO_PATHS:
5282 return btrfs_ioctl_ino_to_path(root, argp);
5283 case BTRFS_IOC_LOGICAL_INO:
5284 return btrfs_ioctl_logical_to_ino(root, argp);
5285 case BTRFS_IOC_SPACE_INFO:
5286 return btrfs_ioctl_space_info(root, argp);
5287 case BTRFS_IOC_SYNC: {
5288 int ret;
5289
5290 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5291 if (ret)
5292 return ret;
5293 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
5294 /*
5295 * The transaction thread may want to do more work,
5296 * namely it pokes the cleaner ktread that will start
5297 * processing uncleaned subvols.
5298 */
5299 wake_up_process(root->fs_info->transaction_kthread);
5300 return ret;
5301 }
5302 case BTRFS_IOC_START_SYNC:
5303 return btrfs_ioctl_start_sync(root, argp);
5304 case BTRFS_IOC_WAIT_SYNC:
5305 return btrfs_ioctl_wait_sync(root, argp);
5306 case BTRFS_IOC_SCRUB:
5307 return btrfs_ioctl_scrub(file, argp);
5308 case BTRFS_IOC_SCRUB_CANCEL:
5309 return btrfs_ioctl_scrub_cancel(root, argp);
5310 case BTRFS_IOC_SCRUB_PROGRESS:
5311 return btrfs_ioctl_scrub_progress(root, argp);
5312 case BTRFS_IOC_BALANCE_V2:
5313 return btrfs_ioctl_balance(file, argp);
5314 case BTRFS_IOC_BALANCE_CTL:
5315 return btrfs_ioctl_balance_ctl(root, arg);
5316 case BTRFS_IOC_BALANCE_PROGRESS:
5317 return btrfs_ioctl_balance_progress(root, argp);
5318 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5319 return btrfs_ioctl_set_received_subvol(file, argp);
5320#ifdef CONFIG_64BIT
5321 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5322 return btrfs_ioctl_set_received_subvol_32(file, argp);
5323#endif
5324 case BTRFS_IOC_SEND:
5325 return btrfs_ioctl_send(file, argp);
5326 case BTRFS_IOC_GET_DEV_STATS:
5327 return btrfs_ioctl_get_dev_stats(root, argp);
5328 case BTRFS_IOC_QUOTA_CTL:
5329 return btrfs_ioctl_quota_ctl(file, argp);
5330 case BTRFS_IOC_QGROUP_ASSIGN:
5331 return btrfs_ioctl_qgroup_assign(file, argp);
5332 case BTRFS_IOC_QGROUP_CREATE:
5333 return btrfs_ioctl_qgroup_create(file, argp);
5334 case BTRFS_IOC_QGROUP_LIMIT:
5335 return btrfs_ioctl_qgroup_limit(file, argp);
5336 case BTRFS_IOC_QUOTA_RESCAN:
5337 return btrfs_ioctl_quota_rescan(file, argp);
5338 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5339 return btrfs_ioctl_quota_rescan_status(file, argp);
5340 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5341 return btrfs_ioctl_quota_rescan_wait(file, argp);
5342 case BTRFS_IOC_DEV_REPLACE:
5343 return btrfs_ioctl_dev_replace(root, argp);
5344 case BTRFS_IOC_GET_FSLABEL:
5345 return btrfs_ioctl_get_fslabel(file, argp);
5346 case BTRFS_IOC_SET_FSLABEL:
5347 return btrfs_ioctl_set_fslabel(file, argp);
5348 case BTRFS_IOC_FILE_EXTENT_SAME:
5349 return btrfs_ioctl_file_extent_same(file, argp);
5350 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5351 return btrfs_ioctl_get_supported_features(file, argp);
5352 case BTRFS_IOC_GET_FEATURES:
5353 return btrfs_ioctl_get_features(file, argp);
5354 case BTRFS_IOC_SET_FEATURES:
5355 return btrfs_ioctl_set_features(file, argp);
5356 }
5357
5358 return -ENOTTY;
5359}
Linux 6.x block layer and io_uring tree(s)