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/iversion.h>
19 : #include <linux/writeback.h>
20 : #include <linux/buffer_head.h> /* sync_mapping_buffers */
21 : #include <linux/fs_context.h>
22 : #include <linux/pseudo_fs.h>
23 : #include <linux/fsnotify.h>
24 : #include <linux/unicode.h>
25 : #include <linux/fscrypt.h>
26 :
27 : #include <linux/uaccess.h>
28 :
29 : #include "internal.h"
30 :
31 0 : int simple_getattr(struct mnt_idmap *idmap, const struct path *path,
32 : struct kstat *stat, u32 request_mask,
33 : unsigned int query_flags)
34 : {
35 0 : struct inode *inode = d_inode(path->dentry);
36 0 : generic_fillattr(&nop_mnt_idmap, inode, stat);
37 0 : stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
38 0 : return 0;
39 : }
40 : EXPORT_SYMBOL(simple_getattr);
41 :
42 267186 : int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
43 : {
44 267186 : buf->f_type = dentry->d_sb->s_magic;
45 267186 : buf->f_bsize = PAGE_SIZE;
46 267186 : buf->f_namelen = NAME_MAX;
47 267186 : return 0;
48 : }
49 : EXPORT_SYMBOL(simple_statfs);
50 :
51 : /*
52 : * Retaining negative dentries for an in-memory filesystem just wastes
53 : * memory and lookup time: arrange for them to be deleted immediately.
54 : */
55 737182121 : int always_delete_dentry(const struct dentry *dentry)
56 : {
57 737182121 : return 1;
58 : }
59 : EXPORT_SYMBOL(always_delete_dentry);
60 :
61 : const struct dentry_operations simple_dentry_operations = {
62 : .d_delete = always_delete_dentry,
63 : };
64 : EXPORT_SYMBOL(simple_dentry_operations);
65 :
66 : /*
67 : * Lookup the data. This is trivial - if the dentry didn't already
68 : * exist, we know it is negative. Set d_op to delete negative dentries.
69 : */
70 737767548 : struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
71 : {
72 737767548 : if (dentry->d_name.len > NAME_MAX)
73 : return ERR_PTR(-ENAMETOOLONG);
74 737763834 : if (!dentry->d_sb->s_d_op)
75 737655074 : d_set_d_op(dentry, &simple_dentry_operations);
76 737788254 : d_add(dentry, NULL);
77 737788254 : return NULL;
78 : }
79 : EXPORT_SYMBOL(simple_lookup);
80 :
81 5663804 : int dcache_dir_open(struct inode *inode, struct file *file)
82 : {
83 5663804 : file->private_data = d_alloc_cursor(file->f_path.dentry);
84 :
85 5663581 : return file->private_data ? 0 : -ENOMEM;
86 : }
87 : EXPORT_SYMBOL(dcache_dir_open);
88 :
89 5661925 : int dcache_dir_close(struct inode *inode, struct file *file)
90 : {
91 5661925 : dput(file->private_data);
92 5662917 : return 0;
93 : }
94 : EXPORT_SYMBOL(dcache_dir_close);
95 :
96 : /* parent is locked at least shared */
97 : /*
98 : * Returns an element of siblings' list.
99 : * We are looking for <count>th positive after <p>; if
100 : * found, dentry is grabbed and returned to caller.
101 : * If no such element exists, NULL is returned.
102 : */
103 263399721 : static struct dentry *scan_positives(struct dentry *cursor,
104 : struct list_head *p,
105 : loff_t count,
106 : struct dentry *last)
107 : {
108 263399721 : struct dentry *dentry = cursor->d_parent, *found = NULL;
109 :
110 263399721 : spin_lock(&dentry->d_lock);
111 263416777 : while ((p = p->next) != &dentry->d_subdirs) {
112 257766958 : struct dentry *d = list_entry(p, struct dentry, d_child);
113 : // we must at least skip cursors, to avoid livelocks
114 257766958 : if (d->d_flags & DCACHE_DENTRY_CURSOR)
115 1 : continue;
116 515516594 : if (simple_positive(d) && !--count) {
117 257749539 : spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
118 257749846 : if (simple_positive(d))
119 257749846 : found = dget_dlock(d);
120 257749846 : spin_unlock(&d->d_lock);
121 257750382 : if (likely(found))
122 : break;
123 : count = 1;
124 : }
125 17418 : if (need_resched()) {
126 9 : list_move(&cursor->d_child, p);
127 9 : p = &cursor->d_child;
128 9 : spin_unlock(&dentry->d_lock);
129 9 : cond_resched();
130 9 : spin_lock(&dentry->d_lock);
131 : }
132 : }
133 263400201 : spin_unlock(&dentry->d_lock);
134 263400329 : dput(last);
135 263399900 : return found;
136 : }
137 :
138 8 : loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
139 : {
140 8 : struct dentry *dentry = file->f_path.dentry;
141 8 : switch (whence) {
142 0 : case 1:
143 0 : offset += file->f_pos;
144 8 : fallthrough;
145 8 : case 0:
146 8 : if (offset >= 0)
147 : break;
148 : fallthrough;
149 : default:
150 : return -EINVAL;
151 : }
152 8 : if (offset != file->f_pos) {
153 8 : struct dentry *cursor = file->private_data;
154 8 : struct dentry *to = NULL;
155 :
156 8 : inode_lock_shared(dentry->d_inode);
157 :
158 8 : if (offset > 2)
159 0 : to = scan_positives(cursor, &dentry->d_subdirs,
160 : offset - 2, NULL);
161 8 : spin_lock(&dentry->d_lock);
162 8 : if (to)
163 0 : list_move(&cursor->d_child, &to->d_child);
164 : else
165 8 : list_del_init(&cursor->d_child);
166 8 : spin_unlock(&dentry->d_lock);
167 8 : dput(to);
168 :
169 8 : file->f_pos = offset;
170 :
171 8 : inode_unlock_shared(dentry->d_inode);
172 : }
173 : return offset;
174 : }
175 : EXPORT_SYMBOL(dcache_dir_lseek);
176 :
177 : /*
178 : * Directory is locked and all positive dentries in it are safe, since
179 : * for ramfs-type trees they can't go away without unlink() or rmdir(),
180 : * both impossible due to the lock on directory.
181 : */
182 :
183 11212804 : int dcache_readdir(struct file *file, struct dir_context *ctx)
184 : {
185 11212804 : struct dentry *dentry = file->f_path.dentry;
186 11212804 : struct dentry *cursor = file->private_data;
187 11212804 : struct list_head *anchor = &dentry->d_subdirs;
188 11212804 : struct dentry *next = NULL;
189 11212804 : struct list_head *p;
190 :
191 11212804 : if (!dir_emit_dots(file, ctx))
192 : return 0;
193 :
194 11211900 : if (ctx->pos == 2)
195 : p = anchor;
196 5562846 : else if (!list_empty(&cursor->d_child))
197 : p = &cursor->d_child;
198 : else
199 : return 0;
200 :
201 263399455 : while ((next = scan_positives(cursor, p, 1, next)) != NULL) {
202 257750431 : if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
203 257749687 : d_inode(next)->i_ino,
204 257750431 : fs_umode_to_dtype(d_inode(next)->i_mode)))
205 : break;
206 257744963 : ctx->pos++;
207 257744963 : p = &next->d_child;
208 : }
209 5655213 : spin_lock(&dentry->d_lock);
210 5655340 : if (next)
211 5441 : list_move_tail(&cursor->d_child, &next->d_child);
212 : else
213 5649899 : list_del_init(&cursor->d_child);
214 5654945 : spin_unlock(&dentry->d_lock);
215 5655136 : dput(next);
216 :
217 5655136 : return 0;
218 : }
219 : EXPORT_SYMBOL(dcache_readdir);
220 :
221 320696745 : ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
222 : {
223 320696745 : return -EISDIR;
224 : }
225 : EXPORT_SYMBOL(generic_read_dir);
226 :
227 : const struct file_operations simple_dir_operations = {
228 : .open = dcache_dir_open,
229 : .release = dcache_dir_close,
230 : .llseek = dcache_dir_lseek,
231 : .read = generic_read_dir,
232 : .iterate_shared = dcache_readdir,
233 : .fsync = noop_fsync,
234 : };
235 : EXPORT_SYMBOL(simple_dir_operations);
236 :
237 : const struct inode_operations simple_dir_inode_operations = {
238 : .lookup = simple_lookup,
239 : };
240 : EXPORT_SYMBOL(simple_dir_inode_operations);
241 :
242 773863 : static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev)
243 : {
244 773863 : struct dentry *child = NULL;
245 773863 : struct list_head *p = prev ? &prev->d_child : &parent->d_subdirs;
246 :
247 773863 : spin_lock(&parent->d_lock);
248 773863 : while ((p = p->next) != &parent->d_subdirs) {
249 348926 : struct dentry *d = container_of(p, struct dentry, d_child);
250 348926 : if (simple_positive(d)) {
251 348926 : spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
252 348926 : if (simple_positive(d))
253 348926 : child = dget_dlock(d);
254 348926 : spin_unlock(&d->d_lock);
255 348926 : if (likely(child))
256 : break;
257 : }
258 : }
259 773863 : spin_unlock(&parent->d_lock);
260 773863 : dput(prev);
261 773863 : return child;
262 : }
263 :
264 76011 : void simple_recursive_removal(struct dentry *dentry,
265 : void (*callback)(struct dentry *))
266 : {
267 76011 : struct dentry *this = dget(dentry);
268 1122789 : while (true) {
269 424937 : struct dentry *victim = NULL, *child;
270 424937 : struct inode *inode = this->d_inode;
271 :
272 424937 : inode_lock(inode);
273 773131 : if (d_is_dir(this))
274 76743 : inode->i_flags |= S_DEAD;
275 773863 : while ((child = find_next_child(this, victim)) == NULL) {
276 : // kill and ascend
277 : // update metadata while it's still locked
278 424937 : inode->i_ctime = current_time(inode);
279 424937 : clear_nlink(inode);
280 424937 : inode_unlock(inode);
281 424937 : victim = this;
282 424937 : this = this->d_parent;
283 424937 : inode = this->d_inode;
284 424937 : inode_lock(inode);
285 424937 : if (simple_positive(victim)) {
286 424937 : d_invalidate(victim); // avoid lost mounts
287 773131 : if (d_is_dir(victim))
288 76743 : fsnotify_rmdir(inode, victim);
289 : else
290 348194 : fsnotify_unlink(inode, victim);
291 424937 : if (callback)
292 424937 : callback(victim);
293 424937 : dput(victim); // unpin it
294 : }
295 424937 : if (victim == dentry) {
296 152022 : inode->i_ctime = inode->i_mtime =
297 76011 : current_time(inode);
298 76011 : if (d_is_dir(dentry))
299 76011 : drop_nlink(inode);
300 76011 : inode_unlock(inode);
301 76011 : dput(dentry);
302 76011 : return;
303 : }
304 : }
305 348926 : inode_unlock(inode);
306 348926 : this = child;
307 : }
308 : }
309 : EXPORT_SYMBOL(simple_recursive_removal);
310 :
311 : static const struct super_operations simple_super_operations = {
312 : .statfs = simple_statfs,
313 : };
314 :
315 0 : static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc)
316 : {
317 0 : struct pseudo_fs_context *ctx = fc->fs_private;
318 0 : struct inode *root;
319 :
320 0 : s->s_maxbytes = MAX_LFS_FILESIZE;
321 0 : s->s_blocksize = PAGE_SIZE;
322 0 : s->s_blocksize_bits = PAGE_SHIFT;
323 0 : s->s_magic = ctx->magic;
324 0 : s->s_op = ctx->ops ?: &simple_super_operations;
325 0 : s->s_xattr = ctx->xattr;
326 0 : s->s_time_gran = 1;
327 0 : root = new_inode(s);
328 0 : if (!root)
329 : return -ENOMEM;
330 :
331 : /*
332 : * since this is the first inode, make it number 1. New inodes created
333 : * after this must take care not to collide with it (by passing
334 : * max_reserved of 1 to iunique).
335 : */
336 0 : root->i_ino = 1;
337 0 : root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
338 0 : root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
339 0 : s->s_root = d_make_root(root);
340 0 : if (!s->s_root)
341 : return -ENOMEM;
342 0 : s->s_d_op = ctx->dops;
343 0 : return 0;
344 : }
345 :
346 0 : static int pseudo_fs_get_tree(struct fs_context *fc)
347 : {
348 0 : return get_tree_nodev(fc, pseudo_fs_fill_super);
349 : }
350 :
351 0 : static void pseudo_fs_free(struct fs_context *fc)
352 : {
353 0 : kfree(fc->fs_private);
354 0 : }
355 :
356 : static const struct fs_context_operations pseudo_fs_context_ops = {
357 : .free = pseudo_fs_free,
358 : .get_tree = pseudo_fs_get_tree,
359 : };
360 :
361 : /*
362 : * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
363 : * will never be mountable)
364 : */
365 0 : struct pseudo_fs_context *init_pseudo(struct fs_context *fc,
366 : unsigned long magic)
367 : {
368 0 : struct pseudo_fs_context *ctx;
369 :
370 0 : ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL);
371 0 : if (likely(ctx)) {
372 0 : ctx->magic = magic;
373 0 : fc->fs_private = ctx;
374 0 : fc->ops = &pseudo_fs_context_ops;
375 0 : fc->sb_flags |= SB_NOUSER;
376 0 : fc->global = true;
377 : }
378 0 : return ctx;
379 : }
380 : EXPORT_SYMBOL(init_pseudo);
381 :
382 0 : int simple_open(struct inode *inode, struct file *file)
383 : {
384 0 : if (inode->i_private)
385 0 : file->private_data = inode->i_private;
386 0 : return 0;
387 : }
388 : EXPORT_SYMBOL(simple_open);
389 :
390 0 : int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
391 : {
392 0 : struct inode *inode = d_inode(old_dentry);
393 :
394 0 : inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
395 0 : inc_nlink(inode);
396 0 : ihold(inode);
397 0 : dget(dentry);
398 0 : d_instantiate(dentry, inode);
399 0 : return 0;
400 : }
401 : EXPORT_SYMBOL(simple_link);
402 :
403 2505801 : int simple_empty(struct dentry *dentry)
404 : {
405 2505801 : struct dentry *child;
406 2505801 : int ret = 0;
407 :
408 2505801 : spin_lock(&dentry->d_lock);
409 2505950 : list_for_each_entry(child, &dentry->d_subdirs, d_child) {
410 98473 : spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
411 98484 : if (simple_positive(child)) {
412 98484 : spin_unlock(&child->d_lock);
413 98484 : goto out;
414 : }
415 0 : spin_unlock(&child->d_lock);
416 : }
417 : ret = 1;
418 2505961 : out:
419 2505961 : spin_unlock(&dentry->d_lock);
420 2506024 : return ret;
421 : }
422 : EXPORT_SYMBOL(simple_empty);
423 :
424 0 : int simple_unlink(struct inode *dir, struct dentry *dentry)
425 : {
426 0 : struct inode *inode = d_inode(dentry);
427 :
428 0 : inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
429 0 : drop_nlink(inode);
430 0 : dput(dentry);
431 0 : return 0;
432 : }
433 : EXPORT_SYMBOL(simple_unlink);
434 :
435 0 : int simple_rmdir(struct inode *dir, struct dentry *dentry)
436 : {
437 0 : if (!simple_empty(dentry))
438 : return -ENOTEMPTY;
439 :
440 0 : drop_nlink(d_inode(dentry));
441 0 : simple_unlink(dir, dentry);
442 0 : drop_nlink(dir);
443 0 : return 0;
444 : }
445 : EXPORT_SYMBOL(simple_rmdir);
446 :
447 0 : int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
448 : struct inode *new_dir, struct dentry *new_dentry)
449 : {
450 0 : bool old_is_dir = d_is_dir(old_dentry);
451 0 : bool new_is_dir = d_is_dir(new_dentry);
452 :
453 0 : if (old_dir != new_dir && old_is_dir != new_is_dir) {
454 0 : if (old_is_dir) {
455 0 : drop_nlink(old_dir);
456 0 : inc_nlink(new_dir);
457 : } else {
458 0 : drop_nlink(new_dir);
459 0 : inc_nlink(old_dir);
460 : }
461 : }
462 0 : old_dir->i_ctime = old_dir->i_mtime =
463 0 : new_dir->i_ctime = new_dir->i_mtime =
464 0 : d_inode(old_dentry)->i_ctime =
465 0 : d_inode(new_dentry)->i_ctime = current_time(old_dir);
466 :
467 0 : return 0;
468 : }
469 : EXPORT_SYMBOL_GPL(simple_rename_exchange);
470 :
471 0 : int simple_rename(struct mnt_idmap *idmap, struct inode *old_dir,
472 : struct dentry *old_dentry, struct inode *new_dir,
473 : struct dentry *new_dentry, unsigned int flags)
474 : {
475 0 : struct inode *inode = d_inode(old_dentry);
476 0 : int they_are_dirs = d_is_dir(old_dentry);
477 :
478 0 : if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE))
479 : return -EINVAL;
480 :
481 0 : if (flags & RENAME_EXCHANGE)
482 0 : return simple_rename_exchange(old_dir, old_dentry, new_dir, new_dentry);
483 :
484 0 : if (!simple_empty(new_dentry))
485 : return -ENOTEMPTY;
486 :
487 0 : if (d_really_is_positive(new_dentry)) {
488 0 : simple_unlink(new_dir, new_dentry);
489 0 : if (they_are_dirs) {
490 0 : drop_nlink(d_inode(new_dentry));
491 0 : drop_nlink(old_dir);
492 : }
493 0 : } else if (they_are_dirs) {
494 0 : drop_nlink(old_dir);
495 0 : inc_nlink(new_dir);
496 : }
497 :
498 0 : old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
499 0 : new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
500 :
501 0 : return 0;
502 : }
503 : EXPORT_SYMBOL(simple_rename);
504 :
505 : /**
506 : * simple_setattr - setattr for simple filesystem
507 : * @idmap: idmap of the target mount
508 : * @dentry: dentry
509 : * @iattr: iattr structure
510 : *
511 : * Returns 0 on success, -error on failure.
512 : *
513 : * simple_setattr is a simple ->setattr implementation without a proper
514 : * implementation of size changes.
515 : *
516 : * It can either be used for in-memory filesystems or special files
517 : * on simple regular filesystems. Anything that needs to change on-disk
518 : * or wire state on size changes needs its own setattr method.
519 : */
520 15798 : int simple_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
521 : struct iattr *iattr)
522 : {
523 15798 : struct inode *inode = d_inode(dentry);
524 15798 : int error;
525 :
526 15798 : error = setattr_prepare(idmap, dentry, iattr);
527 15798 : if (error)
528 : return error;
529 :
530 15798 : if (iattr->ia_valid & ATTR_SIZE)
531 15798 : truncate_setsize(inode, iattr->ia_size);
532 15798 : setattr_copy(idmap, inode, iattr);
533 15798 : mark_inode_dirty(inode);
534 15798 : return 0;
535 : }
536 : EXPORT_SYMBOL(simple_setattr);
537 :
538 0 : static int simple_read_folio(struct file *file, struct folio *folio)
539 : {
540 0 : folio_zero_range(folio, 0, folio_size(folio));
541 0 : flush_dcache_folio(folio);
542 0 : folio_mark_uptodate(folio);
543 0 : folio_unlock(folio);
544 0 : return 0;
545 : }
546 :
547 0 : int simple_write_begin(struct file *file, struct address_space *mapping,
548 : loff_t pos, unsigned len,
549 : struct page **pagep, void **fsdata)
550 : {
551 0 : struct page *page;
552 0 : pgoff_t index;
553 :
554 0 : index = pos >> PAGE_SHIFT;
555 :
556 0 : page = grab_cache_page_write_begin(mapping, index);
557 0 : if (!page)
558 : return -ENOMEM;
559 :
560 0 : *pagep = page;
561 :
562 0 : if (!PageUptodate(page) && (len != PAGE_SIZE)) {
563 0 : unsigned from = pos & (PAGE_SIZE - 1);
564 :
565 0 : zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
566 : }
567 : return 0;
568 : }
569 : EXPORT_SYMBOL(simple_write_begin);
570 :
571 : /**
572 : * simple_write_end - .write_end helper for non-block-device FSes
573 : * @file: See .write_end of address_space_operations
574 : * @mapping: "
575 : * @pos: "
576 : * @len: "
577 : * @copied: "
578 : * @page: "
579 : * @fsdata: "
580 : *
581 : * simple_write_end does the minimum needed for updating a page after writing is
582 : * done. It has the same API signature as the .write_end of
583 : * address_space_operations vector. So it can just be set onto .write_end for
584 : * FSes that don't need any other processing. i_mutex is assumed to be held.
585 : * Block based filesystems should use generic_write_end().
586 : * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
587 : * is not called, so a filesystem that actually does store data in .write_inode
588 : * should extend on what's done here with a call to mark_inode_dirty() in the
589 : * case that i_size has changed.
590 : *
591 : * Use *ONLY* with simple_read_folio()
592 : */
593 0 : static int simple_write_end(struct file *file, struct address_space *mapping,
594 : loff_t pos, unsigned len, unsigned copied,
595 : struct page *page, void *fsdata)
596 : {
597 0 : struct inode *inode = page->mapping->host;
598 0 : loff_t last_pos = pos + copied;
599 :
600 : /* zero the stale part of the page if we did a short copy */
601 0 : if (!PageUptodate(page)) {
602 0 : if (copied < len) {
603 0 : unsigned from = pos & (PAGE_SIZE - 1);
604 :
605 0 : zero_user(page, from + copied, len - copied);
606 : }
607 0 : SetPageUptodate(page);
608 : }
609 : /*
610 : * No need to use i_size_read() here, the i_size
611 : * cannot change under us because we hold the i_mutex.
612 : */
613 0 : if (last_pos > inode->i_size)
614 0 : i_size_write(inode, last_pos);
615 :
616 0 : set_page_dirty(page);
617 0 : unlock_page(page);
618 0 : put_page(page);
619 :
620 0 : return copied;
621 : }
622 :
623 : /*
624 : * Provides ramfs-style behavior: data in the pagecache, but no writeback.
625 : */
626 : const struct address_space_operations ram_aops = {
627 : .read_folio = simple_read_folio,
628 : .write_begin = simple_write_begin,
629 : .write_end = simple_write_end,
630 : .dirty_folio = noop_dirty_folio,
631 : };
632 : EXPORT_SYMBOL(ram_aops);
633 :
634 : /*
635 : * the inodes created here are not hashed. If you use iunique to generate
636 : * unique inode values later for this filesystem, then you must take care
637 : * to pass it an appropriate max_reserved value to avoid collisions.
638 : */
639 0 : int simple_fill_super(struct super_block *s, unsigned long magic,
640 : const struct tree_descr *files)
641 : {
642 0 : struct inode *inode;
643 0 : struct dentry *root;
644 0 : struct dentry *dentry;
645 0 : int i;
646 :
647 0 : s->s_blocksize = PAGE_SIZE;
648 0 : s->s_blocksize_bits = PAGE_SHIFT;
649 0 : s->s_magic = magic;
650 0 : s->s_op = &simple_super_operations;
651 0 : s->s_time_gran = 1;
652 :
653 0 : inode = new_inode(s);
654 0 : if (!inode)
655 : return -ENOMEM;
656 : /*
657 : * because the root inode is 1, the files array must not contain an
658 : * entry at index 1
659 : */
660 0 : inode->i_ino = 1;
661 0 : inode->i_mode = S_IFDIR | 0755;
662 0 : inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
663 0 : inode->i_op = &simple_dir_inode_operations;
664 0 : inode->i_fop = &simple_dir_operations;
665 0 : set_nlink(inode, 2);
666 0 : root = d_make_root(inode);
667 0 : if (!root)
668 : return -ENOMEM;
669 0 : for (i = 0; !files->name || files->name[0]; i++, files++) {
670 0 : if (!files->name)
671 0 : continue;
672 :
673 : /* warn if it tries to conflict with the root inode */
674 0 : if (unlikely(i == 1))
675 0 : printk(KERN_WARNING "%s: %s passed in a files array"
676 : "with an index of 1!\n", __func__,
677 : s->s_type->name);
678 :
679 0 : dentry = d_alloc_name(root, files->name);
680 0 : if (!dentry)
681 0 : goto out;
682 0 : inode = new_inode(s);
683 0 : if (!inode) {
684 0 : dput(dentry);
685 0 : goto out;
686 : }
687 0 : inode->i_mode = S_IFREG | files->mode;
688 0 : inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
689 0 : inode->i_fop = files->ops;
690 0 : inode->i_ino = i;
691 0 : d_add(dentry, inode);
692 : }
693 0 : s->s_root = root;
694 0 : return 0;
695 0 : out:
696 0 : d_genocide(root);
697 0 : shrink_dcache_parent(root);
698 0 : dput(root);
699 0 : return -ENOMEM;
700 : }
701 : EXPORT_SYMBOL(simple_fill_super);
702 :
703 : static DEFINE_SPINLOCK(pin_fs_lock);
704 :
705 524035 : int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
706 : {
707 524035 : struct vfsmount *mnt = NULL;
708 524035 : spin_lock(&pin_fs_lock);
709 524035 : if (unlikely(!*mount)) {
710 0 : spin_unlock(&pin_fs_lock);
711 0 : mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
712 0 : if (IS_ERR(mnt))
713 0 : return PTR_ERR(mnt);
714 0 : spin_lock(&pin_fs_lock);
715 0 : if (!*mount)
716 0 : *mount = mnt;
717 : }
718 524035 : mntget(*mount);
719 524035 : ++*count;
720 524035 : spin_unlock(&pin_fs_lock);
721 524035 : mntput(mnt);
722 524035 : return 0;
723 : }
724 : EXPORT_SYMBOL(simple_pin_fs);
725 :
726 500948 : void simple_release_fs(struct vfsmount **mount, int *count)
727 : {
728 500948 : struct vfsmount *mnt;
729 500948 : spin_lock(&pin_fs_lock);
730 500948 : mnt = *mount;
731 500948 : if (!--*count)
732 0 : *mount = NULL;
733 500948 : spin_unlock(&pin_fs_lock);
734 500948 : mntput(mnt);
735 500948 : }
736 : EXPORT_SYMBOL(simple_release_fs);
737 :
738 : /**
739 : * simple_read_from_buffer - copy data from the buffer to user space
740 : * @to: the user space buffer to read to
741 : * @count: the maximum number of bytes to read
742 : * @ppos: the current position in the buffer
743 : * @from: the buffer to read from
744 : * @available: the size of the buffer
745 : *
746 : * The simple_read_from_buffer() function reads up to @count bytes from the
747 : * buffer @from at offset @ppos into the user space address starting at @to.
748 : *
749 : * On success, the number of bytes read is returned and the offset @ppos is
750 : * advanced by this number, or negative value is returned on error.
751 : **/
752 144246 : ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
753 : const void *from, size_t available)
754 : {
755 144246 : loff_t pos = *ppos;
756 144246 : size_t ret;
757 :
758 144246 : if (pos < 0)
759 : return -EINVAL;
760 144246 : if (pos >= available || !count)
761 : return 0;
762 144246 : if (count > available - pos)
763 : count = available - pos;
764 144246 : ret = copy_to_user(to, from + pos, count);
765 144246 : if (ret == count)
766 : return -EFAULT;
767 144246 : count -= ret;
768 144246 : *ppos = pos + count;
769 144246 : return count;
770 : }
771 : EXPORT_SYMBOL(simple_read_from_buffer);
772 :
773 : /**
774 : * simple_write_to_buffer - copy data from user space to the buffer
775 : * @to: the buffer to write to
776 : * @available: the size of the buffer
777 : * @ppos: the current position in the buffer
778 : * @from: the user space buffer to read from
779 : * @count: the maximum number of bytes to read
780 : *
781 : * The simple_write_to_buffer() function reads up to @count bytes from the user
782 : * space address starting at @from into the buffer @to at offset @ppos.
783 : *
784 : * On success, the number of bytes written is returned and the offset @ppos is
785 : * advanced by this number, or negative value is returned on error.
786 : **/
787 0 : ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
788 : const void __user *from, size_t count)
789 : {
790 0 : loff_t pos = *ppos;
791 0 : size_t res;
792 :
793 0 : if (pos < 0)
794 : return -EINVAL;
795 0 : if (pos >= available || !count)
796 : return 0;
797 0 : if (count > available - pos)
798 : count = available - pos;
799 0 : res = copy_from_user(to + pos, from, count);
800 0 : if (res == count)
801 : return -EFAULT;
802 0 : count -= res;
803 0 : *ppos = pos + count;
804 0 : return count;
805 : }
806 : EXPORT_SYMBOL(simple_write_to_buffer);
807 :
808 : /**
809 : * memory_read_from_buffer - copy data from the buffer
810 : * @to: the kernel space buffer to read to
811 : * @count: the maximum number of bytes to read
812 : * @ppos: the current position in the buffer
813 : * @from: the buffer to read from
814 : * @available: the size of the buffer
815 : *
816 : * The memory_read_from_buffer() function reads up to @count bytes from the
817 : * buffer @from at offset @ppos into the kernel space address starting at @to.
818 : *
819 : * On success, the number of bytes read is returned and the offset @ppos is
820 : * advanced by this number, or negative value is returned on error.
821 : **/
822 1206 : ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
823 : const void *from, size_t available)
824 : {
825 1206 : loff_t pos = *ppos;
826 :
827 1206 : if (pos < 0)
828 : return -EINVAL;
829 1206 : if (pos >= available)
830 : return 0;
831 1206 : if (count > available - pos)
832 : count = available - pos;
833 2412 : memcpy(to, from + pos, count);
834 1206 : *ppos = pos + count;
835 :
836 1206 : return count;
837 : }
838 : EXPORT_SYMBOL(memory_read_from_buffer);
839 :
840 : /*
841 : * Transaction based IO.
842 : * The file expects a single write which triggers the transaction, and then
843 : * possibly a read which collects the result - which is stored in a
844 : * file-local buffer.
845 : */
846 :
847 0 : void simple_transaction_set(struct file *file, size_t n)
848 : {
849 0 : struct simple_transaction_argresp *ar = file->private_data;
850 :
851 0 : BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
852 :
853 : /*
854 : * The barrier ensures that ar->size will really remain zero until
855 : * ar->data is ready for reading.
856 : */
857 0 : smp_mb();
858 0 : ar->size = n;
859 0 : }
860 : EXPORT_SYMBOL(simple_transaction_set);
861 :
862 0 : char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
863 : {
864 0 : struct simple_transaction_argresp *ar;
865 0 : static DEFINE_SPINLOCK(simple_transaction_lock);
866 :
867 0 : if (size > SIMPLE_TRANSACTION_LIMIT - 1)
868 : return ERR_PTR(-EFBIG);
869 :
870 0 : ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
871 0 : if (!ar)
872 : return ERR_PTR(-ENOMEM);
873 :
874 0 : spin_lock(&simple_transaction_lock);
875 :
876 : /* only one write allowed per open */
877 0 : if (file->private_data) {
878 0 : spin_unlock(&simple_transaction_lock);
879 0 : free_page((unsigned long)ar);
880 0 : return ERR_PTR(-EBUSY);
881 : }
882 :
883 0 : file->private_data = ar;
884 :
885 0 : spin_unlock(&simple_transaction_lock);
886 :
887 0 : if (copy_from_user(ar->data, buf, size))
888 0 : return ERR_PTR(-EFAULT);
889 :
890 : return ar->data;
891 : }
892 : EXPORT_SYMBOL(simple_transaction_get);
893 :
894 0 : ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
895 : {
896 0 : struct simple_transaction_argresp *ar = file->private_data;
897 :
898 0 : if (!ar)
899 : return 0;
900 0 : return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
901 : }
902 : EXPORT_SYMBOL(simple_transaction_read);
903 :
904 0 : int simple_transaction_release(struct inode *inode, struct file *file)
905 : {
906 0 : free_page((unsigned long)file->private_data);
907 0 : return 0;
908 : }
909 : EXPORT_SYMBOL(simple_transaction_release);
910 :
911 : /* Simple attribute files */
912 :
913 : struct simple_attr {
914 : int (*get)(void *, u64 *);
915 : int (*set)(void *, u64);
916 : char get_buf[24]; /* enough to store a u64 and "\n\0" */
917 : char set_buf[24];
918 : void *data;
919 : const char *fmt; /* format for read operation */
920 : struct mutex mutex; /* protects access to these buffers */
921 : };
922 :
923 : /* simple_attr_open is called by an actual attribute open file operation
924 : * to set the attribute specific access operations. */
925 15768 : int simple_attr_open(struct inode *inode, struct file *file,
926 : int (*get)(void *, u64 *), int (*set)(void *, u64),
927 : const char *fmt)
928 : {
929 15768 : struct simple_attr *attr;
930 :
931 15768 : attr = kzalloc(sizeof(*attr), GFP_KERNEL);
932 15768 : if (!attr)
933 : return -ENOMEM;
934 :
935 15768 : attr->get = get;
936 15768 : attr->set = set;
937 15768 : attr->data = inode->i_private;
938 15768 : attr->fmt = fmt;
939 15768 : mutex_init(&attr->mutex);
940 :
941 15768 : file->private_data = attr;
942 :
943 15768 : return nonseekable_open(inode, file);
944 : }
945 : EXPORT_SYMBOL_GPL(simple_attr_open);
946 :
947 15768 : int simple_attr_release(struct inode *inode, struct file *file)
948 : {
949 15768 : kfree(file->private_data);
950 15768 : return 0;
951 : }
952 : EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
953 :
954 : /* read from the buffer that is filled with the get function */
955 0 : ssize_t simple_attr_read(struct file *file, char __user *buf,
956 : size_t len, loff_t *ppos)
957 : {
958 0 : struct simple_attr *attr;
959 0 : size_t size;
960 0 : ssize_t ret;
961 :
962 0 : attr = file->private_data;
963 :
964 0 : if (!attr->get)
965 : return -EACCES;
966 :
967 0 : ret = mutex_lock_interruptible(&attr->mutex);
968 0 : if (ret)
969 : return ret;
970 :
971 0 : if (*ppos && attr->get_buf[0]) {
972 : /* continued read */
973 0 : size = strlen(attr->get_buf);
974 : } else {
975 : /* first read */
976 0 : u64 val;
977 0 : ret = attr->get(attr->data, &val);
978 0 : if (ret)
979 0 : goto out;
980 :
981 0 : size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
982 : attr->fmt, (unsigned long long)val);
983 : }
984 :
985 0 : ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
986 0 : out:
987 0 : mutex_unlock(&attr->mutex);
988 0 : return ret;
989 : }
990 : EXPORT_SYMBOL_GPL(simple_attr_read);
991 :
992 : /* interpret the buffer as a number to call the set function with */
993 15768 : static ssize_t simple_attr_write_xsigned(struct file *file, const char __user *buf,
994 : size_t len, loff_t *ppos, bool is_signed)
995 : {
996 15768 : struct simple_attr *attr;
997 15768 : unsigned long long val;
998 15768 : size_t size;
999 15768 : ssize_t ret;
1000 :
1001 15768 : attr = file->private_data;
1002 15768 : if (!attr->set)
1003 : return -EACCES;
1004 :
1005 15768 : ret = mutex_lock_interruptible(&attr->mutex);
1006 15768 : if (ret)
1007 : return ret;
1008 :
1009 15768 : ret = -EFAULT;
1010 15768 : size = min(sizeof(attr->set_buf) - 1, len);
1011 31536 : if (copy_from_user(attr->set_buf, buf, size))
1012 0 : goto out;
1013 :
1014 15768 : attr->set_buf[size] = '\0';
1015 15768 : if (is_signed)
1016 4 : ret = kstrtoll(attr->set_buf, 0, &val);
1017 : else
1018 15764 : ret = kstrtoull(attr->set_buf, 0, &val);
1019 15768 : if (ret)
1020 0 : goto out;
1021 15768 : ret = attr->set(attr->data, val);
1022 15768 : if (ret == 0)
1023 15768 : ret = len; /* on success, claim we got the whole input */
1024 0 : out:
1025 15768 : mutex_unlock(&attr->mutex);
1026 15768 : return ret;
1027 : }
1028 :
1029 15764 : ssize_t simple_attr_write(struct file *file, const char __user *buf,
1030 : size_t len, loff_t *ppos)
1031 : {
1032 15764 : return simple_attr_write_xsigned(file, buf, len, ppos, false);
1033 : }
1034 : EXPORT_SYMBOL_GPL(simple_attr_write);
1035 :
1036 4 : ssize_t simple_attr_write_signed(struct file *file, const char __user *buf,
1037 : size_t len, loff_t *ppos)
1038 : {
1039 4 : return simple_attr_write_xsigned(file, buf, len, ppos, true);
1040 : }
1041 : EXPORT_SYMBOL_GPL(simple_attr_write_signed);
1042 :
1043 : /**
1044 : * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
1045 : * @sb: filesystem to do the file handle conversion on
1046 : * @fid: file handle to convert
1047 : * @fh_len: length of the file handle in bytes
1048 : * @fh_type: type of file handle
1049 : * @get_inode: filesystem callback to retrieve inode
1050 : *
1051 : * This function decodes @fid as long as it has one of the well-known
1052 : * Linux filehandle types and calls @get_inode on it to retrieve the
1053 : * inode for the object specified in the file handle.
1054 : */
1055 4223 : struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
1056 : int fh_len, int fh_type, struct inode *(*get_inode)
1057 : (struct super_block *sb, u64 ino, u32 gen))
1058 : {
1059 4223 : struct inode *inode = NULL;
1060 :
1061 4223 : if (fh_len < 2)
1062 : return NULL;
1063 :
1064 4223 : switch (fh_type) {
1065 4223 : case FILEID_INO32_GEN:
1066 : case FILEID_INO32_GEN_PARENT:
1067 4223 : inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
1068 4223 : break;
1069 : }
1070 :
1071 4223 : return d_obtain_alias(inode);
1072 : }
1073 : EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
1074 :
1075 : /**
1076 : * generic_fh_to_parent - generic helper for the fh_to_parent export operation
1077 : * @sb: filesystem to do the file handle conversion on
1078 : * @fid: file handle to convert
1079 : * @fh_len: length of the file handle in bytes
1080 : * @fh_type: type of file handle
1081 : * @get_inode: filesystem callback to retrieve inode
1082 : *
1083 : * This function decodes @fid as long as it has one of the well-known
1084 : * Linux filehandle types and calls @get_inode on it to retrieve the
1085 : * inode for the _parent_ object specified in the file handle if it
1086 : * is specified in the file handle, or NULL otherwise.
1087 : */
1088 0 : struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
1089 : int fh_len, int fh_type, struct inode *(*get_inode)
1090 : (struct super_block *sb, u64 ino, u32 gen))
1091 : {
1092 0 : struct inode *inode = NULL;
1093 :
1094 0 : if (fh_len <= 2)
1095 : return NULL;
1096 :
1097 0 : switch (fh_type) {
1098 0 : case FILEID_INO32_GEN_PARENT:
1099 0 : inode = get_inode(sb, fid->i32.parent_ino,
1100 : (fh_len > 3 ? fid->i32.parent_gen : 0));
1101 0 : break;
1102 : }
1103 :
1104 0 : return d_obtain_alias(inode);
1105 : }
1106 : EXPORT_SYMBOL_GPL(generic_fh_to_parent);
1107 :
1108 : /**
1109 : * __generic_file_fsync - generic fsync implementation for simple filesystems
1110 : *
1111 : * @file: file to synchronize
1112 : * @start: start offset in bytes
1113 : * @end: end offset in bytes (inclusive)
1114 : * @datasync: only synchronize essential metadata if true
1115 : *
1116 : * This is a generic implementation of the fsync method for simple
1117 : * filesystems which track all non-inode metadata in the buffers list
1118 : * hanging off the address_space structure.
1119 : */
1120 0 : int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
1121 : int datasync)
1122 : {
1123 0 : struct inode *inode = file->f_mapping->host;
1124 0 : int err;
1125 0 : int ret;
1126 :
1127 0 : err = file_write_and_wait_range(file, start, end);
1128 0 : if (err)
1129 : return err;
1130 :
1131 0 : inode_lock(inode);
1132 0 : ret = sync_mapping_buffers(inode->i_mapping);
1133 0 : if (!(inode->i_state & I_DIRTY_ALL))
1134 0 : goto out;
1135 0 : if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
1136 0 : goto out;
1137 :
1138 0 : err = sync_inode_metadata(inode, 1);
1139 0 : if (ret == 0)
1140 0 : ret = err;
1141 :
1142 0 : out:
1143 0 : inode_unlock(inode);
1144 : /* check and advance again to catch errors after syncing out buffers */
1145 0 : err = file_check_and_advance_wb_err(file);
1146 0 : if (ret == 0)
1147 0 : ret = err;
1148 : return ret;
1149 : }
1150 : EXPORT_SYMBOL(__generic_file_fsync);
1151 :
1152 : /**
1153 : * generic_file_fsync - generic fsync implementation for simple filesystems
1154 : * with flush
1155 : * @file: file to synchronize
1156 : * @start: start offset in bytes
1157 : * @end: end offset in bytes (inclusive)
1158 : * @datasync: only synchronize essential metadata if true
1159 : *
1160 : */
1161 :
1162 0 : int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1163 : int datasync)
1164 : {
1165 0 : struct inode *inode = file->f_mapping->host;
1166 0 : int err;
1167 :
1168 0 : err = __generic_file_fsync(file, start, end, datasync);
1169 0 : if (err)
1170 : return err;
1171 0 : return blkdev_issue_flush(inode->i_sb->s_bdev);
1172 : }
1173 : EXPORT_SYMBOL(generic_file_fsync);
1174 :
1175 : /**
1176 : * generic_check_addressable - Check addressability of file system
1177 : * @blocksize_bits: log of file system block size
1178 : * @num_blocks: number of blocks in file system
1179 : *
1180 : * Determine whether a file system with @num_blocks blocks (and a
1181 : * block size of 2**@blocksize_bits) is addressable by the sector_t
1182 : * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1183 : */
1184 2525 : int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1185 : {
1186 2525 : u64 last_fs_block = num_blocks - 1;
1187 2525 : u64 last_fs_page =
1188 2525 : last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1189 :
1190 2525 : if (unlikely(num_blocks == 0))
1191 : return 0;
1192 :
1193 2525 : if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1194 : return -EINVAL;
1195 :
1196 2525 : if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1197 : (last_fs_page > (pgoff_t)(~0ULL))) {
1198 0 : return -EFBIG;
1199 : }
1200 : return 0;
1201 : }
1202 : EXPORT_SYMBOL(generic_check_addressable);
1203 :
1204 : /*
1205 : * No-op implementation of ->fsync for in-memory filesystems.
1206 : */
1207 17021 : int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1208 : {
1209 17021 : return 0;
1210 : }
1211 : EXPORT_SYMBOL(noop_fsync);
1212 :
1213 0 : ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1214 : {
1215 : /*
1216 : * iomap based filesystems support direct I/O without need for
1217 : * this callback. However, it still needs to be set in
1218 : * inode->a_ops so that open/fcntl know that direct I/O is
1219 : * generally supported.
1220 : */
1221 0 : return -EINVAL;
1222 : }
1223 : EXPORT_SYMBOL_GPL(noop_direct_IO);
1224 :
1225 : /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1226 274844135 : void kfree_link(void *p)
1227 : {
1228 274844135 : kfree(p);
1229 274593855 : }
1230 : EXPORT_SYMBOL(kfree_link);
1231 :
1232 993278 : struct inode *alloc_anon_inode(struct super_block *s)
1233 : {
1234 993278 : static const struct address_space_operations anon_aops = {
1235 : .dirty_folio = noop_dirty_folio,
1236 : };
1237 993278 : struct inode *inode = new_inode_pseudo(s);
1238 :
1239 993059 : if (!inode)
1240 : return ERR_PTR(-ENOMEM);
1241 :
1242 993059 : inode->i_ino = get_next_ino();
1243 992750 : inode->i_mapping->a_ops = &anon_aops;
1244 :
1245 : /*
1246 : * Mark the inode dirty from the very beginning,
1247 : * that way it will never be moved to the dirty
1248 : * list because mark_inode_dirty() will think
1249 : * that it already _is_ on the dirty list.
1250 : */
1251 992750 : inode->i_state = I_DIRTY;
1252 992750 : inode->i_mode = S_IRUSR | S_IWUSR;
1253 992750 : inode->i_uid = current_fsuid();
1254 992750 : inode->i_gid = current_fsgid();
1255 992750 : inode->i_flags |= S_PRIVATE;
1256 992750 : inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1257 992504 : return inode;
1258 : }
1259 : EXPORT_SYMBOL(alloc_anon_inode);
1260 :
1261 : /**
1262 : * simple_nosetlease - generic helper for prohibiting leases
1263 : * @filp: file pointer
1264 : * @arg: type of lease to obtain
1265 : * @flp: new lease supplied for insertion
1266 : * @priv: private data for lm_setup operation
1267 : *
1268 : * Generic helper for filesystems that do not wish to allow leases to be set.
1269 : * All arguments are ignored and it just returns -EINVAL.
1270 : */
1271 : int
1272 0 : simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1273 : void **priv)
1274 : {
1275 0 : return -EINVAL;
1276 : }
1277 : EXPORT_SYMBOL(simple_nosetlease);
1278 :
1279 : /**
1280 : * simple_get_link - generic helper to get the target of "fast" symlinks
1281 : * @dentry: not used here
1282 : * @inode: the symlink inode
1283 : * @done: not used here
1284 : *
1285 : * Generic helper for filesystems to use for symlink inodes where a pointer to
1286 : * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1287 : * since as an optimization the path lookup code uses any non-NULL ->i_link
1288 : * directly, without calling ->get_link(). But ->get_link() still must be set,
1289 : * to mark the inode_operations as being for a symlink.
1290 : *
1291 : * Return: the symlink target
1292 : */
1293 26 : const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1294 : struct delayed_call *done)
1295 : {
1296 26 : return inode->i_link;
1297 : }
1298 : EXPORT_SYMBOL(simple_get_link);
1299 :
1300 : const struct inode_operations simple_symlink_inode_operations = {
1301 : .get_link = simple_get_link,
1302 : };
1303 : EXPORT_SYMBOL(simple_symlink_inode_operations);
1304 :
1305 : /*
1306 : * Operations for a permanently empty directory.
1307 : */
1308 0 : static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1309 : {
1310 0 : return ERR_PTR(-ENOENT);
1311 : }
1312 :
1313 24 : static int empty_dir_getattr(struct mnt_idmap *idmap,
1314 : const struct path *path, struct kstat *stat,
1315 : u32 request_mask, unsigned int query_flags)
1316 : {
1317 24 : struct inode *inode = d_inode(path->dentry);
1318 24 : generic_fillattr(&nop_mnt_idmap, inode, stat);
1319 24 : return 0;
1320 : }
1321 :
1322 0 : static int empty_dir_setattr(struct mnt_idmap *idmap,
1323 : struct dentry *dentry, struct iattr *attr)
1324 : {
1325 0 : return -EPERM;
1326 : }
1327 :
1328 0 : static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1329 : {
1330 0 : return -EOPNOTSUPP;
1331 : }
1332 :
1333 : static const struct inode_operations empty_dir_inode_operations = {
1334 : .lookup = empty_dir_lookup,
1335 : .permission = generic_permission,
1336 : .setattr = empty_dir_setattr,
1337 : .getattr = empty_dir_getattr,
1338 : .listxattr = empty_dir_listxattr,
1339 : };
1340 :
1341 0 : static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1342 : {
1343 : /* An empty directory has two entries . and .. at offsets 0 and 1 */
1344 0 : return generic_file_llseek_size(file, offset, whence, 2, 2);
1345 : }
1346 :
1347 24 : static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1348 : {
1349 24 : dir_emit_dots(file, ctx);
1350 24 : return 0;
1351 : }
1352 :
1353 : static const struct file_operations empty_dir_operations = {
1354 : .llseek = empty_dir_llseek,
1355 : .read = generic_read_dir,
1356 : .iterate_shared = empty_dir_readdir,
1357 : .fsync = noop_fsync,
1358 : };
1359 :
1360 :
1361 12 : void make_empty_dir_inode(struct inode *inode)
1362 : {
1363 12 : set_nlink(inode, 2);
1364 12 : inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1365 12 : inode->i_uid = GLOBAL_ROOT_UID;
1366 12 : inode->i_gid = GLOBAL_ROOT_GID;
1367 12 : inode->i_rdev = 0;
1368 12 : inode->i_size = 0;
1369 12 : inode->i_blkbits = PAGE_SHIFT;
1370 12 : inode->i_blocks = 0;
1371 :
1372 12 : inode->i_op = &empty_dir_inode_operations;
1373 12 : inode->i_opflags &= ~IOP_XATTR;
1374 12 : inode->i_fop = &empty_dir_operations;
1375 12 : }
1376 :
1377 0 : bool is_empty_dir_inode(struct inode *inode)
1378 : {
1379 0 : return (inode->i_fop == &empty_dir_operations) &&
1380 0 : (inode->i_op == &empty_dir_inode_operations);
1381 : }
1382 :
1383 : #if IS_ENABLED(CONFIG_UNICODE)
1384 : /*
1385 : * Determine if the name of a dentry should be casefolded.
1386 : *
1387 : * Return: if names will need casefolding
1388 : */
1389 : static bool needs_casefold(const struct inode *dir)
1390 : {
1391 4604 : return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding;
1392 : }
1393 :
1394 : /**
1395 : * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
1396 : * @dentry: dentry whose name we are checking against
1397 : * @len: len of name of dentry
1398 : * @str: str pointer to name of dentry
1399 : * @name: Name to compare against
1400 : *
1401 : * Return: 0 if names match, 1 if mismatch, or -ERRNO
1402 : */
1403 237 : static int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
1404 : const char *str, const struct qstr *name)
1405 : {
1406 237 : const struct dentry *parent = READ_ONCE(dentry->d_parent);
1407 237 : const struct inode *dir = READ_ONCE(parent->d_inode);
1408 237 : const struct super_block *sb = dentry->d_sb;
1409 237 : const struct unicode_map *um = sb->s_encoding;
1410 237 : struct qstr qstr = QSTR_INIT(str, len);
1411 237 : char strbuf[DNAME_INLINE_LEN];
1412 237 : int ret;
1413 :
1414 474 : if (!dir || !needs_casefold(dir))
1415 29 : goto fallback;
1416 : /*
1417 : * If the dentry name is stored in-line, then it may be concurrently
1418 : * modified by a rename. If this happens, the VFS will eventually retry
1419 : * the lookup, so it doesn't matter what ->d_compare() returns.
1420 : * However, it's unsafe to call utf8_strncasecmp() with an unstable
1421 : * string. Therefore, we have to copy the name into a temporary buffer.
1422 : */
1423 208 : if (len <= DNAME_INLINE_LEN - 1) {
1424 208 : memcpy(strbuf, str, len);
1425 208 : strbuf[len] = 0;
1426 208 : qstr.name = strbuf;
1427 : /* prevent compiler from optimizing out the temporary buffer */
1428 208 : barrier();
1429 : }
1430 208 : ret = utf8_strncasecmp(um, name, &qstr);
1431 208 : if (ret >= 0)
1432 : return ret;
1433 :
1434 0 : if (sb_has_strict_encoding(sb))
1435 : return -EINVAL;
1436 0 : fallback:
1437 29 : if (len != name->len)
1438 : return 1;
1439 58 : return !!memcmp(str, name->name, len);
1440 : }
1441 :
1442 : /**
1443 : * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
1444 : * @dentry: dentry of the parent directory
1445 : * @str: qstr of name whose hash we should fill in
1446 : *
1447 : * Return: 0 if hash was successful or unchanged, and -EINVAL on error
1448 : */
1449 4367 : static int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str)
1450 : {
1451 4367 : const struct inode *dir = READ_ONCE(dentry->d_inode);
1452 4367 : struct super_block *sb = dentry->d_sb;
1453 4367 : const struct unicode_map *um = sb->s_encoding;
1454 4367 : int ret = 0;
1455 :
1456 8734 : if (!dir || !needs_casefold(dir))
1457 : return 0;
1458 :
1459 4332 : ret = utf8_casefold_hash(um, dentry, str);
1460 4332 : if (ret < 0 && sb_has_strict_encoding(sb))
1461 4 : return -EINVAL;
1462 : return 0;
1463 : }
1464 :
1465 : static const struct dentry_operations generic_ci_dentry_ops = {
1466 : .d_hash = generic_ci_d_hash,
1467 : .d_compare = generic_ci_d_compare,
1468 : };
1469 : #endif
1470 :
1471 : #ifdef CONFIG_FS_ENCRYPTION
1472 : static const struct dentry_operations generic_encrypted_dentry_ops = {
1473 : .d_revalidate = fscrypt_d_revalidate,
1474 : };
1475 : #endif
1476 :
1477 : #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
1478 : static const struct dentry_operations generic_encrypted_ci_dentry_ops = {
1479 : .d_hash = generic_ci_d_hash,
1480 : .d_compare = generic_ci_d_compare,
1481 : .d_revalidate = fscrypt_d_revalidate,
1482 : };
1483 : #endif
1484 :
1485 : /**
1486 : * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry
1487 : * @dentry: dentry to set ops on
1488 : *
1489 : * Casefolded directories need d_hash and d_compare set, so that the dentries
1490 : * contained in them are handled case-insensitively. Note that these operations
1491 : * are needed on the parent directory rather than on the dentries in it, and
1492 : * while the casefolding flag can be toggled on and off on an empty directory,
1493 : * dentry_operations can't be changed later. As a result, if the filesystem has
1494 : * casefolding support enabled at all, we have to give all dentries the
1495 : * casefolding operations even if their inode doesn't have the casefolding flag
1496 : * currently (and thus the casefolding ops would be no-ops for now).
1497 : *
1498 : * Encryption works differently in that the only dentry operation it needs is
1499 : * d_revalidate, which it only needs on dentries that have the no-key name flag.
1500 : * The no-key flag can't be set "later", so we don't have to worry about that.
1501 : *
1502 : * Finally, to maximize compatibility with overlayfs (which isn't compatible
1503 : * with certain dentry operations) and to avoid taking an unnecessary
1504 : * performance hit, we use custom dentry_operations for each possible
1505 : * combination rather than always installing all operations.
1506 : */
1507 3043207 : void generic_set_encrypted_ci_d_ops(struct dentry *dentry)
1508 : {
1509 : #ifdef CONFIG_FS_ENCRYPTION
1510 : bool needs_encrypt_ops = dentry->d_flags & DCACHE_NOKEY_NAME;
1511 : #endif
1512 : #if IS_ENABLED(CONFIG_UNICODE)
1513 3043207 : bool needs_ci_ops = dentry->d_sb->s_encoding;
1514 : #endif
1515 : #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
1516 : if (needs_encrypt_ops && needs_ci_ops) {
1517 : d_set_d_op(dentry, &generic_encrypted_ci_dentry_ops);
1518 : return;
1519 : }
1520 : #endif
1521 : #ifdef CONFIG_FS_ENCRYPTION
1522 : if (needs_encrypt_ops) {
1523 : d_set_d_op(dentry, &generic_encrypted_dentry_ops);
1524 : return;
1525 : }
1526 : #endif
1527 : #if IS_ENABLED(CONFIG_UNICODE)
1528 3043207 : if (needs_ci_ops) {
1529 4150 : d_set_d_op(dentry, &generic_ci_dentry_ops);
1530 4150 : return;
1531 : }
1532 : #endif
1533 : }
1534 : EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops);
1535 :
1536 : /**
1537 : * inode_maybe_inc_iversion - increments i_version
1538 : * @inode: inode with the i_version that should be updated
1539 : * @force: increment the counter even if it's not necessary?
1540 : *
1541 : * Every time the inode is modified, the i_version field must be seen to have
1542 : * changed by any observer.
1543 : *
1544 : * If "force" is set or the QUERIED flag is set, then ensure that we increment
1545 : * the value, and clear the queried flag.
1546 : *
1547 : * In the common case where neither is set, then we can return "false" without
1548 : * updating i_version.
1549 : *
1550 : * If this function returns false, and no other metadata has changed, then we
1551 : * can avoid logging the metadata.
1552 : */
1553 1822191288 : bool inode_maybe_inc_iversion(struct inode *inode, bool force)
1554 : {
1555 1822191288 : u64 cur, new;
1556 :
1557 : /*
1558 : * The i_version field is not strictly ordered with any other inode
1559 : * information, but the legacy inode_inc_iversion code used a spinlock
1560 : * to serialize increments.
1561 : *
1562 : * Here, we add full memory barriers to ensure that any de-facto
1563 : * ordering with other info is preserved.
1564 : *
1565 : * This barrier pairs with the barrier in inode_query_iversion()
1566 : */
1567 1822191288 : smp_mb();
1568 1822613917 : cur = inode_peek_iversion_raw(inode);
1569 1822614175 : do {
1570 : /* If flag is clear then we needn't do anything */
1571 1822614175 : if (!force && !(cur & I_VERSION_QUERIED))
1572 : return false;
1573 :
1574 : /* Since lowest bit is flag, add 2 to avoid it */
1575 1678377002 : new = (cur & ~I_VERSION_QUERIED) + I_VERSION_INCREMENT;
1576 3356628635 : } while (!atomic64_try_cmpxchg(&inode->i_version, &cur, new));
1577 : return true;
1578 : }
1579 : EXPORT_SYMBOL(inode_maybe_inc_iversion);
1580 :
1581 : /**
1582 : * inode_query_iversion - read i_version for later use
1583 : * @inode: inode from which i_version should be read
1584 : *
1585 : * Read the inode i_version counter. This should be used by callers that wish
1586 : * to store the returned i_version for later comparison. This will guarantee
1587 : * that a later query of the i_version will result in a different value if
1588 : * anything has changed.
1589 : *
1590 : * In this implementation, we fetch the current value, set the QUERIED flag and
1591 : * then try to swap it into place with a cmpxchg, if it wasn't already set. If
1592 : * that fails, we try again with the newly fetched value from the cmpxchg.
1593 : */
1594 805750 : u64 inode_query_iversion(struct inode *inode)
1595 : {
1596 805750 : u64 cur, new;
1597 :
1598 805750 : cur = inode_peek_iversion_raw(inode);
1599 805750 : do {
1600 : /* If flag is already set, then no need to swap */
1601 805750 : if (cur & I_VERSION_QUERIED) {
1602 : /*
1603 : * This barrier (and the implicit barrier in the
1604 : * cmpxchg below) pairs with the barrier in
1605 : * inode_maybe_inc_iversion().
1606 : */
1607 761053 : smp_mb();
1608 761053 : break;
1609 : }
1610 :
1611 44697 : new = cur | I_VERSION_QUERIED;
1612 89418 : } while (!atomic64_try_cmpxchg(&inode->i_version, &cur, new));
1613 805796 : return cur >> I_VERSION_QUERIED_SHIFT;
1614 : }
1615 : EXPORT_SYMBOL(inode_query_iversion);
1616 :
1617 0 : ssize_t direct_write_fallback(struct kiocb *iocb, struct iov_iter *iter,
1618 : ssize_t direct_written, ssize_t buffered_written)
1619 : {
1620 0 : struct address_space *mapping = iocb->ki_filp->f_mapping;
1621 0 : loff_t pos = iocb->ki_pos - buffered_written;
1622 0 : loff_t end = iocb->ki_pos - 1;
1623 0 : int err;
1624 :
1625 : /*
1626 : * If the buffered write fallback returned an error, we want to return
1627 : * the number of bytes which were written by direct I/O, or the error
1628 : * code if that was zero.
1629 : *
1630 : * Note that this differs from normal direct-io semantics, which will
1631 : * return -EFOO even if some bytes were written.
1632 : */
1633 0 : if (unlikely(buffered_written < 0)) {
1634 0 : if (direct_written)
1635 : return direct_written;
1636 0 : return buffered_written;
1637 : }
1638 :
1639 : /*
1640 : * We need to ensure that the page cache pages are written to disk and
1641 : * invalidated to preserve the expected O_DIRECT semantics.
1642 : */
1643 0 : err = filemap_write_and_wait_range(mapping, pos, end);
1644 0 : if (err < 0) {
1645 : /*
1646 : * We don't know how much we wrote, so just return the number of
1647 : * bytes which were direct-written
1648 : */
1649 0 : if (direct_written)
1650 : return direct_written;
1651 0 : return err;
1652 : }
1653 0 : invalidate_mapping_pages(mapping, pos >> PAGE_SHIFT, end >> PAGE_SHIFT);
1654 0 : return direct_written + buffered_written;
1655 : }
1656 : EXPORT_SYMBOL_GPL(direct_write_fallback);
|
