]>
Commit | Line | Data |
---|---|---|
f46b5a66 CH |
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> | |
cb8e7090 | 24 | #include <linux/fsnotify.h> |
f46b5a66 CH |
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> | |
f46b5a66 | 30 | #include <linux/backing-dev.h> |
cb8e7090 | 31 | #include <linux/mount.h> |
f46b5a66 | 32 | #include <linux/mpage.h> |
cb8e7090 | 33 | #include <linux/namei.h> |
f46b5a66 CH |
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> | |
cb8e7090 | 39 | #include <linux/security.h> |
f46b5a66 | 40 | #include <linux/xattr.h> |
7ea394f1 | 41 | #include <linux/vmalloc.h> |
5a0e3ad6 | 42 | #include <linux/slab.h> |
f7039b1d | 43 | #include <linux/blkdev.h> |
4b4e25f2 | 44 | #include "compat.h" |
f46b5a66 CH |
45 | #include "ctree.h" |
46 | #include "disk-io.h" | |
47 | #include "transaction.h" | |
48 | #include "btrfs_inode.h" | |
49 | #include "ioctl.h" | |
50 | #include "print-tree.h" | |
51 | #include "volumes.h" | |
925baedd | 52 | #include "locking.h" |
581bb050 | 53 | #include "inode-map.h" |
f46b5a66 | 54 | |
6cbff00f CH |
55 | /* Mask out flags that are inappropriate for the given type of inode. */ |
56 | static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) | |
57 | { | |
58 | if (S_ISDIR(mode)) | |
59 | return flags; | |
60 | else if (S_ISREG(mode)) | |
61 | return flags & ~FS_DIRSYNC_FL; | |
62 | else | |
63 | return flags & (FS_NODUMP_FL | FS_NOATIME_FL); | |
64 | } | |
65 | ||
66 | /* | |
67 | * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. | |
68 | */ | |
69 | static unsigned int btrfs_flags_to_ioctl(unsigned int flags) | |
70 | { | |
71 | unsigned int iflags = 0; | |
72 | ||
73 | if (flags & BTRFS_INODE_SYNC) | |
74 | iflags |= FS_SYNC_FL; | |
75 | if (flags & BTRFS_INODE_IMMUTABLE) | |
76 | iflags |= FS_IMMUTABLE_FL; | |
77 | if (flags & BTRFS_INODE_APPEND) | |
78 | iflags |= FS_APPEND_FL; | |
79 | if (flags & BTRFS_INODE_NODUMP) | |
80 | iflags |= FS_NODUMP_FL; | |
81 | if (flags & BTRFS_INODE_NOATIME) | |
82 | iflags |= FS_NOATIME_FL; | |
83 | if (flags & BTRFS_INODE_DIRSYNC) | |
84 | iflags |= FS_DIRSYNC_FL; | |
d0092bdd LZ |
85 | if (flags & BTRFS_INODE_NODATACOW) |
86 | iflags |= FS_NOCOW_FL; | |
87 | ||
88 | if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS)) | |
89 | iflags |= FS_COMPR_FL; | |
90 | else if (flags & BTRFS_INODE_NOCOMPRESS) | |
91 | iflags |= FS_NOCOMP_FL; | |
6cbff00f CH |
92 | |
93 | return iflags; | |
94 | } | |
95 | ||
96 | /* | |
97 | * Update inode->i_flags based on the btrfs internal flags. | |
98 | */ | |
99 | void btrfs_update_iflags(struct inode *inode) | |
100 | { | |
101 | struct btrfs_inode *ip = BTRFS_I(inode); | |
102 | ||
103 | inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC); | |
104 | ||
105 | if (ip->flags & BTRFS_INODE_SYNC) | |
106 | inode->i_flags |= S_SYNC; | |
107 | if (ip->flags & BTRFS_INODE_IMMUTABLE) | |
108 | inode->i_flags |= S_IMMUTABLE; | |
109 | if (ip->flags & BTRFS_INODE_APPEND) | |
110 | inode->i_flags |= S_APPEND; | |
111 | if (ip->flags & BTRFS_INODE_NOATIME) | |
112 | inode->i_flags |= S_NOATIME; | |
113 | if (ip->flags & BTRFS_INODE_DIRSYNC) | |
114 | inode->i_flags |= S_DIRSYNC; | |
115 | } | |
116 | ||
117 | /* | |
118 | * Inherit flags from the parent inode. | |
119 | * | |
120 | * Unlike extN we don't have any flags we don't want to inherit currently. | |
121 | */ | |
122 | void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) | |
123 | { | |
0b4dcea5 CM |
124 | unsigned int flags; |
125 | ||
126 | if (!dir) | |
127 | return; | |
128 | ||
129 | flags = BTRFS_I(dir)->flags; | |
6cbff00f CH |
130 | |
131 | if (S_ISREG(inode->i_mode)) | |
132 | flags &= ~BTRFS_INODE_DIRSYNC; | |
133 | else if (!S_ISDIR(inode->i_mode)) | |
134 | flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME); | |
135 | ||
136 | BTRFS_I(inode)->flags = flags; | |
137 | btrfs_update_iflags(inode); | |
138 | } | |
139 | ||
140 | static int btrfs_ioctl_getflags(struct file *file, void __user *arg) | |
141 | { | |
142 | struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode); | |
143 | unsigned int flags = btrfs_flags_to_ioctl(ip->flags); | |
144 | ||
145 | if (copy_to_user(arg, &flags, sizeof(flags))) | |
146 | return -EFAULT; | |
147 | return 0; | |
148 | } | |
149 | ||
75e7cb7f LB |
150 | static int check_flags(unsigned int flags) |
151 | { | |
152 | if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ | |
153 | FS_NOATIME_FL | FS_NODUMP_FL | \ | |
154 | FS_SYNC_FL | FS_DIRSYNC_FL | \ | |
e1e8fb6a LZ |
155 | FS_NOCOMP_FL | FS_COMPR_FL | |
156 | FS_NOCOW_FL)) | |
75e7cb7f LB |
157 | return -EOPNOTSUPP; |
158 | ||
159 | if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) | |
160 | return -EINVAL; | |
161 | ||
75e7cb7f LB |
162 | return 0; |
163 | } | |
164 | ||
6cbff00f CH |
165 | static int btrfs_ioctl_setflags(struct file *file, void __user *arg) |
166 | { | |
167 | struct inode *inode = file->f_path.dentry->d_inode; | |
168 | struct btrfs_inode *ip = BTRFS_I(inode); | |
169 | struct btrfs_root *root = ip->root; | |
170 | struct btrfs_trans_handle *trans; | |
171 | unsigned int flags, oldflags; | |
172 | int ret; | |
173 | ||
b83cc969 LZ |
174 | if (btrfs_root_readonly(root)) |
175 | return -EROFS; | |
176 | ||
6cbff00f CH |
177 | if (copy_from_user(&flags, arg, sizeof(flags))) |
178 | return -EFAULT; | |
179 | ||
75e7cb7f LB |
180 | ret = check_flags(flags); |
181 | if (ret) | |
182 | return ret; | |
f46b5a66 | 183 | |
2e149670 | 184 | if (!inode_owner_or_capable(inode)) |
6cbff00f CH |
185 | return -EACCES; |
186 | ||
187 | mutex_lock(&inode->i_mutex); | |
188 | ||
189 | flags = btrfs_mask_flags(inode->i_mode, flags); | |
190 | oldflags = btrfs_flags_to_ioctl(ip->flags); | |
191 | if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { | |
192 | if (!capable(CAP_LINUX_IMMUTABLE)) { | |
193 | ret = -EPERM; | |
194 | goto out_unlock; | |
195 | } | |
196 | } | |
197 | ||
198 | ret = mnt_want_write(file->f_path.mnt); | |
199 | if (ret) | |
200 | goto out_unlock; | |
201 | ||
202 | if (flags & FS_SYNC_FL) | |
203 | ip->flags |= BTRFS_INODE_SYNC; | |
204 | else | |
205 | ip->flags &= ~BTRFS_INODE_SYNC; | |
206 | if (flags & FS_IMMUTABLE_FL) | |
207 | ip->flags |= BTRFS_INODE_IMMUTABLE; | |
208 | else | |
209 | ip->flags &= ~BTRFS_INODE_IMMUTABLE; | |
210 | if (flags & FS_APPEND_FL) | |
211 | ip->flags |= BTRFS_INODE_APPEND; | |
212 | else | |
213 | ip->flags &= ~BTRFS_INODE_APPEND; | |
214 | if (flags & FS_NODUMP_FL) | |
215 | ip->flags |= BTRFS_INODE_NODUMP; | |
216 | else | |
217 | ip->flags &= ~BTRFS_INODE_NODUMP; | |
218 | if (flags & FS_NOATIME_FL) | |
219 | ip->flags |= BTRFS_INODE_NOATIME; | |
220 | else | |
221 | ip->flags &= ~BTRFS_INODE_NOATIME; | |
222 | if (flags & FS_DIRSYNC_FL) | |
223 | ip->flags |= BTRFS_INODE_DIRSYNC; | |
224 | else | |
225 | ip->flags &= ~BTRFS_INODE_DIRSYNC; | |
e1e8fb6a LZ |
226 | if (flags & FS_NOCOW_FL) |
227 | ip->flags |= BTRFS_INODE_NODATACOW; | |
228 | else | |
229 | ip->flags &= ~BTRFS_INODE_NODATACOW; | |
6cbff00f | 230 | |
75e7cb7f LB |
231 | /* |
232 | * The COMPRESS flag can only be changed by users, while the NOCOMPRESS | |
233 | * flag may be changed automatically if compression code won't make | |
234 | * things smaller. | |
235 | */ | |
236 | if (flags & FS_NOCOMP_FL) { | |
237 | ip->flags &= ~BTRFS_INODE_COMPRESS; | |
238 | ip->flags |= BTRFS_INODE_NOCOMPRESS; | |
239 | } else if (flags & FS_COMPR_FL) { | |
240 | ip->flags |= BTRFS_INODE_COMPRESS; | |
241 | ip->flags &= ~BTRFS_INODE_NOCOMPRESS; | |
ebcb904d LZ |
242 | } else { |
243 | ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); | |
75e7cb7f | 244 | } |
6cbff00f | 245 | |
7a7eaa40 | 246 | trans = btrfs_join_transaction(root); |
3612b495 | 247 | BUG_ON(IS_ERR(trans)); |
6cbff00f CH |
248 | |
249 | ret = btrfs_update_inode(trans, root, inode); | |
250 | BUG_ON(ret); | |
251 | ||
252 | btrfs_update_iflags(inode); | |
253 | inode->i_ctime = CURRENT_TIME; | |
254 | btrfs_end_transaction(trans, root); | |
255 | ||
256 | mnt_drop_write(file->f_path.mnt); | |
2d4e6f6a | 257 | |
258 | ret = 0; | |
6cbff00f CH |
259 | out_unlock: |
260 | mutex_unlock(&inode->i_mutex); | |
2d4e6f6a | 261 | return ret; |
6cbff00f CH |
262 | } |
263 | ||
264 | static int btrfs_ioctl_getversion(struct file *file, int __user *arg) | |
265 | { | |
266 | struct inode *inode = file->f_path.dentry->d_inode; | |
267 | ||
268 | return put_user(inode->i_generation, arg); | |
269 | } | |
f46b5a66 | 270 | |
f7039b1d LD |
271 | static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) |
272 | { | |
273 | struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info; | |
274 | struct btrfs_fs_info *fs_info = root->fs_info; | |
275 | struct btrfs_device *device; | |
276 | struct request_queue *q; | |
277 | struct fstrim_range range; | |
278 | u64 minlen = ULLONG_MAX; | |
279 | u64 num_devices = 0; | |
280 | int ret; | |
281 | ||
282 | if (!capable(CAP_SYS_ADMIN)) | |
283 | return -EPERM; | |
284 | ||
1f78160c XG |
285 | rcu_read_lock(); |
286 | list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, | |
287 | dev_list) { | |
f7039b1d LD |
288 | if (!device->bdev) |
289 | continue; | |
290 | q = bdev_get_queue(device->bdev); | |
291 | if (blk_queue_discard(q)) { | |
292 | num_devices++; | |
293 | minlen = min((u64)q->limits.discard_granularity, | |
294 | minlen); | |
295 | } | |
296 | } | |
1f78160c | 297 | rcu_read_unlock(); |
f7039b1d LD |
298 | if (!num_devices) |
299 | return -EOPNOTSUPP; | |
300 | ||
301 | if (copy_from_user(&range, arg, sizeof(range))) | |
302 | return -EFAULT; | |
303 | ||
304 | range.minlen = max(range.minlen, minlen); | |
305 | ret = btrfs_trim_fs(root, &range); | |
306 | if (ret < 0) | |
307 | return ret; | |
308 | ||
309 | if (copy_to_user(arg, &range, sizeof(range))) | |
310 | return -EFAULT; | |
311 | ||
312 | return 0; | |
313 | } | |
314 | ||
cb8e7090 CH |
315 | static noinline int create_subvol(struct btrfs_root *root, |
316 | struct dentry *dentry, | |
72fd032e SW |
317 | char *name, int namelen, |
318 | u64 *async_transid) | |
f46b5a66 CH |
319 | { |
320 | struct btrfs_trans_handle *trans; | |
321 | struct btrfs_key key; | |
322 | struct btrfs_root_item root_item; | |
323 | struct btrfs_inode_item *inode_item; | |
324 | struct extent_buffer *leaf; | |
76dda93c | 325 | struct btrfs_root *new_root; |
2fbe8c8a | 326 | struct dentry *parent = dentry->d_parent; |
6a912213 | 327 | struct inode *dir; |
f46b5a66 CH |
328 | int ret; |
329 | int err; | |
330 | u64 objectid; | |
331 | u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; | |
3de4586c | 332 | u64 index = 0; |
f46b5a66 | 333 | |
581bb050 | 334 | ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); |
2fbe8c8a | 335 | if (ret) |
a22285a6 | 336 | return ret; |
6a912213 JB |
337 | |
338 | dir = parent->d_inode; | |
339 | ||
9ed74f2d JB |
340 | /* |
341 | * 1 - inode item | |
342 | * 2 - refs | |
343 | * 1 - root item | |
344 | * 2 - dir items | |
345 | */ | |
a22285a6 | 346 | trans = btrfs_start_transaction(root, 6); |
2fbe8c8a | 347 | if (IS_ERR(trans)) |
a22285a6 | 348 | return PTR_ERR(trans); |
f46b5a66 | 349 | |
5d4f98a2 YZ |
350 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, |
351 | 0, objectid, NULL, 0, 0, 0); | |
8e8a1e31 JB |
352 | if (IS_ERR(leaf)) { |
353 | ret = PTR_ERR(leaf); | |
354 | goto fail; | |
355 | } | |
f46b5a66 | 356 | |
5d4f98a2 | 357 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
f46b5a66 CH |
358 | btrfs_set_header_bytenr(leaf, leaf->start); |
359 | btrfs_set_header_generation(leaf, trans->transid); | |
5d4f98a2 | 360 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); |
f46b5a66 CH |
361 | btrfs_set_header_owner(leaf, objectid); |
362 | ||
363 | write_extent_buffer(leaf, root->fs_info->fsid, | |
364 | (unsigned long)btrfs_header_fsid(leaf), | |
365 | BTRFS_FSID_SIZE); | |
5d4f98a2 YZ |
366 | write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, |
367 | (unsigned long)btrfs_header_chunk_tree_uuid(leaf), | |
368 | BTRFS_UUID_SIZE); | |
f46b5a66 CH |
369 | btrfs_mark_buffer_dirty(leaf); |
370 | ||
371 | inode_item = &root_item.inode; | |
372 | memset(inode_item, 0, sizeof(*inode_item)); | |
373 | inode_item->generation = cpu_to_le64(1); | |
374 | inode_item->size = cpu_to_le64(3); | |
375 | inode_item->nlink = cpu_to_le32(1); | |
a76a3cd4 | 376 | inode_item->nbytes = cpu_to_le64(root->leafsize); |
f46b5a66 CH |
377 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); |
378 | ||
08fe4db1 LZ |
379 | root_item.flags = 0; |
380 | root_item.byte_limit = 0; | |
381 | inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT); | |
382 | ||
f46b5a66 | 383 | btrfs_set_root_bytenr(&root_item, leaf->start); |
84234f3a | 384 | btrfs_set_root_generation(&root_item, trans->transid); |
f46b5a66 CH |
385 | btrfs_set_root_level(&root_item, 0); |
386 | btrfs_set_root_refs(&root_item, 1); | |
86b9f2ec | 387 | btrfs_set_root_used(&root_item, leaf->len); |
80ff3856 | 388 | btrfs_set_root_last_snapshot(&root_item, 0); |
f46b5a66 CH |
389 | |
390 | memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); | |
391 | root_item.drop_level = 0; | |
392 | ||
925baedd | 393 | btrfs_tree_unlock(leaf); |
f46b5a66 CH |
394 | free_extent_buffer(leaf); |
395 | leaf = NULL; | |
396 | ||
397 | btrfs_set_root_dirid(&root_item, new_dirid); | |
398 | ||
399 | key.objectid = objectid; | |
5d4f98a2 | 400 | key.offset = 0; |
f46b5a66 CH |
401 | btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); |
402 | ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, | |
403 | &root_item); | |
404 | if (ret) | |
405 | goto fail; | |
406 | ||
76dda93c YZ |
407 | key.offset = (u64)-1; |
408 | new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); | |
409 | BUG_ON(IS_ERR(new_root)); | |
410 | ||
411 | btrfs_record_root_in_trans(trans, new_root); | |
412 | ||
d82a6f1d | 413 | ret = btrfs_create_subvol_root(trans, new_root, new_dirid); |
f46b5a66 CH |
414 | /* |
415 | * insert the directory item | |
416 | */ | |
3de4586c CM |
417 | ret = btrfs_set_inode_index(dir, &index); |
418 | BUG_ON(ret); | |
419 | ||
420 | ret = btrfs_insert_dir_item(trans, root, | |
16cdcec7 | 421 | name, namelen, dir, &key, |
3de4586c | 422 | BTRFS_FT_DIR, index); |
f46b5a66 CH |
423 | if (ret) |
424 | goto fail; | |
0660b5af | 425 | |
52c26179 YZ |
426 | btrfs_i_size_write(dir, dir->i_size + namelen * 2); |
427 | ret = btrfs_update_inode(trans, root, dir); | |
428 | BUG_ON(ret); | |
429 | ||
0660b5af | 430 | ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, |
4df27c4d | 431 | objectid, root->root_key.objectid, |
33345d01 | 432 | btrfs_ino(dir), index, name, namelen); |
0660b5af | 433 | |
76dda93c | 434 | BUG_ON(ret); |
f46b5a66 | 435 | |
76dda93c | 436 | d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry)); |
f46b5a66 | 437 | fail: |
72fd032e SW |
438 | if (async_transid) { |
439 | *async_transid = trans->transid; | |
440 | err = btrfs_commit_transaction_async(trans, root, 1); | |
441 | } else { | |
442 | err = btrfs_commit_transaction(trans, root); | |
443 | } | |
f46b5a66 CH |
444 | if (err && !ret) |
445 | ret = err; | |
f46b5a66 CH |
446 | return ret; |
447 | } | |
448 | ||
72fd032e | 449 | static int create_snapshot(struct btrfs_root *root, struct dentry *dentry, |
b83cc969 LZ |
450 | char *name, int namelen, u64 *async_transid, |
451 | bool readonly) | |
f46b5a66 | 452 | { |
2e4bfab9 | 453 | struct inode *inode; |
f46b5a66 CH |
454 | struct btrfs_pending_snapshot *pending_snapshot; |
455 | struct btrfs_trans_handle *trans; | |
2e4bfab9 | 456 | int ret; |
f46b5a66 CH |
457 | |
458 | if (!root->ref_cows) | |
459 | return -EINVAL; | |
460 | ||
3de4586c | 461 | pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); |
a22285a6 YZ |
462 | if (!pending_snapshot) |
463 | return -ENOMEM; | |
464 | ||
465 | btrfs_init_block_rsv(&pending_snapshot->block_rsv); | |
3de4586c | 466 | pending_snapshot->dentry = dentry; |
f46b5a66 | 467 | pending_snapshot->root = root; |
b83cc969 | 468 | pending_snapshot->readonly = readonly; |
a22285a6 YZ |
469 | |
470 | trans = btrfs_start_transaction(root->fs_info->extent_root, 5); | |
471 | if (IS_ERR(trans)) { | |
472 | ret = PTR_ERR(trans); | |
473 | goto fail; | |
474 | } | |
475 | ||
476 | ret = btrfs_snap_reserve_metadata(trans, pending_snapshot); | |
477 | BUG_ON(ret); | |
478 | ||
8351583e | 479 | spin_lock(&root->fs_info->trans_lock); |
f46b5a66 CH |
480 | list_add(&pending_snapshot->list, |
481 | &trans->transaction->pending_snapshots); | |
8351583e | 482 | spin_unlock(&root->fs_info->trans_lock); |
72fd032e SW |
483 | if (async_transid) { |
484 | *async_transid = trans->transid; | |
485 | ret = btrfs_commit_transaction_async(trans, | |
486 | root->fs_info->extent_root, 1); | |
487 | } else { | |
488 | ret = btrfs_commit_transaction(trans, | |
489 | root->fs_info->extent_root); | |
490 | } | |
2e4bfab9 | 491 | BUG_ON(ret); |
a22285a6 YZ |
492 | |
493 | ret = pending_snapshot->error; | |
494 | if (ret) | |
495 | goto fail; | |
496 | ||
66b4ffd1 JB |
497 | ret = btrfs_orphan_cleanup(pending_snapshot->snap); |
498 | if (ret) | |
499 | goto fail; | |
f46b5a66 | 500 | |
2fbe8c8a | 501 | inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry); |
2e4bfab9 YZ |
502 | if (IS_ERR(inode)) { |
503 | ret = PTR_ERR(inode); | |
504 | goto fail; | |
505 | } | |
506 | BUG_ON(!inode); | |
507 | d_instantiate(dentry, inode); | |
508 | ret = 0; | |
509 | fail: | |
a22285a6 | 510 | kfree(pending_snapshot); |
f46b5a66 CH |
511 | return ret; |
512 | } | |
513 | ||
4260f7c7 SW |
514 | /* copy of check_sticky in fs/namei.c() |
515 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
516 | * minimal. | |
517 | */ | |
518 | static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode) | |
519 | { | |
520 | uid_t fsuid = current_fsuid(); | |
521 | ||
522 | if (!(dir->i_mode & S_ISVTX)) | |
523 | return 0; | |
524 | if (inode->i_uid == fsuid) | |
525 | return 0; | |
526 | if (dir->i_uid == fsuid) | |
527 | return 0; | |
528 | return !capable(CAP_FOWNER); | |
529 | } | |
530 | ||
531 | /* copy of may_delete in fs/namei.c() | |
532 | * Check whether we can remove a link victim from directory dir, check | |
533 | * whether the type of victim is right. | |
534 | * 1. We can't do it if dir is read-only (done in permission()) | |
535 | * 2. We should have write and exec permissions on dir | |
536 | * 3. We can't remove anything from append-only dir | |
537 | * 4. We can't do anything with immutable dir (done in permission()) | |
538 | * 5. If the sticky bit on dir is set we should either | |
539 | * a. be owner of dir, or | |
540 | * b. be owner of victim, or | |
541 | * c. have CAP_FOWNER capability | |
542 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
543 | * links pointing to it. | |
544 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
545 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
546 | * 9. We can't remove a root or mountpoint. | |
547 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
548 | * nfs_async_unlink(). | |
549 | */ | |
550 | ||
551 | static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir) | |
552 | { | |
553 | int error; | |
554 | ||
555 | if (!victim->d_inode) | |
556 | return -ENOENT; | |
557 | ||
558 | BUG_ON(victim->d_parent->d_inode != dir); | |
559 | audit_inode_child(victim, dir); | |
560 | ||
561 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
562 | if (error) | |
563 | return error; | |
564 | if (IS_APPEND(dir)) | |
565 | return -EPERM; | |
566 | if (btrfs_check_sticky(dir, victim->d_inode)|| | |
567 | IS_APPEND(victim->d_inode)|| | |
568 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) | |
569 | return -EPERM; | |
570 | if (isdir) { | |
571 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
572 | return -ENOTDIR; | |
573 | if (IS_ROOT(victim)) | |
574 | return -EBUSY; | |
575 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
576 | return -EISDIR; | |
577 | if (IS_DEADDIR(dir)) | |
578 | return -ENOENT; | |
579 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
580 | return -EBUSY; | |
581 | return 0; | |
582 | } | |
583 | ||
cb8e7090 CH |
584 | /* copy of may_create in fs/namei.c() */ |
585 | static inline int btrfs_may_create(struct inode *dir, struct dentry *child) | |
586 | { | |
587 | if (child->d_inode) | |
588 | return -EEXIST; | |
589 | if (IS_DEADDIR(dir)) | |
590 | return -ENOENT; | |
591 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
592 | } | |
593 | ||
594 | /* | |
595 | * Create a new subvolume below @parent. This is largely modeled after | |
596 | * sys_mkdirat and vfs_mkdir, but we only do a single component lookup | |
597 | * inside this filesystem so it's quite a bit simpler. | |
598 | */ | |
76dda93c YZ |
599 | static noinline int btrfs_mksubvol(struct path *parent, |
600 | char *name, int namelen, | |
72fd032e | 601 | struct btrfs_root *snap_src, |
b83cc969 | 602 | u64 *async_transid, bool readonly) |
cb8e7090 | 603 | { |
76dda93c | 604 | struct inode *dir = parent->dentry->d_inode; |
cb8e7090 CH |
605 | struct dentry *dentry; |
606 | int error; | |
607 | ||
76dda93c | 608 | mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); |
cb8e7090 CH |
609 | |
610 | dentry = lookup_one_len(name, parent->dentry, namelen); | |
611 | error = PTR_ERR(dentry); | |
612 | if (IS_ERR(dentry)) | |
613 | goto out_unlock; | |
614 | ||
615 | error = -EEXIST; | |
616 | if (dentry->d_inode) | |
617 | goto out_dput; | |
618 | ||
cb8e7090 CH |
619 | error = mnt_want_write(parent->mnt); |
620 | if (error) | |
621 | goto out_dput; | |
622 | ||
76dda93c | 623 | error = btrfs_may_create(dir, dentry); |
cb8e7090 CH |
624 | if (error) |
625 | goto out_drop_write; | |
626 | ||
76dda93c YZ |
627 | down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); |
628 | ||
629 | if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) | |
630 | goto out_up_read; | |
631 | ||
3de4586c | 632 | if (snap_src) { |
72fd032e | 633 | error = create_snapshot(snap_src, dentry, |
b83cc969 | 634 | name, namelen, async_transid, readonly); |
3de4586c | 635 | } else { |
76dda93c | 636 | error = create_subvol(BTRFS_I(dir)->root, dentry, |
72fd032e | 637 | name, namelen, async_transid); |
3de4586c | 638 | } |
76dda93c YZ |
639 | if (!error) |
640 | fsnotify_mkdir(dir, dentry); | |
641 | out_up_read: | |
642 | up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); | |
cb8e7090 CH |
643 | out_drop_write: |
644 | mnt_drop_write(parent->mnt); | |
645 | out_dput: | |
646 | dput(dentry); | |
647 | out_unlock: | |
76dda93c | 648 | mutex_unlock(&dir->i_mutex); |
cb8e7090 CH |
649 | return error; |
650 | } | |
651 | ||
4cb5300b CM |
652 | /* |
653 | * When we're defragging a range, we don't want to kick it off again | |
654 | * if it is really just waiting for delalloc to send it down. | |
655 | * If we find a nice big extent or delalloc range for the bytes in the | |
656 | * file you want to defrag, we return 0 to let you know to skip this | |
657 | * part of the file | |
658 | */ | |
659 | static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh) | |
660 | { | |
661 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
662 | struct extent_map *em = NULL; | |
663 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
664 | u64 end; | |
665 | ||
666 | read_lock(&em_tree->lock); | |
667 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
668 | read_unlock(&em_tree->lock); | |
669 | ||
670 | if (em) { | |
671 | end = extent_map_end(em); | |
672 | free_extent_map(em); | |
673 | if (end - offset > thresh) | |
674 | return 0; | |
675 | } | |
676 | /* if we already have a nice delalloc here, just stop */ | |
677 | thresh /= 2; | |
678 | end = count_range_bits(io_tree, &offset, offset + thresh, | |
679 | thresh, EXTENT_DELALLOC, 1); | |
680 | if (end >= thresh) | |
681 | return 0; | |
682 | return 1; | |
683 | } | |
684 | ||
685 | /* | |
686 | * helper function to walk through a file and find extents | |
687 | * newer than a specific transid, and smaller than thresh. | |
688 | * | |
689 | * This is used by the defragging code to find new and small | |
690 | * extents | |
691 | */ | |
692 | static int find_new_extents(struct btrfs_root *root, | |
693 | struct inode *inode, u64 newer_than, | |
694 | u64 *off, int thresh) | |
695 | { | |
696 | struct btrfs_path *path; | |
697 | struct btrfs_key min_key; | |
698 | struct btrfs_key max_key; | |
699 | struct extent_buffer *leaf; | |
700 | struct btrfs_file_extent_item *extent; | |
701 | int type; | |
702 | int ret; | |
a4689d2b | 703 | u64 ino = btrfs_ino(inode); |
4cb5300b CM |
704 | |
705 | path = btrfs_alloc_path(); | |
706 | if (!path) | |
707 | return -ENOMEM; | |
708 | ||
a4689d2b | 709 | min_key.objectid = ino; |
4cb5300b CM |
710 | min_key.type = BTRFS_EXTENT_DATA_KEY; |
711 | min_key.offset = *off; | |
712 | ||
a4689d2b | 713 | max_key.objectid = ino; |
4cb5300b CM |
714 | max_key.type = (u8)-1; |
715 | max_key.offset = (u64)-1; | |
716 | ||
717 | path->keep_locks = 1; | |
718 | ||
719 | while(1) { | |
720 | ret = btrfs_search_forward(root, &min_key, &max_key, | |
721 | path, 0, newer_than); | |
722 | if (ret != 0) | |
723 | goto none; | |
a4689d2b | 724 | if (min_key.objectid != ino) |
4cb5300b CM |
725 | goto none; |
726 | if (min_key.type != BTRFS_EXTENT_DATA_KEY) | |
727 | goto none; | |
728 | ||
729 | leaf = path->nodes[0]; | |
730 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
731 | struct btrfs_file_extent_item); | |
732 | ||
733 | type = btrfs_file_extent_type(leaf, extent); | |
734 | if (type == BTRFS_FILE_EXTENT_REG && | |
735 | btrfs_file_extent_num_bytes(leaf, extent) < thresh && | |
736 | check_defrag_in_cache(inode, min_key.offset, thresh)) { | |
737 | *off = min_key.offset; | |
738 | btrfs_free_path(path); | |
739 | return 0; | |
740 | } | |
741 | ||
742 | if (min_key.offset == (u64)-1) | |
743 | goto none; | |
744 | ||
745 | min_key.offset++; | |
746 | btrfs_release_path(path); | |
747 | } | |
748 | none: | |
749 | btrfs_free_path(path); | |
750 | return -ENOENT; | |
751 | } | |
752 | ||
940100a4 | 753 | static int should_defrag_range(struct inode *inode, u64 start, u64 len, |
1e701a32 CM |
754 | int thresh, u64 *last_len, u64 *skip, |
755 | u64 *defrag_end) | |
940100a4 CM |
756 | { |
757 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
758 | struct extent_map *em = NULL; | |
759 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
760 | int ret = 1; | |
761 | ||
762 | /* | |
763 | * make sure that once we start defragging and extent, we keep on | |
764 | * defragging it | |
765 | */ | |
766 | if (start < *defrag_end) | |
767 | return 1; | |
768 | ||
769 | *skip = 0; | |
770 | ||
771 | /* | |
772 | * hopefully we have this extent in the tree already, try without | |
773 | * the full extent lock | |
774 | */ | |
775 | read_lock(&em_tree->lock); | |
776 | em = lookup_extent_mapping(em_tree, start, len); | |
777 | read_unlock(&em_tree->lock); | |
778 | ||
779 | if (!em) { | |
780 | /* get the big lock and read metadata off disk */ | |
781 | lock_extent(io_tree, start, start + len - 1, GFP_NOFS); | |
782 | em = btrfs_get_extent(inode, NULL, 0, start, len, 0); | |
783 | unlock_extent(io_tree, start, start + len - 1, GFP_NOFS); | |
784 | ||
6cf8bfbf | 785 | if (IS_ERR(em)) |
940100a4 CM |
786 | return 0; |
787 | } | |
788 | ||
789 | /* this will cover holes, and inline extents */ | |
790 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) | |
791 | ret = 0; | |
792 | ||
793 | /* | |
794 | * we hit a real extent, if it is big don't bother defragging it again | |
795 | */ | |
1e701a32 | 796 | if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh) |
940100a4 CM |
797 | ret = 0; |
798 | ||
799 | /* | |
800 | * last_len ends up being a counter of how many bytes we've defragged. | |
801 | * every time we choose not to defrag an extent, we reset *last_len | |
802 | * so that the next tiny extent will force a defrag. | |
803 | * | |
804 | * The end result of this is that tiny extents before a single big | |
805 | * extent will force at least part of that big extent to be defragged. | |
806 | */ | |
807 | if (ret) { | |
808 | *last_len += len; | |
809 | *defrag_end = extent_map_end(em); | |
810 | } else { | |
811 | *last_len = 0; | |
812 | *skip = extent_map_end(em); | |
813 | *defrag_end = 0; | |
814 | } | |
815 | ||
816 | free_extent_map(em); | |
817 | return ret; | |
818 | } | |
819 | ||
4cb5300b CM |
820 | /* |
821 | * it doesn't do much good to defrag one or two pages | |
822 | * at a time. This pulls in a nice chunk of pages | |
823 | * to COW and defrag. | |
824 | * | |
825 | * It also makes sure the delalloc code has enough | |
826 | * dirty data to avoid making new small extents as part | |
827 | * of the defrag | |
828 | * | |
829 | * It's a good idea to start RA on this range | |
830 | * before calling this. | |
831 | */ | |
832 | static int cluster_pages_for_defrag(struct inode *inode, | |
833 | struct page **pages, | |
834 | unsigned long start_index, | |
835 | int num_pages) | |
f46b5a66 | 836 | { |
4cb5300b CM |
837 | unsigned long file_end; |
838 | u64 isize = i_size_read(inode); | |
839 | u64 page_start; | |
840 | u64 page_end; | |
841 | int ret; | |
842 | int i; | |
843 | int i_done; | |
3eaa2885 | 844 | struct btrfs_ordered_extent *ordered; |
4cb5300b CM |
845 | struct extent_state *cached_state = NULL; |
846 | ||
847 | if (isize == 0) | |
848 | return 0; | |
849 | file_end = (isize - 1) >> PAGE_CACHE_SHIFT; | |
850 | ||
851 | ret = btrfs_delalloc_reserve_space(inode, | |
852 | num_pages << PAGE_CACHE_SHIFT); | |
853 | if (ret) | |
854 | return ret; | |
855 | again: | |
856 | ret = 0; | |
857 | i_done = 0; | |
858 | ||
859 | /* step one, lock all the pages */ | |
860 | for (i = 0; i < num_pages; i++) { | |
861 | struct page *page; | |
a94733d0 JB |
862 | page = find_or_create_page(inode->i_mapping, |
863 | start_index + i, GFP_NOFS); | |
4cb5300b CM |
864 | if (!page) |
865 | break; | |
866 | ||
867 | if (!PageUptodate(page)) { | |
868 | btrfs_readpage(NULL, page); | |
869 | lock_page(page); | |
870 | if (!PageUptodate(page)) { | |
871 | unlock_page(page); | |
872 | page_cache_release(page); | |
873 | ret = -EIO; | |
874 | break; | |
875 | } | |
876 | } | |
877 | isize = i_size_read(inode); | |
878 | file_end = (isize - 1) >> PAGE_CACHE_SHIFT; | |
879 | if (!isize || page->index > file_end || | |
880 | page->mapping != inode->i_mapping) { | |
881 | /* whoops, we blew past eof, skip this page */ | |
882 | unlock_page(page); | |
883 | page_cache_release(page); | |
884 | break; | |
885 | } | |
886 | pages[i] = page; | |
887 | i_done++; | |
888 | } | |
889 | if (!i_done || ret) | |
890 | goto out; | |
891 | ||
892 | if (!(inode->i_sb->s_flags & MS_ACTIVE)) | |
893 | goto out; | |
894 | ||
895 | /* | |
896 | * so now we have a nice long stream of locked | |
897 | * and up to date pages, lets wait on them | |
898 | */ | |
899 | for (i = 0; i < i_done; i++) | |
900 | wait_on_page_writeback(pages[i]); | |
901 | ||
902 | page_start = page_offset(pages[0]); | |
903 | page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE; | |
904 | ||
905 | lock_extent_bits(&BTRFS_I(inode)->io_tree, | |
906 | page_start, page_end - 1, 0, &cached_state, | |
907 | GFP_NOFS); | |
908 | ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1); | |
909 | if (ordered && | |
910 | ordered->file_offset + ordered->len > page_start && | |
911 | ordered->file_offset < page_end) { | |
912 | btrfs_put_ordered_extent(ordered); | |
913 | unlock_extent_cached(&BTRFS_I(inode)->io_tree, | |
914 | page_start, page_end - 1, | |
915 | &cached_state, GFP_NOFS); | |
916 | for (i = 0; i < i_done; i++) { | |
917 | unlock_page(pages[i]); | |
918 | page_cache_release(pages[i]); | |
919 | } | |
920 | btrfs_wait_ordered_range(inode, page_start, | |
921 | page_end - page_start); | |
922 | goto again; | |
923 | } | |
924 | if (ordered) | |
925 | btrfs_put_ordered_extent(ordered); | |
926 | ||
927 | clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, | |
928 | page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | | |
929 | EXTENT_DO_ACCOUNTING, 0, 0, &cached_state, | |
930 | GFP_NOFS); | |
931 | ||
932 | if (i_done != num_pages) { | |
9e0baf60 JB |
933 | spin_lock(&BTRFS_I(inode)->lock); |
934 | BTRFS_I(inode)->outstanding_extents++; | |
935 | spin_unlock(&BTRFS_I(inode)->lock); | |
4cb5300b CM |
936 | btrfs_delalloc_release_space(inode, |
937 | (num_pages - i_done) << PAGE_CACHE_SHIFT); | |
938 | } | |
939 | ||
940 | ||
941 | btrfs_set_extent_delalloc(inode, page_start, page_end - 1, | |
942 | &cached_state); | |
943 | ||
944 | unlock_extent_cached(&BTRFS_I(inode)->io_tree, | |
945 | page_start, page_end - 1, &cached_state, | |
946 | GFP_NOFS); | |
947 | ||
948 | for (i = 0; i < i_done; i++) { | |
949 | clear_page_dirty_for_io(pages[i]); | |
950 | ClearPageChecked(pages[i]); | |
951 | set_page_extent_mapped(pages[i]); | |
952 | set_page_dirty(pages[i]); | |
953 | unlock_page(pages[i]); | |
954 | page_cache_release(pages[i]); | |
955 | } | |
956 | return i_done; | |
957 | out: | |
958 | for (i = 0; i < i_done; i++) { | |
959 | unlock_page(pages[i]); | |
960 | page_cache_release(pages[i]); | |
961 | } | |
962 | btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT); | |
963 | return ret; | |
964 | ||
965 | } | |
966 | ||
967 | int btrfs_defrag_file(struct inode *inode, struct file *file, | |
968 | struct btrfs_ioctl_defrag_range_args *range, | |
969 | u64 newer_than, unsigned long max_to_defrag) | |
970 | { | |
971 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
1a419d85 | 972 | struct btrfs_super_block *disk_super; |
4cb5300b | 973 | struct file_ra_state *ra = NULL; |
f46b5a66 | 974 | unsigned long last_index; |
1a419d85 | 975 | u64 features; |
940100a4 CM |
976 | u64 last_len = 0; |
977 | u64 skip = 0; | |
978 | u64 defrag_end = 0; | |
4cb5300b CM |
979 | u64 newer_off = range->start; |
980 | int newer_left = 0; | |
f46b5a66 CH |
981 | unsigned long i; |
982 | int ret; | |
4cb5300b | 983 | int defrag_count = 0; |
1a419d85 | 984 | int compress_type = BTRFS_COMPRESS_ZLIB; |
4cb5300b CM |
985 | int extent_thresh = range->extent_thresh; |
986 | int newer_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT; | |
987 | u64 new_align = ~((u64)128 * 1024 - 1); | |
988 | struct page **pages = NULL; | |
989 | ||
990 | if (extent_thresh == 0) | |
991 | extent_thresh = 256 * 1024; | |
1a419d85 LZ |
992 | |
993 | if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { | |
994 | if (range->compress_type > BTRFS_COMPRESS_TYPES) | |
995 | return -EINVAL; | |
996 | if (range->compress_type) | |
997 | compress_type = range->compress_type; | |
998 | } | |
f46b5a66 | 999 | |
940100a4 CM |
1000 | if (inode->i_size == 0) |
1001 | return 0; | |
1002 | ||
4cb5300b CM |
1003 | /* |
1004 | * if we were not given a file, allocate a readahead | |
1005 | * context | |
1006 | */ | |
1007 | if (!file) { | |
1008 | ra = kzalloc(sizeof(*ra), GFP_NOFS); | |
1009 | if (!ra) | |
1010 | return -ENOMEM; | |
1011 | file_ra_state_init(ra, inode->i_mapping); | |
1012 | } else { | |
1013 | ra = &file->f_ra; | |
1014 | } | |
1015 | ||
1016 | pages = kmalloc(sizeof(struct page *) * newer_cluster, | |
1017 | GFP_NOFS); | |
1018 | if (!pages) { | |
1019 | ret = -ENOMEM; | |
1020 | goto out_ra; | |
1021 | } | |
1022 | ||
1023 | /* find the last page to defrag */ | |
1e701a32 CM |
1024 | if (range->start + range->len > range->start) { |
1025 | last_index = min_t(u64, inode->i_size - 1, | |
1026 | range->start + range->len - 1) >> PAGE_CACHE_SHIFT; | |
1027 | } else { | |
1028 | last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT; | |
1029 | } | |
1030 | ||
4cb5300b CM |
1031 | if (newer_than) { |
1032 | ret = find_new_extents(root, inode, newer_than, | |
1033 | &newer_off, 64 * 1024); | |
1034 | if (!ret) { | |
1035 | range->start = newer_off; | |
1036 | /* | |
1037 | * we always align our defrag to help keep | |
1038 | * the extents in the file evenly spaced | |
1039 | */ | |
1040 | i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; | |
1041 | newer_left = newer_cluster; | |
1042 | } else | |
1043 | goto out_ra; | |
1044 | } else { | |
1045 | i = range->start >> PAGE_CACHE_SHIFT; | |
1046 | } | |
1047 | if (!max_to_defrag) | |
1048 | max_to_defrag = last_index - 1; | |
1049 | ||
1050 | while (i <= last_index && defrag_count < max_to_defrag) { | |
1051 | /* | |
1052 | * make sure we stop running if someone unmounts | |
1053 | * the FS | |
1054 | */ | |
1055 | if (!(inode->i_sb->s_flags & MS_ACTIVE)) | |
1056 | break; | |
1057 | ||
1058 | if (!newer_than && | |
1059 | !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, | |
1e701a32 | 1060 | PAGE_CACHE_SIZE, |
4cb5300b | 1061 | extent_thresh, |
1e701a32 | 1062 | &last_len, &skip, |
940100a4 CM |
1063 | &defrag_end)) { |
1064 | unsigned long next; | |
1065 | /* | |
1066 | * the should_defrag function tells us how much to skip | |
1067 | * bump our counter by the suggested amount | |
1068 | */ | |
1069 | next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; | |
1070 | i = max(i + 1, next); | |
1071 | continue; | |
1072 | } | |
1e701a32 | 1073 | if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) |
1a419d85 | 1074 | BTRFS_I(inode)->force_compress = compress_type; |
940100a4 | 1075 | |
4cb5300b | 1076 | btrfs_force_ra(inode->i_mapping, ra, file, i, newer_cluster); |
940100a4 | 1077 | |
4cb5300b CM |
1078 | ret = cluster_pages_for_defrag(inode, pages, i, newer_cluster); |
1079 | if (ret < 0) | |
1080 | goto out_ra; | |
1081 | ||
1082 | defrag_count += ret; | |
1083 | balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret); | |
1084 | i += ret; | |
1085 | ||
1086 | if (newer_than) { | |
1087 | if (newer_off == (u64)-1) | |
1088 | break; | |
1089 | ||
1090 | newer_off = max(newer_off + 1, | |
1091 | (u64)i << PAGE_CACHE_SHIFT); | |
1092 | ||
1093 | ret = find_new_extents(root, inode, | |
1094 | newer_than, &newer_off, | |
1095 | 64 * 1024); | |
1096 | if (!ret) { | |
1097 | range->start = newer_off; | |
1098 | i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; | |
1099 | newer_left = newer_cluster; | |
1100 | } else { | |
1101 | break; | |
f46b5a66 | 1102 | } |
4cb5300b CM |
1103 | } else { |
1104 | i++; | |
f46b5a66 | 1105 | } |
f46b5a66 CH |
1106 | } |
1107 | ||
1e701a32 CM |
1108 | if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) |
1109 | filemap_flush(inode->i_mapping); | |
1110 | ||
1111 | if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
1112 | /* the filemap_flush will queue IO into the worker threads, but | |
1113 | * we have to make sure the IO is actually started and that | |
1114 | * ordered extents get created before we return | |
1115 | */ | |
1116 | atomic_inc(&root->fs_info->async_submit_draining); | |
1117 | while (atomic_read(&root->fs_info->nr_async_submits) || | |
1118 | atomic_read(&root->fs_info->async_delalloc_pages)) { | |
1119 | wait_event(root->fs_info->async_submit_wait, | |
1120 | (atomic_read(&root->fs_info->nr_async_submits) == 0 && | |
1121 | atomic_read(&root->fs_info->async_delalloc_pages) == 0)); | |
1122 | } | |
1123 | atomic_dec(&root->fs_info->async_submit_draining); | |
1124 | ||
1125 | mutex_lock(&inode->i_mutex); | |
261507a0 | 1126 | BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; |
1e701a32 CM |
1127 | mutex_unlock(&inode->i_mutex); |
1128 | } | |
1129 | ||
1a419d85 LZ |
1130 | disk_super = &root->fs_info->super_copy; |
1131 | features = btrfs_super_incompat_flags(disk_super); | |
1132 | if (range->compress_type == BTRFS_COMPRESS_LZO) { | |
1133 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; | |
1134 | btrfs_set_super_incompat_flags(disk_super, features); | |
1135 | } | |
1136 | ||
4cb5300b CM |
1137 | if (!file) |
1138 | kfree(ra); | |
1139 | return defrag_count; | |
940100a4 | 1140 | |
4cb5300b CM |
1141 | out_ra: |
1142 | if (!file) | |
1143 | kfree(ra); | |
1144 | kfree(pages); | |
940100a4 | 1145 | return ret; |
f46b5a66 CH |
1146 | } |
1147 | ||
76dda93c YZ |
1148 | static noinline int btrfs_ioctl_resize(struct btrfs_root *root, |
1149 | void __user *arg) | |
f46b5a66 CH |
1150 | { |
1151 | u64 new_size; | |
1152 | u64 old_size; | |
1153 | u64 devid = 1; | |
1154 | struct btrfs_ioctl_vol_args *vol_args; | |
1155 | struct btrfs_trans_handle *trans; | |
1156 | struct btrfs_device *device = NULL; | |
1157 | char *sizestr; | |
1158 | char *devstr = NULL; | |
1159 | int ret = 0; | |
f46b5a66 CH |
1160 | int mod = 0; |
1161 | ||
c146afad YZ |
1162 | if (root->fs_info->sb->s_flags & MS_RDONLY) |
1163 | return -EROFS; | |
1164 | ||
e441d54d CM |
1165 | if (!capable(CAP_SYS_ADMIN)) |
1166 | return -EPERM; | |
1167 | ||
dae7b665 LZ |
1168 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1169 | if (IS_ERR(vol_args)) | |
1170 | return PTR_ERR(vol_args); | |
5516e595 MF |
1171 | |
1172 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
f46b5a66 | 1173 | |
7d9eb12c | 1174 | mutex_lock(&root->fs_info->volume_mutex); |
f46b5a66 CH |
1175 | sizestr = vol_args->name; |
1176 | devstr = strchr(sizestr, ':'); | |
1177 | if (devstr) { | |
1178 | char *end; | |
1179 | sizestr = devstr + 1; | |
1180 | *devstr = '\0'; | |
1181 | devstr = vol_args->name; | |
1182 | devid = simple_strtoull(devstr, &end, 10); | |
21380931 JB |
1183 | printk(KERN_INFO "resizing devid %llu\n", |
1184 | (unsigned long long)devid); | |
f46b5a66 | 1185 | } |
2b82032c | 1186 | device = btrfs_find_device(root, devid, NULL, NULL); |
f46b5a66 | 1187 | if (!device) { |
21380931 JB |
1188 | printk(KERN_INFO "resizer unable to find device %llu\n", |
1189 | (unsigned long long)devid); | |
f46b5a66 CH |
1190 | ret = -EINVAL; |
1191 | goto out_unlock; | |
1192 | } | |
1193 | if (!strcmp(sizestr, "max")) | |
1194 | new_size = device->bdev->bd_inode->i_size; | |
1195 | else { | |
1196 | if (sizestr[0] == '-') { | |
1197 | mod = -1; | |
1198 | sizestr++; | |
1199 | } else if (sizestr[0] == '+') { | |
1200 | mod = 1; | |
1201 | sizestr++; | |
1202 | } | |
91748467 | 1203 | new_size = memparse(sizestr, NULL); |
f46b5a66 CH |
1204 | if (new_size == 0) { |
1205 | ret = -EINVAL; | |
1206 | goto out_unlock; | |
1207 | } | |
1208 | } | |
1209 | ||
1210 | old_size = device->total_bytes; | |
1211 | ||
1212 | if (mod < 0) { | |
1213 | if (new_size > old_size) { | |
1214 | ret = -EINVAL; | |
1215 | goto out_unlock; | |
1216 | } | |
1217 | new_size = old_size - new_size; | |
1218 | } else if (mod > 0) { | |
1219 | new_size = old_size + new_size; | |
1220 | } | |
1221 | ||
1222 | if (new_size < 256 * 1024 * 1024) { | |
1223 | ret = -EINVAL; | |
1224 | goto out_unlock; | |
1225 | } | |
1226 | if (new_size > device->bdev->bd_inode->i_size) { | |
1227 | ret = -EFBIG; | |
1228 | goto out_unlock; | |
1229 | } | |
1230 | ||
1231 | do_div(new_size, root->sectorsize); | |
1232 | new_size *= root->sectorsize; | |
1233 | ||
1234 | printk(KERN_INFO "new size for %s is %llu\n", | |
1235 | device->name, (unsigned long long)new_size); | |
1236 | ||
1237 | if (new_size > old_size) { | |
a22285a6 | 1238 | trans = btrfs_start_transaction(root, 0); |
98d5dc13 TI |
1239 | if (IS_ERR(trans)) { |
1240 | ret = PTR_ERR(trans); | |
1241 | goto out_unlock; | |
1242 | } | |
f46b5a66 CH |
1243 | ret = btrfs_grow_device(trans, device, new_size); |
1244 | btrfs_commit_transaction(trans, root); | |
1245 | } else { | |
1246 | ret = btrfs_shrink_device(device, new_size); | |
1247 | } | |
1248 | ||
1249 | out_unlock: | |
7d9eb12c | 1250 | mutex_unlock(&root->fs_info->volume_mutex); |
f46b5a66 CH |
1251 | kfree(vol_args); |
1252 | return ret; | |
1253 | } | |
1254 | ||
72fd032e SW |
1255 | static noinline int btrfs_ioctl_snap_create_transid(struct file *file, |
1256 | char *name, | |
1257 | unsigned long fd, | |
1258 | int subvol, | |
b83cc969 LZ |
1259 | u64 *transid, |
1260 | bool readonly) | |
f46b5a66 | 1261 | { |
cb8e7090 | 1262 | struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; |
3de4586c | 1263 | struct file *src_file; |
f46b5a66 | 1264 | int namelen; |
3de4586c | 1265 | int ret = 0; |
f46b5a66 | 1266 | |
c146afad YZ |
1267 | if (root->fs_info->sb->s_flags & MS_RDONLY) |
1268 | return -EROFS; | |
1269 | ||
72fd032e SW |
1270 | namelen = strlen(name); |
1271 | if (strchr(name, '/')) { | |
f46b5a66 CH |
1272 | ret = -EINVAL; |
1273 | goto out; | |
1274 | } | |
1275 | ||
3de4586c | 1276 | if (subvol) { |
72fd032e | 1277 | ret = btrfs_mksubvol(&file->f_path, name, namelen, |
b83cc969 | 1278 | NULL, transid, readonly); |
cb8e7090 | 1279 | } else { |
3de4586c | 1280 | struct inode *src_inode; |
72fd032e | 1281 | src_file = fget(fd); |
3de4586c CM |
1282 | if (!src_file) { |
1283 | ret = -EINVAL; | |
1284 | goto out; | |
1285 | } | |
1286 | ||
1287 | src_inode = src_file->f_path.dentry->d_inode; | |
1288 | if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) { | |
d397712b CM |
1289 | printk(KERN_INFO "btrfs: Snapshot src from " |
1290 | "another FS\n"); | |
3de4586c CM |
1291 | ret = -EINVAL; |
1292 | fput(src_file); | |
1293 | goto out; | |
1294 | } | |
72fd032e SW |
1295 | ret = btrfs_mksubvol(&file->f_path, name, namelen, |
1296 | BTRFS_I(src_inode)->root, | |
b83cc969 | 1297 | transid, readonly); |
3de4586c | 1298 | fput(src_file); |
cb8e7090 | 1299 | } |
f46b5a66 | 1300 | out: |
72fd032e SW |
1301 | return ret; |
1302 | } | |
1303 | ||
1304 | static noinline int btrfs_ioctl_snap_create(struct file *file, | |
fa0d2b9b | 1305 | void __user *arg, int subvol) |
72fd032e | 1306 | { |
fa0d2b9b | 1307 | struct btrfs_ioctl_vol_args *vol_args; |
72fd032e SW |
1308 | int ret; |
1309 | ||
fa0d2b9b LZ |
1310 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1311 | if (IS_ERR(vol_args)) | |
1312 | return PTR_ERR(vol_args); | |
1313 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
72fd032e | 1314 | |
fa0d2b9b | 1315 | ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, |
b83cc969 LZ |
1316 | vol_args->fd, subvol, |
1317 | NULL, false); | |
fdfb1e4f | 1318 | |
fa0d2b9b LZ |
1319 | kfree(vol_args); |
1320 | return ret; | |
1321 | } | |
fdfb1e4f | 1322 | |
fa0d2b9b LZ |
1323 | static noinline int btrfs_ioctl_snap_create_v2(struct file *file, |
1324 | void __user *arg, int subvol) | |
1325 | { | |
1326 | struct btrfs_ioctl_vol_args_v2 *vol_args; | |
1327 | int ret; | |
1328 | u64 transid = 0; | |
1329 | u64 *ptr = NULL; | |
b83cc969 | 1330 | bool readonly = false; |
75eaa0e2 | 1331 | |
fa0d2b9b LZ |
1332 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1333 | if (IS_ERR(vol_args)) | |
1334 | return PTR_ERR(vol_args); | |
1335 | vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; | |
75eaa0e2 | 1336 | |
b83cc969 LZ |
1337 | if (vol_args->flags & |
1338 | ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) { | |
1339 | ret = -EOPNOTSUPP; | |
fa0d2b9b | 1340 | goto out; |
72fd032e | 1341 | } |
fa0d2b9b LZ |
1342 | |
1343 | if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) | |
1344 | ptr = &transid; | |
b83cc969 LZ |
1345 | if (vol_args->flags & BTRFS_SUBVOL_RDONLY) |
1346 | readonly = true; | |
fa0d2b9b LZ |
1347 | |
1348 | ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, | |
b83cc969 LZ |
1349 | vol_args->fd, subvol, |
1350 | ptr, readonly); | |
fa0d2b9b LZ |
1351 | |
1352 | if (ret == 0 && ptr && | |
1353 | copy_to_user(arg + | |
1354 | offsetof(struct btrfs_ioctl_vol_args_v2, | |
1355 | transid), ptr, sizeof(*ptr))) | |
1356 | ret = -EFAULT; | |
fdfb1e4f | 1357 | out: |
f46b5a66 CH |
1358 | kfree(vol_args); |
1359 | return ret; | |
1360 | } | |
1361 | ||
0caa102d LZ |
1362 | static noinline int btrfs_ioctl_subvol_getflags(struct file *file, |
1363 | void __user *arg) | |
1364 | { | |
1365 | struct inode *inode = fdentry(file)->d_inode; | |
1366 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
1367 | int ret = 0; | |
1368 | u64 flags = 0; | |
1369 | ||
33345d01 | 1370 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) |
0caa102d LZ |
1371 | return -EINVAL; |
1372 | ||
1373 | down_read(&root->fs_info->subvol_sem); | |
1374 | if (btrfs_root_readonly(root)) | |
1375 | flags |= BTRFS_SUBVOL_RDONLY; | |
1376 | up_read(&root->fs_info->subvol_sem); | |
1377 | ||
1378 | if (copy_to_user(arg, &flags, sizeof(flags))) | |
1379 | ret = -EFAULT; | |
1380 | ||
1381 | return ret; | |
1382 | } | |
1383 | ||
1384 | static noinline int btrfs_ioctl_subvol_setflags(struct file *file, | |
1385 | void __user *arg) | |
1386 | { | |
1387 | struct inode *inode = fdentry(file)->d_inode; | |
1388 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
1389 | struct btrfs_trans_handle *trans; | |
1390 | u64 root_flags; | |
1391 | u64 flags; | |
1392 | int ret = 0; | |
1393 | ||
1394 | if (root->fs_info->sb->s_flags & MS_RDONLY) | |
1395 | return -EROFS; | |
1396 | ||
33345d01 | 1397 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) |
0caa102d LZ |
1398 | return -EINVAL; |
1399 | ||
1400 | if (copy_from_user(&flags, arg, sizeof(flags))) | |
1401 | return -EFAULT; | |
1402 | ||
b4dc2b8c | 1403 | if (flags & BTRFS_SUBVOL_CREATE_ASYNC) |
0caa102d LZ |
1404 | return -EINVAL; |
1405 | ||
1406 | if (flags & ~BTRFS_SUBVOL_RDONLY) | |
1407 | return -EOPNOTSUPP; | |
1408 | ||
2e149670 | 1409 | if (!inode_owner_or_capable(inode)) |
b4dc2b8c LZ |
1410 | return -EACCES; |
1411 | ||
0caa102d LZ |
1412 | down_write(&root->fs_info->subvol_sem); |
1413 | ||
1414 | /* nothing to do */ | |
1415 | if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) | |
1416 | goto out; | |
1417 | ||
1418 | root_flags = btrfs_root_flags(&root->root_item); | |
1419 | if (flags & BTRFS_SUBVOL_RDONLY) | |
1420 | btrfs_set_root_flags(&root->root_item, | |
1421 | root_flags | BTRFS_ROOT_SUBVOL_RDONLY); | |
1422 | else | |
1423 | btrfs_set_root_flags(&root->root_item, | |
1424 | root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); | |
1425 | ||
1426 | trans = btrfs_start_transaction(root, 1); | |
1427 | if (IS_ERR(trans)) { | |
1428 | ret = PTR_ERR(trans); | |
1429 | goto out_reset; | |
1430 | } | |
1431 | ||
b4dc2b8c | 1432 | ret = btrfs_update_root(trans, root->fs_info->tree_root, |
0caa102d LZ |
1433 | &root->root_key, &root->root_item); |
1434 | ||
1435 | btrfs_commit_transaction(trans, root); | |
1436 | out_reset: | |
1437 | if (ret) | |
1438 | btrfs_set_root_flags(&root->root_item, root_flags); | |
1439 | out: | |
1440 | up_write(&root->fs_info->subvol_sem); | |
1441 | return ret; | |
1442 | } | |
1443 | ||
76dda93c YZ |
1444 | /* |
1445 | * helper to check if the subvolume references other subvolumes | |
1446 | */ | |
1447 | static noinline int may_destroy_subvol(struct btrfs_root *root) | |
1448 | { | |
1449 | struct btrfs_path *path; | |
1450 | struct btrfs_key key; | |
1451 | int ret; | |
1452 | ||
1453 | path = btrfs_alloc_path(); | |
1454 | if (!path) | |
1455 | return -ENOMEM; | |
1456 | ||
1457 | key.objectid = root->root_key.objectid; | |
1458 | key.type = BTRFS_ROOT_REF_KEY; | |
1459 | key.offset = (u64)-1; | |
1460 | ||
1461 | ret = btrfs_search_slot(NULL, root->fs_info->tree_root, | |
1462 | &key, path, 0, 0); | |
1463 | if (ret < 0) | |
1464 | goto out; | |
1465 | BUG_ON(ret == 0); | |
1466 | ||
1467 | ret = 0; | |
1468 | if (path->slots[0] > 0) { | |
1469 | path->slots[0]--; | |
1470 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
1471 | if (key.objectid == root->root_key.objectid && | |
1472 | key.type == BTRFS_ROOT_REF_KEY) | |
1473 | ret = -ENOTEMPTY; | |
1474 | } | |
1475 | out: | |
1476 | btrfs_free_path(path); | |
1477 | return ret; | |
1478 | } | |
1479 | ||
ac8e9819 CM |
1480 | static noinline int key_in_sk(struct btrfs_key *key, |
1481 | struct btrfs_ioctl_search_key *sk) | |
1482 | { | |
abc6e134 CM |
1483 | struct btrfs_key test; |
1484 | int ret; | |
1485 | ||
1486 | test.objectid = sk->min_objectid; | |
1487 | test.type = sk->min_type; | |
1488 | test.offset = sk->min_offset; | |
1489 | ||
1490 | ret = btrfs_comp_cpu_keys(key, &test); | |
1491 | if (ret < 0) | |
ac8e9819 | 1492 | return 0; |
abc6e134 CM |
1493 | |
1494 | test.objectid = sk->max_objectid; | |
1495 | test.type = sk->max_type; | |
1496 | test.offset = sk->max_offset; | |
1497 | ||
1498 | ret = btrfs_comp_cpu_keys(key, &test); | |
1499 | if (ret > 0) | |
ac8e9819 CM |
1500 | return 0; |
1501 | return 1; | |
1502 | } | |
1503 | ||
1504 | static noinline int copy_to_sk(struct btrfs_root *root, | |
1505 | struct btrfs_path *path, | |
1506 | struct btrfs_key *key, | |
1507 | struct btrfs_ioctl_search_key *sk, | |
1508 | char *buf, | |
1509 | unsigned long *sk_offset, | |
1510 | int *num_found) | |
1511 | { | |
1512 | u64 found_transid; | |
1513 | struct extent_buffer *leaf; | |
1514 | struct btrfs_ioctl_search_header sh; | |
1515 | unsigned long item_off; | |
1516 | unsigned long item_len; | |
1517 | int nritems; | |
1518 | int i; | |
1519 | int slot; | |
ac8e9819 CM |
1520 | int ret = 0; |
1521 | ||
1522 | leaf = path->nodes[0]; | |
1523 | slot = path->slots[0]; | |
1524 | nritems = btrfs_header_nritems(leaf); | |
1525 | ||
1526 | if (btrfs_header_generation(leaf) > sk->max_transid) { | |
1527 | i = nritems; | |
1528 | goto advance_key; | |
1529 | } | |
1530 | found_transid = btrfs_header_generation(leaf); | |
1531 | ||
1532 | for (i = slot; i < nritems; i++) { | |
1533 | item_off = btrfs_item_ptr_offset(leaf, i); | |
1534 | item_len = btrfs_item_size_nr(leaf, i); | |
1535 | ||
1536 | if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE) | |
1537 | item_len = 0; | |
1538 | ||
1539 | if (sizeof(sh) + item_len + *sk_offset > | |
1540 | BTRFS_SEARCH_ARGS_BUFSIZE) { | |
1541 | ret = 1; | |
1542 | goto overflow; | |
1543 | } | |
1544 | ||
1545 | btrfs_item_key_to_cpu(leaf, key, i); | |
1546 | if (!key_in_sk(key, sk)) | |
1547 | continue; | |
1548 | ||
1549 | sh.objectid = key->objectid; | |
1550 | sh.offset = key->offset; | |
1551 | sh.type = key->type; | |
1552 | sh.len = item_len; | |
1553 | sh.transid = found_transid; | |
1554 | ||
1555 | /* copy search result header */ | |
1556 | memcpy(buf + *sk_offset, &sh, sizeof(sh)); | |
1557 | *sk_offset += sizeof(sh); | |
1558 | ||
1559 | if (item_len) { | |
1560 | char *p = buf + *sk_offset; | |
1561 | /* copy the item */ | |
1562 | read_extent_buffer(leaf, p, | |
1563 | item_off, item_len); | |
1564 | *sk_offset += item_len; | |
ac8e9819 | 1565 | } |
e2156867 | 1566 | (*num_found)++; |
ac8e9819 CM |
1567 | |
1568 | if (*num_found >= sk->nr_items) | |
1569 | break; | |
1570 | } | |
1571 | advance_key: | |
abc6e134 CM |
1572 | ret = 0; |
1573 | if (key->offset < (u64)-1 && key->offset < sk->max_offset) | |
ac8e9819 | 1574 | key->offset++; |
abc6e134 CM |
1575 | else if (key->type < (u8)-1 && key->type < sk->max_type) { |
1576 | key->offset = 0; | |
ac8e9819 | 1577 | key->type++; |
abc6e134 CM |
1578 | } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) { |
1579 | key->offset = 0; | |
1580 | key->type = 0; | |
ac8e9819 | 1581 | key->objectid++; |
abc6e134 CM |
1582 | } else |
1583 | ret = 1; | |
ac8e9819 | 1584 | overflow: |
ac8e9819 CM |
1585 | return ret; |
1586 | } | |
1587 | ||
1588 | static noinline int search_ioctl(struct inode *inode, | |
1589 | struct btrfs_ioctl_search_args *args) | |
1590 | { | |
1591 | struct btrfs_root *root; | |
1592 | struct btrfs_key key; | |
1593 | struct btrfs_key max_key; | |
1594 | struct btrfs_path *path; | |
1595 | struct btrfs_ioctl_search_key *sk = &args->key; | |
1596 | struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; | |
1597 | int ret; | |
1598 | int num_found = 0; | |
1599 | unsigned long sk_offset = 0; | |
1600 | ||
1601 | path = btrfs_alloc_path(); | |
1602 | if (!path) | |
1603 | return -ENOMEM; | |
1604 | ||
1605 | if (sk->tree_id == 0) { | |
1606 | /* search the root of the inode that was passed */ | |
1607 | root = BTRFS_I(inode)->root; | |
1608 | } else { | |
1609 | key.objectid = sk->tree_id; | |
1610 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1611 | key.offset = (u64)-1; | |
1612 | root = btrfs_read_fs_root_no_name(info, &key); | |
1613 | if (IS_ERR(root)) { | |
1614 | printk(KERN_ERR "could not find root %llu\n", | |
1615 | sk->tree_id); | |
1616 | btrfs_free_path(path); | |
1617 | return -ENOENT; | |
1618 | } | |
1619 | } | |
1620 | ||
1621 | key.objectid = sk->min_objectid; | |
1622 | key.type = sk->min_type; | |
1623 | key.offset = sk->min_offset; | |
1624 | ||
1625 | max_key.objectid = sk->max_objectid; | |
1626 | max_key.type = sk->max_type; | |
1627 | max_key.offset = sk->max_offset; | |
1628 | ||
1629 | path->keep_locks = 1; | |
1630 | ||
1631 | while(1) { | |
1632 | ret = btrfs_search_forward(root, &key, &max_key, path, 0, | |
1633 | sk->min_transid); | |
1634 | if (ret != 0) { | |
1635 | if (ret > 0) | |
1636 | ret = 0; | |
1637 | goto err; | |
1638 | } | |
1639 | ret = copy_to_sk(root, path, &key, sk, args->buf, | |
1640 | &sk_offset, &num_found); | |
b3b4aa74 | 1641 | btrfs_release_path(path); |
ac8e9819 CM |
1642 | if (ret || num_found >= sk->nr_items) |
1643 | break; | |
1644 | ||
1645 | } | |
1646 | ret = 0; | |
1647 | err: | |
1648 | sk->nr_items = num_found; | |
1649 | btrfs_free_path(path); | |
1650 | return ret; | |
1651 | } | |
1652 | ||
1653 | static noinline int btrfs_ioctl_tree_search(struct file *file, | |
1654 | void __user *argp) | |
1655 | { | |
1656 | struct btrfs_ioctl_search_args *args; | |
1657 | struct inode *inode; | |
1658 | int ret; | |
1659 | ||
1660 | if (!capable(CAP_SYS_ADMIN)) | |
1661 | return -EPERM; | |
1662 | ||
2354d08f JL |
1663 | args = memdup_user(argp, sizeof(*args)); |
1664 | if (IS_ERR(args)) | |
1665 | return PTR_ERR(args); | |
ac8e9819 | 1666 | |
ac8e9819 CM |
1667 | inode = fdentry(file)->d_inode; |
1668 | ret = search_ioctl(inode, args); | |
1669 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) | |
1670 | ret = -EFAULT; | |
1671 | kfree(args); | |
1672 | return ret; | |
1673 | } | |
1674 | ||
98d377a0 | 1675 | /* |
ac8e9819 CM |
1676 | * Search INODE_REFs to identify path name of 'dirid' directory |
1677 | * in a 'tree_id' tree. and sets path name to 'name'. | |
1678 | */ | |
98d377a0 TH |
1679 | static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, |
1680 | u64 tree_id, u64 dirid, char *name) | |
1681 | { | |
1682 | struct btrfs_root *root; | |
1683 | struct btrfs_key key; | |
ac8e9819 | 1684 | char *ptr; |
98d377a0 TH |
1685 | int ret = -1; |
1686 | int slot; | |
1687 | int len; | |
1688 | int total_len = 0; | |
1689 | struct btrfs_inode_ref *iref; | |
1690 | struct extent_buffer *l; | |
1691 | struct btrfs_path *path; | |
1692 | ||
1693 | if (dirid == BTRFS_FIRST_FREE_OBJECTID) { | |
1694 | name[0]='\0'; | |
1695 | return 0; | |
1696 | } | |
1697 | ||
1698 | path = btrfs_alloc_path(); | |
1699 | if (!path) | |
1700 | return -ENOMEM; | |
1701 | ||
ac8e9819 | 1702 | ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; |
98d377a0 TH |
1703 | |
1704 | key.objectid = tree_id; | |
1705 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1706 | key.offset = (u64)-1; | |
1707 | root = btrfs_read_fs_root_no_name(info, &key); | |
1708 | if (IS_ERR(root)) { | |
1709 | printk(KERN_ERR "could not find root %llu\n", tree_id); | |
8ad6fcab CM |
1710 | ret = -ENOENT; |
1711 | goto out; | |
98d377a0 TH |
1712 | } |
1713 | ||
1714 | key.objectid = dirid; | |
1715 | key.type = BTRFS_INODE_REF_KEY; | |
8ad6fcab | 1716 | key.offset = (u64)-1; |
98d377a0 TH |
1717 | |
1718 | while(1) { | |
1719 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
1720 | if (ret < 0) | |
1721 | goto out; | |
1722 | ||
1723 | l = path->nodes[0]; | |
1724 | slot = path->slots[0]; | |
8ad6fcab CM |
1725 | if (ret > 0 && slot > 0) |
1726 | slot--; | |
98d377a0 TH |
1727 | btrfs_item_key_to_cpu(l, &key, slot); |
1728 | ||
1729 | if (ret > 0 && (key.objectid != dirid || | |
ac8e9819 CM |
1730 | key.type != BTRFS_INODE_REF_KEY)) { |
1731 | ret = -ENOENT; | |
98d377a0 | 1732 | goto out; |
ac8e9819 | 1733 | } |
98d377a0 TH |
1734 | |
1735 | iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); | |
1736 | len = btrfs_inode_ref_name_len(l, iref); | |
1737 | ptr -= len + 1; | |
1738 | total_len += len + 1; | |
ac8e9819 | 1739 | if (ptr < name) |
98d377a0 TH |
1740 | goto out; |
1741 | ||
1742 | *(ptr + len) = '/'; | |
1743 | read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len); | |
1744 | ||
1745 | if (key.offset == BTRFS_FIRST_FREE_OBJECTID) | |
1746 | break; | |
1747 | ||
b3b4aa74 | 1748 | btrfs_release_path(path); |
98d377a0 | 1749 | key.objectid = key.offset; |
8ad6fcab | 1750 | key.offset = (u64)-1; |
98d377a0 | 1751 | dirid = key.objectid; |
98d377a0 | 1752 | } |
ac8e9819 | 1753 | if (ptr < name) |
98d377a0 | 1754 | goto out; |
77906a50 | 1755 | memmove(name, ptr, total_len); |
98d377a0 TH |
1756 | name[total_len]='\0'; |
1757 | ret = 0; | |
1758 | out: | |
1759 | btrfs_free_path(path); | |
ac8e9819 CM |
1760 | return ret; |
1761 | } | |
1762 | ||
1763 | static noinline int btrfs_ioctl_ino_lookup(struct file *file, | |
1764 | void __user *argp) | |
1765 | { | |
1766 | struct btrfs_ioctl_ino_lookup_args *args; | |
1767 | struct inode *inode; | |
1768 | int ret; | |
1769 | ||
1770 | if (!capable(CAP_SYS_ADMIN)) | |
1771 | return -EPERM; | |
1772 | ||
2354d08f JL |
1773 | args = memdup_user(argp, sizeof(*args)); |
1774 | if (IS_ERR(args)) | |
1775 | return PTR_ERR(args); | |
c2b96929 | 1776 | |
ac8e9819 CM |
1777 | inode = fdentry(file)->d_inode; |
1778 | ||
1b53ac4d CM |
1779 | if (args->treeid == 0) |
1780 | args->treeid = BTRFS_I(inode)->root->root_key.objectid; | |
1781 | ||
ac8e9819 CM |
1782 | ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, |
1783 | args->treeid, args->objectid, | |
1784 | args->name); | |
1785 | ||
1786 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) | |
1787 | ret = -EFAULT; | |
1788 | ||
1789 | kfree(args); | |
98d377a0 TH |
1790 | return ret; |
1791 | } | |
1792 | ||
76dda93c YZ |
1793 | static noinline int btrfs_ioctl_snap_destroy(struct file *file, |
1794 | void __user *arg) | |
1795 | { | |
1796 | struct dentry *parent = fdentry(file); | |
1797 | struct dentry *dentry; | |
1798 | struct inode *dir = parent->d_inode; | |
1799 | struct inode *inode; | |
1800 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
1801 | struct btrfs_root *dest = NULL; | |
1802 | struct btrfs_ioctl_vol_args *vol_args; | |
1803 | struct btrfs_trans_handle *trans; | |
1804 | int namelen; | |
1805 | int ret; | |
1806 | int err = 0; | |
1807 | ||
76dda93c YZ |
1808 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1809 | if (IS_ERR(vol_args)) | |
1810 | return PTR_ERR(vol_args); | |
1811 | ||
1812 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
1813 | namelen = strlen(vol_args->name); | |
1814 | if (strchr(vol_args->name, '/') || | |
1815 | strncmp(vol_args->name, "..", namelen) == 0) { | |
1816 | err = -EINVAL; | |
1817 | goto out; | |
1818 | } | |
1819 | ||
1820 | err = mnt_want_write(file->f_path.mnt); | |
1821 | if (err) | |
1822 | goto out; | |
1823 | ||
1824 | mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); | |
1825 | dentry = lookup_one_len(vol_args->name, parent, namelen); | |
1826 | if (IS_ERR(dentry)) { | |
1827 | err = PTR_ERR(dentry); | |
1828 | goto out_unlock_dir; | |
1829 | } | |
1830 | ||
1831 | if (!dentry->d_inode) { | |
1832 | err = -ENOENT; | |
1833 | goto out_dput; | |
1834 | } | |
1835 | ||
1836 | inode = dentry->d_inode; | |
4260f7c7 SW |
1837 | dest = BTRFS_I(inode)->root; |
1838 | if (!capable(CAP_SYS_ADMIN)){ | |
1839 | /* | |
1840 | * Regular user. Only allow this with a special mount | |
1841 | * option, when the user has write+exec access to the | |
1842 | * subvol root, and when rmdir(2) would have been | |
1843 | * allowed. | |
1844 | * | |
1845 | * Note that this is _not_ check that the subvol is | |
1846 | * empty or doesn't contain data that we wouldn't | |
1847 | * otherwise be able to delete. | |
1848 | * | |
1849 | * Users who want to delete empty subvols should try | |
1850 | * rmdir(2). | |
1851 | */ | |
1852 | err = -EPERM; | |
1853 | if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) | |
1854 | goto out_dput; | |
1855 | ||
1856 | /* | |
1857 | * Do not allow deletion if the parent dir is the same | |
1858 | * as the dir to be deleted. That means the ioctl | |
1859 | * must be called on the dentry referencing the root | |
1860 | * of the subvol, not a random directory contained | |
1861 | * within it. | |
1862 | */ | |
1863 | err = -EINVAL; | |
1864 | if (root == dest) | |
1865 | goto out_dput; | |
1866 | ||
1867 | err = inode_permission(inode, MAY_WRITE | MAY_EXEC); | |
1868 | if (err) | |
1869 | goto out_dput; | |
1870 | ||
1871 | /* check if subvolume may be deleted by a non-root user */ | |
1872 | err = btrfs_may_delete(dir, dentry, 1); | |
1873 | if (err) | |
1874 | goto out_dput; | |
1875 | } | |
1876 | ||
33345d01 | 1877 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { |
76dda93c YZ |
1878 | err = -EINVAL; |
1879 | goto out_dput; | |
1880 | } | |
1881 | ||
76dda93c YZ |
1882 | mutex_lock(&inode->i_mutex); |
1883 | err = d_invalidate(dentry); | |
1884 | if (err) | |
1885 | goto out_unlock; | |
1886 | ||
1887 | down_write(&root->fs_info->subvol_sem); | |
1888 | ||
1889 | err = may_destroy_subvol(dest); | |
1890 | if (err) | |
1891 | goto out_up_write; | |
1892 | ||
a22285a6 YZ |
1893 | trans = btrfs_start_transaction(root, 0); |
1894 | if (IS_ERR(trans)) { | |
1895 | err = PTR_ERR(trans); | |
d327099a | 1896 | goto out_up_write; |
a22285a6 YZ |
1897 | } |
1898 | trans->block_rsv = &root->fs_info->global_block_rsv; | |
1899 | ||
76dda93c YZ |
1900 | ret = btrfs_unlink_subvol(trans, root, dir, |
1901 | dest->root_key.objectid, | |
1902 | dentry->d_name.name, | |
1903 | dentry->d_name.len); | |
1904 | BUG_ON(ret); | |
1905 | ||
1906 | btrfs_record_root_in_trans(trans, dest); | |
1907 | ||
1908 | memset(&dest->root_item.drop_progress, 0, | |
1909 | sizeof(dest->root_item.drop_progress)); | |
1910 | dest->root_item.drop_level = 0; | |
1911 | btrfs_set_root_refs(&dest->root_item, 0); | |
1912 | ||
d68fc57b YZ |
1913 | if (!xchg(&dest->orphan_item_inserted, 1)) { |
1914 | ret = btrfs_insert_orphan_item(trans, | |
1915 | root->fs_info->tree_root, | |
1916 | dest->root_key.objectid); | |
1917 | BUG_ON(ret); | |
1918 | } | |
76dda93c | 1919 | |
531cb13f | 1920 | ret = btrfs_end_transaction(trans, root); |
76dda93c YZ |
1921 | BUG_ON(ret); |
1922 | inode->i_flags |= S_DEAD; | |
1923 | out_up_write: | |
1924 | up_write(&root->fs_info->subvol_sem); | |
1925 | out_unlock: | |
1926 | mutex_unlock(&inode->i_mutex); | |
1927 | if (!err) { | |
efefb143 | 1928 | shrink_dcache_sb(root->fs_info->sb); |
76dda93c YZ |
1929 | btrfs_invalidate_inodes(dest); |
1930 | d_delete(dentry); | |
1931 | } | |
1932 | out_dput: | |
1933 | dput(dentry); | |
1934 | out_unlock_dir: | |
1935 | mutex_unlock(&dir->i_mutex); | |
1936 | mnt_drop_write(file->f_path.mnt); | |
1937 | out: | |
1938 | kfree(vol_args); | |
1939 | return err; | |
1940 | } | |
1941 | ||
1e701a32 | 1942 | static int btrfs_ioctl_defrag(struct file *file, void __user *argp) |
f46b5a66 CH |
1943 | { |
1944 | struct inode *inode = fdentry(file)->d_inode; | |
1945 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
1e701a32 | 1946 | struct btrfs_ioctl_defrag_range_args *range; |
c146afad YZ |
1947 | int ret; |
1948 | ||
b83cc969 LZ |
1949 | if (btrfs_root_readonly(root)) |
1950 | return -EROFS; | |
1951 | ||
c146afad YZ |
1952 | ret = mnt_want_write(file->f_path.mnt); |
1953 | if (ret) | |
1954 | return ret; | |
f46b5a66 CH |
1955 | |
1956 | switch (inode->i_mode & S_IFMT) { | |
1957 | case S_IFDIR: | |
e441d54d CM |
1958 | if (!capable(CAP_SYS_ADMIN)) { |
1959 | ret = -EPERM; | |
1960 | goto out; | |
1961 | } | |
8929ecfa YZ |
1962 | ret = btrfs_defrag_root(root, 0); |
1963 | if (ret) | |
1964 | goto out; | |
1965 | ret = btrfs_defrag_root(root->fs_info->extent_root, 0); | |
f46b5a66 CH |
1966 | break; |
1967 | case S_IFREG: | |
e441d54d CM |
1968 | if (!(file->f_mode & FMODE_WRITE)) { |
1969 | ret = -EINVAL; | |
1970 | goto out; | |
1971 | } | |
1e701a32 CM |
1972 | |
1973 | range = kzalloc(sizeof(*range), GFP_KERNEL); | |
1974 | if (!range) { | |
1975 | ret = -ENOMEM; | |
1976 | goto out; | |
1977 | } | |
1978 | ||
1979 | if (argp) { | |
1980 | if (copy_from_user(range, argp, | |
1981 | sizeof(*range))) { | |
1982 | ret = -EFAULT; | |
1983 | kfree(range); | |
683be16e | 1984 | goto out; |
1e701a32 CM |
1985 | } |
1986 | /* compression requires us to start the IO */ | |
1987 | if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
1988 | range->flags |= BTRFS_DEFRAG_RANGE_START_IO; | |
1989 | range->extent_thresh = (u32)-1; | |
1990 | } | |
1991 | } else { | |
1992 | /* the rest are all set to zero by kzalloc */ | |
1993 | range->len = (u64)-1; | |
1994 | } | |
4cb5300b CM |
1995 | ret = btrfs_defrag_file(fdentry(file)->d_inode, file, |
1996 | range, 0, 0); | |
1997 | if (ret > 0) | |
1998 | ret = 0; | |
1e701a32 | 1999 | kfree(range); |
f46b5a66 | 2000 | break; |
8929ecfa YZ |
2001 | default: |
2002 | ret = -EINVAL; | |
f46b5a66 | 2003 | } |
e441d54d | 2004 | out: |
ab67b7c1 | 2005 | mnt_drop_write(file->f_path.mnt); |
e441d54d | 2006 | return ret; |
f46b5a66 CH |
2007 | } |
2008 | ||
b2950863 | 2009 | static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) |
f46b5a66 CH |
2010 | { |
2011 | struct btrfs_ioctl_vol_args *vol_args; | |
2012 | int ret; | |
2013 | ||
e441d54d CM |
2014 | if (!capable(CAP_SYS_ADMIN)) |
2015 | return -EPERM; | |
2016 | ||
dae7b665 LZ |
2017 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
2018 | if (IS_ERR(vol_args)) | |
2019 | return PTR_ERR(vol_args); | |
f46b5a66 | 2020 | |
5516e595 | 2021 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 CH |
2022 | ret = btrfs_init_new_device(root, vol_args->name); |
2023 | ||
f46b5a66 CH |
2024 | kfree(vol_args); |
2025 | return ret; | |
2026 | } | |
2027 | ||
b2950863 | 2028 | static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) |
f46b5a66 CH |
2029 | { |
2030 | struct btrfs_ioctl_vol_args *vol_args; | |
2031 | int ret; | |
2032 | ||
e441d54d CM |
2033 | if (!capable(CAP_SYS_ADMIN)) |
2034 | return -EPERM; | |
2035 | ||
c146afad YZ |
2036 | if (root->fs_info->sb->s_flags & MS_RDONLY) |
2037 | return -EROFS; | |
2038 | ||
dae7b665 LZ |
2039 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
2040 | if (IS_ERR(vol_args)) | |
2041 | return PTR_ERR(vol_args); | |
f46b5a66 | 2042 | |
5516e595 | 2043 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 CH |
2044 | ret = btrfs_rm_device(root, vol_args->name); |
2045 | ||
f46b5a66 CH |
2046 | kfree(vol_args); |
2047 | return ret; | |
2048 | } | |
2049 | ||
475f6387 JS |
2050 | static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) |
2051 | { | |
027ed2f0 | 2052 | struct btrfs_ioctl_fs_info_args *fi_args; |
475f6387 JS |
2053 | struct btrfs_device *device; |
2054 | struct btrfs_device *next; | |
2055 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
027ed2f0 | 2056 | int ret = 0; |
475f6387 JS |
2057 | |
2058 | if (!capable(CAP_SYS_ADMIN)) | |
2059 | return -EPERM; | |
2060 | ||
027ed2f0 LZ |
2061 | fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); |
2062 | if (!fi_args) | |
2063 | return -ENOMEM; | |
2064 | ||
2065 | fi_args->num_devices = fs_devices->num_devices; | |
2066 | memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); | |
475f6387 JS |
2067 | |
2068 | mutex_lock(&fs_devices->device_list_mutex); | |
2069 | list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) { | |
027ed2f0 LZ |
2070 | if (device->devid > fi_args->max_id) |
2071 | fi_args->max_id = device->devid; | |
475f6387 JS |
2072 | } |
2073 | mutex_unlock(&fs_devices->device_list_mutex); | |
2074 | ||
027ed2f0 LZ |
2075 | if (copy_to_user(arg, fi_args, sizeof(*fi_args))) |
2076 | ret = -EFAULT; | |
475f6387 | 2077 | |
027ed2f0 LZ |
2078 | kfree(fi_args); |
2079 | return ret; | |
475f6387 JS |
2080 | } |
2081 | ||
2082 | static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) | |
2083 | { | |
2084 | struct btrfs_ioctl_dev_info_args *di_args; | |
2085 | struct btrfs_device *dev; | |
2086 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
2087 | int ret = 0; | |
2088 | char *s_uuid = NULL; | |
2089 | char empty_uuid[BTRFS_UUID_SIZE] = {0}; | |
2090 | ||
2091 | if (!capable(CAP_SYS_ADMIN)) | |
2092 | return -EPERM; | |
2093 | ||
2094 | di_args = memdup_user(arg, sizeof(*di_args)); | |
2095 | if (IS_ERR(di_args)) | |
2096 | return PTR_ERR(di_args); | |
2097 | ||
2098 | if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0) | |
2099 | s_uuid = di_args->uuid; | |
2100 | ||
2101 | mutex_lock(&fs_devices->device_list_mutex); | |
2102 | dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL); | |
2103 | mutex_unlock(&fs_devices->device_list_mutex); | |
2104 | ||
2105 | if (!dev) { | |
2106 | ret = -ENODEV; | |
2107 | goto out; | |
2108 | } | |
2109 | ||
2110 | di_args->devid = dev->devid; | |
2111 | di_args->bytes_used = dev->bytes_used; | |
2112 | di_args->total_bytes = dev->total_bytes; | |
2113 | memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); | |
2114 | strncpy(di_args->path, dev->name, sizeof(di_args->path)); | |
2115 | ||
2116 | out: | |
2117 | if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) | |
2118 | ret = -EFAULT; | |
2119 | ||
2120 | kfree(di_args); | |
2121 | return ret; | |
2122 | } | |
2123 | ||
76dda93c YZ |
2124 | static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, |
2125 | u64 off, u64 olen, u64 destoff) | |
f46b5a66 CH |
2126 | { |
2127 | struct inode *inode = fdentry(file)->d_inode; | |
2128 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2129 | struct file *src_file; | |
2130 | struct inode *src; | |
2131 | struct btrfs_trans_handle *trans; | |
f46b5a66 | 2132 | struct btrfs_path *path; |
f46b5a66 | 2133 | struct extent_buffer *leaf; |
ae01a0ab YZ |
2134 | char *buf; |
2135 | struct btrfs_key key; | |
f46b5a66 CH |
2136 | u32 nritems; |
2137 | int slot; | |
ae01a0ab | 2138 | int ret; |
c5c9cd4d SW |
2139 | u64 len = olen; |
2140 | u64 bs = root->fs_info->sb->s_blocksize; | |
2141 | u64 hint_byte; | |
d20f7043 | 2142 | |
c5c9cd4d SW |
2143 | /* |
2144 | * TODO: | |
2145 | * - split compressed inline extents. annoying: we need to | |
2146 | * decompress into destination's address_space (the file offset | |
2147 | * may change, so source mapping won't do), then recompress (or | |
2148 | * otherwise reinsert) a subrange. | |
2149 | * - allow ranges within the same file to be cloned (provided | |
2150 | * they don't overlap)? | |
2151 | */ | |
2152 | ||
e441d54d | 2153 | /* the destination must be opened for writing */ |
2ebc3464 | 2154 | if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) |
e441d54d CM |
2155 | return -EINVAL; |
2156 | ||
b83cc969 LZ |
2157 | if (btrfs_root_readonly(root)) |
2158 | return -EROFS; | |
2159 | ||
c146afad YZ |
2160 | ret = mnt_want_write(file->f_path.mnt); |
2161 | if (ret) | |
2162 | return ret; | |
2163 | ||
c5c9cd4d | 2164 | src_file = fget(srcfd); |
ab67b7c1 YZ |
2165 | if (!src_file) { |
2166 | ret = -EBADF; | |
2167 | goto out_drop_write; | |
2168 | } | |
5dc64164 | 2169 | |
f46b5a66 CH |
2170 | src = src_file->f_dentry->d_inode; |
2171 | ||
c5c9cd4d SW |
2172 | ret = -EINVAL; |
2173 | if (src == inode) | |
2174 | goto out_fput; | |
2175 | ||
5dc64164 DR |
2176 | /* the src must be open for reading */ |
2177 | if (!(src_file->f_mode & FMODE_READ)) | |
2178 | goto out_fput; | |
2179 | ||
ae01a0ab YZ |
2180 | ret = -EISDIR; |
2181 | if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) | |
2182 | goto out_fput; | |
2183 | ||
f46b5a66 | 2184 | ret = -EXDEV; |
ae01a0ab YZ |
2185 | if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root) |
2186 | goto out_fput; | |
2187 | ||
2188 | ret = -ENOMEM; | |
2189 | buf = vmalloc(btrfs_level_size(root, 0)); | |
2190 | if (!buf) | |
2191 | goto out_fput; | |
2192 | ||
2193 | path = btrfs_alloc_path(); | |
2194 | if (!path) { | |
2195 | vfree(buf); | |
f46b5a66 | 2196 | goto out_fput; |
ae01a0ab YZ |
2197 | } |
2198 | path->reada = 2; | |
f46b5a66 CH |
2199 | |
2200 | if (inode < src) { | |
fccdae43 SW |
2201 | mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT); |
2202 | mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD); | |
f46b5a66 | 2203 | } else { |
fccdae43 SW |
2204 | mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT); |
2205 | mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD); | |
f46b5a66 CH |
2206 | } |
2207 | ||
c5c9cd4d SW |
2208 | /* determine range to clone */ |
2209 | ret = -EINVAL; | |
2ebc3464 | 2210 | if (off + len > src->i_size || off + len < off) |
f46b5a66 | 2211 | goto out_unlock; |
c5c9cd4d SW |
2212 | if (len == 0) |
2213 | olen = len = src->i_size - off; | |
2214 | /* if we extend to eof, continue to block boundary */ | |
2215 | if (off + len == src->i_size) | |
2a6b8dae | 2216 | len = ALIGN(src->i_size, bs) - off; |
c5c9cd4d SW |
2217 | |
2218 | /* verify the end result is block aligned */ | |
2a6b8dae LZ |
2219 | if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || |
2220 | !IS_ALIGNED(destoff, bs)) | |
c5c9cd4d SW |
2221 | goto out_unlock; |
2222 | ||
f46b5a66 CH |
2223 | /* do any pending delalloc/csum calc on src, one way or |
2224 | another, and lock file content */ | |
2225 | while (1) { | |
31840ae1 | 2226 | struct btrfs_ordered_extent *ordered; |
c5c9cd4d | 2227 | lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); |
9a019196 SW |
2228 | ordered = btrfs_lookup_first_ordered_extent(src, off+len); |
2229 | if (!ordered && | |
2230 | !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len, | |
2231 | EXTENT_DELALLOC, 0, NULL)) | |
f46b5a66 | 2232 | break; |
c5c9cd4d | 2233 | unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); |
ae01a0ab YZ |
2234 | if (ordered) |
2235 | btrfs_put_ordered_extent(ordered); | |
9a019196 | 2236 | btrfs_wait_ordered_range(src, off, len); |
f46b5a66 CH |
2237 | } |
2238 | ||
c5c9cd4d | 2239 | /* clone data */ |
33345d01 | 2240 | key.objectid = btrfs_ino(src); |
ae01a0ab YZ |
2241 | key.type = BTRFS_EXTENT_DATA_KEY; |
2242 | key.offset = 0; | |
f46b5a66 CH |
2243 | |
2244 | while (1) { | |
2245 | /* | |
2246 | * note the key will change type as we walk through the | |
2247 | * tree. | |
2248 | */ | |
a22285a6 | 2249 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
f46b5a66 CH |
2250 | if (ret < 0) |
2251 | goto out; | |
2252 | ||
ae01a0ab YZ |
2253 | nritems = btrfs_header_nritems(path->nodes[0]); |
2254 | if (path->slots[0] >= nritems) { | |
f46b5a66 CH |
2255 | ret = btrfs_next_leaf(root, path); |
2256 | if (ret < 0) | |
2257 | goto out; | |
2258 | if (ret > 0) | |
2259 | break; | |
ae01a0ab | 2260 | nritems = btrfs_header_nritems(path->nodes[0]); |
f46b5a66 CH |
2261 | } |
2262 | leaf = path->nodes[0]; | |
2263 | slot = path->slots[0]; | |
f46b5a66 | 2264 | |
ae01a0ab | 2265 | btrfs_item_key_to_cpu(leaf, &key, slot); |
d20f7043 | 2266 | if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || |
33345d01 | 2267 | key.objectid != btrfs_ino(src)) |
f46b5a66 CH |
2268 | break; |
2269 | ||
c5c9cd4d SW |
2270 | if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { |
2271 | struct btrfs_file_extent_item *extent; | |
2272 | int type; | |
31840ae1 ZY |
2273 | u32 size; |
2274 | struct btrfs_key new_key; | |
c5c9cd4d SW |
2275 | u64 disko = 0, diskl = 0; |
2276 | u64 datao = 0, datal = 0; | |
2277 | u8 comp; | |
b5384d48 | 2278 | u64 endoff; |
31840ae1 ZY |
2279 | |
2280 | size = btrfs_item_size_nr(leaf, slot); | |
2281 | read_extent_buffer(leaf, buf, | |
2282 | btrfs_item_ptr_offset(leaf, slot), | |
2283 | size); | |
c5c9cd4d SW |
2284 | |
2285 | extent = btrfs_item_ptr(leaf, slot, | |
2286 | struct btrfs_file_extent_item); | |
2287 | comp = btrfs_file_extent_compression(leaf, extent); | |
2288 | type = btrfs_file_extent_type(leaf, extent); | |
c8a894d7 CM |
2289 | if (type == BTRFS_FILE_EXTENT_REG || |
2290 | type == BTRFS_FILE_EXTENT_PREALLOC) { | |
d397712b CM |
2291 | disko = btrfs_file_extent_disk_bytenr(leaf, |
2292 | extent); | |
2293 | diskl = btrfs_file_extent_disk_num_bytes(leaf, | |
2294 | extent); | |
c5c9cd4d | 2295 | datao = btrfs_file_extent_offset(leaf, extent); |
d397712b CM |
2296 | datal = btrfs_file_extent_num_bytes(leaf, |
2297 | extent); | |
c5c9cd4d SW |
2298 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
2299 | /* take upper bound, may be compressed */ | |
2300 | datal = btrfs_file_extent_ram_bytes(leaf, | |
2301 | extent); | |
2302 | } | |
b3b4aa74 | 2303 | btrfs_release_path(path); |
31840ae1 | 2304 | |
050006a7 | 2305 | if (key.offset + datal <= off || |
c5c9cd4d SW |
2306 | key.offset >= off+len) |
2307 | goto next; | |
2308 | ||
31840ae1 | 2309 | memcpy(&new_key, &key, sizeof(new_key)); |
33345d01 | 2310 | new_key.objectid = btrfs_ino(inode); |
4d728ec7 LZ |
2311 | if (off <= key.offset) |
2312 | new_key.offset = key.offset + destoff - off; | |
2313 | else | |
2314 | new_key.offset = destoff; | |
31840ae1 | 2315 | |
a22285a6 YZ |
2316 | trans = btrfs_start_transaction(root, 1); |
2317 | if (IS_ERR(trans)) { | |
2318 | ret = PTR_ERR(trans); | |
2319 | goto out; | |
2320 | } | |
2321 | ||
c8a894d7 CM |
2322 | if (type == BTRFS_FILE_EXTENT_REG || |
2323 | type == BTRFS_FILE_EXTENT_PREALLOC) { | |
a22285a6 YZ |
2324 | if (off > key.offset) { |
2325 | datao += off - key.offset; | |
2326 | datal -= off - key.offset; | |
2327 | } | |
2328 | ||
2329 | if (key.offset + datal > off + len) | |
2330 | datal = off + len - key.offset; | |
2331 | ||
2332 | ret = btrfs_drop_extents(trans, inode, | |
2333 | new_key.offset, | |
2334 | new_key.offset + datal, | |
2335 | &hint_byte, 1); | |
2336 | BUG_ON(ret); | |
2337 | ||
c5c9cd4d SW |
2338 | ret = btrfs_insert_empty_item(trans, root, path, |
2339 | &new_key, size); | |
a22285a6 | 2340 | BUG_ON(ret); |
c5c9cd4d SW |
2341 | |
2342 | leaf = path->nodes[0]; | |
2343 | slot = path->slots[0]; | |
2344 | write_extent_buffer(leaf, buf, | |
31840ae1 ZY |
2345 | btrfs_item_ptr_offset(leaf, slot), |
2346 | size); | |
ae01a0ab | 2347 | |
c5c9cd4d | 2348 | extent = btrfs_item_ptr(leaf, slot, |
f46b5a66 | 2349 | struct btrfs_file_extent_item); |
c5c9cd4d | 2350 | |
c5c9cd4d SW |
2351 | /* disko == 0 means it's a hole */ |
2352 | if (!disko) | |
2353 | datao = 0; | |
c5c9cd4d SW |
2354 | |
2355 | btrfs_set_file_extent_offset(leaf, extent, | |
2356 | datao); | |
2357 | btrfs_set_file_extent_num_bytes(leaf, extent, | |
2358 | datal); | |
2359 | if (disko) { | |
2360 | inode_add_bytes(inode, datal); | |
ae01a0ab | 2361 | ret = btrfs_inc_extent_ref(trans, root, |
5d4f98a2 YZ |
2362 | disko, diskl, 0, |
2363 | root->root_key.objectid, | |
33345d01 | 2364 | btrfs_ino(inode), |
5d4f98a2 | 2365 | new_key.offset - datao); |
31840ae1 | 2366 | BUG_ON(ret); |
f46b5a66 | 2367 | } |
c5c9cd4d SW |
2368 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
2369 | u64 skip = 0; | |
2370 | u64 trim = 0; | |
2371 | if (off > key.offset) { | |
2372 | skip = off - key.offset; | |
2373 | new_key.offset += skip; | |
2374 | } | |
d397712b | 2375 | |
c5c9cd4d SW |
2376 | if (key.offset + datal > off+len) |
2377 | trim = key.offset + datal - (off+len); | |
d397712b | 2378 | |
c5c9cd4d | 2379 | if (comp && (skip || trim)) { |
c5c9cd4d | 2380 | ret = -EINVAL; |
a22285a6 | 2381 | btrfs_end_transaction(trans, root); |
c5c9cd4d SW |
2382 | goto out; |
2383 | } | |
2384 | size -= skip + trim; | |
2385 | datal -= skip + trim; | |
a22285a6 YZ |
2386 | |
2387 | ret = btrfs_drop_extents(trans, inode, | |
2388 | new_key.offset, | |
2389 | new_key.offset + datal, | |
2390 | &hint_byte, 1); | |
2391 | BUG_ON(ret); | |
2392 | ||
c5c9cd4d SW |
2393 | ret = btrfs_insert_empty_item(trans, root, path, |
2394 | &new_key, size); | |
a22285a6 | 2395 | BUG_ON(ret); |
c5c9cd4d SW |
2396 | |
2397 | if (skip) { | |
d397712b CM |
2398 | u32 start = |
2399 | btrfs_file_extent_calc_inline_size(0); | |
c5c9cd4d SW |
2400 | memmove(buf+start, buf+start+skip, |
2401 | datal); | |
2402 | } | |
2403 | ||
2404 | leaf = path->nodes[0]; | |
2405 | slot = path->slots[0]; | |
2406 | write_extent_buffer(leaf, buf, | |
2407 | btrfs_item_ptr_offset(leaf, slot), | |
2408 | size); | |
2409 | inode_add_bytes(inode, datal); | |
f46b5a66 | 2410 | } |
c5c9cd4d SW |
2411 | |
2412 | btrfs_mark_buffer_dirty(leaf); | |
b3b4aa74 | 2413 | btrfs_release_path(path); |
c5c9cd4d | 2414 | |
a22285a6 | 2415 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; |
b5384d48 SW |
2416 | |
2417 | /* | |
2418 | * we round up to the block size at eof when | |
2419 | * determining which extents to clone above, | |
2420 | * but shouldn't round up the file size | |
2421 | */ | |
2422 | endoff = new_key.offset + datal; | |
5f3888ff LZ |
2423 | if (endoff > destoff+olen) |
2424 | endoff = destoff+olen; | |
b5384d48 SW |
2425 | if (endoff > inode->i_size) |
2426 | btrfs_i_size_write(inode, endoff); | |
2427 | ||
a22285a6 YZ |
2428 | BTRFS_I(inode)->flags = BTRFS_I(src)->flags; |
2429 | ret = btrfs_update_inode(trans, root, inode); | |
2430 | BUG_ON(ret); | |
2431 | btrfs_end_transaction(trans, root); | |
2432 | } | |
d397712b | 2433 | next: |
b3b4aa74 | 2434 | btrfs_release_path(path); |
f46b5a66 | 2435 | key.offset++; |
f46b5a66 | 2436 | } |
f46b5a66 CH |
2437 | ret = 0; |
2438 | out: | |
b3b4aa74 | 2439 | btrfs_release_path(path); |
c5c9cd4d | 2440 | unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); |
f46b5a66 CH |
2441 | out_unlock: |
2442 | mutex_unlock(&src->i_mutex); | |
2443 | mutex_unlock(&inode->i_mutex); | |
ae01a0ab YZ |
2444 | vfree(buf); |
2445 | btrfs_free_path(path); | |
f46b5a66 CH |
2446 | out_fput: |
2447 | fput(src_file); | |
ab67b7c1 YZ |
2448 | out_drop_write: |
2449 | mnt_drop_write(file->f_path.mnt); | |
f46b5a66 CH |
2450 | return ret; |
2451 | } | |
2452 | ||
7a865e8a | 2453 | static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) |
c5c9cd4d SW |
2454 | { |
2455 | struct btrfs_ioctl_clone_range_args args; | |
2456 | ||
7a865e8a | 2457 | if (copy_from_user(&args, argp, sizeof(args))) |
c5c9cd4d SW |
2458 | return -EFAULT; |
2459 | return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, | |
2460 | args.src_length, args.dest_offset); | |
2461 | } | |
2462 | ||
f46b5a66 CH |
2463 | /* |
2464 | * there are many ways the trans_start and trans_end ioctls can lead | |
2465 | * to deadlocks. They should only be used by applications that | |
2466 | * basically own the machine, and have a very in depth understanding | |
2467 | * of all the possible deadlocks and enospc problems. | |
2468 | */ | |
b2950863 | 2469 | static long btrfs_ioctl_trans_start(struct file *file) |
f46b5a66 CH |
2470 | { |
2471 | struct inode *inode = fdentry(file)->d_inode; | |
2472 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2473 | struct btrfs_trans_handle *trans; | |
1ab86aed | 2474 | int ret; |
f46b5a66 | 2475 | |
1ab86aed | 2476 | ret = -EPERM; |
df5b5520 | 2477 | if (!capable(CAP_SYS_ADMIN)) |
1ab86aed | 2478 | goto out; |
df5b5520 | 2479 | |
1ab86aed SW |
2480 | ret = -EINPROGRESS; |
2481 | if (file->private_data) | |
f46b5a66 | 2482 | goto out; |
9ca9ee09 | 2483 | |
b83cc969 LZ |
2484 | ret = -EROFS; |
2485 | if (btrfs_root_readonly(root)) | |
2486 | goto out; | |
2487 | ||
c146afad YZ |
2488 | ret = mnt_want_write(file->f_path.mnt); |
2489 | if (ret) | |
2490 | goto out; | |
2491 | ||
a4abeea4 | 2492 | atomic_inc(&root->fs_info->open_ioctl_trans); |
9ca9ee09 | 2493 | |
1ab86aed | 2494 | ret = -ENOMEM; |
7a7eaa40 | 2495 | trans = btrfs_start_ioctl_transaction(root); |
abd30bb0 | 2496 | if (IS_ERR(trans)) |
1ab86aed SW |
2497 | goto out_drop; |
2498 | ||
2499 | file->private_data = trans; | |
2500 | return 0; | |
2501 | ||
2502 | out_drop: | |
a4abeea4 | 2503 | atomic_dec(&root->fs_info->open_ioctl_trans); |
1ab86aed | 2504 | mnt_drop_write(file->f_path.mnt); |
f46b5a66 | 2505 | out: |
f46b5a66 CH |
2506 | return ret; |
2507 | } | |
2508 | ||
6ef5ed0d JB |
2509 | static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) |
2510 | { | |
2511 | struct inode *inode = fdentry(file)->d_inode; | |
2512 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2513 | struct btrfs_root *new_root; | |
2514 | struct btrfs_dir_item *di; | |
2515 | struct btrfs_trans_handle *trans; | |
2516 | struct btrfs_path *path; | |
2517 | struct btrfs_key location; | |
2518 | struct btrfs_disk_key disk_key; | |
2519 | struct btrfs_super_block *disk_super; | |
2520 | u64 features; | |
2521 | u64 objectid = 0; | |
2522 | u64 dir_id; | |
2523 | ||
2524 | if (!capable(CAP_SYS_ADMIN)) | |
2525 | return -EPERM; | |
2526 | ||
2527 | if (copy_from_user(&objectid, argp, sizeof(objectid))) | |
2528 | return -EFAULT; | |
2529 | ||
2530 | if (!objectid) | |
2531 | objectid = root->root_key.objectid; | |
2532 | ||
2533 | location.objectid = objectid; | |
2534 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2535 | location.offset = (u64)-1; | |
2536 | ||
2537 | new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); | |
2538 | if (IS_ERR(new_root)) | |
2539 | return PTR_ERR(new_root); | |
2540 | ||
2541 | if (btrfs_root_refs(&new_root->root_item) == 0) | |
2542 | return -ENOENT; | |
2543 | ||
2544 | path = btrfs_alloc_path(); | |
2545 | if (!path) | |
2546 | return -ENOMEM; | |
2547 | path->leave_spinning = 1; | |
2548 | ||
2549 | trans = btrfs_start_transaction(root, 1); | |
98d5dc13 | 2550 | if (IS_ERR(trans)) { |
6ef5ed0d | 2551 | btrfs_free_path(path); |
98d5dc13 | 2552 | return PTR_ERR(trans); |
6ef5ed0d JB |
2553 | } |
2554 | ||
2555 | dir_id = btrfs_super_root_dir(&root->fs_info->super_copy); | |
2556 | di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, | |
2557 | dir_id, "default", 7, 1); | |
cf1e99a4 | 2558 | if (IS_ERR_OR_NULL(di)) { |
6ef5ed0d JB |
2559 | btrfs_free_path(path); |
2560 | btrfs_end_transaction(trans, root); | |
2561 | printk(KERN_ERR "Umm, you don't have the default dir item, " | |
2562 | "this isn't going to work\n"); | |
2563 | return -ENOENT; | |
2564 | } | |
2565 | ||
2566 | btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); | |
2567 | btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); | |
2568 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
2569 | btrfs_free_path(path); | |
2570 | ||
2571 | disk_super = &root->fs_info->super_copy; | |
2572 | features = btrfs_super_incompat_flags(disk_super); | |
2573 | if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) { | |
2574 | features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL; | |
2575 | btrfs_set_super_incompat_flags(disk_super, features); | |
2576 | } | |
2577 | btrfs_end_transaction(trans, root); | |
2578 | ||
2579 | return 0; | |
2580 | } | |
2581 | ||
bf5fc093 JB |
2582 | static void get_block_group_info(struct list_head *groups_list, |
2583 | struct btrfs_ioctl_space_info *space) | |
2584 | { | |
2585 | struct btrfs_block_group_cache *block_group; | |
2586 | ||
2587 | space->total_bytes = 0; | |
2588 | space->used_bytes = 0; | |
2589 | space->flags = 0; | |
2590 | list_for_each_entry(block_group, groups_list, list) { | |
2591 | space->flags = block_group->flags; | |
2592 | space->total_bytes += block_group->key.offset; | |
2593 | space->used_bytes += | |
2594 | btrfs_block_group_used(&block_group->item); | |
2595 | } | |
2596 | } | |
2597 | ||
1406e432 JB |
2598 | long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) |
2599 | { | |
2600 | struct btrfs_ioctl_space_args space_args; | |
2601 | struct btrfs_ioctl_space_info space; | |
2602 | struct btrfs_ioctl_space_info *dest; | |
7fde62bf | 2603 | struct btrfs_ioctl_space_info *dest_orig; |
13f2696f | 2604 | struct btrfs_ioctl_space_info __user *user_dest; |
1406e432 | 2605 | struct btrfs_space_info *info; |
bf5fc093 JB |
2606 | u64 types[] = {BTRFS_BLOCK_GROUP_DATA, |
2607 | BTRFS_BLOCK_GROUP_SYSTEM, | |
2608 | BTRFS_BLOCK_GROUP_METADATA, | |
2609 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; | |
2610 | int num_types = 4; | |
7fde62bf | 2611 | int alloc_size; |
1406e432 | 2612 | int ret = 0; |
51788b1b | 2613 | u64 slot_count = 0; |
bf5fc093 | 2614 | int i, c; |
1406e432 JB |
2615 | |
2616 | if (copy_from_user(&space_args, | |
2617 | (struct btrfs_ioctl_space_args __user *)arg, | |
2618 | sizeof(space_args))) | |
2619 | return -EFAULT; | |
2620 | ||
bf5fc093 JB |
2621 | for (i = 0; i < num_types; i++) { |
2622 | struct btrfs_space_info *tmp; | |
2623 | ||
2624 | info = NULL; | |
2625 | rcu_read_lock(); | |
2626 | list_for_each_entry_rcu(tmp, &root->fs_info->space_info, | |
2627 | list) { | |
2628 | if (tmp->flags == types[i]) { | |
2629 | info = tmp; | |
2630 | break; | |
2631 | } | |
2632 | } | |
2633 | rcu_read_unlock(); | |
2634 | ||
2635 | if (!info) | |
2636 | continue; | |
2637 | ||
2638 | down_read(&info->groups_sem); | |
2639 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
2640 | if (!list_empty(&info->block_groups[c])) | |
2641 | slot_count++; | |
2642 | } | |
2643 | up_read(&info->groups_sem); | |
2644 | } | |
7fde62bf CM |
2645 | |
2646 | /* space_slots == 0 means they are asking for a count */ | |
2647 | if (space_args.space_slots == 0) { | |
2648 | space_args.total_spaces = slot_count; | |
2649 | goto out; | |
2650 | } | |
bf5fc093 | 2651 | |
51788b1b | 2652 | slot_count = min_t(u64, space_args.space_slots, slot_count); |
bf5fc093 | 2653 | |
7fde62bf | 2654 | alloc_size = sizeof(*dest) * slot_count; |
bf5fc093 | 2655 | |
7fde62bf CM |
2656 | /* we generally have at most 6 or so space infos, one for each raid |
2657 | * level. So, a whole page should be more than enough for everyone | |
2658 | */ | |
2659 | if (alloc_size > PAGE_CACHE_SIZE) | |
2660 | return -ENOMEM; | |
2661 | ||
1406e432 | 2662 | space_args.total_spaces = 0; |
7fde62bf CM |
2663 | dest = kmalloc(alloc_size, GFP_NOFS); |
2664 | if (!dest) | |
2665 | return -ENOMEM; | |
2666 | dest_orig = dest; | |
1406e432 | 2667 | |
7fde62bf | 2668 | /* now we have a buffer to copy into */ |
bf5fc093 JB |
2669 | for (i = 0; i < num_types; i++) { |
2670 | struct btrfs_space_info *tmp; | |
2671 | ||
51788b1b DR |
2672 | if (!slot_count) |
2673 | break; | |
2674 | ||
bf5fc093 JB |
2675 | info = NULL; |
2676 | rcu_read_lock(); | |
2677 | list_for_each_entry_rcu(tmp, &root->fs_info->space_info, | |
2678 | list) { | |
2679 | if (tmp->flags == types[i]) { | |
2680 | info = tmp; | |
2681 | break; | |
2682 | } | |
2683 | } | |
2684 | rcu_read_unlock(); | |
7fde62bf | 2685 | |
bf5fc093 JB |
2686 | if (!info) |
2687 | continue; | |
2688 | down_read(&info->groups_sem); | |
2689 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
2690 | if (!list_empty(&info->block_groups[c])) { | |
2691 | get_block_group_info(&info->block_groups[c], | |
2692 | &space); | |
2693 | memcpy(dest, &space, sizeof(space)); | |
2694 | dest++; | |
2695 | space_args.total_spaces++; | |
51788b1b | 2696 | slot_count--; |
bf5fc093 | 2697 | } |
51788b1b DR |
2698 | if (!slot_count) |
2699 | break; | |
bf5fc093 JB |
2700 | } |
2701 | up_read(&info->groups_sem); | |
1406e432 | 2702 | } |
1406e432 | 2703 | |
7fde62bf CM |
2704 | user_dest = (struct btrfs_ioctl_space_info *) |
2705 | (arg + sizeof(struct btrfs_ioctl_space_args)); | |
2706 | ||
2707 | if (copy_to_user(user_dest, dest_orig, alloc_size)) | |
2708 | ret = -EFAULT; | |
2709 | ||
2710 | kfree(dest_orig); | |
2711 | out: | |
2712 | if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) | |
1406e432 JB |
2713 | ret = -EFAULT; |
2714 | ||
2715 | return ret; | |
2716 | } | |
2717 | ||
f46b5a66 CH |
2718 | /* |
2719 | * there are many ways the trans_start and trans_end ioctls can lead | |
2720 | * to deadlocks. They should only be used by applications that | |
2721 | * basically own the machine, and have a very in depth understanding | |
2722 | * of all the possible deadlocks and enospc problems. | |
2723 | */ | |
2724 | long btrfs_ioctl_trans_end(struct file *file) | |
2725 | { | |
2726 | struct inode *inode = fdentry(file)->d_inode; | |
2727 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2728 | struct btrfs_trans_handle *trans; | |
f46b5a66 | 2729 | |
f46b5a66 | 2730 | trans = file->private_data; |
1ab86aed SW |
2731 | if (!trans) |
2732 | return -EINVAL; | |
b214107e | 2733 | file->private_data = NULL; |
9ca9ee09 | 2734 | |
1ab86aed SW |
2735 | btrfs_end_transaction(trans, root); |
2736 | ||
a4abeea4 | 2737 | atomic_dec(&root->fs_info->open_ioctl_trans); |
9ca9ee09 | 2738 | |
cfc8ea87 | 2739 | mnt_drop_write(file->f_path.mnt); |
1ab86aed | 2740 | return 0; |
f46b5a66 CH |
2741 | } |
2742 | ||
46204592 SW |
2743 | static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp) |
2744 | { | |
2745 | struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root; | |
2746 | struct btrfs_trans_handle *trans; | |
2747 | u64 transid; | |
db5b493a | 2748 | int ret; |
46204592 SW |
2749 | |
2750 | trans = btrfs_start_transaction(root, 0); | |
98d5dc13 TI |
2751 | if (IS_ERR(trans)) |
2752 | return PTR_ERR(trans); | |
46204592 | 2753 | transid = trans->transid; |
db5b493a | 2754 | ret = btrfs_commit_transaction_async(trans, root, 0); |
8b2b2d3c TI |
2755 | if (ret) { |
2756 | btrfs_end_transaction(trans, root); | |
db5b493a | 2757 | return ret; |
8b2b2d3c | 2758 | } |
46204592 SW |
2759 | |
2760 | if (argp) | |
2761 | if (copy_to_user(argp, &transid, sizeof(transid))) | |
2762 | return -EFAULT; | |
2763 | return 0; | |
2764 | } | |
2765 | ||
2766 | static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp) | |
2767 | { | |
2768 | struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root; | |
2769 | u64 transid; | |
2770 | ||
2771 | if (argp) { | |
2772 | if (copy_from_user(&transid, argp, sizeof(transid))) | |
2773 | return -EFAULT; | |
2774 | } else { | |
2775 | transid = 0; /* current trans */ | |
2776 | } | |
2777 | return btrfs_wait_for_commit(root, transid); | |
2778 | } | |
2779 | ||
475f6387 JS |
2780 | static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg) |
2781 | { | |
2782 | int ret; | |
2783 | struct btrfs_ioctl_scrub_args *sa; | |
2784 | ||
2785 | if (!capable(CAP_SYS_ADMIN)) | |
2786 | return -EPERM; | |
2787 | ||
2788 | sa = memdup_user(arg, sizeof(*sa)); | |
2789 | if (IS_ERR(sa)) | |
2790 | return PTR_ERR(sa); | |
2791 | ||
2792 | ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end, | |
8628764e | 2793 | &sa->progress, sa->flags & BTRFS_SCRUB_READONLY); |
475f6387 JS |
2794 | |
2795 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
2796 | ret = -EFAULT; | |
2797 | ||
2798 | kfree(sa); | |
2799 | return ret; | |
2800 | } | |
2801 | ||
2802 | static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) | |
2803 | { | |
2804 | if (!capable(CAP_SYS_ADMIN)) | |
2805 | return -EPERM; | |
2806 | ||
2807 | return btrfs_scrub_cancel(root); | |
2808 | } | |
2809 | ||
2810 | static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, | |
2811 | void __user *arg) | |
2812 | { | |
2813 | struct btrfs_ioctl_scrub_args *sa; | |
2814 | int ret; | |
2815 | ||
2816 | if (!capable(CAP_SYS_ADMIN)) | |
2817 | return -EPERM; | |
2818 | ||
2819 | sa = memdup_user(arg, sizeof(*sa)); | |
2820 | if (IS_ERR(sa)) | |
2821 | return PTR_ERR(sa); | |
2822 | ||
2823 | ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); | |
2824 | ||
2825 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
2826 | ret = -EFAULT; | |
2827 | ||
2828 | kfree(sa); | |
2829 | return ret; | |
2830 | } | |
2831 | ||
f46b5a66 CH |
2832 | long btrfs_ioctl(struct file *file, unsigned int |
2833 | cmd, unsigned long arg) | |
2834 | { | |
2835 | struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; | |
4bcabaa3 | 2836 | void __user *argp = (void __user *)arg; |
f46b5a66 CH |
2837 | |
2838 | switch (cmd) { | |
6cbff00f CH |
2839 | case FS_IOC_GETFLAGS: |
2840 | return btrfs_ioctl_getflags(file, argp); | |
2841 | case FS_IOC_SETFLAGS: | |
2842 | return btrfs_ioctl_setflags(file, argp); | |
2843 | case FS_IOC_GETVERSION: | |
2844 | return btrfs_ioctl_getversion(file, argp); | |
f7039b1d LD |
2845 | case FITRIM: |
2846 | return btrfs_ioctl_fitrim(file, argp); | |
f46b5a66 | 2847 | case BTRFS_IOC_SNAP_CREATE: |
fa0d2b9b | 2848 | return btrfs_ioctl_snap_create(file, argp, 0); |
fdfb1e4f | 2849 | case BTRFS_IOC_SNAP_CREATE_V2: |
fa0d2b9b | 2850 | return btrfs_ioctl_snap_create_v2(file, argp, 0); |
3de4586c | 2851 | case BTRFS_IOC_SUBVOL_CREATE: |
fa0d2b9b | 2852 | return btrfs_ioctl_snap_create(file, argp, 1); |
76dda93c YZ |
2853 | case BTRFS_IOC_SNAP_DESTROY: |
2854 | return btrfs_ioctl_snap_destroy(file, argp); | |
0caa102d LZ |
2855 | case BTRFS_IOC_SUBVOL_GETFLAGS: |
2856 | return btrfs_ioctl_subvol_getflags(file, argp); | |
2857 | case BTRFS_IOC_SUBVOL_SETFLAGS: | |
2858 | return btrfs_ioctl_subvol_setflags(file, argp); | |
6ef5ed0d JB |
2859 | case BTRFS_IOC_DEFAULT_SUBVOL: |
2860 | return btrfs_ioctl_default_subvol(file, argp); | |
f46b5a66 | 2861 | case BTRFS_IOC_DEFRAG: |
1e701a32 CM |
2862 | return btrfs_ioctl_defrag(file, NULL); |
2863 | case BTRFS_IOC_DEFRAG_RANGE: | |
2864 | return btrfs_ioctl_defrag(file, argp); | |
f46b5a66 | 2865 | case BTRFS_IOC_RESIZE: |
4bcabaa3 | 2866 | return btrfs_ioctl_resize(root, argp); |
f46b5a66 | 2867 | case BTRFS_IOC_ADD_DEV: |
4bcabaa3 | 2868 | return btrfs_ioctl_add_dev(root, argp); |
f46b5a66 | 2869 | case BTRFS_IOC_RM_DEV: |
4bcabaa3 | 2870 | return btrfs_ioctl_rm_dev(root, argp); |
475f6387 JS |
2871 | case BTRFS_IOC_FS_INFO: |
2872 | return btrfs_ioctl_fs_info(root, argp); | |
2873 | case BTRFS_IOC_DEV_INFO: | |
2874 | return btrfs_ioctl_dev_info(root, argp); | |
f46b5a66 CH |
2875 | case BTRFS_IOC_BALANCE: |
2876 | return btrfs_balance(root->fs_info->dev_root); | |
2877 | case BTRFS_IOC_CLONE: | |
c5c9cd4d SW |
2878 | return btrfs_ioctl_clone(file, arg, 0, 0, 0); |
2879 | case BTRFS_IOC_CLONE_RANGE: | |
7a865e8a | 2880 | return btrfs_ioctl_clone_range(file, argp); |
f46b5a66 CH |
2881 | case BTRFS_IOC_TRANS_START: |
2882 | return btrfs_ioctl_trans_start(file); | |
2883 | case BTRFS_IOC_TRANS_END: | |
2884 | return btrfs_ioctl_trans_end(file); | |
ac8e9819 CM |
2885 | case BTRFS_IOC_TREE_SEARCH: |
2886 | return btrfs_ioctl_tree_search(file, argp); | |
2887 | case BTRFS_IOC_INO_LOOKUP: | |
2888 | return btrfs_ioctl_ino_lookup(file, argp); | |
1406e432 JB |
2889 | case BTRFS_IOC_SPACE_INFO: |
2890 | return btrfs_ioctl_space_info(root, argp); | |
f46b5a66 CH |
2891 | case BTRFS_IOC_SYNC: |
2892 | btrfs_sync_fs(file->f_dentry->d_sb, 1); | |
2893 | return 0; | |
46204592 SW |
2894 | case BTRFS_IOC_START_SYNC: |
2895 | return btrfs_ioctl_start_sync(file, argp); | |
2896 | case BTRFS_IOC_WAIT_SYNC: | |
2897 | return btrfs_ioctl_wait_sync(file, argp); | |
475f6387 JS |
2898 | case BTRFS_IOC_SCRUB: |
2899 | return btrfs_ioctl_scrub(root, argp); | |
2900 | case BTRFS_IOC_SCRUB_CANCEL: | |
2901 | return btrfs_ioctl_scrub_cancel(root, argp); | |
2902 | case BTRFS_IOC_SCRUB_PROGRESS: | |
2903 | return btrfs_ioctl_scrub_progress(root, argp); | |
f46b5a66 CH |
2904 | } |
2905 | ||
2906 | return -ENOTTY; | |
2907 | } |