xref: /openbmc/linux/fs/libfs.c (revision 6a1a90ad1b0edb556a7550a6ef8a8756f0304dd5)
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 int get_sb_pseudo(struct file_system_type *fs_type, char *name,
205 	const struct super_operations *ops, unsigned long magic,
206 	struct vfsmount *mnt)
207 {
208 	struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
209 	struct dentry *dentry;
210 	struct inode *root;
211 	struct qstr d_name = {.name = name, .len = strlen(name)};
212 
213 	if (IS_ERR(s))
214 		return PTR_ERR(s);
215 
216 	s->s_flags = MS_NOUSER;
217 	s->s_maxbytes = MAX_LFS_FILESIZE;
218 	s->s_blocksize = PAGE_SIZE;
219 	s->s_blocksize_bits = PAGE_SHIFT;
220 	s->s_magic = magic;
221 	s->s_op = ops ? ops : &simple_super_operations;
222 	s->s_time_gran = 1;
223 	root = new_inode(s);
224 	if (!root)
225 		goto Enomem;
226 	/*
227 	 * since this is the first inode, make it number 1. New inodes created
228 	 * after this must take care not to collide with it (by passing
229 	 * max_reserved of 1 to iunique).
230 	 */
231 	root->i_ino = 1;
232 	root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
233 	root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
234 	dentry = d_alloc(NULL, &d_name);
235 	if (!dentry) {
236 		iput(root);
237 		goto Enomem;
238 	}
239 	dentry->d_sb = s;
240 	dentry->d_parent = dentry;
241 	d_instantiate(dentry, root);
242 	s->s_root = dentry;
243 	s->s_flags |= MS_ACTIVE;
244 	simple_set_mnt(mnt, s);
245 	return 0;
246 
247 Enomem:
248 	deactivate_locked_super(s);
249 	return -ENOMEM;
250 }
251 
252 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
253 {
254 	struct inode *inode = old_dentry->d_inode;
255 
256 	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
257 	inc_nlink(inode);
258 	atomic_inc(&inode->i_count);
259 	dget(dentry);
260 	d_instantiate(dentry, inode);
261 	return 0;
262 }
263 
264 static inline int simple_positive(struct dentry *dentry)
265 {
266 	return dentry->d_inode && !d_unhashed(dentry);
267 }
268 
269 int simple_empty(struct dentry *dentry)
270 {
271 	struct dentry *child;
272 	int ret = 0;
273 
274 	spin_lock(&dcache_lock);
275 	list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child)
276 		if (simple_positive(child))
277 			goto out;
278 	ret = 1;
279 out:
280 	spin_unlock(&dcache_lock);
281 	return ret;
282 }
283 
284 int simple_unlink(struct inode *dir, struct dentry *dentry)
285 {
286 	struct inode *inode = dentry->d_inode;
287 
288 	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
289 	drop_nlink(inode);
290 	dput(dentry);
291 	return 0;
292 }
293 
294 int simple_rmdir(struct inode *dir, struct dentry *dentry)
295 {
296 	if (!simple_empty(dentry))
297 		return -ENOTEMPTY;
298 
299 	drop_nlink(dentry->d_inode);
300 	simple_unlink(dir, dentry);
301 	drop_nlink(dir);
302 	return 0;
303 }
304 
305 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
306 		struct inode *new_dir, struct dentry *new_dentry)
307 {
308 	struct inode *inode = old_dentry->d_inode;
309 	int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
310 
311 	if (!simple_empty(new_dentry))
312 		return -ENOTEMPTY;
313 
314 	if (new_dentry->d_inode) {
315 		simple_unlink(new_dir, new_dentry);
316 		if (they_are_dirs)
317 			drop_nlink(old_dir);
318 	} else if (they_are_dirs) {
319 		drop_nlink(old_dir);
320 		inc_nlink(new_dir);
321 	}
322 
323 	old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
324 		new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
325 
326 	return 0;
327 }
328 
329 /**
330  * simple_setsize - handle core mm and vfs requirements for file size change
331  * @inode: inode
332  * @newsize: new file size
333  *
334  * Returns 0 on success, -error on failure.
335  *
336  * simple_setsize must be called with inode_mutex held.
337  *
338  * simple_setsize will check that the requested new size is OK (see
339  * inode_newsize_ok), and then will perform the necessary i_size update
340  * and pagecache truncation (if necessary). It will be typically be called
341  * from the filesystem's setattr function when ATTR_SIZE is passed in.
342  *
343  * The inode itself must have correct permissions and attributes to allow
344  * i_size to be changed, this function then just checks that the new size
345  * requested is valid.
346  *
347  * In the case of simple in-memory filesystems with inodes stored solely
348  * in the inode cache, and file data in the pagecache, nothing more needs
349  * to be done to satisfy a truncate request. Filesystems with on-disk
350  * blocks for example will need to free them in the case of truncate, in
351  * that case it may be easier not to use simple_setsize (but each of its
352  * components will likely be required at some point to update pagecache
353  * and inode etc).
354  */
355 int simple_setsize(struct inode *inode, loff_t newsize)
356 {
357 	loff_t oldsize;
358 	int error;
359 
360 	error = inode_newsize_ok(inode, newsize);
361 	if (error)
362 		return error;
363 
364 	oldsize = inode->i_size;
365 	i_size_write(inode, newsize);
366 	truncate_pagecache(inode, oldsize, newsize);
367 
368 	return error;
369 }
370 EXPORT_SYMBOL(simple_setsize);
371 
372 /**
373  * simple_setattr - setattr for simple in-memory filesystem
374  * @dentry: dentry
375  * @iattr: iattr structure
376  *
377  * Returns 0 on success, -error on failure.
378  *
379  * simple_setattr implements setattr for an in-memory filesystem which
380  * does not store its own file data or metadata (eg. uses the page cache
381  * and inode cache as its data store).
382  */
383 int simple_setattr(struct dentry *dentry, struct iattr *iattr)
384 {
385 	struct inode *inode = dentry->d_inode;
386 	int error;
387 
388 	error = inode_change_ok(inode, iattr);
389 	if (error)
390 		return error;
391 
392 	if (iattr->ia_valid & ATTR_SIZE) {
393 		error = simple_setsize(inode, iattr->ia_size);
394 		if (error)
395 			return error;
396 	}
397 
398 	setattr_copy(inode, iattr);
399 	return error;
400 }
401 EXPORT_SYMBOL(simple_setattr);
402 
403 int simple_readpage(struct file *file, struct page *page)
404 {
405 	clear_highpage(page);
406 	flush_dcache_page(page);
407 	SetPageUptodate(page);
408 	unlock_page(page);
409 	return 0;
410 }
411 
412 int simple_write_begin(struct file *file, struct address_space *mapping,
413 			loff_t pos, unsigned len, unsigned flags,
414 			struct page **pagep, void **fsdata)
415 {
416 	struct page *page;
417 	pgoff_t index;
418 
419 	index = pos >> PAGE_CACHE_SHIFT;
420 
421 	page = grab_cache_page_write_begin(mapping, index, flags);
422 	if (!page)
423 		return -ENOMEM;
424 
425 	*pagep = page;
426 
427 	if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
428 		unsigned from = pos & (PAGE_CACHE_SIZE - 1);
429 
430 		zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
431 	}
432 	return 0;
433 }
434 
435 /**
436  * simple_write_end - .write_end helper for non-block-device FSes
437  * @available: See .write_end of address_space_operations
438  * @file: 		"
439  * @mapping: 		"
440  * @pos: 		"
441  * @len: 		"
442  * @copied: 		"
443  * @page: 		"
444  * @fsdata: 		"
445  *
446  * simple_write_end does the minimum needed for updating a page after writing is
447  * done. It has the same API signature as the .write_end of
448  * address_space_operations vector. So it can just be set onto .write_end for
449  * FSes that don't need any other processing. i_mutex is assumed to be held.
450  * Block based filesystems should use generic_write_end().
451  * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
452  * is not called, so a filesystem that actually does store data in .write_inode
453  * should extend on what's done here with a call to mark_inode_dirty() in the
454  * case that i_size has changed.
455  */
456 int simple_write_end(struct file *file, struct address_space *mapping,
457 			loff_t pos, unsigned len, unsigned copied,
458 			struct page *page, void *fsdata)
459 {
460 	struct inode *inode = page->mapping->host;
461 	loff_t last_pos = pos + copied;
462 
463 	/* zero the stale part of the page if we did a short copy */
464 	if (copied < len) {
465 		unsigned from = pos & (PAGE_CACHE_SIZE - 1);
466 
467 		zero_user(page, from + copied, len - copied);
468 	}
469 
470 	if (!PageUptodate(page))
471 		SetPageUptodate(page);
472 	/*
473 	 * No need to use i_size_read() here, the i_size
474 	 * cannot change under us because we hold the i_mutex.
475 	 */
476 	if (last_pos > inode->i_size)
477 		i_size_write(inode, last_pos);
478 
479 	set_page_dirty(page);
480 	unlock_page(page);
481 	page_cache_release(page);
482 
483 	return copied;
484 }
485 
486 /*
487  * the inodes created here are not hashed. If you use iunique to generate
488  * unique inode values later for this filesystem, then you must take care
489  * to pass it an appropriate max_reserved value to avoid collisions.
490  */
491 int simple_fill_super(struct super_block *s, unsigned long magic,
492 		      struct tree_descr *files)
493 {
494 	struct inode *inode;
495 	struct dentry *root;
496 	struct dentry *dentry;
497 	int i;
498 
499 	s->s_blocksize = PAGE_CACHE_SIZE;
500 	s->s_blocksize_bits = PAGE_CACHE_SHIFT;
501 	s->s_magic = magic;
502 	s->s_op = &simple_super_operations;
503 	s->s_time_gran = 1;
504 
505 	inode = new_inode(s);
506 	if (!inode)
507 		return -ENOMEM;
508 	/*
509 	 * because the root inode is 1, the files array must not contain an
510 	 * entry at index 1
511 	 */
512 	inode->i_ino = 1;
513 	inode->i_mode = S_IFDIR | 0755;
514 	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
515 	inode->i_op = &simple_dir_inode_operations;
516 	inode->i_fop = &simple_dir_operations;
517 	inode->i_nlink = 2;
518 	root = d_alloc_root(inode);
519 	if (!root) {
520 		iput(inode);
521 		return -ENOMEM;
522 	}
523 	for (i = 0; !files->name || files->name[0]; i++, files++) {
524 		if (!files->name)
525 			continue;
526 
527 		/* warn if it tries to conflict with the root inode */
528 		if (unlikely(i == 1))
529 			printk(KERN_WARNING "%s: %s passed in a files array"
530 				"with an index of 1!\n", __func__,
531 				s->s_type->name);
532 
533 		dentry = d_alloc_name(root, files->name);
534 		if (!dentry)
535 			goto out;
536 		inode = new_inode(s);
537 		if (!inode)
538 			goto out;
539 		inode->i_mode = S_IFREG | files->mode;
540 		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
541 		inode->i_fop = files->ops;
542 		inode->i_ino = i;
543 		d_add(dentry, inode);
544 	}
545 	s->s_root = root;
546 	return 0;
547 out:
548 	d_genocide(root);
549 	dput(root);
550 	return -ENOMEM;
551 }
552 
553 static DEFINE_SPINLOCK(pin_fs_lock);
554 
555 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
556 {
557 	struct vfsmount *mnt = NULL;
558 	spin_lock(&pin_fs_lock);
559 	if (unlikely(!*mount)) {
560 		spin_unlock(&pin_fs_lock);
561 		mnt = vfs_kern_mount(type, 0, type->name, NULL);
562 		if (IS_ERR(mnt))
563 			return PTR_ERR(mnt);
564 		spin_lock(&pin_fs_lock);
565 		if (!*mount)
566 			*mount = mnt;
567 	}
568 	mntget(*mount);
569 	++*count;
570 	spin_unlock(&pin_fs_lock);
571 	mntput(mnt);
572 	return 0;
573 }
574 
575 void simple_release_fs(struct vfsmount **mount, int *count)
576 {
577 	struct vfsmount *mnt;
578 	spin_lock(&pin_fs_lock);
579 	mnt = *mount;
580 	if (!--*count)
581 		*mount = NULL;
582 	spin_unlock(&pin_fs_lock);
583 	mntput(mnt);
584 }
585 
586 /**
587  * simple_read_from_buffer - copy data from the buffer to user space
588  * @to: the user space buffer to read to
589  * @count: the maximum number of bytes to read
590  * @ppos: the current position in the buffer
591  * @from: the buffer to read from
592  * @available: the size of the buffer
593  *
594  * The simple_read_from_buffer() function reads up to @count bytes from the
595  * buffer @from at offset @ppos into the user space address starting at @to.
596  *
597  * On success, the number of bytes read is returned and the offset @ppos is
598  * advanced by this number, or negative value is returned on error.
599  **/
600 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
601 				const void *from, size_t available)
602 {
603 	loff_t pos = *ppos;
604 	size_t ret;
605 
606 	if (pos < 0)
607 		return -EINVAL;
608 	if (pos >= available || !count)
609 		return 0;
610 	if (count > available - pos)
611 		count = available - pos;
612 	ret = copy_to_user(to, from + pos, count);
613 	if (ret == count)
614 		return -EFAULT;
615 	count -= ret;
616 	*ppos = pos + count;
617 	return count;
618 }
619 
620 /**
621  * simple_write_to_buffer - copy data from user space to the buffer
622  * @to: the buffer to write to
623  * @available: the size of the buffer
624  * @ppos: the current position in the buffer
625  * @from: the user space buffer to read from
626  * @count: the maximum number of bytes to read
627  *
628  * The simple_write_to_buffer() function reads up to @count bytes from the user
629  * space address starting at @from into the buffer @to at offset @ppos.
630  *
631  * On success, the number of bytes written is returned and the offset @ppos is
632  * advanced by this number, or negative value is returned on error.
633  **/
634 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
635 		const void __user *from, size_t count)
636 {
637 	loff_t pos = *ppos;
638 	size_t res;
639 
640 	if (pos < 0)
641 		return -EINVAL;
642 	if (pos >= available || !count)
643 		return 0;
644 	if (count > available - pos)
645 		count = available - pos;
646 	res = copy_from_user(to + pos, from, count);
647 	if (res == count)
648 		return -EFAULT;
649 	count -= res;
650 	*ppos = pos + count;
651 	return count;
652 }
653 
654 /**
655  * memory_read_from_buffer - copy data from the buffer
656  * @to: the kernel space buffer to read to
657  * @count: the maximum number of bytes to read
658  * @ppos: the current position in the buffer
659  * @from: the buffer to read from
660  * @available: the size of the buffer
661  *
662  * The memory_read_from_buffer() function reads up to @count bytes from the
663  * buffer @from at offset @ppos into the kernel space address starting at @to.
664  *
665  * On success, the number of bytes read is returned and the offset @ppos is
666  * advanced by this number, or negative value is returned on error.
667  **/
668 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
669 				const void *from, size_t available)
670 {
671 	loff_t pos = *ppos;
672 
673 	if (pos < 0)
674 		return -EINVAL;
675 	if (pos >= available)
676 		return 0;
677 	if (count > available - pos)
678 		count = available - pos;
679 	memcpy(to, from + pos, count);
680 	*ppos = pos + count;
681 
682 	return count;
683 }
684 
685 /*
686  * Transaction based IO.
687  * The file expects a single write which triggers the transaction, and then
688  * possibly a read which collects the result - which is stored in a
689  * file-local buffer.
690  */
691 
692 void simple_transaction_set(struct file *file, size_t n)
693 {
694 	struct simple_transaction_argresp *ar = file->private_data;
695 
696 	BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
697 
698 	/*
699 	 * The barrier ensures that ar->size will really remain zero until
700 	 * ar->data is ready for reading.
701 	 */
702 	smp_mb();
703 	ar->size = n;
704 }
705 
706 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
707 {
708 	struct simple_transaction_argresp *ar;
709 	static DEFINE_SPINLOCK(simple_transaction_lock);
710 
711 	if (size > SIMPLE_TRANSACTION_LIMIT - 1)
712 		return ERR_PTR(-EFBIG);
713 
714 	ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
715 	if (!ar)
716 		return ERR_PTR(-ENOMEM);
717 
718 	spin_lock(&simple_transaction_lock);
719 
720 	/* only one write allowed per open */
721 	if (file->private_data) {
722 		spin_unlock(&simple_transaction_lock);
723 		free_page((unsigned long)ar);
724 		return ERR_PTR(-EBUSY);
725 	}
726 
727 	file->private_data = ar;
728 
729 	spin_unlock(&simple_transaction_lock);
730 
731 	if (copy_from_user(ar->data, buf, size))
732 		return ERR_PTR(-EFAULT);
733 
734 	return ar->data;
735 }
736 
737 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
738 {
739 	struct simple_transaction_argresp *ar = file->private_data;
740 
741 	if (!ar)
742 		return 0;
743 	return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
744 }
745 
746 int simple_transaction_release(struct inode *inode, struct file *file)
747 {
748 	free_page((unsigned long)file->private_data);
749 	return 0;
750 }
751 
752 /* Simple attribute files */
753 
754 struct simple_attr {
755 	int (*get)(void *, u64 *);
756 	int (*set)(void *, u64);
757 	char get_buf[24];	/* enough to store a u64 and "\n\0" */
758 	char set_buf[24];
759 	void *data;
760 	const char *fmt;	/* format for read operation */
761 	struct mutex mutex;	/* protects access to these buffers */
762 };
763 
764 /* simple_attr_open is called by an actual attribute open file operation
765  * to set the attribute specific access operations. */
766 int simple_attr_open(struct inode *inode, struct file *file,
767 		     int (*get)(void *, u64 *), int (*set)(void *, u64),
768 		     const char *fmt)
769 {
770 	struct simple_attr *attr;
771 
772 	attr = kmalloc(sizeof(*attr), GFP_KERNEL);
773 	if (!attr)
774 		return -ENOMEM;
775 
776 	attr->get = get;
777 	attr->set = set;
778 	attr->data = inode->i_private;
779 	attr->fmt = fmt;
780 	mutex_init(&attr->mutex);
781 
782 	file->private_data = attr;
783 
784 	return nonseekable_open(inode, file);
785 }
786 
787 int simple_attr_release(struct inode *inode, struct file *file)
788 {
789 	kfree(file->private_data);
790 	return 0;
791 }
792 
793 /* read from the buffer that is filled with the get function */
794 ssize_t simple_attr_read(struct file *file, char __user *buf,
795 			 size_t len, loff_t *ppos)
796 {
797 	struct simple_attr *attr;
798 	size_t size;
799 	ssize_t ret;
800 
801 	attr = file->private_data;
802 
803 	if (!attr->get)
804 		return -EACCES;
805 
806 	ret = mutex_lock_interruptible(&attr->mutex);
807 	if (ret)
808 		return ret;
809 
810 	if (*ppos) {		/* continued read */
811 		size = strlen(attr->get_buf);
812 	} else {		/* first read */
813 		u64 val;
814 		ret = attr->get(attr->data, &val);
815 		if (ret)
816 			goto out;
817 
818 		size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
819 				 attr->fmt, (unsigned long long)val);
820 	}
821 
822 	ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
823 out:
824 	mutex_unlock(&attr->mutex);
825 	return ret;
826 }
827 
828 /* interpret the buffer as a number to call the set function with */
829 ssize_t simple_attr_write(struct file *file, const char __user *buf,
830 			  size_t len, loff_t *ppos)
831 {
832 	struct simple_attr *attr;
833 	u64 val;
834 	size_t size;
835 	ssize_t ret;
836 
837 	attr = file->private_data;
838 	if (!attr->set)
839 		return -EACCES;
840 
841 	ret = mutex_lock_interruptible(&attr->mutex);
842 	if (ret)
843 		return ret;
844 
845 	ret = -EFAULT;
846 	size = min(sizeof(attr->set_buf) - 1, len);
847 	if (copy_from_user(attr->set_buf, buf, size))
848 		goto out;
849 
850 	attr->set_buf[size] = '\0';
851 	val = simple_strtol(attr->set_buf, NULL, 0);
852 	ret = attr->set(attr->data, val);
853 	if (ret == 0)
854 		ret = len; /* on success, claim we got the whole input */
855 out:
856 	mutex_unlock(&attr->mutex);
857 	return ret;
858 }
859 
860 /**
861  * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
862  * @sb:		filesystem to do the file handle conversion on
863  * @fid:	file handle to convert
864  * @fh_len:	length of the file handle in bytes
865  * @fh_type:	type of file handle
866  * @get_inode:	filesystem callback to retrieve inode
867  *
868  * This function decodes @fid as long as it has one of the well-known
869  * Linux filehandle types and calls @get_inode on it to retrieve the
870  * inode for the object specified in the file handle.
871  */
872 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
873 		int fh_len, int fh_type, struct inode *(*get_inode)
874 			(struct super_block *sb, u64 ino, u32 gen))
875 {
876 	struct inode *inode = NULL;
877 
878 	if (fh_len < 2)
879 		return NULL;
880 
881 	switch (fh_type) {
882 	case FILEID_INO32_GEN:
883 	case FILEID_INO32_GEN_PARENT:
884 		inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
885 		break;
886 	}
887 
888 	return d_obtain_alias(inode);
889 }
890 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
891 
892 /**
893  * generic_fh_to_dentry - generic helper for the fh_to_parent export operation
894  * @sb:		filesystem to do the file handle conversion on
895  * @fid:	file handle to convert
896  * @fh_len:	length of the file handle in bytes
897  * @fh_type:	type of file handle
898  * @get_inode:	filesystem callback to retrieve inode
899  *
900  * This function decodes @fid as long as it has one of the well-known
901  * Linux filehandle types and calls @get_inode on it to retrieve the
902  * inode for the _parent_ object specified in the file handle if it
903  * is specified in the file handle, or NULL otherwise.
904  */
905 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
906 		int fh_len, int fh_type, struct inode *(*get_inode)
907 			(struct super_block *sb, u64 ino, u32 gen))
908 {
909 	struct inode *inode = NULL;
910 
911 	if (fh_len <= 2)
912 		return NULL;
913 
914 	switch (fh_type) {
915 	case FILEID_INO32_GEN_PARENT:
916 		inode = get_inode(sb, fid->i32.parent_ino,
917 				  (fh_len > 3 ? fid->i32.parent_gen : 0));
918 		break;
919 	}
920 
921 	return d_obtain_alias(inode);
922 }
923 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
924 
925 /**
926  * generic_file_fsync - generic fsync implementation for simple filesystems
927  * @file:	file to synchronize
928  * @datasync:	only synchronize essential metadata if true
929  *
930  * This is a generic implementation of the fsync method for simple
931  * filesystems which track all non-inode metadata in the buffers list
932  * hanging off the address_space structure.
933  */
934 int generic_file_fsync(struct file *file, int datasync)
935 {
936 	struct writeback_control wbc = {
937 		.sync_mode = WB_SYNC_ALL,
938 		.nr_to_write = 0, /* metadata-only; caller takes care of data */
939 	};
940 	struct inode *inode = file->f_mapping->host;
941 	int err;
942 	int ret;
943 
944 	ret = sync_mapping_buffers(inode->i_mapping);
945 	if (!(inode->i_state & I_DIRTY))
946 		return ret;
947 	if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
948 		return ret;
949 
950 	err = sync_inode(inode, &wbc);
951 	if (ret == 0)
952 		ret = err;
953 	return ret;
954 }
955 EXPORT_SYMBOL(generic_file_fsync);
956 
957 /*
958  * No-op implementation of ->fsync for in-memory filesystems.
959  */
960 int noop_fsync(struct file *file, int datasync)
961 {
962 	return 0;
963 }
964 
965 EXPORT_SYMBOL(dcache_dir_close);
966 EXPORT_SYMBOL(dcache_dir_lseek);
967 EXPORT_SYMBOL(dcache_dir_open);
968 EXPORT_SYMBOL(dcache_readdir);
969 EXPORT_SYMBOL(generic_read_dir);
970 EXPORT_SYMBOL(get_sb_pseudo);
971 EXPORT_SYMBOL(simple_write_begin);
972 EXPORT_SYMBOL(simple_write_end);
973 EXPORT_SYMBOL(simple_dir_inode_operations);
974 EXPORT_SYMBOL(simple_dir_operations);
975 EXPORT_SYMBOL(simple_empty);
976 EXPORT_SYMBOL(simple_fill_super);
977 EXPORT_SYMBOL(simple_getattr);
978 EXPORT_SYMBOL(simple_link);
979 EXPORT_SYMBOL(simple_lookup);
980 EXPORT_SYMBOL(simple_pin_fs);
981 EXPORT_SYMBOL(simple_readpage);
982 EXPORT_SYMBOL(simple_release_fs);
983 EXPORT_SYMBOL(simple_rename);
984 EXPORT_SYMBOL(simple_rmdir);
985 EXPORT_SYMBOL(simple_statfs);
986 EXPORT_SYMBOL(noop_fsync);
987 EXPORT_SYMBOL(simple_unlink);
988 EXPORT_SYMBOL(simple_read_from_buffer);
989 EXPORT_SYMBOL(simple_write_to_buffer);
990 EXPORT_SYMBOL(memory_read_from_buffer);
991 EXPORT_SYMBOL(simple_transaction_set);
992 EXPORT_SYMBOL(simple_transaction_get);
993 EXPORT_SYMBOL(simple_transaction_read);
994 EXPORT_SYMBOL(simple_transaction_release);
995 EXPORT_SYMBOL_GPL(simple_attr_open);
996 EXPORT_SYMBOL_GPL(simple_attr_release);
997 EXPORT_SYMBOL_GPL(simple_attr_read);
998 EXPORT_SYMBOL_GPL(simple_attr_write);
999