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 
simple_getattr(struct user_namespace * mnt_userns,const struct path * path,struct kstat * stat,u32 request_mask,unsigned int query_flags)31 int simple_getattr(struct user_namespace *mnt_userns, const struct path *path,
32 		   struct kstat *stat, u32 request_mask,
33 		   unsigned int query_flags)
34 {
35 	struct inode *inode = d_inode(path->dentry);
36 	generic_fillattr(&init_user_ns, inode, stat);
37 	stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
38 	return 0;
39 }
40 EXPORT_SYMBOL(simple_getattr);
41 
simple_statfs(struct dentry * dentry,struct kstatfs * buf)42 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
43 {
44 	buf->f_type = dentry->d_sb->s_magic;
45 	buf->f_bsize = PAGE_SIZE;
46 	buf->f_namelen = NAME_MAX;
47 	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  */
always_delete_dentry(const struct dentry * dentry)55 int always_delete_dentry(const struct dentry *dentry)
56 {
57 	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  */
simple_lookup(struct inode * dir,struct dentry * dentry,unsigned int flags)70 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
71 {
72 	if (dentry->d_name.len > NAME_MAX)
73 		return ERR_PTR(-ENAMETOOLONG);
74 	if (!dentry->d_sb->s_d_op)
75 		d_set_d_op(dentry, &simple_dentry_operations);
76 	d_add(dentry, NULL);
77 	return NULL;
78 }
79 EXPORT_SYMBOL(simple_lookup);
80 
dcache_dir_open(struct inode * inode,struct file * file)81 int dcache_dir_open(struct inode *inode, struct file *file)
82 {
83 	file->private_data = d_alloc_cursor(file->f_path.dentry);
84 
85 	return file->private_data ? 0 : -ENOMEM;
86 }
87 EXPORT_SYMBOL(dcache_dir_open);
88 
dcache_dir_close(struct inode * inode,struct file * file)89 int dcache_dir_close(struct inode *inode, struct file *file)
90 {
91 	dput(file->private_data);
92 	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  */
scan_positives(struct dentry * cursor,struct list_head * p,loff_t count,struct dentry * last)103 static struct dentry *scan_positives(struct dentry *cursor,
104 					struct list_head *p,
105 					loff_t count,
106 					struct dentry *last)
107 {
108 	struct dentry *dentry = cursor->d_parent, *found = NULL;
109 
110 	spin_lock(&dentry->d_lock);
111 	while ((p = p->next) != &dentry->d_subdirs) {
112 		struct dentry *d = list_entry(p, struct dentry, d_child);
113 		// we must at least skip cursors, to avoid livelocks
114 		if (d->d_flags & DCACHE_DENTRY_CURSOR)
115 			continue;
116 		if (simple_positive(d) && !--count) {
117 			spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
118 			if (simple_positive(d))
119 				found = dget_dlock(d);
120 			spin_unlock(&d->d_lock);
121 			if (likely(found))
122 				break;
123 			count = 1;
124 		}
125 		if (need_resched()) {
126 			list_move(&cursor->d_child, p);
127 			p = &cursor->d_child;
128 			spin_unlock(&dentry->d_lock);
129 			cond_resched();
130 			spin_lock(&dentry->d_lock);
131 		}
132 	}
133 	spin_unlock(&dentry->d_lock);
134 	dput(last);
135 	return found;
136 }
137 
dcache_dir_lseek(struct file * file,loff_t offset,int whence)138 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
139 {
140 	struct dentry *dentry = file->f_path.dentry;
141 	switch (whence) {
142 		case 1:
143 			offset += file->f_pos;
144 			fallthrough;
145 		case 0:
146 			if (offset >= 0)
147 				break;
148 			fallthrough;
149 		default:
150 			return -EINVAL;
151 	}
152 	if (offset != file->f_pos) {
153 		struct dentry *cursor = file->private_data;
154 		struct dentry *to = NULL;
155 
156 		inode_lock_shared(dentry->d_inode);
157 
158 		if (offset > 2)
159 			to = scan_positives(cursor, &dentry->d_subdirs,
160 					    offset - 2, NULL);
161 		spin_lock(&dentry->d_lock);
162 		if (to)
163 			list_move(&cursor->d_child, &to->d_child);
164 		else
165 			list_del_init(&cursor->d_child);
166 		spin_unlock(&dentry->d_lock);
167 		dput(to);
168 
169 		file->f_pos = offset;
170 
171 		inode_unlock_shared(dentry->d_inode);
172 	}
173 	return offset;
174 }
175 EXPORT_SYMBOL(dcache_dir_lseek);
176 
177 /* Relationship between i_mode and the DT_xxx types */
dt_type(struct inode * inode)178 static inline unsigned char dt_type(struct inode *inode)
179 {
180 	return (inode->i_mode >> 12) & 15;
181 }
182 
183 /*
184  * Directory is locked and all positive dentries in it are safe, since
185  * for ramfs-type trees they can't go away without unlink() or rmdir(),
186  * both impossible due to the lock on directory.
187  */
188 
dcache_readdir(struct file * file,struct dir_context * ctx)189 int dcache_readdir(struct file *file, struct dir_context *ctx)
190 {
191 	struct dentry *dentry = file->f_path.dentry;
192 	struct dentry *cursor = file->private_data;
193 	struct list_head *anchor = &dentry->d_subdirs;
194 	struct dentry *next = NULL;
195 	struct list_head *p;
196 
197 	if (!dir_emit_dots(file, ctx))
198 		return 0;
199 
200 	if (ctx->pos == 2)
201 		p = anchor;
202 	else if (!list_empty(&cursor->d_child))
203 		p = &cursor->d_child;
204 	else
205 		return 0;
206 
207 	while ((next = scan_positives(cursor, p, 1, next)) != NULL) {
208 		if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
209 			      d_inode(next)->i_ino, dt_type(d_inode(next))))
210 			break;
211 		ctx->pos++;
212 		p = &next->d_child;
213 	}
214 	spin_lock(&dentry->d_lock);
215 	if (next)
216 		list_move_tail(&cursor->d_child, &next->d_child);
217 	else
218 		list_del_init(&cursor->d_child);
219 	spin_unlock(&dentry->d_lock);
220 	dput(next);
221 
222 	return 0;
223 }
224 EXPORT_SYMBOL(dcache_readdir);
225 
generic_read_dir(struct file * filp,char __user * buf,size_t siz,loff_t * ppos)226 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
227 {
228 	return -EISDIR;
229 }
230 EXPORT_SYMBOL(generic_read_dir);
231 
232 const struct file_operations simple_dir_operations = {
233 	.open		= dcache_dir_open,
234 	.release	= dcache_dir_close,
235 	.llseek		= dcache_dir_lseek,
236 	.read		= generic_read_dir,
237 	.iterate_shared	= dcache_readdir,
238 	.fsync		= noop_fsync,
239 };
240 EXPORT_SYMBOL(simple_dir_operations);
241 
242 const struct inode_operations simple_dir_inode_operations = {
243 	.lookup		= simple_lookup,
244 };
245 EXPORT_SYMBOL(simple_dir_inode_operations);
246 
find_next_child(struct dentry * parent,struct dentry * prev)247 static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev)
248 {
249 	struct dentry *child = NULL;
250 	struct list_head *p = prev ? &prev->d_child : &parent->d_subdirs;
251 
252 	spin_lock(&parent->d_lock);
253 	while ((p = p->next) != &parent->d_subdirs) {
254 		struct dentry *d = container_of(p, struct dentry, d_child);
255 		if (simple_positive(d)) {
256 			spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
257 			if (simple_positive(d))
258 				child = dget_dlock(d);
259 			spin_unlock(&d->d_lock);
260 			if (likely(child))
261 				break;
262 		}
263 	}
264 	spin_unlock(&parent->d_lock);
265 	dput(prev);
266 	return child;
267 }
268 
simple_recursive_removal(struct dentry * dentry,void (* callback)(struct dentry *))269 void simple_recursive_removal(struct dentry *dentry,
270                               void (*callback)(struct dentry *))
271 {
272 	struct dentry *this = dget(dentry);
273 	while (true) {
274 		struct dentry *victim = NULL, *child;
275 		struct inode *inode = this->d_inode;
276 
277 		inode_lock(inode);
278 		if (d_is_dir(this))
279 			inode->i_flags |= S_DEAD;
280 		while ((child = find_next_child(this, victim)) == NULL) {
281 			// kill and ascend
282 			// update metadata while it's still locked
283 			inode->i_ctime = current_time(inode);
284 			clear_nlink(inode);
285 			inode_unlock(inode);
286 			victim = this;
287 			this = this->d_parent;
288 			inode = this->d_inode;
289 			inode_lock(inode);
290 			if (simple_positive(victim)) {
291 				d_invalidate(victim);	// avoid lost mounts
292 				if (d_is_dir(victim))
293 					fsnotify_rmdir(inode, victim);
294 				else
295 					fsnotify_unlink(inode, victim);
296 				if (callback)
297 					callback(victim);
298 				dput(victim);		// unpin it
299 			}
300 			if (victim == dentry) {
301 				inode->i_ctime = inode->i_mtime =
302 					current_time(inode);
303 				if (d_is_dir(dentry))
304 					drop_nlink(inode);
305 				inode_unlock(inode);
306 				dput(dentry);
307 				return;
308 			}
309 		}
310 		inode_unlock(inode);
311 		this = child;
312 	}
313 }
314 EXPORT_SYMBOL(simple_recursive_removal);
315 
316 static const struct super_operations simple_super_operations = {
317 	.statfs		= simple_statfs,
318 };
319 
pseudo_fs_fill_super(struct super_block * s,struct fs_context * fc)320 static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc)
321 {
322 	struct pseudo_fs_context *ctx = fc->fs_private;
323 	struct inode *root;
324 
325 	s->s_maxbytes = MAX_LFS_FILESIZE;
326 	s->s_blocksize = PAGE_SIZE;
327 	s->s_blocksize_bits = PAGE_SHIFT;
328 	s->s_magic = ctx->magic;
329 	s->s_op = ctx->ops ?: &simple_super_operations;
330 	s->s_xattr = ctx->xattr;
331 	s->s_time_gran = 1;
332 	root = new_inode(s);
333 	if (!root)
334 		return -ENOMEM;
335 
336 	/*
337 	 * since this is the first inode, make it number 1. New inodes created
338 	 * after this must take care not to collide with it (by passing
339 	 * max_reserved of 1 to iunique).
340 	 */
341 	root->i_ino = 1;
342 	root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
343 	root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
344 	s->s_root = d_make_root(root);
345 	if (!s->s_root)
346 		return -ENOMEM;
347 	s->s_d_op = ctx->dops;
348 	return 0;
349 }
350 
pseudo_fs_get_tree(struct fs_context * fc)351 static int pseudo_fs_get_tree(struct fs_context *fc)
352 {
353 	return get_tree_nodev(fc, pseudo_fs_fill_super);
354 }
355 
pseudo_fs_free(struct fs_context * fc)356 static void pseudo_fs_free(struct fs_context *fc)
357 {
358 	kfree(fc->fs_private);
359 }
360 
361 static const struct fs_context_operations pseudo_fs_context_ops = {
362 	.free		= pseudo_fs_free,
363 	.get_tree	= pseudo_fs_get_tree,
364 };
365 
366 /*
367  * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
368  * will never be mountable)
369  */
init_pseudo(struct fs_context * fc,unsigned long magic)370 struct pseudo_fs_context *init_pseudo(struct fs_context *fc,
371 					unsigned long magic)
372 {
373 	struct pseudo_fs_context *ctx;
374 
375 	ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL);
376 	if (likely(ctx)) {
377 		ctx->magic = magic;
378 		fc->fs_private = ctx;
379 		fc->ops = &pseudo_fs_context_ops;
380 		fc->sb_flags |= SB_NOUSER;
381 		fc->global = true;
382 	}
383 	return ctx;
384 }
385 EXPORT_SYMBOL(init_pseudo);
386 
simple_open(struct inode * inode,struct file * file)387 int simple_open(struct inode *inode, struct file *file)
388 {
389 	if (inode->i_private)
390 		file->private_data = inode->i_private;
391 	return 0;
392 }
393 EXPORT_SYMBOL(simple_open);
394 
simple_link(struct dentry * old_dentry,struct inode * dir,struct dentry * dentry)395 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
396 {
397 	struct inode *inode = d_inode(old_dentry);
398 
399 	inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
400 	inc_nlink(inode);
401 	ihold(inode);
402 	dget(dentry);
403 	d_instantiate(dentry, inode);
404 	return 0;
405 }
406 EXPORT_SYMBOL(simple_link);
407 
simple_empty(struct dentry * dentry)408 int simple_empty(struct dentry *dentry)
409 {
410 	struct dentry *child;
411 	int ret = 0;
412 
413 	spin_lock(&dentry->d_lock);
414 	list_for_each_entry(child, &dentry->d_subdirs, d_child) {
415 		spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
416 		if (simple_positive(child)) {
417 			spin_unlock(&child->d_lock);
418 			goto out;
419 		}
420 		spin_unlock(&child->d_lock);
421 	}
422 	ret = 1;
423 out:
424 	spin_unlock(&dentry->d_lock);
425 	return ret;
426 }
427 EXPORT_SYMBOL(simple_empty);
428 
simple_unlink(struct inode * dir,struct dentry * dentry)429 int simple_unlink(struct inode *dir, struct dentry *dentry)
430 {
431 	struct inode *inode = d_inode(dentry);
432 
433 	inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
434 	drop_nlink(inode);
435 	dput(dentry);
436 	return 0;
437 }
438 EXPORT_SYMBOL(simple_unlink);
439 
simple_rmdir(struct inode * dir,struct dentry * dentry)440 int simple_rmdir(struct inode *dir, struct dentry *dentry)
441 {
442 	if (!simple_empty(dentry))
443 		return -ENOTEMPTY;
444 
445 	drop_nlink(d_inode(dentry));
446 	simple_unlink(dir, dentry);
447 	drop_nlink(dir);
448 	return 0;
449 }
450 EXPORT_SYMBOL(simple_rmdir);
451 
simple_rename_exchange(struct inode * old_dir,struct dentry * old_dentry,struct inode * new_dir,struct dentry * new_dentry)452 int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
453 			   struct inode *new_dir, struct dentry *new_dentry)
454 {
455 	bool old_is_dir = d_is_dir(old_dentry);
456 	bool new_is_dir = d_is_dir(new_dentry);
457 
458 	if (old_dir != new_dir && old_is_dir != new_is_dir) {
459 		if (old_is_dir) {
460 			drop_nlink(old_dir);
461 			inc_nlink(new_dir);
462 		} else {
463 			drop_nlink(new_dir);
464 			inc_nlink(old_dir);
465 		}
466 	}
467 	old_dir->i_ctime = old_dir->i_mtime =
468 	new_dir->i_ctime = new_dir->i_mtime =
469 	d_inode(old_dentry)->i_ctime =
470 	d_inode(new_dentry)->i_ctime = current_time(old_dir);
471 
472 	return 0;
473 }
474 EXPORT_SYMBOL_GPL(simple_rename_exchange);
475 
simple_rename(struct user_namespace * mnt_userns,struct inode * old_dir,struct dentry * old_dentry,struct inode * new_dir,struct dentry * new_dentry,unsigned int flags)476 int simple_rename(struct user_namespace *mnt_userns, struct inode *old_dir,
477 		  struct dentry *old_dentry, struct inode *new_dir,
478 		  struct dentry *new_dentry, unsigned int flags)
479 {
480 	struct inode *inode = d_inode(old_dentry);
481 	int they_are_dirs = d_is_dir(old_dentry);
482 
483 	if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE))
484 		return -EINVAL;
485 
486 	if (flags & RENAME_EXCHANGE)
487 		return simple_rename_exchange(old_dir, old_dentry, new_dir, new_dentry);
488 
489 	if (!simple_empty(new_dentry))
490 		return -ENOTEMPTY;
491 
492 	if (d_really_is_positive(new_dentry)) {
493 		simple_unlink(new_dir, new_dentry);
494 		if (they_are_dirs) {
495 			drop_nlink(d_inode(new_dentry));
496 			drop_nlink(old_dir);
497 		}
498 	} else if (they_are_dirs) {
499 		drop_nlink(old_dir);
500 		inc_nlink(new_dir);
501 	}
502 
503 	old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
504 		new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
505 
506 	return 0;
507 }
508 EXPORT_SYMBOL(simple_rename);
509 
510 /**
511  * simple_setattr - setattr for simple filesystem
512  * @mnt_userns: user namespace of the target mount
513  * @dentry: dentry
514  * @iattr: iattr structure
515  *
516  * Returns 0 on success, -error on failure.
517  *
518  * simple_setattr is a simple ->setattr implementation without a proper
519  * implementation of size changes.
520  *
521  * It can either be used for in-memory filesystems or special files
522  * on simple regular filesystems.  Anything that needs to change on-disk
523  * or wire state on size changes needs its own setattr method.
524  */
simple_setattr(struct user_namespace * mnt_userns,struct dentry * dentry,struct iattr * iattr)525 int simple_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
526 		   struct iattr *iattr)
527 {
528 	struct inode *inode = d_inode(dentry);
529 	int error;
530 
531 	error = setattr_prepare(mnt_userns, dentry, iattr);
532 	if (error)
533 		return error;
534 
535 	if (iattr->ia_valid & ATTR_SIZE)
536 		truncate_setsize(inode, iattr->ia_size);
537 	setattr_copy(mnt_userns, inode, iattr);
538 	mark_inode_dirty(inode);
539 	return 0;
540 }
541 EXPORT_SYMBOL(simple_setattr);
542 
simple_read_folio(struct file * file,struct folio * folio)543 static int simple_read_folio(struct file *file, struct folio *folio)
544 {
545 	folio_zero_range(folio, 0, folio_size(folio));
546 	flush_dcache_folio(folio);
547 	folio_mark_uptodate(folio);
548 	folio_unlock(folio);
549 	return 0;
550 }
551 
simple_write_begin(struct file * file,struct address_space * mapping,loff_t pos,unsigned len,struct page ** pagep,void ** fsdata)552 int simple_write_begin(struct file *file, struct address_space *mapping,
553 			loff_t pos, unsigned len,
554 			struct page **pagep, void **fsdata)
555 {
556 	struct page *page;
557 	pgoff_t index;
558 
559 	index = pos >> PAGE_SHIFT;
560 
561 	page = grab_cache_page_write_begin(mapping, index);
562 	if (!page)
563 		return -ENOMEM;
564 
565 	*pagep = page;
566 
567 	if (!PageUptodate(page) && (len != PAGE_SIZE)) {
568 		unsigned from = pos & (PAGE_SIZE - 1);
569 
570 		zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
571 	}
572 	return 0;
573 }
574 EXPORT_SYMBOL(simple_write_begin);
575 
576 /**
577  * simple_write_end - .write_end helper for non-block-device FSes
578  * @file: See .write_end of address_space_operations
579  * @mapping: 		"
580  * @pos: 		"
581  * @len: 		"
582  * @copied: 		"
583  * @page: 		"
584  * @fsdata: 		"
585  *
586  * simple_write_end does the minimum needed for updating a page after writing is
587  * done. It has the same API signature as the .write_end of
588  * address_space_operations vector. So it can just be set onto .write_end for
589  * FSes that don't need any other processing. i_mutex is assumed to be held.
590  * Block based filesystems should use generic_write_end().
591  * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
592  * is not called, so a filesystem that actually does store data in .write_inode
593  * should extend on what's done here with a call to mark_inode_dirty() in the
594  * case that i_size has changed.
595  *
596  * Use *ONLY* with simple_read_folio()
597  */
simple_write_end(struct file * file,struct address_space * mapping,loff_t pos,unsigned len,unsigned copied,struct page * page,void * fsdata)598 static int simple_write_end(struct file *file, struct address_space *mapping,
599 			loff_t pos, unsigned len, unsigned copied,
600 			struct page *page, void *fsdata)
601 {
602 	struct inode *inode = page->mapping->host;
603 	loff_t last_pos = pos + copied;
604 
605 	/* zero the stale part of the page if we did a short copy */
606 	if (!PageUptodate(page)) {
607 		if (copied < len) {
608 			unsigned from = pos & (PAGE_SIZE - 1);
609 
610 			zero_user(page, from + copied, len - copied);
611 		}
612 		SetPageUptodate(page);
613 	}
614 	/*
615 	 * No need to use i_size_read() here, the i_size
616 	 * cannot change under us because we hold the i_mutex.
617 	 */
618 	if (last_pos > inode->i_size)
619 		i_size_write(inode, last_pos);
620 
621 	set_page_dirty(page);
622 	unlock_page(page);
623 	put_page(page);
624 
625 	return copied;
626 }
627 
628 /*
629  * Provides ramfs-style behavior: data in the pagecache, but no writeback.
630  */
631 const struct address_space_operations ram_aops = {
632 	.read_folio	= simple_read_folio,
633 	.write_begin	= simple_write_begin,
634 	.write_end	= simple_write_end,
635 	.dirty_folio	= noop_dirty_folio,
636 };
637 EXPORT_SYMBOL(ram_aops);
638 
639 /*
640  * the inodes created here are not hashed. If you use iunique to generate
641  * unique inode values later for this filesystem, then you must take care
642  * to pass it an appropriate max_reserved value to avoid collisions.
643  */
simple_fill_super(struct super_block * s,unsigned long magic,const struct tree_descr * files)644 int simple_fill_super(struct super_block *s, unsigned long magic,
645 		      const struct tree_descr *files)
646 {
647 	struct inode *inode;
648 	struct dentry *root;
649 	struct dentry *dentry;
650 	int i;
651 
652 	s->s_blocksize = PAGE_SIZE;
653 	s->s_blocksize_bits = PAGE_SHIFT;
654 	s->s_magic = magic;
655 	s->s_op = &simple_super_operations;
656 	s->s_time_gran = 1;
657 
658 	inode = new_inode(s);
659 	if (!inode)
660 		return -ENOMEM;
661 	/*
662 	 * because the root inode is 1, the files array must not contain an
663 	 * entry at index 1
664 	 */
665 	inode->i_ino = 1;
666 	inode->i_mode = S_IFDIR | 0755;
667 	inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
668 	inode->i_op = &simple_dir_inode_operations;
669 	inode->i_fop = &simple_dir_operations;
670 	set_nlink(inode, 2);
671 	root = d_make_root(inode);
672 	if (!root)
673 		return -ENOMEM;
674 	for (i = 0; !files->name || files->name[0]; i++, files++) {
675 		if (!files->name)
676 			continue;
677 
678 		/* warn if it tries to conflict with the root inode */
679 		if (unlikely(i == 1))
680 			printk(KERN_WARNING "%s: %s passed in a files array"
681 				"with an index of 1!\n", __func__,
682 				s->s_type->name);
683 
684 		dentry = d_alloc_name(root, files->name);
685 		if (!dentry)
686 			goto out;
687 		inode = new_inode(s);
688 		if (!inode) {
689 			dput(dentry);
690 			goto out;
691 		}
692 		inode->i_mode = S_IFREG | files->mode;
693 		inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
694 		inode->i_fop = files->ops;
695 		inode->i_ino = i;
696 		d_add(dentry, inode);
697 	}
698 	s->s_root = root;
699 	return 0;
700 out:
701 	d_genocide(root);
702 	shrink_dcache_parent(root);
703 	dput(root);
704 	return -ENOMEM;
705 }
706 EXPORT_SYMBOL(simple_fill_super);
707 
708 static DEFINE_SPINLOCK(pin_fs_lock);
709 
simple_pin_fs(struct file_system_type * type,struct vfsmount ** mount,int * count)710 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
711 {
712 	struct vfsmount *mnt = NULL;
713 	spin_lock(&pin_fs_lock);
714 	if (unlikely(!*mount)) {
715 		spin_unlock(&pin_fs_lock);
716 		mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
717 		if (IS_ERR(mnt))
718 			return PTR_ERR(mnt);
719 		spin_lock(&pin_fs_lock);
720 		if (!*mount)
721 			*mount = mnt;
722 	}
723 	mntget(*mount);
724 	++*count;
725 	spin_unlock(&pin_fs_lock);
726 	mntput(mnt);
727 	return 0;
728 }
729 EXPORT_SYMBOL(simple_pin_fs);
730 
simple_release_fs(struct vfsmount ** mount,int * count)731 void simple_release_fs(struct vfsmount **mount, int *count)
732 {
733 	struct vfsmount *mnt;
734 	spin_lock(&pin_fs_lock);
735 	mnt = *mount;
736 	if (!--*count)
737 		*mount = NULL;
738 	spin_unlock(&pin_fs_lock);
739 	mntput(mnt);
740 }
741 EXPORT_SYMBOL(simple_release_fs);
742 
743 /**
744  * simple_read_from_buffer - copy data from the buffer to user space
745  * @to: the user space buffer to read to
746  * @count: the maximum number of bytes to read
747  * @ppos: the current position in the buffer
748  * @from: the buffer to read from
749  * @available: the size of the buffer
750  *
751  * The simple_read_from_buffer() function reads up to @count bytes from the
752  * buffer @from at offset @ppos into the user space address starting at @to.
753  *
754  * On success, the number of bytes read is returned and the offset @ppos is
755  * advanced by this number, or negative value is returned on error.
756  **/
simple_read_from_buffer(void __user * to,size_t count,loff_t * ppos,const void * from,size_t available)757 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
758 				const void *from, size_t available)
759 {
760 	loff_t pos = *ppos;
761 	size_t ret;
762 
763 	if (pos < 0)
764 		return -EINVAL;
765 	if (pos >= available || !count)
766 		return 0;
767 	if (count > available - pos)
768 		count = available - pos;
769 	ret = copy_to_user(to, from + pos, count);
770 	if (ret == count)
771 		return -EFAULT;
772 	count -= ret;
773 	*ppos = pos + count;
774 	return count;
775 }
776 EXPORT_SYMBOL(simple_read_from_buffer);
777 
778 /**
779  * simple_write_to_buffer - copy data from user space to the buffer
780  * @to: the buffer to write to
781  * @available: the size of the buffer
782  * @ppos: the current position in the buffer
783  * @from: the user space buffer to read from
784  * @count: the maximum number of bytes to read
785  *
786  * The simple_write_to_buffer() function reads up to @count bytes from the user
787  * space address starting at @from into the buffer @to at offset @ppos.
788  *
789  * On success, the number of bytes written is returned and the offset @ppos is
790  * advanced by this number, or negative value is returned on error.
791  **/
simple_write_to_buffer(void * to,size_t available,loff_t * ppos,const void __user * from,size_t count)792 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
793 		const void __user *from, size_t count)
794 {
795 	loff_t pos = *ppos;
796 	size_t res;
797 
798 	if (pos < 0)
799 		return -EINVAL;
800 	if (pos >= available || !count)
801 		return 0;
802 	if (count > available - pos)
803 		count = available - pos;
804 	res = copy_from_user(to + pos, from, count);
805 	if (res == count)
806 		return -EFAULT;
807 	count -= res;
808 	*ppos = pos + count;
809 	return count;
810 }
811 EXPORT_SYMBOL(simple_write_to_buffer);
812 
813 /**
814  * memory_read_from_buffer - copy data from the buffer
815  * @to: the kernel space buffer to read to
816  * @count: the maximum number of bytes to read
817  * @ppos: the current position in the buffer
818  * @from: the buffer to read from
819  * @available: the size of the buffer
820  *
821  * The memory_read_from_buffer() function reads up to @count bytes from the
822  * buffer @from at offset @ppos into the kernel space address starting at @to.
823  *
824  * On success, the number of bytes read is returned and the offset @ppos is
825  * advanced by this number, or negative value is returned on error.
826  **/
memory_read_from_buffer(void * to,size_t count,loff_t * ppos,const void * from,size_t available)827 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
828 				const void *from, size_t available)
829 {
830 	loff_t pos = *ppos;
831 
832 	if (pos < 0)
833 		return -EINVAL;
834 	if (pos >= available)
835 		return 0;
836 	if (count > available - pos)
837 		count = available - pos;
838 	memcpy(to, from + pos, count);
839 	*ppos = pos + count;
840 
841 	return count;
842 }
843 EXPORT_SYMBOL(memory_read_from_buffer);
844 
845 /*
846  * Transaction based IO.
847  * The file expects a single write which triggers the transaction, and then
848  * possibly a read which collects the result - which is stored in a
849  * file-local buffer.
850  */
851 
simple_transaction_set(struct file * file,size_t n)852 void simple_transaction_set(struct file *file, size_t n)
853 {
854 	struct simple_transaction_argresp *ar = file->private_data;
855 
856 	BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
857 
858 	/*
859 	 * The barrier ensures that ar->size will really remain zero until
860 	 * ar->data is ready for reading.
861 	 */
862 	smp_mb();
863 	ar->size = n;
864 }
865 EXPORT_SYMBOL(simple_transaction_set);
866 
simple_transaction_get(struct file * file,const char __user * buf,size_t size)867 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
868 {
869 	struct simple_transaction_argresp *ar;
870 	static DEFINE_SPINLOCK(simple_transaction_lock);
871 
872 	if (size > SIMPLE_TRANSACTION_LIMIT - 1)
873 		return ERR_PTR(-EFBIG);
874 
875 	ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
876 	if (!ar)
877 		return ERR_PTR(-ENOMEM);
878 
879 	spin_lock(&simple_transaction_lock);
880 
881 	/* only one write allowed per open */
882 	if (file->private_data) {
883 		spin_unlock(&simple_transaction_lock);
884 		free_page((unsigned long)ar);
885 		return ERR_PTR(-EBUSY);
886 	}
887 
888 	file->private_data = ar;
889 
890 	spin_unlock(&simple_transaction_lock);
891 
892 	if (copy_from_user(ar->data, buf, size))
893 		return ERR_PTR(-EFAULT);
894 
895 	return ar->data;
896 }
897 EXPORT_SYMBOL(simple_transaction_get);
898 
simple_transaction_read(struct file * file,char __user * buf,size_t size,loff_t * pos)899 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
900 {
901 	struct simple_transaction_argresp *ar = file->private_data;
902 
903 	if (!ar)
904 		return 0;
905 	return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
906 }
907 EXPORT_SYMBOL(simple_transaction_read);
908 
simple_transaction_release(struct inode * inode,struct file * file)909 int simple_transaction_release(struct inode *inode, struct file *file)
910 {
911 	free_page((unsigned long)file->private_data);
912 	return 0;
913 }
914 EXPORT_SYMBOL(simple_transaction_release);
915 
916 /* Simple attribute files */
917 
918 struct simple_attr {
919 	int (*get)(void *, u64 *);
920 	int (*set)(void *, u64);
921 	char get_buf[24];	/* enough to store a u64 and "\n\0" */
922 	char set_buf[24];
923 	void *data;
924 	const char *fmt;	/* format for read operation */
925 	struct mutex mutex;	/* protects access to these buffers */
926 };
927 
928 /* simple_attr_open is called by an actual attribute open file operation
929  * to set the attribute specific access operations. */
simple_attr_open(struct inode * inode,struct file * file,int (* get)(void *,u64 *),int (* set)(void *,u64),const char * fmt)930 int simple_attr_open(struct inode *inode, struct file *file,
931 		     int (*get)(void *, u64 *), int (*set)(void *, u64),
932 		     const char *fmt)
933 {
934 	struct simple_attr *attr;
935 
936 	attr = kzalloc(sizeof(*attr), GFP_KERNEL);
937 	if (!attr)
938 		return -ENOMEM;
939 
940 	attr->get = get;
941 	attr->set = set;
942 	attr->data = inode->i_private;
943 	attr->fmt = fmt;
944 	mutex_init(&attr->mutex);
945 
946 	file->private_data = attr;
947 
948 	return nonseekable_open(inode, file);
949 }
950 EXPORT_SYMBOL_GPL(simple_attr_open);
951 
simple_attr_release(struct inode * inode,struct file * file)952 int simple_attr_release(struct inode *inode, struct file *file)
953 {
954 	kfree(file->private_data);
955 	return 0;
956 }
957 EXPORT_SYMBOL_GPL(simple_attr_release);	/* GPL-only?  This?  Really? */
958 
959 /* read from the buffer that is filled with the get function */
simple_attr_read(struct file * file,char __user * buf,size_t len,loff_t * ppos)960 ssize_t simple_attr_read(struct file *file, char __user *buf,
961 			 size_t len, loff_t *ppos)
962 {
963 	struct simple_attr *attr;
964 	size_t size;
965 	ssize_t ret;
966 
967 	attr = file->private_data;
968 
969 	if (!attr->get)
970 		return -EACCES;
971 
972 	ret = mutex_lock_interruptible(&attr->mutex);
973 	if (ret)
974 		return ret;
975 
976 	if (*ppos && attr->get_buf[0]) {
977 		/* continued read */
978 		size = strlen(attr->get_buf);
979 	} else {
980 		/* first read */
981 		u64 val;
982 		ret = attr->get(attr->data, &val);
983 		if (ret)
984 			goto out;
985 
986 		size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
987 				 attr->fmt, (unsigned long long)val);
988 	}
989 
990 	ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
991 out:
992 	mutex_unlock(&attr->mutex);
993 	return ret;
994 }
995 EXPORT_SYMBOL_GPL(simple_attr_read);
996 
997 /* interpret the buffer as a number to call the set function with */
simple_attr_write_xsigned(struct file * file,const char __user * buf,size_t len,loff_t * ppos,bool is_signed)998 static ssize_t simple_attr_write_xsigned(struct file *file, const char __user *buf,
999 			  size_t len, loff_t *ppos, bool is_signed)
1000 {
1001 	struct simple_attr *attr;
1002 	unsigned long long val;
1003 	size_t size;
1004 	ssize_t ret;
1005 
1006 	attr = file->private_data;
1007 	if (!attr->set)
1008 		return -EACCES;
1009 
1010 	ret = mutex_lock_interruptible(&attr->mutex);
1011 	if (ret)
1012 		return ret;
1013 
1014 	ret = -EFAULT;
1015 	size = min(sizeof(attr->set_buf) - 1, len);
1016 	if (copy_from_user(attr->set_buf, buf, size))
1017 		goto out;
1018 
1019 	attr->set_buf[size] = '\0';
1020 	if (is_signed)
1021 		ret = kstrtoll(attr->set_buf, 0, &val);
1022 	else
1023 		ret = kstrtoull(attr->set_buf, 0, &val);
1024 	if (ret)
1025 		goto out;
1026 	ret = attr->set(attr->data, val);
1027 	if (ret == 0)
1028 		ret = len; /* on success, claim we got the whole input */
1029 out:
1030 	mutex_unlock(&attr->mutex);
1031 	return ret;
1032 }
1033 
simple_attr_write(struct file * file,const char __user * buf,size_t len,loff_t * ppos)1034 ssize_t simple_attr_write(struct file *file, const char __user *buf,
1035 			  size_t len, loff_t *ppos)
1036 {
1037 	return simple_attr_write_xsigned(file, buf, len, ppos, false);
1038 }
1039 EXPORT_SYMBOL_GPL(simple_attr_write);
1040 
simple_attr_write_signed(struct file * file,const char __user * buf,size_t len,loff_t * ppos)1041 ssize_t simple_attr_write_signed(struct file *file, const char __user *buf,
1042 			  size_t len, loff_t *ppos)
1043 {
1044 	return simple_attr_write_xsigned(file, buf, len, ppos, true);
1045 }
1046 EXPORT_SYMBOL_GPL(simple_attr_write_signed);
1047 
1048 /**
1049  * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
1050  * @sb:		filesystem to do the file handle conversion on
1051  * @fid:	file handle to convert
1052  * @fh_len:	length of the file handle in bytes
1053  * @fh_type:	type of file handle
1054  * @get_inode:	filesystem callback to retrieve inode
1055  *
1056  * This function decodes @fid as long as it has one of the well-known
1057  * Linux filehandle types and calls @get_inode on it to retrieve the
1058  * inode for the object specified in the file handle.
1059  */
generic_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type,struct inode * (* get_inode)(struct super_block * sb,u64 ino,u32 gen))1060 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
1061 		int fh_len, int fh_type, struct inode *(*get_inode)
1062 			(struct super_block *sb, u64 ino, u32 gen))
1063 {
1064 	struct inode *inode = NULL;
1065 
1066 	if (fh_len < 2)
1067 		return NULL;
1068 
1069 	switch (fh_type) {
1070 	case FILEID_INO32_GEN:
1071 	case FILEID_INO32_GEN_PARENT:
1072 		inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
1073 		break;
1074 	}
1075 
1076 	return d_obtain_alias(inode);
1077 }
1078 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
1079 
1080 /**
1081  * generic_fh_to_parent - generic helper for the fh_to_parent export operation
1082  * @sb:		filesystem to do the file handle conversion on
1083  * @fid:	file handle to convert
1084  * @fh_len:	length of the file handle in bytes
1085  * @fh_type:	type of file handle
1086  * @get_inode:	filesystem callback to retrieve inode
1087  *
1088  * This function decodes @fid as long as it has one of the well-known
1089  * Linux filehandle types and calls @get_inode on it to retrieve the
1090  * inode for the _parent_ object specified in the file handle if it
1091  * is specified in the file handle, or NULL otherwise.
1092  */
generic_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type,struct inode * (* get_inode)(struct super_block * sb,u64 ino,u32 gen))1093 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
1094 		int fh_len, int fh_type, struct inode *(*get_inode)
1095 			(struct super_block *sb, u64 ino, u32 gen))
1096 {
1097 	struct inode *inode = NULL;
1098 
1099 	if (fh_len <= 2)
1100 		return NULL;
1101 
1102 	switch (fh_type) {
1103 	case FILEID_INO32_GEN_PARENT:
1104 		inode = get_inode(sb, fid->i32.parent_ino,
1105 				  (fh_len > 3 ? fid->i32.parent_gen : 0));
1106 		break;
1107 	}
1108 
1109 	return d_obtain_alias(inode);
1110 }
1111 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
1112 
1113 /**
1114  * __generic_file_fsync - generic fsync implementation for simple filesystems
1115  *
1116  * @file:	file to synchronize
1117  * @start:	start offset in bytes
1118  * @end:	end offset in bytes (inclusive)
1119  * @datasync:	only synchronize essential metadata if true
1120  *
1121  * This is a generic implementation of the fsync method for simple
1122  * filesystems which track all non-inode metadata in the buffers list
1123  * hanging off the address_space structure.
1124  */
__generic_file_fsync(struct file * file,loff_t start,loff_t end,int datasync)1125 int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
1126 				 int datasync)
1127 {
1128 	struct inode *inode = file->f_mapping->host;
1129 	int err;
1130 	int ret;
1131 
1132 	err = file_write_and_wait_range(file, start, end);
1133 	if (err)
1134 		return err;
1135 
1136 	inode_lock(inode);
1137 	ret = sync_mapping_buffers(inode->i_mapping);
1138 	if (!(inode->i_state & I_DIRTY_ALL))
1139 		goto out;
1140 	if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
1141 		goto out;
1142 
1143 	err = sync_inode_metadata(inode, 1);
1144 	if (ret == 0)
1145 		ret = err;
1146 
1147 out:
1148 	inode_unlock(inode);
1149 	/* check and advance again to catch errors after syncing out buffers */
1150 	err = file_check_and_advance_wb_err(file);
1151 	if (ret == 0)
1152 		ret = err;
1153 	return ret;
1154 }
1155 EXPORT_SYMBOL(__generic_file_fsync);
1156 
1157 /**
1158  * generic_file_fsync - generic fsync implementation for simple filesystems
1159  *			with flush
1160  * @file:	file to synchronize
1161  * @start:	start offset in bytes
1162  * @end:	end offset in bytes (inclusive)
1163  * @datasync:	only synchronize essential metadata if true
1164  *
1165  */
1166 
generic_file_fsync(struct file * file,loff_t start,loff_t end,int datasync)1167 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1168 		       int datasync)
1169 {
1170 	struct inode *inode = file->f_mapping->host;
1171 	int err;
1172 
1173 	err = __generic_file_fsync(file, start, end, datasync);
1174 	if (err)
1175 		return err;
1176 	return blkdev_issue_flush(inode->i_sb->s_bdev);
1177 }
1178 EXPORT_SYMBOL(generic_file_fsync);
1179 
1180 /**
1181  * generic_check_addressable - Check addressability of file system
1182  * @blocksize_bits:	log of file system block size
1183  * @num_blocks:		number of blocks in file system
1184  *
1185  * Determine whether a file system with @num_blocks blocks (and a
1186  * block size of 2**@blocksize_bits) is addressable by the sector_t
1187  * and page cache of the system.  Return 0 if so and -EFBIG otherwise.
1188  */
generic_check_addressable(unsigned blocksize_bits,u64 num_blocks)1189 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1190 {
1191 	u64 last_fs_block = num_blocks - 1;
1192 	u64 last_fs_page =
1193 		last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1194 
1195 	if (unlikely(num_blocks == 0))
1196 		return 0;
1197 
1198 	if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1199 		return -EINVAL;
1200 
1201 	if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1202 	    (last_fs_page > (pgoff_t)(~0ULL))) {
1203 		return -EFBIG;
1204 	}
1205 	return 0;
1206 }
1207 EXPORT_SYMBOL(generic_check_addressable);
1208 
1209 /*
1210  * No-op implementation of ->fsync for in-memory filesystems.
1211  */
noop_fsync(struct file * file,loff_t start,loff_t end,int datasync)1212 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1213 {
1214 	return 0;
1215 }
1216 EXPORT_SYMBOL(noop_fsync);
1217 
noop_direct_IO(struct kiocb * iocb,struct iov_iter * iter)1218 ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1219 {
1220 	/*
1221 	 * iomap based filesystems support direct I/O without need for
1222 	 * this callback. However, it still needs to be set in
1223 	 * inode->a_ops so that open/fcntl know that direct I/O is
1224 	 * generally supported.
1225 	 */
1226 	return -EINVAL;
1227 }
1228 EXPORT_SYMBOL_GPL(noop_direct_IO);
1229 
1230 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
kfree_link(void * p)1231 void kfree_link(void *p)
1232 {
1233 	kfree(p);
1234 }
1235 EXPORT_SYMBOL(kfree_link);
1236 
alloc_anon_inode(struct super_block * s)1237 struct inode *alloc_anon_inode(struct super_block *s)
1238 {
1239 	static const struct address_space_operations anon_aops = {
1240 		.dirty_folio	= noop_dirty_folio,
1241 	};
1242 	struct inode *inode = new_inode_pseudo(s);
1243 
1244 	if (!inode)
1245 		return ERR_PTR(-ENOMEM);
1246 
1247 	inode->i_ino = get_next_ino();
1248 	inode->i_mapping->a_ops = &anon_aops;
1249 
1250 	/*
1251 	 * Mark the inode dirty from the very beginning,
1252 	 * that way it will never be moved to the dirty
1253 	 * list because mark_inode_dirty() will think
1254 	 * that it already _is_ on the dirty list.
1255 	 */
1256 	inode->i_state = I_DIRTY;
1257 	inode->i_mode = S_IRUSR | S_IWUSR;
1258 	inode->i_uid = current_fsuid();
1259 	inode->i_gid = current_fsgid();
1260 	inode->i_flags |= S_PRIVATE;
1261 	inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1262 	return inode;
1263 }
1264 EXPORT_SYMBOL(alloc_anon_inode);
1265 
1266 /**
1267  * simple_nosetlease - generic helper for prohibiting leases
1268  * @filp: file pointer
1269  * @arg: type of lease to obtain
1270  * @flp: new lease supplied for insertion
1271  * @priv: private data for lm_setup operation
1272  *
1273  * Generic helper for filesystems that do not wish to allow leases to be set.
1274  * All arguments are ignored and it just returns -EINVAL.
1275  */
1276 int
simple_nosetlease(struct file * filp,long arg,struct file_lock ** flp,void ** priv)1277 simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1278 		  void **priv)
1279 {
1280 	return -EINVAL;
1281 }
1282 EXPORT_SYMBOL(simple_nosetlease);
1283 
1284 /**
1285  * simple_get_link - generic helper to get the target of "fast" symlinks
1286  * @dentry: not used here
1287  * @inode: the symlink inode
1288  * @done: not used here
1289  *
1290  * Generic helper for filesystems to use for symlink inodes where a pointer to
1291  * the symlink target is stored in ->i_link.  NOTE: this isn't normally called,
1292  * since as an optimization the path lookup code uses any non-NULL ->i_link
1293  * directly, without calling ->get_link().  But ->get_link() still must be set,
1294  * to mark the inode_operations as being for a symlink.
1295  *
1296  * Return: the symlink target
1297  */
simple_get_link(struct dentry * dentry,struct inode * inode,struct delayed_call * done)1298 const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1299 			    struct delayed_call *done)
1300 {
1301 	return inode->i_link;
1302 }
1303 EXPORT_SYMBOL(simple_get_link);
1304 
1305 const struct inode_operations simple_symlink_inode_operations = {
1306 	.get_link = simple_get_link,
1307 };
1308 EXPORT_SYMBOL(simple_symlink_inode_operations);
1309 
1310 /*
1311  * Operations for a permanently empty directory.
1312  */
empty_dir_lookup(struct inode * dir,struct dentry * dentry,unsigned int flags)1313 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1314 {
1315 	return ERR_PTR(-ENOENT);
1316 }
1317 
empty_dir_getattr(struct user_namespace * mnt_userns,const struct path * path,struct kstat * stat,u32 request_mask,unsigned int query_flags)1318 static int empty_dir_getattr(struct user_namespace *mnt_userns,
1319 			     const struct path *path, struct kstat *stat,
1320 			     u32 request_mask, unsigned int query_flags)
1321 {
1322 	struct inode *inode = d_inode(path->dentry);
1323 	generic_fillattr(&init_user_ns, inode, stat);
1324 	return 0;
1325 }
1326 
empty_dir_setattr(struct user_namespace * mnt_userns,struct dentry * dentry,struct iattr * attr)1327 static int empty_dir_setattr(struct user_namespace *mnt_userns,
1328 			     struct dentry *dentry, struct iattr *attr)
1329 {
1330 	return -EPERM;
1331 }
1332 
empty_dir_listxattr(struct dentry * dentry,char * list,size_t size)1333 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1334 {
1335 	return -EOPNOTSUPP;
1336 }
1337 
1338 static const struct inode_operations empty_dir_inode_operations = {
1339 	.lookup		= empty_dir_lookup,
1340 	.permission	= generic_permission,
1341 	.setattr	= empty_dir_setattr,
1342 	.getattr	= empty_dir_getattr,
1343 	.listxattr	= empty_dir_listxattr,
1344 };
1345 
empty_dir_llseek(struct file * file,loff_t offset,int whence)1346 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1347 {
1348 	/* An empty directory has two entries . and .. at offsets 0 and 1 */
1349 	return generic_file_llseek_size(file, offset, whence, 2, 2);
1350 }
1351 
empty_dir_readdir(struct file * file,struct dir_context * ctx)1352 static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1353 {
1354 	dir_emit_dots(file, ctx);
1355 	return 0;
1356 }
1357 
1358 static const struct file_operations empty_dir_operations = {
1359 	.llseek		= empty_dir_llseek,
1360 	.read		= generic_read_dir,
1361 	.iterate_shared	= empty_dir_readdir,
1362 	.fsync		= noop_fsync,
1363 };
1364 
1365 
make_empty_dir_inode(struct inode * inode)1366 void make_empty_dir_inode(struct inode *inode)
1367 {
1368 	set_nlink(inode, 2);
1369 	inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1370 	inode->i_uid = GLOBAL_ROOT_UID;
1371 	inode->i_gid = GLOBAL_ROOT_GID;
1372 	inode->i_rdev = 0;
1373 	inode->i_size = 0;
1374 	inode->i_blkbits = PAGE_SHIFT;
1375 	inode->i_blocks = 0;
1376 
1377 	inode->i_op = &empty_dir_inode_operations;
1378 	inode->i_opflags &= ~IOP_XATTR;
1379 	inode->i_fop = &empty_dir_operations;
1380 }
1381 
is_empty_dir_inode(struct inode * inode)1382 bool is_empty_dir_inode(struct inode *inode)
1383 {
1384 	return (inode->i_fop == &empty_dir_operations) &&
1385 		(inode->i_op == &empty_dir_inode_operations);
1386 }
1387 
1388 #if IS_ENABLED(CONFIG_UNICODE)
1389 /*
1390  * Determine if the name of a dentry should be casefolded.
1391  *
1392  * Return: if names will need casefolding
1393  */
needs_casefold(const struct inode * dir)1394 static bool needs_casefold(const struct inode *dir)
1395 {
1396 	return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding;
1397 }
1398 
1399 /**
1400  * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
1401  * @dentry:	dentry whose name we are checking against
1402  * @len:	len of name of dentry
1403  * @str:	str pointer to name of dentry
1404  * @name:	Name to compare against
1405  *
1406  * Return: 0 if names match, 1 if mismatch, or -ERRNO
1407  */
generic_ci_d_compare(const struct dentry * dentry,unsigned int len,const char * str,const struct qstr * name)1408 static int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
1409 				const char *str, const struct qstr *name)
1410 {
1411 	const struct dentry *parent = READ_ONCE(dentry->d_parent);
1412 	const struct inode *dir = READ_ONCE(parent->d_inode);
1413 	const struct super_block *sb = dentry->d_sb;
1414 	const struct unicode_map *um = sb->s_encoding;
1415 	struct qstr qstr = QSTR_INIT(str, len);
1416 	char strbuf[DNAME_INLINE_LEN];
1417 	int ret;
1418 
1419 	if (!dir || !needs_casefold(dir))
1420 		goto fallback;
1421 	/*
1422 	 * If the dentry name is stored in-line, then it may be concurrently
1423 	 * modified by a rename.  If this happens, the VFS will eventually retry
1424 	 * the lookup, so it doesn't matter what ->d_compare() returns.
1425 	 * However, it's unsafe to call utf8_strncasecmp() with an unstable
1426 	 * string.  Therefore, we have to copy the name into a temporary buffer.
1427 	 */
1428 	if (len <= DNAME_INLINE_LEN - 1) {
1429 		memcpy(strbuf, str, len);
1430 		strbuf[len] = 0;
1431 		qstr.name = strbuf;
1432 		/* prevent compiler from optimizing out the temporary buffer */
1433 		barrier();
1434 	}
1435 	ret = utf8_strncasecmp(um, name, &qstr);
1436 	if (ret >= 0)
1437 		return ret;
1438 
1439 	if (sb_has_strict_encoding(sb))
1440 		return -EINVAL;
1441 fallback:
1442 	if (len != name->len)
1443 		return 1;
1444 	return !!memcmp(str, name->name, len);
1445 }
1446 
1447 /**
1448  * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
1449  * @dentry:	dentry of the parent directory
1450  * @str:	qstr of name whose hash we should fill in
1451  *
1452  * Return: 0 if hash was successful or unchanged, and -EINVAL on error
1453  */
generic_ci_d_hash(const struct dentry * dentry,struct qstr * str)1454 static int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str)
1455 {
1456 	const struct inode *dir = READ_ONCE(dentry->d_inode);
1457 	struct super_block *sb = dentry->d_sb;
1458 	const struct unicode_map *um = sb->s_encoding;
1459 	int ret = 0;
1460 
1461 	if (!dir || !needs_casefold(dir))
1462 		return 0;
1463 
1464 	ret = utf8_casefold_hash(um, dentry, str);
1465 	if (ret < 0 && sb_has_strict_encoding(sb))
1466 		return -EINVAL;
1467 	return 0;
1468 }
1469 
1470 static const struct dentry_operations generic_ci_dentry_ops = {
1471 	.d_hash = generic_ci_d_hash,
1472 	.d_compare = generic_ci_d_compare,
1473 };
1474 #endif
1475 
1476 #ifdef CONFIG_FS_ENCRYPTION
1477 static const struct dentry_operations generic_encrypted_dentry_ops = {
1478 	.d_revalidate = fscrypt_d_revalidate,
1479 };
1480 #endif
1481 
1482 #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
1483 static const struct dentry_operations generic_encrypted_ci_dentry_ops = {
1484 	.d_hash = generic_ci_d_hash,
1485 	.d_compare = generic_ci_d_compare,
1486 	.d_revalidate = fscrypt_d_revalidate,
1487 };
1488 #endif
1489 
1490 /**
1491  * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry
1492  * @dentry:	dentry to set ops on
1493  *
1494  * Casefolded directories need d_hash and d_compare set, so that the dentries
1495  * contained in them are handled case-insensitively.  Note that these operations
1496  * are needed on the parent directory rather than on the dentries in it, and
1497  * while the casefolding flag can be toggled on and off on an empty directory,
1498  * dentry_operations can't be changed later.  As a result, if the filesystem has
1499  * casefolding support enabled at all, we have to give all dentries the
1500  * casefolding operations even if their inode doesn't have the casefolding flag
1501  * currently (and thus the casefolding ops would be no-ops for now).
1502  *
1503  * Encryption works differently in that the only dentry operation it needs is
1504  * d_revalidate, which it only needs on dentries that have the no-key name flag.
1505  * The no-key flag can't be set "later", so we don't have to worry about that.
1506  *
1507  * Finally, to maximize compatibility with overlayfs (which isn't compatible
1508  * with certain dentry operations) and to avoid taking an unnecessary
1509  * performance hit, we use custom dentry_operations for each possible
1510  * combination rather than always installing all operations.
1511  */
generic_set_encrypted_ci_d_ops(struct dentry * dentry)1512 void generic_set_encrypted_ci_d_ops(struct dentry *dentry)
1513 {
1514 #ifdef CONFIG_FS_ENCRYPTION
1515 	bool needs_encrypt_ops = dentry->d_flags & DCACHE_NOKEY_NAME;
1516 #endif
1517 #if IS_ENABLED(CONFIG_UNICODE)
1518 	bool needs_ci_ops = dentry->d_sb->s_encoding;
1519 #endif
1520 #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
1521 	if (needs_encrypt_ops && needs_ci_ops) {
1522 		d_set_d_op(dentry, &generic_encrypted_ci_dentry_ops);
1523 		return;
1524 	}
1525 #endif
1526 #ifdef CONFIG_FS_ENCRYPTION
1527 	if (needs_encrypt_ops) {
1528 		d_set_d_op(dentry, &generic_encrypted_dentry_ops);
1529 		return;
1530 	}
1531 #endif
1532 #if IS_ENABLED(CONFIG_UNICODE)
1533 	if (needs_ci_ops) {
1534 		d_set_d_op(dentry, &generic_ci_dentry_ops);
1535 		return;
1536 	}
1537 #endif
1538 }
1539 EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops);
1540 
1541 /**
1542  * inode_maybe_inc_iversion - increments i_version
1543  * @inode: inode with the i_version that should be updated
1544  * @force: increment the counter even if it's not necessary?
1545  *
1546  * Every time the inode is modified, the i_version field must be seen to have
1547  * changed by any observer.
1548  *
1549  * If "force" is set or the QUERIED flag is set, then ensure that we increment
1550  * the value, and clear the queried flag.
1551  *
1552  * In the common case where neither is set, then we can return "false" without
1553  * updating i_version.
1554  *
1555  * If this function returns false, and no other metadata has changed, then we
1556  * can avoid logging the metadata.
1557  */
inode_maybe_inc_iversion(struct inode * inode,bool force)1558 bool inode_maybe_inc_iversion(struct inode *inode, bool force)
1559 {
1560 	u64 cur, new;
1561 
1562 	/*
1563 	 * The i_version field is not strictly ordered with any other inode
1564 	 * information, but the legacy inode_inc_iversion code used a spinlock
1565 	 * to serialize increments.
1566 	 *
1567 	 * Here, we add full memory barriers to ensure that any de-facto
1568 	 * ordering with other info is preserved.
1569 	 *
1570 	 * This barrier pairs with the barrier in inode_query_iversion()
1571 	 */
1572 	smp_mb();
1573 	cur = inode_peek_iversion_raw(inode);
1574 	do {
1575 		/* If flag is clear then we needn't do anything */
1576 		if (!force && !(cur & I_VERSION_QUERIED))
1577 			return false;
1578 
1579 		/* Since lowest bit is flag, add 2 to avoid it */
1580 		new = (cur & ~I_VERSION_QUERIED) + I_VERSION_INCREMENT;
1581 	} while (!atomic64_try_cmpxchg(&inode->i_version, &cur, new));
1582 	return true;
1583 }
1584 EXPORT_SYMBOL(inode_maybe_inc_iversion);
1585