xref: /openbmc/linux/fs/libfs.c (revision 7b6d864b)
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, unsigned int flags)
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 = QSTR_INIT(".", 1);
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 whence)
85 {
86 	struct dentry *dentry = file->f_path.dentry;
87 	mutex_lock(&dentry->d_inode->i_mutex);
88 	switch (whence) {
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 *file, struct dir_context *ctx)
139 {
140 	struct dentry *dentry = file->f_path.dentry;
141 	struct dentry *cursor = file->private_data;
142 	struct list_head *p, *q = &cursor->d_u.d_child;
143 
144 	if (!dir_emit_dots(file, ctx))
145 		return 0;
146 	spin_lock(&dentry->d_lock);
147 	if (ctx->pos == 2)
148 		list_move(q, &dentry->d_subdirs);
149 
150 	for (p = q->next; p != &dentry->d_subdirs; p = p->next) {
151 		struct dentry *next = list_entry(p, struct dentry, d_u.d_child);
152 		spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
153 		if (!simple_positive(next)) {
154 			spin_unlock(&next->d_lock);
155 			continue;
156 		}
157 
158 		spin_unlock(&next->d_lock);
159 		spin_unlock(&dentry->d_lock);
160 		if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
161 			      next->d_inode->i_ino, dt_type(next->d_inode)))
162 			return 0;
163 		spin_lock(&dentry->d_lock);
164 		spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
165 		/* next is still alive */
166 		list_move(q, p);
167 		spin_unlock(&next->d_lock);
168 		p = q;
169 		ctx->pos++;
170 	}
171 	spin_unlock(&dentry->d_lock);
172 	return 0;
173 }
174 
175 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
176 {
177 	return -EISDIR;
178 }
179 
180 const struct file_operations simple_dir_operations = {
181 	.open		= dcache_dir_open,
182 	.release	= dcache_dir_close,
183 	.llseek		= dcache_dir_lseek,
184 	.read		= generic_read_dir,
185 	.iterate	= dcache_readdir,
186 	.fsync		= noop_fsync,
187 };
188 
189 const struct inode_operations simple_dir_inode_operations = {
190 	.lookup		= simple_lookup,
191 };
192 
193 static const struct super_operations simple_super_operations = {
194 	.statfs		= simple_statfs,
195 };
196 
197 /*
198  * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
199  * will never be mountable)
200  */
201 struct dentry *mount_pseudo(struct file_system_type *fs_type, char *name,
202 	const struct super_operations *ops,
203 	const struct dentry_operations *dops, unsigned long magic)
204 {
205 	struct super_block *s;
206 	struct dentry *dentry;
207 	struct inode *root;
208 	struct qstr d_name = QSTR_INIT(name, strlen(name));
209 
210 	s = sget(fs_type, NULL, set_anon_super, MS_NOUSER, NULL);
211 	if (IS_ERR(s))
212 		return ERR_CAST(s);
213 
214 	s->s_maxbytes = MAX_LFS_FILESIZE;
215 	s->s_blocksize = PAGE_SIZE;
216 	s->s_blocksize_bits = PAGE_SHIFT;
217 	s->s_magic = magic;
218 	s->s_op = ops ? ops : &simple_super_operations;
219 	s->s_time_gran = 1;
220 	root = new_inode(s);
221 	if (!root)
222 		goto Enomem;
223 	/*
224 	 * since this is the first inode, make it number 1. New inodes created
225 	 * after this must take care not to collide with it (by passing
226 	 * max_reserved of 1 to iunique).
227 	 */
228 	root->i_ino = 1;
229 	root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
230 	root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
231 	dentry = __d_alloc(s, &d_name);
232 	if (!dentry) {
233 		iput(root);
234 		goto Enomem;
235 	}
236 	d_instantiate(dentry, root);
237 	s->s_root = dentry;
238 	s->s_d_op = dops;
239 	s->s_flags |= MS_ACTIVE;
240 	return dget(s->s_root);
241 
242 Enomem:
243 	deactivate_locked_super(s);
244 	return ERR_PTR(-ENOMEM);
245 }
246 
247 int simple_open(struct inode *inode, struct file *file)
248 {
249 	if (inode->i_private)
250 		file->private_data = inode->i_private;
251 	return 0;
252 }
253 
254 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
255 {
256 	struct inode *inode = old_dentry->d_inode;
257 
258 	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
259 	inc_nlink(inode);
260 	ihold(inode);
261 	dget(dentry);
262 	d_instantiate(dentry, inode);
263 	return 0;
264 }
265 
266 int simple_empty(struct dentry *dentry)
267 {
268 	struct dentry *child;
269 	int ret = 0;
270 
271 	spin_lock(&dentry->d_lock);
272 	list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child) {
273 		spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
274 		if (simple_positive(child)) {
275 			spin_unlock(&child->d_lock);
276 			goto out;
277 		}
278 		spin_unlock(&child->d_lock);
279 	}
280 	ret = 1;
281 out:
282 	spin_unlock(&dentry->d_lock);
283 	return ret;
284 }
285 
286 int simple_unlink(struct inode *dir, struct dentry *dentry)
287 {
288 	struct inode *inode = dentry->d_inode;
289 
290 	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
291 	drop_nlink(inode);
292 	dput(dentry);
293 	return 0;
294 }
295 
296 int simple_rmdir(struct inode *dir, struct dentry *dentry)
297 {
298 	if (!simple_empty(dentry))
299 		return -ENOTEMPTY;
300 
301 	drop_nlink(dentry->d_inode);
302 	simple_unlink(dir, dentry);
303 	drop_nlink(dir);
304 	return 0;
305 }
306 
307 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
308 		struct inode *new_dir, struct dentry *new_dentry)
309 {
310 	struct inode *inode = old_dentry->d_inode;
311 	int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
312 
313 	if (!simple_empty(new_dentry))
314 		return -ENOTEMPTY;
315 
316 	if (new_dentry->d_inode) {
317 		simple_unlink(new_dir, new_dentry);
318 		if (they_are_dirs) {
319 			drop_nlink(new_dentry->d_inode);
320 			drop_nlink(old_dir);
321 		}
322 	} else if (they_are_dirs) {
323 		drop_nlink(old_dir);
324 		inc_nlink(new_dir);
325 	}
326 
327 	old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
328 		new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
329 
330 	return 0;
331 }
332 
333 /**
334  * simple_setattr - setattr for simple filesystem
335  * @dentry: dentry
336  * @iattr: iattr structure
337  *
338  * Returns 0 on success, -error on failure.
339  *
340  * simple_setattr is a simple ->setattr implementation without a proper
341  * implementation of size changes.
342  *
343  * It can either be used for in-memory filesystems or special files
344  * on simple regular filesystems.  Anything that needs to change on-disk
345  * or wire state on size changes needs its own setattr method.
346  */
347 int simple_setattr(struct dentry *dentry, struct iattr *iattr)
348 {
349 	struct inode *inode = dentry->d_inode;
350 	int error;
351 
352 	error = inode_change_ok(inode, iattr);
353 	if (error)
354 		return error;
355 
356 	if (iattr->ia_valid & ATTR_SIZE)
357 		truncate_setsize(inode, iattr->ia_size);
358 	setattr_copy(inode, iattr);
359 	mark_inode_dirty(inode);
360 	return 0;
361 }
362 EXPORT_SYMBOL(simple_setattr);
363 
364 int simple_readpage(struct file *file, struct page *page)
365 {
366 	clear_highpage(page);
367 	flush_dcache_page(page);
368 	SetPageUptodate(page);
369 	unlock_page(page);
370 	return 0;
371 }
372 
373 int simple_write_begin(struct file *file, struct address_space *mapping,
374 			loff_t pos, unsigned len, unsigned flags,
375 			struct page **pagep, void **fsdata)
376 {
377 	struct page *page;
378 	pgoff_t index;
379 
380 	index = pos >> PAGE_CACHE_SHIFT;
381 
382 	page = grab_cache_page_write_begin(mapping, index, flags);
383 	if (!page)
384 		return -ENOMEM;
385 
386 	*pagep = page;
387 
388 	if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
389 		unsigned from = pos & (PAGE_CACHE_SIZE - 1);
390 
391 		zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
392 	}
393 	return 0;
394 }
395 
396 /**
397  * simple_write_end - .write_end helper for non-block-device FSes
398  * @available: See .write_end of address_space_operations
399  * @file: 		"
400  * @mapping: 		"
401  * @pos: 		"
402  * @len: 		"
403  * @copied: 		"
404  * @page: 		"
405  * @fsdata: 		"
406  *
407  * simple_write_end does the minimum needed for updating a page after writing is
408  * done. It has the same API signature as the .write_end of
409  * address_space_operations vector. So it can just be set onto .write_end for
410  * FSes that don't need any other processing. i_mutex is assumed to be held.
411  * Block based filesystems should use generic_write_end().
412  * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
413  * is not called, so a filesystem that actually does store data in .write_inode
414  * should extend on what's done here with a call to mark_inode_dirty() in the
415  * case that i_size has changed.
416  */
417 int simple_write_end(struct file *file, struct address_space *mapping,
418 			loff_t pos, unsigned len, unsigned copied,
419 			struct page *page, void *fsdata)
420 {
421 	struct inode *inode = page->mapping->host;
422 	loff_t last_pos = pos + copied;
423 
424 	/* zero the stale part of the page if we did a short copy */
425 	if (copied < len) {
426 		unsigned from = pos & (PAGE_CACHE_SIZE - 1);
427 
428 		zero_user(page, from + copied, len - copied);
429 	}
430 
431 	if (!PageUptodate(page))
432 		SetPageUptodate(page);
433 	/*
434 	 * No need to use i_size_read() here, the i_size
435 	 * cannot change under us because we hold the i_mutex.
436 	 */
437 	if (last_pos > inode->i_size)
438 		i_size_write(inode, last_pos);
439 
440 	set_page_dirty(page);
441 	unlock_page(page);
442 	page_cache_release(page);
443 
444 	return copied;
445 }
446 
447 /*
448  * the inodes created here are not hashed. If you use iunique to generate
449  * unique inode values later for this filesystem, then you must take care
450  * to pass it an appropriate max_reserved value to avoid collisions.
451  */
452 int simple_fill_super(struct super_block *s, unsigned long magic,
453 		      struct tree_descr *files)
454 {
455 	struct inode *inode;
456 	struct dentry *root;
457 	struct dentry *dentry;
458 	int i;
459 
460 	s->s_blocksize = PAGE_CACHE_SIZE;
461 	s->s_blocksize_bits = PAGE_CACHE_SHIFT;
462 	s->s_magic = magic;
463 	s->s_op = &simple_super_operations;
464 	s->s_time_gran = 1;
465 
466 	inode = new_inode(s);
467 	if (!inode)
468 		return -ENOMEM;
469 	/*
470 	 * because the root inode is 1, the files array must not contain an
471 	 * entry at index 1
472 	 */
473 	inode->i_ino = 1;
474 	inode->i_mode = S_IFDIR | 0755;
475 	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
476 	inode->i_op = &simple_dir_inode_operations;
477 	inode->i_fop = &simple_dir_operations;
478 	set_nlink(inode, 2);
479 	root = d_make_root(inode);
480 	if (!root)
481 		return -ENOMEM;
482 	for (i = 0; !files->name || files->name[0]; i++, files++) {
483 		if (!files->name)
484 			continue;
485 
486 		/* warn if it tries to conflict with the root inode */
487 		if (unlikely(i == 1))
488 			printk(KERN_WARNING "%s: %s passed in a files array"
489 				"with an index of 1!\n", __func__,
490 				s->s_type->name);
491 
492 		dentry = d_alloc_name(root, files->name);
493 		if (!dentry)
494 			goto out;
495 		inode = new_inode(s);
496 		if (!inode) {
497 			dput(dentry);
498 			goto out;
499 		}
500 		inode->i_mode = S_IFREG | files->mode;
501 		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
502 		inode->i_fop = files->ops;
503 		inode->i_ino = i;
504 		d_add(dentry, inode);
505 	}
506 	s->s_root = root;
507 	return 0;
508 out:
509 	d_genocide(root);
510 	shrink_dcache_parent(root);
511 	dput(root);
512 	return -ENOMEM;
513 }
514 
515 static DEFINE_SPINLOCK(pin_fs_lock);
516 
517 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
518 {
519 	struct vfsmount *mnt = NULL;
520 	spin_lock(&pin_fs_lock);
521 	if (unlikely(!*mount)) {
522 		spin_unlock(&pin_fs_lock);
523 		mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, NULL);
524 		if (IS_ERR(mnt))
525 			return PTR_ERR(mnt);
526 		spin_lock(&pin_fs_lock);
527 		if (!*mount)
528 			*mount = mnt;
529 	}
530 	mntget(*mount);
531 	++*count;
532 	spin_unlock(&pin_fs_lock);
533 	mntput(mnt);
534 	return 0;
535 }
536 
537 void simple_release_fs(struct vfsmount **mount, int *count)
538 {
539 	struct vfsmount *mnt;
540 	spin_lock(&pin_fs_lock);
541 	mnt = *mount;
542 	if (!--*count)
543 		*mount = NULL;
544 	spin_unlock(&pin_fs_lock);
545 	mntput(mnt);
546 }
547 
548 /**
549  * simple_read_from_buffer - copy data from the buffer to user space
550  * @to: the user space buffer to read to
551  * @count: the maximum number of bytes to read
552  * @ppos: the current position in the buffer
553  * @from: the buffer to read from
554  * @available: the size of the buffer
555  *
556  * The simple_read_from_buffer() function reads up to @count bytes from the
557  * buffer @from at offset @ppos into the user space address starting at @to.
558  *
559  * On success, the number of bytes read is returned and the offset @ppos is
560  * advanced by this number, or negative value is returned on error.
561  **/
562 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
563 				const void *from, size_t available)
564 {
565 	loff_t pos = *ppos;
566 	size_t ret;
567 
568 	if (pos < 0)
569 		return -EINVAL;
570 	if (pos >= available || !count)
571 		return 0;
572 	if (count > available - pos)
573 		count = available - pos;
574 	ret = copy_to_user(to, from + pos, count);
575 	if (ret == count)
576 		return -EFAULT;
577 	count -= ret;
578 	*ppos = pos + count;
579 	return count;
580 }
581 
582 /**
583  * simple_write_to_buffer - copy data from user space to the buffer
584  * @to: the buffer to write to
585  * @available: the size of the buffer
586  * @ppos: the current position in the buffer
587  * @from: the user space buffer to read from
588  * @count: the maximum number of bytes to read
589  *
590  * The simple_write_to_buffer() function reads up to @count bytes from the user
591  * space address starting at @from into the buffer @to at offset @ppos.
592  *
593  * On success, the number of bytes written is returned and the offset @ppos is
594  * advanced by this number, or negative value is returned on error.
595  **/
596 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
597 		const void __user *from, size_t count)
598 {
599 	loff_t pos = *ppos;
600 	size_t res;
601 
602 	if (pos < 0)
603 		return -EINVAL;
604 	if (pos >= available || !count)
605 		return 0;
606 	if (count > available - pos)
607 		count = available - pos;
608 	res = copy_from_user(to + pos, from, count);
609 	if (res == count)
610 		return -EFAULT;
611 	count -= res;
612 	*ppos = pos + count;
613 	return count;
614 }
615 
616 /**
617  * memory_read_from_buffer - copy data from the buffer
618  * @to: the kernel space buffer to read to
619  * @count: the maximum number of bytes to read
620  * @ppos: the current position in the buffer
621  * @from: the buffer to read from
622  * @available: the size of the buffer
623  *
624  * The memory_read_from_buffer() function reads up to @count bytes from the
625  * buffer @from at offset @ppos into the kernel space address starting at @to.
626  *
627  * On success, the number of bytes read is returned and the offset @ppos is
628  * advanced by this number, or negative value is returned on error.
629  **/
630 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
631 				const void *from, size_t available)
632 {
633 	loff_t pos = *ppos;
634 
635 	if (pos < 0)
636 		return -EINVAL;
637 	if (pos >= available)
638 		return 0;
639 	if (count > available - pos)
640 		count = available - pos;
641 	memcpy(to, from + pos, count);
642 	*ppos = pos + count;
643 
644 	return count;
645 }
646 
647 /*
648  * Transaction based IO.
649  * The file expects a single write which triggers the transaction, and then
650  * possibly a read which collects the result - which is stored in a
651  * file-local buffer.
652  */
653 
654 void simple_transaction_set(struct file *file, size_t n)
655 {
656 	struct simple_transaction_argresp *ar = file->private_data;
657 
658 	BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
659 
660 	/*
661 	 * The barrier ensures that ar->size will really remain zero until
662 	 * ar->data is ready for reading.
663 	 */
664 	smp_mb();
665 	ar->size = n;
666 }
667 
668 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
669 {
670 	struct simple_transaction_argresp *ar;
671 	static DEFINE_SPINLOCK(simple_transaction_lock);
672 
673 	if (size > SIMPLE_TRANSACTION_LIMIT - 1)
674 		return ERR_PTR(-EFBIG);
675 
676 	ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
677 	if (!ar)
678 		return ERR_PTR(-ENOMEM);
679 
680 	spin_lock(&simple_transaction_lock);
681 
682 	/* only one write allowed per open */
683 	if (file->private_data) {
684 		spin_unlock(&simple_transaction_lock);
685 		free_page((unsigned long)ar);
686 		return ERR_PTR(-EBUSY);
687 	}
688 
689 	file->private_data = ar;
690 
691 	spin_unlock(&simple_transaction_lock);
692 
693 	if (copy_from_user(ar->data, buf, size))
694 		return ERR_PTR(-EFAULT);
695 
696 	return ar->data;
697 }
698 
699 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
700 {
701 	struct simple_transaction_argresp *ar = file->private_data;
702 
703 	if (!ar)
704 		return 0;
705 	return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
706 }
707 
708 int simple_transaction_release(struct inode *inode, struct file *file)
709 {
710 	free_page((unsigned long)file->private_data);
711 	return 0;
712 }
713 
714 /* Simple attribute files */
715 
716 struct simple_attr {
717 	int (*get)(void *, u64 *);
718 	int (*set)(void *, u64);
719 	char get_buf[24];	/* enough to store a u64 and "\n\0" */
720 	char set_buf[24];
721 	void *data;
722 	const char *fmt;	/* format for read operation */
723 	struct mutex mutex;	/* protects access to these buffers */
724 };
725 
726 /* simple_attr_open is called by an actual attribute open file operation
727  * to set the attribute specific access operations. */
728 int simple_attr_open(struct inode *inode, struct file *file,
729 		     int (*get)(void *, u64 *), int (*set)(void *, u64),
730 		     const char *fmt)
731 {
732 	struct simple_attr *attr;
733 
734 	attr = kmalloc(sizeof(*attr), GFP_KERNEL);
735 	if (!attr)
736 		return -ENOMEM;
737 
738 	attr->get = get;
739 	attr->set = set;
740 	attr->data = inode->i_private;
741 	attr->fmt = fmt;
742 	mutex_init(&attr->mutex);
743 
744 	file->private_data = attr;
745 
746 	return nonseekable_open(inode, file);
747 }
748 
749 int simple_attr_release(struct inode *inode, struct file *file)
750 {
751 	kfree(file->private_data);
752 	return 0;
753 }
754 
755 /* read from the buffer that is filled with the get function */
756 ssize_t simple_attr_read(struct file *file, char __user *buf,
757 			 size_t len, loff_t *ppos)
758 {
759 	struct simple_attr *attr;
760 	size_t size;
761 	ssize_t ret;
762 
763 	attr = file->private_data;
764 
765 	if (!attr->get)
766 		return -EACCES;
767 
768 	ret = mutex_lock_interruptible(&attr->mutex);
769 	if (ret)
770 		return ret;
771 
772 	if (*ppos) {		/* continued read */
773 		size = strlen(attr->get_buf);
774 	} else {		/* first read */
775 		u64 val;
776 		ret = attr->get(attr->data, &val);
777 		if (ret)
778 			goto out;
779 
780 		size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
781 				 attr->fmt, (unsigned long long)val);
782 	}
783 
784 	ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
785 out:
786 	mutex_unlock(&attr->mutex);
787 	return ret;
788 }
789 
790 /* interpret the buffer as a number to call the set function with */
791 ssize_t simple_attr_write(struct file *file, const char __user *buf,
792 			  size_t len, loff_t *ppos)
793 {
794 	struct simple_attr *attr;
795 	u64 val;
796 	size_t size;
797 	ssize_t ret;
798 
799 	attr = file->private_data;
800 	if (!attr->set)
801 		return -EACCES;
802 
803 	ret = mutex_lock_interruptible(&attr->mutex);
804 	if (ret)
805 		return ret;
806 
807 	ret = -EFAULT;
808 	size = min(sizeof(attr->set_buf) - 1, len);
809 	if (copy_from_user(attr->set_buf, buf, size))
810 		goto out;
811 
812 	attr->set_buf[size] = '\0';
813 	val = simple_strtoll(attr->set_buf, NULL, 0);
814 	ret = attr->set(attr->data, val);
815 	if (ret == 0)
816 		ret = len; /* on success, claim we got the whole input */
817 out:
818 	mutex_unlock(&attr->mutex);
819 	return ret;
820 }
821 
822 /**
823  * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
824  * @sb:		filesystem to do the file handle conversion on
825  * @fid:	file handle to convert
826  * @fh_len:	length of the file handle in bytes
827  * @fh_type:	type of file handle
828  * @get_inode:	filesystem callback to retrieve inode
829  *
830  * This function decodes @fid as long as it has one of the well-known
831  * Linux filehandle types and calls @get_inode on it to retrieve the
832  * inode for the object specified in the file handle.
833  */
834 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
835 		int fh_len, int fh_type, struct inode *(*get_inode)
836 			(struct super_block *sb, u64 ino, u32 gen))
837 {
838 	struct inode *inode = NULL;
839 
840 	if (fh_len < 2)
841 		return NULL;
842 
843 	switch (fh_type) {
844 	case FILEID_INO32_GEN:
845 	case FILEID_INO32_GEN_PARENT:
846 		inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
847 		break;
848 	}
849 
850 	return d_obtain_alias(inode);
851 }
852 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
853 
854 /**
855  * generic_fh_to_parent - generic helper for the fh_to_parent export operation
856  * @sb:		filesystem to do the file handle conversion on
857  * @fid:	file handle to convert
858  * @fh_len:	length of the file handle in bytes
859  * @fh_type:	type of file handle
860  * @get_inode:	filesystem callback to retrieve inode
861  *
862  * This function decodes @fid as long as it has one of the well-known
863  * Linux filehandle types and calls @get_inode on it to retrieve the
864  * inode for the _parent_ object specified in the file handle if it
865  * is specified in the file handle, or NULL otherwise.
866  */
867 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
868 		int fh_len, int fh_type, struct inode *(*get_inode)
869 			(struct super_block *sb, u64 ino, u32 gen))
870 {
871 	struct inode *inode = NULL;
872 
873 	if (fh_len <= 2)
874 		return NULL;
875 
876 	switch (fh_type) {
877 	case FILEID_INO32_GEN_PARENT:
878 		inode = get_inode(sb, fid->i32.parent_ino,
879 				  (fh_len > 3 ? fid->i32.parent_gen : 0));
880 		break;
881 	}
882 
883 	return d_obtain_alias(inode);
884 }
885 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
886 
887 /**
888  * generic_file_fsync - generic fsync implementation for simple filesystems
889  * @file:	file to synchronize
890  * @datasync:	only synchronize essential metadata if true
891  *
892  * This is a generic implementation of the fsync method for simple
893  * filesystems which track all non-inode metadata in the buffers list
894  * hanging off the address_space structure.
895  */
896 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
897 		       int datasync)
898 {
899 	struct inode *inode = file->f_mapping->host;
900 	int err;
901 	int ret;
902 
903 	err = filemap_write_and_wait_range(inode->i_mapping, start, end);
904 	if (err)
905 		return err;
906 
907 	mutex_lock(&inode->i_mutex);
908 	ret = sync_mapping_buffers(inode->i_mapping);
909 	if (!(inode->i_state & I_DIRTY))
910 		goto out;
911 	if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
912 		goto out;
913 
914 	err = sync_inode_metadata(inode, 1);
915 	if (ret == 0)
916 		ret = err;
917 out:
918 	mutex_unlock(&inode->i_mutex);
919 	return ret;
920 }
921 EXPORT_SYMBOL(generic_file_fsync);
922 
923 /**
924  * generic_check_addressable - Check addressability of file system
925  * @blocksize_bits:	log of file system block size
926  * @num_blocks:		number of blocks in file system
927  *
928  * Determine whether a file system with @num_blocks blocks (and a
929  * block size of 2**@blocksize_bits) is addressable by the sector_t
930  * and page cache of the system.  Return 0 if so and -EFBIG otherwise.
931  */
932 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
933 {
934 	u64 last_fs_block = num_blocks - 1;
935 	u64 last_fs_page =
936 		last_fs_block >> (PAGE_CACHE_SHIFT - blocksize_bits);
937 
938 	if (unlikely(num_blocks == 0))
939 		return 0;
940 
941 	if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT))
942 		return -EINVAL;
943 
944 	if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
945 	    (last_fs_page > (pgoff_t)(~0ULL))) {
946 		return -EFBIG;
947 	}
948 	return 0;
949 }
950 EXPORT_SYMBOL(generic_check_addressable);
951 
952 /*
953  * No-op implementation of ->fsync for in-memory filesystems.
954  */
955 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
956 {
957 	return 0;
958 }
959 
960 EXPORT_SYMBOL(dcache_dir_close);
961 EXPORT_SYMBOL(dcache_dir_lseek);
962 EXPORT_SYMBOL(dcache_dir_open);
963 EXPORT_SYMBOL(dcache_readdir);
964 EXPORT_SYMBOL(generic_read_dir);
965 EXPORT_SYMBOL(mount_pseudo);
966 EXPORT_SYMBOL(simple_write_begin);
967 EXPORT_SYMBOL(simple_write_end);
968 EXPORT_SYMBOL(simple_dir_inode_operations);
969 EXPORT_SYMBOL(simple_dir_operations);
970 EXPORT_SYMBOL(simple_empty);
971 EXPORT_SYMBOL(simple_fill_super);
972 EXPORT_SYMBOL(simple_getattr);
973 EXPORT_SYMBOL(simple_open);
974 EXPORT_SYMBOL(simple_link);
975 EXPORT_SYMBOL(simple_lookup);
976 EXPORT_SYMBOL(simple_pin_fs);
977 EXPORT_SYMBOL(simple_readpage);
978 EXPORT_SYMBOL(simple_release_fs);
979 EXPORT_SYMBOL(simple_rename);
980 EXPORT_SYMBOL(simple_rmdir);
981 EXPORT_SYMBOL(simple_statfs);
982 EXPORT_SYMBOL(noop_fsync);
983 EXPORT_SYMBOL(simple_unlink);
984 EXPORT_SYMBOL(simple_read_from_buffer);
985 EXPORT_SYMBOL(simple_write_to_buffer);
986 EXPORT_SYMBOL(memory_read_from_buffer);
987 EXPORT_SYMBOL(simple_transaction_set);
988 EXPORT_SYMBOL(simple_transaction_get);
989 EXPORT_SYMBOL(simple_transaction_read);
990 EXPORT_SYMBOL(simple_transaction_release);
991 EXPORT_SYMBOL_GPL(simple_attr_open);
992 EXPORT_SYMBOL_GPL(simple_attr_release);
993 EXPORT_SYMBOL_GPL(simple_attr_read);
994 EXPORT_SYMBOL_GPL(simple_attr_write);
995