xref: /openbmc/linux/fs/libfs.c (revision c1d45424)
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 	if (!dentry->d_sb->s_d_op)
65 		d_set_d_op(dentry, &simple_dentry_operations);
66 	d_add(dentry, NULL);
67 	return NULL;
68 }
69 
70 int dcache_dir_open(struct inode *inode, struct file *file)
71 {
72 	static struct qstr cursor_name = QSTR_INIT(".", 1);
73 
74 	file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
75 
76 	return file->private_data ? 0 : -ENOMEM;
77 }
78 
79 int dcache_dir_close(struct inode *inode, struct file *file)
80 {
81 	dput(file->private_data);
82 	return 0;
83 }
84 
85 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
86 {
87 	struct dentry *dentry = file->f_path.dentry;
88 	mutex_lock(&dentry->d_inode->i_mutex);
89 	switch (whence) {
90 		case 1:
91 			offset += file->f_pos;
92 		case 0:
93 			if (offset >= 0)
94 				break;
95 		default:
96 			mutex_unlock(&dentry->d_inode->i_mutex);
97 			return -EINVAL;
98 	}
99 	if (offset != file->f_pos) {
100 		file->f_pos = offset;
101 		if (file->f_pos >= 2) {
102 			struct list_head *p;
103 			struct dentry *cursor = file->private_data;
104 			loff_t n = file->f_pos - 2;
105 
106 			spin_lock(&dentry->d_lock);
107 			/* d_lock not required for cursor */
108 			list_del(&cursor->d_u.d_child);
109 			p = dentry->d_subdirs.next;
110 			while (n && p != &dentry->d_subdirs) {
111 				struct dentry *next;
112 				next = list_entry(p, struct dentry, d_u.d_child);
113 				spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
114 				if (simple_positive(next))
115 					n--;
116 				spin_unlock(&next->d_lock);
117 				p = p->next;
118 			}
119 			list_add_tail(&cursor->d_u.d_child, p);
120 			spin_unlock(&dentry->d_lock);
121 		}
122 	}
123 	mutex_unlock(&dentry->d_inode->i_mutex);
124 	return offset;
125 }
126 
127 /* Relationship between i_mode and the DT_xxx types */
128 static inline unsigned char dt_type(struct inode *inode)
129 {
130 	return (inode->i_mode >> 12) & 15;
131 }
132 
133 /*
134  * Directory is locked and all positive dentries in it are safe, since
135  * for ramfs-type trees they can't go away without unlink() or rmdir(),
136  * both impossible due to the lock on directory.
137  */
138 
139 int dcache_readdir(struct file *file, struct dir_context *ctx)
140 {
141 	struct dentry *dentry = file->f_path.dentry;
142 	struct dentry *cursor = file->private_data;
143 	struct list_head *p, *q = &cursor->d_u.d_child;
144 
145 	if (!dir_emit_dots(file, ctx))
146 		return 0;
147 	spin_lock(&dentry->d_lock);
148 	if (ctx->pos == 2)
149 		list_move(q, &dentry->d_subdirs);
150 
151 	for (p = q->next; p != &dentry->d_subdirs; p = p->next) {
152 		struct dentry *next = list_entry(p, struct dentry, d_u.d_child);
153 		spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
154 		if (!simple_positive(next)) {
155 			spin_unlock(&next->d_lock);
156 			continue;
157 		}
158 
159 		spin_unlock(&next->d_lock);
160 		spin_unlock(&dentry->d_lock);
161 		if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
162 			      next->d_inode->i_ino, dt_type(next->d_inode)))
163 			return 0;
164 		spin_lock(&dentry->d_lock);
165 		spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
166 		/* next is still alive */
167 		list_move(q, p);
168 		spin_unlock(&next->d_lock);
169 		p = q;
170 		ctx->pos++;
171 	}
172 	spin_unlock(&dentry->d_lock);
173 	return 0;
174 }
175 
176 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
177 {
178 	return -EISDIR;
179 }
180 
181 const struct file_operations simple_dir_operations = {
182 	.open		= dcache_dir_open,
183 	.release	= dcache_dir_close,
184 	.llseek		= dcache_dir_lseek,
185 	.read		= generic_read_dir,
186 	.iterate	= dcache_readdir,
187 	.fsync		= noop_fsync,
188 };
189 
190 const struct inode_operations simple_dir_inode_operations = {
191 	.lookup		= simple_lookup,
192 };
193 
194 static const struct super_operations simple_super_operations = {
195 	.statfs		= simple_statfs,
196 };
197 
198 /*
199  * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
200  * will never be mountable)
201  */
202 struct dentry *mount_pseudo(struct file_system_type *fs_type, char *name,
203 	const struct super_operations *ops,
204 	const struct dentry_operations *dops, unsigned long magic)
205 {
206 	struct super_block *s;
207 	struct dentry *dentry;
208 	struct inode *root;
209 	struct qstr d_name = QSTR_INIT(name, strlen(name));
210 
211 	s = sget(fs_type, NULL, set_anon_super, MS_NOUSER, NULL);
212 	if (IS_ERR(s))
213 		return ERR_CAST(s);
214 
215 	s->s_maxbytes = MAX_LFS_FILESIZE;
216 	s->s_blocksize = PAGE_SIZE;
217 	s->s_blocksize_bits = PAGE_SHIFT;
218 	s->s_magic = magic;
219 	s->s_op = ops ? ops : &simple_super_operations;
220 	s->s_time_gran = 1;
221 	root = new_inode(s);
222 	if (!root)
223 		goto Enomem;
224 	/*
225 	 * since this is the first inode, make it number 1. New inodes created
226 	 * after this must take care not to collide with it (by passing
227 	 * max_reserved of 1 to iunique).
228 	 */
229 	root->i_ino = 1;
230 	root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
231 	root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
232 	dentry = __d_alloc(s, &d_name);
233 	if (!dentry) {
234 		iput(root);
235 		goto Enomem;
236 	}
237 	d_instantiate(dentry, root);
238 	s->s_root = dentry;
239 	s->s_d_op = dops;
240 	s->s_flags |= MS_ACTIVE;
241 	return dget(s->s_root);
242 
243 Enomem:
244 	deactivate_locked_super(s);
245 	return ERR_PTR(-ENOMEM);
246 }
247 
248 int simple_open(struct inode *inode, struct file *file)
249 {
250 	if (inode->i_private)
251 		file->private_data = inode->i_private;
252 	return 0;
253 }
254 
255 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
256 {
257 	struct inode *inode = old_dentry->d_inode;
258 
259 	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
260 	inc_nlink(inode);
261 	ihold(inode);
262 	dget(dentry);
263 	d_instantiate(dentry, inode);
264 	return 0;
265 }
266 
267 int simple_empty(struct dentry *dentry)
268 {
269 	struct dentry *child;
270 	int ret = 0;
271 
272 	spin_lock(&dentry->d_lock);
273 	list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child) {
274 		spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
275 		if (simple_positive(child)) {
276 			spin_unlock(&child->d_lock);
277 			goto out;
278 		}
279 		spin_unlock(&child->d_lock);
280 	}
281 	ret = 1;
282 out:
283 	spin_unlock(&dentry->d_lock);
284 	return ret;
285 }
286 
287 int simple_unlink(struct inode *dir, struct dentry *dentry)
288 {
289 	struct inode *inode = dentry->d_inode;
290 
291 	inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
292 	drop_nlink(inode);
293 	dput(dentry);
294 	return 0;
295 }
296 
297 int simple_rmdir(struct inode *dir, struct dentry *dentry)
298 {
299 	if (!simple_empty(dentry))
300 		return -ENOTEMPTY;
301 
302 	drop_nlink(dentry->d_inode);
303 	simple_unlink(dir, dentry);
304 	drop_nlink(dir);
305 	return 0;
306 }
307 
308 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
309 		struct inode *new_dir, struct dentry *new_dentry)
310 {
311 	struct inode *inode = old_dentry->d_inode;
312 	int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
313 
314 	if (!simple_empty(new_dentry))
315 		return -ENOTEMPTY;
316 
317 	if (new_dentry->d_inode) {
318 		simple_unlink(new_dir, new_dentry);
319 		if (they_are_dirs) {
320 			drop_nlink(new_dentry->d_inode);
321 			drop_nlink(old_dir);
322 		}
323 	} else if (they_are_dirs) {
324 		drop_nlink(old_dir);
325 		inc_nlink(new_dir);
326 	}
327 
328 	old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
329 		new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
330 
331 	return 0;
332 }
333 
334 /**
335  * simple_setattr - setattr for simple filesystem
336  * @dentry: dentry
337  * @iattr: iattr structure
338  *
339  * Returns 0 on success, -error on failure.
340  *
341  * simple_setattr is a simple ->setattr implementation without a proper
342  * implementation of size changes.
343  *
344  * It can either be used for in-memory filesystems or special files
345  * on simple regular filesystems.  Anything that needs to change on-disk
346  * or wire state on size changes needs its own setattr method.
347  */
348 int simple_setattr(struct dentry *dentry, struct iattr *iattr)
349 {
350 	struct inode *inode = dentry->d_inode;
351 	int error;
352 
353 	error = inode_change_ok(inode, iattr);
354 	if (error)
355 		return error;
356 
357 	if (iattr->ia_valid & ATTR_SIZE)
358 		truncate_setsize(inode, iattr->ia_size);
359 	setattr_copy(inode, iattr);
360 	mark_inode_dirty(inode);
361 	return 0;
362 }
363 EXPORT_SYMBOL(simple_setattr);
364 
365 int simple_readpage(struct file *file, struct page *page)
366 {
367 	clear_highpage(page);
368 	flush_dcache_page(page);
369 	SetPageUptodate(page);
370 	unlock_page(page);
371 	return 0;
372 }
373 
374 int simple_write_begin(struct file *file, struct address_space *mapping,
375 			loff_t pos, unsigned len, unsigned flags,
376 			struct page **pagep, void **fsdata)
377 {
378 	struct page *page;
379 	pgoff_t index;
380 
381 	index = pos >> PAGE_CACHE_SHIFT;
382 
383 	page = grab_cache_page_write_begin(mapping, index, flags);
384 	if (!page)
385 		return -ENOMEM;
386 
387 	*pagep = page;
388 
389 	if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
390 		unsigned from = pos & (PAGE_CACHE_SIZE - 1);
391 
392 		zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
393 	}
394 	return 0;
395 }
396 
397 /**
398  * simple_write_end - .write_end helper for non-block-device FSes
399  * @available: See .write_end of address_space_operations
400  * @file: 		"
401  * @mapping: 		"
402  * @pos: 		"
403  * @len: 		"
404  * @copied: 		"
405  * @page: 		"
406  * @fsdata: 		"
407  *
408  * simple_write_end does the minimum needed for updating a page after writing is
409  * done. It has the same API signature as the .write_end of
410  * address_space_operations vector. So it can just be set onto .write_end for
411  * FSes that don't need any other processing. i_mutex is assumed to be held.
412  * Block based filesystems should use generic_write_end().
413  * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
414  * is not called, so a filesystem that actually does store data in .write_inode
415  * should extend on what's done here with a call to mark_inode_dirty() in the
416  * case that i_size has changed.
417  */
418 int simple_write_end(struct file *file, struct address_space *mapping,
419 			loff_t pos, unsigned len, unsigned copied,
420 			struct page *page, void *fsdata)
421 {
422 	struct inode *inode = page->mapping->host;
423 	loff_t last_pos = pos + copied;
424 
425 	/* zero the stale part of the page if we did a short copy */
426 	if (copied < len) {
427 		unsigned from = pos & (PAGE_CACHE_SIZE - 1);
428 
429 		zero_user(page, from + copied, len - copied);
430 	}
431 
432 	if (!PageUptodate(page))
433 		SetPageUptodate(page);
434 	/*
435 	 * No need to use i_size_read() here, the i_size
436 	 * cannot change under us because we hold the i_mutex.
437 	 */
438 	if (last_pos > inode->i_size)
439 		i_size_write(inode, last_pos);
440 
441 	set_page_dirty(page);
442 	unlock_page(page);
443 	page_cache_release(page);
444 
445 	return copied;
446 }
447 
448 /*
449  * the inodes created here are not hashed. If you use iunique to generate
450  * unique inode values later for this filesystem, then you must take care
451  * to pass it an appropriate max_reserved value to avoid collisions.
452  */
453 int simple_fill_super(struct super_block *s, unsigned long magic,
454 		      struct tree_descr *files)
455 {
456 	struct inode *inode;
457 	struct dentry *root;
458 	struct dentry *dentry;
459 	int i;
460 
461 	s->s_blocksize = PAGE_CACHE_SIZE;
462 	s->s_blocksize_bits = PAGE_CACHE_SHIFT;
463 	s->s_magic = magic;
464 	s->s_op = &simple_super_operations;
465 	s->s_time_gran = 1;
466 
467 	inode = new_inode(s);
468 	if (!inode)
469 		return -ENOMEM;
470 	/*
471 	 * because the root inode is 1, the files array must not contain an
472 	 * entry at index 1
473 	 */
474 	inode->i_ino = 1;
475 	inode->i_mode = S_IFDIR | 0755;
476 	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
477 	inode->i_op = &simple_dir_inode_operations;
478 	inode->i_fop = &simple_dir_operations;
479 	set_nlink(inode, 2);
480 	root = d_make_root(inode);
481 	if (!root)
482 		return -ENOMEM;
483 	for (i = 0; !files->name || files->name[0]; i++, files++) {
484 		if (!files->name)
485 			continue;
486 
487 		/* warn if it tries to conflict with the root inode */
488 		if (unlikely(i == 1))
489 			printk(KERN_WARNING "%s: %s passed in a files array"
490 				"with an index of 1!\n", __func__,
491 				s->s_type->name);
492 
493 		dentry = d_alloc_name(root, files->name);
494 		if (!dentry)
495 			goto out;
496 		inode = new_inode(s);
497 		if (!inode) {
498 			dput(dentry);
499 			goto out;
500 		}
501 		inode->i_mode = S_IFREG | files->mode;
502 		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
503 		inode->i_fop = files->ops;
504 		inode->i_ino = i;
505 		d_add(dentry, inode);
506 	}
507 	s->s_root = root;
508 	return 0;
509 out:
510 	d_genocide(root);
511 	shrink_dcache_parent(root);
512 	dput(root);
513 	return -ENOMEM;
514 }
515 
516 static DEFINE_SPINLOCK(pin_fs_lock);
517 
518 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
519 {
520 	struct vfsmount *mnt = NULL;
521 	spin_lock(&pin_fs_lock);
522 	if (unlikely(!*mount)) {
523 		spin_unlock(&pin_fs_lock);
524 		mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, NULL);
525 		if (IS_ERR(mnt))
526 			return PTR_ERR(mnt);
527 		spin_lock(&pin_fs_lock);
528 		if (!*mount)
529 			*mount = mnt;
530 	}
531 	mntget(*mount);
532 	++*count;
533 	spin_unlock(&pin_fs_lock);
534 	mntput(mnt);
535 	return 0;
536 }
537 
538 void simple_release_fs(struct vfsmount **mount, int *count)
539 {
540 	struct vfsmount *mnt;
541 	spin_lock(&pin_fs_lock);
542 	mnt = *mount;
543 	if (!--*count)
544 		*mount = NULL;
545 	spin_unlock(&pin_fs_lock);
546 	mntput(mnt);
547 }
548 
549 /**
550  * simple_read_from_buffer - copy data from the buffer to user space
551  * @to: the user space buffer to read to
552  * @count: the maximum number of bytes to read
553  * @ppos: the current position in the buffer
554  * @from: the buffer to read from
555  * @available: the size of the buffer
556  *
557  * The simple_read_from_buffer() function reads up to @count bytes from the
558  * buffer @from at offset @ppos into the user space address starting at @to.
559  *
560  * On success, the number of bytes read is returned and the offset @ppos is
561  * advanced by this number, or negative value is returned on error.
562  **/
563 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
564 				const void *from, size_t available)
565 {
566 	loff_t pos = *ppos;
567 	size_t ret;
568 
569 	if (pos < 0)
570 		return -EINVAL;
571 	if (pos >= available || !count)
572 		return 0;
573 	if (count > available - pos)
574 		count = available - pos;
575 	ret = copy_to_user(to, from + pos, count);
576 	if (ret == count)
577 		return -EFAULT;
578 	count -= ret;
579 	*ppos = pos + count;
580 	return count;
581 }
582 
583 /**
584  * simple_write_to_buffer - copy data from user space to the buffer
585  * @to: the buffer to write to
586  * @available: the size of the buffer
587  * @ppos: the current position in the buffer
588  * @from: the user space buffer to read from
589  * @count: the maximum number of bytes to read
590  *
591  * The simple_write_to_buffer() function reads up to @count bytes from the user
592  * space address starting at @from into the buffer @to at offset @ppos.
593  *
594  * On success, the number of bytes written is returned and the offset @ppos is
595  * advanced by this number, or negative value is returned on error.
596  **/
597 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
598 		const void __user *from, size_t count)
599 {
600 	loff_t pos = *ppos;
601 	size_t res;
602 
603 	if (pos < 0)
604 		return -EINVAL;
605 	if (pos >= available || !count)
606 		return 0;
607 	if (count > available - pos)
608 		count = available - pos;
609 	res = copy_from_user(to + pos, from, count);
610 	if (res == count)
611 		return -EFAULT;
612 	count -= res;
613 	*ppos = pos + count;
614 	return count;
615 }
616 
617 /**
618  * memory_read_from_buffer - copy data from the buffer
619  * @to: the kernel space buffer to read to
620  * @count: the maximum number of bytes to read
621  * @ppos: the current position in the buffer
622  * @from: the buffer to read from
623  * @available: the size of the buffer
624  *
625  * The memory_read_from_buffer() function reads up to @count bytes from the
626  * buffer @from at offset @ppos into the kernel space address starting at @to.
627  *
628  * On success, the number of bytes read is returned and the offset @ppos is
629  * advanced by this number, or negative value is returned on error.
630  **/
631 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
632 				const void *from, size_t available)
633 {
634 	loff_t pos = *ppos;
635 
636 	if (pos < 0)
637 		return -EINVAL;
638 	if (pos >= available)
639 		return 0;
640 	if (count > available - pos)
641 		count = available - pos;
642 	memcpy(to, from + pos, count);
643 	*ppos = pos + count;
644 
645 	return count;
646 }
647 
648 /*
649  * Transaction based IO.
650  * The file expects a single write which triggers the transaction, and then
651  * possibly a read which collects the result - which is stored in a
652  * file-local buffer.
653  */
654 
655 void simple_transaction_set(struct file *file, size_t n)
656 {
657 	struct simple_transaction_argresp *ar = file->private_data;
658 
659 	BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
660 
661 	/*
662 	 * The barrier ensures that ar->size will really remain zero until
663 	 * ar->data is ready for reading.
664 	 */
665 	smp_mb();
666 	ar->size = n;
667 }
668 
669 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
670 {
671 	struct simple_transaction_argresp *ar;
672 	static DEFINE_SPINLOCK(simple_transaction_lock);
673 
674 	if (size > SIMPLE_TRANSACTION_LIMIT - 1)
675 		return ERR_PTR(-EFBIG);
676 
677 	ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
678 	if (!ar)
679 		return ERR_PTR(-ENOMEM);
680 
681 	spin_lock(&simple_transaction_lock);
682 
683 	/* only one write allowed per open */
684 	if (file->private_data) {
685 		spin_unlock(&simple_transaction_lock);
686 		free_page((unsigned long)ar);
687 		return ERR_PTR(-EBUSY);
688 	}
689 
690 	file->private_data = ar;
691 
692 	spin_unlock(&simple_transaction_lock);
693 
694 	if (copy_from_user(ar->data, buf, size))
695 		return ERR_PTR(-EFAULT);
696 
697 	return ar->data;
698 }
699 
700 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
701 {
702 	struct simple_transaction_argresp *ar = file->private_data;
703 
704 	if (!ar)
705 		return 0;
706 	return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
707 }
708 
709 int simple_transaction_release(struct inode *inode, struct file *file)
710 {
711 	free_page((unsigned long)file->private_data);
712 	return 0;
713 }
714 
715 /* Simple attribute files */
716 
717 struct simple_attr {
718 	int (*get)(void *, u64 *);
719 	int (*set)(void *, u64);
720 	char get_buf[24];	/* enough to store a u64 and "\n\0" */
721 	char set_buf[24];
722 	void *data;
723 	const char *fmt;	/* format for read operation */
724 	struct mutex mutex;	/* protects access to these buffers */
725 };
726 
727 /* simple_attr_open is called by an actual attribute open file operation
728  * to set the attribute specific access operations. */
729 int simple_attr_open(struct inode *inode, struct file *file,
730 		     int (*get)(void *, u64 *), int (*set)(void *, u64),
731 		     const char *fmt)
732 {
733 	struct simple_attr *attr;
734 
735 	attr = kmalloc(sizeof(*attr), GFP_KERNEL);
736 	if (!attr)
737 		return -ENOMEM;
738 
739 	attr->get = get;
740 	attr->set = set;
741 	attr->data = inode->i_private;
742 	attr->fmt = fmt;
743 	mutex_init(&attr->mutex);
744 
745 	file->private_data = attr;
746 
747 	return nonseekable_open(inode, file);
748 }
749 
750 int simple_attr_release(struct inode *inode, struct file *file)
751 {
752 	kfree(file->private_data);
753 	return 0;
754 }
755 
756 /* read from the buffer that is filled with the get function */
757 ssize_t simple_attr_read(struct file *file, char __user *buf,
758 			 size_t len, loff_t *ppos)
759 {
760 	struct simple_attr *attr;
761 	size_t size;
762 	ssize_t ret;
763 
764 	attr = file->private_data;
765 
766 	if (!attr->get)
767 		return -EACCES;
768 
769 	ret = mutex_lock_interruptible(&attr->mutex);
770 	if (ret)
771 		return ret;
772 
773 	if (*ppos) {		/* continued read */
774 		size = strlen(attr->get_buf);
775 	} else {		/* first read */
776 		u64 val;
777 		ret = attr->get(attr->data, &val);
778 		if (ret)
779 			goto out;
780 
781 		size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
782 				 attr->fmt, (unsigned long long)val);
783 	}
784 
785 	ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
786 out:
787 	mutex_unlock(&attr->mutex);
788 	return ret;
789 }
790 
791 /* interpret the buffer as a number to call the set function with */
792 ssize_t simple_attr_write(struct file *file, const char __user *buf,
793 			  size_t len, loff_t *ppos)
794 {
795 	struct simple_attr *attr;
796 	u64 val;
797 	size_t size;
798 	ssize_t ret;
799 
800 	attr = file->private_data;
801 	if (!attr->set)
802 		return -EACCES;
803 
804 	ret = mutex_lock_interruptible(&attr->mutex);
805 	if (ret)
806 		return ret;
807 
808 	ret = -EFAULT;
809 	size = min(sizeof(attr->set_buf) - 1, len);
810 	if (copy_from_user(attr->set_buf, buf, size))
811 		goto out;
812 
813 	attr->set_buf[size] = '\0';
814 	val = simple_strtoll(attr->set_buf, NULL, 0);
815 	ret = attr->set(attr->data, val);
816 	if (ret == 0)
817 		ret = len; /* on success, claim we got the whole input */
818 out:
819 	mutex_unlock(&attr->mutex);
820 	return ret;
821 }
822 
823 /**
824  * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
825  * @sb:		filesystem to do the file handle conversion on
826  * @fid:	file handle to convert
827  * @fh_len:	length of the file handle in bytes
828  * @fh_type:	type of file handle
829  * @get_inode:	filesystem callback to retrieve inode
830  *
831  * This function decodes @fid as long as it has one of the well-known
832  * Linux filehandle types and calls @get_inode on it to retrieve the
833  * inode for the object specified in the file handle.
834  */
835 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
836 		int fh_len, int fh_type, struct inode *(*get_inode)
837 			(struct super_block *sb, u64 ino, u32 gen))
838 {
839 	struct inode *inode = NULL;
840 
841 	if (fh_len < 2)
842 		return NULL;
843 
844 	switch (fh_type) {
845 	case FILEID_INO32_GEN:
846 	case FILEID_INO32_GEN_PARENT:
847 		inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
848 		break;
849 	}
850 
851 	return d_obtain_alias(inode);
852 }
853 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
854 
855 /**
856  * generic_fh_to_parent - generic helper for the fh_to_parent export operation
857  * @sb:		filesystem to do the file handle conversion on
858  * @fid:	file handle to convert
859  * @fh_len:	length of the file handle in bytes
860  * @fh_type:	type of file handle
861  * @get_inode:	filesystem callback to retrieve inode
862  *
863  * This function decodes @fid as long as it has one of the well-known
864  * Linux filehandle types and calls @get_inode on it to retrieve the
865  * inode for the _parent_ object specified in the file handle if it
866  * is specified in the file handle, or NULL otherwise.
867  */
868 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
869 		int fh_len, int fh_type, struct inode *(*get_inode)
870 			(struct super_block *sb, u64 ino, u32 gen))
871 {
872 	struct inode *inode = NULL;
873 
874 	if (fh_len <= 2)
875 		return NULL;
876 
877 	switch (fh_type) {
878 	case FILEID_INO32_GEN_PARENT:
879 		inode = get_inode(sb, fid->i32.parent_ino,
880 				  (fh_len > 3 ? fid->i32.parent_gen : 0));
881 		break;
882 	}
883 
884 	return d_obtain_alias(inode);
885 }
886 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
887 
888 /**
889  * generic_file_fsync - generic fsync implementation for simple filesystems
890  * @file:	file to synchronize
891  * @datasync:	only synchronize essential metadata if true
892  *
893  * This is a generic implementation of the fsync method for simple
894  * filesystems which track all non-inode metadata in the buffers list
895  * hanging off the address_space structure.
896  */
897 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
898 		       int datasync)
899 {
900 	struct inode *inode = file->f_mapping->host;
901 	int err;
902 	int ret;
903 
904 	err = filemap_write_and_wait_range(inode->i_mapping, start, end);
905 	if (err)
906 		return err;
907 
908 	mutex_lock(&inode->i_mutex);
909 	ret = sync_mapping_buffers(inode->i_mapping);
910 	if (!(inode->i_state & I_DIRTY))
911 		goto out;
912 	if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
913 		goto out;
914 
915 	err = sync_inode_metadata(inode, 1);
916 	if (ret == 0)
917 		ret = err;
918 out:
919 	mutex_unlock(&inode->i_mutex);
920 	return ret;
921 }
922 EXPORT_SYMBOL(generic_file_fsync);
923 
924 /**
925  * generic_check_addressable - Check addressability of file system
926  * @blocksize_bits:	log of file system block size
927  * @num_blocks:		number of blocks in file system
928  *
929  * Determine whether a file system with @num_blocks blocks (and a
930  * block size of 2**@blocksize_bits) is addressable by the sector_t
931  * and page cache of the system.  Return 0 if so and -EFBIG otherwise.
932  */
933 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
934 {
935 	u64 last_fs_block = num_blocks - 1;
936 	u64 last_fs_page =
937 		last_fs_block >> (PAGE_CACHE_SHIFT - blocksize_bits);
938 
939 	if (unlikely(num_blocks == 0))
940 		return 0;
941 
942 	if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT))
943 		return -EINVAL;
944 
945 	if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
946 	    (last_fs_page > (pgoff_t)(~0ULL))) {
947 		return -EFBIG;
948 	}
949 	return 0;
950 }
951 EXPORT_SYMBOL(generic_check_addressable);
952 
953 /*
954  * No-op implementation of ->fsync for in-memory filesystems.
955  */
956 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
957 {
958 	return 0;
959 }
960 
961 EXPORT_SYMBOL(dcache_dir_close);
962 EXPORT_SYMBOL(dcache_dir_lseek);
963 EXPORT_SYMBOL(dcache_dir_open);
964 EXPORT_SYMBOL(dcache_readdir);
965 EXPORT_SYMBOL(generic_read_dir);
966 EXPORT_SYMBOL(mount_pseudo);
967 EXPORT_SYMBOL(simple_write_begin);
968 EXPORT_SYMBOL(simple_write_end);
969 EXPORT_SYMBOL(simple_dir_inode_operations);
970 EXPORT_SYMBOL(simple_dir_operations);
971 EXPORT_SYMBOL(simple_empty);
972 EXPORT_SYMBOL(simple_fill_super);
973 EXPORT_SYMBOL(simple_getattr);
974 EXPORT_SYMBOL(simple_open);
975 EXPORT_SYMBOL(simple_link);
976 EXPORT_SYMBOL(simple_lookup);
977 EXPORT_SYMBOL(simple_pin_fs);
978 EXPORT_SYMBOL(simple_readpage);
979 EXPORT_SYMBOL(simple_release_fs);
980 EXPORT_SYMBOL(simple_rename);
981 EXPORT_SYMBOL(simple_rmdir);
982 EXPORT_SYMBOL(simple_statfs);
983 EXPORT_SYMBOL(noop_fsync);
984 EXPORT_SYMBOL(simple_unlink);
985 EXPORT_SYMBOL(simple_read_from_buffer);
986 EXPORT_SYMBOL(simple_write_to_buffer);
987 EXPORT_SYMBOL(memory_read_from_buffer);
988 EXPORT_SYMBOL(simple_transaction_set);
989 EXPORT_SYMBOL(simple_transaction_get);
990 EXPORT_SYMBOL(simple_transaction_read);
991 EXPORT_SYMBOL(simple_transaction_release);
992 EXPORT_SYMBOL_GPL(simple_attr_open);
993 EXPORT_SYMBOL_GPL(simple_attr_release);
994 EXPORT_SYMBOL_GPL(simple_attr_read);
995 EXPORT_SYMBOL_GPL(simple_attr_write);
996