Line data Source code
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 0 : int simple_getattr(struct user_namespace *mnt_userns, const struct path *path,
31 : struct kstat *stat, u32 request_mask,
32 : unsigned int query_flags)
33 : {
34 0 : struct inode *inode = d_inode(path->dentry);
35 0 : generic_fillattr(&init_user_ns, inode, stat);
36 0 : stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
37 0 : return 0;
38 : }
39 : EXPORT_SYMBOL(simple_getattr);
40 :
41 45 : int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
42 : {
43 45 : buf->f_type = dentry->d_sb->s_magic;
44 45 : buf->f_bsize = PAGE_SIZE;
45 45 : buf->f_namelen = NAME_MAX;
46 45 : return 0;
47 : }
48 : EXPORT_SYMBOL(simple_statfs);
49 :
50 : /*
51 : * Retaining negative dentries for an in-memory filesystem just wastes
52 : * memory and lookup time: arrange for them to be deleted immediately.
53 : */
54 6994 : int always_delete_dentry(const struct dentry *dentry)
55 : {
56 6994 : return 1;
57 : }
58 : EXPORT_SYMBOL(always_delete_dentry);
59 :
60 : const struct dentry_operations simple_dentry_operations = {
61 : .d_delete = always_delete_dentry,
62 : };
63 : EXPORT_SYMBOL(simple_dentry_operations);
64 :
65 : /*
66 : * Lookup the data. This is trivial - if the dentry didn't already
67 : * exist, we know it is negative. Set d_op to delete negative dentries.
68 : */
69 13440 : struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
70 : {
71 13440 : if (dentry->d_name.len > NAME_MAX)
72 13440 : return ERR_PTR(-ENAMETOOLONG);
73 13440 : if (!dentry->d_sb->s_d_op)
74 11372 : d_set_d_op(dentry, &simple_dentry_operations);
75 13440 : d_add(dentry, NULL);
76 13440 : return NULL;
77 : }
78 : EXPORT_SYMBOL(simple_lookup);
79 :
80 288 : int dcache_dir_open(struct inode *inode, struct file *file)
81 : {
82 288 : file->private_data = d_alloc_cursor(file->f_path.dentry);
83 :
84 288 : return file->private_data ? 0 : -ENOMEM;
85 : }
86 : EXPORT_SYMBOL(dcache_dir_open);
87 :
88 284 : int dcache_dir_close(struct inode *inode, struct file *file)
89 : {
90 284 : dput(file->private_data);
91 284 : return 0;
92 : }
93 : EXPORT_SYMBOL(dcache_dir_close);
94 :
95 : /* parent is locked at least shared */
96 : /*
97 : * Returns an element of siblings' list.
98 : * We are looking for <count>th positive after <p>; if
99 : * found, dentry is grabbed and returned to caller.
100 : * If no such element exists, NULL is returned.
101 : */
102 3233 : static struct dentry *scan_positives(struct dentry *cursor,
103 : struct list_head *p,
104 : loff_t count,
105 : struct dentry *last)
106 : {
107 3233 : struct dentry *dentry = cursor->d_parent, *found = NULL;
108 :
109 3233 : spin_lock(&dentry->d_lock);
110 3233 : while ((p = p->next) != &dentry->d_subdirs) {
111 3054 : struct dentry *d = list_entry(p, struct dentry, d_child);
112 : // we must at least skip cursors, to avoid livelocks
113 3054 : if (d->d_flags & DCACHE_DENTRY_CURSOR)
114 0 : continue;
115 6108 : if (simple_positive(d) && !--count) {
116 3054 : spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
117 3054 : if (simple_positive(d))
118 3054 : found = dget_dlock(d);
119 3054 : spin_unlock(&d->d_lock);
120 3054 : if (likely(found))
121 : break;
122 : count = 1;
123 : }
124 0 : if (need_resched()) {
125 0 : list_move(&cursor->d_child, p);
126 0 : p = &cursor->d_child;
127 0 : spin_unlock(&dentry->d_lock);
128 0 : cond_resched();
129 3233 : spin_lock(&dentry->d_lock);
130 : }
131 : }
132 3233 : spin_unlock(&dentry->d_lock);
133 3233 : dput(last);
134 3233 : return found;
135 : }
136 :
137 0 : loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
138 : {
139 0 : struct dentry *dentry = file->f_path.dentry;
140 0 : switch (whence) {
141 0 : case 1:
142 0 : offset += file->f_pos;
143 0 : fallthrough;
144 0 : case 0:
145 0 : if (offset >= 0)
146 : break;
147 : fallthrough;
148 : default:
149 : return -EINVAL;
150 : }
151 0 : if (offset != file->f_pos) {
152 0 : struct dentry *cursor = file->private_data;
153 0 : struct dentry *to = NULL;
154 :
155 0 : inode_lock_shared(dentry->d_inode);
156 :
157 0 : if (offset > 2)
158 0 : to = scan_positives(cursor, &dentry->d_subdirs,
159 : offset - 2, NULL);
160 0 : spin_lock(&dentry->d_lock);
161 0 : if (to)
162 0 : list_move(&cursor->d_child, &to->d_child);
163 : else
164 0 : list_del_init(&cursor->d_child);
165 0 : spin_unlock(&dentry->d_lock);
166 0 : dput(to);
167 :
168 0 : file->f_pos = offset;
169 :
170 0 : inode_unlock_shared(dentry->d_inode);
171 : }
172 : return offset;
173 : }
174 : EXPORT_SYMBOL(dcache_dir_lseek);
175 :
176 : /* Relationship between i_mode and the DT_xxx types */
177 3054 : static inline unsigned char dt_type(struct inode *inode)
178 : {
179 3054 : return (inode->i_mode >> 12) & 15;
180 : }
181 :
182 : /*
183 : * Directory is locked and all positive dentries in it are safe, since
184 : * for ramfs-type trees they can't go away without unlink() or rmdir(),
185 : * both impossible due to the lock on directory.
186 : */
187 :
188 334 : int dcache_readdir(struct file *file, struct dir_context *ctx)
189 : {
190 334 : struct dentry *dentry = file->f_path.dentry;
191 334 : struct dentry *cursor = file->private_data;
192 334 : struct list_head *anchor = &dentry->d_subdirs;
193 334 : struct dentry *next = NULL;
194 334 : struct list_head *p;
195 :
196 334 : if (!dir_emit_dots(file, ctx))
197 : return 0;
198 :
199 334 : if (ctx->pos == 2)
200 : p = anchor;
201 155 : else if (!list_empty(&cursor->d_child))
202 : p = &cursor->d_child;
203 : else
204 : return 0;
205 :
206 3233 : while ((next = scan_positives(cursor, p, 1, next)) != NULL) {
207 3054 : if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
208 3054 : d_inode(next)->i_ino, dt_type(d_inode(next))))
209 : break;
210 3054 : ctx->pos++;
211 3054 : p = &next->d_child;
212 : }
213 179 : spin_lock(&dentry->d_lock);
214 179 : if (next)
215 0 : list_move_tail(&cursor->d_child, &next->d_child);
216 : else
217 179 : list_del_init(&cursor->d_child);
218 179 : spin_unlock(&dentry->d_lock);
219 179 : dput(next);
220 :
221 179 : return 0;
222 : }
223 : EXPORT_SYMBOL(dcache_readdir);
224 :
225 0 : ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
226 : {
227 0 : return -EISDIR;
228 : }
229 : EXPORT_SYMBOL(generic_read_dir);
230 :
231 : const struct file_operations simple_dir_operations = {
232 : .open = dcache_dir_open,
233 : .release = dcache_dir_close,
234 : .llseek = dcache_dir_lseek,
235 : .read = generic_read_dir,
236 : .iterate_shared = dcache_readdir,
237 : .fsync = noop_fsync,
238 : };
239 : EXPORT_SYMBOL(simple_dir_operations);
240 :
241 : const struct inode_operations simple_dir_inode_operations = {
242 : .lookup = simple_lookup,
243 : };
244 : EXPORT_SYMBOL(simple_dir_inode_operations);
245 :
246 0 : static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev)
247 : {
248 0 : struct dentry *child = NULL;
249 0 : struct list_head *p = prev ? &prev->d_child : &parent->d_subdirs;
250 :
251 0 : spin_lock(&parent->d_lock);
252 0 : while ((p = p->next) != &parent->d_subdirs) {
253 0 : struct dentry *d = container_of(p, struct dentry, d_child);
254 0 : if (simple_positive(d)) {
255 0 : spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
256 0 : if (simple_positive(d))
257 0 : child = dget_dlock(d);
258 0 : spin_unlock(&d->d_lock);
259 0 : if (likely(child))
260 : break;
261 : }
262 : }
263 0 : spin_unlock(&parent->d_lock);
264 0 : dput(prev);
265 0 : return child;
266 : }
267 :
268 0 : void simple_recursive_removal(struct dentry *dentry,
269 : void (*callback)(struct dentry *))
270 : {
271 0 : struct dentry *this = dget(dentry);
272 0 : while (true) {
273 0 : struct dentry *victim = NULL, *child;
274 0 : struct inode *inode = this->d_inode;
275 :
276 0 : inode_lock(inode);
277 0 : if (d_is_dir(this))
278 0 : inode->i_flags |= S_DEAD;
279 0 : while ((child = find_next_child(this, victim)) == NULL) {
280 : // kill and ascend
281 : // update metadata while it's still locked
282 0 : inode->i_ctime = current_time(inode);
283 0 : clear_nlink(inode);
284 0 : inode_unlock(inode);
285 0 : victim = this;
286 0 : this = this->d_parent;
287 0 : inode = this->d_inode;
288 0 : inode_lock(inode);
289 0 : if (simple_positive(victim)) {
290 0 : d_invalidate(victim); // avoid lost mounts
291 0 : if (d_is_dir(victim))
292 0 : fsnotify_rmdir(inode, victim);
293 : else
294 0 : fsnotify_unlink(inode, victim);
295 0 : if (callback)
296 0 : callback(victim);
297 0 : dput(victim); // unpin it
298 : }
299 0 : if (victim == dentry) {
300 0 : inode->i_ctime = inode->i_mtime =
301 0 : current_time(inode);
302 0 : if (d_is_dir(dentry))
303 0 : drop_nlink(inode);
304 0 : inode_unlock(inode);
305 0 : dput(dentry);
306 0 : return;
307 : }
308 : }
309 0 : inode_unlock(inode);
310 0 : this = child;
311 : }
312 : }
313 : EXPORT_SYMBOL(simple_recursive_removal);
314 :
315 : static const struct super_operations simple_super_operations = {
316 : .statfs = simple_statfs,
317 : };
318 :
319 6 : static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc)
320 : {
321 6 : struct pseudo_fs_context *ctx = fc->fs_private;
322 6 : struct inode *root;
323 :
324 6 : s->s_maxbytes = MAX_LFS_FILESIZE;
325 6 : s->s_blocksize = PAGE_SIZE;
326 6 : s->s_blocksize_bits = PAGE_SHIFT;
327 6 : s->s_magic = ctx->magic;
328 6 : s->s_op = ctx->ops ?: &simple_super_operations;
329 6 : s->s_xattr = ctx->xattr;
330 6 : s->s_time_gran = 1;
331 6 : root = new_inode(s);
332 6 : if (!root)
333 : return -ENOMEM;
334 :
335 : /*
336 : * since this is the first inode, make it number 1. New inodes created
337 : * after this must take care not to collide with it (by passing
338 : * max_reserved of 1 to iunique).
339 : */
340 6 : root->i_ino = 1;
341 6 : root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
342 6 : root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
343 6 : s->s_root = d_make_root(root);
344 6 : if (!s->s_root)
345 : return -ENOMEM;
346 6 : s->s_d_op = ctx->dops;
347 6 : return 0;
348 : }
349 :
350 6 : static int pseudo_fs_get_tree(struct fs_context *fc)
351 : {
352 6 : return get_tree_nodev(fc, pseudo_fs_fill_super);
353 : }
354 :
355 6 : static void pseudo_fs_free(struct fs_context *fc)
356 : {
357 6 : kfree(fc->fs_private);
358 6 : }
359 :
360 : static const struct fs_context_operations pseudo_fs_context_ops = {
361 : .free = pseudo_fs_free,
362 : .get_tree = pseudo_fs_get_tree,
363 : };
364 :
365 : /*
366 : * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
367 : * will never be mountable)
368 : */
369 6 : struct pseudo_fs_context *init_pseudo(struct fs_context *fc,
370 : unsigned long magic)
371 : {
372 6 : struct pseudo_fs_context *ctx;
373 :
374 6 : ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL);
375 6 : if (likely(ctx)) {
376 6 : ctx->magic = magic;
377 6 : fc->fs_private = ctx;
378 6 : fc->ops = &pseudo_fs_context_ops;
379 6 : fc->sb_flags |= SB_NOUSER;
380 6 : fc->global = true;
381 : }
382 6 : return ctx;
383 : }
384 : EXPORT_SYMBOL(init_pseudo);
385 :
386 0 : int simple_open(struct inode *inode, struct file *file)
387 : {
388 0 : if (inode->i_private)
389 0 : file->private_data = inode->i_private;
390 0 : return 0;
391 : }
392 : EXPORT_SYMBOL(simple_open);
393 :
394 0 : int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
395 : {
396 0 : struct inode *inode = d_inode(old_dentry);
397 :
398 0 : inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
399 0 : inc_nlink(inode);
400 0 : ihold(inode);
401 0 : dget(dentry);
402 0 : d_instantiate(dentry, inode);
403 0 : return 0;
404 : }
405 : EXPORT_SYMBOL(simple_link);
406 :
407 1042 : int simple_empty(struct dentry *dentry)
408 : {
409 1042 : struct dentry *child;
410 1042 : int ret = 0;
411 :
412 1042 : spin_lock(&dentry->d_lock);
413 1071 : list_for_each_entry(child, &dentry->d_subdirs, d_child) {
414 361 : spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
415 361 : if (simple_positive(child)) {
416 332 : spin_unlock(&child->d_lock);
417 332 : goto out;
418 : }
419 29 : spin_unlock(&child->d_lock);
420 : }
421 : ret = 1;
422 1042 : out:
423 1042 : spin_unlock(&dentry->d_lock);
424 1042 : return ret;
425 : }
426 : EXPORT_SYMBOL(simple_empty);
427 :
428 0 : int simple_unlink(struct inode *dir, struct dentry *dentry)
429 : {
430 0 : struct inode *inode = d_inode(dentry);
431 :
432 0 : inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
433 0 : drop_nlink(inode);
434 0 : dput(dentry);
435 0 : return 0;
436 : }
437 : EXPORT_SYMBOL(simple_unlink);
438 :
439 0 : int simple_rmdir(struct inode *dir, struct dentry *dentry)
440 : {
441 0 : if (!simple_empty(dentry))
442 : return -ENOTEMPTY;
443 :
444 0 : drop_nlink(d_inode(dentry));
445 0 : simple_unlink(dir, dentry);
446 0 : drop_nlink(dir);
447 0 : return 0;
448 : }
449 : EXPORT_SYMBOL(simple_rmdir);
450 :
451 0 : int simple_rename(struct user_namespace *mnt_userns, struct inode *old_dir,
452 : struct dentry *old_dentry, struct inode *new_dir,
453 : struct dentry *new_dentry, unsigned int flags)
454 : {
455 0 : struct inode *inode = d_inode(old_dentry);
456 0 : int they_are_dirs = d_is_dir(old_dentry);
457 :
458 0 : if (flags & ~RENAME_NOREPLACE)
459 : return -EINVAL;
460 :
461 0 : if (!simple_empty(new_dentry))
462 : return -ENOTEMPTY;
463 :
464 0 : if (d_really_is_positive(new_dentry)) {
465 0 : simple_unlink(new_dir, new_dentry);
466 0 : if (they_are_dirs) {
467 0 : drop_nlink(d_inode(new_dentry));
468 0 : drop_nlink(old_dir);
469 : }
470 0 : } else if (they_are_dirs) {
471 0 : drop_nlink(old_dir);
472 0 : inc_nlink(new_dir);
473 : }
474 :
475 0 : old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
476 0 : new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
477 :
478 0 : return 0;
479 : }
480 : EXPORT_SYMBOL(simple_rename);
481 :
482 : /**
483 : * simple_setattr - setattr for simple filesystem
484 : * @dentry: dentry
485 : * @iattr: iattr structure
486 : *
487 : * Returns 0 on success, -error on failure.
488 : *
489 : * simple_setattr is a simple ->setattr implementation without a proper
490 : * implementation of size changes.
491 : *
492 : * It can either be used for in-memory filesystems or special files
493 : * on simple regular filesystems. Anything that needs to change on-disk
494 : * or wire state on size changes needs its own setattr method.
495 : */
496 505 : int simple_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
497 : struct iattr *iattr)
498 : {
499 505 : struct inode *inode = d_inode(dentry);
500 505 : int error;
501 :
502 505 : error = setattr_prepare(mnt_userns, dentry, iattr);
503 505 : if (error)
504 : return error;
505 :
506 505 : if (iattr->ia_valid & ATTR_SIZE)
507 0 : truncate_setsize(inode, iattr->ia_size);
508 505 : setattr_copy(mnt_userns, inode, iattr);
509 505 : mark_inode_dirty(inode);
510 505 : return 0;
511 : }
512 : EXPORT_SYMBOL(simple_setattr);
513 :
514 0 : int simple_readpage(struct file *file, struct page *page)
515 : {
516 0 : clear_highpage(page);
517 0 : flush_dcache_page(page);
518 0 : SetPageUptodate(page);
519 0 : unlock_page(page);
520 0 : return 0;
521 : }
522 : EXPORT_SYMBOL(simple_readpage);
523 :
524 0 : int simple_write_begin(struct file *file, struct address_space *mapping,
525 : loff_t pos, unsigned len, unsigned flags,
526 : struct page **pagep, void **fsdata)
527 : {
528 0 : struct page *page;
529 0 : pgoff_t index;
530 :
531 0 : index = pos >> PAGE_SHIFT;
532 :
533 0 : page = grab_cache_page_write_begin(mapping, index, flags);
534 0 : if (!page)
535 : return -ENOMEM;
536 :
537 0 : *pagep = page;
538 :
539 0 : if (!PageUptodate(page) && (len != PAGE_SIZE)) {
540 0 : unsigned from = pos & (PAGE_SIZE - 1);
541 :
542 0 : zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
543 : }
544 : return 0;
545 : }
546 : EXPORT_SYMBOL(simple_write_begin);
547 :
548 : /**
549 : * simple_write_end - .write_end helper for non-block-device FSes
550 : * @file: See .write_end of address_space_operations
551 : * @mapping: "
552 : * @pos: "
553 : * @len: "
554 : * @copied: "
555 : * @page: "
556 : * @fsdata: "
557 : *
558 : * simple_write_end does the minimum needed for updating a page after writing is
559 : * done. It has the same API signature as the .write_end of
560 : * address_space_operations vector. So it can just be set onto .write_end for
561 : * FSes that don't need any other processing. i_mutex is assumed to be held.
562 : * Block based filesystems should use generic_write_end().
563 : * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
564 : * is not called, so a filesystem that actually does store data in .write_inode
565 : * should extend on what's done here with a call to mark_inode_dirty() in the
566 : * case that i_size has changed.
567 : *
568 : * Use *ONLY* with simple_readpage()
569 : */
570 0 : int simple_write_end(struct file *file, struct address_space *mapping,
571 : loff_t pos, unsigned len, unsigned copied,
572 : struct page *page, void *fsdata)
573 : {
574 0 : struct inode *inode = page->mapping->host;
575 0 : loff_t last_pos = pos + copied;
576 :
577 : /* zero the stale part of the page if we did a short copy */
578 0 : if (!PageUptodate(page)) {
579 0 : if (copied < len) {
580 0 : unsigned from = pos & (PAGE_SIZE - 1);
581 :
582 0 : zero_user(page, from + copied, len - copied);
583 : }
584 0 : SetPageUptodate(page);
585 : }
586 : /*
587 : * No need to use i_size_read() here, the i_size
588 : * cannot change under us because we hold the i_mutex.
589 : */
590 0 : if (last_pos > inode->i_size)
591 0 : i_size_write(inode, last_pos);
592 :
593 0 : set_page_dirty(page);
594 0 : unlock_page(page);
595 0 : put_page(page);
596 :
597 0 : return copied;
598 : }
599 : EXPORT_SYMBOL(simple_write_end);
600 :
601 : /*
602 : * the inodes created here are not hashed. If you use iunique to generate
603 : * unique inode values later for this filesystem, then you must take care
604 : * to pass it an appropriate max_reserved value to avoid collisions.
605 : */
606 2 : int simple_fill_super(struct super_block *s, unsigned long magic,
607 : const struct tree_descr *files)
608 : {
609 2 : struct inode *inode;
610 2 : struct dentry *root;
611 2 : struct dentry *dentry;
612 2 : int i;
613 :
614 2 : s->s_blocksize = PAGE_SIZE;
615 2 : s->s_blocksize_bits = PAGE_SHIFT;
616 2 : s->s_magic = magic;
617 2 : s->s_op = &simple_super_operations;
618 2 : s->s_time_gran = 1;
619 :
620 2 : inode = new_inode(s);
621 2 : if (!inode)
622 : return -ENOMEM;
623 : /*
624 : * because the root inode is 1, the files array must not contain an
625 : * entry at index 1
626 : */
627 2 : inode->i_ino = 1;
628 2 : inode->i_mode = S_IFDIR | 0755;
629 2 : inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
630 2 : inode->i_op = &simple_dir_inode_operations;
631 2 : inode->i_fop = &simple_dir_operations;
632 2 : set_nlink(inode, 2);
633 2 : root = d_make_root(inode);
634 2 : if (!root)
635 : return -ENOMEM;
636 2 : for (i = 0; !files->name || files->name[0]; i++, files++) {
637 0 : if (!files->name)
638 0 : continue;
639 :
640 : /* warn if it tries to conflict with the root inode */
641 0 : if (unlikely(i == 1))
642 0 : printk(KERN_WARNING "%s: %s passed in a files array"
643 : "with an index of 1!\n", __func__,
644 0 : s->s_type->name);
645 :
646 0 : dentry = d_alloc_name(root, files->name);
647 0 : if (!dentry)
648 0 : goto out;
649 0 : inode = new_inode(s);
650 0 : if (!inode) {
651 0 : dput(dentry);
652 0 : goto out;
653 : }
654 0 : inode->i_mode = S_IFREG | files->mode;
655 0 : inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
656 0 : inode->i_fop = files->ops;
657 0 : inode->i_ino = i;
658 0 : d_add(dentry, inode);
659 : }
660 2 : s->s_root = root;
661 2 : return 0;
662 0 : out:
663 0 : d_genocide(root);
664 0 : shrink_dcache_parent(root);
665 0 : dput(root);
666 0 : return -ENOMEM;
667 : }
668 : EXPORT_SYMBOL(simple_fill_super);
669 :
670 : static DEFINE_SPINLOCK(pin_fs_lock);
671 :
672 5519 : int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
673 : {
674 5519 : struct vfsmount *mnt = NULL;
675 5519 : spin_lock(&pin_fs_lock);
676 5519 : if (unlikely(!*mount)) {
677 3 : spin_unlock(&pin_fs_lock);
678 3 : mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
679 3 : if (IS_ERR(mnt))
680 0 : return PTR_ERR(mnt);
681 3 : spin_lock(&pin_fs_lock);
682 3 : if (!*mount)
683 3 : *mount = mnt;
684 : }
685 5519 : mntget(*mount);
686 5519 : ++*count;
687 5519 : spin_unlock(&pin_fs_lock);
688 5519 : mntput(mnt);
689 5519 : return 0;
690 : }
691 : EXPORT_SYMBOL(simple_pin_fs);
692 :
693 0 : void simple_release_fs(struct vfsmount **mount, int *count)
694 : {
695 0 : struct vfsmount *mnt;
696 0 : spin_lock(&pin_fs_lock);
697 0 : mnt = *mount;
698 0 : if (!--*count)
699 0 : *mount = NULL;
700 0 : spin_unlock(&pin_fs_lock);
701 0 : mntput(mnt);
702 0 : }
703 : EXPORT_SYMBOL(simple_release_fs);
704 :
705 : /**
706 : * simple_read_from_buffer - copy data from the buffer to user space
707 : * @to: the user space buffer to read to
708 : * @count: the maximum number of bytes to read
709 : * @ppos: the current position in the buffer
710 : * @from: the buffer to read from
711 : * @available: the size of the buffer
712 : *
713 : * The simple_read_from_buffer() function reads up to @count bytes from the
714 : * buffer @from at offset @ppos into the user space address starting at @to.
715 : *
716 : * On success, the number of bytes read is returned and the offset @ppos is
717 : * advanced by this number, or negative value is returned on error.
718 : **/
719 3 : ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
720 : const void *from, size_t available)
721 : {
722 3 : loff_t pos = *ppos;
723 3 : size_t ret;
724 :
725 3 : if (pos < 0)
726 : return -EINVAL;
727 3 : if (pos >= available || !count)
728 : return 0;
729 2 : if (count > available - pos)
730 : count = available - pos;
731 2 : ret = copy_to_user(to, from + pos, count);
732 2 : if (ret == count)
733 : return -EFAULT;
734 2 : count -= ret;
735 2 : *ppos = pos + count;
736 2 : return count;
737 : }
738 : EXPORT_SYMBOL(simple_read_from_buffer);
739 :
740 : /**
741 : * simple_write_to_buffer - copy data from user space to the buffer
742 : * @to: the buffer to write to
743 : * @available: the size of the buffer
744 : * @ppos: the current position in the buffer
745 : * @from: the user space buffer to read from
746 : * @count: the maximum number of bytes to read
747 : *
748 : * The simple_write_to_buffer() function reads up to @count bytes from the user
749 : * space address starting at @from into the buffer @to at offset @ppos.
750 : *
751 : * On success, the number of bytes written is returned and the offset @ppos is
752 : * advanced by this number, or negative value is returned on error.
753 : **/
754 0 : ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
755 : const void __user *from, size_t count)
756 : {
757 0 : loff_t pos = *ppos;
758 0 : size_t res;
759 :
760 0 : if (pos < 0)
761 : return -EINVAL;
762 0 : if (pos >= available || !count)
763 : return 0;
764 0 : if (count > available - pos)
765 : count = available - pos;
766 0 : res = copy_from_user(to + pos, from, count);
767 0 : if (res == count)
768 : return -EFAULT;
769 0 : count -= res;
770 0 : *ppos = pos + count;
771 0 : return count;
772 : }
773 : EXPORT_SYMBOL(simple_write_to_buffer);
774 :
775 : /**
776 : * memory_read_from_buffer - copy data from the buffer
777 : * @to: the kernel space buffer to read to
778 : * @count: the maximum number of bytes to read
779 : * @ppos: the current position in the buffer
780 : * @from: the buffer to read from
781 : * @available: the size of the buffer
782 : *
783 : * The memory_read_from_buffer() function reads up to @count bytes from the
784 : * buffer @from at offset @ppos into the kernel space address starting at @to.
785 : *
786 : * On success, the number of bytes read is returned and the offset @ppos is
787 : * advanced by this number, or negative value is returned on error.
788 : **/
789 0 : ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
790 : const void *from, size_t available)
791 : {
792 0 : loff_t pos = *ppos;
793 :
794 0 : if (pos < 0)
795 : return -EINVAL;
796 0 : if (pos >= available)
797 : return 0;
798 0 : if (count > available - pos)
799 : count = available - pos;
800 0 : memcpy(to, from + pos, count);
801 0 : *ppos = pos + count;
802 :
803 0 : return count;
804 : }
805 : EXPORT_SYMBOL(memory_read_from_buffer);
806 :
807 : /*
808 : * Transaction based IO.
809 : * The file expects a single write which triggers the transaction, and then
810 : * possibly a read which collects the result - which is stored in a
811 : * file-local buffer.
812 : */
813 :
814 0 : void simple_transaction_set(struct file *file, size_t n)
815 : {
816 0 : struct simple_transaction_argresp *ar = file->private_data;
817 :
818 0 : BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
819 :
820 : /*
821 : * The barrier ensures that ar->size will really remain zero until
822 : * ar->data is ready for reading.
823 : */
824 0 : smp_mb();
825 0 : ar->size = n;
826 0 : }
827 : EXPORT_SYMBOL(simple_transaction_set);
828 :
829 0 : char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
830 : {
831 0 : struct simple_transaction_argresp *ar;
832 0 : static DEFINE_SPINLOCK(simple_transaction_lock);
833 :
834 0 : if (size > SIMPLE_TRANSACTION_LIMIT - 1)
835 0 : return ERR_PTR(-EFBIG);
836 :
837 0 : ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
838 0 : if (!ar)
839 0 : return ERR_PTR(-ENOMEM);
840 :
841 0 : spin_lock(&simple_transaction_lock);
842 :
843 : /* only one write allowed per open */
844 0 : if (file->private_data) {
845 0 : spin_unlock(&simple_transaction_lock);
846 0 : free_page((unsigned long)ar);
847 0 : return ERR_PTR(-EBUSY);
848 : }
849 :
850 0 : file->private_data = ar;
851 :
852 0 : spin_unlock(&simple_transaction_lock);
853 :
854 0 : if (copy_from_user(ar->data, buf, size))
855 0 : return ERR_PTR(-EFAULT);
856 :
857 : return ar->data;
858 : }
859 : EXPORT_SYMBOL(simple_transaction_get);
860 :
861 0 : ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
862 : {
863 0 : struct simple_transaction_argresp *ar = file->private_data;
864 :
865 0 : if (!ar)
866 : return 0;
867 0 : return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
868 : }
869 : EXPORT_SYMBOL(simple_transaction_read);
870 :
871 0 : int simple_transaction_release(struct inode *inode, struct file *file)
872 : {
873 0 : free_page((unsigned long)file->private_data);
874 0 : return 0;
875 : }
876 : EXPORT_SYMBOL(simple_transaction_release);
877 :
878 : /* Simple attribute files */
879 :
880 : struct simple_attr {
881 : int (*get)(void *, u64 *);
882 : int (*set)(void *, u64);
883 : char get_buf[24]; /* enough to store a u64 and "\n\0" */
884 : char set_buf[24];
885 : void *data;
886 : const char *fmt; /* format for read operation */
887 : struct mutex mutex; /* protects access to these buffers */
888 : };
889 :
890 : /* simple_attr_open is called by an actual attribute open file operation
891 : * to set the attribute specific access operations. */
892 0 : int simple_attr_open(struct inode *inode, struct file *file,
893 : int (*get)(void *, u64 *), int (*set)(void *, u64),
894 : const char *fmt)
895 : {
896 0 : struct simple_attr *attr;
897 :
898 0 : attr = kzalloc(sizeof(*attr), GFP_KERNEL);
899 0 : if (!attr)
900 : return -ENOMEM;
901 :
902 0 : attr->get = get;
903 0 : attr->set = set;
904 0 : attr->data = inode->i_private;
905 0 : attr->fmt = fmt;
906 0 : mutex_init(&attr->mutex);
907 :
908 0 : file->private_data = attr;
909 :
910 0 : return nonseekable_open(inode, file);
911 : }
912 : EXPORT_SYMBOL_GPL(simple_attr_open);
913 :
914 0 : int simple_attr_release(struct inode *inode, struct file *file)
915 : {
916 0 : kfree(file->private_data);
917 0 : return 0;
918 : }
919 : EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
920 :
921 : /* read from the buffer that is filled with the get function */
922 0 : ssize_t simple_attr_read(struct file *file, char __user *buf,
923 : size_t len, loff_t *ppos)
924 : {
925 0 : struct simple_attr *attr;
926 0 : size_t size;
927 0 : ssize_t ret;
928 :
929 0 : attr = file->private_data;
930 :
931 0 : if (!attr->get)
932 : return -EACCES;
933 :
934 0 : ret = mutex_lock_interruptible(&attr->mutex);
935 0 : if (ret)
936 : return ret;
937 :
938 0 : if (*ppos && attr->get_buf[0]) {
939 : /* continued read */
940 0 : size = strlen(attr->get_buf);
941 : } else {
942 : /* first read */
943 0 : u64 val;
944 0 : ret = attr->get(attr->data, &val);
945 0 : if (ret)
946 0 : goto out;
947 :
948 0 : size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
949 : attr->fmt, (unsigned long long)val);
950 : }
951 :
952 0 : ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
953 0 : out:
954 0 : mutex_unlock(&attr->mutex);
955 0 : return ret;
956 : }
957 : EXPORT_SYMBOL_GPL(simple_attr_read);
958 :
959 : /* interpret the buffer as a number to call the set function with */
960 0 : ssize_t simple_attr_write(struct file *file, const char __user *buf,
961 : size_t len, loff_t *ppos)
962 : {
963 0 : struct simple_attr *attr;
964 0 : unsigned long long val;
965 0 : size_t size;
966 0 : ssize_t ret;
967 :
968 0 : attr = file->private_data;
969 0 : if (!attr->set)
970 : return -EACCES;
971 :
972 0 : ret = mutex_lock_interruptible(&attr->mutex);
973 0 : if (ret)
974 : return ret;
975 :
976 0 : ret = -EFAULT;
977 0 : size = min(sizeof(attr->set_buf) - 1, len);
978 0 : if (copy_from_user(attr->set_buf, buf, size))
979 0 : goto out;
980 :
981 0 : attr->set_buf[size] = '\0';
982 0 : ret = kstrtoull(attr->set_buf, 0, &val);
983 0 : if (ret)
984 0 : goto out;
985 0 : ret = attr->set(attr->data, val);
986 0 : if (ret == 0)
987 0 : ret = len; /* on success, claim we got the whole input */
988 0 : out:
989 0 : mutex_unlock(&attr->mutex);
990 0 : return ret;
991 : }
992 : EXPORT_SYMBOL_GPL(simple_attr_write);
993 :
994 : /**
995 : * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
996 : * @sb: filesystem to do the file handle conversion on
997 : * @fid: file handle to convert
998 : * @fh_len: length of the file handle in bytes
999 : * @fh_type: type of file handle
1000 : * @get_inode: filesystem callback to retrieve inode
1001 : *
1002 : * This function decodes @fid as long as it has one of the well-known
1003 : * Linux filehandle types and calls @get_inode on it to retrieve the
1004 : * inode for the object specified in the file handle.
1005 : */
1006 0 : struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
1007 : int fh_len, int fh_type, struct inode *(*get_inode)
1008 : (struct super_block *sb, u64 ino, u32 gen))
1009 : {
1010 0 : struct inode *inode = NULL;
1011 :
1012 0 : if (fh_len < 2)
1013 : return NULL;
1014 :
1015 0 : switch (fh_type) {
1016 0 : case FILEID_INO32_GEN:
1017 : case FILEID_INO32_GEN_PARENT:
1018 0 : inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
1019 0 : break;
1020 : }
1021 :
1022 0 : return d_obtain_alias(inode);
1023 : }
1024 : EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
1025 :
1026 : /**
1027 : * generic_fh_to_parent - generic helper for the fh_to_parent export operation
1028 : * @sb: filesystem to do the file handle conversion on
1029 : * @fid: file handle to convert
1030 : * @fh_len: length of the file handle in bytes
1031 : * @fh_type: type of file handle
1032 : * @get_inode: filesystem callback to retrieve inode
1033 : *
1034 : * This function decodes @fid as long as it has one of the well-known
1035 : * Linux filehandle types and calls @get_inode on it to retrieve the
1036 : * inode for the _parent_ object specified in the file handle if it
1037 : * is specified in the file handle, or NULL otherwise.
1038 : */
1039 0 : struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
1040 : int fh_len, int fh_type, struct inode *(*get_inode)
1041 : (struct super_block *sb, u64 ino, u32 gen))
1042 : {
1043 0 : struct inode *inode = NULL;
1044 :
1045 0 : if (fh_len <= 2)
1046 : return NULL;
1047 :
1048 0 : switch (fh_type) {
1049 0 : case FILEID_INO32_GEN_PARENT:
1050 0 : inode = get_inode(sb, fid->i32.parent_ino,
1051 : (fh_len > 3 ? fid->i32.parent_gen : 0));
1052 0 : break;
1053 : }
1054 :
1055 0 : return d_obtain_alias(inode);
1056 : }
1057 : EXPORT_SYMBOL_GPL(generic_fh_to_parent);
1058 :
1059 : /**
1060 : * __generic_file_fsync - generic fsync implementation for simple filesystems
1061 : *
1062 : * @file: file to synchronize
1063 : * @start: start offset in bytes
1064 : * @end: end offset in bytes (inclusive)
1065 : * @datasync: only synchronize essential metadata if true
1066 : *
1067 : * This is a generic implementation of the fsync method for simple
1068 : * filesystems which track all non-inode metadata in the buffers list
1069 : * hanging off the address_space structure.
1070 : */
1071 0 : int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
1072 : int datasync)
1073 : {
1074 0 : struct inode *inode = file->f_mapping->host;
1075 0 : int err;
1076 0 : int ret;
1077 :
1078 0 : err = file_write_and_wait_range(file, start, end);
1079 0 : if (err)
1080 : return err;
1081 :
1082 0 : inode_lock(inode);
1083 0 : ret = sync_mapping_buffers(inode->i_mapping);
1084 0 : if (!(inode->i_state & I_DIRTY_ALL))
1085 0 : goto out;
1086 0 : if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
1087 0 : goto out;
1088 :
1089 0 : err = sync_inode_metadata(inode, 1);
1090 0 : if (ret == 0)
1091 0 : ret = err;
1092 :
1093 0 : out:
1094 0 : inode_unlock(inode);
1095 : /* check and advance again to catch errors after syncing out buffers */
1096 0 : err = file_check_and_advance_wb_err(file);
1097 0 : if (ret == 0)
1098 0 : ret = err;
1099 : return ret;
1100 : }
1101 : EXPORT_SYMBOL(__generic_file_fsync);
1102 :
1103 : /**
1104 : * generic_file_fsync - generic fsync implementation for simple filesystems
1105 : * with flush
1106 : * @file: file to synchronize
1107 : * @start: start offset in bytes
1108 : * @end: end offset in bytes (inclusive)
1109 : * @datasync: only synchronize essential metadata if true
1110 : *
1111 : */
1112 :
1113 0 : int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1114 : int datasync)
1115 : {
1116 0 : struct inode *inode = file->f_mapping->host;
1117 0 : int err;
1118 :
1119 0 : err = __generic_file_fsync(file, start, end, datasync);
1120 0 : if (err)
1121 : return err;
1122 0 : return blkdev_issue_flush(inode->i_sb->s_bdev);
1123 : }
1124 : EXPORT_SYMBOL(generic_file_fsync);
1125 :
1126 : /**
1127 : * generic_check_addressable - Check addressability of file system
1128 : * @blocksize_bits: log of file system block size
1129 : * @num_blocks: number of blocks in file system
1130 : *
1131 : * Determine whether a file system with @num_blocks blocks (and a
1132 : * block size of 2**@blocksize_bits) is addressable by the sector_t
1133 : * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1134 : */
1135 1 : int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1136 : {
1137 1 : u64 last_fs_block = num_blocks - 1;
1138 1 : u64 last_fs_page =
1139 1 : last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1140 :
1141 1 : if (unlikely(num_blocks == 0))
1142 : return 0;
1143 :
1144 1 : if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1145 : return -EINVAL;
1146 :
1147 1 : if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1148 : (last_fs_page > (pgoff_t)(~0ULL))) {
1149 0 : return -EFBIG;
1150 : }
1151 : return 0;
1152 : }
1153 : EXPORT_SYMBOL(generic_check_addressable);
1154 :
1155 : /*
1156 : * No-op implementation of ->fsync for in-memory filesystems.
1157 : */
1158 3 : int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1159 : {
1160 3 : return 0;
1161 : }
1162 : EXPORT_SYMBOL(noop_fsync);
1163 :
1164 0 : int noop_set_page_dirty(struct page *page)
1165 : {
1166 : /*
1167 : * Unlike __set_page_dirty_no_writeback that handles dirty page
1168 : * tracking in the page object, dax does all dirty tracking in
1169 : * the inode address_space in response to mkwrite faults. In the
1170 : * dax case we only need to worry about potentially dirty CPU
1171 : * caches, not dirty page cache pages to write back.
1172 : *
1173 : * This callback is defined to prevent fallback to
1174 : * __set_page_dirty_buffers() in set_page_dirty().
1175 : */
1176 0 : return 0;
1177 : }
1178 : EXPORT_SYMBOL_GPL(noop_set_page_dirty);
1179 :
1180 0 : void noop_invalidatepage(struct page *page, unsigned int offset,
1181 : unsigned int length)
1182 : {
1183 : /*
1184 : * There is no page cache to invalidate in the dax case, however
1185 : * we need this callback defined to prevent falling back to
1186 : * block_invalidatepage() in do_invalidatepage().
1187 : */
1188 0 : }
1189 : EXPORT_SYMBOL_GPL(noop_invalidatepage);
1190 :
1191 0 : ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1192 : {
1193 : /*
1194 : * iomap based filesystems support direct I/O without need for
1195 : * this callback. However, it still needs to be set in
1196 : * inode->a_ops so that open/fcntl know that direct I/O is
1197 : * generally supported.
1198 : */
1199 0 : return -EINVAL;
1200 : }
1201 : EXPORT_SYMBOL_GPL(noop_direct_IO);
1202 :
1203 : /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1204 1729 : void kfree_link(void *p)
1205 : {
1206 1729 : kfree(p);
1207 1729 : }
1208 : EXPORT_SYMBOL(kfree_link);
1209 :
1210 : /*
1211 : * nop .set_page_dirty method so that people can use .page_mkwrite on
1212 : * anon inodes.
1213 : */
1214 0 : static int anon_set_page_dirty(struct page *page)
1215 : {
1216 0 : return 0;
1217 : };
1218 :
1219 2 : struct inode *alloc_anon_inode(struct super_block *s)
1220 : {
1221 2 : static const struct address_space_operations anon_aops = {
1222 : .set_page_dirty = anon_set_page_dirty,
1223 : };
1224 2 : struct inode *inode = new_inode_pseudo(s);
1225 :
1226 2 : if (!inode)
1227 2 : return ERR_PTR(-ENOMEM);
1228 :
1229 2 : inode->i_ino = get_next_ino();
1230 2 : inode->i_mapping->a_ops = &anon_aops;
1231 :
1232 : /*
1233 : * Mark the inode dirty from the very beginning,
1234 : * that way it will never be moved to the dirty
1235 : * list because mark_inode_dirty() will think
1236 : * that it already _is_ on the dirty list.
1237 : */
1238 2 : inode->i_state = I_DIRTY;
1239 2 : inode->i_mode = S_IRUSR | S_IWUSR;
1240 2 : inode->i_uid = current_fsuid();
1241 2 : inode->i_gid = current_fsgid();
1242 2 : inode->i_flags |= S_PRIVATE;
1243 2 : inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1244 2 : return inode;
1245 : }
1246 : EXPORT_SYMBOL(alloc_anon_inode);
1247 :
1248 : /**
1249 : * simple_nosetlease - generic helper for prohibiting leases
1250 : * @filp: file pointer
1251 : * @arg: type of lease to obtain
1252 : * @flp: new lease supplied for insertion
1253 : * @priv: private data for lm_setup operation
1254 : *
1255 : * Generic helper for filesystems that do not wish to allow leases to be set.
1256 : * All arguments are ignored and it just returns -EINVAL.
1257 : */
1258 : int
1259 0 : simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1260 : void **priv)
1261 : {
1262 0 : return -EINVAL;
1263 : }
1264 : EXPORT_SYMBOL(simple_nosetlease);
1265 :
1266 : /**
1267 : * simple_get_link - generic helper to get the target of "fast" symlinks
1268 : * @dentry: not used here
1269 : * @inode: the symlink inode
1270 : * @done: not used here
1271 : *
1272 : * Generic helper for filesystems to use for symlink inodes where a pointer to
1273 : * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1274 : * since as an optimization the path lookup code uses any non-NULL ->i_link
1275 : * directly, without calling ->get_link(). But ->get_link() still must be set,
1276 : * to mark the inode_operations as being for a symlink.
1277 : *
1278 : * Return: the symlink target
1279 : */
1280 0 : const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1281 : struct delayed_call *done)
1282 : {
1283 0 : return inode->i_link;
1284 : }
1285 : EXPORT_SYMBOL(simple_get_link);
1286 :
1287 : const struct inode_operations simple_symlink_inode_operations = {
1288 : .get_link = simple_get_link,
1289 : };
1290 : EXPORT_SYMBOL(simple_symlink_inode_operations);
1291 :
1292 : /*
1293 : * Operations for a permanently empty directory.
1294 : */
1295 0 : static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1296 : {
1297 0 : return ERR_PTR(-ENOENT);
1298 : }
1299 :
1300 9 : static int empty_dir_getattr(struct user_namespace *mnt_userns,
1301 : const struct path *path, struct kstat *stat,
1302 : u32 request_mask, unsigned int query_flags)
1303 : {
1304 9 : struct inode *inode = d_inode(path->dentry);
1305 9 : generic_fillattr(&init_user_ns, inode, stat);
1306 9 : return 0;
1307 : }
1308 :
1309 0 : static int empty_dir_setattr(struct user_namespace *mnt_userns,
1310 : struct dentry *dentry, struct iattr *attr)
1311 : {
1312 0 : return -EPERM;
1313 : }
1314 :
1315 0 : static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1316 : {
1317 0 : return -EOPNOTSUPP;
1318 : }
1319 :
1320 : static const struct inode_operations empty_dir_inode_operations = {
1321 : .lookup = empty_dir_lookup,
1322 : .permission = generic_permission,
1323 : .setattr = empty_dir_setattr,
1324 : .getattr = empty_dir_getattr,
1325 : .listxattr = empty_dir_listxattr,
1326 : };
1327 :
1328 0 : static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1329 : {
1330 : /* An empty directory has two entries . and .. at offsets 0 and 1 */
1331 0 : return generic_file_llseek_size(file, offset, whence, 2, 2);
1332 : }
1333 :
1334 2 : static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1335 : {
1336 2 : dir_emit_dots(file, ctx);
1337 2 : return 0;
1338 : }
1339 :
1340 : static const struct file_operations empty_dir_operations = {
1341 : .llseek = empty_dir_llseek,
1342 : .read = generic_read_dir,
1343 : .iterate_shared = empty_dir_readdir,
1344 : .fsync = noop_fsync,
1345 : };
1346 :
1347 :
1348 4 : void make_empty_dir_inode(struct inode *inode)
1349 : {
1350 4 : set_nlink(inode, 2);
1351 4 : inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1352 4 : inode->i_uid = GLOBAL_ROOT_UID;
1353 4 : inode->i_gid = GLOBAL_ROOT_GID;
1354 4 : inode->i_rdev = 0;
1355 4 : inode->i_size = 0;
1356 4 : inode->i_blkbits = PAGE_SHIFT;
1357 4 : inode->i_blocks = 0;
1358 :
1359 4 : inode->i_op = &empty_dir_inode_operations;
1360 4 : inode->i_opflags &= ~IOP_XATTR;
1361 4 : inode->i_fop = &empty_dir_operations;
1362 4 : }
1363 :
1364 0 : bool is_empty_dir_inode(struct inode *inode)
1365 : {
1366 0 : return (inode->i_fop == &empty_dir_operations) &&
1367 0 : (inode->i_op == &empty_dir_inode_operations);
1368 : }
1369 :
1370 : #ifdef CONFIG_UNICODE
1371 : /*
1372 : * Determine if the name of a dentry should be casefolded.
1373 : *
1374 : * Return: if names will need casefolding
1375 : */
1376 : static bool needs_casefold(const struct inode *dir)
1377 : {
1378 : return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding;
1379 : }
1380 :
1381 : /**
1382 : * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
1383 : * @dentry: dentry whose name we are checking against
1384 : * @len: len of name of dentry
1385 : * @str: str pointer to name of dentry
1386 : * @name: Name to compare against
1387 : *
1388 : * Return: 0 if names match, 1 if mismatch, or -ERRNO
1389 : */
1390 : static int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
1391 : const char *str, const struct qstr *name)
1392 : {
1393 : const struct dentry *parent = READ_ONCE(dentry->d_parent);
1394 : const struct inode *dir = READ_ONCE(parent->d_inode);
1395 : const struct super_block *sb = dentry->d_sb;
1396 : const struct unicode_map *um = sb->s_encoding;
1397 : struct qstr qstr = QSTR_INIT(str, len);
1398 : char strbuf[DNAME_INLINE_LEN];
1399 : int ret;
1400 :
1401 : if (!dir || !needs_casefold(dir))
1402 : goto fallback;
1403 : /*
1404 : * If the dentry name is stored in-line, then it may be concurrently
1405 : * modified by a rename. If this happens, the VFS will eventually retry
1406 : * the lookup, so it doesn't matter what ->d_compare() returns.
1407 : * However, it's unsafe to call utf8_strncasecmp() with an unstable
1408 : * string. Therefore, we have to copy the name into a temporary buffer.
1409 : */
1410 : if (len <= DNAME_INLINE_LEN - 1) {
1411 : memcpy(strbuf, str, len);
1412 : strbuf[len] = 0;
1413 : qstr.name = strbuf;
1414 : /* prevent compiler from optimizing out the temporary buffer */
1415 : barrier();
1416 : }
1417 : ret = utf8_strncasecmp(um, name, &qstr);
1418 : if (ret >= 0)
1419 : return ret;
1420 :
1421 : if (sb_has_strict_encoding(sb))
1422 : return -EINVAL;
1423 : fallback:
1424 : if (len != name->len)
1425 : return 1;
1426 : return !!memcmp(str, name->name, len);
1427 : }
1428 :
1429 : /**
1430 : * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
1431 : * @dentry: dentry of the parent directory
1432 : * @str: qstr of name whose hash we should fill in
1433 : *
1434 : * Return: 0 if hash was successful or unchanged, and -EINVAL on error
1435 : */
1436 : static int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str)
1437 : {
1438 : const struct inode *dir = READ_ONCE(dentry->d_inode);
1439 : struct super_block *sb = dentry->d_sb;
1440 : const struct unicode_map *um = sb->s_encoding;
1441 : int ret = 0;
1442 :
1443 : if (!dir || !needs_casefold(dir))
1444 : return 0;
1445 :
1446 : ret = utf8_casefold_hash(um, dentry, str);
1447 : if (ret < 0 && sb_has_strict_encoding(sb))
1448 : return -EINVAL;
1449 : return 0;
1450 : }
1451 :
1452 : static const struct dentry_operations generic_ci_dentry_ops = {
1453 : .d_hash = generic_ci_d_hash,
1454 : .d_compare = generic_ci_d_compare,
1455 : };
1456 : #endif
1457 :
1458 : #ifdef CONFIG_FS_ENCRYPTION
1459 : static const struct dentry_operations generic_encrypted_dentry_ops = {
1460 : .d_revalidate = fscrypt_d_revalidate,
1461 : };
1462 : #endif
1463 :
1464 : #if defined(CONFIG_FS_ENCRYPTION) && defined(CONFIG_UNICODE)
1465 : static const struct dentry_operations generic_encrypted_ci_dentry_ops = {
1466 : .d_hash = generic_ci_d_hash,
1467 : .d_compare = generic_ci_d_compare,
1468 : .d_revalidate = fscrypt_d_revalidate,
1469 : };
1470 : #endif
1471 :
1472 : /**
1473 : * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry
1474 : * @dentry: dentry to set ops on
1475 : *
1476 : * Casefolded directories need d_hash and d_compare set, so that the dentries
1477 : * contained in them are handled case-insensitively. Note that these operations
1478 : * are needed on the parent directory rather than on the dentries in it, and
1479 : * while the casefolding flag can be toggled on and off on an empty directory,
1480 : * dentry_operations can't be changed later. As a result, if the filesystem has
1481 : * casefolding support enabled at all, we have to give all dentries the
1482 : * casefolding operations even if their inode doesn't have the casefolding flag
1483 : * currently (and thus the casefolding ops would be no-ops for now).
1484 : *
1485 : * Encryption works differently in that the only dentry operation it needs is
1486 : * d_revalidate, which it only needs on dentries that have the no-key name flag.
1487 : * The no-key flag can't be set "later", so we don't have to worry about that.
1488 : *
1489 : * Finally, to maximize compatibility with overlayfs (which isn't compatible
1490 : * with certain dentry operations) and to avoid taking an unnecessary
1491 : * performance hit, we use custom dentry_operations for each possible
1492 : * combination rather than always installing all operations.
1493 : */
1494 6930 : void generic_set_encrypted_ci_d_ops(struct dentry *dentry)
1495 : {
1496 : #ifdef CONFIG_FS_ENCRYPTION
1497 : bool needs_encrypt_ops = dentry->d_flags & DCACHE_NOKEY_NAME;
1498 : #endif
1499 : #ifdef CONFIG_UNICODE
1500 : bool needs_ci_ops = dentry->d_sb->s_encoding;
1501 : #endif
1502 : #if defined(CONFIG_FS_ENCRYPTION) && defined(CONFIG_UNICODE)
1503 : if (needs_encrypt_ops && needs_ci_ops) {
1504 : d_set_d_op(dentry, &generic_encrypted_ci_dentry_ops);
1505 : return;
1506 : }
1507 : #endif
1508 : #ifdef CONFIG_FS_ENCRYPTION
1509 : if (needs_encrypt_ops) {
1510 : d_set_d_op(dentry, &generic_encrypted_dentry_ops);
1511 : return;
1512 : }
1513 : #endif
1514 : #ifdef CONFIG_UNICODE
1515 : if (needs_ci_ops) {
1516 : d_set_d_op(dentry, &generic_ci_dentry_ops);
1517 : return;
1518 : }
1519 : #endif
1520 6930 : }
1521 : EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops);
|
