1/*
2 *	fs/libfs.c
3 *	Library for filesystems writers.
4 */
5
6#include <linux/blkdev.h>
7#include <linux/export.h>
8#include <linux/pagemap.h>
9#include <linux/slab.h>
10#include <linux/mount.h>
11#include <linux/vfs.h>
12#include <linux/quotaops.h>
13#include <linux/mutex.h>
14#include <linux/namei.h>
15#include <linux/exportfs.h>
16#include <linux/writeback.h>
17#include <linux/buffer_head.h> /* sync_mapping_buffers */
18
19#include <asm/uaccess.h>
20
21#include "internal.h"
22
23static inline int simple_positive(struct dentry *dentry)
24{
25	return d_really_is_positive(dentry) && !d_unhashed(dentry);
26}
27
28int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
29		   struct kstat *stat)
30{
31	struct inode *inode = d_inode(dentry);
32	generic_fillattr(inode, stat);
33	stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_SHIFT - 9);
34	return 0;
35}
36EXPORT_SYMBOL(simple_getattr);
37
38int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
39{
40	buf->f_type = dentry->d_sb->s_magic;
41	buf->f_bsize = PAGE_CACHE_SIZE;
42	buf->f_namelen = NAME_MAX;
43	return 0;
44}
45EXPORT_SYMBOL(simple_statfs);
46
47/*
48 * Retaining negative dentries for an in-memory filesystem just wastes
49 * memory and lookup time: arrange for them to be deleted immediately.
50 */
51int always_delete_dentry(const struct dentry *dentry)
52{
53	return 1;
54}
55EXPORT_SYMBOL(always_delete_dentry);
56
57const struct dentry_operations simple_dentry_operations = {
58	.d_delete = always_delete_dentry,
59};
60EXPORT_SYMBOL(simple_dentry_operations);
61
62/*
63 * Lookup the data. This is trivial - if the dentry didn't already
64 * exist, we know it is negative.  Set d_op to delete negative dentries.
65 */
66struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
67{
68	if (dentry->d_name.len > NAME_MAX)
69		return ERR_PTR(-ENAMETOOLONG);
70	if (!dentry->d_sb->s_d_op)
71		d_set_d_op(dentry, &simple_dentry_operations);
72	d_add(dentry, NULL);
73	return NULL;
74}
75EXPORT_SYMBOL(simple_lookup);
76
77int dcache_dir_open(struct inode *inode, struct file *file)
78{
79	static struct qstr cursor_name = QSTR_INIT(".", 1);
80
81	file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
82
83	return file->private_data ? 0 : -ENOMEM;
84}
85EXPORT_SYMBOL(dcache_dir_open);
86
87int dcache_dir_close(struct inode *inode, struct file *file)
88{
89	dput(file->private_data);
90	return 0;
91}
92EXPORT_SYMBOL(dcache_dir_close);
93
94loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
95{
96	struct dentry *dentry = file->f_path.dentry;
97	mutex_lock(&d_inode(dentry)->i_mutex);
98	switch (whence) {
99		case 1:
100			offset += file->f_pos;
101		case 0:
102			if (offset >= 0)
103				break;
104		default:
105			mutex_unlock(&d_inode(dentry)->i_mutex);
106			return -EINVAL;
107	}
108	if (offset != file->f_pos) {
109		file->f_pos = offset;
110		if (file->f_pos >= 2) {
111			struct list_head *p;
112			struct dentry *cursor = file->private_data;
113			loff_t n = file->f_pos - 2;
114
115			spin_lock(&dentry->d_lock);
116			/* d_lock not required for cursor */
117			list_del(&cursor->d_child);
118			p = dentry->d_subdirs.next;
119			while (n && p != &dentry->d_subdirs) {
120				struct dentry *next;
121				next = list_entry(p, struct dentry, d_child);
122				spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
123				if (simple_positive(next))
124					n--;
125				spin_unlock(&next->d_lock);
126				p = p->next;
127			}
128			list_add_tail(&cursor->d_child, p);
129			spin_unlock(&dentry->d_lock);
130		}
131	}
132	mutex_unlock(&d_inode(dentry)->i_mutex);
133	return offset;
134}
135EXPORT_SYMBOL(dcache_dir_lseek);
136
137/* Relationship between i_mode and the DT_xxx types */
138static inline unsigned char dt_type(struct inode *inode)
139{
140	return (inode->i_mode >> 12) & 15;
141}
142
143/*
144 * Directory is locked and all positive dentries in it are safe, since
145 * for ramfs-type trees they can't go away without unlink() or rmdir(),
146 * both impossible due to the lock on directory.
147 */
148
149int dcache_readdir(struct file *file, struct dir_context *ctx)
150{
151	struct dentry *dentry = file->f_path.dentry;
152	struct dentry *cursor = file->private_data;
153	struct list_head *p, *q = &cursor->d_child;
154
155	if (!dir_emit_dots(file, ctx))
156		return 0;
157	spin_lock(&dentry->d_lock);
158	if (ctx->pos == 2)
159		list_move(q, &dentry->d_subdirs);
160
161	for (p = q->next; p != &dentry->d_subdirs; p = p->next) {
162		struct dentry *next = list_entry(p, struct dentry, d_child);
163		spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
164		if (!simple_positive(next)) {
165			spin_unlock(&next->d_lock);
166			continue;
167		}
168
169		spin_unlock(&next->d_lock);
170		spin_unlock(&dentry->d_lock);
171		if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
172			      d_inode(next)->i_ino, dt_type(d_inode(next))))
173			return 0;
174		spin_lock(&dentry->d_lock);
175		spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
176		/* next is still alive */
177		list_move(q, p);
178		spin_unlock(&next->d_lock);
179		p = q;
180		ctx->pos++;
181	}
182	spin_unlock(&dentry->d_lock);
183	return 0;
184}
185EXPORT_SYMBOL(dcache_readdir);
186
187ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
188{
189	return -EISDIR;
190}
191EXPORT_SYMBOL(generic_read_dir);
192
193const struct file_operations simple_dir_operations = {
194	.open		= dcache_dir_open,
195	.release	= dcache_dir_close,
196	.llseek		= dcache_dir_lseek,
197	.read		= generic_read_dir,
198	.iterate	= dcache_readdir,
199	.fsync		= noop_fsync,
200};
201EXPORT_SYMBOL(simple_dir_operations);
202
203const struct inode_operations simple_dir_inode_operations = {
204	.lookup		= simple_lookup,
205};
206EXPORT_SYMBOL(simple_dir_inode_operations);
207
208static const struct super_operations simple_super_operations = {
209	.statfs		= simple_statfs,
210};
211
212/*
213 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
214 * will never be mountable)
215 */
216struct dentry *mount_pseudo(struct file_system_type *fs_type, char *name,
217	const struct super_operations *ops,
218	const struct dentry_operations *dops, unsigned long magic)
219{
220	struct super_block *s;
221	struct dentry *dentry;
222	struct inode *root;
223	struct qstr d_name = QSTR_INIT(name, strlen(name));
224
225	s = sget(fs_type, NULL, set_anon_super, MS_NOUSER, NULL);
226	if (IS_ERR(s))
227		return ERR_CAST(s);
228
229	s->s_maxbytes = MAX_LFS_FILESIZE;
230	s->s_blocksize = PAGE_SIZE;
231	s->s_blocksize_bits = PAGE_SHIFT;
232	s->s_magic = magic;
233	s->s_op = ops ? ops : &simple_super_operations;
234	s->s_time_gran = 1;
235	root = new_inode(s);
236	if (!root)
237		goto Enomem;
238	/*
239	 * since this is the first inode, make it number 1. New inodes created
240	 * after this must take care not to collide with it (by passing
241	 * max_reserved of 1 to iunique).
242	 */
243	root->i_ino = 1;
244	root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
245	root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
246	dentry = __d_alloc(s, &d_name);
247	if (!dentry) {
248		iput(root);
249		goto Enomem;
250	}
251	d_instantiate(dentry, root);
252	s->s_root = dentry;
253	s->s_d_op = dops;
254	s->s_flags |= MS_ACTIVE;
255	return dget(s->s_root);
256
257Enomem:
258	deactivate_locked_super(s);
259	return ERR_PTR(-ENOMEM);
260}
261EXPORT_SYMBOL(mount_pseudo);
262
263int simple_open(struct inode *inode, struct file *file)
264{
265	if (inode->i_private)
266		file->private_data = inode->i_private;
267	return 0;
268}
269EXPORT_SYMBOL(simple_open);
270
271int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
272{
273	struct inode *inode = d_inode(old_dentry);
274
275	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
276	inc_nlink(inode);
277	ihold(inode);
278	dget(dentry);
279	d_instantiate(dentry, inode);
280	return 0;
281}
282EXPORT_SYMBOL(simple_link);
283
284int simple_empty(struct dentry *dentry)
285{
286	struct dentry *child;
287	int ret = 0;
288
289	spin_lock(&dentry->d_lock);
290	list_for_each_entry(child, &dentry->d_subdirs, d_child) {
291		spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
292		if (simple_positive(child)) {
293			spin_unlock(&child->d_lock);
294			goto out;
295		}
296		spin_unlock(&child->d_lock);
297	}
298	ret = 1;
299out:
300	spin_unlock(&dentry->d_lock);
301	return ret;
302}
303EXPORT_SYMBOL(simple_empty);
304
305int simple_unlink(struct inode *dir, struct dentry *dentry)
306{
307	struct inode *inode = d_inode(dentry);
308
309	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
310	drop_nlink(inode);
311	dput(dentry);
312	return 0;
313}
314EXPORT_SYMBOL(simple_unlink);
315
316int simple_rmdir(struct inode *dir, struct dentry *dentry)
317{
318	if (!simple_empty(dentry))
319		return -ENOTEMPTY;
320
321	drop_nlink(d_inode(dentry));
322	simple_unlink(dir, dentry);
323	drop_nlink(dir);
324	return 0;
325}
326EXPORT_SYMBOL(simple_rmdir);
327
328int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
329		struct inode *new_dir, struct dentry *new_dentry)
330{
331	struct inode *inode = d_inode(old_dentry);
332	int they_are_dirs = d_is_dir(old_dentry);
333
334	if (!simple_empty(new_dentry))
335		return -ENOTEMPTY;
336
337	if (d_really_is_positive(new_dentry)) {
338		simple_unlink(new_dir, new_dentry);
339		if (they_are_dirs) {
340			drop_nlink(d_inode(new_dentry));
341			drop_nlink(old_dir);
342		}
343	} else if (they_are_dirs) {
344		drop_nlink(old_dir);
345		inc_nlink(new_dir);
346	}
347
348	old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
349		new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
350
351	return 0;
352}
353EXPORT_SYMBOL(simple_rename);
354
355/**
356 * simple_setattr - setattr for simple filesystem
357 * @dentry: dentry
358 * @iattr: iattr structure
359 *
360 * Returns 0 on success, -error on failure.
361 *
362 * simple_setattr is a simple ->setattr implementation without a proper
363 * implementation of size changes.
364 *
365 * It can either be used for in-memory filesystems or special files
366 * on simple regular filesystems.  Anything that needs to change on-disk
367 * or wire state on size changes needs its own setattr method.
368 */
369int simple_setattr(struct dentry *dentry, struct iattr *iattr)
370{
371	struct inode *inode = d_inode(dentry);
372	int error;
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}
384EXPORT_SYMBOL(simple_setattr);
385
386int 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}
394EXPORT_SYMBOL(simple_readpage);
395
396int simple_write_begin(struct file *file, struct address_space *mapping,
397			loff_t pos, unsigned len, unsigned flags,
398			struct page **pagep, void **fsdata)
399{
400	struct page *page;
401	pgoff_t index;
402
403	index = pos >> PAGE_CACHE_SHIFT;
404
405	page = grab_cache_page_write_begin(mapping, index, flags);
406	if (!page)
407		return -ENOMEM;
408
409	*pagep = page;
410
411	if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
412		unsigned from = pos & (PAGE_CACHE_SIZE - 1);
413
414		zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
415	}
416	return 0;
417}
418EXPORT_SYMBOL(simple_write_begin);
419
420/**
421 * simple_write_end - .write_end helper for non-block-device FSes
422 * @available: See .write_end of address_space_operations
423 * @file: 		"
424 * @mapping: 		"
425 * @pos: 		"
426 * @len: 		"
427 * @copied: 		"
428 * @page: 		"
429 * @fsdata: 		"
430 *
431 * simple_write_end does the minimum needed for updating a page after writing is
432 * done. It has the same API signature as the .write_end of
433 * address_space_operations vector. So it can just be set onto .write_end for
434 * FSes that don't need any other processing. i_mutex is assumed to be held.
435 * Block based filesystems should use generic_write_end().
436 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
437 * is not called, so a filesystem that actually does store data in .write_inode
438 * should extend on what's done here with a call to mark_inode_dirty() in the
439 * case that i_size has changed.
440 */
441int simple_write_end(struct file *file, struct address_space *mapping,
442			loff_t pos, unsigned len, unsigned copied,
443			struct page *page, void *fsdata)
444{
445	struct inode *inode = page->mapping->host;
446	loff_t last_pos = pos + copied;
447
448	/* zero the stale part of the page if we did a short copy */
449	if (copied < len) {
450		unsigned from = pos & (PAGE_CACHE_SIZE - 1);
451
452		zero_user(page, from + copied, len - copied);
453	}
454
455	if (!PageUptodate(page))
456		SetPageUptodate(page);
457	/*
458	 * No need to use i_size_read() here, the i_size
459	 * cannot change under us because we hold the i_mutex.
460	 */
461	if (last_pos > inode->i_size)
462		i_size_write(inode, last_pos);
463
464	set_page_dirty(page);
465	unlock_page(page);
466	page_cache_release(page);
467
468	return copied;
469}
470EXPORT_SYMBOL(simple_write_end);
471
472/*
473 * the inodes created here are not hashed. If you use iunique to generate
474 * unique inode values later for this filesystem, then you must take care
475 * to pass it an appropriate max_reserved value to avoid collisions.
476 */
477int simple_fill_super(struct super_block *s, unsigned long magic,
478		      struct tree_descr *files)
479{
480	struct inode *inode;
481	struct dentry *root;
482	struct dentry *dentry;
483	int i;
484
485	s->s_blocksize = PAGE_CACHE_SIZE;
486	s->s_blocksize_bits = PAGE_CACHE_SHIFT;
487	s->s_magic = magic;
488	s->s_op = &simple_super_operations;
489	s->s_time_gran = 1;
490
491	inode = new_inode(s);
492	if (!inode)
493		return -ENOMEM;
494	/*
495	 * because the root inode is 1, the files array must not contain an
496	 * entry at index 1
497	 */
498	inode->i_ino = 1;
499	inode->i_mode = S_IFDIR | 0755;
500	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
501	inode->i_op = &simple_dir_inode_operations;
502	inode->i_fop = &simple_dir_operations;
503	set_nlink(inode, 2);
504	root = d_make_root(inode);
505	if (!root)
506		return -ENOMEM;
507	for (i = 0; !files->name || files->name[0]; i++, files++) {
508		if (!files->name)
509			continue;
510
511		/* warn if it tries to conflict with the root inode */
512		if (unlikely(i == 1))
513			printk(KERN_WARNING "%s: %s passed in a files array"
514				"with an index of 1!\n", __func__,
515				s->s_type->name);
516
517		dentry = d_alloc_name(root, files->name);
518		if (!dentry)
519			goto out;
520		inode = new_inode(s);
521		if (!inode) {
522			dput(dentry);
523			goto out;
524		}
525		inode->i_mode = S_IFREG | files->mode;
526		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
527		inode->i_fop = files->ops;
528		inode->i_ino = i;
529		d_add(dentry, inode);
530	}
531	s->s_root = root;
532	return 0;
533out:
534	d_genocide(root);
535	shrink_dcache_parent(root);
536	dput(root);
537	return -ENOMEM;
538}
539EXPORT_SYMBOL(simple_fill_super);
540
541static DEFINE_SPINLOCK(pin_fs_lock);
542
543int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
544{
545	struct vfsmount *mnt = NULL;
546	spin_lock(&pin_fs_lock);
547	if (unlikely(!*mount)) {
548		spin_unlock(&pin_fs_lock);
549		mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, NULL);
550		if (IS_ERR(mnt))
551			return PTR_ERR(mnt);
552		spin_lock(&pin_fs_lock);
553		if (!*mount)
554			*mount = mnt;
555	}
556	mntget(*mount);
557	++*count;
558	spin_unlock(&pin_fs_lock);
559	mntput(mnt);
560	return 0;
561}
562EXPORT_SYMBOL(simple_pin_fs);
563
564void simple_release_fs(struct vfsmount **mount, int *count)
565{
566	struct vfsmount *mnt;
567	spin_lock(&pin_fs_lock);
568	mnt = *mount;
569	if (!--*count)
570		*mount = NULL;
571	spin_unlock(&pin_fs_lock);
572	mntput(mnt);
573}
574EXPORT_SYMBOL(simple_release_fs);
575
576/**
577 * simple_read_from_buffer - copy data from the buffer to user space
578 * @to: the user space buffer to read to
579 * @count: the maximum number of bytes to read
580 * @ppos: the current position in the buffer
581 * @from: the buffer to read from
582 * @available: the size of the buffer
583 *
584 * The simple_read_from_buffer() function reads up to @count bytes from the
585 * buffer @from at offset @ppos into the user space address starting at @to.
586 *
587 * On success, the number of bytes read is returned and the offset @ppos is
588 * advanced by this number, or negative value is returned on error.
589 **/
590ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
591				const void *from, size_t available)
592{
593	loff_t pos = *ppos;
594	size_t ret;
595
596	if (pos < 0)
597		return -EINVAL;
598	if (pos >= available || !count)
599		return 0;
600	if (count > available - pos)
601		count = available - pos;
602	ret = copy_to_user(to, from + pos, count);
603	if (ret == count)
604		return -EFAULT;
605	count -= ret;
606	*ppos = pos + count;
607	return count;
608}
609EXPORT_SYMBOL(simple_read_from_buffer);
610
611/**
612 * simple_write_to_buffer - copy data from user space to the buffer
613 * @to: the buffer to write to
614 * @available: the size of the buffer
615 * @ppos: the current position in the buffer
616 * @from: the user space buffer to read from
617 * @count: the maximum number of bytes to read
618 *
619 * The simple_write_to_buffer() function reads up to @count bytes from the user
620 * space address starting at @from into the buffer @to at offset @ppos.
621 *
622 * On success, the number of bytes written is returned and the offset @ppos is
623 * advanced by this number, or negative value is returned on error.
624 **/
625ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
626		const void __user *from, size_t count)
627{
628	loff_t pos = *ppos;
629	size_t res;
630
631	if (pos < 0)
632		return -EINVAL;
633	if (pos >= available || !count)
634		return 0;
635	if (count > available - pos)
636		count = available - pos;
637	res = copy_from_user(to + pos, from, count);
638	if (res == count)
639		return -EFAULT;
640	count -= res;
641	*ppos = pos + count;
642	return count;
643}
644EXPORT_SYMBOL(simple_write_to_buffer);
645
646/**
647 * memory_read_from_buffer - copy data from the buffer
648 * @to: the kernel space buffer to read to
649 * @count: the maximum number of bytes to read
650 * @ppos: the current position in the buffer
651 * @from: the buffer to read from
652 * @available: the size of the buffer
653 *
654 * The memory_read_from_buffer() function reads up to @count bytes from the
655 * buffer @from at offset @ppos into the kernel space address starting at @to.
656 *
657 * On success, the number of bytes read is returned and the offset @ppos is
658 * advanced by this number, or negative value is returned on error.
659 **/
660ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
661				const void *from, size_t available)
662{
663	loff_t pos = *ppos;
664
665	if (pos < 0)
666		return -EINVAL;
667	if (pos >= available)
668		return 0;
669	if (count > available - pos)
670		count = available - pos;
671	memcpy(to, from + pos, count);
672	*ppos = pos + count;
673
674	return count;
675}
676EXPORT_SYMBOL(memory_read_from_buffer);
677
678/*
679 * Transaction based IO.
680 * The file expects a single write which triggers the transaction, and then
681 * possibly a read which collects the result - which is stored in a
682 * file-local buffer.
683 */
684
685void simple_transaction_set(struct file *file, size_t n)
686{
687	struct simple_transaction_argresp *ar = file->private_data;
688
689	BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
690
691	/*
692	 * The barrier ensures that ar->size will really remain zero until
693	 * ar->data is ready for reading.
694	 */
695	smp_mb();
696	ar->size = n;
697}
698EXPORT_SYMBOL(simple_transaction_set);
699
700char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
701{
702	struct simple_transaction_argresp *ar;
703	static DEFINE_SPINLOCK(simple_transaction_lock);
704
705	if (size > SIMPLE_TRANSACTION_LIMIT - 1)
706		return ERR_PTR(-EFBIG);
707
708	ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
709	if (!ar)
710		return ERR_PTR(-ENOMEM);
711
712	spin_lock(&simple_transaction_lock);
713
714	/* only one write allowed per open */
715	if (file->private_data) {
716		spin_unlock(&simple_transaction_lock);
717		free_page((unsigned long)ar);
718		return ERR_PTR(-EBUSY);
719	}
720
721	file->private_data = ar;
722
723	spin_unlock(&simple_transaction_lock);
724
725	if (copy_from_user(ar->data, buf, size))
726		return ERR_PTR(-EFAULT);
727
728	return ar->data;
729}
730EXPORT_SYMBOL(simple_transaction_get);
731
732ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
733{
734	struct simple_transaction_argresp *ar = file->private_data;
735
736	if (!ar)
737		return 0;
738	return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
739}
740EXPORT_SYMBOL(simple_transaction_read);
741
742int simple_transaction_release(struct inode *inode, struct file *file)
743{
744	free_page((unsigned long)file->private_data);
745	return 0;
746}
747EXPORT_SYMBOL(simple_transaction_release);
748
749/* Simple attribute files */
750
751struct simple_attr {
752	int (*get)(void *, u64 *);
753	int (*set)(void *, u64);
754	char get_buf[24];	/* enough to store a u64 and "\n\0" */
755	char set_buf[24];
756	void *data;
757	const char *fmt;	/* format for read operation */
758	struct mutex mutex;	/* protects access to these buffers */
759};
760
761/* simple_attr_open is called by an actual attribute open file operation
762 * to set the attribute specific access operations. */
763int simple_attr_open(struct inode *inode, struct file *file,
764		     int (*get)(void *, u64 *), int (*set)(void *, u64),
765		     const char *fmt)
766{
767	struct simple_attr *attr;
768
769	attr = kmalloc(sizeof(*attr), GFP_KERNEL);
770	if (!attr)
771		return -ENOMEM;
772
773	attr->get = get;
774	attr->set = set;
775	attr->data = inode->i_private;
776	attr->fmt = fmt;
777	mutex_init(&attr->mutex);
778
779	file->private_data = attr;
780
781	return nonseekable_open(inode, file);
782}
783EXPORT_SYMBOL_GPL(simple_attr_open);
784
785int simple_attr_release(struct inode *inode, struct file *file)
786{
787	kfree(file->private_data);
788	return 0;
789}
790EXPORT_SYMBOL_GPL(simple_attr_release);	/* GPL-only?  This?  Really? */
791
792/* read from the buffer that is filled with the get function */
793ssize_t simple_attr_read(struct file *file, char __user *buf,
794			 size_t len, loff_t *ppos)
795{
796	struct simple_attr *attr;
797	size_t size;
798	ssize_t ret;
799
800	attr = file->private_data;
801
802	if (!attr->get)
803		return -EACCES;
804
805	ret = mutex_lock_interruptible(&attr->mutex);
806	if (ret)
807		return ret;
808
809	if (*ppos) {		/* continued read */
810		size = strlen(attr->get_buf);
811	} else {		/* first read */
812		u64 val;
813		ret = attr->get(attr->data, &val);
814		if (ret)
815			goto out;
816
817		size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
818				 attr->fmt, (unsigned long long)val);
819	}
820
821	ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
822out:
823	mutex_unlock(&attr->mutex);
824	return ret;
825}
826EXPORT_SYMBOL_GPL(simple_attr_read);
827
828/* interpret the buffer as a number to call the set function with */
829ssize_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_strtoll(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 */
855out:
856	mutex_unlock(&attr->mutex);
857	return ret;
858}
859EXPORT_SYMBOL_GPL(simple_attr_write);
860
861/**
862 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
863 * @sb:		filesystem to do the file handle conversion on
864 * @fid:	file handle to convert
865 * @fh_len:	length of the file handle in bytes
866 * @fh_type:	type of file handle
867 * @get_inode:	filesystem callback to retrieve inode
868 *
869 * This function decodes @fid as long as it has one of the well-known
870 * Linux filehandle types and calls @get_inode on it to retrieve the
871 * inode for the object specified in the file handle.
872 */
873struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
874		int fh_len, int fh_type, struct inode *(*get_inode)
875			(struct super_block *sb, u64 ino, u32 gen))
876{
877	struct inode *inode = NULL;
878
879	if (fh_len < 2)
880		return NULL;
881
882	switch (fh_type) {
883	case FILEID_INO32_GEN:
884	case FILEID_INO32_GEN_PARENT:
885		inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
886		break;
887	}
888
889	return d_obtain_alias(inode);
890}
891EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
892
893/**
894 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
895 * @sb:		filesystem to do the file handle conversion on
896 * @fid:	file handle to convert
897 * @fh_len:	length of the file handle in bytes
898 * @fh_type:	type of file handle
899 * @get_inode:	filesystem callback to retrieve inode
900 *
901 * This function decodes @fid as long as it has one of the well-known
902 * Linux filehandle types and calls @get_inode on it to retrieve the
903 * inode for the _parent_ object specified in the file handle if it
904 * is specified in the file handle, or NULL otherwise.
905 */
906struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
907		int fh_len, int fh_type, struct inode *(*get_inode)
908			(struct super_block *sb, u64 ino, u32 gen))
909{
910	struct inode *inode = NULL;
911
912	if (fh_len <= 2)
913		return NULL;
914
915	switch (fh_type) {
916	case FILEID_INO32_GEN_PARENT:
917		inode = get_inode(sb, fid->i32.parent_ino,
918				  (fh_len > 3 ? fid->i32.parent_gen : 0));
919		break;
920	}
921
922	return d_obtain_alias(inode);
923}
924EXPORT_SYMBOL_GPL(generic_fh_to_parent);
925
926/**
927 * __generic_file_fsync - generic fsync implementation for simple filesystems
928 *
929 * @file:	file to synchronize
930 * @start:	start offset in bytes
931 * @end:	end offset in bytes (inclusive)
932 * @datasync:	only synchronize essential metadata if true
933 *
934 * This is a generic implementation of the fsync method for simple
935 * filesystems which track all non-inode metadata in the buffers list
936 * hanging off the address_space structure.
937 */
938int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
939				 int datasync)
940{
941	struct inode *inode = file->f_mapping->host;
942	int err;
943	int ret;
944
945	err = filemap_write_and_wait_range(inode->i_mapping, start, end);
946	if (err)
947		return err;
948
949	mutex_lock(&inode->i_mutex);
950	ret = sync_mapping_buffers(inode->i_mapping);
951	if (!(inode->i_state & I_DIRTY_ALL))
952		goto out;
953	if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
954		goto out;
955
956	err = sync_inode_metadata(inode, 1);
957	if (ret == 0)
958		ret = err;
959
960out:
961	mutex_unlock(&inode->i_mutex);
962	return ret;
963}
964EXPORT_SYMBOL(__generic_file_fsync);
965
966/**
967 * generic_file_fsync - generic fsync implementation for simple filesystems
968 *			with flush
969 * @file:	file to synchronize
970 * @start:	start offset in bytes
971 * @end:	end offset in bytes (inclusive)
972 * @datasync:	only synchronize essential metadata if true
973 *
974 */
975
976int generic_file_fsync(struct file *file, loff_t start, loff_t end,
977		       int datasync)
978{
979	struct inode *inode = file->f_mapping->host;
980	int err;
981
982	err = __generic_file_fsync(file, start, end, datasync);
983	if (err)
984		return err;
985	return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
986}
987EXPORT_SYMBOL(generic_file_fsync);
988
989/**
990 * generic_check_addressable - Check addressability of file system
991 * @blocksize_bits:	log of file system block size
992 * @num_blocks:		number of blocks in file system
993 *
994 * Determine whether a file system with @num_blocks blocks (and a
995 * block size of 2**@blocksize_bits) is addressable by the sector_t
996 * and page cache of the system.  Return 0 if so and -EFBIG otherwise.
997 */
998int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
999{
1000	u64 last_fs_block = num_blocks - 1;
1001	u64 last_fs_page =
1002		last_fs_block >> (PAGE_CACHE_SHIFT - blocksize_bits);
1003
1004	if (unlikely(num_blocks == 0))
1005		return 0;
1006
1007	if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT))
1008		return -EINVAL;
1009
1010	if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1011	    (last_fs_page > (pgoff_t)(~0ULL))) {
1012		return -EFBIG;
1013	}
1014	return 0;
1015}
1016EXPORT_SYMBOL(generic_check_addressable);
1017
1018/*
1019 * No-op implementation of ->fsync for in-memory filesystems.
1020 */
1021int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1022{
1023	return 0;
1024}
1025EXPORT_SYMBOL(noop_fsync);
1026
1027void kfree_put_link(struct dentry *dentry, struct nameidata *nd,
1028				void *cookie)
1029{
1030	char *s = nd_get_link(nd);
1031	if (!IS_ERR(s))
1032		kfree(s);
1033}
1034EXPORT_SYMBOL(kfree_put_link);
1035
1036/*
1037 * nop .set_page_dirty method so that people can use .page_mkwrite on
1038 * anon inodes.
1039 */
1040static int anon_set_page_dirty(struct page *page)
1041{
1042	return 0;
1043};
1044
1045/*
1046 * A single inode exists for all anon_inode files. Contrary to pipes,
1047 * anon_inode inodes have no associated per-instance data, so we need
1048 * only allocate one of them.
1049 */
1050struct inode *alloc_anon_inode(struct super_block *s)
1051{
1052	static const struct address_space_operations anon_aops = {
1053		.set_page_dirty = anon_set_page_dirty,
1054	};
1055	struct inode *inode = new_inode_pseudo(s);
1056
1057	if (!inode)
1058		return ERR_PTR(-ENOMEM);
1059
1060	inode->i_ino = get_next_ino();
1061	inode->i_mapping->a_ops = &anon_aops;
1062
1063	/*
1064	 * Mark the inode dirty from the very beginning,
1065	 * that way it will never be moved to the dirty
1066	 * list because mark_inode_dirty() will think
1067	 * that it already _is_ on the dirty list.
1068	 */
1069	inode->i_state = I_DIRTY;
1070	inode->i_mode = S_IRUSR | S_IWUSR;
1071	inode->i_uid = current_fsuid();
1072	inode->i_gid = current_fsgid();
1073	inode->i_flags |= S_PRIVATE;
1074	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1075	return inode;
1076}
1077EXPORT_SYMBOL(alloc_anon_inode);
1078
1079/**
1080 * simple_nosetlease - generic helper for prohibiting leases
1081 * @filp: file pointer
1082 * @arg: type of lease to obtain
1083 * @flp: new lease supplied for insertion
1084 * @priv: private data for lm_setup operation
1085 *
1086 * Generic helper for filesystems that do not wish to allow leases to be set.
1087 * All arguments are ignored and it just returns -EINVAL.
1088 */
1089int
1090simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1091		  void **priv)
1092{
1093	return -EINVAL;
1094}
1095EXPORT_SYMBOL(simple_nosetlease);
1096
1097
1098/*
1099 * Operations for a permanently empty directory.
1100 */
1101static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1102{
1103	return ERR_PTR(-ENOENT);
1104}
1105
1106static int empty_dir_getattr(struct vfsmount *mnt, struct dentry *dentry,
1107				 struct kstat *stat)
1108{
1109	struct inode *inode = d_inode(dentry);
1110	generic_fillattr(inode, stat);
1111	return 0;
1112}
1113
1114static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1115{
1116	return -EPERM;
1117}
1118
1119static int empty_dir_setxattr(struct dentry *dentry, const char *name,
1120			      const void *value, size_t size, int flags)
1121{
1122	return -EOPNOTSUPP;
1123}
1124
1125static ssize_t empty_dir_getxattr(struct dentry *dentry, const char *name,
1126				  void *value, size_t size)
1127{
1128	return -EOPNOTSUPP;
1129}
1130
1131static int empty_dir_removexattr(struct dentry *dentry, const char *name)
1132{
1133	return -EOPNOTSUPP;
1134}
1135
1136static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1137{
1138	return -EOPNOTSUPP;
1139}
1140
1141static const struct inode_operations empty_dir_inode_operations = {
1142	.lookup		= empty_dir_lookup,
1143	.permission	= generic_permission,
1144	.setattr	= empty_dir_setattr,
1145	.getattr	= empty_dir_getattr,
1146	.setxattr	= empty_dir_setxattr,
1147	.getxattr	= empty_dir_getxattr,
1148	.removexattr	= empty_dir_removexattr,
1149	.listxattr	= empty_dir_listxattr,
1150};
1151
1152static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1153{
1154	/* An empty directory has two entries . and .. at offsets 0 and 1 */
1155	return generic_file_llseek_size(file, offset, whence, 2, 2);
1156}
1157
1158static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1159{
1160	dir_emit_dots(file, ctx);
1161	return 0;
1162}
1163
1164static const struct file_operations empty_dir_operations = {
1165	.llseek		= empty_dir_llseek,
1166	.read		= generic_read_dir,
1167	.iterate	= empty_dir_readdir,
1168	.fsync		= noop_fsync,
1169};
1170
1171
1172void make_empty_dir_inode(struct inode *inode)
1173{
1174	set_nlink(inode, 2);
1175	inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1176	inode->i_uid = GLOBAL_ROOT_UID;
1177	inode->i_gid = GLOBAL_ROOT_GID;
1178	inode->i_rdev = 0;
1179	inode->i_size = 0;
1180	inode->i_blkbits = PAGE_SHIFT;
1181	inode->i_blocks = 0;
1182
1183	inode->i_op = &empty_dir_inode_operations;
1184	inode->i_fop = &empty_dir_operations;
1185}
1186
1187bool is_empty_dir_inode(struct inode *inode)
1188{
1189	return (inode->i_fop == &empty_dir_operations) &&
1190		(inode->i_op == &empty_dir_inode_operations);
1191}
1192