xref: /openbmc/linux/fs/libfs.c (revision b595076a)
1 /*
2  *	fs/libfs.c
3  *	Library for filesystems writers.
4  */
5 
6 #include <linux/module.h>
7 #include <linux/pagemap.h>
8 #include <linux/slab.h>
9 #include <linux/mount.h>
10 #include <linux/vfs.h>
11 #include <linux/quotaops.h>
12 #include <linux/mutex.h>
13 #include <linux/exportfs.h>
14 #include <linux/writeback.h>
15 #include <linux/buffer_head.h>
16 
17 #include <asm/uaccess.h>
18 
19 int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
20 		   struct kstat *stat)
21 {
22 	struct inode *inode = dentry->d_inode;
23 	generic_fillattr(inode, stat);
24 	stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_SHIFT - 9);
25 	return 0;
26 }
27 
28 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
29 {
30 	buf->f_type = dentry->d_sb->s_magic;
31 	buf->f_bsize = PAGE_CACHE_SIZE;
32 	buf->f_namelen = NAME_MAX;
33 	return 0;
34 }
35 
36 /*
37  * Retaining negative dentries for an in-memory filesystem just wastes
38  * memory and lookup time: arrange for them to be deleted immediately.
39  */
40 static int simple_delete_dentry(struct dentry *dentry)
41 {
42 	return 1;
43 }
44 
45 /*
46  * Lookup the data. This is trivial - if the dentry didn't already
47  * exist, we know it is negative.  Set d_op to delete negative dentries.
48  */
49 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
50 {
51 	static const struct dentry_operations simple_dentry_operations = {
52 		.d_delete = simple_delete_dentry,
53 	};
54 
55 	if (dentry->d_name.len > NAME_MAX)
56 		return ERR_PTR(-ENAMETOOLONG);
57 	dentry->d_op = &simple_dentry_operations;
58 	d_add(dentry, NULL);
59 	return NULL;
60 }
61 
62 int dcache_dir_open(struct inode *inode, struct file *file)
63 {
64 	static struct qstr cursor_name = {.len = 1, .name = "."};
65 
66 	file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
67 
68 	return file->private_data ? 0 : -ENOMEM;
69 }
70 
71 int dcache_dir_close(struct inode *inode, struct file *file)
72 {
73 	dput(file->private_data);
74 	return 0;
75 }
76 
77 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
78 {
79 	mutex_lock(&file->f_path.dentry->d_inode->i_mutex);
80 	switch (origin) {
81 		case 1:
82 			offset += file->f_pos;
83 		case 0:
84 			if (offset >= 0)
85 				break;
86 		default:
87 			mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
88 			return -EINVAL;
89 	}
90 	if (offset != file->f_pos) {
91 		file->f_pos = offset;
92 		if (file->f_pos >= 2) {
93 			struct list_head *p;
94 			struct dentry *cursor = file->private_data;
95 			loff_t n = file->f_pos - 2;
96 
97 			spin_lock(&dcache_lock);
98 			list_del(&cursor->d_u.d_child);
99 			p = file->f_path.dentry->d_subdirs.next;
100 			while (n && p != &file->f_path.dentry->d_subdirs) {
101 				struct dentry *next;
102 				next = list_entry(p, struct dentry, d_u.d_child);
103 				if (!d_unhashed(next) && next->d_inode)
104 					n--;
105 				p = p->next;
106 			}
107 			list_add_tail(&cursor->d_u.d_child, p);
108 			spin_unlock(&dcache_lock);
109 		}
110 	}
111 	mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
112 	return offset;
113 }
114 
115 /* Relationship between i_mode and the DT_xxx types */
116 static inline unsigned char dt_type(struct inode *inode)
117 {
118 	return (inode->i_mode >> 12) & 15;
119 }
120 
121 /*
122  * Directory is locked and all positive dentries in it are safe, since
123  * for ramfs-type trees they can't go away without unlink() or rmdir(),
124  * both impossible due to the lock on directory.
125  */
126 
127 int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
128 {
129 	struct dentry *dentry = filp->f_path.dentry;
130 	struct dentry *cursor = filp->private_data;
131 	struct list_head *p, *q = &cursor->d_u.d_child;
132 	ino_t ino;
133 	int i = filp->f_pos;
134 
135 	switch (i) {
136 		case 0:
137 			ino = dentry->d_inode->i_ino;
138 			if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
139 				break;
140 			filp->f_pos++;
141 			i++;
142 			/* fallthrough */
143 		case 1:
144 			ino = parent_ino(dentry);
145 			if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
146 				break;
147 			filp->f_pos++;
148 			i++;
149 			/* fallthrough */
150 		default:
151 			spin_lock(&dcache_lock);
152 			if (filp->f_pos == 2)
153 				list_move(q, &dentry->d_subdirs);
154 
155 			for (p=q->next; p != &dentry->d_subdirs; p=p->next) {
156 				struct dentry *next;
157 				next = list_entry(p, struct dentry, d_u.d_child);
158 				if (d_unhashed(next) || !next->d_inode)
159 					continue;
160 
161 				spin_unlock(&dcache_lock);
162 				if (filldir(dirent, next->d_name.name,
163 					    next->d_name.len, filp->f_pos,
164 					    next->d_inode->i_ino,
165 					    dt_type(next->d_inode)) < 0)
166 					return 0;
167 				spin_lock(&dcache_lock);
168 				/* next is still alive */
169 				list_move(q, p);
170 				p = q;
171 				filp->f_pos++;
172 			}
173 			spin_unlock(&dcache_lock);
174 	}
175 	return 0;
176 }
177 
178 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
179 {
180 	return -EISDIR;
181 }
182 
183 const struct file_operations simple_dir_operations = {
184 	.open		= dcache_dir_open,
185 	.release	= dcache_dir_close,
186 	.llseek		= dcache_dir_lseek,
187 	.read		= generic_read_dir,
188 	.readdir	= dcache_readdir,
189 	.fsync		= noop_fsync,
190 };
191 
192 const struct inode_operations simple_dir_inode_operations = {
193 	.lookup		= simple_lookup,
194 };
195 
196 static const struct super_operations simple_super_operations = {
197 	.statfs		= simple_statfs,
198 };
199 
200 /*
201  * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
202  * will never be mountable)
203  */
204 struct dentry *mount_pseudo(struct file_system_type *fs_type, char *name,
205 	const struct super_operations *ops, unsigned long magic)
206 {
207 	struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
208 	struct dentry *dentry;
209 	struct inode *root;
210 	struct qstr d_name = {.name = name, .len = strlen(name)};
211 
212 	if (IS_ERR(s))
213 		return ERR_CAST(s);
214 
215 	s->s_flags = MS_NOUSER;
216 	s->s_maxbytes = MAX_LFS_FILESIZE;
217 	s->s_blocksize = PAGE_SIZE;
218 	s->s_blocksize_bits = PAGE_SHIFT;
219 	s->s_magic = magic;
220 	s->s_op = ops ? ops : &simple_super_operations;
221 	s->s_time_gran = 1;
222 	root = new_inode(s);
223 	if (!root)
224 		goto Enomem;
225 	/*
226 	 * since this is the first inode, make it number 1. New inodes created
227 	 * after this must take care not to collide with it (by passing
228 	 * max_reserved of 1 to iunique).
229 	 */
230 	root->i_ino = 1;
231 	root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
232 	root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
233 	dentry = d_alloc(NULL, &d_name);
234 	if (!dentry) {
235 		iput(root);
236 		goto Enomem;
237 	}
238 	dentry->d_sb = s;
239 	dentry->d_parent = dentry;
240 	d_instantiate(dentry, root);
241 	s->s_root = dentry;
242 	s->s_flags |= MS_ACTIVE;
243 	return dget(s->s_root);
244 
245 Enomem:
246 	deactivate_locked_super(s);
247 	return ERR_PTR(-ENOMEM);
248 }
249 
250 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
251 {
252 	struct inode *inode = old_dentry->d_inode;
253 
254 	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
255 	inc_nlink(inode);
256 	ihold(inode);
257 	dget(dentry);
258 	d_instantiate(dentry, inode);
259 	return 0;
260 }
261 
262 static inline int simple_positive(struct dentry *dentry)
263 {
264 	return dentry->d_inode && !d_unhashed(dentry);
265 }
266 
267 int simple_empty(struct dentry *dentry)
268 {
269 	struct dentry *child;
270 	int ret = 0;
271 
272 	spin_lock(&dcache_lock);
273 	list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child)
274 		if (simple_positive(child))
275 			goto out;
276 	ret = 1;
277 out:
278 	spin_unlock(&dcache_lock);
279 	return ret;
280 }
281 
282 int simple_unlink(struct inode *dir, struct dentry *dentry)
283 {
284 	struct inode *inode = dentry->d_inode;
285 
286 	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
287 	drop_nlink(inode);
288 	dput(dentry);
289 	return 0;
290 }
291 
292 int simple_rmdir(struct inode *dir, struct dentry *dentry)
293 {
294 	if (!simple_empty(dentry))
295 		return -ENOTEMPTY;
296 
297 	drop_nlink(dentry->d_inode);
298 	simple_unlink(dir, dentry);
299 	drop_nlink(dir);
300 	return 0;
301 }
302 
303 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
304 		struct inode *new_dir, struct dentry *new_dentry)
305 {
306 	struct inode *inode = old_dentry->d_inode;
307 	int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
308 
309 	if (!simple_empty(new_dentry))
310 		return -ENOTEMPTY;
311 
312 	if (new_dentry->d_inode) {
313 		simple_unlink(new_dir, new_dentry);
314 		if (they_are_dirs)
315 			drop_nlink(old_dir);
316 	} else if (they_are_dirs) {
317 		drop_nlink(old_dir);
318 		inc_nlink(new_dir);
319 	}
320 
321 	old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
322 		new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
323 
324 	return 0;
325 }
326 
327 /**
328  * simple_setattr - setattr for simple filesystem
329  * @dentry: dentry
330  * @iattr: iattr structure
331  *
332  * Returns 0 on success, -error on failure.
333  *
334  * simple_setattr is a simple ->setattr implementation without a proper
335  * implementation of size changes.
336  *
337  * It can either be used for in-memory filesystems or special files
338  * on simple regular filesystems.  Anything that needs to change on-disk
339  * or wire state on size changes needs its own setattr method.
340  */
341 int simple_setattr(struct dentry *dentry, struct iattr *iattr)
342 {
343 	struct inode *inode = dentry->d_inode;
344 	int error;
345 
346 	WARN_ON_ONCE(inode->i_op->truncate);
347 
348 	error = inode_change_ok(inode, iattr);
349 	if (error)
350 		return error;
351 
352 	if (iattr->ia_valid & ATTR_SIZE)
353 		truncate_setsize(inode, iattr->ia_size);
354 	setattr_copy(inode, iattr);
355 	mark_inode_dirty(inode);
356 	return 0;
357 }
358 EXPORT_SYMBOL(simple_setattr);
359 
360 int simple_readpage(struct file *file, struct page *page)
361 {
362 	clear_highpage(page);
363 	flush_dcache_page(page);
364 	SetPageUptodate(page);
365 	unlock_page(page);
366 	return 0;
367 }
368 
369 int simple_write_begin(struct file *file, struct address_space *mapping,
370 			loff_t pos, unsigned len, unsigned flags,
371 			struct page **pagep, void **fsdata)
372 {
373 	struct page *page;
374 	pgoff_t index;
375 
376 	index = pos >> PAGE_CACHE_SHIFT;
377 
378 	page = grab_cache_page_write_begin(mapping, index, flags);
379 	if (!page)
380 		return -ENOMEM;
381 
382 	*pagep = page;
383 
384 	if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
385 		unsigned from = pos & (PAGE_CACHE_SIZE - 1);
386 
387 		zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
388 	}
389 	return 0;
390 }
391 
392 /**
393  * simple_write_end - .write_end helper for non-block-device FSes
394  * @available: See .write_end of address_space_operations
395  * @file: 		"
396  * @mapping: 		"
397  * @pos: 		"
398  * @len: 		"
399  * @copied: 		"
400  * @page: 		"
401  * @fsdata: 		"
402  *
403  * simple_write_end does the minimum needed for updating a page after writing is
404  * done. It has the same API signature as the .write_end of
405  * address_space_operations vector. So it can just be set onto .write_end for
406  * FSes that don't need any other processing. i_mutex is assumed to be held.
407  * Block based filesystems should use generic_write_end().
408  * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
409  * is not called, so a filesystem that actually does store data in .write_inode
410  * should extend on what's done here with a call to mark_inode_dirty() in the
411  * case that i_size has changed.
412  */
413 int simple_write_end(struct file *file, struct address_space *mapping,
414 			loff_t pos, unsigned len, unsigned copied,
415 			struct page *page, void *fsdata)
416 {
417 	struct inode *inode = page->mapping->host;
418 	loff_t last_pos = pos + copied;
419 
420 	/* zero the stale part of the page if we did a short copy */
421 	if (copied < len) {
422 		unsigned from = pos & (PAGE_CACHE_SIZE - 1);
423 
424 		zero_user(page, from + copied, len - copied);
425 	}
426 
427 	if (!PageUptodate(page))
428 		SetPageUptodate(page);
429 	/*
430 	 * No need to use i_size_read() here, the i_size
431 	 * cannot change under us because we hold the i_mutex.
432 	 */
433 	if (last_pos > inode->i_size)
434 		i_size_write(inode, last_pos);
435 
436 	set_page_dirty(page);
437 	unlock_page(page);
438 	page_cache_release(page);
439 
440 	return copied;
441 }
442 
443 /*
444  * the inodes created here are not hashed. If you use iunique to generate
445  * unique inode values later for this filesystem, then you must take care
446  * to pass it an appropriate max_reserved value to avoid collisions.
447  */
448 int simple_fill_super(struct super_block *s, unsigned long magic,
449 		      struct tree_descr *files)
450 {
451 	struct inode *inode;
452 	struct dentry *root;
453 	struct dentry *dentry;
454 	int i;
455 
456 	s->s_blocksize = PAGE_CACHE_SIZE;
457 	s->s_blocksize_bits = PAGE_CACHE_SHIFT;
458 	s->s_magic = magic;
459 	s->s_op = &simple_super_operations;
460 	s->s_time_gran = 1;
461 
462 	inode = new_inode(s);
463 	if (!inode)
464 		return -ENOMEM;
465 	/*
466 	 * because the root inode is 1, the files array must not contain an
467 	 * entry at index 1
468 	 */
469 	inode->i_ino = 1;
470 	inode->i_mode = S_IFDIR | 0755;
471 	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
472 	inode->i_op = &simple_dir_inode_operations;
473 	inode->i_fop = &simple_dir_operations;
474 	inode->i_nlink = 2;
475 	root = d_alloc_root(inode);
476 	if (!root) {
477 		iput(inode);
478 		return -ENOMEM;
479 	}
480 	for (i = 0; !files->name || files->name[0]; i++, files++) {
481 		if (!files->name)
482 			continue;
483 
484 		/* warn if it tries to conflict with the root inode */
485 		if (unlikely(i == 1))
486 			printk(KERN_WARNING "%s: %s passed in a files array"
487 				"with an index of 1!\n", __func__,
488 				s->s_type->name);
489 
490 		dentry = d_alloc_name(root, files->name);
491 		if (!dentry)
492 			goto out;
493 		inode = new_inode(s);
494 		if (!inode)
495 			goto out;
496 		inode->i_mode = S_IFREG | files->mode;
497 		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
498 		inode->i_fop = files->ops;
499 		inode->i_ino = i;
500 		d_add(dentry, inode);
501 	}
502 	s->s_root = root;
503 	return 0;
504 out:
505 	d_genocide(root);
506 	dput(root);
507 	return -ENOMEM;
508 }
509 
510 static DEFINE_SPINLOCK(pin_fs_lock);
511 
512 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
513 {
514 	struct vfsmount *mnt = NULL;
515 	spin_lock(&pin_fs_lock);
516 	if (unlikely(!*mount)) {
517 		spin_unlock(&pin_fs_lock);
518 		mnt = vfs_kern_mount(type, 0, type->name, NULL);
519 		if (IS_ERR(mnt))
520 			return PTR_ERR(mnt);
521 		spin_lock(&pin_fs_lock);
522 		if (!*mount)
523 			*mount = mnt;
524 	}
525 	mntget(*mount);
526 	++*count;
527 	spin_unlock(&pin_fs_lock);
528 	mntput(mnt);
529 	return 0;
530 }
531 
532 void simple_release_fs(struct vfsmount **mount, int *count)
533 {
534 	struct vfsmount *mnt;
535 	spin_lock(&pin_fs_lock);
536 	mnt = *mount;
537 	if (!--*count)
538 		*mount = NULL;
539 	spin_unlock(&pin_fs_lock);
540 	mntput(mnt);
541 }
542 
543 /**
544  * simple_read_from_buffer - copy data from the buffer to user space
545  * @to: the user space buffer to read to
546  * @count: the maximum number of bytes to read
547  * @ppos: the current position in the buffer
548  * @from: the buffer to read from
549  * @available: the size of the buffer
550  *
551  * The simple_read_from_buffer() function reads up to @count bytes from the
552  * buffer @from at offset @ppos into the user space address starting at @to.
553  *
554  * On success, the number of bytes read is returned and the offset @ppos is
555  * advanced by this number, or negative value is returned on error.
556  **/
557 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
558 				const void *from, size_t available)
559 {
560 	loff_t pos = *ppos;
561 	size_t ret;
562 
563 	if (pos < 0)
564 		return -EINVAL;
565 	if (pos >= available || !count)
566 		return 0;
567 	if (count > available - pos)
568 		count = available - pos;
569 	ret = copy_to_user(to, from + pos, count);
570 	if (ret == count)
571 		return -EFAULT;
572 	count -= ret;
573 	*ppos = pos + count;
574 	return count;
575 }
576 
577 /**
578  * simple_write_to_buffer - copy data from user space to the buffer
579  * @to: the buffer to write to
580  * @available: the size of the buffer
581  * @ppos: the current position in the buffer
582  * @from: the user space buffer to read from
583  * @count: the maximum number of bytes to read
584  *
585  * The simple_write_to_buffer() function reads up to @count bytes from the user
586  * space address starting at @from into the buffer @to at offset @ppos.
587  *
588  * On success, the number of bytes written is returned and the offset @ppos is
589  * advanced by this number, or negative value is returned on error.
590  **/
591 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
592 		const void __user *from, size_t count)
593 {
594 	loff_t pos = *ppos;
595 	size_t res;
596 
597 	if (pos < 0)
598 		return -EINVAL;
599 	if (pos >= available || !count)
600 		return 0;
601 	if (count > available - pos)
602 		count = available - pos;
603 	res = copy_from_user(to + pos, from, count);
604 	if (res == count)
605 		return -EFAULT;
606 	count -= res;
607 	*ppos = pos + count;
608 	return count;
609 }
610 
611 /**
612  * memory_read_from_buffer - copy data from the buffer
613  * @to: the kernel space buffer to read to
614  * @count: the maximum number of bytes to read
615  * @ppos: the current position in the buffer
616  * @from: the buffer to read from
617  * @available: the size of the buffer
618  *
619  * The memory_read_from_buffer() function reads up to @count bytes from the
620  * buffer @from at offset @ppos into the kernel space address starting at @to.
621  *
622  * On success, the number of bytes read is returned and the offset @ppos is
623  * advanced by this number, or negative value is returned on error.
624  **/
625 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
626 				const void *from, size_t available)
627 {
628 	loff_t pos = *ppos;
629 
630 	if (pos < 0)
631 		return -EINVAL;
632 	if (pos >= available)
633 		return 0;
634 	if (count > available - pos)
635 		count = available - pos;
636 	memcpy(to, from + pos, count);
637 	*ppos = pos + count;
638 
639 	return count;
640 }
641 
642 /*
643  * Transaction based IO.
644  * The file expects a single write which triggers the transaction, and then
645  * possibly a read which collects the result - which is stored in a
646  * file-local buffer.
647  */
648 
649 void simple_transaction_set(struct file *file, size_t n)
650 {
651 	struct simple_transaction_argresp *ar = file->private_data;
652 
653 	BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
654 
655 	/*
656 	 * The barrier ensures that ar->size will really remain zero until
657 	 * ar->data is ready for reading.
658 	 */
659 	smp_mb();
660 	ar->size = n;
661 }
662 
663 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
664 {
665 	struct simple_transaction_argresp *ar;
666 	static DEFINE_SPINLOCK(simple_transaction_lock);
667 
668 	if (size > SIMPLE_TRANSACTION_LIMIT - 1)
669 		return ERR_PTR(-EFBIG);
670 
671 	ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
672 	if (!ar)
673 		return ERR_PTR(-ENOMEM);
674 
675 	spin_lock(&simple_transaction_lock);
676 
677 	/* only one write allowed per open */
678 	if (file->private_data) {
679 		spin_unlock(&simple_transaction_lock);
680 		free_page((unsigned long)ar);
681 		return ERR_PTR(-EBUSY);
682 	}
683 
684 	file->private_data = ar;
685 
686 	spin_unlock(&simple_transaction_lock);
687 
688 	if (copy_from_user(ar->data, buf, size))
689 		return ERR_PTR(-EFAULT);
690 
691 	return ar->data;
692 }
693 
694 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
695 {
696 	struct simple_transaction_argresp *ar = file->private_data;
697 
698 	if (!ar)
699 		return 0;
700 	return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
701 }
702 
703 int simple_transaction_release(struct inode *inode, struct file *file)
704 {
705 	free_page((unsigned long)file->private_data);
706 	return 0;
707 }
708 
709 /* Simple attribute files */
710 
711 struct simple_attr {
712 	int (*get)(void *, u64 *);
713 	int (*set)(void *, u64);
714 	char get_buf[24];	/* enough to store a u64 and "\n\0" */
715 	char set_buf[24];
716 	void *data;
717 	const char *fmt;	/* format for read operation */
718 	struct mutex mutex;	/* protects access to these buffers */
719 };
720 
721 /* simple_attr_open is called by an actual attribute open file operation
722  * to set the attribute specific access operations. */
723 int simple_attr_open(struct inode *inode, struct file *file,
724 		     int (*get)(void *, u64 *), int (*set)(void *, u64),
725 		     const char *fmt)
726 {
727 	struct simple_attr *attr;
728 
729 	attr = kmalloc(sizeof(*attr), GFP_KERNEL);
730 	if (!attr)
731 		return -ENOMEM;
732 
733 	attr->get = get;
734 	attr->set = set;
735 	attr->data = inode->i_private;
736 	attr->fmt = fmt;
737 	mutex_init(&attr->mutex);
738 
739 	file->private_data = attr;
740 
741 	return nonseekable_open(inode, file);
742 }
743 
744 int simple_attr_release(struct inode *inode, struct file *file)
745 {
746 	kfree(file->private_data);
747 	return 0;
748 }
749 
750 /* read from the buffer that is filled with the get function */
751 ssize_t simple_attr_read(struct file *file, char __user *buf,
752 			 size_t len, loff_t *ppos)
753 {
754 	struct simple_attr *attr;
755 	size_t size;
756 	ssize_t ret;
757 
758 	attr = file->private_data;
759 
760 	if (!attr->get)
761 		return -EACCES;
762 
763 	ret = mutex_lock_interruptible(&attr->mutex);
764 	if (ret)
765 		return ret;
766 
767 	if (*ppos) {		/* continued read */
768 		size = strlen(attr->get_buf);
769 	} else {		/* first read */
770 		u64 val;
771 		ret = attr->get(attr->data, &val);
772 		if (ret)
773 			goto out;
774 
775 		size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
776 				 attr->fmt, (unsigned long long)val);
777 	}
778 
779 	ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
780 out:
781 	mutex_unlock(&attr->mutex);
782 	return ret;
783 }
784 
785 /* interpret the buffer as a number to call the set function with */
786 ssize_t simple_attr_write(struct file *file, const char __user *buf,
787 			  size_t len, loff_t *ppos)
788 {
789 	struct simple_attr *attr;
790 	u64 val;
791 	size_t size;
792 	ssize_t ret;
793 
794 	attr = file->private_data;
795 	if (!attr->set)
796 		return -EACCES;
797 
798 	ret = mutex_lock_interruptible(&attr->mutex);
799 	if (ret)
800 		return ret;
801 
802 	ret = -EFAULT;
803 	size = min(sizeof(attr->set_buf) - 1, len);
804 	if (copy_from_user(attr->set_buf, buf, size))
805 		goto out;
806 
807 	attr->set_buf[size] = '\0';
808 	val = simple_strtol(attr->set_buf, NULL, 0);
809 	ret = attr->set(attr->data, val);
810 	if (ret == 0)
811 		ret = len; /* on success, claim we got the whole input */
812 out:
813 	mutex_unlock(&attr->mutex);
814 	return ret;
815 }
816 
817 /**
818  * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
819  * @sb:		filesystem to do the file handle conversion on
820  * @fid:	file handle to convert
821  * @fh_len:	length of the file handle in bytes
822  * @fh_type:	type of file handle
823  * @get_inode:	filesystem callback to retrieve inode
824  *
825  * This function decodes @fid as long as it has one of the well-known
826  * Linux filehandle types and calls @get_inode on it to retrieve the
827  * inode for the object specified in the file handle.
828  */
829 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
830 		int fh_len, int fh_type, struct inode *(*get_inode)
831 			(struct super_block *sb, u64 ino, u32 gen))
832 {
833 	struct inode *inode = NULL;
834 
835 	if (fh_len < 2)
836 		return NULL;
837 
838 	switch (fh_type) {
839 	case FILEID_INO32_GEN:
840 	case FILEID_INO32_GEN_PARENT:
841 		inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
842 		break;
843 	}
844 
845 	return d_obtain_alias(inode);
846 }
847 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
848 
849 /**
850  * generic_fh_to_dentry - generic helper for the fh_to_parent export operation
851  * @sb:		filesystem to do the file handle conversion on
852  * @fid:	file handle to convert
853  * @fh_len:	length of the file handle in bytes
854  * @fh_type:	type of file handle
855  * @get_inode:	filesystem callback to retrieve inode
856  *
857  * This function decodes @fid as long as it has one of the well-known
858  * Linux filehandle types and calls @get_inode on it to retrieve the
859  * inode for the _parent_ object specified in the file handle if it
860  * is specified in the file handle, or NULL otherwise.
861  */
862 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
863 		int fh_len, int fh_type, struct inode *(*get_inode)
864 			(struct super_block *sb, u64 ino, u32 gen))
865 {
866 	struct inode *inode = NULL;
867 
868 	if (fh_len <= 2)
869 		return NULL;
870 
871 	switch (fh_type) {
872 	case FILEID_INO32_GEN_PARENT:
873 		inode = get_inode(sb, fid->i32.parent_ino,
874 				  (fh_len > 3 ? fid->i32.parent_gen : 0));
875 		break;
876 	}
877 
878 	return d_obtain_alias(inode);
879 }
880 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
881 
882 /**
883  * generic_file_fsync - generic fsync implementation for simple filesystems
884  * @file:	file to synchronize
885  * @datasync:	only synchronize essential metadata if true
886  *
887  * This is a generic implementation of the fsync method for simple
888  * filesystems which track all non-inode metadata in the buffers list
889  * hanging off the address_space structure.
890  */
891 int generic_file_fsync(struct file *file, int datasync)
892 {
893 	struct inode *inode = file->f_mapping->host;
894 	int err;
895 	int ret;
896 
897 	ret = sync_mapping_buffers(inode->i_mapping);
898 	if (!(inode->i_state & I_DIRTY))
899 		return ret;
900 	if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
901 		return ret;
902 
903 	err = sync_inode_metadata(inode, 1);
904 	if (ret == 0)
905 		ret = err;
906 	return ret;
907 }
908 EXPORT_SYMBOL(generic_file_fsync);
909 
910 /**
911  * generic_check_addressable - Check addressability of file system
912  * @blocksize_bits:	log of file system block size
913  * @num_blocks:		number of blocks in file system
914  *
915  * Determine whether a file system with @num_blocks blocks (and a
916  * block size of 2**@blocksize_bits) is addressable by the sector_t
917  * and page cache of the system.  Return 0 if so and -EFBIG otherwise.
918  */
919 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
920 {
921 	u64 last_fs_block = num_blocks - 1;
922 	u64 last_fs_page =
923 		last_fs_block >> (PAGE_CACHE_SHIFT - blocksize_bits);
924 
925 	if (unlikely(num_blocks == 0))
926 		return 0;
927 
928 	if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT))
929 		return -EINVAL;
930 
931 	if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
932 	    (last_fs_page > (pgoff_t)(~0ULL))) {
933 		return -EFBIG;
934 	}
935 	return 0;
936 }
937 EXPORT_SYMBOL(generic_check_addressable);
938 
939 /*
940  * No-op implementation of ->fsync for in-memory filesystems.
941  */
942 int noop_fsync(struct file *file, int datasync)
943 {
944 	return 0;
945 }
946 
947 EXPORT_SYMBOL(dcache_dir_close);
948 EXPORT_SYMBOL(dcache_dir_lseek);
949 EXPORT_SYMBOL(dcache_dir_open);
950 EXPORT_SYMBOL(dcache_readdir);
951 EXPORT_SYMBOL(generic_read_dir);
952 EXPORT_SYMBOL(mount_pseudo);
953 EXPORT_SYMBOL(simple_write_begin);
954 EXPORT_SYMBOL(simple_write_end);
955 EXPORT_SYMBOL(simple_dir_inode_operations);
956 EXPORT_SYMBOL(simple_dir_operations);
957 EXPORT_SYMBOL(simple_empty);
958 EXPORT_SYMBOL(simple_fill_super);
959 EXPORT_SYMBOL(simple_getattr);
960 EXPORT_SYMBOL(simple_link);
961 EXPORT_SYMBOL(simple_lookup);
962 EXPORT_SYMBOL(simple_pin_fs);
963 EXPORT_SYMBOL(simple_readpage);
964 EXPORT_SYMBOL(simple_release_fs);
965 EXPORT_SYMBOL(simple_rename);
966 EXPORT_SYMBOL(simple_rmdir);
967 EXPORT_SYMBOL(simple_statfs);
968 EXPORT_SYMBOL(noop_fsync);
969 EXPORT_SYMBOL(simple_unlink);
970 EXPORT_SYMBOL(simple_read_from_buffer);
971 EXPORT_SYMBOL(simple_write_to_buffer);
972 EXPORT_SYMBOL(memory_read_from_buffer);
973 EXPORT_SYMBOL(simple_transaction_set);
974 EXPORT_SYMBOL(simple_transaction_get);
975 EXPORT_SYMBOL(simple_transaction_read);
976 EXPORT_SYMBOL(simple_transaction_release);
977 EXPORT_SYMBOL_GPL(simple_attr_open);
978 EXPORT_SYMBOL_GPL(simple_attr_release);
979 EXPORT_SYMBOL_GPL(simple_attr_read);
980 EXPORT_SYMBOL_GPL(simple_attr_write);
981