]>
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> |
4b4e25f2 | 42 | #include "compat.h" |
f46b5a66 CH |
43 | #include "ctree.h" |
44 | #include "disk-io.h" | |
45 | #include "transaction.h" | |
46 | #include "btrfs_inode.h" | |
47 | #include "ioctl.h" | |
48 | #include "print-tree.h" | |
49 | #include "volumes.h" | |
925baedd | 50 | #include "locking.h" |
98d377a0 | 51 | #include "ctree.h" |
f46b5a66 | 52 | |
6cbff00f CH |
53 | /* Mask out flags that are inappropriate for the given type of inode. */ |
54 | static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) | |
55 | { | |
56 | if (S_ISDIR(mode)) | |
57 | return flags; | |
58 | else if (S_ISREG(mode)) | |
59 | return flags & ~FS_DIRSYNC_FL; | |
60 | else | |
61 | return flags & (FS_NODUMP_FL | FS_NOATIME_FL); | |
62 | } | |
63 | ||
64 | /* | |
65 | * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. | |
66 | */ | |
67 | static unsigned int btrfs_flags_to_ioctl(unsigned int flags) | |
68 | { | |
69 | unsigned int iflags = 0; | |
70 | ||
71 | if (flags & BTRFS_INODE_SYNC) | |
72 | iflags |= FS_SYNC_FL; | |
73 | if (flags & BTRFS_INODE_IMMUTABLE) | |
74 | iflags |= FS_IMMUTABLE_FL; | |
75 | if (flags & BTRFS_INODE_APPEND) | |
76 | iflags |= FS_APPEND_FL; | |
77 | if (flags & BTRFS_INODE_NODUMP) | |
78 | iflags |= FS_NODUMP_FL; | |
79 | if (flags & BTRFS_INODE_NOATIME) | |
80 | iflags |= FS_NOATIME_FL; | |
81 | if (flags & BTRFS_INODE_DIRSYNC) | |
82 | iflags |= FS_DIRSYNC_FL; | |
83 | ||
84 | return iflags; | |
85 | } | |
86 | ||
87 | /* | |
88 | * Update inode->i_flags based on the btrfs internal flags. | |
89 | */ | |
90 | void btrfs_update_iflags(struct inode *inode) | |
91 | { | |
92 | struct btrfs_inode *ip = BTRFS_I(inode); | |
93 | ||
94 | inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC); | |
95 | ||
96 | if (ip->flags & BTRFS_INODE_SYNC) | |
97 | inode->i_flags |= S_SYNC; | |
98 | if (ip->flags & BTRFS_INODE_IMMUTABLE) | |
99 | inode->i_flags |= S_IMMUTABLE; | |
100 | if (ip->flags & BTRFS_INODE_APPEND) | |
101 | inode->i_flags |= S_APPEND; | |
102 | if (ip->flags & BTRFS_INODE_NOATIME) | |
103 | inode->i_flags |= S_NOATIME; | |
104 | if (ip->flags & BTRFS_INODE_DIRSYNC) | |
105 | inode->i_flags |= S_DIRSYNC; | |
106 | } | |
107 | ||
108 | /* | |
109 | * Inherit flags from the parent inode. | |
110 | * | |
111 | * Unlike extN we don't have any flags we don't want to inherit currently. | |
112 | */ | |
113 | void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) | |
114 | { | |
0b4dcea5 CM |
115 | unsigned int flags; |
116 | ||
117 | if (!dir) | |
118 | return; | |
119 | ||
120 | flags = BTRFS_I(dir)->flags; | |
6cbff00f CH |
121 | |
122 | if (S_ISREG(inode->i_mode)) | |
123 | flags &= ~BTRFS_INODE_DIRSYNC; | |
124 | else if (!S_ISDIR(inode->i_mode)) | |
125 | flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME); | |
126 | ||
127 | BTRFS_I(inode)->flags = flags; | |
128 | btrfs_update_iflags(inode); | |
129 | } | |
130 | ||
131 | static int btrfs_ioctl_getflags(struct file *file, void __user *arg) | |
132 | { | |
133 | struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode); | |
134 | unsigned int flags = btrfs_flags_to_ioctl(ip->flags); | |
135 | ||
136 | if (copy_to_user(arg, &flags, sizeof(flags))) | |
137 | return -EFAULT; | |
138 | return 0; | |
139 | } | |
140 | ||
141 | static int btrfs_ioctl_setflags(struct file *file, void __user *arg) | |
142 | { | |
143 | struct inode *inode = file->f_path.dentry->d_inode; | |
144 | struct btrfs_inode *ip = BTRFS_I(inode); | |
145 | struct btrfs_root *root = ip->root; | |
146 | struct btrfs_trans_handle *trans; | |
147 | unsigned int flags, oldflags; | |
148 | int ret; | |
149 | ||
150 | if (copy_from_user(&flags, arg, sizeof(flags))) | |
151 | return -EFAULT; | |
152 | ||
153 | if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ | |
154 | FS_NOATIME_FL | FS_NODUMP_FL | \ | |
155 | FS_SYNC_FL | FS_DIRSYNC_FL)) | |
156 | return -EOPNOTSUPP; | |
f46b5a66 | 157 | |
6cbff00f CH |
158 | if (!is_owner_or_cap(inode)) |
159 | return -EACCES; | |
160 | ||
161 | mutex_lock(&inode->i_mutex); | |
162 | ||
163 | flags = btrfs_mask_flags(inode->i_mode, flags); | |
164 | oldflags = btrfs_flags_to_ioctl(ip->flags); | |
165 | if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { | |
166 | if (!capable(CAP_LINUX_IMMUTABLE)) { | |
167 | ret = -EPERM; | |
168 | goto out_unlock; | |
169 | } | |
170 | } | |
171 | ||
172 | ret = mnt_want_write(file->f_path.mnt); | |
173 | if (ret) | |
174 | goto out_unlock; | |
175 | ||
176 | if (flags & FS_SYNC_FL) | |
177 | ip->flags |= BTRFS_INODE_SYNC; | |
178 | else | |
179 | ip->flags &= ~BTRFS_INODE_SYNC; | |
180 | if (flags & FS_IMMUTABLE_FL) | |
181 | ip->flags |= BTRFS_INODE_IMMUTABLE; | |
182 | else | |
183 | ip->flags &= ~BTRFS_INODE_IMMUTABLE; | |
184 | if (flags & FS_APPEND_FL) | |
185 | ip->flags |= BTRFS_INODE_APPEND; | |
186 | else | |
187 | ip->flags &= ~BTRFS_INODE_APPEND; | |
188 | if (flags & FS_NODUMP_FL) | |
189 | ip->flags |= BTRFS_INODE_NODUMP; | |
190 | else | |
191 | ip->flags &= ~BTRFS_INODE_NODUMP; | |
192 | if (flags & FS_NOATIME_FL) | |
193 | ip->flags |= BTRFS_INODE_NOATIME; | |
194 | else | |
195 | ip->flags &= ~BTRFS_INODE_NOATIME; | |
196 | if (flags & FS_DIRSYNC_FL) | |
197 | ip->flags |= BTRFS_INODE_DIRSYNC; | |
198 | else | |
199 | ip->flags &= ~BTRFS_INODE_DIRSYNC; | |
200 | ||
201 | ||
202 | trans = btrfs_join_transaction(root, 1); | |
203 | BUG_ON(!trans); | |
204 | ||
205 | ret = btrfs_update_inode(trans, root, inode); | |
206 | BUG_ON(ret); | |
207 | ||
208 | btrfs_update_iflags(inode); | |
209 | inode->i_ctime = CURRENT_TIME; | |
210 | btrfs_end_transaction(trans, root); | |
211 | ||
212 | mnt_drop_write(file->f_path.mnt); | |
213 | out_unlock: | |
214 | mutex_unlock(&inode->i_mutex); | |
215 | return 0; | |
216 | } | |
217 | ||
218 | static int btrfs_ioctl_getversion(struct file *file, int __user *arg) | |
219 | { | |
220 | struct inode *inode = file->f_path.dentry->d_inode; | |
221 | ||
222 | return put_user(inode->i_generation, arg); | |
223 | } | |
f46b5a66 | 224 | |
cb8e7090 CH |
225 | static noinline int create_subvol(struct btrfs_root *root, |
226 | struct dentry *dentry, | |
227 | char *name, int namelen) | |
f46b5a66 CH |
228 | { |
229 | struct btrfs_trans_handle *trans; | |
230 | struct btrfs_key key; | |
231 | struct btrfs_root_item root_item; | |
232 | struct btrfs_inode_item *inode_item; | |
233 | struct extent_buffer *leaf; | |
76dda93c YZ |
234 | struct btrfs_root *new_root; |
235 | struct inode *dir = dentry->d_parent->d_inode; | |
f46b5a66 CH |
236 | int ret; |
237 | int err; | |
238 | u64 objectid; | |
239 | u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; | |
3de4586c | 240 | u64 index = 0; |
f46b5a66 | 241 | |
9ed74f2d JB |
242 | /* |
243 | * 1 - inode item | |
244 | * 2 - refs | |
245 | * 1 - root item | |
246 | * 2 - dir items | |
247 | */ | |
248 | ret = btrfs_reserve_metadata_space(root, 6); | |
f46b5a66 | 249 | if (ret) |
76dda93c | 250 | return ret; |
f46b5a66 CH |
251 | |
252 | trans = btrfs_start_transaction(root, 1); | |
253 | BUG_ON(!trans); | |
254 | ||
255 | ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, | |
256 | 0, &objectid); | |
257 | if (ret) | |
258 | goto fail; | |
259 | ||
5d4f98a2 YZ |
260 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, |
261 | 0, objectid, NULL, 0, 0, 0); | |
8e8a1e31 JB |
262 | if (IS_ERR(leaf)) { |
263 | ret = PTR_ERR(leaf); | |
264 | goto fail; | |
265 | } | |
f46b5a66 | 266 | |
5d4f98a2 | 267 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
f46b5a66 CH |
268 | btrfs_set_header_bytenr(leaf, leaf->start); |
269 | btrfs_set_header_generation(leaf, trans->transid); | |
5d4f98a2 | 270 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); |
f46b5a66 CH |
271 | btrfs_set_header_owner(leaf, objectid); |
272 | ||
273 | write_extent_buffer(leaf, root->fs_info->fsid, | |
274 | (unsigned long)btrfs_header_fsid(leaf), | |
275 | BTRFS_FSID_SIZE); | |
5d4f98a2 YZ |
276 | write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, |
277 | (unsigned long)btrfs_header_chunk_tree_uuid(leaf), | |
278 | BTRFS_UUID_SIZE); | |
f46b5a66 CH |
279 | btrfs_mark_buffer_dirty(leaf); |
280 | ||
281 | inode_item = &root_item.inode; | |
282 | memset(inode_item, 0, sizeof(*inode_item)); | |
283 | inode_item->generation = cpu_to_le64(1); | |
284 | inode_item->size = cpu_to_le64(3); | |
285 | inode_item->nlink = cpu_to_le32(1); | |
a76a3cd4 | 286 | inode_item->nbytes = cpu_to_le64(root->leafsize); |
f46b5a66 CH |
287 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); |
288 | ||
289 | btrfs_set_root_bytenr(&root_item, leaf->start); | |
84234f3a | 290 | btrfs_set_root_generation(&root_item, trans->transid); |
f46b5a66 CH |
291 | btrfs_set_root_level(&root_item, 0); |
292 | btrfs_set_root_refs(&root_item, 1); | |
86b9f2ec | 293 | btrfs_set_root_used(&root_item, leaf->len); |
80ff3856 | 294 | btrfs_set_root_last_snapshot(&root_item, 0); |
f46b5a66 CH |
295 | |
296 | memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); | |
297 | root_item.drop_level = 0; | |
298 | ||
925baedd | 299 | btrfs_tree_unlock(leaf); |
f46b5a66 CH |
300 | free_extent_buffer(leaf); |
301 | leaf = NULL; | |
302 | ||
303 | btrfs_set_root_dirid(&root_item, new_dirid); | |
304 | ||
305 | key.objectid = objectid; | |
5d4f98a2 | 306 | key.offset = 0; |
f46b5a66 CH |
307 | btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); |
308 | ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, | |
309 | &root_item); | |
310 | if (ret) | |
311 | goto fail; | |
312 | ||
76dda93c YZ |
313 | key.offset = (u64)-1; |
314 | new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); | |
315 | BUG_ON(IS_ERR(new_root)); | |
316 | ||
317 | btrfs_record_root_in_trans(trans, new_root); | |
318 | ||
319 | ret = btrfs_create_subvol_root(trans, new_root, new_dirid, | |
320 | BTRFS_I(dir)->block_group); | |
f46b5a66 CH |
321 | /* |
322 | * insert the directory item | |
323 | */ | |
3de4586c CM |
324 | ret = btrfs_set_inode_index(dir, &index); |
325 | BUG_ON(ret); | |
326 | ||
327 | ret = btrfs_insert_dir_item(trans, root, | |
f46b5a66 | 328 | name, namelen, dir->i_ino, &key, |
3de4586c | 329 | BTRFS_FT_DIR, index); |
f46b5a66 CH |
330 | if (ret) |
331 | goto fail; | |
0660b5af | 332 | |
52c26179 YZ |
333 | btrfs_i_size_write(dir, dir->i_size + namelen * 2); |
334 | ret = btrfs_update_inode(trans, root, dir); | |
335 | BUG_ON(ret); | |
336 | ||
0660b5af | 337 | ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, |
4df27c4d | 338 | objectid, root->root_key.objectid, |
0660b5af | 339 | dir->i_ino, index, name, namelen); |
0660b5af | 340 | |
76dda93c | 341 | BUG_ON(ret); |
f46b5a66 | 342 | |
76dda93c | 343 | d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry)); |
f46b5a66 | 344 | fail: |
76dda93c | 345 | err = btrfs_commit_transaction(trans, root); |
f46b5a66 CH |
346 | if (err && !ret) |
347 | ret = err; | |
9ed74f2d JB |
348 | |
349 | btrfs_unreserve_metadata_space(root, 6); | |
f46b5a66 CH |
350 | return ret; |
351 | } | |
352 | ||
3de4586c CM |
353 | static int create_snapshot(struct btrfs_root *root, struct dentry *dentry, |
354 | char *name, int namelen) | |
f46b5a66 | 355 | { |
2e4bfab9 | 356 | struct inode *inode; |
f46b5a66 CH |
357 | struct btrfs_pending_snapshot *pending_snapshot; |
358 | struct btrfs_trans_handle *trans; | |
2e4bfab9 | 359 | int ret; |
f46b5a66 CH |
360 | |
361 | if (!root->ref_cows) | |
362 | return -EINVAL; | |
363 | ||
9ed74f2d JB |
364 | /* |
365 | * 1 - inode item | |
366 | * 2 - refs | |
367 | * 1 - root item | |
368 | * 2 - dir items | |
369 | */ | |
370 | ret = btrfs_reserve_metadata_space(root, 6); | |
f46b5a66 | 371 | if (ret) |
2e4bfab9 | 372 | goto fail; |
f46b5a66 | 373 | |
3de4586c | 374 | pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); |
f46b5a66 CH |
375 | if (!pending_snapshot) { |
376 | ret = -ENOMEM; | |
9ed74f2d | 377 | btrfs_unreserve_metadata_space(root, 6); |
2e4bfab9 | 378 | goto fail; |
f46b5a66 CH |
379 | } |
380 | pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS); | |
381 | if (!pending_snapshot->name) { | |
382 | ret = -ENOMEM; | |
383 | kfree(pending_snapshot); | |
9ed74f2d | 384 | btrfs_unreserve_metadata_space(root, 6); |
2e4bfab9 | 385 | goto fail; |
f46b5a66 CH |
386 | } |
387 | memcpy(pending_snapshot->name, name, namelen); | |
388 | pending_snapshot->name[namelen] = '\0'; | |
3de4586c | 389 | pending_snapshot->dentry = dentry; |
f46b5a66 CH |
390 | trans = btrfs_start_transaction(root, 1); |
391 | BUG_ON(!trans); | |
392 | pending_snapshot->root = root; | |
393 | list_add(&pending_snapshot->list, | |
394 | &trans->transaction->pending_snapshots); | |
2e4bfab9 YZ |
395 | ret = btrfs_commit_transaction(trans, root); |
396 | BUG_ON(ret); | |
397 | btrfs_unreserve_metadata_space(root, 6); | |
f46b5a66 | 398 | |
2e4bfab9 YZ |
399 | inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry); |
400 | if (IS_ERR(inode)) { | |
401 | ret = PTR_ERR(inode); | |
402 | goto fail; | |
403 | } | |
404 | BUG_ON(!inode); | |
405 | d_instantiate(dentry, inode); | |
406 | ret = 0; | |
407 | fail: | |
f46b5a66 CH |
408 | return ret; |
409 | } | |
410 | ||
cb8e7090 CH |
411 | /* copy of may_create in fs/namei.c() */ |
412 | static inline int btrfs_may_create(struct inode *dir, struct dentry *child) | |
413 | { | |
414 | if (child->d_inode) | |
415 | return -EEXIST; | |
416 | if (IS_DEADDIR(dir)) | |
417 | return -ENOENT; | |
418 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
419 | } | |
420 | ||
421 | /* | |
422 | * Create a new subvolume below @parent. This is largely modeled after | |
423 | * sys_mkdirat and vfs_mkdir, but we only do a single component lookup | |
424 | * inside this filesystem so it's quite a bit simpler. | |
425 | */ | |
76dda93c YZ |
426 | static noinline int btrfs_mksubvol(struct path *parent, |
427 | char *name, int namelen, | |
3de4586c | 428 | struct btrfs_root *snap_src) |
cb8e7090 | 429 | { |
76dda93c | 430 | struct inode *dir = parent->dentry->d_inode; |
cb8e7090 CH |
431 | struct dentry *dentry; |
432 | int error; | |
433 | ||
76dda93c | 434 | mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); |
cb8e7090 CH |
435 | |
436 | dentry = lookup_one_len(name, parent->dentry, namelen); | |
437 | error = PTR_ERR(dentry); | |
438 | if (IS_ERR(dentry)) | |
439 | goto out_unlock; | |
440 | ||
441 | error = -EEXIST; | |
442 | if (dentry->d_inode) | |
443 | goto out_dput; | |
444 | ||
cb8e7090 CH |
445 | error = mnt_want_write(parent->mnt); |
446 | if (error) | |
447 | goto out_dput; | |
448 | ||
76dda93c | 449 | error = btrfs_may_create(dir, dentry); |
cb8e7090 CH |
450 | if (error) |
451 | goto out_drop_write; | |
452 | ||
76dda93c YZ |
453 | down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); |
454 | ||
455 | if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) | |
456 | goto out_up_read; | |
457 | ||
3de4586c | 458 | if (snap_src) { |
76dda93c YZ |
459 | error = create_snapshot(snap_src, dentry, |
460 | name, namelen); | |
3de4586c | 461 | } else { |
76dda93c YZ |
462 | error = create_subvol(BTRFS_I(dir)->root, dentry, |
463 | name, namelen); | |
3de4586c | 464 | } |
76dda93c YZ |
465 | if (!error) |
466 | fsnotify_mkdir(dir, dentry); | |
467 | out_up_read: | |
468 | up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); | |
cb8e7090 CH |
469 | out_drop_write: |
470 | mnt_drop_write(parent->mnt); | |
471 | out_dput: | |
472 | dput(dentry); | |
473 | out_unlock: | |
76dda93c | 474 | mutex_unlock(&dir->i_mutex); |
cb8e7090 CH |
475 | return error; |
476 | } | |
477 | ||
940100a4 | 478 | static int should_defrag_range(struct inode *inode, u64 start, u64 len, |
1e701a32 CM |
479 | int thresh, u64 *last_len, u64 *skip, |
480 | u64 *defrag_end) | |
940100a4 CM |
481 | { |
482 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
483 | struct extent_map *em = NULL; | |
484 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
485 | int ret = 1; | |
486 | ||
1e701a32 CM |
487 | |
488 | if (thresh == 0) | |
489 | thresh = 256 * 1024; | |
490 | ||
940100a4 CM |
491 | /* |
492 | * make sure that once we start defragging and extent, we keep on | |
493 | * defragging it | |
494 | */ | |
495 | if (start < *defrag_end) | |
496 | return 1; | |
497 | ||
498 | *skip = 0; | |
499 | ||
500 | /* | |
501 | * hopefully we have this extent in the tree already, try without | |
502 | * the full extent lock | |
503 | */ | |
504 | read_lock(&em_tree->lock); | |
505 | em = lookup_extent_mapping(em_tree, start, len); | |
506 | read_unlock(&em_tree->lock); | |
507 | ||
508 | if (!em) { | |
509 | /* get the big lock and read metadata off disk */ | |
510 | lock_extent(io_tree, start, start + len - 1, GFP_NOFS); | |
511 | em = btrfs_get_extent(inode, NULL, 0, start, len, 0); | |
512 | unlock_extent(io_tree, start, start + len - 1, GFP_NOFS); | |
513 | ||
514 | if (!em) | |
515 | return 0; | |
516 | } | |
517 | ||
518 | /* this will cover holes, and inline extents */ | |
519 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) | |
520 | ret = 0; | |
521 | ||
522 | /* | |
523 | * we hit a real extent, if it is big don't bother defragging it again | |
524 | */ | |
1e701a32 | 525 | if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh) |
940100a4 CM |
526 | ret = 0; |
527 | ||
528 | /* | |
529 | * last_len ends up being a counter of how many bytes we've defragged. | |
530 | * every time we choose not to defrag an extent, we reset *last_len | |
531 | * so that the next tiny extent will force a defrag. | |
532 | * | |
533 | * The end result of this is that tiny extents before a single big | |
534 | * extent will force at least part of that big extent to be defragged. | |
535 | */ | |
536 | if (ret) { | |
537 | *last_len += len; | |
538 | *defrag_end = extent_map_end(em); | |
539 | } else { | |
540 | *last_len = 0; | |
541 | *skip = extent_map_end(em); | |
542 | *defrag_end = 0; | |
543 | } | |
544 | ||
545 | free_extent_map(em); | |
546 | return ret; | |
547 | } | |
548 | ||
1e701a32 CM |
549 | static int btrfs_defrag_file(struct file *file, |
550 | struct btrfs_ioctl_defrag_range_args *range) | |
f46b5a66 CH |
551 | { |
552 | struct inode *inode = fdentry(file)->d_inode; | |
553 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
554 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
3eaa2885 | 555 | struct btrfs_ordered_extent *ordered; |
f46b5a66 CH |
556 | struct page *page; |
557 | unsigned long last_index; | |
558 | unsigned long ra_pages = root->fs_info->bdi.ra_pages; | |
559 | unsigned long total_read = 0; | |
560 | u64 page_start; | |
561 | u64 page_end; | |
940100a4 CM |
562 | u64 last_len = 0; |
563 | u64 skip = 0; | |
564 | u64 defrag_end = 0; | |
f46b5a66 CH |
565 | unsigned long i; |
566 | int ret; | |
567 | ||
940100a4 CM |
568 | if (inode->i_size == 0) |
569 | return 0; | |
570 | ||
1e701a32 CM |
571 | if (range->start + range->len > range->start) { |
572 | last_index = min_t(u64, inode->i_size - 1, | |
573 | range->start + range->len - 1) >> PAGE_CACHE_SHIFT; | |
574 | } else { | |
575 | last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT; | |
576 | } | |
577 | ||
578 | i = range->start >> PAGE_CACHE_SHIFT; | |
940100a4 CM |
579 | while (i <= last_index) { |
580 | if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, | |
1e701a32 CM |
581 | PAGE_CACHE_SIZE, |
582 | range->extent_thresh, | |
583 | &last_len, &skip, | |
940100a4 CM |
584 | &defrag_end)) { |
585 | unsigned long next; | |
586 | /* | |
587 | * the should_defrag function tells us how much to skip | |
588 | * bump our counter by the suggested amount | |
589 | */ | |
590 | next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; | |
591 | i = max(i + 1, next); | |
592 | continue; | |
593 | } | |
f46b5a66 | 594 | |
f46b5a66 CH |
595 | if (total_read % ra_pages == 0) { |
596 | btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i, | |
597 | min(last_index, i + ra_pages - 1)); | |
598 | } | |
599 | total_read++; | |
940100a4 | 600 | mutex_lock(&inode->i_mutex); |
1e701a32 CM |
601 | if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) |
602 | BTRFS_I(inode)->force_compress = 1; | |
940100a4 CM |
603 | |
604 | ret = btrfs_check_data_free_space(root, inode, PAGE_CACHE_SIZE); | |
605 | if (ret) { | |
606 | ret = -ENOSPC; | |
607 | break; | |
608 | } | |
609 | ||
610 | ret = btrfs_reserve_metadata_for_delalloc(root, inode, 1); | |
611 | if (ret) { | |
612 | btrfs_free_reserved_data_space(root, inode, | |
613 | PAGE_CACHE_SIZE); | |
614 | ret = -ENOSPC; | |
615 | break; | |
616 | } | |
3eaa2885 | 617 | again: |
940100a4 CM |
618 | if (inode->i_size == 0 || |
619 | i > ((inode->i_size - 1) >> PAGE_CACHE_SHIFT)) { | |
620 | ret = 0; | |
621 | goto err_reservations; | |
622 | } | |
623 | ||
f46b5a66 CH |
624 | page = grab_cache_page(inode->i_mapping, i); |
625 | if (!page) | |
940100a4 CM |
626 | goto err_reservations; |
627 | ||
f46b5a66 CH |
628 | if (!PageUptodate(page)) { |
629 | btrfs_readpage(NULL, page); | |
630 | lock_page(page); | |
631 | if (!PageUptodate(page)) { | |
632 | unlock_page(page); | |
633 | page_cache_release(page); | |
940100a4 | 634 | goto err_reservations; |
f46b5a66 CH |
635 | } |
636 | } | |
637 | ||
940100a4 CM |
638 | if (page->mapping != inode->i_mapping) { |
639 | unlock_page(page); | |
640 | page_cache_release(page); | |
641 | goto again; | |
642 | } | |
643 | ||
f46b5a66 | 644 | wait_on_page_writeback(page); |
f46b5a66 | 645 | |
940100a4 CM |
646 | if (PageDirty(page)) { |
647 | btrfs_free_reserved_data_space(root, inode, | |
648 | PAGE_CACHE_SIZE); | |
649 | goto loop_unlock; | |
650 | } | |
651 | ||
f46b5a66 CH |
652 | page_start = (u64)page->index << PAGE_CACHE_SHIFT; |
653 | page_end = page_start + PAGE_CACHE_SIZE - 1; | |
f46b5a66 | 654 | lock_extent(io_tree, page_start, page_end, GFP_NOFS); |
3eaa2885 CM |
655 | |
656 | ordered = btrfs_lookup_ordered_extent(inode, page_start); | |
657 | if (ordered) { | |
658 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); | |
659 | unlock_page(page); | |
660 | page_cache_release(page); | |
661 | btrfs_start_ordered_extent(inode, ordered, 1); | |
662 | btrfs_put_ordered_extent(ordered); | |
663 | goto again; | |
664 | } | |
665 | set_page_extent_mapped(page); | |
666 | ||
f87f057b CM |
667 | /* |
668 | * this makes sure page_mkwrite is called on the | |
669 | * page if it is dirtied again later | |
670 | */ | |
671 | clear_page_dirty_for_io(page); | |
940100a4 CM |
672 | clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start, |
673 | page_end, EXTENT_DIRTY | EXTENT_DELALLOC | | |
674 | EXTENT_DO_ACCOUNTING, GFP_NOFS); | |
f87f057b | 675 | |
2ac55d41 | 676 | btrfs_set_extent_delalloc(inode, page_start, page_end, NULL); |
940100a4 | 677 | ClearPageChecked(page); |
f46b5a66 | 678 | set_page_dirty(page); |
a1ed835e | 679 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); |
940100a4 CM |
680 | |
681 | loop_unlock: | |
f46b5a66 CH |
682 | unlock_page(page); |
683 | page_cache_release(page); | |
940100a4 CM |
684 | mutex_unlock(&inode->i_mutex); |
685 | ||
686 | btrfs_unreserve_metadata_for_delalloc(root, inode, 1); | |
f46b5a66 | 687 | balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1); |
940100a4 | 688 | i++; |
f46b5a66 CH |
689 | } |
690 | ||
1e701a32 CM |
691 | if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) |
692 | filemap_flush(inode->i_mapping); | |
693 | ||
694 | if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
695 | /* the filemap_flush will queue IO into the worker threads, but | |
696 | * we have to make sure the IO is actually started and that | |
697 | * ordered extents get created before we return | |
698 | */ | |
699 | atomic_inc(&root->fs_info->async_submit_draining); | |
700 | while (atomic_read(&root->fs_info->nr_async_submits) || | |
701 | atomic_read(&root->fs_info->async_delalloc_pages)) { | |
702 | wait_event(root->fs_info->async_submit_wait, | |
703 | (atomic_read(&root->fs_info->nr_async_submits) == 0 && | |
704 | atomic_read(&root->fs_info->async_delalloc_pages) == 0)); | |
705 | } | |
706 | atomic_dec(&root->fs_info->async_submit_draining); | |
707 | ||
708 | mutex_lock(&inode->i_mutex); | |
709 | BTRFS_I(inode)->force_compress = 0; | |
710 | mutex_unlock(&inode->i_mutex); | |
711 | } | |
712 | ||
f46b5a66 | 713 | return 0; |
940100a4 CM |
714 | |
715 | err_reservations: | |
716 | mutex_unlock(&inode->i_mutex); | |
717 | btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE); | |
718 | btrfs_unreserve_metadata_for_delalloc(root, inode, 1); | |
719 | return ret; | |
f46b5a66 CH |
720 | } |
721 | ||
76dda93c YZ |
722 | static noinline int btrfs_ioctl_resize(struct btrfs_root *root, |
723 | void __user *arg) | |
f46b5a66 CH |
724 | { |
725 | u64 new_size; | |
726 | u64 old_size; | |
727 | u64 devid = 1; | |
728 | struct btrfs_ioctl_vol_args *vol_args; | |
729 | struct btrfs_trans_handle *trans; | |
730 | struct btrfs_device *device = NULL; | |
731 | char *sizestr; | |
732 | char *devstr = NULL; | |
733 | int ret = 0; | |
734 | int namelen; | |
735 | int mod = 0; | |
736 | ||
c146afad YZ |
737 | if (root->fs_info->sb->s_flags & MS_RDONLY) |
738 | return -EROFS; | |
739 | ||
e441d54d CM |
740 | if (!capable(CAP_SYS_ADMIN)) |
741 | return -EPERM; | |
742 | ||
dae7b665 LZ |
743 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
744 | if (IS_ERR(vol_args)) | |
745 | return PTR_ERR(vol_args); | |
5516e595 MF |
746 | |
747 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
f46b5a66 | 748 | namelen = strlen(vol_args->name); |
f46b5a66 | 749 | |
7d9eb12c | 750 | mutex_lock(&root->fs_info->volume_mutex); |
f46b5a66 CH |
751 | sizestr = vol_args->name; |
752 | devstr = strchr(sizestr, ':'); | |
753 | if (devstr) { | |
754 | char *end; | |
755 | sizestr = devstr + 1; | |
756 | *devstr = '\0'; | |
757 | devstr = vol_args->name; | |
758 | devid = simple_strtoull(devstr, &end, 10); | |
21380931 JB |
759 | printk(KERN_INFO "resizing devid %llu\n", |
760 | (unsigned long long)devid); | |
f46b5a66 | 761 | } |
2b82032c | 762 | device = btrfs_find_device(root, devid, NULL, NULL); |
f46b5a66 | 763 | if (!device) { |
21380931 JB |
764 | printk(KERN_INFO "resizer unable to find device %llu\n", |
765 | (unsigned long long)devid); | |
f46b5a66 CH |
766 | ret = -EINVAL; |
767 | goto out_unlock; | |
768 | } | |
769 | if (!strcmp(sizestr, "max")) | |
770 | new_size = device->bdev->bd_inode->i_size; | |
771 | else { | |
772 | if (sizestr[0] == '-') { | |
773 | mod = -1; | |
774 | sizestr++; | |
775 | } else if (sizestr[0] == '+') { | |
776 | mod = 1; | |
777 | sizestr++; | |
778 | } | |
91748467 | 779 | new_size = memparse(sizestr, NULL); |
f46b5a66 CH |
780 | if (new_size == 0) { |
781 | ret = -EINVAL; | |
782 | goto out_unlock; | |
783 | } | |
784 | } | |
785 | ||
786 | old_size = device->total_bytes; | |
787 | ||
788 | if (mod < 0) { | |
789 | if (new_size > old_size) { | |
790 | ret = -EINVAL; | |
791 | goto out_unlock; | |
792 | } | |
793 | new_size = old_size - new_size; | |
794 | } else if (mod > 0) { | |
795 | new_size = old_size + new_size; | |
796 | } | |
797 | ||
798 | if (new_size < 256 * 1024 * 1024) { | |
799 | ret = -EINVAL; | |
800 | goto out_unlock; | |
801 | } | |
802 | if (new_size > device->bdev->bd_inode->i_size) { | |
803 | ret = -EFBIG; | |
804 | goto out_unlock; | |
805 | } | |
806 | ||
807 | do_div(new_size, root->sectorsize); | |
808 | new_size *= root->sectorsize; | |
809 | ||
810 | printk(KERN_INFO "new size for %s is %llu\n", | |
811 | device->name, (unsigned long long)new_size); | |
812 | ||
813 | if (new_size > old_size) { | |
814 | trans = btrfs_start_transaction(root, 1); | |
815 | ret = btrfs_grow_device(trans, device, new_size); | |
816 | btrfs_commit_transaction(trans, root); | |
817 | } else { | |
818 | ret = btrfs_shrink_device(device, new_size); | |
819 | } | |
820 | ||
821 | out_unlock: | |
7d9eb12c | 822 | mutex_unlock(&root->fs_info->volume_mutex); |
f46b5a66 CH |
823 | kfree(vol_args); |
824 | return ret; | |
825 | } | |
826 | ||
cb8e7090 | 827 | static noinline int btrfs_ioctl_snap_create(struct file *file, |
3de4586c | 828 | void __user *arg, int subvol) |
f46b5a66 | 829 | { |
cb8e7090 | 830 | struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; |
f46b5a66 | 831 | struct btrfs_ioctl_vol_args *vol_args; |
3de4586c | 832 | struct file *src_file; |
f46b5a66 | 833 | int namelen; |
3de4586c | 834 | int ret = 0; |
f46b5a66 | 835 | |
c146afad YZ |
836 | if (root->fs_info->sb->s_flags & MS_RDONLY) |
837 | return -EROFS; | |
838 | ||
dae7b665 LZ |
839 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
840 | if (IS_ERR(vol_args)) | |
841 | return PTR_ERR(vol_args); | |
f46b5a66 | 842 | |
5516e595 | 843 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 | 844 | namelen = strlen(vol_args->name); |
f46b5a66 CH |
845 | if (strchr(vol_args->name, '/')) { |
846 | ret = -EINVAL; | |
847 | goto out; | |
848 | } | |
849 | ||
3de4586c | 850 | if (subvol) { |
76dda93c YZ |
851 | ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen, |
852 | NULL); | |
cb8e7090 | 853 | } else { |
3de4586c CM |
854 | struct inode *src_inode; |
855 | src_file = fget(vol_args->fd); | |
856 | if (!src_file) { | |
857 | ret = -EINVAL; | |
858 | goto out; | |
859 | } | |
860 | ||
861 | src_inode = src_file->f_path.dentry->d_inode; | |
862 | if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) { | |
d397712b CM |
863 | printk(KERN_INFO "btrfs: Snapshot src from " |
864 | "another FS\n"); | |
3de4586c CM |
865 | ret = -EINVAL; |
866 | fput(src_file); | |
867 | goto out; | |
868 | } | |
76dda93c YZ |
869 | ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen, |
870 | BTRFS_I(src_inode)->root); | |
3de4586c | 871 | fput(src_file); |
cb8e7090 | 872 | } |
f46b5a66 CH |
873 | out: |
874 | kfree(vol_args); | |
875 | return ret; | |
876 | } | |
877 | ||
76dda93c YZ |
878 | /* |
879 | * helper to check if the subvolume references other subvolumes | |
880 | */ | |
881 | static noinline int may_destroy_subvol(struct btrfs_root *root) | |
882 | { | |
883 | struct btrfs_path *path; | |
884 | struct btrfs_key key; | |
885 | int ret; | |
886 | ||
887 | path = btrfs_alloc_path(); | |
888 | if (!path) | |
889 | return -ENOMEM; | |
890 | ||
891 | key.objectid = root->root_key.objectid; | |
892 | key.type = BTRFS_ROOT_REF_KEY; | |
893 | key.offset = (u64)-1; | |
894 | ||
895 | ret = btrfs_search_slot(NULL, root->fs_info->tree_root, | |
896 | &key, path, 0, 0); | |
897 | if (ret < 0) | |
898 | goto out; | |
899 | BUG_ON(ret == 0); | |
900 | ||
901 | ret = 0; | |
902 | if (path->slots[0] > 0) { | |
903 | path->slots[0]--; | |
904 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
905 | if (key.objectid == root->root_key.objectid && | |
906 | key.type == BTRFS_ROOT_REF_KEY) | |
907 | ret = -ENOTEMPTY; | |
908 | } | |
909 | out: | |
910 | btrfs_free_path(path); | |
911 | return ret; | |
912 | } | |
913 | ||
ac8e9819 CM |
914 | static noinline int key_in_sk(struct btrfs_key *key, |
915 | struct btrfs_ioctl_search_key *sk) | |
916 | { | |
abc6e134 CM |
917 | struct btrfs_key test; |
918 | int ret; | |
919 | ||
920 | test.objectid = sk->min_objectid; | |
921 | test.type = sk->min_type; | |
922 | test.offset = sk->min_offset; | |
923 | ||
924 | ret = btrfs_comp_cpu_keys(key, &test); | |
925 | if (ret < 0) | |
ac8e9819 | 926 | return 0; |
abc6e134 CM |
927 | |
928 | test.objectid = sk->max_objectid; | |
929 | test.type = sk->max_type; | |
930 | test.offset = sk->max_offset; | |
931 | ||
932 | ret = btrfs_comp_cpu_keys(key, &test); | |
933 | if (ret > 0) | |
ac8e9819 CM |
934 | return 0; |
935 | return 1; | |
936 | } | |
937 | ||
938 | static noinline int copy_to_sk(struct btrfs_root *root, | |
939 | struct btrfs_path *path, | |
940 | struct btrfs_key *key, | |
941 | struct btrfs_ioctl_search_key *sk, | |
942 | char *buf, | |
943 | unsigned long *sk_offset, | |
944 | int *num_found) | |
945 | { | |
946 | u64 found_transid; | |
947 | struct extent_buffer *leaf; | |
948 | struct btrfs_ioctl_search_header sh; | |
949 | unsigned long item_off; | |
950 | unsigned long item_len; | |
951 | int nritems; | |
952 | int i; | |
953 | int slot; | |
954 | int found = 0; | |
955 | int ret = 0; | |
956 | ||
957 | leaf = path->nodes[0]; | |
958 | slot = path->slots[0]; | |
959 | nritems = btrfs_header_nritems(leaf); | |
960 | ||
961 | if (btrfs_header_generation(leaf) > sk->max_transid) { | |
962 | i = nritems; | |
963 | goto advance_key; | |
964 | } | |
965 | found_transid = btrfs_header_generation(leaf); | |
966 | ||
967 | for (i = slot; i < nritems; i++) { | |
968 | item_off = btrfs_item_ptr_offset(leaf, i); | |
969 | item_len = btrfs_item_size_nr(leaf, i); | |
970 | ||
971 | if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE) | |
972 | item_len = 0; | |
973 | ||
974 | if (sizeof(sh) + item_len + *sk_offset > | |
975 | BTRFS_SEARCH_ARGS_BUFSIZE) { | |
976 | ret = 1; | |
977 | goto overflow; | |
978 | } | |
979 | ||
980 | btrfs_item_key_to_cpu(leaf, key, i); | |
981 | if (!key_in_sk(key, sk)) | |
982 | continue; | |
983 | ||
984 | sh.objectid = key->objectid; | |
985 | sh.offset = key->offset; | |
986 | sh.type = key->type; | |
987 | sh.len = item_len; | |
988 | sh.transid = found_transid; | |
989 | ||
990 | /* copy search result header */ | |
991 | memcpy(buf + *sk_offset, &sh, sizeof(sh)); | |
992 | *sk_offset += sizeof(sh); | |
993 | ||
994 | if (item_len) { | |
995 | char *p = buf + *sk_offset; | |
996 | /* copy the item */ | |
997 | read_extent_buffer(leaf, p, | |
998 | item_off, item_len); | |
999 | *sk_offset += item_len; | |
ac8e9819 | 1000 | } |
90fdde14 | 1001 | found++; |
ac8e9819 CM |
1002 | |
1003 | if (*num_found >= sk->nr_items) | |
1004 | break; | |
1005 | } | |
1006 | advance_key: | |
abc6e134 CM |
1007 | ret = 0; |
1008 | if (key->offset < (u64)-1 && key->offset < sk->max_offset) | |
ac8e9819 | 1009 | key->offset++; |
abc6e134 CM |
1010 | else if (key->type < (u8)-1 && key->type < sk->max_type) { |
1011 | key->offset = 0; | |
ac8e9819 | 1012 | key->type++; |
abc6e134 CM |
1013 | } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) { |
1014 | key->offset = 0; | |
1015 | key->type = 0; | |
ac8e9819 | 1016 | key->objectid++; |
abc6e134 CM |
1017 | } else |
1018 | ret = 1; | |
ac8e9819 CM |
1019 | overflow: |
1020 | *num_found += found; | |
1021 | return ret; | |
1022 | } | |
1023 | ||
1024 | static noinline int search_ioctl(struct inode *inode, | |
1025 | struct btrfs_ioctl_search_args *args) | |
1026 | { | |
1027 | struct btrfs_root *root; | |
1028 | struct btrfs_key key; | |
1029 | struct btrfs_key max_key; | |
1030 | struct btrfs_path *path; | |
1031 | struct btrfs_ioctl_search_key *sk = &args->key; | |
1032 | struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; | |
1033 | int ret; | |
1034 | int num_found = 0; | |
1035 | unsigned long sk_offset = 0; | |
1036 | ||
1037 | path = btrfs_alloc_path(); | |
1038 | if (!path) | |
1039 | return -ENOMEM; | |
1040 | ||
1041 | if (sk->tree_id == 0) { | |
1042 | /* search the root of the inode that was passed */ | |
1043 | root = BTRFS_I(inode)->root; | |
1044 | } else { | |
1045 | key.objectid = sk->tree_id; | |
1046 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1047 | key.offset = (u64)-1; | |
1048 | root = btrfs_read_fs_root_no_name(info, &key); | |
1049 | if (IS_ERR(root)) { | |
1050 | printk(KERN_ERR "could not find root %llu\n", | |
1051 | sk->tree_id); | |
1052 | btrfs_free_path(path); | |
1053 | return -ENOENT; | |
1054 | } | |
1055 | } | |
1056 | ||
1057 | key.objectid = sk->min_objectid; | |
1058 | key.type = sk->min_type; | |
1059 | key.offset = sk->min_offset; | |
1060 | ||
1061 | max_key.objectid = sk->max_objectid; | |
1062 | max_key.type = sk->max_type; | |
1063 | max_key.offset = sk->max_offset; | |
1064 | ||
1065 | path->keep_locks = 1; | |
1066 | ||
1067 | while(1) { | |
1068 | ret = btrfs_search_forward(root, &key, &max_key, path, 0, | |
1069 | sk->min_transid); | |
1070 | if (ret != 0) { | |
1071 | if (ret > 0) | |
1072 | ret = 0; | |
1073 | goto err; | |
1074 | } | |
1075 | ret = copy_to_sk(root, path, &key, sk, args->buf, | |
1076 | &sk_offset, &num_found); | |
1077 | btrfs_release_path(root, path); | |
1078 | if (ret || num_found >= sk->nr_items) | |
1079 | break; | |
1080 | ||
1081 | } | |
1082 | ret = 0; | |
1083 | err: | |
1084 | sk->nr_items = num_found; | |
1085 | btrfs_free_path(path); | |
1086 | return ret; | |
1087 | } | |
1088 | ||
1089 | static noinline int btrfs_ioctl_tree_search(struct file *file, | |
1090 | void __user *argp) | |
1091 | { | |
1092 | struct btrfs_ioctl_search_args *args; | |
1093 | struct inode *inode; | |
1094 | int ret; | |
1095 | ||
1096 | if (!capable(CAP_SYS_ADMIN)) | |
1097 | return -EPERM; | |
1098 | ||
1099 | args = kmalloc(sizeof(*args), GFP_KERNEL); | |
1100 | if (!args) | |
1101 | return -ENOMEM; | |
1102 | ||
1103 | if (copy_from_user(args, argp, sizeof(*args))) { | |
1104 | kfree(args); | |
1105 | return -EFAULT; | |
1106 | } | |
1107 | inode = fdentry(file)->d_inode; | |
1108 | ret = search_ioctl(inode, args); | |
1109 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) | |
1110 | ret = -EFAULT; | |
1111 | kfree(args); | |
1112 | return ret; | |
1113 | } | |
1114 | ||
98d377a0 | 1115 | /* |
ac8e9819 CM |
1116 | * Search INODE_REFs to identify path name of 'dirid' directory |
1117 | * in a 'tree_id' tree. and sets path name to 'name'. | |
1118 | */ | |
98d377a0 TH |
1119 | static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, |
1120 | u64 tree_id, u64 dirid, char *name) | |
1121 | { | |
1122 | struct btrfs_root *root; | |
1123 | struct btrfs_key key; | |
ac8e9819 | 1124 | char *ptr; |
98d377a0 TH |
1125 | int ret = -1; |
1126 | int slot; | |
1127 | int len; | |
1128 | int total_len = 0; | |
1129 | struct btrfs_inode_ref *iref; | |
1130 | struct extent_buffer *l; | |
1131 | struct btrfs_path *path; | |
1132 | ||
1133 | if (dirid == BTRFS_FIRST_FREE_OBJECTID) { | |
1134 | name[0]='\0'; | |
1135 | return 0; | |
1136 | } | |
1137 | ||
1138 | path = btrfs_alloc_path(); | |
1139 | if (!path) | |
1140 | return -ENOMEM; | |
1141 | ||
ac8e9819 | 1142 | ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; |
98d377a0 TH |
1143 | |
1144 | key.objectid = tree_id; | |
1145 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1146 | key.offset = (u64)-1; | |
1147 | root = btrfs_read_fs_root_no_name(info, &key); | |
1148 | if (IS_ERR(root)) { | |
1149 | printk(KERN_ERR "could not find root %llu\n", tree_id); | |
1150 | return -ENOENT; | |
1151 | } | |
1152 | ||
1153 | key.objectid = dirid; | |
1154 | key.type = BTRFS_INODE_REF_KEY; | |
1155 | key.offset = 0; | |
1156 | ||
1157 | while(1) { | |
1158 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
1159 | if (ret < 0) | |
1160 | goto out; | |
1161 | ||
1162 | l = path->nodes[0]; | |
1163 | slot = path->slots[0]; | |
1164 | btrfs_item_key_to_cpu(l, &key, slot); | |
1165 | ||
1166 | if (ret > 0 && (key.objectid != dirid || | |
ac8e9819 CM |
1167 | key.type != BTRFS_INODE_REF_KEY)) { |
1168 | ret = -ENOENT; | |
98d377a0 | 1169 | goto out; |
ac8e9819 | 1170 | } |
98d377a0 TH |
1171 | |
1172 | iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); | |
1173 | len = btrfs_inode_ref_name_len(l, iref); | |
1174 | ptr -= len + 1; | |
1175 | total_len += len + 1; | |
ac8e9819 | 1176 | if (ptr < name) |
98d377a0 TH |
1177 | goto out; |
1178 | ||
1179 | *(ptr + len) = '/'; | |
1180 | read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len); | |
1181 | ||
1182 | if (key.offset == BTRFS_FIRST_FREE_OBJECTID) | |
1183 | break; | |
1184 | ||
1185 | btrfs_release_path(root, path); | |
1186 | key.objectid = key.offset; | |
1187 | key.offset = 0; | |
1188 | dirid = key.objectid; | |
1189 | ||
1190 | } | |
ac8e9819 | 1191 | if (ptr < name) |
98d377a0 | 1192 | goto out; |
ac8e9819 | 1193 | memcpy(name, ptr, total_len); |
98d377a0 TH |
1194 | name[total_len]='\0'; |
1195 | ret = 0; | |
1196 | out: | |
1197 | btrfs_free_path(path); | |
ac8e9819 CM |
1198 | return ret; |
1199 | } | |
1200 | ||
1201 | static noinline int btrfs_ioctl_ino_lookup(struct file *file, | |
1202 | void __user *argp) | |
1203 | { | |
1204 | struct btrfs_ioctl_ino_lookup_args *args; | |
1205 | struct inode *inode; | |
1206 | int ret; | |
1207 | ||
1208 | if (!capable(CAP_SYS_ADMIN)) | |
1209 | return -EPERM; | |
1210 | ||
1211 | args = kmalloc(sizeof(*args), GFP_KERNEL); | |
1212 | if (copy_from_user(args, argp, sizeof(*args))) { | |
1213 | kfree(args); | |
1214 | return -EFAULT; | |
1215 | } | |
1216 | inode = fdentry(file)->d_inode; | |
1217 | ||
1b53ac4d CM |
1218 | if (args->treeid == 0) |
1219 | args->treeid = BTRFS_I(inode)->root->root_key.objectid; | |
1220 | ||
ac8e9819 CM |
1221 | ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, |
1222 | args->treeid, args->objectid, | |
1223 | args->name); | |
1224 | ||
1225 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) | |
1226 | ret = -EFAULT; | |
1227 | ||
1228 | kfree(args); | |
98d377a0 TH |
1229 | return ret; |
1230 | } | |
1231 | ||
76dda93c YZ |
1232 | static noinline int btrfs_ioctl_snap_destroy(struct file *file, |
1233 | void __user *arg) | |
1234 | { | |
1235 | struct dentry *parent = fdentry(file); | |
1236 | struct dentry *dentry; | |
1237 | struct inode *dir = parent->d_inode; | |
1238 | struct inode *inode; | |
1239 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
1240 | struct btrfs_root *dest = NULL; | |
1241 | struct btrfs_ioctl_vol_args *vol_args; | |
1242 | struct btrfs_trans_handle *trans; | |
1243 | int namelen; | |
1244 | int ret; | |
1245 | int err = 0; | |
1246 | ||
1247 | if (!capable(CAP_SYS_ADMIN)) | |
1248 | return -EPERM; | |
1249 | ||
1250 | vol_args = memdup_user(arg, sizeof(*vol_args)); | |
1251 | if (IS_ERR(vol_args)) | |
1252 | return PTR_ERR(vol_args); | |
1253 | ||
1254 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
1255 | namelen = strlen(vol_args->name); | |
1256 | if (strchr(vol_args->name, '/') || | |
1257 | strncmp(vol_args->name, "..", namelen) == 0) { | |
1258 | err = -EINVAL; | |
1259 | goto out; | |
1260 | } | |
1261 | ||
1262 | err = mnt_want_write(file->f_path.mnt); | |
1263 | if (err) | |
1264 | goto out; | |
1265 | ||
1266 | mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); | |
1267 | dentry = lookup_one_len(vol_args->name, parent, namelen); | |
1268 | if (IS_ERR(dentry)) { | |
1269 | err = PTR_ERR(dentry); | |
1270 | goto out_unlock_dir; | |
1271 | } | |
1272 | ||
1273 | if (!dentry->d_inode) { | |
1274 | err = -ENOENT; | |
1275 | goto out_dput; | |
1276 | } | |
1277 | ||
1278 | inode = dentry->d_inode; | |
1279 | if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) { | |
1280 | err = -EINVAL; | |
1281 | goto out_dput; | |
1282 | } | |
1283 | ||
1284 | dest = BTRFS_I(inode)->root; | |
1285 | ||
1286 | mutex_lock(&inode->i_mutex); | |
1287 | err = d_invalidate(dentry); | |
1288 | if (err) | |
1289 | goto out_unlock; | |
1290 | ||
1291 | down_write(&root->fs_info->subvol_sem); | |
1292 | ||
1293 | err = may_destroy_subvol(dest); | |
1294 | if (err) | |
1295 | goto out_up_write; | |
1296 | ||
1297 | trans = btrfs_start_transaction(root, 1); | |
1298 | ret = btrfs_unlink_subvol(trans, root, dir, | |
1299 | dest->root_key.objectid, | |
1300 | dentry->d_name.name, | |
1301 | dentry->d_name.len); | |
1302 | BUG_ON(ret); | |
1303 | ||
1304 | btrfs_record_root_in_trans(trans, dest); | |
1305 | ||
1306 | memset(&dest->root_item.drop_progress, 0, | |
1307 | sizeof(dest->root_item.drop_progress)); | |
1308 | dest->root_item.drop_level = 0; | |
1309 | btrfs_set_root_refs(&dest->root_item, 0); | |
1310 | ||
1311 | ret = btrfs_insert_orphan_item(trans, | |
1312 | root->fs_info->tree_root, | |
1313 | dest->root_key.objectid); | |
1314 | BUG_ON(ret); | |
1315 | ||
1316 | ret = btrfs_commit_transaction(trans, root); | |
1317 | BUG_ON(ret); | |
1318 | inode->i_flags |= S_DEAD; | |
1319 | out_up_write: | |
1320 | up_write(&root->fs_info->subvol_sem); | |
1321 | out_unlock: | |
1322 | mutex_unlock(&inode->i_mutex); | |
1323 | if (!err) { | |
efefb143 | 1324 | shrink_dcache_sb(root->fs_info->sb); |
76dda93c YZ |
1325 | btrfs_invalidate_inodes(dest); |
1326 | d_delete(dentry); | |
1327 | } | |
1328 | out_dput: | |
1329 | dput(dentry); | |
1330 | out_unlock_dir: | |
1331 | mutex_unlock(&dir->i_mutex); | |
1332 | mnt_drop_write(file->f_path.mnt); | |
1333 | out: | |
1334 | kfree(vol_args); | |
1335 | return err; | |
1336 | } | |
1337 | ||
1e701a32 | 1338 | static int btrfs_ioctl_defrag(struct file *file, void __user *argp) |
f46b5a66 CH |
1339 | { |
1340 | struct inode *inode = fdentry(file)->d_inode; | |
1341 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
1e701a32 | 1342 | struct btrfs_ioctl_defrag_range_args *range; |
c146afad YZ |
1343 | int ret; |
1344 | ||
1345 | ret = mnt_want_write(file->f_path.mnt); | |
1346 | if (ret) | |
1347 | return ret; | |
f46b5a66 CH |
1348 | |
1349 | switch (inode->i_mode & S_IFMT) { | |
1350 | case S_IFDIR: | |
e441d54d CM |
1351 | if (!capable(CAP_SYS_ADMIN)) { |
1352 | ret = -EPERM; | |
1353 | goto out; | |
1354 | } | |
f46b5a66 CH |
1355 | btrfs_defrag_root(root, 0); |
1356 | btrfs_defrag_root(root->fs_info->extent_root, 0); | |
f46b5a66 CH |
1357 | break; |
1358 | case S_IFREG: | |
e441d54d CM |
1359 | if (!(file->f_mode & FMODE_WRITE)) { |
1360 | ret = -EINVAL; | |
1361 | goto out; | |
1362 | } | |
1e701a32 CM |
1363 | |
1364 | range = kzalloc(sizeof(*range), GFP_KERNEL); | |
1365 | if (!range) { | |
1366 | ret = -ENOMEM; | |
1367 | goto out; | |
1368 | } | |
1369 | ||
1370 | if (argp) { | |
1371 | if (copy_from_user(range, argp, | |
1372 | sizeof(*range))) { | |
1373 | ret = -EFAULT; | |
1374 | kfree(range); | |
1375 | } | |
1376 | /* compression requires us to start the IO */ | |
1377 | if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
1378 | range->flags |= BTRFS_DEFRAG_RANGE_START_IO; | |
1379 | range->extent_thresh = (u32)-1; | |
1380 | } | |
1381 | } else { | |
1382 | /* the rest are all set to zero by kzalloc */ | |
1383 | range->len = (u64)-1; | |
1384 | } | |
1385 | btrfs_defrag_file(file, range); | |
1386 | kfree(range); | |
f46b5a66 CH |
1387 | break; |
1388 | } | |
e441d54d | 1389 | out: |
ab67b7c1 | 1390 | mnt_drop_write(file->f_path.mnt); |
e441d54d | 1391 | return ret; |
f46b5a66 CH |
1392 | } |
1393 | ||
b2950863 | 1394 | static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) |
f46b5a66 CH |
1395 | { |
1396 | struct btrfs_ioctl_vol_args *vol_args; | |
1397 | int ret; | |
1398 | ||
e441d54d CM |
1399 | if (!capable(CAP_SYS_ADMIN)) |
1400 | return -EPERM; | |
1401 | ||
dae7b665 LZ |
1402 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1403 | if (IS_ERR(vol_args)) | |
1404 | return PTR_ERR(vol_args); | |
f46b5a66 | 1405 | |
5516e595 | 1406 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 CH |
1407 | ret = btrfs_init_new_device(root, vol_args->name); |
1408 | ||
f46b5a66 CH |
1409 | kfree(vol_args); |
1410 | return ret; | |
1411 | } | |
1412 | ||
b2950863 | 1413 | static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) |
f46b5a66 CH |
1414 | { |
1415 | struct btrfs_ioctl_vol_args *vol_args; | |
1416 | int ret; | |
1417 | ||
e441d54d CM |
1418 | if (!capable(CAP_SYS_ADMIN)) |
1419 | return -EPERM; | |
1420 | ||
c146afad YZ |
1421 | if (root->fs_info->sb->s_flags & MS_RDONLY) |
1422 | return -EROFS; | |
1423 | ||
dae7b665 LZ |
1424 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1425 | if (IS_ERR(vol_args)) | |
1426 | return PTR_ERR(vol_args); | |
f46b5a66 | 1427 | |
5516e595 | 1428 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 CH |
1429 | ret = btrfs_rm_device(root, vol_args->name); |
1430 | ||
f46b5a66 CH |
1431 | kfree(vol_args); |
1432 | return ret; | |
1433 | } | |
1434 | ||
76dda93c YZ |
1435 | static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, |
1436 | u64 off, u64 olen, u64 destoff) | |
f46b5a66 CH |
1437 | { |
1438 | struct inode *inode = fdentry(file)->d_inode; | |
1439 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
1440 | struct file *src_file; | |
1441 | struct inode *src; | |
1442 | struct btrfs_trans_handle *trans; | |
f46b5a66 | 1443 | struct btrfs_path *path; |
f46b5a66 | 1444 | struct extent_buffer *leaf; |
ae01a0ab YZ |
1445 | char *buf; |
1446 | struct btrfs_key key; | |
f46b5a66 CH |
1447 | u32 nritems; |
1448 | int slot; | |
ae01a0ab | 1449 | int ret; |
c5c9cd4d SW |
1450 | u64 len = olen; |
1451 | u64 bs = root->fs_info->sb->s_blocksize; | |
1452 | u64 hint_byte; | |
d20f7043 | 1453 | |
c5c9cd4d SW |
1454 | /* |
1455 | * TODO: | |
1456 | * - split compressed inline extents. annoying: we need to | |
1457 | * decompress into destination's address_space (the file offset | |
1458 | * may change, so source mapping won't do), then recompress (or | |
1459 | * otherwise reinsert) a subrange. | |
1460 | * - allow ranges within the same file to be cloned (provided | |
1461 | * they don't overlap)? | |
1462 | */ | |
1463 | ||
e441d54d CM |
1464 | /* the destination must be opened for writing */ |
1465 | if (!(file->f_mode & FMODE_WRITE)) | |
1466 | return -EINVAL; | |
1467 | ||
c146afad YZ |
1468 | ret = mnt_want_write(file->f_path.mnt); |
1469 | if (ret) | |
1470 | return ret; | |
1471 | ||
c5c9cd4d | 1472 | src_file = fget(srcfd); |
ab67b7c1 YZ |
1473 | if (!src_file) { |
1474 | ret = -EBADF; | |
1475 | goto out_drop_write; | |
1476 | } | |
f46b5a66 CH |
1477 | src = src_file->f_dentry->d_inode; |
1478 | ||
c5c9cd4d SW |
1479 | ret = -EINVAL; |
1480 | if (src == inode) | |
1481 | goto out_fput; | |
1482 | ||
ae01a0ab YZ |
1483 | ret = -EISDIR; |
1484 | if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) | |
1485 | goto out_fput; | |
1486 | ||
f46b5a66 | 1487 | ret = -EXDEV; |
ae01a0ab YZ |
1488 | if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root) |
1489 | goto out_fput; | |
1490 | ||
1491 | ret = -ENOMEM; | |
1492 | buf = vmalloc(btrfs_level_size(root, 0)); | |
1493 | if (!buf) | |
1494 | goto out_fput; | |
1495 | ||
1496 | path = btrfs_alloc_path(); | |
1497 | if (!path) { | |
1498 | vfree(buf); | |
f46b5a66 | 1499 | goto out_fput; |
ae01a0ab YZ |
1500 | } |
1501 | path->reada = 2; | |
f46b5a66 CH |
1502 | |
1503 | if (inode < src) { | |
1504 | mutex_lock(&inode->i_mutex); | |
1505 | mutex_lock(&src->i_mutex); | |
1506 | } else { | |
1507 | mutex_lock(&src->i_mutex); | |
1508 | mutex_lock(&inode->i_mutex); | |
1509 | } | |
1510 | ||
c5c9cd4d SW |
1511 | /* determine range to clone */ |
1512 | ret = -EINVAL; | |
1513 | if (off >= src->i_size || off + len > src->i_size) | |
f46b5a66 | 1514 | goto out_unlock; |
c5c9cd4d SW |
1515 | if (len == 0) |
1516 | olen = len = src->i_size - off; | |
1517 | /* if we extend to eof, continue to block boundary */ | |
1518 | if (off + len == src->i_size) | |
1519 | len = ((src->i_size + bs-1) & ~(bs-1)) | |
1520 | - off; | |
1521 | ||
1522 | /* verify the end result is block aligned */ | |
1523 | if ((off & (bs-1)) || | |
1524 | ((off + len) & (bs-1))) | |
1525 | goto out_unlock; | |
1526 | ||
f46b5a66 CH |
1527 | /* do any pending delalloc/csum calc on src, one way or |
1528 | another, and lock file content */ | |
1529 | while (1) { | |
31840ae1 | 1530 | struct btrfs_ordered_extent *ordered; |
c5c9cd4d SW |
1531 | lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); |
1532 | ordered = btrfs_lookup_first_ordered_extent(inode, off+len); | |
ae01a0ab | 1533 | if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered) |
f46b5a66 | 1534 | break; |
c5c9cd4d | 1535 | unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); |
ae01a0ab YZ |
1536 | if (ordered) |
1537 | btrfs_put_ordered_extent(ordered); | |
c5c9cd4d | 1538 | btrfs_wait_ordered_range(src, off, off+len); |
f46b5a66 CH |
1539 | } |
1540 | ||
ae01a0ab YZ |
1541 | trans = btrfs_start_transaction(root, 1); |
1542 | BUG_ON(!trans); | |
1543 | ||
c5c9cd4d | 1544 | /* punch hole in destination first */ |
920bbbfb | 1545 | btrfs_drop_extents(trans, inode, off, off + len, &hint_byte, 1); |
c5c9cd4d SW |
1546 | |
1547 | /* clone data */ | |
f46b5a66 | 1548 | key.objectid = src->i_ino; |
ae01a0ab YZ |
1549 | key.type = BTRFS_EXTENT_DATA_KEY; |
1550 | key.offset = 0; | |
f46b5a66 CH |
1551 | |
1552 | while (1) { | |
1553 | /* | |
1554 | * note the key will change type as we walk through the | |
1555 | * tree. | |
1556 | */ | |
1557 | ret = btrfs_search_slot(trans, root, &key, path, 0, 0); | |
1558 | if (ret < 0) | |
1559 | goto out; | |
1560 | ||
ae01a0ab YZ |
1561 | nritems = btrfs_header_nritems(path->nodes[0]); |
1562 | if (path->slots[0] >= nritems) { | |
f46b5a66 CH |
1563 | ret = btrfs_next_leaf(root, path); |
1564 | if (ret < 0) | |
1565 | goto out; | |
1566 | if (ret > 0) | |
1567 | break; | |
ae01a0ab | 1568 | nritems = btrfs_header_nritems(path->nodes[0]); |
f46b5a66 CH |
1569 | } |
1570 | leaf = path->nodes[0]; | |
1571 | slot = path->slots[0]; | |
f46b5a66 | 1572 | |
ae01a0ab | 1573 | btrfs_item_key_to_cpu(leaf, &key, slot); |
d20f7043 | 1574 | if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || |
f46b5a66 CH |
1575 | key.objectid != src->i_ino) |
1576 | break; | |
1577 | ||
c5c9cd4d SW |
1578 | if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { |
1579 | struct btrfs_file_extent_item *extent; | |
1580 | int type; | |
31840ae1 ZY |
1581 | u32 size; |
1582 | struct btrfs_key new_key; | |
c5c9cd4d SW |
1583 | u64 disko = 0, diskl = 0; |
1584 | u64 datao = 0, datal = 0; | |
1585 | u8 comp; | |
31840ae1 ZY |
1586 | |
1587 | size = btrfs_item_size_nr(leaf, slot); | |
1588 | read_extent_buffer(leaf, buf, | |
1589 | btrfs_item_ptr_offset(leaf, slot), | |
1590 | size); | |
c5c9cd4d SW |
1591 | |
1592 | extent = btrfs_item_ptr(leaf, slot, | |
1593 | struct btrfs_file_extent_item); | |
1594 | comp = btrfs_file_extent_compression(leaf, extent); | |
1595 | type = btrfs_file_extent_type(leaf, extent); | |
c8a894d7 CM |
1596 | if (type == BTRFS_FILE_EXTENT_REG || |
1597 | type == BTRFS_FILE_EXTENT_PREALLOC) { | |
d397712b CM |
1598 | disko = btrfs_file_extent_disk_bytenr(leaf, |
1599 | extent); | |
1600 | diskl = btrfs_file_extent_disk_num_bytes(leaf, | |
1601 | extent); | |
c5c9cd4d | 1602 | datao = btrfs_file_extent_offset(leaf, extent); |
d397712b CM |
1603 | datal = btrfs_file_extent_num_bytes(leaf, |
1604 | extent); | |
c5c9cd4d SW |
1605 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
1606 | /* take upper bound, may be compressed */ | |
1607 | datal = btrfs_file_extent_ram_bytes(leaf, | |
1608 | extent); | |
1609 | } | |
31840ae1 ZY |
1610 | btrfs_release_path(root, path); |
1611 | ||
c5c9cd4d SW |
1612 | if (key.offset + datal < off || |
1613 | key.offset >= off+len) | |
1614 | goto next; | |
1615 | ||
31840ae1 ZY |
1616 | memcpy(&new_key, &key, sizeof(new_key)); |
1617 | new_key.objectid = inode->i_ino; | |
c5c9cd4d | 1618 | new_key.offset = key.offset + destoff - off; |
31840ae1 | 1619 | |
c8a894d7 CM |
1620 | if (type == BTRFS_FILE_EXTENT_REG || |
1621 | type == BTRFS_FILE_EXTENT_PREALLOC) { | |
c5c9cd4d SW |
1622 | ret = btrfs_insert_empty_item(trans, root, path, |
1623 | &new_key, size); | |
1624 | if (ret) | |
1625 | goto out; | |
1626 | ||
1627 | leaf = path->nodes[0]; | |
1628 | slot = path->slots[0]; | |
1629 | write_extent_buffer(leaf, buf, | |
31840ae1 ZY |
1630 | btrfs_item_ptr_offset(leaf, slot), |
1631 | size); | |
ae01a0ab | 1632 | |
c5c9cd4d | 1633 | extent = btrfs_item_ptr(leaf, slot, |
f46b5a66 | 1634 | struct btrfs_file_extent_item); |
c5c9cd4d SW |
1635 | |
1636 | if (off > key.offset) { | |
1637 | datao += off - key.offset; | |
1638 | datal -= off - key.offset; | |
1639 | } | |
ac6889cb CM |
1640 | |
1641 | if (key.offset + datal > off + len) | |
1642 | datal = off + len - key.offset; | |
1643 | ||
c5c9cd4d SW |
1644 | /* disko == 0 means it's a hole */ |
1645 | if (!disko) | |
1646 | datao = 0; | |
c5c9cd4d SW |
1647 | |
1648 | btrfs_set_file_extent_offset(leaf, extent, | |
1649 | datao); | |
1650 | btrfs_set_file_extent_num_bytes(leaf, extent, | |
1651 | datal); | |
1652 | if (disko) { | |
1653 | inode_add_bytes(inode, datal); | |
ae01a0ab | 1654 | ret = btrfs_inc_extent_ref(trans, root, |
5d4f98a2 YZ |
1655 | disko, diskl, 0, |
1656 | root->root_key.objectid, | |
1657 | inode->i_ino, | |
1658 | new_key.offset - datao); | |
31840ae1 | 1659 | BUG_ON(ret); |
f46b5a66 | 1660 | } |
c5c9cd4d SW |
1661 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
1662 | u64 skip = 0; | |
1663 | u64 trim = 0; | |
1664 | if (off > key.offset) { | |
1665 | skip = off - key.offset; | |
1666 | new_key.offset += skip; | |
1667 | } | |
d397712b | 1668 | |
c5c9cd4d SW |
1669 | if (key.offset + datal > off+len) |
1670 | trim = key.offset + datal - (off+len); | |
d397712b | 1671 | |
c5c9cd4d | 1672 | if (comp && (skip || trim)) { |
c5c9cd4d SW |
1673 | ret = -EINVAL; |
1674 | goto out; | |
1675 | } | |
1676 | size -= skip + trim; | |
1677 | datal -= skip + trim; | |
1678 | ret = btrfs_insert_empty_item(trans, root, path, | |
1679 | &new_key, size); | |
1680 | if (ret) | |
1681 | goto out; | |
1682 | ||
1683 | if (skip) { | |
d397712b CM |
1684 | u32 start = |
1685 | btrfs_file_extent_calc_inline_size(0); | |
c5c9cd4d SW |
1686 | memmove(buf+start, buf+start+skip, |
1687 | datal); | |
1688 | } | |
1689 | ||
1690 | leaf = path->nodes[0]; | |
1691 | slot = path->slots[0]; | |
1692 | write_extent_buffer(leaf, buf, | |
1693 | btrfs_item_ptr_offset(leaf, slot), | |
1694 | size); | |
1695 | inode_add_bytes(inode, datal); | |
f46b5a66 | 1696 | } |
c5c9cd4d SW |
1697 | |
1698 | btrfs_mark_buffer_dirty(leaf); | |
ae01a0ab | 1699 | } |
c5c9cd4d | 1700 | |
d397712b | 1701 | next: |
31840ae1 | 1702 | btrfs_release_path(root, path); |
f46b5a66 | 1703 | key.offset++; |
f46b5a66 | 1704 | } |
f46b5a66 CH |
1705 | ret = 0; |
1706 | out: | |
ae01a0ab YZ |
1707 | btrfs_release_path(root, path); |
1708 | if (ret == 0) { | |
1709 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
c5c9cd4d SW |
1710 | if (destoff + olen > inode->i_size) |
1711 | btrfs_i_size_write(inode, destoff + olen); | |
ae01a0ab YZ |
1712 | BTRFS_I(inode)->flags = BTRFS_I(src)->flags; |
1713 | ret = btrfs_update_inode(trans, root, inode); | |
1714 | } | |
f46b5a66 | 1715 | btrfs_end_transaction(trans, root); |
c5c9cd4d | 1716 | unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); |
ae01a0ab YZ |
1717 | if (ret) |
1718 | vmtruncate(inode, 0); | |
f46b5a66 CH |
1719 | out_unlock: |
1720 | mutex_unlock(&src->i_mutex); | |
1721 | mutex_unlock(&inode->i_mutex); | |
ae01a0ab YZ |
1722 | vfree(buf); |
1723 | btrfs_free_path(path); | |
f46b5a66 CH |
1724 | out_fput: |
1725 | fput(src_file); | |
ab67b7c1 YZ |
1726 | out_drop_write: |
1727 | mnt_drop_write(file->f_path.mnt); | |
f46b5a66 CH |
1728 | return ret; |
1729 | } | |
1730 | ||
7a865e8a | 1731 | static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) |
c5c9cd4d SW |
1732 | { |
1733 | struct btrfs_ioctl_clone_range_args args; | |
1734 | ||
7a865e8a | 1735 | if (copy_from_user(&args, argp, sizeof(args))) |
c5c9cd4d SW |
1736 | return -EFAULT; |
1737 | return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, | |
1738 | args.src_length, args.dest_offset); | |
1739 | } | |
1740 | ||
f46b5a66 CH |
1741 | /* |
1742 | * there are many ways the trans_start and trans_end ioctls can lead | |
1743 | * to deadlocks. They should only be used by applications that | |
1744 | * basically own the machine, and have a very in depth understanding | |
1745 | * of all the possible deadlocks and enospc problems. | |
1746 | */ | |
b2950863 | 1747 | static long btrfs_ioctl_trans_start(struct file *file) |
f46b5a66 CH |
1748 | { |
1749 | struct inode *inode = fdentry(file)->d_inode; | |
1750 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
1751 | struct btrfs_trans_handle *trans; | |
1ab86aed | 1752 | int ret; |
f46b5a66 | 1753 | |
1ab86aed | 1754 | ret = -EPERM; |
df5b5520 | 1755 | if (!capable(CAP_SYS_ADMIN)) |
1ab86aed | 1756 | goto out; |
df5b5520 | 1757 | |
1ab86aed SW |
1758 | ret = -EINPROGRESS; |
1759 | if (file->private_data) | |
f46b5a66 | 1760 | goto out; |
9ca9ee09 | 1761 | |
c146afad YZ |
1762 | ret = mnt_want_write(file->f_path.mnt); |
1763 | if (ret) | |
1764 | goto out; | |
1765 | ||
9ca9ee09 SW |
1766 | mutex_lock(&root->fs_info->trans_mutex); |
1767 | root->fs_info->open_ioctl_trans++; | |
1768 | mutex_unlock(&root->fs_info->trans_mutex); | |
1769 | ||
1ab86aed | 1770 | ret = -ENOMEM; |
9ca9ee09 | 1771 | trans = btrfs_start_ioctl_transaction(root, 0); |
1ab86aed SW |
1772 | if (!trans) |
1773 | goto out_drop; | |
1774 | ||
1775 | file->private_data = trans; | |
1776 | return 0; | |
1777 | ||
1778 | out_drop: | |
1779 | mutex_lock(&root->fs_info->trans_mutex); | |
1780 | root->fs_info->open_ioctl_trans--; | |
1781 | mutex_unlock(&root->fs_info->trans_mutex); | |
1782 | mnt_drop_write(file->f_path.mnt); | |
f46b5a66 | 1783 | out: |
f46b5a66 CH |
1784 | return ret; |
1785 | } | |
1786 | ||
6ef5ed0d JB |
1787 | static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) |
1788 | { | |
1789 | struct inode *inode = fdentry(file)->d_inode; | |
1790 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
1791 | struct btrfs_root *new_root; | |
1792 | struct btrfs_dir_item *di; | |
1793 | struct btrfs_trans_handle *trans; | |
1794 | struct btrfs_path *path; | |
1795 | struct btrfs_key location; | |
1796 | struct btrfs_disk_key disk_key; | |
1797 | struct btrfs_super_block *disk_super; | |
1798 | u64 features; | |
1799 | u64 objectid = 0; | |
1800 | u64 dir_id; | |
1801 | ||
1802 | if (!capable(CAP_SYS_ADMIN)) | |
1803 | return -EPERM; | |
1804 | ||
1805 | if (copy_from_user(&objectid, argp, sizeof(objectid))) | |
1806 | return -EFAULT; | |
1807 | ||
1808 | if (!objectid) | |
1809 | objectid = root->root_key.objectid; | |
1810 | ||
1811 | location.objectid = objectid; | |
1812 | location.type = BTRFS_ROOT_ITEM_KEY; | |
1813 | location.offset = (u64)-1; | |
1814 | ||
1815 | new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); | |
1816 | if (IS_ERR(new_root)) | |
1817 | return PTR_ERR(new_root); | |
1818 | ||
1819 | if (btrfs_root_refs(&new_root->root_item) == 0) | |
1820 | return -ENOENT; | |
1821 | ||
1822 | path = btrfs_alloc_path(); | |
1823 | if (!path) | |
1824 | return -ENOMEM; | |
1825 | path->leave_spinning = 1; | |
1826 | ||
1827 | trans = btrfs_start_transaction(root, 1); | |
1828 | if (!trans) { | |
1829 | btrfs_free_path(path); | |
1830 | return -ENOMEM; | |
1831 | } | |
1832 | ||
1833 | dir_id = btrfs_super_root_dir(&root->fs_info->super_copy); | |
1834 | di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, | |
1835 | dir_id, "default", 7, 1); | |
1836 | if (!di) { | |
1837 | btrfs_free_path(path); | |
1838 | btrfs_end_transaction(trans, root); | |
1839 | printk(KERN_ERR "Umm, you don't have the default dir item, " | |
1840 | "this isn't going to work\n"); | |
1841 | return -ENOENT; | |
1842 | } | |
1843 | ||
1844 | btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); | |
1845 | btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); | |
1846 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
1847 | btrfs_free_path(path); | |
1848 | ||
1849 | disk_super = &root->fs_info->super_copy; | |
1850 | features = btrfs_super_incompat_flags(disk_super); | |
1851 | if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) { | |
1852 | features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL; | |
1853 | btrfs_set_super_incompat_flags(disk_super, features); | |
1854 | } | |
1855 | btrfs_end_transaction(trans, root); | |
1856 | ||
1857 | return 0; | |
1858 | } | |
1859 | ||
1406e432 JB |
1860 | long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) |
1861 | { | |
1862 | struct btrfs_ioctl_space_args space_args; | |
1863 | struct btrfs_ioctl_space_info space; | |
1864 | struct btrfs_ioctl_space_info *dest; | |
7fde62bf CM |
1865 | struct btrfs_ioctl_space_info *dest_orig; |
1866 | struct btrfs_ioctl_space_info *user_dest; | |
1406e432 | 1867 | struct btrfs_space_info *info; |
7fde62bf | 1868 | int alloc_size; |
1406e432 | 1869 | int ret = 0; |
7fde62bf | 1870 | int slot_count = 0; |
1406e432 JB |
1871 | |
1872 | if (copy_from_user(&space_args, | |
1873 | (struct btrfs_ioctl_space_args __user *)arg, | |
1874 | sizeof(space_args))) | |
1875 | return -EFAULT; | |
1876 | ||
7fde62bf CM |
1877 | /* first we count slots */ |
1878 | rcu_read_lock(); | |
1879 | list_for_each_entry_rcu(info, &root->fs_info->space_info, list) | |
1880 | slot_count++; | |
1881 | rcu_read_unlock(); | |
1882 | ||
1883 | /* space_slots == 0 means they are asking for a count */ | |
1884 | if (space_args.space_slots == 0) { | |
1885 | space_args.total_spaces = slot_count; | |
1886 | goto out; | |
1887 | } | |
1888 | alloc_size = sizeof(*dest) * slot_count; | |
1889 | /* we generally have at most 6 or so space infos, one for each raid | |
1890 | * level. So, a whole page should be more than enough for everyone | |
1891 | */ | |
1892 | if (alloc_size > PAGE_CACHE_SIZE) | |
1893 | return -ENOMEM; | |
1894 | ||
1406e432 | 1895 | space_args.total_spaces = 0; |
7fde62bf CM |
1896 | dest = kmalloc(alloc_size, GFP_NOFS); |
1897 | if (!dest) | |
1898 | return -ENOMEM; | |
1899 | dest_orig = dest; | |
1406e432 | 1900 | |
7fde62bf | 1901 | /* now we have a buffer to copy into */ |
1406e432 JB |
1902 | rcu_read_lock(); |
1903 | list_for_each_entry_rcu(info, &root->fs_info->space_info, list) { | |
7fde62bf CM |
1904 | /* make sure we don't copy more than we allocated |
1905 | * in our buffer | |
1906 | */ | |
1907 | if (slot_count == 0) | |
1908 | break; | |
1909 | slot_count--; | |
1910 | ||
1911 | /* make sure userland has enough room in their buffer */ | |
1406e432 JB |
1912 | if (space_args.total_spaces >= space_args.space_slots) |
1913 | break; | |
7fde62bf | 1914 | |
1406e432 JB |
1915 | space.flags = info->flags; |
1916 | space.total_bytes = info->total_bytes; | |
1917 | space.used_bytes = info->bytes_used; | |
7fde62bf | 1918 | memcpy(dest, &space, sizeof(space)); |
1406e432 JB |
1919 | dest++; |
1920 | space_args.total_spaces++; | |
1921 | } | |
1922 | rcu_read_unlock(); | |
1923 | ||
7fde62bf CM |
1924 | user_dest = (struct btrfs_ioctl_space_info *) |
1925 | (arg + sizeof(struct btrfs_ioctl_space_args)); | |
1926 | ||
1927 | if (copy_to_user(user_dest, dest_orig, alloc_size)) | |
1928 | ret = -EFAULT; | |
1929 | ||
1930 | kfree(dest_orig); | |
1931 | out: | |
1932 | if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) | |
1406e432 JB |
1933 | ret = -EFAULT; |
1934 | ||
1935 | return ret; | |
1936 | } | |
1937 | ||
f46b5a66 CH |
1938 | /* |
1939 | * there are many ways the trans_start and trans_end ioctls can lead | |
1940 | * to deadlocks. They should only be used by applications that | |
1941 | * basically own the machine, and have a very in depth understanding | |
1942 | * of all the possible deadlocks and enospc problems. | |
1943 | */ | |
1944 | long btrfs_ioctl_trans_end(struct file *file) | |
1945 | { | |
1946 | struct inode *inode = fdentry(file)->d_inode; | |
1947 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
1948 | struct btrfs_trans_handle *trans; | |
f46b5a66 | 1949 | |
f46b5a66 | 1950 | trans = file->private_data; |
1ab86aed SW |
1951 | if (!trans) |
1952 | return -EINVAL; | |
b214107e | 1953 | file->private_data = NULL; |
9ca9ee09 | 1954 | |
1ab86aed SW |
1955 | btrfs_end_transaction(trans, root); |
1956 | ||
9ca9ee09 SW |
1957 | mutex_lock(&root->fs_info->trans_mutex); |
1958 | root->fs_info->open_ioctl_trans--; | |
1959 | mutex_unlock(&root->fs_info->trans_mutex); | |
1960 | ||
cfc8ea87 | 1961 | mnt_drop_write(file->f_path.mnt); |
1ab86aed | 1962 | return 0; |
f46b5a66 CH |
1963 | } |
1964 | ||
1965 | long btrfs_ioctl(struct file *file, unsigned int | |
1966 | cmd, unsigned long arg) | |
1967 | { | |
1968 | struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; | |
4bcabaa3 | 1969 | void __user *argp = (void __user *)arg; |
f46b5a66 CH |
1970 | |
1971 | switch (cmd) { | |
6cbff00f CH |
1972 | case FS_IOC_GETFLAGS: |
1973 | return btrfs_ioctl_getflags(file, argp); | |
1974 | case FS_IOC_SETFLAGS: | |
1975 | return btrfs_ioctl_setflags(file, argp); | |
1976 | case FS_IOC_GETVERSION: | |
1977 | return btrfs_ioctl_getversion(file, argp); | |
f46b5a66 | 1978 | case BTRFS_IOC_SNAP_CREATE: |
4bcabaa3 | 1979 | return btrfs_ioctl_snap_create(file, argp, 0); |
3de4586c | 1980 | case BTRFS_IOC_SUBVOL_CREATE: |
4bcabaa3 | 1981 | return btrfs_ioctl_snap_create(file, argp, 1); |
76dda93c YZ |
1982 | case BTRFS_IOC_SNAP_DESTROY: |
1983 | return btrfs_ioctl_snap_destroy(file, argp); | |
6ef5ed0d JB |
1984 | case BTRFS_IOC_DEFAULT_SUBVOL: |
1985 | return btrfs_ioctl_default_subvol(file, argp); | |
f46b5a66 | 1986 | case BTRFS_IOC_DEFRAG: |
1e701a32 CM |
1987 | return btrfs_ioctl_defrag(file, NULL); |
1988 | case BTRFS_IOC_DEFRAG_RANGE: | |
1989 | return btrfs_ioctl_defrag(file, argp); | |
f46b5a66 | 1990 | case BTRFS_IOC_RESIZE: |
4bcabaa3 | 1991 | return btrfs_ioctl_resize(root, argp); |
f46b5a66 | 1992 | case BTRFS_IOC_ADD_DEV: |
4bcabaa3 | 1993 | return btrfs_ioctl_add_dev(root, argp); |
f46b5a66 | 1994 | case BTRFS_IOC_RM_DEV: |
4bcabaa3 | 1995 | return btrfs_ioctl_rm_dev(root, argp); |
f46b5a66 CH |
1996 | case BTRFS_IOC_BALANCE: |
1997 | return btrfs_balance(root->fs_info->dev_root); | |
1998 | case BTRFS_IOC_CLONE: | |
c5c9cd4d SW |
1999 | return btrfs_ioctl_clone(file, arg, 0, 0, 0); |
2000 | case BTRFS_IOC_CLONE_RANGE: | |
7a865e8a | 2001 | return btrfs_ioctl_clone_range(file, argp); |
f46b5a66 CH |
2002 | case BTRFS_IOC_TRANS_START: |
2003 | return btrfs_ioctl_trans_start(file); | |
2004 | case BTRFS_IOC_TRANS_END: | |
2005 | return btrfs_ioctl_trans_end(file); | |
ac8e9819 CM |
2006 | case BTRFS_IOC_TREE_SEARCH: |
2007 | return btrfs_ioctl_tree_search(file, argp); | |
2008 | case BTRFS_IOC_INO_LOOKUP: | |
2009 | return btrfs_ioctl_ino_lookup(file, argp); | |
1406e432 JB |
2010 | case BTRFS_IOC_SPACE_INFO: |
2011 | return btrfs_ioctl_space_info(root, argp); | |
f46b5a66 CH |
2012 | case BTRFS_IOC_SYNC: |
2013 | btrfs_sync_fs(file->f_dentry->d_sb, 1); | |
2014 | return 0; | |
2015 | } | |
2016 | ||
2017 | return -ENOTTY; | |
2018 | } |