Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * fs/libfs.c | |
3 | * Library for filesystems writers. | |
4 | */ | |
5 | ||
6 | #include <linux/module.h> | |
7 | #include <linux/pagemap.h> | |
5a0e3ad6 | 8 | #include <linux/slab.h> |
1da177e4 LT |
9 | #include <linux/mount.h> |
10 | #include <linux/vfs.h> | |
7bb46a67 | 11 | #include <linux/quotaops.h> |
7cf34c76 | 12 | #include <linux/mutex.h> |
2596110a | 13 | #include <linux/exportfs.h> |
d5aacad5 AV |
14 | #include <linux/writeback.h> |
15 | #include <linux/buffer_head.h> | |
7cf34c76 | 16 | |
1da177e4 LT |
17 | #include <asm/uaccess.h> |
18 | ||
da502956 NP |
19 | static inline int simple_positive(struct dentry *dentry) |
20 | { | |
21 | return dentry->d_inode && !d_unhashed(dentry); | |
22 | } | |
23 | ||
1da177e4 LT |
24 | int simple_getattr(struct vfsmount *mnt, struct dentry *dentry, |
25 | struct kstat *stat) | |
26 | { | |
27 | struct inode *inode = dentry->d_inode; | |
28 | generic_fillattr(inode, stat); | |
29 | stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_SHIFT - 9); | |
30 | return 0; | |
31 | } | |
32 | ||
726c3342 | 33 | int simple_statfs(struct dentry *dentry, struct kstatfs *buf) |
1da177e4 | 34 | { |
726c3342 | 35 | buf->f_type = dentry->d_sb->s_magic; |
1da177e4 LT |
36 | buf->f_bsize = PAGE_CACHE_SIZE; |
37 | buf->f_namelen = NAME_MAX; | |
38 | return 0; | |
39 | } | |
40 | ||
41 | /* | |
42 | * Retaining negative dentries for an in-memory filesystem just wastes | |
43 | * memory and lookup time: arrange for them to be deleted immediately. | |
44 | */ | |
fe15ce44 | 45 | static int simple_delete_dentry(const struct dentry *dentry) |
1da177e4 LT |
46 | { |
47 | return 1; | |
48 | } | |
49 | ||
50 | /* | |
51 | * Lookup the data. This is trivial - if the dentry didn't already | |
52 | * exist, we know it is negative. Set d_op to delete negative dentries. | |
53 | */ | |
54 | struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd) | |
55 | { | |
3ba13d17 | 56 | static const struct dentry_operations simple_dentry_operations = { |
1da177e4 LT |
57 | .d_delete = simple_delete_dentry, |
58 | }; | |
59 | ||
60 | if (dentry->d_name.len > NAME_MAX) | |
61 | return ERR_PTR(-ENAMETOOLONG); | |
62 | dentry->d_op = &simple_dentry_operations; | |
63 | d_add(dentry, NULL); | |
64 | return NULL; | |
65 | } | |
66 | ||
1da177e4 LT |
67 | int dcache_dir_open(struct inode *inode, struct file *file) |
68 | { | |
69 | static struct qstr cursor_name = {.len = 1, .name = "."}; | |
70 | ||
0f7fc9e4 | 71 | file->private_data = d_alloc(file->f_path.dentry, &cursor_name); |
1da177e4 LT |
72 | |
73 | return file->private_data ? 0 : -ENOMEM; | |
74 | } | |
75 | ||
76 | int dcache_dir_close(struct inode *inode, struct file *file) | |
77 | { | |
78 | dput(file->private_data); | |
79 | return 0; | |
80 | } | |
81 | ||
82 | loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin) | |
83 | { | |
0f7fc9e4 | 84 | mutex_lock(&file->f_path.dentry->d_inode->i_mutex); |
1da177e4 LT |
85 | switch (origin) { |
86 | case 1: | |
87 | offset += file->f_pos; | |
88 | case 0: | |
89 | if (offset >= 0) | |
90 | break; | |
91 | default: | |
0f7fc9e4 | 92 | mutex_unlock(&file->f_path.dentry->d_inode->i_mutex); |
1da177e4 LT |
93 | return -EINVAL; |
94 | } | |
95 | if (offset != file->f_pos) { | |
96 | file->f_pos = offset; | |
97 | if (file->f_pos >= 2) { | |
98 | struct list_head *p; | |
99 | struct dentry *cursor = file->private_data; | |
100 | loff_t n = file->f_pos - 2; | |
101 | ||
102 | spin_lock(&dcache_lock); | |
5160ee6f | 103 | list_del(&cursor->d_u.d_child); |
0f7fc9e4 JJS |
104 | p = file->f_path.dentry->d_subdirs.next; |
105 | while (n && p != &file->f_path.dentry->d_subdirs) { | |
1da177e4 | 106 | struct dentry *next; |
5160ee6f | 107 | next = list_entry(p, struct dentry, d_u.d_child); |
da502956 NP |
108 | spin_lock(&next->d_lock); |
109 | if (simple_positive(next)) | |
1da177e4 | 110 | n--; |
da502956 | 111 | spin_unlock(&next->d_lock); |
1da177e4 LT |
112 | p = p->next; |
113 | } | |
5160ee6f | 114 | list_add_tail(&cursor->d_u.d_child, p); |
1da177e4 LT |
115 | spin_unlock(&dcache_lock); |
116 | } | |
117 | } | |
0f7fc9e4 | 118 | mutex_unlock(&file->f_path.dentry->d_inode->i_mutex); |
1da177e4 LT |
119 | return offset; |
120 | } | |
121 | ||
122 | /* Relationship between i_mode and the DT_xxx types */ | |
123 | static inline unsigned char dt_type(struct inode *inode) | |
124 | { | |
125 | return (inode->i_mode >> 12) & 15; | |
126 | } | |
127 | ||
128 | /* | |
129 | * Directory is locked and all positive dentries in it are safe, since | |
130 | * for ramfs-type trees they can't go away without unlink() or rmdir(), | |
131 | * both impossible due to the lock on directory. | |
132 | */ | |
133 | ||
134 | int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir) | |
135 | { | |
0f7fc9e4 | 136 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 137 | struct dentry *cursor = filp->private_data; |
5160ee6f | 138 | struct list_head *p, *q = &cursor->d_u.d_child; |
1da177e4 LT |
139 | ino_t ino; |
140 | int i = filp->f_pos; | |
141 | ||
142 | switch (i) { | |
143 | case 0: | |
144 | ino = dentry->d_inode->i_ino; | |
145 | if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0) | |
146 | break; | |
147 | filp->f_pos++; | |
148 | i++; | |
149 | /* fallthrough */ | |
150 | case 1: | |
151 | ino = parent_ino(dentry); | |
152 | if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0) | |
153 | break; | |
154 | filp->f_pos++; | |
155 | i++; | |
156 | /* fallthrough */ | |
157 | default: | |
158 | spin_lock(&dcache_lock); | |
1bfba4e8 AM |
159 | if (filp->f_pos == 2) |
160 | list_move(q, &dentry->d_subdirs); | |
161 | ||
1da177e4 LT |
162 | for (p=q->next; p != &dentry->d_subdirs; p=p->next) { |
163 | struct dentry *next; | |
5160ee6f | 164 | next = list_entry(p, struct dentry, d_u.d_child); |
da502956 NP |
165 | spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED); |
166 | if (!simple_positive(next)) { | |
167 | spin_unlock(&next->d_lock); | |
1da177e4 | 168 | continue; |
da502956 | 169 | } |
1da177e4 | 170 | |
da502956 | 171 | spin_unlock(&next->d_lock); |
1da177e4 | 172 | spin_unlock(&dcache_lock); |
0f8952c2 RN |
173 | if (filldir(dirent, next->d_name.name, |
174 | next->d_name.len, filp->f_pos, | |
175 | next->d_inode->i_ino, | |
176 | dt_type(next->d_inode)) < 0) | |
1da177e4 LT |
177 | return 0; |
178 | spin_lock(&dcache_lock); | |
179 | /* next is still alive */ | |
1bfba4e8 | 180 | list_move(q, p); |
1da177e4 LT |
181 | p = q; |
182 | filp->f_pos++; | |
183 | } | |
184 | spin_unlock(&dcache_lock); | |
185 | } | |
186 | return 0; | |
187 | } | |
188 | ||
189 | ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos) | |
190 | { | |
191 | return -EISDIR; | |
192 | } | |
193 | ||
4b6f5d20 | 194 | const struct file_operations simple_dir_operations = { |
1da177e4 LT |
195 | .open = dcache_dir_open, |
196 | .release = dcache_dir_close, | |
197 | .llseek = dcache_dir_lseek, | |
198 | .read = generic_read_dir, | |
199 | .readdir = dcache_readdir, | |
1b061d92 | 200 | .fsync = noop_fsync, |
1da177e4 LT |
201 | }; |
202 | ||
92e1d5be | 203 | const struct inode_operations simple_dir_inode_operations = { |
1da177e4 LT |
204 | .lookup = simple_lookup, |
205 | }; | |
206 | ||
759b9775 HD |
207 | static const struct super_operations simple_super_operations = { |
208 | .statfs = simple_statfs, | |
209 | }; | |
210 | ||
1da177e4 LT |
211 | /* |
212 | * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that | |
213 | * will never be mountable) | |
214 | */ | |
51139ada AV |
215 | struct dentry *mount_pseudo(struct file_system_type *fs_type, char *name, |
216 | const struct super_operations *ops, unsigned long magic) | |
1da177e4 LT |
217 | { |
218 | struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL); | |
1da177e4 LT |
219 | struct dentry *dentry; |
220 | struct inode *root; | |
221 | struct qstr d_name = {.name = name, .len = strlen(name)}; | |
222 | ||
223 | if (IS_ERR(s)) | |
51139ada | 224 | return ERR_CAST(s); |
1da177e4 LT |
225 | |
226 | s->s_flags = MS_NOUSER; | |
89a4eb4b | 227 | s->s_maxbytes = MAX_LFS_FILESIZE; |
3971e1a9 AN |
228 | s->s_blocksize = PAGE_SIZE; |
229 | s->s_blocksize_bits = PAGE_SHIFT; | |
1da177e4 | 230 | s->s_magic = magic; |
759b9775 | 231 | s->s_op = ops ? ops : &simple_super_operations; |
1da177e4 LT |
232 | s->s_time_gran = 1; |
233 | root = new_inode(s); | |
234 | if (!root) | |
235 | goto Enomem; | |
1a1c9bb4 JL |
236 | /* |
237 | * since this is the first inode, make it number 1. New inodes created | |
238 | * after this must take care not to collide with it (by passing | |
239 | * max_reserved of 1 to iunique). | |
240 | */ | |
241 | root->i_ino = 1; | |
1da177e4 | 242 | root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR; |
1da177e4 LT |
243 | root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME; |
244 | dentry = d_alloc(NULL, &d_name); | |
245 | if (!dentry) { | |
246 | iput(root); | |
247 | goto Enomem; | |
248 | } | |
249 | dentry->d_sb = s; | |
250 | dentry->d_parent = dentry; | |
251 | d_instantiate(dentry, root); | |
252 | s->s_root = dentry; | |
253 | s->s_flags |= MS_ACTIVE; | |
51139ada | 254 | return dget(s->s_root); |
1da177e4 LT |
255 | |
256 | Enomem: | |
6f5bbff9 | 257 | deactivate_locked_super(s); |
51139ada | 258 | return ERR_PTR(-ENOMEM); |
1da177e4 LT |
259 | } |
260 | ||
261 | int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry) | |
262 | { | |
263 | struct inode *inode = old_dentry->d_inode; | |
264 | ||
265 | inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME; | |
d8c76e6f | 266 | inc_nlink(inode); |
7de9c6ee | 267 | ihold(inode); |
1da177e4 LT |
268 | dget(dentry); |
269 | d_instantiate(dentry, inode); | |
270 | return 0; | |
271 | } | |
272 | ||
1da177e4 LT |
273 | int simple_empty(struct dentry *dentry) |
274 | { | |
275 | struct dentry *child; | |
276 | int ret = 0; | |
277 | ||
278 | spin_lock(&dcache_lock); | |
da502956 NP |
279 | list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child) { |
280 | spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED); | |
281 | if (simple_positive(child)) { | |
282 | spin_unlock(&child->d_lock); | |
1da177e4 | 283 | goto out; |
da502956 NP |
284 | } |
285 | spin_unlock(&child->d_lock); | |
286 | } | |
1da177e4 LT |
287 | ret = 1; |
288 | out: | |
289 | spin_unlock(&dcache_lock); | |
290 | return ret; | |
291 | } | |
292 | ||
293 | int simple_unlink(struct inode *dir, struct dentry *dentry) | |
294 | { | |
295 | struct inode *inode = dentry->d_inode; | |
296 | ||
297 | inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME; | |
9a53c3a7 | 298 | drop_nlink(inode); |
1da177e4 LT |
299 | dput(dentry); |
300 | return 0; | |
301 | } | |
302 | ||
303 | int simple_rmdir(struct inode *dir, struct dentry *dentry) | |
304 | { | |
305 | if (!simple_empty(dentry)) | |
306 | return -ENOTEMPTY; | |
307 | ||
9a53c3a7 | 308 | drop_nlink(dentry->d_inode); |
1da177e4 | 309 | simple_unlink(dir, dentry); |
9a53c3a7 | 310 | drop_nlink(dir); |
1da177e4 LT |
311 | return 0; |
312 | } | |
313 | ||
314 | int simple_rename(struct inode *old_dir, struct dentry *old_dentry, | |
315 | struct inode *new_dir, struct dentry *new_dentry) | |
316 | { | |
317 | struct inode *inode = old_dentry->d_inode; | |
318 | int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode); | |
319 | ||
320 | if (!simple_empty(new_dentry)) | |
321 | return -ENOTEMPTY; | |
322 | ||
323 | if (new_dentry->d_inode) { | |
324 | simple_unlink(new_dir, new_dentry); | |
325 | if (they_are_dirs) | |
9a53c3a7 | 326 | drop_nlink(old_dir); |
1da177e4 | 327 | } else if (they_are_dirs) { |
9a53c3a7 | 328 | drop_nlink(old_dir); |
d8c76e6f | 329 | inc_nlink(new_dir); |
1da177e4 LT |
330 | } |
331 | ||
332 | old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime = | |
333 | new_dir->i_mtime = inode->i_ctime = CURRENT_TIME; | |
334 | ||
335 | return 0; | |
336 | } | |
337 | ||
7bb46a67 | 338 | /** |
eef2380c | 339 | * simple_setattr - setattr for simple filesystem |
7bb46a67 | 340 | * @dentry: dentry |
341 | * @iattr: iattr structure | |
342 | * | |
343 | * Returns 0 on success, -error on failure. | |
344 | * | |
eef2380c CH |
345 | * simple_setattr is a simple ->setattr implementation without a proper |
346 | * implementation of size changes. | |
347 | * | |
348 | * It can either be used for in-memory filesystems or special files | |
349 | * on simple regular filesystems. Anything that needs to change on-disk | |
350 | * or wire state on size changes needs its own setattr method. | |
7bb46a67 | 351 | */ |
352 | int simple_setattr(struct dentry *dentry, struct iattr *iattr) | |
353 | { | |
354 | struct inode *inode = dentry->d_inode; | |
355 | int error; | |
356 | ||
eef2380c CH |
357 | WARN_ON_ONCE(inode->i_op->truncate); |
358 | ||
7bb46a67 | 359 | error = inode_change_ok(inode, iattr); |
360 | if (error) | |
361 | return error; | |
362 | ||
2c27c65e CH |
363 | if (iattr->ia_valid & ATTR_SIZE) |
364 | truncate_setsize(inode, iattr->ia_size); | |
6a1a90ad | 365 | setattr_copy(inode, iattr); |
eef2380c CH |
366 | mark_inode_dirty(inode); |
367 | return 0; | |
7bb46a67 | 368 | } |
369 | EXPORT_SYMBOL(simple_setattr); | |
370 | ||
1da177e4 LT |
371 | int simple_readpage(struct file *file, struct page *page) |
372 | { | |
c0d92cbc | 373 | clear_highpage(page); |
1da177e4 LT |
374 | flush_dcache_page(page); |
375 | SetPageUptodate(page); | |
1da177e4 LT |
376 | unlock_page(page); |
377 | return 0; | |
378 | } | |
379 | ||
afddba49 NP |
380 | int simple_write_begin(struct file *file, struct address_space *mapping, |
381 | loff_t pos, unsigned len, unsigned flags, | |
382 | struct page **pagep, void **fsdata) | |
383 | { | |
384 | struct page *page; | |
385 | pgoff_t index; | |
afddba49 NP |
386 | |
387 | index = pos >> PAGE_CACHE_SHIFT; | |
afddba49 | 388 | |
54566b2c | 389 | page = grab_cache_page_write_begin(mapping, index, flags); |
afddba49 NP |
390 | if (!page) |
391 | return -ENOMEM; | |
392 | ||
393 | *pagep = page; | |
394 | ||
193cf4b9 BH |
395 | if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) { |
396 | unsigned from = pos & (PAGE_CACHE_SIZE - 1); | |
397 | ||
398 | zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE); | |
399 | } | |
400 | return 0; | |
afddba49 NP |
401 | } |
402 | ||
ad2a722f BH |
403 | /** |
404 | * simple_write_end - .write_end helper for non-block-device FSes | |
405 | * @available: See .write_end of address_space_operations | |
406 | * @file: " | |
407 | * @mapping: " | |
408 | * @pos: " | |
409 | * @len: " | |
410 | * @copied: " | |
411 | * @page: " | |
412 | * @fsdata: " | |
413 | * | |
414 | * simple_write_end does the minimum needed for updating a page after writing is | |
415 | * done. It has the same API signature as the .write_end of | |
416 | * address_space_operations vector. So it can just be set onto .write_end for | |
417 | * FSes that don't need any other processing. i_mutex is assumed to be held. | |
418 | * Block based filesystems should use generic_write_end(). | |
419 | * NOTE: Even though i_size might get updated by this function, mark_inode_dirty | |
420 | * is not called, so a filesystem that actually does store data in .write_inode | |
421 | * should extend on what's done here with a call to mark_inode_dirty() in the | |
422 | * case that i_size has changed. | |
423 | */ | |
afddba49 NP |
424 | int simple_write_end(struct file *file, struct address_space *mapping, |
425 | loff_t pos, unsigned len, unsigned copied, | |
426 | struct page *page, void *fsdata) | |
427 | { | |
ad2a722f BH |
428 | struct inode *inode = page->mapping->host; |
429 | loff_t last_pos = pos + copied; | |
afddba49 NP |
430 | |
431 | /* zero the stale part of the page if we did a short copy */ | |
432 | if (copied < len) { | |
ad2a722f BH |
433 | unsigned from = pos & (PAGE_CACHE_SIZE - 1); |
434 | ||
435 | zero_user(page, from + copied, len - copied); | |
afddba49 NP |
436 | } |
437 | ||
ad2a722f BH |
438 | if (!PageUptodate(page)) |
439 | SetPageUptodate(page); | |
440 | /* | |
441 | * No need to use i_size_read() here, the i_size | |
442 | * cannot change under us because we hold the i_mutex. | |
443 | */ | |
444 | if (last_pos > inode->i_size) | |
445 | i_size_write(inode, last_pos); | |
afddba49 | 446 | |
ad2a722f | 447 | set_page_dirty(page); |
afddba49 NP |
448 | unlock_page(page); |
449 | page_cache_release(page); | |
450 | ||
451 | return copied; | |
452 | } | |
453 | ||
1a1c9bb4 JL |
454 | /* |
455 | * the inodes created here are not hashed. If you use iunique to generate | |
456 | * unique inode values later for this filesystem, then you must take care | |
457 | * to pass it an appropriate max_reserved value to avoid collisions. | |
458 | */ | |
7d683a09 RS |
459 | int simple_fill_super(struct super_block *s, unsigned long magic, |
460 | struct tree_descr *files) | |
1da177e4 | 461 | { |
1da177e4 LT |
462 | struct inode *inode; |
463 | struct dentry *root; | |
464 | struct dentry *dentry; | |
465 | int i; | |
466 | ||
467 | s->s_blocksize = PAGE_CACHE_SIZE; | |
468 | s->s_blocksize_bits = PAGE_CACHE_SHIFT; | |
469 | s->s_magic = magic; | |
759b9775 | 470 | s->s_op = &simple_super_operations; |
1da177e4 LT |
471 | s->s_time_gran = 1; |
472 | ||
473 | inode = new_inode(s); | |
474 | if (!inode) | |
475 | return -ENOMEM; | |
1a1c9bb4 JL |
476 | /* |
477 | * because the root inode is 1, the files array must not contain an | |
478 | * entry at index 1 | |
479 | */ | |
480 | inode->i_ino = 1; | |
1da177e4 | 481 | inode->i_mode = S_IFDIR | 0755; |
1da177e4 LT |
482 | inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME; |
483 | inode->i_op = &simple_dir_inode_operations; | |
484 | inode->i_fop = &simple_dir_operations; | |
7656f328 | 485 | inode->i_nlink = 2; |
1da177e4 LT |
486 | root = d_alloc_root(inode); |
487 | if (!root) { | |
488 | iput(inode); | |
489 | return -ENOMEM; | |
490 | } | |
491 | for (i = 0; !files->name || files->name[0]; i++, files++) { | |
492 | if (!files->name) | |
493 | continue; | |
1a1c9bb4 JL |
494 | |
495 | /* warn if it tries to conflict with the root inode */ | |
496 | if (unlikely(i == 1)) | |
497 | printk(KERN_WARNING "%s: %s passed in a files array" | |
498 | "with an index of 1!\n", __func__, | |
499 | s->s_type->name); | |
500 | ||
1da177e4 LT |
501 | dentry = d_alloc_name(root, files->name); |
502 | if (!dentry) | |
503 | goto out; | |
504 | inode = new_inode(s); | |
505 | if (!inode) | |
506 | goto out; | |
507 | inode->i_mode = S_IFREG | files->mode; | |
1da177e4 LT |
508 | inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME; |
509 | inode->i_fop = files->ops; | |
510 | inode->i_ino = i; | |
511 | d_add(dentry, inode); | |
512 | } | |
513 | s->s_root = root; | |
514 | return 0; | |
515 | out: | |
516 | d_genocide(root); | |
517 | dput(root); | |
518 | return -ENOMEM; | |
519 | } | |
520 | ||
521 | static DEFINE_SPINLOCK(pin_fs_lock); | |
522 | ||
1f5ce9e9 | 523 | int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count) |
1da177e4 LT |
524 | { |
525 | struct vfsmount *mnt = NULL; | |
526 | spin_lock(&pin_fs_lock); | |
527 | if (unlikely(!*mount)) { | |
528 | spin_unlock(&pin_fs_lock); | |
1f5ce9e9 | 529 | mnt = vfs_kern_mount(type, 0, type->name, NULL); |
1da177e4 LT |
530 | if (IS_ERR(mnt)) |
531 | return PTR_ERR(mnt); | |
532 | spin_lock(&pin_fs_lock); | |
533 | if (!*mount) | |
534 | *mount = mnt; | |
535 | } | |
536 | mntget(*mount); | |
537 | ++*count; | |
538 | spin_unlock(&pin_fs_lock); | |
539 | mntput(mnt); | |
540 | return 0; | |
541 | } | |
542 | ||
543 | void simple_release_fs(struct vfsmount **mount, int *count) | |
544 | { | |
545 | struct vfsmount *mnt; | |
546 | spin_lock(&pin_fs_lock); | |
547 | mnt = *mount; | |
548 | if (!--*count) | |
549 | *mount = NULL; | |
550 | spin_unlock(&pin_fs_lock); | |
551 | mntput(mnt); | |
552 | } | |
553 | ||
6d1029b5 AM |
554 | /** |
555 | * simple_read_from_buffer - copy data from the buffer to user space | |
556 | * @to: the user space buffer to read to | |
557 | * @count: the maximum number of bytes to read | |
558 | * @ppos: the current position in the buffer | |
559 | * @from: the buffer to read from | |
560 | * @available: the size of the buffer | |
561 | * | |
562 | * The simple_read_from_buffer() function reads up to @count bytes from the | |
563 | * buffer @from at offset @ppos into the user space address starting at @to. | |
564 | * | |
565 | * On success, the number of bytes read is returned and the offset @ppos is | |
566 | * advanced by this number, or negative value is returned on error. | |
567 | **/ | |
1da177e4 LT |
568 | ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos, |
569 | const void *from, size_t available) | |
570 | { | |
571 | loff_t pos = *ppos; | |
14be2746 SR |
572 | size_t ret; |
573 | ||
1da177e4 LT |
574 | if (pos < 0) |
575 | return -EINVAL; | |
14be2746 | 576 | if (pos >= available || !count) |
1da177e4 LT |
577 | return 0; |
578 | if (count > available - pos) | |
579 | count = available - pos; | |
14be2746 SR |
580 | ret = copy_to_user(to, from + pos, count); |
581 | if (ret == count) | |
1da177e4 | 582 | return -EFAULT; |
14be2746 | 583 | count -= ret; |
1da177e4 LT |
584 | *ppos = pos + count; |
585 | return count; | |
586 | } | |
587 | ||
6a727b43 JS |
588 | /** |
589 | * simple_write_to_buffer - copy data from user space to the buffer | |
590 | * @to: the buffer to write to | |
591 | * @available: the size of the buffer | |
592 | * @ppos: the current position in the buffer | |
593 | * @from: the user space buffer to read from | |
594 | * @count: the maximum number of bytes to read | |
595 | * | |
596 | * The simple_write_to_buffer() function reads up to @count bytes from the user | |
597 | * space address starting at @from into the buffer @to at offset @ppos. | |
598 | * | |
599 | * On success, the number of bytes written is returned and the offset @ppos is | |
600 | * advanced by this number, or negative value is returned on error. | |
601 | **/ | |
602 | ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, | |
603 | const void __user *from, size_t count) | |
604 | { | |
605 | loff_t pos = *ppos; | |
606 | size_t res; | |
607 | ||
608 | if (pos < 0) | |
609 | return -EINVAL; | |
610 | if (pos >= available || !count) | |
611 | return 0; | |
612 | if (count > available - pos) | |
613 | count = available - pos; | |
614 | res = copy_from_user(to + pos, from, count); | |
615 | if (res == count) | |
616 | return -EFAULT; | |
617 | count -= res; | |
618 | *ppos = pos + count; | |
619 | return count; | |
620 | } | |
621 | ||
6d1029b5 AM |
622 | /** |
623 | * memory_read_from_buffer - copy data from the buffer | |
624 | * @to: the kernel space buffer to read to | |
625 | * @count: the maximum number of bytes to read | |
626 | * @ppos: the current position in the buffer | |
627 | * @from: the buffer to read from | |
628 | * @available: the size of the buffer | |
629 | * | |
630 | * The memory_read_from_buffer() function reads up to @count bytes from the | |
631 | * buffer @from at offset @ppos into the kernel space address starting at @to. | |
632 | * | |
633 | * On success, the number of bytes read is returned and the offset @ppos is | |
634 | * advanced by this number, or negative value is returned on error. | |
635 | **/ | |
93b07113 AM |
636 | ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos, |
637 | const void *from, size_t available) | |
638 | { | |
639 | loff_t pos = *ppos; | |
640 | ||
641 | if (pos < 0) | |
642 | return -EINVAL; | |
643 | if (pos >= available) | |
644 | return 0; | |
645 | if (count > available - pos) | |
646 | count = available - pos; | |
647 | memcpy(to, from + pos, count); | |
648 | *ppos = pos + count; | |
649 | ||
650 | return count; | |
651 | } | |
652 | ||
1da177e4 LT |
653 | /* |
654 | * Transaction based IO. | |
655 | * The file expects a single write which triggers the transaction, and then | |
656 | * possibly a read which collects the result - which is stored in a | |
657 | * file-local buffer. | |
658 | */ | |
76791ab2 IM |
659 | |
660 | void simple_transaction_set(struct file *file, size_t n) | |
661 | { | |
662 | struct simple_transaction_argresp *ar = file->private_data; | |
663 | ||
664 | BUG_ON(n > SIMPLE_TRANSACTION_LIMIT); | |
665 | ||
666 | /* | |
667 | * The barrier ensures that ar->size will really remain zero until | |
668 | * ar->data is ready for reading. | |
669 | */ | |
670 | smp_mb(); | |
671 | ar->size = n; | |
672 | } | |
673 | ||
1da177e4 LT |
674 | char *simple_transaction_get(struct file *file, const char __user *buf, size_t size) |
675 | { | |
676 | struct simple_transaction_argresp *ar; | |
677 | static DEFINE_SPINLOCK(simple_transaction_lock); | |
678 | ||
679 | if (size > SIMPLE_TRANSACTION_LIMIT - 1) | |
680 | return ERR_PTR(-EFBIG); | |
681 | ||
682 | ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL); | |
683 | if (!ar) | |
684 | return ERR_PTR(-ENOMEM); | |
685 | ||
686 | spin_lock(&simple_transaction_lock); | |
687 | ||
688 | /* only one write allowed per open */ | |
689 | if (file->private_data) { | |
690 | spin_unlock(&simple_transaction_lock); | |
691 | free_page((unsigned long)ar); | |
692 | return ERR_PTR(-EBUSY); | |
693 | } | |
694 | ||
695 | file->private_data = ar; | |
696 | ||
697 | spin_unlock(&simple_transaction_lock); | |
698 | ||
699 | if (copy_from_user(ar->data, buf, size)) | |
700 | return ERR_PTR(-EFAULT); | |
701 | ||
702 | return ar->data; | |
703 | } | |
704 | ||
705 | ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos) | |
706 | { | |
707 | struct simple_transaction_argresp *ar = file->private_data; | |
708 | ||
709 | if (!ar) | |
710 | return 0; | |
711 | return simple_read_from_buffer(buf, size, pos, ar->data, ar->size); | |
712 | } | |
713 | ||
714 | int simple_transaction_release(struct inode *inode, struct file *file) | |
715 | { | |
716 | free_page((unsigned long)file->private_data); | |
717 | return 0; | |
718 | } | |
719 | ||
acaefc25 AB |
720 | /* Simple attribute files */ |
721 | ||
722 | struct simple_attr { | |
8b88b099 CH |
723 | int (*get)(void *, u64 *); |
724 | int (*set)(void *, u64); | |
acaefc25 AB |
725 | char get_buf[24]; /* enough to store a u64 and "\n\0" */ |
726 | char set_buf[24]; | |
727 | void *data; | |
728 | const char *fmt; /* format for read operation */ | |
7cf34c76 | 729 | struct mutex mutex; /* protects access to these buffers */ |
acaefc25 AB |
730 | }; |
731 | ||
732 | /* simple_attr_open is called by an actual attribute open file operation | |
733 | * to set the attribute specific access operations. */ | |
734 | int simple_attr_open(struct inode *inode, struct file *file, | |
8b88b099 | 735 | int (*get)(void *, u64 *), int (*set)(void *, u64), |
acaefc25 AB |
736 | const char *fmt) |
737 | { | |
738 | struct simple_attr *attr; | |
739 | ||
740 | attr = kmalloc(sizeof(*attr), GFP_KERNEL); | |
741 | if (!attr) | |
742 | return -ENOMEM; | |
743 | ||
744 | attr->get = get; | |
745 | attr->set = set; | |
8e18e294 | 746 | attr->data = inode->i_private; |
acaefc25 | 747 | attr->fmt = fmt; |
7cf34c76 | 748 | mutex_init(&attr->mutex); |
acaefc25 AB |
749 | |
750 | file->private_data = attr; | |
751 | ||
752 | return nonseekable_open(inode, file); | |
753 | } | |
754 | ||
74bedc4d | 755 | int simple_attr_release(struct inode *inode, struct file *file) |
acaefc25 AB |
756 | { |
757 | kfree(file->private_data); | |
758 | return 0; | |
759 | } | |
760 | ||
761 | /* read from the buffer that is filled with the get function */ | |
762 | ssize_t simple_attr_read(struct file *file, char __user *buf, | |
763 | size_t len, loff_t *ppos) | |
764 | { | |
765 | struct simple_attr *attr; | |
766 | size_t size; | |
767 | ssize_t ret; | |
768 | ||
769 | attr = file->private_data; | |
770 | ||
771 | if (!attr->get) | |
772 | return -EACCES; | |
773 | ||
9261303a CH |
774 | ret = mutex_lock_interruptible(&attr->mutex); |
775 | if (ret) | |
776 | return ret; | |
777 | ||
8b88b099 | 778 | if (*ppos) { /* continued read */ |
acaefc25 | 779 | size = strlen(attr->get_buf); |
8b88b099 CH |
780 | } else { /* first read */ |
781 | u64 val; | |
782 | ret = attr->get(attr->data, &val); | |
783 | if (ret) | |
784 | goto out; | |
785 | ||
acaefc25 | 786 | size = scnprintf(attr->get_buf, sizeof(attr->get_buf), |
8b88b099 CH |
787 | attr->fmt, (unsigned long long)val); |
788 | } | |
acaefc25 AB |
789 | |
790 | ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size); | |
8b88b099 | 791 | out: |
7cf34c76 | 792 | mutex_unlock(&attr->mutex); |
acaefc25 AB |
793 | return ret; |
794 | } | |
795 | ||
796 | /* interpret the buffer as a number to call the set function with */ | |
797 | ssize_t simple_attr_write(struct file *file, const char __user *buf, | |
798 | size_t len, loff_t *ppos) | |
799 | { | |
800 | struct simple_attr *attr; | |
801 | u64 val; | |
802 | size_t size; | |
803 | ssize_t ret; | |
804 | ||
805 | attr = file->private_data; | |
acaefc25 AB |
806 | if (!attr->set) |
807 | return -EACCES; | |
808 | ||
9261303a CH |
809 | ret = mutex_lock_interruptible(&attr->mutex); |
810 | if (ret) | |
811 | return ret; | |
812 | ||
acaefc25 AB |
813 | ret = -EFAULT; |
814 | size = min(sizeof(attr->set_buf) - 1, len); | |
815 | if (copy_from_user(attr->set_buf, buf, size)) | |
816 | goto out; | |
817 | ||
acaefc25 AB |
818 | attr->set_buf[size] = '\0'; |
819 | val = simple_strtol(attr->set_buf, NULL, 0); | |
05cc0cee WF |
820 | ret = attr->set(attr->data, val); |
821 | if (ret == 0) | |
822 | ret = len; /* on success, claim we got the whole input */ | |
acaefc25 | 823 | out: |
7cf34c76 | 824 | mutex_unlock(&attr->mutex); |
acaefc25 AB |
825 | return ret; |
826 | } | |
827 | ||
2596110a CH |
828 | /** |
829 | * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation | |
830 | * @sb: filesystem to do the file handle conversion on | |
831 | * @fid: file handle to convert | |
832 | * @fh_len: length of the file handle in bytes | |
833 | * @fh_type: type of file handle | |
834 | * @get_inode: filesystem callback to retrieve inode | |
835 | * | |
836 | * This function decodes @fid as long as it has one of the well-known | |
837 | * Linux filehandle types and calls @get_inode on it to retrieve the | |
838 | * inode for the object specified in the file handle. | |
839 | */ | |
840 | struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
841 | int fh_len, int fh_type, struct inode *(*get_inode) | |
842 | (struct super_block *sb, u64 ino, u32 gen)) | |
843 | { | |
844 | struct inode *inode = NULL; | |
845 | ||
846 | if (fh_len < 2) | |
847 | return NULL; | |
848 | ||
849 | switch (fh_type) { | |
850 | case FILEID_INO32_GEN: | |
851 | case FILEID_INO32_GEN_PARENT: | |
852 | inode = get_inode(sb, fid->i32.ino, fid->i32.gen); | |
853 | break; | |
854 | } | |
855 | ||
4ea3ada2 | 856 | return d_obtain_alias(inode); |
2596110a CH |
857 | } |
858 | EXPORT_SYMBOL_GPL(generic_fh_to_dentry); | |
859 | ||
860 | /** | |
861 | * generic_fh_to_dentry - generic helper for the fh_to_parent export operation | |
862 | * @sb: filesystem to do the file handle conversion on | |
863 | * @fid: file handle to convert | |
864 | * @fh_len: length of the file handle in bytes | |
865 | * @fh_type: type of file handle | |
866 | * @get_inode: filesystem callback to retrieve inode | |
867 | * | |
868 | * This function decodes @fid as long as it has one of the well-known | |
869 | * Linux filehandle types and calls @get_inode on it to retrieve the | |
870 | * inode for the _parent_ object specified in the file handle if it | |
871 | * is specified in the file handle, or NULL otherwise. | |
872 | */ | |
873 | struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid, | |
874 | int fh_len, int fh_type, struct inode *(*get_inode) | |
875 | (struct super_block *sb, u64 ino, u32 gen)) | |
876 | { | |
877 | struct inode *inode = NULL; | |
878 | ||
879 | if (fh_len <= 2) | |
880 | return NULL; | |
881 | ||
882 | switch (fh_type) { | |
883 | case FILEID_INO32_GEN_PARENT: | |
884 | inode = get_inode(sb, fid->i32.parent_ino, | |
885 | (fh_len > 3 ? fid->i32.parent_gen : 0)); | |
886 | break; | |
887 | } | |
888 | ||
4ea3ada2 | 889 | return d_obtain_alias(inode); |
2596110a CH |
890 | } |
891 | EXPORT_SYMBOL_GPL(generic_fh_to_parent); | |
892 | ||
1b061d92 CH |
893 | /** |
894 | * generic_file_fsync - generic fsync implementation for simple filesystems | |
895 | * @file: file to synchronize | |
896 | * @datasync: only synchronize essential metadata if true | |
897 | * | |
898 | * This is a generic implementation of the fsync method for simple | |
899 | * filesystems which track all non-inode metadata in the buffers list | |
900 | * hanging off the address_space structure. | |
901 | */ | |
902 | int generic_file_fsync(struct file *file, int datasync) | |
d5aacad5 | 903 | { |
7ea80859 | 904 | struct inode *inode = file->f_mapping->host; |
d5aacad5 AV |
905 | int err; |
906 | int ret; | |
907 | ||
908 | ret = sync_mapping_buffers(inode->i_mapping); | |
909 | if (!(inode->i_state & I_DIRTY)) | |
910 | return ret; | |
911 | if (datasync && !(inode->i_state & I_DIRTY_DATASYNC)) | |
912 | return ret; | |
913 | ||
c3765016 | 914 | err = sync_inode_metadata(inode, 1); |
d5aacad5 AV |
915 | if (ret == 0) |
916 | ret = err; | |
917 | return ret; | |
918 | } | |
1b061d92 CH |
919 | EXPORT_SYMBOL(generic_file_fsync); |
920 | ||
30ca22c7 PL |
921 | /** |
922 | * generic_check_addressable - Check addressability of file system | |
923 | * @blocksize_bits: log of file system block size | |
924 | * @num_blocks: number of blocks in file system | |
925 | * | |
926 | * Determine whether a file system with @num_blocks blocks (and a | |
927 | * block size of 2**@blocksize_bits) is addressable by the sector_t | |
928 | * and page cache of the system. Return 0 if so and -EFBIG otherwise. | |
929 | */ | |
930 | int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks) | |
931 | { | |
932 | u64 last_fs_block = num_blocks - 1; | |
a33f13ef JB |
933 | u64 last_fs_page = |
934 | last_fs_block >> (PAGE_CACHE_SHIFT - blocksize_bits); | |
30ca22c7 PL |
935 | |
936 | if (unlikely(num_blocks == 0)) | |
937 | return 0; | |
938 | ||
939 | if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT)) | |
940 | return -EINVAL; | |
941 | ||
a33f13ef JB |
942 | if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) || |
943 | (last_fs_page > (pgoff_t)(~0ULL))) { | |
30ca22c7 PL |
944 | return -EFBIG; |
945 | } | |
946 | return 0; | |
947 | } | |
948 | EXPORT_SYMBOL(generic_check_addressable); | |
949 | ||
1b061d92 CH |
950 | /* |
951 | * No-op implementation of ->fsync for in-memory filesystems. | |
952 | */ | |
953 | int noop_fsync(struct file *file, int datasync) | |
954 | { | |
955 | return 0; | |
956 | } | |
d5aacad5 | 957 | |
1da177e4 LT |
958 | EXPORT_SYMBOL(dcache_dir_close); |
959 | EXPORT_SYMBOL(dcache_dir_lseek); | |
960 | EXPORT_SYMBOL(dcache_dir_open); | |
961 | EXPORT_SYMBOL(dcache_readdir); | |
962 | EXPORT_SYMBOL(generic_read_dir); | |
51139ada | 963 | EXPORT_SYMBOL(mount_pseudo); |
afddba49 NP |
964 | EXPORT_SYMBOL(simple_write_begin); |
965 | EXPORT_SYMBOL(simple_write_end); | |
1da177e4 LT |
966 | EXPORT_SYMBOL(simple_dir_inode_operations); |
967 | EXPORT_SYMBOL(simple_dir_operations); | |
968 | EXPORT_SYMBOL(simple_empty); | |
1da177e4 LT |
969 | EXPORT_SYMBOL(simple_fill_super); |
970 | EXPORT_SYMBOL(simple_getattr); | |
971 | EXPORT_SYMBOL(simple_link); | |
972 | EXPORT_SYMBOL(simple_lookup); | |
973 | EXPORT_SYMBOL(simple_pin_fs); | |
1da177e4 LT |
974 | EXPORT_SYMBOL(simple_readpage); |
975 | EXPORT_SYMBOL(simple_release_fs); | |
976 | EXPORT_SYMBOL(simple_rename); | |
977 | EXPORT_SYMBOL(simple_rmdir); | |
978 | EXPORT_SYMBOL(simple_statfs); | |
1b061d92 | 979 | EXPORT_SYMBOL(noop_fsync); |
1da177e4 LT |
980 | EXPORT_SYMBOL(simple_unlink); |
981 | EXPORT_SYMBOL(simple_read_from_buffer); | |
6a727b43 | 982 | EXPORT_SYMBOL(simple_write_to_buffer); |
93b07113 | 983 | EXPORT_SYMBOL(memory_read_from_buffer); |
76791ab2 | 984 | EXPORT_SYMBOL(simple_transaction_set); |
1da177e4 LT |
985 | EXPORT_SYMBOL(simple_transaction_get); |
986 | EXPORT_SYMBOL(simple_transaction_read); | |
987 | EXPORT_SYMBOL(simple_transaction_release); | |
acaefc25 | 988 | EXPORT_SYMBOL_GPL(simple_attr_open); |
74bedc4d | 989 | EXPORT_SYMBOL_GPL(simple_attr_release); |
acaefc25 AB |
990 | EXPORT_SYMBOL_GPL(simple_attr_read); |
991 | EXPORT_SYMBOL_GPL(simple_attr_write); |