path: root/fs/libfs.c (plain)
blob: 9588780ad43e175c795580275f6ddc62be4f670a

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
23	int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
24	struct kstat *stat)
25	{
26	struct inode *inode = d_inode(dentry);
27	generic_fillattr(inode, stat);
28	stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
29	return 0;
30	}
31	EXPORT_SYMBOL(simple_getattr);
32
33	int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
34	{
35	buf->f_type = dentry->d_sb->s_magic;
36	buf->f_bsize = PAGE_SIZE;
37	buf->f_namelen = NAME_MAX;
38	return 0;
39	}
40	EXPORT_SYMBOL(simple_statfs);
41
42	/*
43	* Retaining negative dentries for an in-memory filesystem just wastes
44	* memory and lookup time: arrange for them to be deleted immediately.
45	*/
46	int always_delete_dentry(const struct dentry *dentry)
47	{
48	return 1;
49	}
50	EXPORT_SYMBOL(always_delete_dentry);
51
52	const struct dentry_operations simple_dentry_operations = {
53	.d_delete = always_delete_dentry,
54	};
55	EXPORT_SYMBOL(simple_dentry_operations);
56
57	/*
58	* Lookup the data. This is trivial - if the dentry didn't already
59	* exist, we know it is negative. Set d_op to delete negative dentries.
60	*/
61	struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
62	{
63	if (dentry->d_name.len > NAME_MAX)
64	return ERR_PTR(-ENAMETOOLONG);
65	if (!dentry->d_sb->s_d_op)
66	d_set_d_op(dentry, &simple_dentry_operations);
67	d_add(dentry, NULL);
68	return NULL;
69	}
70	EXPORT_SYMBOL(simple_lookup);
71
72	int dcache_dir_open(struct inode *inode, struct file *file)
73	{
74	file->private_data = d_alloc_cursor(file->f_path.dentry);
75
76	return file->private_data ? 0 : -ENOMEM;
77	}
78	EXPORT_SYMBOL(dcache_dir_open);
79
80	int dcache_dir_close(struct inode *inode, struct file *file)
81	{
82	dput(file->private_data);
83	return 0;
84	}
85	EXPORT_SYMBOL(dcache_dir_close);
86
87	/* parent is locked at least shared */
88	static struct dentry *next_positive(struct dentry *parent,
89	struct list_head *from,
90	int count)
91	{
92	unsigned *seq = &parent->d_inode->i_dir_seq, n;
93	struct dentry *res;
94	struct list_head *p;
95	bool skipped;
96	int i;
97
98	retry:
99	i = count;
100	skipped = false;
101	n = smp_load_acquire(seq) & ~1;
102	res = NULL;
103	rcu_read_lock();
104	for (p = from->next; p != &parent->d_subdirs; p = p->next) {
105	struct dentry *d = list_entry(p, struct dentry, d_child);
106	if (!simple_positive(d)) {
107	skipped = true;
108	} else if (!--i) {
109	res = d;
110	break;
111	}
112	}
113	rcu_read_unlock();
114	if (skipped) {
115	smp_rmb();
116	if (unlikely(*seq != n))
117	goto retry;
118	}
119	return res;
120	}
121
122	static void move_cursor(struct dentry *cursor, struct list_head *after)
123	{
124	struct dentry *parent = cursor->d_parent;
125	unsigned n, *seq = &parent->d_inode->i_dir_seq;
126	spin_lock(&parent->d_lock);
127	for (;;) {
128	n = *seq;
129	if (!(n & 1) && cmpxchg(seq, n, n + 1) == n)
130	break;
131	cpu_relax();
132	}
133	__list_del(cursor->d_child.prev, cursor->d_child.next);
134	if (after)
135	list_add(&cursor->d_child, after);
136	else
137	list_add_tail(&cursor->d_child, &parent->d_subdirs);
138	smp_store_release(seq, n + 2);
139	spin_unlock(&parent->d_lock);
140	}
141
142	loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
143	{
144	struct dentry *dentry = file->f_path.dentry;
145	switch (whence) {
146	case 1:
147	offset += file->f_pos;
148	case 0:
149	if (offset >= 0)
150	break;
151	default:
152	return -EINVAL;
153	}
154	if (offset != file->f_pos) {
155	file->f_pos = offset;
156	if (file->f_pos >= 2) {
157	struct dentry *cursor = file->private_data;
158	struct dentry *to;
159	loff_t n = file->f_pos - 2;
160
161	inode_lock_shared(dentry->d_inode);
162	to = next_positive(dentry, &dentry->d_subdirs, n);
163	move_cursor(cursor, to ? &to->d_child : NULL);
164	inode_unlock_shared(dentry->d_inode);
165	}
166	}
167	return offset;
168	}
169	EXPORT_SYMBOL(dcache_dir_lseek);
170
171	/* Relationship between i_mode and the DT_xxx types */
172	static inline unsigned char dt_type(struct inode *inode)
173	{
174	return (inode->i_mode >> 12) & 15;
175	}
176
177	/*
178	* Directory is locked and all positive dentries in it are safe, since
179	* for ramfs-type trees they can't go away without unlink() or rmdir(),
180	* both impossible due to the lock on directory.
181	*/
182
183	int dcache_readdir(struct file *file, struct dir_context *ctx)
184	{
185	struct dentry *dentry = file->f_path.dentry;
186	struct dentry *cursor = file->private_data;
187	struct list_head *p = &cursor->d_child;
188	struct dentry *next;
189	bool moved = false;
190
191	if (!dir_emit_dots(file, ctx))
192	return 0;
193
194	if (ctx->pos == 2)
195	p = &dentry->d_subdirs;
196	while ((next = next_positive(dentry, p, 1)) != NULL) {
197	if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
198	d_inode(next)->i_ino, dt_type(d_inode(next))))
199	break;
200	moved = true;
201	p = &next->d_child;
202	ctx->pos++;
203	}
204	if (moved)
205	move_cursor(cursor, p);
206	return 0;
207	}
208	EXPORT_SYMBOL(dcache_readdir);
209
210	ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
211	{
212	return -EISDIR;
213	}
214	EXPORT_SYMBOL(generic_read_dir);
215
216	const struct file_operations simple_dir_operations = {
217	.open = dcache_dir_open,
218	.release = dcache_dir_close,
219	.llseek = dcache_dir_lseek,
220	.read = generic_read_dir,
221	.iterate_shared = dcache_readdir,
222	.fsync = noop_fsync,
223	};
224	EXPORT_SYMBOL(simple_dir_operations);
225
226	const struct inode_operations simple_dir_inode_operations = {
227	.lookup = simple_lookup,
228	};
229	EXPORT_SYMBOL(simple_dir_inode_operations);
230
231	static const struct super_operations simple_super_operations = {
232	.statfs = simple_statfs,
233	};
234
235	/*
236	* Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
237	* will never be mountable)
238	*/
239	struct dentry *mount_pseudo_xattr(struct file_system_type *fs_type, char *name,
240	const struct super_operations *ops, const struct xattr_handler **xattr,
241	const struct dentry_operations *dops, unsigned long magic)
242	{
243	struct super_block *s;
244	struct dentry *dentry;
245	struct inode *root;
246	struct qstr d_name = QSTR_INIT(name, strlen(name));
247
248	s = sget_userns(fs_type, NULL, set_anon_super, MS_KERNMOUNT|MS_NOUSER,
249	&init_user_ns, NULL);
250	if (IS_ERR(s))
251	return ERR_CAST(s);
252
253	s->s_maxbytes = MAX_LFS_FILESIZE;
254	s->s_blocksize = PAGE_SIZE;
255	s->s_blocksize_bits = PAGE_SHIFT;
256	s->s_magic = magic;
257	s->s_op = ops ? ops : &simple_super_operations;
258	s->s_xattr = xattr;
259	s->s_time_gran = 1;
260	root = new_inode(s);
261	if (!root)
262	goto Enomem;
263	/*
264	* since this is the first inode, make it number 1. New inodes created
265	* after this must take care not to collide with it (by passing
266	* max_reserved of 1 to iunique).
267	*/
268	root->i_ino = 1;
269	root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
270	root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
271	dentry = __d_alloc(s, &d_name);
272	if (!dentry) {
273	iput(root);
274	goto Enomem;
275	}
276	d_instantiate(dentry, root);
277	s->s_root = dentry;
278	s->s_d_op = dops;
279	s->s_flags |= MS_ACTIVE;
280	return dget(s->s_root);
281
282	Enomem:
283	deactivate_locked_super(s);
284	return ERR_PTR(-ENOMEM);
285	}
286	EXPORT_SYMBOL(mount_pseudo_xattr);
287
288	int simple_open(struct inode *inode, struct file *file)
289	{
290	if (inode->i_private)
291	file->private_data = inode->i_private;
292	return 0;
293	}
294	EXPORT_SYMBOL(simple_open);
295
296	int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
297	{
298	struct inode *inode = d_inode(old_dentry);
299
300	inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
301	inc_nlink(inode);
302	ihold(inode);
303	dget(dentry);
304	d_instantiate(dentry, inode);
305	return 0;
306	}
307	EXPORT_SYMBOL(simple_link);
308
309	int simple_empty(struct dentry *dentry)
310	{
311	struct dentry *child;
312	int ret = 0;
313
314	spin_lock(&dentry->d_lock);
315	list_for_each_entry(child, &dentry->d_subdirs, d_child) {
316	spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
317	if (simple_positive(child)) {
318	spin_unlock(&child->d_lock);
319	goto out;
320	}
321	spin_unlock(&child->d_lock);
322	}
323	ret = 1;
324	out:
325	spin_unlock(&dentry->d_lock);
326	return ret;
327	}
328	EXPORT_SYMBOL(simple_empty);
329
330	int simple_unlink(struct inode *dir, struct dentry *dentry)
331	{
332	struct inode *inode = d_inode(dentry);
333
334	inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
335	drop_nlink(inode);
336	dput(dentry);
337	return 0;
338	}
339	EXPORT_SYMBOL(simple_unlink);
340
341	int simple_rmdir(struct inode *dir, struct dentry *dentry)
342	{
343	if (!simple_empty(dentry))
344	return -ENOTEMPTY;
345
346	drop_nlink(d_inode(dentry));
347	simple_unlink(dir, dentry);
348	drop_nlink(dir);
349	return 0;
350	}
351	EXPORT_SYMBOL(simple_rmdir);
352
353	int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
354	struct inode *new_dir, struct dentry *new_dentry,
355	unsigned int flags)
356	{
357	struct inode *inode = d_inode(old_dentry);
358	int they_are_dirs = d_is_dir(old_dentry);
359
360	if (flags & ~RENAME_NOREPLACE)
361	return -EINVAL;
362
363	if (!simple_empty(new_dentry))
364	return -ENOTEMPTY;
365
366	if (d_really_is_positive(new_dentry)) {
367	simple_unlink(new_dir, new_dentry);
368	if (they_are_dirs) {
369	drop_nlink(d_inode(new_dentry));
370	drop_nlink(old_dir);
371	}
372	} else if (they_are_dirs) {
373	drop_nlink(old_dir);
374	inc_nlink(new_dir);
375	}
376
377	old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
378	new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
379
380	return 0;
381	}
382	EXPORT_SYMBOL(simple_rename);
383
384	/**
385	* simple_setattr - setattr for simple filesystem
386	* @dentry: dentry
387	* @iattr: iattr structure
388	*
389	* Returns 0 on success, -error on failure.
390	*
391	* simple_setattr is a simple ->setattr implementation without a proper
392	* implementation of size changes.
393	*
394	* It can either be used for in-memory filesystems or special files
395	* on simple regular filesystems. Anything that needs to change on-disk
396	* or wire state on size changes needs its own setattr method.
397	*/
398	int simple_setattr(struct dentry *dentry, struct iattr *iattr)
399	{
400	struct inode *inode = d_inode(dentry);
401	int error;
402
403	error = setattr_prepare(dentry, iattr);
404	if (error)
405	return error;
406
407	if (iattr->ia_valid & ATTR_SIZE)
408	truncate_setsize(inode, iattr->ia_size);
409	setattr_copy(inode, iattr);
410	mark_inode_dirty(inode);
411	return 0;
412	}
413	EXPORT_SYMBOL(simple_setattr);
414
415	int simple_readpage(struct file *file, struct page *page)
416	{
417	clear_highpage(page);
418	flush_dcache_page(page);
419	SetPageUptodate(page);
420	unlock_page(page);
421	return 0;
422	}
423	EXPORT_SYMBOL(simple_readpage);
424
425	int simple_write_begin(struct file *file, struct address_space *mapping,
426	loff_t pos, unsigned len, unsigned flags,
427	struct page **pagep, void **fsdata)
428	{
429	struct page *page;
430	pgoff_t index;
431
432	index = pos >> PAGE_SHIFT;
433
434	page = grab_cache_page_write_begin(mapping, index, flags);
435	if (!page)
436	return -ENOMEM;
437
438	*pagep = page;
439
440	if (!PageUptodate(page) && (len != PAGE_SIZE)) {
441	unsigned from = pos & (PAGE_SIZE - 1);
442
443	zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
444	}
445	return 0;
446	}
447	EXPORT_SYMBOL(simple_write_begin);
448
449	/**
450	* simple_write_end - .write_end helper for non-block-device FSes
451	* @available: See .write_end of address_space_operations
452	* @file: "
453	* @mapping: "
454	* @pos: "
455	* @len: "
456	* @copied: "
457	* @page: "
458	* @fsdata: "
459	*
460	* simple_write_end does the minimum needed for updating a page after writing is
461	* done. It has the same API signature as the .write_end of
462	* address_space_operations vector. So it can just be set onto .write_end for
463	* FSes that don't need any other processing. i_mutex is assumed to be held.
464	* Block based filesystems should use generic_write_end().
465	* NOTE: Even though i_size might get updated by this function, mark_inode_dirty
466	* is not called, so a filesystem that actually does store data in .write_inode
467	* should extend on what's done here with a call to mark_inode_dirty() in the
468	* case that i_size has changed.
469	*/
470	int simple_write_end(struct file *file, struct address_space *mapping,
471	loff_t pos, unsigned len, unsigned copied,
472	struct page *page, void *fsdata)
473	{
474	struct inode *inode = page->mapping->host;
475	loff_t last_pos = pos + copied;
476
477	/* zero the stale part of the page if we did a short copy */
478	if (copied < len) {
479	unsigned from = pos & (PAGE_SIZE - 1);
480
481	zero_user(page, from + copied, len - copied);
482	}
483
484	if (!PageUptodate(page))
485	SetPageUptodate(page);
486	/*
487	* No need to use i_size_read() here, the i_size
488	* cannot change under us because we hold the i_mutex.
489	*/
490	if (last_pos > inode->i_size)
491	i_size_write(inode, last_pos);
492
493	set_page_dirty(page);
494	unlock_page(page);
495	put_page(page);
496
497	return copied;
498	}
499	EXPORT_SYMBOL(simple_write_end);
500
501	/*
502	* the inodes created here are not hashed. If you use iunique to generate
503	* unique inode values later for this filesystem, then you must take care
504	* to pass it an appropriate max_reserved value to avoid collisions.
505	*/
506	int simple_fill_super(struct super_block *s, unsigned long magic,
507	struct tree_descr *files)
508	{
509	struct inode *inode;
510	struct dentry *root;
511	struct dentry *dentry;
512	int i;
513
514	s->s_blocksize = PAGE_SIZE;
515	s->s_blocksize_bits = PAGE_SHIFT;
516	s->s_magic = magic;
517	s->s_op = &simple_super_operations;
518	s->s_time_gran = 1;
519
520	inode = new_inode(s);
521	if (!inode)
522	return -ENOMEM;
523	/*
524	* because the root inode is 1, the files array must not contain an
525	* entry at index 1
526	*/
527	inode->i_ino = 1;
528	inode->i_mode = S_IFDIR | 0755;
529	inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
530	inode->i_op = &simple_dir_inode_operations;
531	inode->i_fop = &simple_dir_operations;
532	set_nlink(inode, 2);
533	root = d_make_root(inode);
534	if (!root)
535	return -ENOMEM;
536	for (i = 0; !files->name || files->name[0]; i++, files++) {
537	if (!files->name)
538	continue;
539
540	/* warn if it tries to conflict with the root inode */
541	if (unlikely(i == 1))
542	printk(KERN_WARNING "%s: %s passed in a files array"
543	"with an index of 1!\n", __func__,
544	s->s_type->name);
545
546	dentry = d_alloc_name(root, files->name);
547	if (!dentry)
548	goto out;
549	inode = new_inode(s);
550	if (!inode) {
551	dput(dentry);
552	goto out;
553	}
554	inode->i_mode = S_IFREG | files->mode;
555	inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
556	inode->i_fop = files->ops;
557	inode->i_ino = i;
558	d_add(dentry, inode);
559	}
560	s->s_root = root;
561	return 0;
562	out:
563	d_genocide(root);
564	shrink_dcache_parent(root);
565	dput(root);
566	return -ENOMEM;
567	}
568	EXPORT_SYMBOL(simple_fill_super);
569
570	static DEFINE_SPINLOCK(pin_fs_lock);
571
572	int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
573	{
574	struct vfsmount *mnt = NULL;
575	spin_lock(&pin_fs_lock);
576	if (unlikely(!*mount)) {
577	spin_unlock(&pin_fs_lock);
578	mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, NULL);
579	if (IS_ERR(mnt))
580	return PTR_ERR(mnt);
581	spin_lock(&pin_fs_lock);
582	if (!*mount)
583	*mount = mnt;
584	}
585	mntget(*mount);
586	++*count;
587	spin_unlock(&pin_fs_lock);
588	mntput(mnt);
589	return 0;
590	}
591	EXPORT_SYMBOL(simple_pin_fs);
592
593	void simple_release_fs(struct vfsmount **mount, int *count)
594	{
595	struct vfsmount *mnt;
596	spin_lock(&pin_fs_lock);
597	mnt = *mount;
598	if (!--*count)
599	*mount = NULL;
600	spin_unlock(&pin_fs_lock);
601	mntput(mnt);
602	}
603	EXPORT_SYMBOL(simple_release_fs);
604
605	/**
606	* simple_read_from_buffer - copy data from the buffer to user space
607	* @to: the user space buffer to read to
608	* @count: the maximum number of bytes to read
609	* @ppos: the current position in the buffer
610	* @from: the buffer to read from
611	* @available: the size of the buffer
612	*
613	* The simple_read_from_buffer() function reads up to @count bytes from the
614	* buffer @from at offset @ppos into the user space address starting at @to.
615	*
616	* On success, the number of bytes read is returned and the offset @ppos is
617	* advanced by this number, or negative value is returned on error.
618	**/
619	ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
620	const void *from, size_t available)
621	{
622	loff_t pos = *ppos;
623	size_t ret;
624
625	if (pos < 0)
626	return -EINVAL;
627	if (pos >= available || !count)
628	return 0;
629	if (count > available - pos)
630	count = available - pos;
631	ret = copy_to_user(to, from + pos, count);
632	if (ret == count)
633	return -EFAULT;
634	count -= ret;
635	*ppos = pos + count;
636	return count;
637	}
638	EXPORT_SYMBOL(simple_read_from_buffer);
639
640	/**
641	* simple_write_to_buffer - copy data from user space to the buffer
642	* @to: the buffer to write to
643	* @available: the size of the buffer
644	* @ppos: the current position in the buffer
645	* @from: the user space buffer to read from
646	* @count: the maximum number of bytes to read
647	*
648	* The simple_write_to_buffer() function reads up to @count bytes from the user
649	* space address starting at @from into the buffer @to at offset @ppos.
650	*
651	* On success, the number of bytes written is returned and the offset @ppos is
652	* advanced by this number, or negative value is returned on error.
653	**/
654	ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
655	const void __user *from, size_t count)
656	{
657	loff_t pos = *ppos;
658	size_t res;
659
660	if (pos < 0)
661	return -EINVAL;
662	if (pos >= available || !count)
663	return 0;
664	if (count > available - pos)
665	count = available - pos;
666	res = copy_from_user(to + pos, from, count);
667	if (res == count)
668	return -EFAULT;
669	count -= res;
670	*ppos = pos + count;
671	return count;
672	}
673	EXPORT_SYMBOL(simple_write_to_buffer);
674
675	/**
676	* memory_read_from_buffer - copy data from the buffer
677	* @to: the kernel space buffer to read to
678	* @count: the maximum number of bytes to read
679	* @ppos: the current position in the buffer
680	* @from: the buffer to read from
681	* @available: the size of the buffer
682	*
683	* The memory_read_from_buffer() function reads up to @count bytes from the
684	* buffer @from at offset @ppos into the kernel space address starting at @to.
685	*
686	* On success, the number of bytes read is returned and the offset @ppos is
687	* advanced by this number, or negative value is returned on error.
688	**/
689	ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
690	const void *from, size_t available)
691	{
692	loff_t pos = *ppos;
693
694	if (pos < 0)
695	return -EINVAL;
696	if (pos >= available)
697	return 0;
698	if (count > available - pos)
699	count = available - pos;
700	memcpy(to, from + pos, count);
701	*ppos = pos + count;
702
703	return count;
704	}
705	EXPORT_SYMBOL(memory_read_from_buffer);
706
707	/*
708	* Transaction based IO.
709	* The file expects a single write which triggers the transaction, and then
710	* possibly a read which collects the result - which is stored in a
711	* file-local buffer.
712	*/
713
714	void simple_transaction_set(struct file *file, size_t n)
715	{
716	struct simple_transaction_argresp *ar = file->private_data;
717
718	BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
719
720	/*
721	* The barrier ensures that ar->size will really remain zero until
722	* ar->data is ready for reading.
723	*/
724	smp_mb();
725	ar->size = n;
726	}
727	EXPORT_SYMBOL(simple_transaction_set);
728
729	char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
730	{
731	struct simple_transaction_argresp *ar;
732	static DEFINE_SPINLOCK(simple_transaction_lock);
733
734	if (size > SIMPLE_TRANSACTION_LIMIT - 1)
735	return ERR_PTR(-EFBIG);
736
737	ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
738	if (!ar)
739	return ERR_PTR(-ENOMEM);
740
741	spin_lock(&simple_transaction_lock);
742
743	/* only one write allowed per open */
744	if (file->private_data) {
745	spin_unlock(&simple_transaction_lock);
746	free_page((unsigned long)ar);
747	return ERR_PTR(-EBUSY);
748	}
749
750	file->private_data = ar;
751
752	spin_unlock(&simple_transaction_lock);
753
754	if (copy_from_user(ar->data, buf, size))
755	return ERR_PTR(-EFAULT);
756
757	return ar->data;
758	}
759	EXPORT_SYMBOL(simple_transaction_get);
760
761	ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
762	{
763	struct simple_transaction_argresp *ar = file->private_data;
764
765	if (!ar)
766	return 0;
767	return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
768	}
769	EXPORT_SYMBOL(simple_transaction_read);
770
771	int simple_transaction_release(struct inode *inode, struct file *file)
772	{
773	free_page((unsigned long)file->private_data);
774	return 0;
775	}
776	EXPORT_SYMBOL(simple_transaction_release);
777
778	/* Simple attribute files */
779
780	struct simple_attr {
781	int (*get)(void *, u64 *);
782	int (*set)(void *, u64);
783	char get_buf[24]; /* enough to store a u64 and "\n\0" */
784	char set_buf[24];
785	void *data;
786	const char *fmt; /* format for read operation */
787	struct mutex mutex; /* protects access to these buffers */
788	};
789
790	/* simple_attr_open is called by an actual attribute open file operation
791	* to set the attribute specific access operations. */
792	int simple_attr_open(struct inode *inode, struct file *file,
793	int (*get)(void *, u64 *), int (*set)(void *, u64),
794	const char *fmt)
795	{
796	struct simple_attr *attr;
797
798	attr = kmalloc(sizeof(*attr), GFP_KERNEL);
799	if (!attr)
800	return -ENOMEM;
801
802	attr->get = get;
803	attr->set = set;
804	attr->data = inode->i_private;
805	attr->fmt = fmt;
806	mutex_init(&attr->mutex);
807
808	file->private_data = attr;
809
810	return nonseekable_open(inode, file);
811	}
812	EXPORT_SYMBOL_GPL(simple_attr_open);
813
814	int simple_attr_release(struct inode *inode, struct file *file)
815	{
816	kfree(file->private_data);
817	return 0;
818	}
819	EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
820
821	/* read from the buffer that is filled with the get function */
822	ssize_t simple_attr_read(struct file *file, char __user *buf,
823	size_t len, loff_t *ppos)
824	{
825	struct simple_attr *attr;
826	size_t size;
827	ssize_t ret;
828
829	attr = file->private_data;
830
831	if (!attr->get)
832	return -EACCES;
833
834	ret = mutex_lock_interruptible(&attr->mutex);
835	if (ret)
836	return ret;
837
838	if (*ppos) { /* continued read */
839	size = strlen(attr->get_buf);
840	} else { /* first read */
841	u64 val;
842	ret = attr->get(attr->data, &val);
843	if (ret)
844	goto out;
845
846	size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
847	attr->fmt, (unsigned long long)val);
848	}
849
850	ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
851	out:
852	mutex_unlock(&attr->mutex);
853	return ret;
854	}
855	EXPORT_SYMBOL_GPL(simple_attr_read);
856
857	/* interpret the buffer as a number to call the set function with */
858	ssize_t simple_attr_write(struct file *file, const char __user *buf,
859	size_t len, loff_t *ppos)
860	{
861	struct simple_attr *attr;
862	u64 val;
863	size_t size;
864	ssize_t ret;
865
866	attr = file->private_data;
867	if (!attr->set)
868	return -EACCES;
869
870	ret = mutex_lock_interruptible(&attr->mutex);
871	if (ret)
872	return ret;
873
874	ret = -EFAULT;
875	size = min(sizeof(attr->set_buf) - 1, len);
876	if (copy_from_user(attr->set_buf, buf, size))
877	goto out;
878
879	attr->set_buf[size] = '\0';
880	val = simple_strtoll(attr->set_buf, NULL, 0);
881	ret = attr->set(attr->data, val);
882	if (ret == 0)
883	ret = len; /* on success, claim we got the whole input */
884	out:
885	mutex_unlock(&attr->mutex);
886	return ret;
887	}
888	EXPORT_SYMBOL_GPL(simple_attr_write);
889
890	/**
891	* generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
892	* @sb: filesystem to do the file handle conversion on
893	* @fid: file handle to convert
894	* @fh_len: length of the file handle in bytes
895	* @fh_type: type of file handle
896	* @get_inode: filesystem callback to retrieve inode
897	*
898	* This function decodes @fid as long as it has one of the well-known
899	* Linux filehandle types and calls @get_inode on it to retrieve the
900	* inode for the object specified in the file handle.
901	*/
902	struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
903	int fh_len, int fh_type, struct inode *(*get_inode)
904	(struct super_block *sb, u64 ino, u32 gen))
905	{
906	struct inode *inode = NULL;
907
908	if (fh_len < 2)
909	return NULL;
910
911	switch (fh_type) {
912	case FILEID_INO32_GEN:
913	case FILEID_INO32_GEN_PARENT:
914	inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
915	break;
916	}
917
918	return d_obtain_alias(inode);
919	}
920	EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
921
922	/**
923	* generic_fh_to_parent - generic helper for the fh_to_parent export operation
924	* @sb: filesystem to do the file handle conversion on
925	* @fid: file handle to convert
926	* @fh_len: length of the file handle in bytes
927	* @fh_type: type of file handle
928	* @get_inode: filesystem callback to retrieve inode
929	*
930	* This function decodes @fid as long as it has one of the well-known
931	* Linux filehandle types and calls @get_inode on it to retrieve the
932	* inode for the _parent_ object specified in the file handle if it
933	* is specified in the file handle, or NULL otherwise.
934	*/
935	struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
936	int fh_len, int fh_type, struct inode *(*get_inode)
937	(struct super_block *sb, u64 ino, u32 gen))
938	{
939	struct inode *inode = NULL;
940
941	if (fh_len <= 2)
942	return NULL;
943
944	switch (fh_type) {
945	case FILEID_INO32_GEN_PARENT:
946	inode = get_inode(sb, fid->i32.parent_ino,
947	(fh_len > 3 ? fid->i32.parent_gen : 0));
948	break;
949	}
950
951	return d_obtain_alias(inode);
952	}
953	EXPORT_SYMBOL_GPL(generic_fh_to_parent);
954
955	/**
956	* __generic_file_fsync - generic fsync implementation for simple filesystems
957	*
958	* @file: file to synchronize
959	* @start: start offset in bytes
960	* @end: end offset in bytes (inclusive)
961	* @datasync: only synchronize essential metadata if true
962	*
963	* This is a generic implementation of the fsync method for simple
964	* filesystems which track all non-inode metadata in the buffers list
965	* hanging off the address_space structure.
966	*/
967	int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
968	int datasync)
969	{
970	struct inode *inode = file->f_mapping->host;
971	int err;
972	int ret;
973
974	err = filemap_write_and_wait_range(inode->i_mapping, start, end);
975	if (err)
976	return err;
977
978	inode_lock(inode);
979	ret = sync_mapping_buffers(inode->i_mapping);
980	if (!(inode->i_state & I_DIRTY_ALL))
981	goto out;
982	if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
983	goto out;
984
985	err = sync_inode_metadata(inode, 1);
986	if (ret == 0)
987	ret = err;
988
989	out:
990	inode_unlock(inode);
991	return ret;
992	}
993	EXPORT_SYMBOL(__generic_file_fsync);
994
995	/**
996	* generic_file_fsync - generic fsync implementation for simple filesystems
997	* with flush
998	* @file: file to synchronize
999	* @start: start offset in bytes
1000	* @end: end offset in bytes (inclusive)
1001	* @datasync: only synchronize essential metadata if true
1002	*
1003	*/
1004
1005	int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1006	int datasync)
1007	{
1008	struct inode *inode = file->f_mapping->host;
1009	int err;
1010
1011	err = __generic_file_fsync(file, start, end, datasync);
1012	if (err)
1013	return err;
1014	return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
1015	}
1016	EXPORT_SYMBOL(generic_file_fsync);
1017
1018	/**
1019	* generic_check_addressable - Check addressability of file system
1020	* @blocksize_bits: log of file system block size
1021	* @num_blocks: number of blocks in file system
1022	*
1023	* Determine whether a file system with @num_blocks blocks (and a
1024	* block size of 2**@blocksize_bits) is addressable by the sector_t
1025	* and page cache of the system. Return 0 if so and -EFBIG otherwise.
1026	*/
1027	int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1028	{
1029	u64 last_fs_block = num_blocks - 1;
1030	u64 last_fs_page =
1031	last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1032
1033	if (unlikely(num_blocks == 0))
1034	return 0;
1035
1036	if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1037	return -EINVAL;
1038
1039	if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1040	(last_fs_page > (pgoff_t)(~0ULL))) {
1041	return -EFBIG;
1042	}
1043	return 0;
1044	}
1045	EXPORT_SYMBOL(generic_check_addressable);
1046
1047	/*
1048	* No-op implementation of ->fsync for in-memory filesystems.
1049	*/
1050	int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1051	{
1052	return 0;
1053	}
1054	EXPORT_SYMBOL(noop_fsync);
1055
1056	/* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1057	void kfree_link(void *p)
1058	{
1059	kfree(p);
1060	}
1061	EXPORT_SYMBOL(kfree_link);
1062
1063	/*
1064	* nop .set_page_dirty method so that people can use .page_mkwrite on
1065	* anon inodes.
1066	*/
1067	static int anon_set_page_dirty(struct page *page)
1068	{
1069	return 0;
1070	};
1071
1072	/*
1073	* A single inode exists for all anon_inode files. Contrary to pipes,
1074	* anon_inode inodes have no associated per-instance data, so we need
1075	* only allocate one of them.
1076	*/
1077	struct inode *alloc_anon_inode(struct super_block *s)
1078	{
1079	static const struct address_space_operations anon_aops = {
1080	.set_page_dirty = anon_set_page_dirty,
1081	};
1082	struct inode *inode = new_inode_pseudo(s);
1083
1084	if (!inode)
1085	return ERR_PTR(-ENOMEM);
1086
1087	inode->i_ino = get_next_ino();
1088	inode->i_mapping->a_ops = &anon_aops;
1089
1090	/*
1091	* Mark the inode dirty from the very beginning,
1092	* that way it will never be moved to the dirty
1093	* list because mark_inode_dirty() will think
1094	* that it already _is_ on the dirty list.
1095	*/
1096	inode->i_state = I_DIRTY;
1097	inode->i_mode = S_IRUSR | S_IWUSR;
1098	inode->i_uid = current_fsuid();
1099	inode->i_gid = current_fsgid();
1100	inode->i_flags |= S_PRIVATE;
1101	inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1102	return inode;
1103	}
1104	EXPORT_SYMBOL(alloc_anon_inode);
1105
1106	/**
1107	* simple_nosetlease - generic helper for prohibiting leases
1108	* @filp: file pointer
1109	* @arg: type of lease to obtain
1110	* @flp: new lease supplied for insertion
1111	* @priv: private data for lm_setup operation
1112	*
1113	* Generic helper for filesystems that do not wish to allow leases to be set.
1114	* All arguments are ignored and it just returns -EINVAL.
1115	*/
1116	int
1117	simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1118	void **priv)
1119	{
1120	return -EINVAL;
1121	}
1122	EXPORT_SYMBOL(simple_nosetlease);
1123
1124	const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1125	struct delayed_call *done)
1126	{
1127	return inode->i_link;
1128	}
1129	EXPORT_SYMBOL(simple_get_link);
1130
1131	const struct inode_operations simple_symlink_inode_operations = {
1132	.get_link = simple_get_link,
1133	.readlink = generic_readlink
1134	};
1135	EXPORT_SYMBOL(simple_symlink_inode_operations);
1136
1137	/*
1138	* Operations for a permanently empty directory.
1139	*/
1140	static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1141	{
1142	return ERR_PTR(-ENOENT);
1143	}
1144
1145	static int empty_dir_getattr(struct vfsmount *mnt, struct dentry *dentry,
1146	struct kstat *stat)
1147	{
1148	struct inode *inode = d_inode(dentry);
1149	generic_fillattr(inode, stat);
1150	return 0;
1151	}
1152
1153	static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1154	{
1155	return -EPERM;
1156	}
1157
1158	static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1159	{
1160	return -EOPNOTSUPP;
1161	}
1162
1163	static const struct inode_operations empty_dir_inode_operations = {
1164	.lookup = empty_dir_lookup,
1165	.permission = generic_permission,
1166	.setattr = empty_dir_setattr,
1167	.getattr = empty_dir_getattr,
1168	.listxattr = empty_dir_listxattr,
1169	};
1170
1171	static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1172	{
1173	/* An empty directory has two entries . and .. at offsets 0 and 1 */
1174	return generic_file_llseek_size(file, offset, whence, 2, 2);
1175	}
1176
1177	static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1178	{
1179	dir_emit_dots(file, ctx);
1180	return 0;
1181	}
1182
1183	static const struct file_operations empty_dir_operations = {
1184	.llseek = empty_dir_llseek,
1185	.read = generic_read_dir,
1186	.iterate_shared = empty_dir_readdir,
1187	.fsync = noop_fsync,
1188	};
1189
1190
1191	void make_empty_dir_inode(struct inode *inode)
1192	{
1193	set_nlink(inode, 2);
1194	inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1195	inode->i_uid = GLOBAL_ROOT_UID;
1196	inode->i_gid = GLOBAL_ROOT_GID;
1197	inode->i_rdev = 0;
1198	inode->i_size = 0;
1199	inode->i_blkbits = PAGE_SHIFT;
1200	inode->i_blocks = 0;
1201
1202	inode->i_op = &empty_dir_inode_operations;
1203	inode->i_opflags &= ~IOP_XATTR;
1204	inode->i_fop = &empty_dir_operations;
1205	}
1206
1207	bool is_empty_dir_inode(struct inode *inode)
1208	{
1209	return (inode->i_fop == &empty_dir_operations) &&
1210	(inode->i_op == &empty_dir_inode_operations);
1211	}
1212