xref: /openbmc/linux/fs/libfs.c (revision 08157984)
1 /*
2  *	fs/libfs.c
3  *	Library for filesystems writers.
4  */
5 
6 #include <linux/export.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> /* sync_mapping_buffers */
16 
17 #include <asm/uaccess.h>
18 
19 #include "internal.h"
20 
21 static inline int simple_positive(struct dentry *dentry)
22 {
23 	return dentry->d_inode && !d_unhashed(dentry);
24 }
25 
26 int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
27 		   struct kstat *stat)
28 {
29 	struct inode *inode = dentry->d_inode;
30 	generic_fillattr(inode, stat);
31 	stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_SHIFT - 9);
32 	return 0;
33 }
34 
35 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
36 {
37 	buf->f_type = dentry->d_sb->s_magic;
38 	buf->f_bsize = PAGE_CACHE_SIZE;
39 	buf->f_namelen = NAME_MAX;
40 	return 0;
41 }
42 
43 /*
44  * Retaining negative dentries for an in-memory filesystem just wastes
45  * memory and lookup time: arrange for them to be deleted immediately.
46  */
47 static int simple_delete_dentry(const struct dentry *dentry)
48 {
49 	return 1;
50 }
51 
52 /*
53  * Lookup the data. This is trivial - if the dentry didn't already
54  * exist, we know it is negative.  Set d_op to delete negative dentries.
55  */
56 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
57 {
58 	static const struct dentry_operations simple_dentry_operations = {
59 		.d_delete = simple_delete_dentry,
60 	};
61 
62 	if (dentry->d_name.len > NAME_MAX)
63 		return ERR_PTR(-ENAMETOOLONG);
64 	d_set_d_op(dentry, &simple_dentry_operations);
65 	d_add(dentry, NULL);
66 	return NULL;
67 }
68 
69 int dcache_dir_open(struct inode *inode, struct file *file)
70 {
71 	static struct qstr cursor_name = {.len = 1, .name = "."};
72 
73 	file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
74 
75 	return file->private_data ? 0 : -ENOMEM;
76 }
77 
78 int dcache_dir_close(struct inode *inode, struct file *file)
79 {
80 	dput(file->private_data);
81 	return 0;
82 }
83 
84 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
85 {
86 	struct dentry *dentry = file->f_path.dentry;
87 	mutex_lock(&dentry->d_inode->i_mutex);
88 	switch (origin) {
89 		case 1:
90 			offset += file->f_pos;
91 		case 0:
92 			if (offset >= 0)
93 				break;
94 		default:
95 			mutex_unlock(&dentry->d_inode->i_mutex);
96 			return -EINVAL;
97 	}
98 	if (offset != file->f_pos) {
99 		file->f_pos = offset;
100 		if (file->f_pos >= 2) {
101 			struct list_head *p;
102 			struct dentry *cursor = file->private_data;
103 			loff_t n = file->f_pos - 2;
104 
105 			spin_lock(&dentry->d_lock);
106 			/* d_lock not required for cursor */
107 			list_del(&cursor->d_u.d_child);
108 			p = dentry->d_subdirs.next;
109 			while (n && p != &dentry->d_subdirs) {
110 				struct dentry *next;
111 				next = list_entry(p, struct dentry, d_u.d_child);
112 				spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
113 				if (simple_positive(next))
114 					n--;
115 				spin_unlock(&next->d_lock);
116 				p = p->next;
117 			}
118 			list_add_tail(&cursor->d_u.d_child, p);
119 			spin_unlock(&dentry->d_lock);
120 		}
121 	}
122 	mutex_unlock(&dentry->d_inode->i_mutex);
123 	return offset;
124 }
125 
126 /* Relationship between i_mode and the DT_xxx types */
127 static inline unsigned char dt_type(struct inode *inode)
128 {
129 	return (inode->i_mode >> 12) & 15;
130 }
131 
132 /*
133  * Directory is locked and all positive dentries in it are safe, since
134  * for ramfs-type trees they can't go away without unlink() or rmdir(),
135  * both impossible due to the lock on directory.
136  */
137 
138 int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
139 {
140 	struct dentry *dentry = filp->f_path.dentry;
141 	struct dentry *cursor = filp->private_data;
142 	struct list_head *p, *q = &cursor->d_u.d_child;
143 	ino_t ino;
144 	int i = filp->f_pos;
145 
146 	switch (i) {
147 		case 0:
148 			ino = dentry->d_inode->i_ino;
149 			if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
150 				break;
151 			filp->f_pos++;
152 			i++;
153 			/* fallthrough */
154 		case 1:
155 			ino = parent_ino(dentry);
156 			if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
157 				break;
158 			filp->f_pos++;
159 			i++;
160 			/* fallthrough */
161 		default:
162 			spin_lock(&dentry->d_lock);
163 			if (filp->f_pos == 2)
164 				list_move(q, &dentry->d_subdirs);
165 
166 			for (p=q->next; p != &dentry->d_subdirs; p=p->next) {
167 				struct dentry *next;
168 				next = list_entry(p, struct dentry, d_u.d_child);
169 				spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
170 				if (!simple_positive(next)) {
171 					spin_unlock(&next->d_lock);
172 					continue;
173 				}
174 
175 				spin_unlock(&next->d_lock);
176 				spin_unlock(&dentry->d_lock);
177 				if (filldir(dirent, next->d_name.name,
178 					    next->d_name.len, filp->f_pos,
179 					    next->d_inode->i_ino,
180 					    dt_type(next->d_inode)) < 0)
181 					return 0;
182 				spin_lock(&dentry->d_lock);
183 				spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
184 				/* next is still alive */
185 				list_move(q, p);
186 				spin_unlock(&next->d_lock);
187 				p = q;
188 				filp->f_pos++;
189 			}
190 			spin_unlock(&dentry->d_lock);
191 	}
192 	return 0;
193 }
194 
195 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
196 {
197 	return -EISDIR;
198 }
199 
200 const struct file_operations simple_dir_operations = {
201 	.open		= dcache_dir_open,
202 	.release	= dcache_dir_close,
203 	.llseek		= dcache_dir_lseek,
204 	.read		= generic_read_dir,
205 	.readdir	= dcache_readdir,
206 	.fsync		= noop_fsync,
207 };
208 
209 const struct inode_operations simple_dir_inode_operations = {
210 	.lookup		= simple_lookup,
211 };
212 
213 static const struct super_operations simple_super_operations = {
214 	.statfs		= simple_statfs,
215 };
216 
217 /*
218  * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
219  * will never be mountable)
220  */
221 struct dentry *mount_pseudo(struct file_system_type *fs_type, char *name,
222 	const struct super_operations *ops,
223 	const struct dentry_operations *dops, unsigned long magic)
224 {
225 	struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
226 	struct dentry *dentry;
227 	struct inode *root;
228 	struct qstr d_name = {.name = name, .len = strlen(name)};
229 
230 	if (IS_ERR(s))
231 		return ERR_CAST(s);
232 
233 	s->s_flags = MS_NOUSER;
234 	s->s_maxbytes = MAX_LFS_FILESIZE;
235 	s->s_blocksize = PAGE_SIZE;
236 	s->s_blocksize_bits = PAGE_SHIFT;
237 	s->s_magic = magic;
238 	s->s_op = ops ? ops : &simple_super_operations;
239 	s->s_time_gran = 1;
240 	root = new_inode(s);
241 	if (!root)
242 		goto Enomem;
243 	/*
244 	 * since this is the first inode, make it number 1. New inodes created
245 	 * after this must take care not to collide with it (by passing
246 	 * max_reserved of 1 to iunique).
247 	 */
248 	root->i_ino = 1;
249 	root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
250 	root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
251 	dentry = __d_alloc(s, &d_name);
252 	if (!dentry) {
253 		iput(root);
254 		goto Enomem;
255 	}
256 	d_instantiate(dentry, root);
257 	s->s_root = dentry;
258 	s->s_d_op = dops;
259 	s->s_flags |= MS_ACTIVE;
260 	return dget(s->s_root);
261 
262 Enomem:
263 	deactivate_locked_super(s);
264 	return ERR_PTR(-ENOMEM);
265 }
266 
267 int simple_open(struct inode *inode, struct file *file)
268 {
269 	if (inode->i_private)
270 		file->private_data = inode->i_private;
271 	return 0;
272 }
273 
274 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
275 {
276 	struct inode *inode = old_dentry->d_inode;
277 
278 	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
279 	inc_nlink(inode);
280 	ihold(inode);
281 	dget(dentry);
282 	d_instantiate(dentry, inode);
283 	return 0;
284 }
285 
286 int simple_empty(struct dentry *dentry)
287 {
288 	struct dentry *child;
289 	int ret = 0;
290 
291 	spin_lock(&dentry->d_lock);
292 	list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child) {
293 		spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
294 		if (simple_positive(child)) {
295 			spin_unlock(&child->d_lock);
296 			goto out;
297 		}
298 		spin_unlock(&child->d_lock);
299 	}
300 	ret = 1;
301 out:
302 	spin_unlock(&dentry->d_lock);
303 	return ret;
304 }
305 
306 int simple_unlink(struct inode *dir, struct dentry *dentry)
307 {
308 	struct inode *inode = dentry->d_inode;
309 
310 	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
311 	drop_nlink(inode);
312 	dput(dentry);
313 	return 0;
314 }
315 
316 int simple_rmdir(struct inode *dir, struct dentry *dentry)
317 {
318 	if (!simple_empty(dentry))
319 		return -ENOTEMPTY;
320 
321 	drop_nlink(dentry->d_inode);
322 	simple_unlink(dir, dentry);
323 	drop_nlink(dir);
324 	return 0;
325 }
326 
327 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
328 		struct inode *new_dir, struct dentry *new_dentry)
329 {
330 	struct inode *inode = old_dentry->d_inode;
331 	int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
332 
333 	if (!simple_empty(new_dentry))
334 		return -ENOTEMPTY;
335 
336 	if (new_dentry->d_inode) {
337 		simple_unlink(new_dir, new_dentry);
338 		if (they_are_dirs) {
339 			drop_nlink(new_dentry->d_inode);
340 			drop_nlink(old_dir);
341 		}
342 	} else if (they_are_dirs) {
343 		drop_nlink(old_dir);
344 		inc_nlink(new_dir);
345 	}
346 
347 	old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
348 		new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
349 
350 	return 0;
351 }
352 
353 /**
354  * simple_setattr - setattr for simple filesystem
355  * @dentry: dentry
356  * @iattr: iattr structure
357  *
358  * Returns 0 on success, -error on failure.
359  *
360  * simple_setattr is a simple ->setattr implementation without a proper
361  * implementation of size changes.
362  *
363  * It can either be used for in-memory filesystems or special files
364  * on simple regular filesystems.  Anything that needs to change on-disk
365  * or wire state on size changes needs its own setattr method.
366  */
367 int simple_setattr(struct dentry *dentry, struct iattr *iattr)
368 {
369 	struct inode *inode = dentry->d_inode;
370 	int error;
371 
372 	WARN_ON_ONCE(inode->i_op->truncate);
373 
374 	error = inode_change_ok(inode, iattr);
375 	if (error)
376 		return error;
377 
378 	if (iattr->ia_valid & ATTR_SIZE)
379 		truncate_setsize(inode, iattr->ia_size);
380 	setattr_copy(inode, iattr);
381 	mark_inode_dirty(inode);
382 	return 0;
383 }
384 EXPORT_SYMBOL(simple_setattr);
385 
386 int simple_readpage(struct file *file, struct page *page)
387 {
388 	clear_highpage(page);
389 	flush_dcache_page(page);
390 	SetPageUptodate(page);
391 	unlock_page(page);
392 	return 0;
393 }
394 
395 int simple_write_begin(struct file *file, struct address_space *mapping,
396 			loff_t pos, unsigned len, unsigned flags,
397 			struct page **pagep, void **fsdata)
398 {
399 	struct page *page;
400 	pgoff_t index;
401 
402 	index = pos >> PAGE_CACHE_SHIFT;
403 
404 	page = grab_cache_page_write_begin(mapping, index, flags);
405 	if (!page)
406 		return -ENOMEM;
407 
408 	*pagep = page;
409 
410 	if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
411 		unsigned from = pos & (PAGE_CACHE_SIZE - 1);
412 
413 		zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
414 	}
415 	return 0;
416 }
417 
418 /**
419  * simple_write_end - .write_end helper for non-block-device FSes
420  * @available: See .write_end of address_space_operations
421  * @file: 		"
422  * @mapping: 		"
423  * @pos: 		"
424  * @len: 		"
425  * @copied: 		"
426  * @page: 		"
427  * @fsdata: 		"
428  *
429  * simple_write_end does the minimum needed for updating a page after writing is
430  * done. It has the same API signature as the .write_end of
431  * address_space_operations vector. So it can just be set onto .write_end for
432  * FSes that don't need any other processing. i_mutex is assumed to be held.
433  * Block based filesystems should use generic_write_end().
434  * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
435  * is not called, so a filesystem that actually does store data in .write_inode
436  * should extend on what's done here with a call to mark_inode_dirty() in the
437  * case that i_size has changed.
438  */
439 int simple_write_end(struct file *file, struct address_space *mapping,
440 			loff_t pos, unsigned len, unsigned copied,
441 			struct page *page, void *fsdata)
442 {
443 	struct inode *inode = page->mapping->host;
444 	loff_t last_pos = pos + copied;
445 
446 	/* zero the stale part of the page if we did a short copy */
447 	if (copied < len) {
448 		unsigned from = pos & (PAGE_CACHE_SIZE - 1);
449 
450 		zero_user(page, from + copied, len - copied);
451 	}
452 
453 	if (!PageUptodate(page))
454 		SetPageUptodate(page);
455 	/*
456 	 * No need to use i_size_read() here, the i_size
457 	 * cannot change under us because we hold the i_mutex.
458 	 */
459 	if (last_pos > inode->i_size)
460 		i_size_write(inode, last_pos);
461 
462 	set_page_dirty(page);
463 	unlock_page(page);
464 	page_cache_release(page);
465 
466 	return copied;
467 }
468 
469 /*
470  * the inodes created here are not hashed. If you use iunique to generate
471  * unique inode values later for this filesystem, then you must take care
472  * to pass it an appropriate max_reserved value to avoid collisions.
473  */
474 int simple_fill_super(struct super_block *s, unsigned long magic,
475 		      struct tree_descr *files)
476 {
477 	struct inode *inode;
478 	struct dentry *root;
479 	struct dentry *dentry;
480 	int i;
481 
482 	s->s_blocksize = PAGE_CACHE_SIZE;
483 	s->s_blocksize_bits = PAGE_CACHE_SHIFT;
484 	s->s_magic = magic;
485 	s->s_op = &simple_super_operations;
486 	s->s_time_gran = 1;
487 
488 	inode = new_inode(s);
489 	if (!inode)
490 		return -ENOMEM;
491 	/*
492 	 * because the root inode is 1, the files array must not contain an
493 	 * entry at index 1
494 	 */
495 	inode->i_ino = 1;
496 	inode->i_mode = S_IFDIR | 0755;
497 	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
498 	inode->i_op = &simple_dir_inode_operations;
499 	inode->i_fop = &simple_dir_operations;
500 	set_nlink(inode, 2);
501 	root = d_make_root(inode);
502 	if (!root)
503 		return -ENOMEM;
504 	for (i = 0; !files->name || files->name[0]; i++, files++) {
505 		if (!files->name)
506 			continue;
507 
508 		/* warn if it tries to conflict with the root inode */
509 		if (unlikely(i == 1))
510 			printk(KERN_WARNING "%s: %s passed in a files array"
511 				"with an index of 1!\n", __func__,
512 				s->s_type->name);
513 
514 		dentry = d_alloc_name(root, files->name);
515 		if (!dentry)
516 			goto out;
517 		inode = new_inode(s);
518 		if (!inode) {
519 			dput(dentry);
520 			goto out;
521 		}
522 		inode->i_mode = S_IFREG | files->mode;
523 		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
524 		inode->i_fop = files->ops;
525 		inode->i_ino = i;
526 		d_add(dentry, inode);
527 	}
528 	s->s_root = root;
529 	return 0;
530 out:
531 	d_genocide(root);
532 	shrink_dcache_parent(root);
533 	dput(root);
534 	return -ENOMEM;
535 }
536 
537 static DEFINE_SPINLOCK(pin_fs_lock);
538 
539 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
540 {
541 	struct vfsmount *mnt = NULL;
542 	spin_lock(&pin_fs_lock);
543 	if (unlikely(!*mount)) {
544 		spin_unlock(&pin_fs_lock);
545 		mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, NULL);
546 		if (IS_ERR(mnt))
547 			return PTR_ERR(mnt);
548 		spin_lock(&pin_fs_lock);
549 		if (!*mount)
550 			*mount = mnt;
551 	}
552 	mntget(*mount);
553 	++*count;
554 	spin_unlock(&pin_fs_lock);
555 	mntput(mnt);
556 	return 0;
557 }
558 
559 void simple_release_fs(struct vfsmount **mount, int *count)
560 {
561 	struct vfsmount *mnt;
562 	spin_lock(&pin_fs_lock);
563 	mnt = *mount;
564 	if (!--*count)
565 		*mount = NULL;
566 	spin_unlock(&pin_fs_lock);
567 	mntput(mnt);
568 }
569 
570 /**
571  * simple_read_from_buffer - copy data from the buffer to user space
572  * @to: the user space buffer to read to
573  * @count: the maximum number of bytes to read
574  * @ppos: the current position in the buffer
575  * @from: the buffer to read from
576  * @available: the size of the buffer
577  *
578  * The simple_read_from_buffer() function reads up to @count bytes from the
579  * buffer @from at offset @ppos into the user space address starting at @to.
580  *
581  * On success, the number of bytes read is returned and the offset @ppos is
582  * advanced by this number, or negative value is returned on error.
583  **/
584 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
585 				const void *from, size_t available)
586 {
587 	loff_t pos = *ppos;
588 	size_t ret;
589 
590 	if (pos < 0)
591 		return -EINVAL;
592 	if (pos >= available || !count)
593 		return 0;
594 	if (count > available - pos)
595 		count = available - pos;
596 	ret = copy_to_user(to, from + pos, count);
597 	if (ret == count)
598 		return -EFAULT;
599 	count -= ret;
600 	*ppos = pos + count;
601 	return count;
602 }
603 
604 /**
605  * simple_write_to_buffer - copy data from user space to the buffer
606  * @to: the buffer to write to
607  * @available: the size of the buffer
608  * @ppos: the current position in the buffer
609  * @from: the user space buffer to read from
610  * @count: the maximum number of bytes to read
611  *
612  * The simple_write_to_buffer() function reads up to @count bytes from the user
613  * space address starting at @from into the buffer @to at offset @ppos.
614  *
615  * On success, the number of bytes written is returned and the offset @ppos is
616  * advanced by this number, or negative value is returned on error.
617  **/
618 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
619 		const void __user *from, size_t count)
620 {
621 	loff_t pos = *ppos;
622 	size_t res;
623 
624 	if (pos < 0)
625 		return -EINVAL;
626 	if (pos >= available || !count)
627 		return 0;
628 	if (count > available - pos)
629 		count = available - pos;
630 	res = copy_from_user(to + pos, from, count);
631 	if (res == count)
632 		return -EFAULT;
633 	count -= res;
634 	*ppos = pos + count;
635 	return count;
636 }
637 
638 /**
639  * memory_read_from_buffer - copy data from the buffer
640  * @to: the kernel space buffer to read to
641  * @count: the maximum number of bytes to read
642  * @ppos: the current position in the buffer
643  * @from: the buffer to read from
644  * @available: the size of the buffer
645  *
646  * The memory_read_from_buffer() function reads up to @count bytes from the
647  * buffer @from at offset @ppos into the kernel space address starting at @to.
648  *
649  * On success, the number of bytes read is returned and the offset @ppos is
650  * advanced by this number, or negative value is returned on error.
651  **/
652 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
653 				const void *from, size_t available)
654 {
655 	loff_t pos = *ppos;
656 
657 	if (pos < 0)
658 		return -EINVAL;
659 	if (pos >= available)
660 		return 0;
661 	if (count > available - pos)
662 		count = available - pos;
663 	memcpy(to, from + pos, count);
664 	*ppos = pos + count;
665 
666 	return count;
667 }
668 
669 /*
670  * Transaction based IO.
671  * The file expects a single write which triggers the transaction, and then
672  * possibly a read which collects the result - which is stored in a
673  * file-local buffer.
674  */
675 
676 void simple_transaction_set(struct file *file, size_t n)
677 {
678 	struct simple_transaction_argresp *ar = file->private_data;
679 
680 	BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
681 
682 	/*
683 	 * The barrier ensures that ar->size will really remain zero until
684 	 * ar->data is ready for reading.
685 	 */
686 	smp_mb();
687 	ar->size = n;
688 }
689 
690 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
691 {
692 	struct simple_transaction_argresp *ar;
693 	static DEFINE_SPINLOCK(simple_transaction_lock);
694 
695 	if (size > SIMPLE_TRANSACTION_LIMIT - 1)
696 		return ERR_PTR(-EFBIG);
697 
698 	ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
699 	if (!ar)
700 		return ERR_PTR(-ENOMEM);
701 
702 	spin_lock(&simple_transaction_lock);
703 
704 	/* only one write allowed per open */
705 	if (file->private_data) {
706 		spin_unlock(&simple_transaction_lock);
707 		free_page((unsigned long)ar);
708 		return ERR_PTR(-EBUSY);
709 	}
710 
711 	file->private_data = ar;
712 
713 	spin_unlock(&simple_transaction_lock);
714 
715 	if (copy_from_user(ar->data, buf, size))
716 		return ERR_PTR(-EFAULT);
717 
718 	return ar->data;
719 }
720 
721 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
722 {
723 	struct simple_transaction_argresp *ar = file->private_data;
724 
725 	if (!ar)
726 		return 0;
727 	return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
728 }
729 
730 int simple_transaction_release(struct inode *inode, struct file *file)
731 {
732 	free_page((unsigned long)file->private_data);
733 	return 0;
734 }
735 
736 /* Simple attribute files */
737 
738 struct simple_attr {
739 	int (*get)(void *, u64 *);
740 	int (*set)(void *, u64);
741 	char get_buf[24];	/* enough to store a u64 and "\n\0" */
742 	char set_buf[24];
743 	void *data;
744 	const char *fmt;	/* format for read operation */
745 	struct mutex mutex;	/* protects access to these buffers */
746 };
747 
748 /* simple_attr_open is called by an actual attribute open file operation
749  * to set the attribute specific access operations. */
750 int simple_attr_open(struct inode *inode, struct file *file,
751 		     int (*get)(void *, u64 *), int (*set)(void *, u64),
752 		     const char *fmt)
753 {
754 	struct simple_attr *attr;
755 
756 	attr = kmalloc(sizeof(*attr), GFP_KERNEL);
757 	if (!attr)
758 		return -ENOMEM;
759 
760 	attr->get = get;
761 	attr->set = set;
762 	attr->data = inode->i_private;
763 	attr->fmt = fmt;
764 	mutex_init(&attr->mutex);
765 
766 	file->private_data = attr;
767 
768 	return nonseekable_open(inode, file);
769 }
770 
771 int simple_attr_release(struct inode *inode, struct file *file)
772 {
773 	kfree(file->private_data);
774 	return 0;
775 }
776 
777 /* read from the buffer that is filled with the get function */
778 ssize_t simple_attr_read(struct file *file, char __user *buf,
779 			 size_t len, loff_t *ppos)
780 {
781 	struct simple_attr *attr;
782 	size_t size;
783 	ssize_t ret;
784 
785 	attr = file->private_data;
786 
787 	if (!attr->get)
788 		return -EACCES;
789 
790 	ret = mutex_lock_interruptible(&attr->mutex);
791 	if (ret)
792 		return ret;
793 
794 	if (*ppos) {		/* continued read */
795 		size = strlen(attr->get_buf);
796 	} else {		/* first read */
797 		u64 val;
798 		ret = attr->get(attr->data, &val);
799 		if (ret)
800 			goto out;
801 
802 		size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
803 				 attr->fmt, (unsigned long long)val);
804 	}
805 
806 	ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
807 out:
808 	mutex_unlock(&attr->mutex);
809 	return ret;
810 }
811 
812 /* interpret the buffer as a number to call the set function with */
813 ssize_t simple_attr_write(struct file *file, const char __user *buf,
814 			  size_t len, loff_t *ppos)
815 {
816 	struct simple_attr *attr;
817 	u64 val;
818 	size_t size;
819 	ssize_t ret;
820 
821 	attr = file->private_data;
822 	if (!attr->set)
823 		return -EACCES;
824 
825 	ret = mutex_lock_interruptible(&attr->mutex);
826 	if (ret)
827 		return ret;
828 
829 	ret = -EFAULT;
830 	size = min(sizeof(attr->set_buf) - 1, len);
831 	if (copy_from_user(attr->set_buf, buf, size))
832 		goto out;
833 
834 	attr->set_buf[size] = '\0';
835 	val = simple_strtoll(attr->set_buf, NULL, 0);
836 	ret = attr->set(attr->data, val);
837 	if (ret == 0)
838 		ret = len; /* on success, claim we got the whole input */
839 out:
840 	mutex_unlock(&attr->mutex);
841 	return ret;
842 }
843 
844 /**
845  * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
846  * @sb:		filesystem to do the file handle conversion on
847  * @fid:	file handle to convert
848  * @fh_len:	length of the file handle in bytes
849  * @fh_type:	type of file handle
850  * @get_inode:	filesystem callback to retrieve inode
851  *
852  * This function decodes @fid as long as it has one of the well-known
853  * Linux filehandle types and calls @get_inode on it to retrieve the
854  * inode for the object specified in the file handle.
855  */
856 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
857 		int fh_len, int fh_type, struct inode *(*get_inode)
858 			(struct super_block *sb, u64 ino, u32 gen))
859 {
860 	struct inode *inode = NULL;
861 
862 	if (fh_len < 2)
863 		return NULL;
864 
865 	switch (fh_type) {
866 	case FILEID_INO32_GEN:
867 	case FILEID_INO32_GEN_PARENT:
868 		inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
869 		break;
870 	}
871 
872 	return d_obtain_alias(inode);
873 }
874 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
875 
876 /**
877  * generic_fh_to_dentry - generic helper for the fh_to_parent export operation
878  * @sb:		filesystem to do the file handle conversion on
879  * @fid:	file handle to convert
880  * @fh_len:	length of the file handle in bytes
881  * @fh_type:	type of file handle
882  * @get_inode:	filesystem callback to retrieve inode
883  *
884  * This function decodes @fid as long as it has one of the well-known
885  * Linux filehandle types and calls @get_inode on it to retrieve the
886  * inode for the _parent_ object specified in the file handle if it
887  * is specified in the file handle, or NULL otherwise.
888  */
889 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
890 		int fh_len, int fh_type, struct inode *(*get_inode)
891 			(struct super_block *sb, u64 ino, u32 gen))
892 {
893 	struct inode *inode = NULL;
894 
895 	if (fh_len <= 2)
896 		return NULL;
897 
898 	switch (fh_type) {
899 	case FILEID_INO32_GEN_PARENT:
900 		inode = get_inode(sb, fid->i32.parent_ino,
901 				  (fh_len > 3 ? fid->i32.parent_gen : 0));
902 		break;
903 	}
904 
905 	return d_obtain_alias(inode);
906 }
907 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
908 
909 /**
910  * generic_file_fsync - generic fsync implementation for simple filesystems
911  * @file:	file to synchronize
912  * @datasync:	only synchronize essential metadata if true
913  *
914  * This is a generic implementation of the fsync method for simple
915  * filesystems which track all non-inode metadata in the buffers list
916  * hanging off the address_space structure.
917  */
918 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
919 		       int datasync)
920 {
921 	struct inode *inode = file->f_mapping->host;
922 	int err;
923 	int ret;
924 
925 	err = filemap_write_and_wait_range(inode->i_mapping, start, end);
926 	if (err)
927 		return err;
928 
929 	mutex_lock(&inode->i_mutex);
930 	ret = sync_mapping_buffers(inode->i_mapping);
931 	if (!(inode->i_state & I_DIRTY))
932 		goto out;
933 	if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
934 		goto out;
935 
936 	err = sync_inode_metadata(inode, 1);
937 	if (ret == 0)
938 		ret = err;
939 out:
940 	mutex_unlock(&inode->i_mutex);
941 	return ret;
942 }
943 EXPORT_SYMBOL(generic_file_fsync);
944 
945 /**
946  * generic_check_addressable - Check addressability of file system
947  * @blocksize_bits:	log of file system block size
948  * @num_blocks:		number of blocks in file system
949  *
950  * Determine whether a file system with @num_blocks blocks (and a
951  * block size of 2**@blocksize_bits) is addressable by the sector_t
952  * and page cache of the system.  Return 0 if so and -EFBIG otherwise.
953  */
954 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
955 {
956 	u64 last_fs_block = num_blocks - 1;
957 	u64 last_fs_page =
958 		last_fs_block >> (PAGE_CACHE_SHIFT - blocksize_bits);
959 
960 	if (unlikely(num_blocks == 0))
961 		return 0;
962 
963 	if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT))
964 		return -EINVAL;
965 
966 	if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
967 	    (last_fs_page > (pgoff_t)(~0ULL))) {
968 		return -EFBIG;
969 	}
970 	return 0;
971 }
972 EXPORT_SYMBOL(generic_check_addressable);
973 
974 /*
975  * No-op implementation of ->fsync for in-memory filesystems.
976  */
977 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
978 {
979 	return 0;
980 }
981 
982 EXPORT_SYMBOL(dcache_dir_close);
983 EXPORT_SYMBOL(dcache_dir_lseek);
984 EXPORT_SYMBOL(dcache_dir_open);
985 EXPORT_SYMBOL(dcache_readdir);
986 EXPORT_SYMBOL(generic_read_dir);
987 EXPORT_SYMBOL(mount_pseudo);
988 EXPORT_SYMBOL(simple_write_begin);
989 EXPORT_SYMBOL(simple_write_end);
990 EXPORT_SYMBOL(simple_dir_inode_operations);
991 EXPORT_SYMBOL(simple_dir_operations);
992 EXPORT_SYMBOL(simple_empty);
993 EXPORT_SYMBOL(simple_fill_super);
994 EXPORT_SYMBOL(simple_getattr);
995 EXPORT_SYMBOL(simple_open);
996 EXPORT_SYMBOL(simple_link);
997 EXPORT_SYMBOL(simple_lookup);
998 EXPORT_SYMBOL(simple_pin_fs);
999 EXPORT_SYMBOL(simple_readpage);
1000 EXPORT_SYMBOL(simple_release_fs);
1001 EXPORT_SYMBOL(simple_rename);
1002 EXPORT_SYMBOL(simple_rmdir);
1003 EXPORT_SYMBOL(simple_statfs);
1004 EXPORT_SYMBOL(noop_fsync);
1005 EXPORT_SYMBOL(simple_unlink);
1006 EXPORT_SYMBOL(simple_read_from_buffer);
1007 EXPORT_SYMBOL(simple_write_to_buffer);
1008 EXPORT_SYMBOL(memory_read_from_buffer);
1009 EXPORT_SYMBOL(simple_transaction_set);
1010 EXPORT_SYMBOL(simple_transaction_get);
1011 EXPORT_SYMBOL(simple_transaction_read);
1012 EXPORT_SYMBOL(simple_transaction_release);
1013 EXPORT_SYMBOL_GPL(simple_attr_open);
1014 EXPORT_SYMBOL_GPL(simple_attr_release);
1015 EXPORT_SYMBOL_GPL(simple_attr_read);
1016 EXPORT_SYMBOL_GPL(simple_attr_write);
1017