]>
Commit | Line | Data |
---|---|---|
1 | // SPDX-License-Identifier: GPL-2.0-only | |
2 | /* | |
3 | * fs/libfs.c | |
4 | * Library for filesystems writers. | |
5 | */ | |
6 | ||
7 | #include <linux/blkdev.h> | |
8 | #include <linux/export.h> | |
9 | #include <linux/pagemap.h> | |
10 | #include <linux/slab.h> | |
11 | #include <linux/cred.h> | |
12 | #include <linux/mount.h> | |
13 | #include <linux/vfs.h> | |
14 | #include <linux/quotaops.h> | |
15 | #include <linux/mutex.h> | |
16 | #include <linux/namei.h> | |
17 | #include <linux/exportfs.h> | |
18 | #include <linux/writeback.h> | |
19 | #include <linux/buffer_head.h> /* sync_mapping_buffers */ | |
20 | #include <linux/fs_context.h> | |
21 | #include <linux/pseudo_fs.h> | |
22 | #include <linux/fsnotify.h> | |
23 | #include <linux/unicode.h> | |
24 | #include <linux/fscrypt.h> | |
25 | ||
26 | #include <linux/uaccess.h> | |
27 | ||
28 | #include "internal.h" | |
29 | ||
30 | int simple_getattr(const struct path *path, struct kstat *stat, | |
31 | u32 request_mask, unsigned int query_flags) | |
32 | { | |
33 | struct inode *inode = d_inode(path->dentry); | |
34 | generic_fillattr(inode, stat); | |
35 | stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9); | |
36 | return 0; | |
37 | } | |
38 | EXPORT_SYMBOL(simple_getattr); | |
39 | ||
40 | int simple_statfs(struct dentry *dentry, struct kstatfs *buf) | |
41 | { | |
42 | buf->f_type = dentry->d_sb->s_magic; | |
43 | buf->f_bsize = PAGE_SIZE; | |
44 | buf->f_namelen = NAME_MAX; | |
45 | return 0; | |
46 | } | |
47 | EXPORT_SYMBOL(simple_statfs); | |
48 | ||
49 | /* | |
50 | * Retaining negative dentries for an in-memory filesystem just wastes | |
51 | * memory and lookup time: arrange for them to be deleted immediately. | |
52 | */ | |
53 | int always_delete_dentry(const struct dentry *dentry) | |
54 | { | |
55 | return 1; | |
56 | } | |
57 | EXPORT_SYMBOL(always_delete_dentry); | |
58 | ||
59 | const struct dentry_operations simple_dentry_operations = { | |
60 | .d_delete = always_delete_dentry, | |
61 | }; | |
62 | EXPORT_SYMBOL(simple_dentry_operations); | |
63 | ||
64 | /* | |
65 | * Lookup the data. This is trivial - if the dentry didn't already | |
66 | * exist, we know it is negative. Set d_op to delete negative dentries. | |
67 | */ | |
68 | struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) | |
69 | { | |
70 | if (dentry->d_name.len > NAME_MAX) | |
71 | return ERR_PTR(-ENAMETOOLONG); | |
72 | if (!dentry->d_sb->s_d_op) | |
73 | d_set_d_op(dentry, &simple_dentry_operations); | |
74 | d_add(dentry, NULL); | |
75 | return NULL; | |
76 | } | |
77 | EXPORT_SYMBOL(simple_lookup); | |
78 | ||
79 | int dcache_dir_open(struct inode *inode, struct file *file) | |
80 | { | |
81 | file->private_data = d_alloc_cursor(file->f_path.dentry); | |
82 | ||
83 | return file->private_data ? 0 : -ENOMEM; | |
84 | } | |
85 | EXPORT_SYMBOL(dcache_dir_open); | |
86 | ||
87 | int dcache_dir_close(struct inode *inode, struct file *file) | |
88 | { | |
89 | dput(file->private_data); | |
90 | return 0; | |
91 | } | |
92 | EXPORT_SYMBOL(dcache_dir_close); | |
93 | ||
94 | /* parent is locked at least shared */ | |
95 | /* | |
96 | * Returns an element of siblings' list. | |
97 | * We are looking for <count>th positive after <p>; if | |
98 | * found, dentry is grabbed and returned to caller. | |
99 | * If no such element exists, NULL is returned. | |
100 | */ | |
101 | static struct dentry *scan_positives(struct dentry *cursor, | |
102 | struct list_head *p, | |
103 | loff_t count, | |
104 | struct dentry *last) | |
105 | { | |
106 | struct dentry *dentry = cursor->d_parent, *found = NULL; | |
107 | ||
108 | spin_lock(&dentry->d_lock); | |
109 | while ((p = p->next) != &dentry->d_subdirs) { | |
110 | struct dentry *d = list_entry(p, struct dentry, d_child); | |
111 | // we must at least skip cursors, to avoid livelocks | |
112 | if (d->d_flags & DCACHE_DENTRY_CURSOR) | |
113 | continue; | |
114 | if (simple_positive(d) && !--count) { | |
115 | spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED); | |
116 | if (simple_positive(d)) | |
117 | found = dget_dlock(d); | |
118 | spin_unlock(&d->d_lock); | |
119 | if (likely(found)) | |
120 | break; | |
121 | count = 1; | |
122 | } | |
123 | if (need_resched()) { | |
124 | list_move(&cursor->d_child, p); | |
125 | p = &cursor->d_child; | |
126 | spin_unlock(&dentry->d_lock); | |
127 | cond_resched(); | |
128 | spin_lock(&dentry->d_lock); | |
129 | } | |
130 | } | |
131 | spin_unlock(&dentry->d_lock); | |
132 | dput(last); | |
133 | return found; | |
134 | } | |
135 | ||
136 | loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence) | |
137 | { | |
138 | struct dentry *dentry = file->f_path.dentry; | |
139 | switch (whence) { | |
140 | case 1: | |
141 | offset += file->f_pos; | |
142 | fallthrough; | |
143 | case 0: | |
144 | if (offset >= 0) | |
145 | break; | |
146 | fallthrough; | |
147 | default: | |
148 | return -EINVAL; | |
149 | } | |
150 | if (offset != file->f_pos) { | |
151 | struct dentry *cursor = file->private_data; | |
152 | struct dentry *to = NULL; | |
153 | ||
154 | inode_lock_shared(dentry->d_inode); | |
155 | ||
156 | if (offset > 2) | |
157 | to = scan_positives(cursor, &dentry->d_subdirs, | |
158 | offset - 2, NULL); | |
159 | spin_lock(&dentry->d_lock); | |
160 | if (to) | |
161 | list_move(&cursor->d_child, &to->d_child); | |
162 | else | |
163 | list_del_init(&cursor->d_child); | |
164 | spin_unlock(&dentry->d_lock); | |
165 | dput(to); | |
166 | ||
167 | file->f_pos = offset; | |
168 | ||
169 | inode_unlock_shared(dentry->d_inode); | |
170 | } | |
171 | return offset; | |
172 | } | |
173 | EXPORT_SYMBOL(dcache_dir_lseek); | |
174 | ||
175 | /* Relationship between i_mode and the DT_xxx types */ | |
176 | static inline unsigned char dt_type(struct inode *inode) | |
177 | { | |
178 | return (inode->i_mode >> 12) & 15; | |
179 | } | |
180 | ||
181 | /* | |
182 | * Directory is locked and all positive dentries in it are safe, since | |
183 | * for ramfs-type trees they can't go away without unlink() or rmdir(), | |
184 | * both impossible due to the lock on directory. | |
185 | */ | |
186 | ||
187 | int dcache_readdir(struct file *file, struct dir_context *ctx) | |
188 | { | |
189 | struct dentry *dentry = file->f_path.dentry; | |
190 | struct dentry *cursor = file->private_data; | |
191 | struct list_head *anchor = &dentry->d_subdirs; | |
192 | struct dentry *next = NULL; | |
193 | struct list_head *p; | |
194 | ||
195 | if (!dir_emit_dots(file, ctx)) | |
196 | return 0; | |
197 | ||
198 | if (ctx->pos == 2) | |
199 | p = anchor; | |
200 | else if (!list_empty(&cursor->d_child)) | |
201 | p = &cursor->d_child; | |
202 | else | |
203 | return 0; | |
204 | ||
205 | while ((next = scan_positives(cursor, p, 1, next)) != NULL) { | |
206 | if (!dir_emit(ctx, next->d_name.name, next->d_name.len, | |
207 | d_inode(next)->i_ino, dt_type(d_inode(next)))) | |
208 | break; | |
209 | ctx->pos++; | |
210 | p = &next->d_child; | |
211 | } | |
212 | spin_lock(&dentry->d_lock); | |
213 | if (next) | |
214 | list_move_tail(&cursor->d_child, &next->d_child); | |
215 | else | |
216 | list_del_init(&cursor->d_child); | |
217 | spin_unlock(&dentry->d_lock); | |
218 | dput(next); | |
219 | ||
220 | return 0; | |
221 | } | |
222 | EXPORT_SYMBOL(dcache_readdir); | |
223 | ||
224 | ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos) | |
225 | { | |
226 | return -EISDIR; | |
227 | } | |
228 | EXPORT_SYMBOL(generic_read_dir); | |
229 | ||
230 | const struct file_operations simple_dir_operations = { | |
231 | .open = dcache_dir_open, | |
232 | .release = dcache_dir_close, | |
233 | .llseek = dcache_dir_lseek, | |
234 | .read = generic_read_dir, | |
235 | .iterate_shared = dcache_readdir, | |
236 | .fsync = noop_fsync, | |
237 | }; | |
238 | EXPORT_SYMBOL(simple_dir_operations); | |
239 | ||
240 | const struct inode_operations simple_dir_inode_operations = { | |
241 | .lookup = simple_lookup, | |
242 | }; | |
243 | EXPORT_SYMBOL(simple_dir_inode_operations); | |
244 | ||
245 | static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev) | |
246 | { | |
247 | struct dentry *child = NULL; | |
248 | struct list_head *p = prev ? &prev->d_child : &parent->d_subdirs; | |
249 | ||
250 | spin_lock(&parent->d_lock); | |
251 | while ((p = p->next) != &parent->d_subdirs) { | |
252 | struct dentry *d = container_of(p, struct dentry, d_child); | |
253 | if (simple_positive(d)) { | |
254 | spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED); | |
255 | if (simple_positive(d)) | |
256 | child = dget_dlock(d); | |
257 | spin_unlock(&d->d_lock); | |
258 | if (likely(child)) | |
259 | break; | |
260 | } | |
261 | } | |
262 | spin_unlock(&parent->d_lock); | |
263 | dput(prev); | |
264 | return child; | |
265 | } | |
266 | ||
267 | void simple_recursive_removal(struct dentry *dentry, | |
268 | void (*callback)(struct dentry *)) | |
269 | { | |
270 | struct dentry *this = dget(dentry); | |
271 | while (true) { | |
272 | struct dentry *victim = NULL, *child; | |
273 | struct inode *inode = this->d_inode; | |
274 | ||
275 | inode_lock(inode); | |
276 | if (d_is_dir(this)) | |
277 | inode->i_flags |= S_DEAD; | |
278 | while ((child = find_next_child(this, victim)) == NULL) { | |
279 | // kill and ascend | |
280 | // update metadata while it's still locked | |
281 | inode->i_ctime = current_time(inode); | |
282 | clear_nlink(inode); | |
283 | inode_unlock(inode); | |
284 | victim = this; | |
285 | this = this->d_parent; | |
286 | inode = this->d_inode; | |
287 | inode_lock(inode); | |
288 | if (simple_positive(victim)) { | |
289 | d_invalidate(victim); // avoid lost mounts | |
290 | if (d_is_dir(victim)) | |
291 | fsnotify_rmdir(inode, victim); | |
292 | else | |
293 | fsnotify_unlink(inode, victim); | |
294 | if (callback) | |
295 | callback(victim); | |
296 | dput(victim); // unpin it | |
297 | } | |
298 | if (victim == dentry) { | |
299 | inode->i_ctime = inode->i_mtime = | |
300 | current_time(inode); | |
301 | if (d_is_dir(dentry)) | |
302 | drop_nlink(inode); | |
303 | inode_unlock(inode); | |
304 | dput(dentry); | |
305 | return; | |
306 | } | |
307 | } | |
308 | inode_unlock(inode); | |
309 | this = child; | |
310 | } | |
311 | } | |
312 | EXPORT_SYMBOL(simple_recursive_removal); | |
313 | ||
314 | static const struct super_operations simple_super_operations = { | |
315 | .statfs = simple_statfs, | |
316 | }; | |
317 | ||
318 | static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc) | |
319 | { | |
320 | struct pseudo_fs_context *ctx = fc->fs_private; | |
321 | struct inode *root; | |
322 | ||
323 | s->s_maxbytes = MAX_LFS_FILESIZE; | |
324 | s->s_blocksize = PAGE_SIZE; | |
325 | s->s_blocksize_bits = PAGE_SHIFT; | |
326 | s->s_magic = ctx->magic; | |
327 | s->s_op = ctx->ops ?: &simple_super_operations; | |
328 | s->s_xattr = ctx->xattr; | |
329 | s->s_time_gran = 1; | |
330 | root = new_inode(s); | |
331 | if (!root) | |
332 | return -ENOMEM; | |
333 | ||
334 | /* | |
335 | * since this is the first inode, make it number 1. New inodes created | |
336 | * after this must take care not to collide with it (by passing | |
337 | * max_reserved of 1 to iunique). | |
338 | */ | |
339 | root->i_ino = 1; | |
340 | root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR; | |
341 | root->i_atime = root->i_mtime = root->i_ctime = current_time(root); | |
342 | s->s_root = d_make_root(root); | |
343 | if (!s->s_root) | |
344 | return -ENOMEM; | |
345 | s->s_d_op = ctx->dops; | |
346 | return 0; | |
347 | } | |
348 | ||
349 | static int pseudo_fs_get_tree(struct fs_context *fc) | |
350 | { | |
351 | return get_tree_nodev(fc, pseudo_fs_fill_super); | |
352 | } | |
353 | ||
354 | static void pseudo_fs_free(struct fs_context *fc) | |
355 | { | |
356 | kfree(fc->fs_private); | |
357 | } | |
358 | ||
359 | static const struct fs_context_operations pseudo_fs_context_ops = { | |
360 | .free = pseudo_fs_free, | |
361 | .get_tree = pseudo_fs_get_tree, | |
362 | }; | |
363 | ||
364 | /* | |
365 | * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that | |
366 | * will never be mountable) | |
367 | */ | |
368 | struct pseudo_fs_context *init_pseudo(struct fs_context *fc, | |
369 | unsigned long magic) | |
370 | { | |
371 | struct pseudo_fs_context *ctx; | |
372 | ||
373 | ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL); | |
374 | if (likely(ctx)) { | |
375 | ctx->magic = magic; | |
376 | fc->fs_private = ctx; | |
377 | fc->ops = &pseudo_fs_context_ops; | |
378 | fc->sb_flags |= SB_NOUSER; | |
379 | fc->global = true; | |
380 | } | |
381 | return ctx; | |
382 | } | |
383 | EXPORT_SYMBOL(init_pseudo); | |
384 | ||
385 | int simple_open(struct inode *inode, struct file *file) | |
386 | { | |
387 | if (inode->i_private) | |
388 | file->private_data = inode->i_private; | |
389 | return 0; | |
390 | } | |
391 | EXPORT_SYMBOL(simple_open); | |
392 | ||
393 | int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry) | |
394 | { | |
395 | struct inode *inode = d_inode(old_dentry); | |
396 | ||
397 | inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode); | |
398 | inc_nlink(inode); | |
399 | ihold(inode); | |
400 | dget(dentry); | |
401 | d_instantiate(dentry, inode); | |
402 | return 0; | |
403 | } | |
404 | EXPORT_SYMBOL(simple_link); | |
405 | ||
406 | int simple_empty(struct dentry *dentry) | |
407 | { | |
408 | struct dentry *child; | |
409 | int ret = 0; | |
410 | ||
411 | spin_lock(&dentry->d_lock); | |
412 | list_for_each_entry(child, &dentry->d_subdirs, d_child) { | |
413 | spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED); | |
414 | if (simple_positive(child)) { | |
415 | spin_unlock(&child->d_lock); | |
416 | goto out; | |
417 | } | |
418 | spin_unlock(&child->d_lock); | |
419 | } | |
420 | ret = 1; | |
421 | out: | |
422 | spin_unlock(&dentry->d_lock); | |
423 | return ret; | |
424 | } | |
425 | EXPORT_SYMBOL(simple_empty); | |
426 | ||
427 | int simple_unlink(struct inode *dir, struct dentry *dentry) | |
428 | { | |
429 | struct inode *inode = d_inode(dentry); | |
430 | ||
431 | inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode); | |
432 | drop_nlink(inode); | |
433 | dput(dentry); | |
434 | return 0; | |
435 | } | |
436 | EXPORT_SYMBOL(simple_unlink); | |
437 | ||
438 | int simple_rmdir(struct inode *dir, struct dentry *dentry) | |
439 | { | |
440 | if (!simple_empty(dentry)) | |
441 | return -ENOTEMPTY; | |
442 | ||
443 | drop_nlink(d_inode(dentry)); | |
444 | simple_unlink(dir, dentry); | |
445 | drop_nlink(dir); | |
446 | return 0; | |
447 | } | |
448 | EXPORT_SYMBOL(simple_rmdir); | |
449 | ||
450 | int simple_rename(struct inode *old_dir, struct dentry *old_dentry, | |
451 | struct inode *new_dir, struct dentry *new_dentry, | |
452 | unsigned int flags) | |
453 | { | |
454 | struct inode *inode = d_inode(old_dentry); | |
455 | int they_are_dirs = d_is_dir(old_dentry); | |
456 | ||
457 | if (flags & ~RENAME_NOREPLACE) | |
458 | return -EINVAL; | |
459 | ||
460 | if (!simple_empty(new_dentry)) | |
461 | return -ENOTEMPTY; | |
462 | ||
463 | if (d_really_is_positive(new_dentry)) { | |
464 | simple_unlink(new_dir, new_dentry); | |
465 | if (they_are_dirs) { | |
466 | drop_nlink(d_inode(new_dentry)); | |
467 | drop_nlink(old_dir); | |
468 | } | |
469 | } else if (they_are_dirs) { | |
470 | drop_nlink(old_dir); | |
471 | inc_nlink(new_dir); | |
472 | } | |
473 | ||
474 | old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime = | |
475 | new_dir->i_mtime = inode->i_ctime = current_time(old_dir); | |
476 | ||
477 | return 0; | |
478 | } | |
479 | EXPORT_SYMBOL(simple_rename); | |
480 | ||
481 | /** | |
482 | * simple_setattr - setattr for simple filesystem | |
483 | * @dentry: dentry | |
484 | * @iattr: iattr structure | |
485 | * | |
486 | * Returns 0 on success, -error on failure. | |
487 | * | |
488 | * simple_setattr is a simple ->setattr implementation without a proper | |
489 | * implementation of size changes. | |
490 | * | |
491 | * It can either be used for in-memory filesystems or special files | |
492 | * on simple regular filesystems. Anything that needs to change on-disk | |
493 | * or wire state on size changes needs its own setattr method. | |
494 | */ | |
495 | int simple_setattr(struct dentry *dentry, struct iattr *iattr) | |
496 | { | |
497 | struct inode *inode = d_inode(dentry); | |
498 | int error; | |
499 | ||
500 | error = setattr_prepare(dentry, iattr); | |
501 | if (error) | |
502 | return error; | |
503 | ||
504 | if (iattr->ia_valid & ATTR_SIZE) | |
505 | truncate_setsize(inode, iattr->ia_size); | |
506 | setattr_copy(inode, iattr); | |
507 | mark_inode_dirty(inode); | |
508 | return 0; | |
509 | } | |
510 | EXPORT_SYMBOL(simple_setattr); | |
511 | ||
512 | int simple_readpage(struct file *file, struct page *page) | |
513 | { | |
514 | clear_highpage(page); | |
515 | flush_dcache_page(page); | |
516 | SetPageUptodate(page); | |
517 | unlock_page(page); | |
518 | return 0; | |
519 | } | |
520 | EXPORT_SYMBOL(simple_readpage); | |
521 | ||
522 | int simple_write_begin(struct file *file, struct address_space *mapping, | |
523 | loff_t pos, unsigned len, unsigned flags, | |
524 | struct page **pagep, void **fsdata) | |
525 | { | |
526 | struct page *page; | |
527 | pgoff_t index; | |
528 | ||
529 | index = pos >> PAGE_SHIFT; | |
530 | ||
531 | page = grab_cache_page_write_begin(mapping, index, flags); | |
532 | if (!page) | |
533 | return -ENOMEM; | |
534 | ||
535 | *pagep = page; | |
536 | ||
537 | if (!PageUptodate(page) && (len != PAGE_SIZE)) { | |
538 | unsigned from = pos & (PAGE_SIZE - 1); | |
539 | ||
540 | zero_user_segments(page, 0, from, from + len, PAGE_SIZE); | |
541 | } | |
542 | return 0; | |
543 | } | |
544 | EXPORT_SYMBOL(simple_write_begin); | |
545 | ||
546 | /** | |
547 | * simple_write_end - .write_end helper for non-block-device FSes | |
548 | * @file: See .write_end of address_space_operations | |
549 | * @mapping: " | |
550 | * @pos: " | |
551 | * @len: " | |
552 | * @copied: " | |
553 | * @page: " | |
554 | * @fsdata: " | |
555 | * | |
556 | * simple_write_end does the minimum needed for updating a page after writing is | |
557 | * done. It has the same API signature as the .write_end of | |
558 | * address_space_operations vector. So it can just be set onto .write_end for | |
559 | * FSes that don't need any other processing. i_mutex is assumed to be held. | |
560 | * Block based filesystems should use generic_write_end(). | |
561 | * NOTE: Even though i_size might get updated by this function, mark_inode_dirty | |
562 | * is not called, so a filesystem that actually does store data in .write_inode | |
563 | * should extend on what's done here with a call to mark_inode_dirty() in the | |
564 | * case that i_size has changed. | |
565 | * | |
566 | * Use *ONLY* with simple_readpage() | |
567 | */ | |
568 | int simple_write_end(struct file *file, struct address_space *mapping, | |
569 | loff_t pos, unsigned len, unsigned copied, | |
570 | struct page *page, void *fsdata) | |
571 | { | |
572 | struct inode *inode = page->mapping->host; | |
573 | loff_t last_pos = pos + copied; | |
574 | ||
575 | /* zero the stale part of the page if we did a short copy */ | |
576 | if (!PageUptodate(page)) { | |
577 | if (copied < len) { | |
578 | unsigned from = pos & (PAGE_SIZE - 1); | |
579 | ||
580 | zero_user(page, from + copied, len - copied); | |
581 | } | |
582 | SetPageUptodate(page); | |
583 | } | |
584 | /* | |
585 | * No need to use i_size_read() here, the i_size | |
586 | * cannot change under us because we hold the i_mutex. | |
587 | */ | |
588 | if (last_pos > inode->i_size) | |
589 | i_size_write(inode, last_pos); | |
590 | ||
591 | set_page_dirty(page); | |
592 | unlock_page(page); | |
593 | put_page(page); | |
594 | ||
595 | return copied; | |
596 | } | |
597 | EXPORT_SYMBOL(simple_write_end); | |
598 | ||
599 | /* | |
600 | * the inodes created here are not hashed. If you use iunique to generate | |
601 | * unique inode values later for this filesystem, then you must take care | |
602 | * to pass it an appropriate max_reserved value to avoid collisions. | |
603 | */ | |
604 | int simple_fill_super(struct super_block *s, unsigned long magic, | |
605 | const struct tree_descr *files) | |
606 | { | |
607 | struct inode *inode; | |
608 | struct dentry *root; | |
609 | struct dentry *dentry; | |
610 | int i; | |
611 | ||
612 | s->s_blocksize = PAGE_SIZE; | |
613 | s->s_blocksize_bits = PAGE_SHIFT; | |
614 | s->s_magic = magic; | |
615 | s->s_op = &simple_super_operations; | |
616 | s->s_time_gran = 1; | |
617 | ||
618 | inode = new_inode(s); | |
619 | if (!inode) | |
620 | return -ENOMEM; | |
621 | /* | |
622 | * because the root inode is 1, the files array must not contain an | |
623 | * entry at index 1 | |
624 | */ | |
625 | inode->i_ino = 1; | |
626 | inode->i_mode = S_IFDIR | 0755; | |
627 | inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); | |
628 | inode->i_op = &simple_dir_inode_operations; | |
629 | inode->i_fop = &simple_dir_operations; | |
630 | set_nlink(inode, 2); | |
631 | root = d_make_root(inode); | |
632 | if (!root) | |
633 | return -ENOMEM; | |
634 | for (i = 0; !files->name || files->name[0]; i++, files++) { | |
635 | if (!files->name) | |
636 | continue; | |
637 | ||
638 | /* warn if it tries to conflict with the root inode */ | |
639 | if (unlikely(i == 1)) | |
640 | printk(KERN_WARNING "%s: %s passed in a files array" | |
641 | "with an index of 1!\n", __func__, | |
642 | s->s_type->name); | |
643 | ||
644 | dentry = d_alloc_name(root, files->name); | |
645 | if (!dentry) | |
646 | goto out; | |
647 | inode = new_inode(s); | |
648 | if (!inode) { | |
649 | dput(dentry); | |
650 | goto out; | |
651 | } | |
652 | inode->i_mode = S_IFREG | files->mode; | |
653 | inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); | |
654 | inode->i_fop = files->ops; | |
655 | inode->i_ino = i; | |
656 | d_add(dentry, inode); | |
657 | } | |
658 | s->s_root = root; | |
659 | return 0; | |
660 | out: | |
661 | d_genocide(root); | |
662 | shrink_dcache_parent(root); | |
663 | dput(root); | |
664 | return -ENOMEM; | |
665 | } | |
666 | EXPORT_SYMBOL(simple_fill_super); | |
667 | ||
668 | static DEFINE_SPINLOCK(pin_fs_lock); | |
669 | ||
670 | int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count) | |
671 | { | |
672 | struct vfsmount *mnt = NULL; | |
673 | spin_lock(&pin_fs_lock); | |
674 | if (unlikely(!*mount)) { | |
675 | spin_unlock(&pin_fs_lock); | |
676 | mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL); | |
677 | if (IS_ERR(mnt)) | |
678 | return PTR_ERR(mnt); | |
679 | spin_lock(&pin_fs_lock); | |
680 | if (!*mount) | |
681 | *mount = mnt; | |
682 | } | |
683 | mntget(*mount); | |
684 | ++*count; | |
685 | spin_unlock(&pin_fs_lock); | |
686 | mntput(mnt); | |
687 | return 0; | |
688 | } | |
689 | EXPORT_SYMBOL(simple_pin_fs); | |
690 | ||
691 | void simple_release_fs(struct vfsmount **mount, int *count) | |
692 | { | |
693 | struct vfsmount *mnt; | |
694 | spin_lock(&pin_fs_lock); | |
695 | mnt = *mount; | |
696 | if (!--*count) | |
697 | *mount = NULL; | |
698 | spin_unlock(&pin_fs_lock); | |
699 | mntput(mnt); | |
700 | } | |
701 | EXPORT_SYMBOL(simple_release_fs); | |
702 | ||
703 | /** | |
704 | * simple_read_from_buffer - copy data from the buffer to user space | |
705 | * @to: the user space buffer to read to | |
706 | * @count: the maximum number of bytes to read | |
707 | * @ppos: the current position in the buffer | |
708 | * @from: the buffer to read from | |
709 | * @available: the size of the buffer | |
710 | * | |
711 | * The simple_read_from_buffer() function reads up to @count bytes from the | |
712 | * buffer @from at offset @ppos into the user space address starting at @to. | |
713 | * | |
714 | * On success, the number of bytes read is returned and the offset @ppos is | |
715 | * advanced by this number, or negative value is returned on error. | |
716 | **/ | |
717 | ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos, | |
718 | const void *from, size_t available) | |
719 | { | |
720 | loff_t pos = *ppos; | |
721 | size_t ret; | |
722 | ||
723 | if (pos < 0) | |
724 | return -EINVAL; | |
725 | if (pos >= available || !count) | |
726 | return 0; | |
727 | if (count > available - pos) | |
728 | count = available - pos; | |
729 | ret = copy_to_user(to, from + pos, count); | |
730 | if (ret == count) | |
731 | return -EFAULT; | |
732 | count -= ret; | |
733 | *ppos = pos + count; | |
734 | return count; | |
735 | } | |
736 | EXPORT_SYMBOL(simple_read_from_buffer); | |
737 | ||
738 | /** | |
739 | * simple_write_to_buffer - copy data from user space to the buffer | |
740 | * @to: the buffer to write to | |
741 | * @available: the size of the buffer | |
742 | * @ppos: the current position in the buffer | |
743 | * @from: the user space buffer to read from | |
744 | * @count: the maximum number of bytes to read | |
745 | * | |
746 | * The simple_write_to_buffer() function reads up to @count bytes from the user | |
747 | * space address starting at @from into the buffer @to at offset @ppos. | |
748 | * | |
749 | * On success, the number of bytes written is returned and the offset @ppos is | |
750 | * advanced by this number, or negative value is returned on error. | |
751 | **/ | |
752 | ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, | |
753 | const void __user *from, size_t count) | |
754 | { | |
755 | loff_t pos = *ppos; | |
756 | size_t res; | |
757 | ||
758 | if (pos < 0) | |
759 | return -EINVAL; | |
760 | if (pos >= available || !count) | |
761 | return 0; | |
762 | if (count > available - pos) | |
763 | count = available - pos; | |
764 | res = copy_from_user(to + pos, from, count); | |
765 | if (res == count) | |
766 | return -EFAULT; | |
767 | count -= res; | |
768 | *ppos = pos + count; | |
769 | return count; | |
770 | } | |
771 | EXPORT_SYMBOL(simple_write_to_buffer); | |
772 | ||
773 | /** | |
774 | * memory_read_from_buffer - copy data from the buffer | |
775 | * @to: the kernel space buffer to read to | |
776 | * @count: the maximum number of bytes to read | |
777 | * @ppos: the current position in the buffer | |
778 | * @from: the buffer to read from | |
779 | * @available: the size of the buffer | |
780 | * | |
781 | * The memory_read_from_buffer() function reads up to @count bytes from the | |
782 | * buffer @from at offset @ppos into the kernel space address starting at @to. | |
783 | * | |
784 | * On success, the number of bytes read is returned and the offset @ppos is | |
785 | * advanced by this number, or negative value is returned on error. | |
786 | **/ | |
787 | ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos, | |
788 | const void *from, size_t available) | |
789 | { | |
790 | loff_t pos = *ppos; | |
791 | ||
792 | if (pos < 0) | |
793 | return -EINVAL; | |
794 | if (pos >= available) | |
795 | return 0; | |
796 | if (count > available - pos) | |
797 | count = available - pos; | |
798 | memcpy(to, from + pos, count); | |
799 | *ppos = pos + count; | |
800 | ||
801 | return count; | |
802 | } | |
803 | EXPORT_SYMBOL(memory_read_from_buffer); | |
804 | ||
805 | /* | |
806 | * Transaction based IO. | |
807 | * The file expects a single write which triggers the transaction, and then | |
808 | * possibly a read which collects the result - which is stored in a | |
809 | * file-local buffer. | |
810 | */ | |
811 | ||
812 | void simple_transaction_set(struct file *file, size_t n) | |
813 | { | |
814 | struct simple_transaction_argresp *ar = file->private_data; | |
815 | ||
816 | BUG_ON(n > SIMPLE_TRANSACTION_LIMIT); | |
817 | ||
818 | /* | |
819 | * The barrier ensures that ar->size will really remain zero until | |
820 | * ar->data is ready for reading. | |
821 | */ | |
822 | smp_mb(); | |
823 | ar->size = n; | |
824 | } | |
825 | EXPORT_SYMBOL(simple_transaction_set); | |
826 | ||
827 | char *simple_transaction_get(struct file *file, const char __user *buf, size_t size) | |
828 | { | |
829 | struct simple_transaction_argresp *ar; | |
830 | static DEFINE_SPINLOCK(simple_transaction_lock); | |
831 | ||
832 | if (size > SIMPLE_TRANSACTION_LIMIT - 1) | |
833 | return ERR_PTR(-EFBIG); | |
834 | ||
835 | ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL); | |
836 | if (!ar) | |
837 | return ERR_PTR(-ENOMEM); | |
838 | ||
839 | spin_lock(&simple_transaction_lock); | |
840 | ||
841 | /* only one write allowed per open */ | |
842 | if (file->private_data) { | |
843 | spin_unlock(&simple_transaction_lock); | |
844 | free_page((unsigned long)ar); | |
845 | return ERR_PTR(-EBUSY); | |
846 | } | |
847 | ||
848 | file->private_data = ar; | |
849 | ||
850 | spin_unlock(&simple_transaction_lock); | |
851 | ||
852 | if (copy_from_user(ar->data, buf, size)) | |
853 | return ERR_PTR(-EFAULT); | |
854 | ||
855 | return ar->data; | |
856 | } | |
857 | EXPORT_SYMBOL(simple_transaction_get); | |
858 | ||
859 | ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos) | |
860 | { | |
861 | struct simple_transaction_argresp *ar = file->private_data; | |
862 | ||
863 | if (!ar) | |
864 | return 0; | |
865 | return simple_read_from_buffer(buf, size, pos, ar->data, ar->size); | |
866 | } | |
867 | EXPORT_SYMBOL(simple_transaction_read); | |
868 | ||
869 | int simple_transaction_release(struct inode *inode, struct file *file) | |
870 | { | |
871 | free_page((unsigned long)file->private_data); | |
872 | return 0; | |
873 | } | |
874 | EXPORT_SYMBOL(simple_transaction_release); | |
875 | ||
876 | /* Simple attribute files */ | |
877 | ||
878 | struct simple_attr { | |
879 | int (*get)(void *, u64 *); | |
880 | int (*set)(void *, u64); | |
881 | char get_buf[24]; /* enough to store a u64 and "\n\0" */ | |
882 | char set_buf[24]; | |
883 | void *data; | |
884 | const char *fmt; /* format for read operation */ | |
885 | struct mutex mutex; /* protects access to these buffers */ | |
886 | }; | |
887 | ||
888 | /* simple_attr_open is called by an actual attribute open file operation | |
889 | * to set the attribute specific access operations. */ | |
890 | int simple_attr_open(struct inode *inode, struct file *file, | |
891 | int (*get)(void *, u64 *), int (*set)(void *, u64), | |
892 | const char *fmt) | |
893 | { | |
894 | struct simple_attr *attr; | |
895 | ||
896 | attr = kzalloc(sizeof(*attr), GFP_KERNEL); | |
897 | if (!attr) | |
898 | return -ENOMEM; | |
899 | ||
900 | attr->get = get; | |
901 | attr->set = set; | |
902 | attr->data = inode->i_private; | |
903 | attr->fmt = fmt; | |
904 | mutex_init(&attr->mutex); | |
905 | ||
906 | file->private_data = attr; | |
907 | ||
908 | return nonseekable_open(inode, file); | |
909 | } | |
910 | EXPORT_SYMBOL_GPL(simple_attr_open); | |
911 | ||
912 | int simple_attr_release(struct inode *inode, struct file *file) | |
913 | { | |
914 | kfree(file->private_data); | |
915 | return 0; | |
916 | } | |
917 | EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */ | |
918 | ||
919 | /* read from the buffer that is filled with the get function */ | |
920 | ssize_t simple_attr_read(struct file *file, char __user *buf, | |
921 | size_t len, loff_t *ppos) | |
922 | { | |
923 | struct simple_attr *attr; | |
924 | size_t size; | |
925 | ssize_t ret; | |
926 | ||
927 | attr = file->private_data; | |
928 | ||
929 | if (!attr->get) | |
930 | return -EACCES; | |
931 | ||
932 | ret = mutex_lock_interruptible(&attr->mutex); | |
933 | if (ret) | |
934 | return ret; | |
935 | ||
936 | if (*ppos && attr->get_buf[0]) { | |
937 | /* continued read */ | |
938 | size = strlen(attr->get_buf); | |
939 | } else { | |
940 | /* first read */ | |
941 | u64 val; | |
942 | ret = attr->get(attr->data, &val); | |
943 | if (ret) | |
944 | goto out; | |
945 | ||
946 | size = scnprintf(attr->get_buf, sizeof(attr->get_buf), | |
947 | attr->fmt, (unsigned long long)val); | |
948 | } | |
949 | ||
950 | ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size); | |
951 | out: | |
952 | mutex_unlock(&attr->mutex); | |
953 | return ret; | |
954 | } | |
955 | EXPORT_SYMBOL_GPL(simple_attr_read); | |
956 | ||
957 | /* interpret the buffer as a number to call the set function with */ | |
958 | ssize_t simple_attr_write(struct file *file, const char __user *buf, | |
959 | size_t len, loff_t *ppos) | |
960 | { | |
961 | struct simple_attr *attr; | |
962 | unsigned long long val; | |
963 | size_t size; | |
964 | ssize_t ret; | |
965 | ||
966 | attr = file->private_data; | |
967 | if (!attr->set) | |
968 | return -EACCES; | |
969 | ||
970 | ret = mutex_lock_interruptible(&attr->mutex); | |
971 | if (ret) | |
972 | return ret; | |
973 | ||
974 | ret = -EFAULT; | |
975 | size = min(sizeof(attr->set_buf) - 1, len); | |
976 | if (copy_from_user(attr->set_buf, buf, size)) | |
977 | goto out; | |
978 | ||
979 | attr->set_buf[size] = '\0'; | |
980 | ret = kstrtoull(attr->set_buf, 0, &val); | |
981 | if (ret) | |
982 | goto out; | |
983 | ret = attr->set(attr->data, val); | |
984 | if (ret == 0) | |
985 | ret = len; /* on success, claim we got the whole input */ | |
986 | out: | |
987 | mutex_unlock(&attr->mutex); | |
988 | return ret; | |
989 | } | |
990 | EXPORT_SYMBOL_GPL(simple_attr_write); | |
991 | ||
992 | /** | |
993 | * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation | |
994 | * @sb: filesystem to do the file handle conversion on | |
995 | * @fid: file handle to convert | |
996 | * @fh_len: length of the file handle in bytes | |
997 | * @fh_type: type of file handle | |
998 | * @get_inode: filesystem callback to retrieve inode | |
999 | * | |
1000 | * This function decodes @fid as long as it has one of the well-known | |
1001 | * Linux filehandle types and calls @get_inode on it to retrieve the | |
1002 | * inode for the object specified in the file handle. | |
1003 | */ | |
1004 | struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
1005 | int fh_len, int fh_type, struct inode *(*get_inode) | |
1006 | (struct super_block *sb, u64 ino, u32 gen)) | |
1007 | { | |
1008 | struct inode *inode = NULL; | |
1009 | ||
1010 | if (fh_len < 2) | |
1011 | return NULL; | |
1012 | ||
1013 | switch (fh_type) { | |
1014 | case FILEID_INO32_GEN: | |
1015 | case FILEID_INO32_GEN_PARENT: | |
1016 | inode = get_inode(sb, fid->i32.ino, fid->i32.gen); | |
1017 | break; | |
1018 | } | |
1019 | ||
1020 | return d_obtain_alias(inode); | |
1021 | } | |
1022 | EXPORT_SYMBOL_GPL(generic_fh_to_dentry); | |
1023 | ||
1024 | /** | |
1025 | * generic_fh_to_parent - generic helper for the fh_to_parent export operation | |
1026 | * @sb: filesystem to do the file handle conversion on | |
1027 | * @fid: file handle to convert | |
1028 | * @fh_len: length of the file handle in bytes | |
1029 | * @fh_type: type of file handle | |
1030 | * @get_inode: filesystem callback to retrieve inode | |
1031 | * | |
1032 | * This function decodes @fid as long as it has one of the well-known | |
1033 | * Linux filehandle types and calls @get_inode on it to retrieve the | |
1034 | * inode for the _parent_ object specified in the file handle if it | |
1035 | * is specified in the file handle, or NULL otherwise. | |
1036 | */ | |
1037 | struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid, | |
1038 | int fh_len, int fh_type, struct inode *(*get_inode) | |
1039 | (struct super_block *sb, u64 ino, u32 gen)) | |
1040 | { | |
1041 | struct inode *inode = NULL; | |
1042 | ||
1043 | if (fh_len <= 2) | |
1044 | return NULL; | |
1045 | ||
1046 | switch (fh_type) { | |
1047 | case FILEID_INO32_GEN_PARENT: | |
1048 | inode = get_inode(sb, fid->i32.parent_ino, | |
1049 | (fh_len > 3 ? fid->i32.parent_gen : 0)); | |
1050 | break; | |
1051 | } | |
1052 | ||
1053 | return d_obtain_alias(inode); | |
1054 | } | |
1055 | EXPORT_SYMBOL_GPL(generic_fh_to_parent); | |
1056 | ||
1057 | /** | |
1058 | * __generic_file_fsync - generic fsync implementation for simple filesystems | |
1059 | * | |
1060 | * @file: file to synchronize | |
1061 | * @start: start offset in bytes | |
1062 | * @end: end offset in bytes (inclusive) | |
1063 | * @datasync: only synchronize essential metadata if true | |
1064 | * | |
1065 | * This is a generic implementation of the fsync method for simple | |
1066 | * filesystems which track all non-inode metadata in the buffers list | |
1067 | * hanging off the address_space structure. | |
1068 | */ | |
1069 | int __generic_file_fsync(struct file *file, loff_t start, loff_t end, | |
1070 | int datasync) | |
1071 | { | |
1072 | struct inode *inode = file->f_mapping->host; | |
1073 | int err; | |
1074 | int ret; | |
1075 | ||
1076 | err = file_write_and_wait_range(file, start, end); | |
1077 | if (err) | |
1078 | return err; | |
1079 | ||
1080 | inode_lock(inode); | |
1081 | ret = sync_mapping_buffers(inode->i_mapping); | |
1082 | if (!(inode->i_state & I_DIRTY_ALL)) | |
1083 | goto out; | |
1084 | if (datasync && !(inode->i_state & I_DIRTY_DATASYNC)) | |
1085 | goto out; | |
1086 | ||
1087 | err = sync_inode_metadata(inode, 1); | |
1088 | if (ret == 0) | |
1089 | ret = err; | |
1090 | ||
1091 | out: | |
1092 | inode_unlock(inode); | |
1093 | /* check and advance again to catch errors after syncing out buffers */ | |
1094 | err = file_check_and_advance_wb_err(file); | |
1095 | if (ret == 0) | |
1096 | ret = err; | |
1097 | return ret; | |
1098 | } | |
1099 | EXPORT_SYMBOL(__generic_file_fsync); | |
1100 | ||
1101 | /** | |
1102 | * generic_file_fsync - generic fsync implementation for simple filesystems | |
1103 | * with flush | |
1104 | * @file: file to synchronize | |
1105 | * @start: start offset in bytes | |
1106 | * @end: end offset in bytes (inclusive) | |
1107 | * @datasync: only synchronize essential metadata if true | |
1108 | * | |
1109 | */ | |
1110 | ||
1111 | int generic_file_fsync(struct file *file, loff_t start, loff_t end, | |
1112 | int datasync) | |
1113 | { | |
1114 | struct inode *inode = file->f_mapping->host; | |
1115 | int err; | |
1116 | ||
1117 | err = __generic_file_fsync(file, start, end, datasync); | |
1118 | if (err) | |
1119 | return err; | |
1120 | return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL); | |
1121 | } | |
1122 | EXPORT_SYMBOL(generic_file_fsync); | |
1123 | ||
1124 | /** | |
1125 | * generic_check_addressable - Check addressability of file system | |
1126 | * @blocksize_bits: log of file system block size | |
1127 | * @num_blocks: number of blocks in file system | |
1128 | * | |
1129 | * Determine whether a file system with @num_blocks blocks (and a | |
1130 | * block size of 2**@blocksize_bits) is addressable by the sector_t | |
1131 | * and page cache of the system. Return 0 if so and -EFBIG otherwise. | |
1132 | */ | |
1133 | int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks) | |
1134 | { | |
1135 | u64 last_fs_block = num_blocks - 1; | |
1136 | u64 last_fs_page = | |
1137 | last_fs_block >> (PAGE_SHIFT - blocksize_bits); | |
1138 | ||
1139 | if (unlikely(num_blocks == 0)) | |
1140 | return 0; | |
1141 | ||
1142 | if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT)) | |
1143 | return -EINVAL; | |
1144 | ||
1145 | if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) || | |
1146 | (last_fs_page > (pgoff_t)(~0ULL))) { | |
1147 | return -EFBIG; | |
1148 | } | |
1149 | return 0; | |
1150 | } | |
1151 | EXPORT_SYMBOL(generic_check_addressable); | |
1152 | ||
1153 | /* | |
1154 | * No-op implementation of ->fsync for in-memory filesystems. | |
1155 | */ | |
1156 | int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync) | |
1157 | { | |
1158 | return 0; | |
1159 | } | |
1160 | EXPORT_SYMBOL(noop_fsync); | |
1161 | ||
1162 | int noop_set_page_dirty(struct page *page) | |
1163 | { | |
1164 | /* | |
1165 | * Unlike __set_page_dirty_no_writeback that handles dirty page | |
1166 | * tracking in the page object, dax does all dirty tracking in | |
1167 | * the inode address_space in response to mkwrite faults. In the | |
1168 | * dax case we only need to worry about potentially dirty CPU | |
1169 | * caches, not dirty page cache pages to write back. | |
1170 | * | |
1171 | * This callback is defined to prevent fallback to | |
1172 | * __set_page_dirty_buffers() in set_page_dirty(). | |
1173 | */ | |
1174 | return 0; | |
1175 | } | |
1176 | EXPORT_SYMBOL_GPL(noop_set_page_dirty); | |
1177 | ||
1178 | void noop_invalidatepage(struct page *page, unsigned int offset, | |
1179 | unsigned int length) | |
1180 | { | |
1181 | /* | |
1182 | * There is no page cache to invalidate in the dax case, however | |
1183 | * we need this callback defined to prevent falling back to | |
1184 | * block_invalidatepage() in do_invalidatepage(). | |
1185 | */ | |
1186 | } | |
1187 | EXPORT_SYMBOL_GPL(noop_invalidatepage); | |
1188 | ||
1189 | ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter) | |
1190 | { | |
1191 | /* | |
1192 | * iomap based filesystems support direct I/O without need for | |
1193 | * this callback. However, it still needs to be set in | |
1194 | * inode->a_ops so that open/fcntl know that direct I/O is | |
1195 | * generally supported. | |
1196 | */ | |
1197 | return -EINVAL; | |
1198 | } | |
1199 | EXPORT_SYMBOL_GPL(noop_direct_IO); | |
1200 | ||
1201 | /* Because kfree isn't assignment-compatible with void(void*) ;-/ */ | |
1202 | void kfree_link(void *p) | |
1203 | { | |
1204 | kfree(p); | |
1205 | } | |
1206 | EXPORT_SYMBOL(kfree_link); | |
1207 | ||
1208 | /* | |
1209 | * nop .set_page_dirty method so that people can use .page_mkwrite on | |
1210 | * anon inodes. | |
1211 | */ | |
1212 | static int anon_set_page_dirty(struct page *page) | |
1213 | { | |
1214 | return 0; | |
1215 | }; | |
1216 | ||
1217 | /* | |
1218 | * A single inode exists for all anon_inode files. Contrary to pipes, | |
1219 | * anon_inode inodes have no associated per-instance data, so we need | |
1220 | * only allocate one of them. | |
1221 | */ | |
1222 | struct inode *alloc_anon_inode(struct super_block *s) | |
1223 | { | |
1224 | static const struct address_space_operations anon_aops = { | |
1225 | .set_page_dirty = anon_set_page_dirty, | |
1226 | }; | |
1227 | struct inode *inode = new_inode_pseudo(s); | |
1228 | ||
1229 | if (!inode) | |
1230 | return ERR_PTR(-ENOMEM); | |
1231 | ||
1232 | inode->i_ino = get_next_ino(); | |
1233 | inode->i_mapping->a_ops = &anon_aops; | |
1234 | ||
1235 | /* | |
1236 | * Mark the inode dirty from the very beginning, | |
1237 | * that way it will never be moved to the dirty | |
1238 | * list because mark_inode_dirty() will think | |
1239 | * that it already _is_ on the dirty list. | |
1240 | */ | |
1241 | inode->i_state = I_DIRTY; | |
1242 | inode->i_mode = S_IRUSR | S_IWUSR; | |
1243 | inode->i_uid = current_fsuid(); | |
1244 | inode->i_gid = current_fsgid(); | |
1245 | inode->i_flags |= S_PRIVATE; | |
1246 | inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); | |
1247 | return inode; | |
1248 | } | |
1249 | EXPORT_SYMBOL(alloc_anon_inode); | |
1250 | ||
1251 | /** | |
1252 | * simple_nosetlease - generic helper for prohibiting leases | |
1253 | * @filp: file pointer | |
1254 | * @arg: type of lease to obtain | |
1255 | * @flp: new lease supplied for insertion | |
1256 | * @priv: private data for lm_setup operation | |
1257 | * | |
1258 | * Generic helper for filesystems that do not wish to allow leases to be set. | |
1259 | * All arguments are ignored and it just returns -EINVAL. | |
1260 | */ | |
1261 | int | |
1262 | simple_nosetlease(struct file *filp, long arg, struct file_lock **flp, | |
1263 | void **priv) | |
1264 | { | |
1265 | return -EINVAL; | |
1266 | } | |
1267 | EXPORT_SYMBOL(simple_nosetlease); | |
1268 | ||
1269 | /** | |
1270 | * simple_get_link - generic helper to get the target of "fast" symlinks | |
1271 | * @dentry: not used here | |
1272 | * @inode: the symlink inode | |
1273 | * @done: not used here | |
1274 | * | |
1275 | * Generic helper for filesystems to use for symlink inodes where a pointer to | |
1276 | * the symlink target is stored in ->i_link. NOTE: this isn't normally called, | |
1277 | * since as an optimization the path lookup code uses any non-NULL ->i_link | |
1278 | * directly, without calling ->get_link(). But ->get_link() still must be set, | |
1279 | * to mark the inode_operations as being for a symlink. | |
1280 | * | |
1281 | * Return: the symlink target | |
1282 | */ | |
1283 | const char *simple_get_link(struct dentry *dentry, struct inode *inode, | |
1284 | struct delayed_call *done) | |
1285 | { | |
1286 | return inode->i_link; | |
1287 | } | |
1288 | EXPORT_SYMBOL(simple_get_link); | |
1289 | ||
1290 | const struct inode_operations simple_symlink_inode_operations = { | |
1291 | .get_link = simple_get_link, | |
1292 | }; | |
1293 | EXPORT_SYMBOL(simple_symlink_inode_operations); | |
1294 | ||
1295 | /* | |
1296 | * Operations for a permanently empty directory. | |
1297 | */ | |
1298 | static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) | |
1299 | { | |
1300 | return ERR_PTR(-ENOENT); | |
1301 | } | |
1302 | ||
1303 | static int empty_dir_getattr(const struct path *path, struct kstat *stat, | |
1304 | u32 request_mask, unsigned int query_flags) | |
1305 | { | |
1306 | struct inode *inode = d_inode(path->dentry); | |
1307 | generic_fillattr(inode, stat); | |
1308 | return 0; | |
1309 | } | |
1310 | ||
1311 | static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr) | |
1312 | { | |
1313 | return -EPERM; | |
1314 | } | |
1315 | ||
1316 | static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size) | |
1317 | { | |
1318 | return -EOPNOTSUPP; | |
1319 | } | |
1320 | ||
1321 | static const struct inode_operations empty_dir_inode_operations = { | |
1322 | .lookup = empty_dir_lookup, | |
1323 | .permission = generic_permission, | |
1324 | .setattr = empty_dir_setattr, | |
1325 | .getattr = empty_dir_getattr, | |
1326 | .listxattr = empty_dir_listxattr, | |
1327 | }; | |
1328 | ||
1329 | static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence) | |
1330 | { | |
1331 | /* An empty directory has two entries . and .. at offsets 0 and 1 */ | |
1332 | return generic_file_llseek_size(file, offset, whence, 2, 2); | |
1333 | } | |
1334 | ||
1335 | static int empty_dir_readdir(struct file *file, struct dir_context *ctx) | |
1336 | { | |
1337 | dir_emit_dots(file, ctx); | |
1338 | return 0; | |
1339 | } | |
1340 | ||
1341 | static const struct file_operations empty_dir_operations = { | |
1342 | .llseek = empty_dir_llseek, | |
1343 | .read = generic_read_dir, | |
1344 | .iterate_shared = empty_dir_readdir, | |
1345 | .fsync = noop_fsync, | |
1346 | }; | |
1347 | ||
1348 | ||
1349 | void make_empty_dir_inode(struct inode *inode) | |
1350 | { | |
1351 | set_nlink(inode, 2); | |
1352 | inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO; | |
1353 | inode->i_uid = GLOBAL_ROOT_UID; | |
1354 | inode->i_gid = GLOBAL_ROOT_GID; | |
1355 | inode->i_rdev = 0; | |
1356 | inode->i_size = 0; | |
1357 | inode->i_blkbits = PAGE_SHIFT; | |
1358 | inode->i_blocks = 0; | |
1359 | ||
1360 | inode->i_op = &empty_dir_inode_operations; | |
1361 | inode->i_opflags &= ~IOP_XATTR; | |
1362 | inode->i_fop = &empty_dir_operations; | |
1363 | } | |
1364 | ||
1365 | bool is_empty_dir_inode(struct inode *inode) | |
1366 | { | |
1367 | return (inode->i_fop == &empty_dir_operations) && | |
1368 | (inode->i_op == &empty_dir_inode_operations); | |
1369 | } | |
1370 | ||
1371 | #ifdef CONFIG_UNICODE | |
1372 | /* | |
1373 | * Determine if the name of a dentry should be casefolded. | |
1374 | * | |
1375 | * Return: if names will need casefolding | |
1376 | */ | |
1377 | static bool needs_casefold(const struct inode *dir) | |
1378 | { | |
1379 | return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding; | |
1380 | } | |
1381 | ||
1382 | /** | |
1383 | * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems | |
1384 | * @dentry: dentry whose name we are checking against | |
1385 | * @len: len of name of dentry | |
1386 | * @str: str pointer to name of dentry | |
1387 | * @name: Name to compare against | |
1388 | * | |
1389 | * Return: 0 if names match, 1 if mismatch, or -ERRNO | |
1390 | */ | |
1391 | int generic_ci_d_compare(const struct dentry *dentry, unsigned int len, | |
1392 | const char *str, const struct qstr *name) | |
1393 | { | |
1394 | const struct dentry *parent = READ_ONCE(dentry->d_parent); | |
1395 | const struct inode *dir = READ_ONCE(parent->d_inode); | |
1396 | const struct super_block *sb = dentry->d_sb; | |
1397 | const struct unicode_map *um = sb->s_encoding; | |
1398 | struct qstr qstr = QSTR_INIT(str, len); | |
1399 | char strbuf[DNAME_INLINE_LEN]; | |
1400 | int ret; | |
1401 | ||
1402 | if (!dir || !needs_casefold(dir)) | |
1403 | goto fallback; | |
1404 | /* | |
1405 | * If the dentry name is stored in-line, then it may be concurrently | |
1406 | * modified by a rename. If this happens, the VFS will eventually retry | |
1407 | * the lookup, so it doesn't matter what ->d_compare() returns. | |
1408 | * However, it's unsafe to call utf8_strncasecmp() with an unstable | |
1409 | * string. Therefore, we have to copy the name into a temporary buffer. | |
1410 | */ | |
1411 | if (len <= DNAME_INLINE_LEN - 1) { | |
1412 | memcpy(strbuf, str, len); | |
1413 | strbuf[len] = 0; | |
1414 | qstr.name = strbuf; | |
1415 | /* prevent compiler from optimizing out the temporary buffer */ | |
1416 | barrier(); | |
1417 | } | |
1418 | ret = utf8_strncasecmp(um, name, &qstr); | |
1419 | if (ret >= 0) | |
1420 | return ret; | |
1421 | ||
1422 | if (sb_has_strict_encoding(sb)) | |
1423 | return -EINVAL; | |
1424 | fallback: | |
1425 | if (len != name->len) | |
1426 | return 1; | |
1427 | return !!memcmp(str, name->name, len); | |
1428 | } | |
1429 | EXPORT_SYMBOL(generic_ci_d_compare); | |
1430 | ||
1431 | /** | |
1432 | * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems | |
1433 | * @dentry: dentry of the parent directory | |
1434 | * @str: qstr of name whose hash we should fill in | |
1435 | * | |
1436 | * Return: 0 if hash was successful or unchanged, and -EINVAL on error | |
1437 | */ | |
1438 | int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str) | |
1439 | { | |
1440 | const struct inode *dir = READ_ONCE(dentry->d_inode); | |
1441 | struct super_block *sb = dentry->d_sb; | |
1442 | const struct unicode_map *um = sb->s_encoding; | |
1443 | int ret = 0; | |
1444 | ||
1445 | if (!dir || !needs_casefold(dir)) | |
1446 | return 0; | |
1447 | ||
1448 | ret = utf8_casefold_hash(um, dentry, str); | |
1449 | if (ret < 0 && sb_has_strict_encoding(sb)) | |
1450 | return -EINVAL; | |
1451 | return 0; | |
1452 | } | |
1453 | EXPORT_SYMBOL(generic_ci_d_hash); | |
1454 | ||
1455 | static const struct dentry_operations generic_ci_dentry_ops = { | |
1456 | .d_hash = generic_ci_d_hash, | |
1457 | .d_compare = generic_ci_d_compare, | |
1458 | }; | |
1459 | #endif | |
1460 | ||
1461 | #ifdef CONFIG_FS_ENCRYPTION | |
1462 | static const struct dentry_operations generic_encrypted_dentry_ops = { | |
1463 | .d_revalidate = fscrypt_d_revalidate, | |
1464 | }; | |
1465 | #endif | |
1466 | ||
1467 | #if defined(CONFIG_FS_ENCRYPTION) && defined(CONFIG_UNICODE) | |
1468 | static const struct dentry_operations generic_encrypted_ci_dentry_ops = { | |
1469 | .d_hash = generic_ci_d_hash, | |
1470 | .d_compare = generic_ci_d_compare, | |
1471 | .d_revalidate = fscrypt_d_revalidate, | |
1472 | }; | |
1473 | #endif | |
1474 | ||
1475 | /** | |
1476 | * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry | |
1477 | * @dentry: dentry to set ops on | |
1478 | * | |
1479 | * Casefolded directories need d_hash and d_compare set, so that the dentries | |
1480 | * contained in them are handled case-insensitively. Note that these operations | |
1481 | * are needed on the parent directory rather than on the dentries in it, and | |
1482 | * while the casefolding flag can be toggled on and off on an empty directory, | |
1483 | * dentry_operations can't be changed later. As a result, if the filesystem has | |
1484 | * casefolding support enabled at all, we have to give all dentries the | |
1485 | * casefolding operations even if their inode doesn't have the casefolding flag | |
1486 | * currently (and thus the casefolding ops would be no-ops for now). | |
1487 | * | |
1488 | * Encryption works differently in that the only dentry operation it needs is | |
1489 | * d_revalidate, which it only needs on dentries that have the no-key name flag. | |
1490 | * The no-key flag can't be set "later", so we don't have to worry about that. | |
1491 | * | |
1492 | * Finally, to maximize compatibility with overlayfs (which isn't compatible | |
1493 | * with certain dentry operations) and to avoid taking an unnecessary | |
1494 | * performance hit, we use custom dentry_operations for each possible | |
1495 | * combination rather than always installing all operations. | |
1496 | */ | |
1497 | void generic_set_encrypted_ci_d_ops(struct dentry *dentry) | |
1498 | { | |
1499 | #ifdef CONFIG_FS_ENCRYPTION | |
1500 | bool needs_encrypt_ops = dentry->d_flags & DCACHE_NOKEY_NAME; | |
1501 | #endif | |
1502 | #ifdef CONFIG_UNICODE | |
1503 | bool needs_ci_ops = dentry->d_sb->s_encoding; | |
1504 | #endif | |
1505 | #if defined(CONFIG_FS_ENCRYPTION) && defined(CONFIG_UNICODE) | |
1506 | if (needs_encrypt_ops && needs_ci_ops) { | |
1507 | d_set_d_op(dentry, &generic_encrypted_ci_dentry_ops); | |
1508 | return; | |
1509 | } | |
1510 | #endif | |
1511 | #ifdef CONFIG_FS_ENCRYPTION | |
1512 | if (needs_encrypt_ops) { | |
1513 | d_set_d_op(dentry, &generic_encrypted_dentry_ops); | |
1514 | return; | |
1515 | } | |
1516 | #endif | |
1517 | #ifdef CONFIG_UNICODE | |
1518 | if (needs_ci_ops) { | |
1519 | d_set_d_op(dentry, &generic_ci_dentry_ops); | |
1520 | return; | |
1521 | } | |
1522 | #endif | |
1523 | } | |
1524 | EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops); |