]>
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> | |
30 | #include <linux/smp_lock.h> | |
31 | #include <linux/backing-dev.h> | |
cb8e7090 | 32 | #include <linux/mount.h> |
f46b5a66 | 33 | #include <linux/mpage.h> |
cb8e7090 | 34 | #include <linux/namei.h> |
f46b5a66 CH |
35 | #include <linux/swap.h> |
36 | #include <linux/writeback.h> | |
37 | #include <linux/statfs.h> | |
38 | #include <linux/compat.h> | |
39 | #include <linux/bit_spinlock.h> | |
cb8e7090 | 40 | #include <linux/security.h> |
f46b5a66 CH |
41 | #include <linux/version.h> |
42 | #include <linux/xattr.h> | |
7ea394f1 | 43 | #include <linux/vmalloc.h> |
f46b5a66 CH |
44 | #include "ctree.h" |
45 | #include "disk-io.h" | |
46 | #include "transaction.h" | |
47 | #include "btrfs_inode.h" | |
48 | #include "ioctl.h" | |
49 | #include "print-tree.h" | |
50 | #include "volumes.h" | |
925baedd | 51 | #include "locking.h" |
f46b5a66 CH |
52 | |
53 | ||
54 | ||
cb8e7090 CH |
55 | static noinline int create_subvol(struct btrfs_root *root, |
56 | struct dentry *dentry, | |
57 | char *name, int namelen) | |
f46b5a66 CH |
58 | { |
59 | struct btrfs_trans_handle *trans; | |
60 | struct btrfs_key key; | |
61 | struct btrfs_root_item root_item; | |
62 | struct btrfs_inode_item *inode_item; | |
63 | struct extent_buffer *leaf; | |
64 | struct btrfs_root *new_root = root; | |
65 | struct inode *dir; | |
66 | int ret; | |
67 | int err; | |
68 | u64 objectid; | |
69 | u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; | |
70 | unsigned long nr = 1; | |
71 | ||
f46b5a66 CH |
72 | ret = btrfs_check_free_space(root, 1, 0); |
73 | if (ret) | |
74 | goto fail_commit; | |
75 | ||
76 | trans = btrfs_start_transaction(root, 1); | |
77 | BUG_ON(!trans); | |
78 | ||
79 | ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, | |
80 | 0, &objectid); | |
81 | if (ret) | |
82 | goto fail; | |
83 | ||
31840ae1 ZY |
84 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
85 | objectid, trans->transid, 0, 0, 0); | |
8e8a1e31 JB |
86 | if (IS_ERR(leaf)) { |
87 | ret = PTR_ERR(leaf); | |
88 | goto fail; | |
89 | } | |
f46b5a66 CH |
90 | |
91 | btrfs_set_header_nritems(leaf, 0); | |
92 | btrfs_set_header_level(leaf, 0); | |
93 | btrfs_set_header_bytenr(leaf, leaf->start); | |
94 | btrfs_set_header_generation(leaf, trans->transid); | |
95 | btrfs_set_header_owner(leaf, objectid); | |
96 | ||
97 | write_extent_buffer(leaf, root->fs_info->fsid, | |
98 | (unsigned long)btrfs_header_fsid(leaf), | |
99 | BTRFS_FSID_SIZE); | |
100 | btrfs_mark_buffer_dirty(leaf); | |
101 | ||
102 | inode_item = &root_item.inode; | |
103 | memset(inode_item, 0, sizeof(*inode_item)); | |
104 | inode_item->generation = cpu_to_le64(1); | |
105 | inode_item->size = cpu_to_le64(3); | |
106 | inode_item->nlink = cpu_to_le32(1); | |
a76a3cd4 | 107 | inode_item->nbytes = cpu_to_le64(root->leafsize); |
f46b5a66 CH |
108 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); |
109 | ||
110 | btrfs_set_root_bytenr(&root_item, leaf->start); | |
84234f3a | 111 | btrfs_set_root_generation(&root_item, trans->transid); |
f46b5a66 CH |
112 | btrfs_set_root_level(&root_item, 0); |
113 | btrfs_set_root_refs(&root_item, 1); | |
114 | btrfs_set_root_used(&root_item, 0); | |
80ff3856 | 115 | btrfs_set_root_last_snapshot(&root_item, 0); |
f46b5a66 CH |
116 | |
117 | memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); | |
118 | root_item.drop_level = 0; | |
119 | ||
925baedd | 120 | btrfs_tree_unlock(leaf); |
f46b5a66 CH |
121 | free_extent_buffer(leaf); |
122 | leaf = NULL; | |
123 | ||
124 | btrfs_set_root_dirid(&root_item, new_dirid); | |
125 | ||
126 | key.objectid = objectid; | |
127 | key.offset = 1; | |
128 | btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); | |
129 | ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, | |
130 | &root_item); | |
131 | if (ret) | |
132 | goto fail; | |
133 | ||
134 | /* | |
135 | * insert the directory item | |
136 | */ | |
137 | key.offset = (u64)-1; | |
138 | dir = root->fs_info->sb->s_root->d_inode; | |
139 | ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, | |
140 | name, namelen, dir->i_ino, &key, | |
aec7477b | 141 | BTRFS_FT_DIR, 0); |
f46b5a66 CH |
142 | if (ret) |
143 | goto fail; | |
144 | ||
145 | ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, | |
146 | name, namelen, objectid, | |
aec7477b | 147 | root->fs_info->sb->s_root->d_inode->i_ino, 0); |
f46b5a66 CH |
148 | if (ret) |
149 | goto fail; | |
150 | ||
151 | ret = btrfs_commit_transaction(trans, root); | |
152 | if (ret) | |
153 | goto fail_commit; | |
154 | ||
155 | new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen); | |
156 | BUG_ON(!new_root); | |
157 | ||
158 | trans = btrfs_start_transaction(new_root, 1); | |
159 | BUG_ON(!trans); | |
160 | ||
cb8e7090 | 161 | ret = btrfs_create_subvol_root(new_root, dentry, trans, new_dirid, |
f46b5a66 CH |
162 | BTRFS_I(dir)->block_group); |
163 | if (ret) | |
164 | goto fail; | |
165 | ||
f46b5a66 CH |
166 | fail: |
167 | nr = trans->blocks_used; | |
168 | err = btrfs_commit_transaction(trans, new_root); | |
169 | if (err && !ret) | |
170 | ret = err; | |
171 | fail_commit: | |
f46b5a66 | 172 | btrfs_btree_balance_dirty(root, nr); |
f46b5a66 CH |
173 | return ret; |
174 | } | |
175 | ||
176 | static int create_snapshot(struct btrfs_root *root, char *name, int namelen) | |
177 | { | |
178 | struct btrfs_pending_snapshot *pending_snapshot; | |
179 | struct btrfs_trans_handle *trans; | |
180 | int ret; | |
181 | int err; | |
182 | unsigned long nr = 0; | |
183 | ||
184 | if (!root->ref_cows) | |
185 | return -EINVAL; | |
186 | ||
f46b5a66 CH |
187 | ret = btrfs_check_free_space(root, 1, 0); |
188 | if (ret) | |
189 | goto fail_unlock; | |
190 | ||
191 | pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS); | |
192 | if (!pending_snapshot) { | |
193 | ret = -ENOMEM; | |
194 | goto fail_unlock; | |
195 | } | |
196 | pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS); | |
197 | if (!pending_snapshot->name) { | |
198 | ret = -ENOMEM; | |
199 | kfree(pending_snapshot); | |
200 | goto fail_unlock; | |
201 | } | |
202 | memcpy(pending_snapshot->name, name, namelen); | |
203 | pending_snapshot->name[namelen] = '\0'; | |
204 | trans = btrfs_start_transaction(root, 1); | |
205 | BUG_ON(!trans); | |
206 | pending_snapshot->root = root; | |
207 | list_add(&pending_snapshot->list, | |
208 | &trans->transaction->pending_snapshots); | |
209 | ret = btrfs_update_inode(trans, root, root->inode); | |
210 | err = btrfs_commit_transaction(trans, root); | |
211 | ||
212 | fail_unlock: | |
f46b5a66 | 213 | btrfs_btree_balance_dirty(root, nr); |
f46b5a66 CH |
214 | return ret; |
215 | } | |
216 | ||
cb8e7090 CH |
217 | /* copy of may_create in fs/namei.c() */ |
218 | static inline int btrfs_may_create(struct inode *dir, struct dentry *child) | |
219 | { | |
220 | if (child->d_inode) | |
221 | return -EEXIST; | |
222 | if (IS_DEADDIR(dir)) | |
223 | return -ENOENT; | |
224 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
225 | } | |
226 | ||
227 | /* | |
228 | * Create a new subvolume below @parent. This is largely modeled after | |
229 | * sys_mkdirat and vfs_mkdir, but we only do a single component lookup | |
230 | * inside this filesystem so it's quite a bit simpler. | |
231 | */ | |
232 | static noinline int btrfs_mksubvol(struct path *parent, char *name, | |
233 | int mode, int namelen) | |
234 | { | |
235 | struct dentry *dentry; | |
236 | int error; | |
237 | ||
238 | mutex_lock_nested(&parent->dentry->d_inode->i_mutex, I_MUTEX_PARENT); | |
239 | ||
240 | dentry = lookup_one_len(name, parent->dentry, namelen); | |
241 | error = PTR_ERR(dentry); | |
242 | if (IS_ERR(dentry)) | |
243 | goto out_unlock; | |
244 | ||
245 | error = -EEXIST; | |
246 | if (dentry->d_inode) | |
247 | goto out_dput; | |
248 | ||
249 | if (!IS_POSIXACL(parent->dentry->d_inode)) | |
250 | mode &= ~current->fs->umask; | |
251 | error = mnt_want_write(parent->mnt); | |
252 | if (error) | |
253 | goto out_dput; | |
254 | ||
255 | error = btrfs_may_create(parent->dentry->d_inode, dentry); | |
256 | if (error) | |
257 | goto out_drop_write; | |
258 | ||
cb8e7090 CH |
259 | /* |
260 | * Actually perform the low-level subvolume creation after all | |
261 | * this VFS fuzz. | |
262 | * | |
263 | * Eventually we want to pass in an inode under which we create this | |
264 | * subvolume, but for now all are under the filesystem root. | |
265 | * | |
266 | * Also we should pass on the mode eventually to allow creating new | |
267 | * subvolume with specific mode bits. | |
268 | */ | |
269 | error = create_subvol(BTRFS_I(parent->dentry->d_inode)->root, dentry, | |
270 | name, namelen); | |
271 | if (error) | |
272 | goto out_drop_write; | |
273 | ||
274 | fsnotify_mkdir(parent->dentry->d_inode, dentry); | |
275 | out_drop_write: | |
276 | mnt_drop_write(parent->mnt); | |
277 | out_dput: | |
278 | dput(dentry); | |
279 | out_unlock: | |
280 | mutex_unlock(&parent->dentry->d_inode->i_mutex); | |
281 | return error; | |
282 | } | |
283 | ||
284 | ||
f46b5a66 CH |
285 | int btrfs_defrag_file(struct file *file) |
286 | { | |
287 | struct inode *inode = fdentry(file)->d_inode; | |
288 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
289 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
3eaa2885 | 290 | struct btrfs_ordered_extent *ordered; |
f46b5a66 CH |
291 | struct page *page; |
292 | unsigned long last_index; | |
293 | unsigned long ra_pages = root->fs_info->bdi.ra_pages; | |
294 | unsigned long total_read = 0; | |
295 | u64 page_start; | |
296 | u64 page_end; | |
297 | unsigned long i; | |
298 | int ret; | |
299 | ||
f46b5a66 | 300 | ret = btrfs_check_free_space(root, inode->i_size, 0); |
f46b5a66 CH |
301 | if (ret) |
302 | return -ENOSPC; | |
303 | ||
304 | mutex_lock(&inode->i_mutex); | |
305 | last_index = inode->i_size >> PAGE_CACHE_SHIFT; | |
306 | for (i = 0; i <= last_index; i++) { | |
307 | if (total_read % ra_pages == 0) { | |
308 | btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i, | |
309 | min(last_index, i + ra_pages - 1)); | |
310 | } | |
311 | total_read++; | |
3eaa2885 | 312 | again: |
f46b5a66 CH |
313 | page = grab_cache_page(inode->i_mapping, i); |
314 | if (!page) | |
315 | goto out_unlock; | |
316 | if (!PageUptodate(page)) { | |
317 | btrfs_readpage(NULL, page); | |
318 | lock_page(page); | |
319 | if (!PageUptodate(page)) { | |
320 | unlock_page(page); | |
321 | page_cache_release(page); | |
322 | goto out_unlock; | |
323 | } | |
324 | } | |
325 | ||
f46b5a66 | 326 | wait_on_page_writeback(page); |
f46b5a66 CH |
327 | |
328 | page_start = (u64)page->index << PAGE_CACHE_SHIFT; | |
329 | page_end = page_start + PAGE_CACHE_SIZE - 1; | |
f46b5a66 | 330 | lock_extent(io_tree, page_start, page_end, GFP_NOFS); |
3eaa2885 CM |
331 | |
332 | ordered = btrfs_lookup_ordered_extent(inode, page_start); | |
333 | if (ordered) { | |
334 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); | |
335 | unlock_page(page); | |
336 | page_cache_release(page); | |
337 | btrfs_start_ordered_extent(inode, ordered, 1); | |
338 | btrfs_put_ordered_extent(ordered); | |
339 | goto again; | |
340 | } | |
341 | set_page_extent_mapped(page); | |
342 | ||
f87f057b CM |
343 | /* |
344 | * this makes sure page_mkwrite is called on the | |
345 | * page if it is dirtied again later | |
346 | */ | |
347 | clear_page_dirty_for_io(page); | |
348 | ||
ea8c2819 | 349 | btrfs_set_extent_delalloc(inode, page_start, page_end); |
f46b5a66 CH |
350 | |
351 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); | |
352 | set_page_dirty(page); | |
353 | unlock_page(page); | |
354 | page_cache_release(page); | |
355 | balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1); | |
356 | } | |
357 | ||
358 | out_unlock: | |
359 | mutex_unlock(&inode->i_mutex); | |
360 | return 0; | |
361 | } | |
362 | ||
363 | /* | |
364 | * Called inside transaction, so use GFP_NOFS | |
365 | */ | |
366 | ||
367 | static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) | |
368 | { | |
369 | u64 new_size; | |
370 | u64 old_size; | |
371 | u64 devid = 1; | |
372 | struct btrfs_ioctl_vol_args *vol_args; | |
373 | struct btrfs_trans_handle *trans; | |
374 | struct btrfs_device *device = NULL; | |
375 | char *sizestr; | |
376 | char *devstr = NULL; | |
377 | int ret = 0; | |
378 | int namelen; | |
379 | int mod = 0; | |
380 | ||
381 | vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); | |
382 | ||
383 | if (!vol_args) | |
384 | return -ENOMEM; | |
385 | ||
386 | if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { | |
387 | ret = -EFAULT; | |
388 | goto out; | |
389 | } | |
5516e595 MF |
390 | |
391 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
f46b5a66 | 392 | namelen = strlen(vol_args->name); |
f46b5a66 | 393 | |
7d9eb12c | 394 | mutex_lock(&root->fs_info->volume_mutex); |
f46b5a66 CH |
395 | sizestr = vol_args->name; |
396 | devstr = strchr(sizestr, ':'); | |
397 | if (devstr) { | |
398 | char *end; | |
399 | sizestr = devstr + 1; | |
400 | *devstr = '\0'; | |
401 | devstr = vol_args->name; | |
402 | devid = simple_strtoull(devstr, &end, 10); | |
403 | printk(KERN_INFO "resizing devid %llu\n", devid); | |
404 | } | |
405 | device = btrfs_find_device(root, devid, NULL); | |
406 | if (!device) { | |
407 | printk(KERN_INFO "resizer unable to find device %llu\n", devid); | |
408 | ret = -EINVAL; | |
409 | goto out_unlock; | |
410 | } | |
411 | if (!strcmp(sizestr, "max")) | |
412 | new_size = device->bdev->bd_inode->i_size; | |
413 | else { | |
414 | if (sizestr[0] == '-') { | |
415 | mod = -1; | |
416 | sizestr++; | |
417 | } else if (sizestr[0] == '+') { | |
418 | mod = 1; | |
419 | sizestr++; | |
420 | } | |
421 | new_size = btrfs_parse_size(sizestr); | |
422 | if (new_size == 0) { | |
423 | ret = -EINVAL; | |
424 | goto out_unlock; | |
425 | } | |
426 | } | |
427 | ||
428 | old_size = device->total_bytes; | |
429 | ||
430 | if (mod < 0) { | |
431 | if (new_size > old_size) { | |
432 | ret = -EINVAL; | |
433 | goto out_unlock; | |
434 | } | |
435 | new_size = old_size - new_size; | |
436 | } else if (mod > 0) { | |
437 | new_size = old_size + new_size; | |
438 | } | |
439 | ||
440 | if (new_size < 256 * 1024 * 1024) { | |
441 | ret = -EINVAL; | |
442 | goto out_unlock; | |
443 | } | |
444 | if (new_size > device->bdev->bd_inode->i_size) { | |
445 | ret = -EFBIG; | |
446 | goto out_unlock; | |
447 | } | |
448 | ||
449 | do_div(new_size, root->sectorsize); | |
450 | new_size *= root->sectorsize; | |
451 | ||
452 | printk(KERN_INFO "new size for %s is %llu\n", | |
453 | device->name, (unsigned long long)new_size); | |
454 | ||
455 | if (new_size > old_size) { | |
456 | trans = btrfs_start_transaction(root, 1); | |
457 | ret = btrfs_grow_device(trans, device, new_size); | |
458 | btrfs_commit_transaction(trans, root); | |
459 | } else { | |
460 | ret = btrfs_shrink_device(device, new_size); | |
461 | } | |
462 | ||
463 | out_unlock: | |
7d9eb12c | 464 | mutex_unlock(&root->fs_info->volume_mutex); |
f46b5a66 CH |
465 | out: |
466 | kfree(vol_args); | |
467 | return ret; | |
468 | } | |
469 | ||
cb8e7090 | 470 | static noinline int btrfs_ioctl_snap_create(struct file *file, |
f46b5a66 CH |
471 | void __user *arg) |
472 | { | |
cb8e7090 | 473 | struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; |
f46b5a66 CH |
474 | struct btrfs_ioctl_vol_args *vol_args; |
475 | struct btrfs_dir_item *di; | |
476 | struct btrfs_path *path; | |
477 | u64 root_dirid; | |
478 | int namelen; | |
479 | int ret; | |
480 | ||
481 | vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); | |
482 | ||
483 | if (!vol_args) | |
484 | return -ENOMEM; | |
485 | ||
486 | if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { | |
487 | ret = -EFAULT; | |
488 | goto out; | |
489 | } | |
490 | ||
5516e595 | 491 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 | 492 | namelen = strlen(vol_args->name); |
f46b5a66 CH |
493 | if (strchr(vol_args->name, '/')) { |
494 | ret = -EINVAL; | |
495 | goto out; | |
496 | } | |
497 | ||
498 | path = btrfs_alloc_path(); | |
499 | if (!path) { | |
500 | ret = -ENOMEM; | |
501 | goto out; | |
502 | } | |
503 | ||
504 | root_dirid = root->fs_info->sb->s_root->d_inode->i_ino, | |
f46b5a66 CH |
505 | di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, |
506 | path, root_dirid, | |
507 | vol_args->name, namelen, 0); | |
f46b5a66 CH |
508 | btrfs_free_path(path); |
509 | ||
510 | if (di && !IS_ERR(di)) { | |
511 | ret = -EEXIST; | |
512 | goto out; | |
513 | } | |
514 | ||
515 | if (IS_ERR(di)) { | |
516 | ret = PTR_ERR(di); | |
517 | goto out; | |
518 | } | |
519 | ||
cb8e7090 CH |
520 | if (root == root->fs_info->tree_root) { |
521 | ret = btrfs_mksubvol(&file->f_path, vol_args->name, | |
522 | file->f_path.dentry->d_inode->i_mode, | |
523 | namelen); | |
524 | } else { | |
f46b5a66 | 525 | ret = create_snapshot(root, vol_args->name, namelen); |
cb8e7090 CH |
526 | } |
527 | ||
f46b5a66 CH |
528 | out: |
529 | kfree(vol_args); | |
530 | return ret; | |
531 | } | |
532 | ||
533 | static int btrfs_ioctl_defrag(struct file *file) | |
534 | { | |
535 | struct inode *inode = fdentry(file)->d_inode; | |
536 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
537 | ||
538 | switch (inode->i_mode & S_IFMT) { | |
539 | case S_IFDIR: | |
f46b5a66 CH |
540 | btrfs_defrag_root(root, 0); |
541 | btrfs_defrag_root(root->fs_info->extent_root, 0); | |
f46b5a66 CH |
542 | break; |
543 | case S_IFREG: | |
544 | btrfs_defrag_file(file); | |
545 | break; | |
546 | } | |
547 | ||
548 | return 0; | |
549 | } | |
550 | ||
551 | long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) | |
552 | { | |
553 | struct btrfs_ioctl_vol_args *vol_args; | |
554 | int ret; | |
555 | ||
556 | vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); | |
557 | ||
558 | if (!vol_args) | |
559 | return -ENOMEM; | |
560 | ||
561 | if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { | |
562 | ret = -EFAULT; | |
563 | goto out; | |
564 | } | |
5516e595 | 565 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 CH |
566 | ret = btrfs_init_new_device(root, vol_args->name); |
567 | ||
568 | out: | |
569 | kfree(vol_args); | |
570 | return ret; | |
571 | } | |
572 | ||
573 | long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) | |
574 | { | |
575 | struct btrfs_ioctl_vol_args *vol_args; | |
576 | int ret; | |
577 | ||
578 | vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); | |
579 | ||
580 | if (!vol_args) | |
581 | return -ENOMEM; | |
582 | ||
583 | if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { | |
584 | ret = -EFAULT; | |
585 | goto out; | |
586 | } | |
5516e595 | 587 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 CH |
588 | ret = btrfs_rm_device(root, vol_args->name); |
589 | ||
590 | out: | |
591 | kfree(vol_args); | |
592 | return ret; | |
593 | } | |
594 | ||
c5c9cd4d SW |
595 | long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, u64 off, |
596 | u64 olen, u64 destoff) | |
f46b5a66 CH |
597 | { |
598 | struct inode *inode = fdentry(file)->d_inode; | |
599 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
600 | struct file *src_file; | |
601 | struct inode *src; | |
602 | struct btrfs_trans_handle *trans; | |
f46b5a66 | 603 | struct btrfs_path *path; |
f46b5a66 | 604 | struct extent_buffer *leaf; |
ae01a0ab YZ |
605 | char *buf; |
606 | struct btrfs_key key; | |
f46b5a66 CH |
607 | u32 nritems; |
608 | int slot; | |
ae01a0ab | 609 | int ret; |
c5c9cd4d SW |
610 | u64 len = olen; |
611 | u64 bs = root->fs_info->sb->s_blocksize; | |
612 | u64 hint_byte; | |
f46b5a66 | 613 | |
c5c9cd4d SW |
614 | /* |
615 | * TODO: | |
616 | * - split compressed inline extents. annoying: we need to | |
617 | * decompress into destination's address_space (the file offset | |
618 | * may change, so source mapping won't do), then recompress (or | |
619 | * otherwise reinsert) a subrange. | |
620 | * - allow ranges within the same file to be cloned (provided | |
621 | * they don't overlap)? | |
622 | */ | |
623 | ||
624 | src_file = fget(srcfd); | |
f46b5a66 CH |
625 | if (!src_file) |
626 | return -EBADF; | |
627 | src = src_file->f_dentry->d_inode; | |
628 | ||
c5c9cd4d SW |
629 | ret = -EINVAL; |
630 | if (src == inode) | |
631 | goto out_fput; | |
632 | ||
ae01a0ab YZ |
633 | ret = -EISDIR; |
634 | if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) | |
635 | goto out_fput; | |
636 | ||
f46b5a66 | 637 | ret = -EXDEV; |
ae01a0ab YZ |
638 | if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root) |
639 | goto out_fput; | |
640 | ||
641 | ret = -ENOMEM; | |
642 | buf = vmalloc(btrfs_level_size(root, 0)); | |
643 | if (!buf) | |
644 | goto out_fput; | |
645 | ||
646 | path = btrfs_alloc_path(); | |
647 | if (!path) { | |
648 | vfree(buf); | |
f46b5a66 | 649 | goto out_fput; |
ae01a0ab YZ |
650 | } |
651 | path->reada = 2; | |
f46b5a66 CH |
652 | |
653 | if (inode < src) { | |
654 | mutex_lock(&inode->i_mutex); | |
655 | mutex_lock(&src->i_mutex); | |
656 | } else { | |
657 | mutex_lock(&src->i_mutex); | |
658 | mutex_lock(&inode->i_mutex); | |
659 | } | |
660 | ||
c5c9cd4d SW |
661 | /* determine range to clone */ |
662 | ret = -EINVAL; | |
663 | if (off >= src->i_size || off + len > src->i_size) | |
f46b5a66 | 664 | goto out_unlock; |
c5c9cd4d SW |
665 | if (len == 0) |
666 | olen = len = src->i_size - off; | |
667 | /* if we extend to eof, continue to block boundary */ | |
668 | if (off + len == src->i_size) | |
669 | len = ((src->i_size + bs-1) & ~(bs-1)) | |
670 | - off; | |
671 | ||
672 | /* verify the end result is block aligned */ | |
673 | if ((off & (bs-1)) || | |
674 | ((off + len) & (bs-1))) | |
675 | goto out_unlock; | |
676 | ||
677 | printk("final src extent is %llu~%llu\n", off, len); | |
678 | printk("final dst extent is %llu~%llu\n", destoff, len); | |
f46b5a66 CH |
679 | |
680 | /* do any pending delalloc/csum calc on src, one way or | |
681 | another, and lock file content */ | |
682 | while (1) { | |
31840ae1 | 683 | struct btrfs_ordered_extent *ordered; |
c5c9cd4d SW |
684 | lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); |
685 | ordered = btrfs_lookup_first_ordered_extent(inode, off+len); | |
ae01a0ab | 686 | if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered) |
f46b5a66 | 687 | break; |
c5c9cd4d | 688 | unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); |
ae01a0ab YZ |
689 | if (ordered) |
690 | btrfs_put_ordered_extent(ordered); | |
c5c9cd4d | 691 | btrfs_wait_ordered_range(src, off, off+len); |
f46b5a66 CH |
692 | } |
693 | ||
ae01a0ab YZ |
694 | trans = btrfs_start_transaction(root, 1); |
695 | BUG_ON(!trans); | |
696 | ||
c5c9cd4d SW |
697 | /* punch hole in destination first */ |
698 | btrfs_drop_extents(trans, root, inode, off, off+len, 0, &hint_byte); | |
699 | ||
700 | /* clone data */ | |
f46b5a66 | 701 | key.objectid = src->i_ino; |
ae01a0ab YZ |
702 | key.type = BTRFS_EXTENT_DATA_KEY; |
703 | key.offset = 0; | |
f46b5a66 CH |
704 | |
705 | while (1) { | |
706 | /* | |
707 | * note the key will change type as we walk through the | |
708 | * tree. | |
709 | */ | |
710 | ret = btrfs_search_slot(trans, root, &key, path, 0, 0); | |
711 | if (ret < 0) | |
712 | goto out; | |
713 | ||
ae01a0ab YZ |
714 | nritems = btrfs_header_nritems(path->nodes[0]); |
715 | if (path->slots[0] >= nritems) { | |
f46b5a66 CH |
716 | ret = btrfs_next_leaf(root, path); |
717 | if (ret < 0) | |
718 | goto out; | |
719 | if (ret > 0) | |
720 | break; | |
ae01a0ab | 721 | nritems = btrfs_header_nritems(path->nodes[0]); |
f46b5a66 CH |
722 | } |
723 | leaf = path->nodes[0]; | |
724 | slot = path->slots[0]; | |
f46b5a66 | 725 | |
ae01a0ab | 726 | btrfs_item_key_to_cpu(leaf, &key, slot); |
f46b5a66 CH |
727 | if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY || |
728 | key.objectid != src->i_ino) | |
729 | break; | |
730 | ||
c5c9cd4d SW |
731 | if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { |
732 | struct btrfs_file_extent_item *extent; | |
733 | int type; | |
31840ae1 ZY |
734 | u32 size; |
735 | struct btrfs_key new_key; | |
c5c9cd4d SW |
736 | u64 disko = 0, diskl = 0; |
737 | u64 datao = 0, datal = 0; | |
738 | u8 comp; | |
31840ae1 ZY |
739 | |
740 | size = btrfs_item_size_nr(leaf, slot); | |
741 | read_extent_buffer(leaf, buf, | |
742 | btrfs_item_ptr_offset(leaf, slot), | |
743 | size); | |
c5c9cd4d SW |
744 | |
745 | extent = btrfs_item_ptr(leaf, slot, | |
746 | struct btrfs_file_extent_item); | |
747 | comp = btrfs_file_extent_compression(leaf, extent); | |
748 | type = btrfs_file_extent_type(leaf, extent); | |
749 | if (type == BTRFS_FILE_EXTENT_REG) { | |
750 | disko = btrfs_file_extent_disk_bytenr(leaf, extent); | |
751 | diskl = btrfs_file_extent_disk_num_bytes(leaf, extent); | |
752 | datao = btrfs_file_extent_offset(leaf, extent); | |
753 | datal = btrfs_file_extent_num_bytes(leaf, extent); | |
754 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { | |
755 | /* take upper bound, may be compressed */ | |
756 | datal = btrfs_file_extent_ram_bytes(leaf, | |
757 | extent); | |
758 | } | |
31840ae1 ZY |
759 | btrfs_release_path(root, path); |
760 | ||
c5c9cd4d SW |
761 | if (key.offset + datal < off || |
762 | key.offset >= off+len) | |
763 | goto next; | |
764 | ||
31840ae1 ZY |
765 | memcpy(&new_key, &key, sizeof(new_key)); |
766 | new_key.objectid = inode->i_ino; | |
c5c9cd4d | 767 | new_key.offset = key.offset + destoff - off; |
31840ae1 | 768 | |
c5c9cd4d SW |
769 | if (type == BTRFS_FILE_EXTENT_REG) { |
770 | ret = btrfs_insert_empty_item(trans, root, path, | |
771 | &new_key, size); | |
772 | if (ret) | |
773 | goto out; | |
774 | ||
775 | leaf = path->nodes[0]; | |
776 | slot = path->slots[0]; | |
777 | write_extent_buffer(leaf, buf, | |
31840ae1 ZY |
778 | btrfs_item_ptr_offset(leaf, slot), |
779 | size); | |
ae01a0ab | 780 | |
c5c9cd4d | 781 | extent = btrfs_item_ptr(leaf, slot, |
f46b5a66 | 782 | struct btrfs_file_extent_item); |
c5c9cd4d SW |
783 | printk(" orig disk %llu~%llu data %llu~%llu\n", |
784 | disko, diskl, datao, datal); | |
785 | ||
786 | if (off > key.offset) { | |
787 | datao += off - key.offset; | |
788 | datal -= off - key.offset; | |
789 | } | |
790 | if (key.offset + datao + datal + key.offset > | |
791 | off + len) | |
792 | datal = off + len - key.offset - datao; | |
793 | /* disko == 0 means it's a hole */ | |
794 | if (!disko) | |
795 | datao = 0; | |
796 | printk(" final disk %llu~%llu data %llu~%llu\n", | |
797 | disko, diskl, datao, datal); | |
798 | ||
799 | btrfs_set_file_extent_offset(leaf, extent, | |
800 | datao); | |
801 | btrfs_set_file_extent_num_bytes(leaf, extent, | |
802 | datal); | |
803 | if (disko) { | |
804 | inode_add_bytes(inode, datal); | |
ae01a0ab | 805 | ret = btrfs_inc_extent_ref(trans, root, |
c5c9cd4d SW |
806 | disko, diskl, leaf->start, |
807 | root->root_key.objectid, | |
808 | trans->transid, | |
809 | inode->i_ino); | |
31840ae1 | 810 | BUG_ON(ret); |
f46b5a66 | 811 | } |
c5c9cd4d SW |
812 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
813 | u64 skip = 0; | |
814 | u64 trim = 0; | |
815 | if (off > key.offset) { | |
816 | skip = off - key.offset; | |
817 | new_key.offset += skip; | |
818 | } | |
819 | if (key.offset + datal > off+len) | |
820 | trim = key.offset + datal - (off+len); | |
821 | printk("len %lld skip %lld trim %lld\n", | |
822 | datal, skip, trim); | |
823 | if (comp && (skip || trim)) { | |
824 | printk("btrfs clone_range can't split compressed inline extents yet\n"); | |
825 | ret = -EINVAL; | |
826 | goto out; | |
827 | } | |
828 | size -= skip + trim; | |
829 | datal -= skip + trim; | |
830 | ret = btrfs_insert_empty_item(trans, root, path, | |
831 | &new_key, size); | |
832 | if (ret) | |
833 | goto out; | |
834 | ||
835 | if (skip) { | |
836 | u32 start = btrfs_file_extent_calc_inline_size(0); | |
837 | memmove(buf+start, buf+start+skip, | |
838 | datal); | |
839 | } | |
840 | ||
841 | leaf = path->nodes[0]; | |
842 | slot = path->slots[0]; | |
843 | write_extent_buffer(leaf, buf, | |
844 | btrfs_item_ptr_offset(leaf, slot), | |
845 | size); | |
846 | inode_add_bytes(inode, datal); | |
f46b5a66 | 847 | } |
c5c9cd4d SW |
848 | |
849 | btrfs_mark_buffer_dirty(leaf); | |
ae01a0ab | 850 | } |
c5c9cd4d SW |
851 | |
852 | if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) { | |
853 | u32 size; | |
854 | struct btrfs_key new_key; | |
855 | u64 coverslen; | |
856 | int coff, clen; | |
857 | ||
858 | size = btrfs_item_size_nr(leaf, slot); | |
859 | coverslen = (size / BTRFS_CRC32_SIZE) << | |
860 | root->fs_info->sb->s_blocksize_bits; | |
861 | printk("csums for %llu~%llu\n", | |
862 | key.offset, coverslen); | |
863 | if (key.offset + coverslen < off || | |
864 | key.offset >= off+len) | |
865 | goto next; | |
866 | ||
867 | read_extent_buffer(leaf, buf, | |
868 | btrfs_item_ptr_offset(leaf, slot), | |
869 | size); | |
870 | btrfs_release_path(root, path); | |
871 | ||
872 | coff = 0; | |
873 | if (off > key.offset) | |
874 | coff = ((off - key.offset) >> | |
875 | root->fs_info->sb->s_blocksize_bits) * | |
876 | BTRFS_CRC32_SIZE; | |
877 | clen = size - coff; | |
878 | if (key.offset + coverslen > off+len) | |
879 | clen -= ((key.offset+coverslen-off-len) >> | |
880 | root->fs_info->sb->s_blocksize_bits) * | |
881 | BTRFS_CRC32_SIZE; | |
882 | printk(" will dup %d~%d of %d\n", | |
883 | coff, clen, size); | |
884 | ||
885 | memcpy(&new_key, &key, sizeof(new_key)); | |
886 | new_key.objectid = inode->i_ino; | |
887 | new_key.offset = key.offset + destoff - off; | |
888 | ||
889 | ret = btrfs_insert_empty_item(trans, root, path, | |
890 | &new_key, clen); | |
891 | if (ret) | |
892 | goto out; | |
893 | ||
894 | leaf = path->nodes[0]; | |
895 | slot = path->slots[0]; | |
896 | write_extent_buffer(leaf, buf + coff, | |
897 | btrfs_item_ptr_offset(leaf, slot), | |
898 | clen); | |
899 | btrfs_mark_buffer_dirty(leaf); | |
900 | } | |
901 | ||
902 | next: | |
31840ae1 | 903 | btrfs_release_path(root, path); |
f46b5a66 | 904 | key.offset++; |
f46b5a66 | 905 | } |
f46b5a66 CH |
906 | ret = 0; |
907 | out: | |
ae01a0ab YZ |
908 | btrfs_release_path(root, path); |
909 | if (ret == 0) { | |
910 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
c5c9cd4d SW |
911 | if (destoff + olen > inode->i_size) |
912 | btrfs_i_size_write(inode, destoff + olen); | |
ae01a0ab YZ |
913 | BTRFS_I(inode)->flags = BTRFS_I(src)->flags; |
914 | ret = btrfs_update_inode(trans, root, inode); | |
915 | } | |
f46b5a66 | 916 | btrfs_end_transaction(trans, root); |
c5c9cd4d | 917 | unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); |
ae01a0ab YZ |
918 | if (ret) |
919 | vmtruncate(inode, 0); | |
f46b5a66 CH |
920 | out_unlock: |
921 | mutex_unlock(&src->i_mutex); | |
922 | mutex_unlock(&inode->i_mutex); | |
ae01a0ab YZ |
923 | vfree(buf); |
924 | btrfs_free_path(path); | |
f46b5a66 CH |
925 | out_fput: |
926 | fput(src_file); | |
927 | return ret; | |
928 | } | |
929 | ||
c5c9cd4d SW |
930 | long btrfs_ioctl_clone_range(struct file *file, unsigned long argptr) |
931 | { | |
932 | struct btrfs_ioctl_clone_range_args args; | |
933 | ||
934 | if (copy_from_user(&args, (void *)argptr, sizeof(args))) | |
935 | return -EFAULT; | |
936 | return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, | |
937 | args.src_length, args.dest_offset); | |
938 | } | |
939 | ||
f46b5a66 CH |
940 | /* |
941 | * there are many ways the trans_start and trans_end ioctls can lead | |
942 | * to deadlocks. They should only be used by applications that | |
943 | * basically own the machine, and have a very in depth understanding | |
944 | * of all the possible deadlocks and enospc problems. | |
945 | */ | |
946 | long btrfs_ioctl_trans_start(struct file *file) | |
947 | { | |
948 | struct inode *inode = fdentry(file)->d_inode; | |
949 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
950 | struct btrfs_trans_handle *trans; | |
951 | int ret = 0; | |
952 | ||
df5b5520 CH |
953 | if (!capable(CAP_SYS_ADMIN)) |
954 | return -EPERM; | |
955 | ||
f46b5a66 CH |
956 | if (file->private_data) { |
957 | ret = -EINPROGRESS; | |
958 | goto out; | |
959 | } | |
9ca9ee09 SW |
960 | |
961 | mutex_lock(&root->fs_info->trans_mutex); | |
962 | root->fs_info->open_ioctl_trans++; | |
963 | mutex_unlock(&root->fs_info->trans_mutex); | |
964 | ||
965 | trans = btrfs_start_ioctl_transaction(root, 0); | |
f46b5a66 CH |
966 | if (trans) |
967 | file->private_data = trans; | |
968 | else | |
969 | ret = -ENOMEM; | |
970 | /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/ | |
971 | out: | |
f46b5a66 CH |
972 | return ret; |
973 | } | |
974 | ||
975 | /* | |
976 | * there are many ways the trans_start and trans_end ioctls can lead | |
977 | * to deadlocks. They should only be used by applications that | |
978 | * basically own the machine, and have a very in depth understanding | |
979 | * of all the possible deadlocks and enospc problems. | |
980 | */ | |
981 | long btrfs_ioctl_trans_end(struct file *file) | |
982 | { | |
983 | struct inode *inode = fdentry(file)->d_inode; | |
984 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
985 | struct btrfs_trans_handle *trans; | |
986 | int ret = 0; | |
987 | ||
f46b5a66 CH |
988 | trans = file->private_data; |
989 | if (!trans) { | |
990 | ret = -EINVAL; | |
991 | goto out; | |
992 | } | |
993 | btrfs_end_transaction(trans, root); | |
b214107e | 994 | file->private_data = NULL; |
9ca9ee09 SW |
995 | |
996 | mutex_lock(&root->fs_info->trans_mutex); | |
997 | root->fs_info->open_ioctl_trans--; | |
998 | mutex_unlock(&root->fs_info->trans_mutex); | |
999 | ||
f46b5a66 | 1000 | out: |
f46b5a66 CH |
1001 | return ret; |
1002 | } | |
1003 | ||
1004 | long btrfs_ioctl(struct file *file, unsigned int | |
1005 | cmd, unsigned long arg) | |
1006 | { | |
1007 | struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; | |
1008 | ||
1009 | switch (cmd) { | |
1010 | case BTRFS_IOC_SNAP_CREATE: | |
cb8e7090 | 1011 | return btrfs_ioctl_snap_create(file, (void __user *)arg); |
f46b5a66 CH |
1012 | case BTRFS_IOC_DEFRAG: |
1013 | return btrfs_ioctl_defrag(file); | |
1014 | case BTRFS_IOC_RESIZE: | |
1015 | return btrfs_ioctl_resize(root, (void __user *)arg); | |
1016 | case BTRFS_IOC_ADD_DEV: | |
1017 | return btrfs_ioctl_add_dev(root, (void __user *)arg); | |
1018 | case BTRFS_IOC_RM_DEV: | |
1019 | return btrfs_ioctl_rm_dev(root, (void __user *)arg); | |
1020 | case BTRFS_IOC_BALANCE: | |
1021 | return btrfs_balance(root->fs_info->dev_root); | |
1022 | case BTRFS_IOC_CLONE: | |
c5c9cd4d SW |
1023 | return btrfs_ioctl_clone(file, arg, 0, 0, 0); |
1024 | case BTRFS_IOC_CLONE_RANGE: | |
1025 | return btrfs_ioctl_clone_range(file, arg); | |
f46b5a66 CH |
1026 | case BTRFS_IOC_TRANS_START: |
1027 | return btrfs_ioctl_trans_start(file); | |
1028 | case BTRFS_IOC_TRANS_END: | |
1029 | return btrfs_ioctl_trans_end(file); | |
1030 | case BTRFS_IOC_SYNC: | |
ea8c2819 | 1031 | btrfs_start_delalloc_inodes(root); |
f46b5a66 CH |
1032 | btrfs_sync_fs(file->f_dentry->d_sb, 1); |
1033 | return 0; | |
1034 | } | |
1035 | ||
1036 | return -ENOTTY; | |
1037 | } |