xref: /openbmc/linux/fs/libfs.c (revision 4da722ca)
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/cred.h>
11 #include <linux/mount.h>
12 #include <linux/vfs.h>
13 #include <linux/quotaops.h>
14 #include <linux/mutex.h>
15 #include <linux/namei.h>
16 #include <linux/exportfs.h>
17 #include <linux/writeback.h>
18 #include <linux/buffer_head.h> /* sync_mapping_buffers */
19 
20 #include <linux/uaccess.h>
21 
22 #include "internal.h"
23 
24 int simple_getattr(const struct path *path, struct kstat *stat,
25 		   u32 request_mask, unsigned int query_flags)
26 {
27 	struct inode *inode = d_inode(path->dentry);
28 	generic_fillattr(inode, stat);
29 	stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
30 	return 0;
31 }
32 EXPORT_SYMBOL(simple_getattr);
33 
34 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
35 {
36 	buf->f_type = dentry->d_sb->s_magic;
37 	buf->f_bsize = PAGE_SIZE;
38 	buf->f_namelen = NAME_MAX;
39 	return 0;
40 }
41 EXPORT_SYMBOL(simple_statfs);
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 int always_delete_dentry(const struct dentry *dentry)
48 {
49 	return 1;
50 }
51 EXPORT_SYMBOL(always_delete_dentry);
52 
53 const struct dentry_operations simple_dentry_operations = {
54 	.d_delete = always_delete_dentry,
55 };
56 EXPORT_SYMBOL(simple_dentry_operations);
57 
58 /*
59  * Lookup the data. This is trivial - if the dentry didn't already
60  * exist, we know it is negative.  Set d_op to delete negative dentries.
61  */
62 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
63 {
64 	if (dentry->d_name.len > NAME_MAX)
65 		return ERR_PTR(-ENAMETOOLONG);
66 	if (!dentry->d_sb->s_d_op)
67 		d_set_d_op(dentry, &simple_dentry_operations);
68 	d_add(dentry, NULL);
69 	return NULL;
70 }
71 EXPORT_SYMBOL(simple_lookup);
72 
73 int dcache_dir_open(struct inode *inode, struct file *file)
74 {
75 	file->private_data = d_alloc_cursor(file->f_path.dentry);
76 
77 	return file->private_data ? 0 : -ENOMEM;
78 }
79 EXPORT_SYMBOL(dcache_dir_open);
80 
81 int dcache_dir_close(struct inode *inode, struct file *file)
82 {
83 	dput(file->private_data);
84 	return 0;
85 }
86 EXPORT_SYMBOL(dcache_dir_close);
87 
88 /* parent is locked at least shared */
89 static struct dentry *next_positive(struct dentry *parent,
90 				    struct list_head *from,
91 				    int count)
92 {
93 	unsigned *seq = &parent->d_inode->i_dir_seq, n;
94 	struct dentry *res;
95 	struct list_head *p;
96 	bool skipped;
97 	int i;
98 
99 retry:
100 	i = count;
101 	skipped = false;
102 	n = smp_load_acquire(seq) & ~1;
103 	res = NULL;
104 	rcu_read_lock();
105 	for (p = from->next; p != &parent->d_subdirs; p = p->next) {
106 		struct dentry *d = list_entry(p, struct dentry, d_child);
107 		if (!simple_positive(d)) {
108 			skipped = true;
109 		} else if (!--i) {
110 			res = d;
111 			break;
112 		}
113 	}
114 	rcu_read_unlock();
115 	if (skipped) {
116 		smp_rmb();
117 		if (unlikely(*seq != n))
118 			goto retry;
119 	}
120 	return res;
121 }
122 
123 static void move_cursor(struct dentry *cursor, struct list_head *after)
124 {
125 	struct dentry *parent = cursor->d_parent;
126 	unsigned n, *seq = &parent->d_inode->i_dir_seq;
127 	spin_lock(&parent->d_lock);
128 	for (;;) {
129 		n = *seq;
130 		if (!(n & 1) && cmpxchg(seq, n, n + 1) == n)
131 			break;
132 		cpu_relax();
133 	}
134 	__list_del(cursor->d_child.prev, cursor->d_child.next);
135 	if (after)
136 		list_add(&cursor->d_child, after);
137 	else
138 		list_add_tail(&cursor->d_child, &parent->d_subdirs);
139 	smp_store_release(seq, n + 2);
140 	spin_unlock(&parent->d_lock);
141 }
142 
143 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
144 {
145 	struct dentry *dentry = file->f_path.dentry;
146 	switch (whence) {
147 		case 1:
148 			offset += file->f_pos;
149 		case 0:
150 			if (offset >= 0)
151 				break;
152 		default:
153 			return -EINVAL;
154 	}
155 	if (offset != file->f_pos) {
156 		file->f_pos = offset;
157 		if (file->f_pos >= 2) {
158 			struct dentry *cursor = file->private_data;
159 			struct dentry *to;
160 			loff_t n = file->f_pos - 2;
161 
162 			inode_lock_shared(dentry->d_inode);
163 			to = next_positive(dentry, &dentry->d_subdirs, n);
164 			move_cursor(cursor, to ? &to->d_child : NULL);
165 			inode_unlock_shared(dentry->d_inode);
166 		}
167 	}
168 	return offset;
169 }
170 EXPORT_SYMBOL(dcache_dir_lseek);
171 
172 /* Relationship between i_mode and the DT_xxx types */
173 static inline unsigned char dt_type(struct inode *inode)
174 {
175 	return (inode->i_mode >> 12) & 15;
176 }
177 
178 /*
179  * Directory is locked and all positive dentries in it are safe, since
180  * for ramfs-type trees they can't go away without unlink() or rmdir(),
181  * both impossible due to the lock on directory.
182  */
183 
184 int dcache_readdir(struct file *file, struct dir_context *ctx)
185 {
186 	struct dentry *dentry = file->f_path.dentry;
187 	struct dentry *cursor = file->private_data;
188 	struct list_head *p = &cursor->d_child;
189 	struct dentry *next;
190 	bool moved = false;
191 
192 	if (!dir_emit_dots(file, ctx))
193 		return 0;
194 
195 	if (ctx->pos == 2)
196 		p = &dentry->d_subdirs;
197 	while ((next = next_positive(dentry, p, 1)) != NULL) {
198 		if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
199 			      d_inode(next)->i_ino, dt_type(d_inode(next))))
200 			break;
201 		moved = true;
202 		p = &next->d_child;
203 		ctx->pos++;
204 	}
205 	if (moved)
206 		move_cursor(cursor, p);
207 	return 0;
208 }
209 EXPORT_SYMBOL(dcache_readdir);
210 
211 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
212 {
213 	return -EISDIR;
214 }
215 EXPORT_SYMBOL(generic_read_dir);
216 
217 const struct file_operations simple_dir_operations = {
218 	.open		= dcache_dir_open,
219 	.release	= dcache_dir_close,
220 	.llseek		= dcache_dir_lseek,
221 	.read		= generic_read_dir,
222 	.iterate_shared	= dcache_readdir,
223 	.fsync		= noop_fsync,
224 };
225 EXPORT_SYMBOL(simple_dir_operations);
226 
227 const struct inode_operations simple_dir_inode_operations = {
228 	.lookup		= simple_lookup,
229 };
230 EXPORT_SYMBOL(simple_dir_inode_operations);
231 
232 static const struct super_operations simple_super_operations = {
233 	.statfs		= simple_statfs,
234 };
235 
236 /*
237  * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
238  * will never be mountable)
239  */
240 struct dentry *mount_pseudo_xattr(struct file_system_type *fs_type, char *name,
241 	const struct super_operations *ops, const struct xattr_handler **xattr,
242 	const struct dentry_operations *dops, unsigned long magic)
243 {
244 	struct super_block *s;
245 	struct dentry *dentry;
246 	struct inode *root;
247 	struct qstr d_name = QSTR_INIT(name, strlen(name));
248 
249 	s = sget_userns(fs_type, NULL, set_anon_super, MS_KERNMOUNT|MS_NOUSER,
250 			&init_user_ns, NULL);
251 	if (IS_ERR(s))
252 		return ERR_CAST(s);
253 
254 	s->s_maxbytes = MAX_LFS_FILESIZE;
255 	s->s_blocksize = PAGE_SIZE;
256 	s->s_blocksize_bits = PAGE_SHIFT;
257 	s->s_magic = magic;
258 	s->s_op = ops ? ops : &simple_super_operations;
259 	s->s_xattr = xattr;
260 	s->s_time_gran = 1;
261 	root = new_inode(s);
262 	if (!root)
263 		goto Enomem;
264 	/*
265 	 * since this is the first inode, make it number 1. New inodes created
266 	 * after this must take care not to collide with it (by passing
267 	 * max_reserved of 1 to iunique).
268 	 */
269 	root->i_ino = 1;
270 	root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
271 	root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
272 	dentry = __d_alloc(s, &d_name);
273 	if (!dentry) {
274 		iput(root);
275 		goto Enomem;
276 	}
277 	d_instantiate(dentry, root);
278 	s->s_root = dentry;
279 	s->s_d_op = dops;
280 	s->s_flags |= MS_ACTIVE;
281 	return dget(s->s_root);
282 
283 Enomem:
284 	deactivate_locked_super(s);
285 	return ERR_PTR(-ENOMEM);
286 }
287 EXPORT_SYMBOL(mount_pseudo_xattr);
288 
289 int simple_open(struct inode *inode, struct file *file)
290 {
291 	if (inode->i_private)
292 		file->private_data = inode->i_private;
293 	return 0;
294 }
295 EXPORT_SYMBOL(simple_open);
296 
297 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
298 {
299 	struct inode *inode = d_inode(old_dentry);
300 
301 	inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
302 	inc_nlink(inode);
303 	ihold(inode);
304 	dget(dentry);
305 	d_instantiate(dentry, inode);
306 	return 0;
307 }
308 EXPORT_SYMBOL(simple_link);
309 
310 int simple_empty(struct dentry *dentry)
311 {
312 	struct dentry *child;
313 	int ret = 0;
314 
315 	spin_lock(&dentry->d_lock);
316 	list_for_each_entry(child, &dentry->d_subdirs, d_child) {
317 		spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
318 		if (simple_positive(child)) {
319 			spin_unlock(&child->d_lock);
320 			goto out;
321 		}
322 		spin_unlock(&child->d_lock);
323 	}
324 	ret = 1;
325 out:
326 	spin_unlock(&dentry->d_lock);
327 	return ret;
328 }
329 EXPORT_SYMBOL(simple_empty);
330 
331 int simple_unlink(struct inode *dir, struct dentry *dentry)
332 {
333 	struct inode *inode = d_inode(dentry);
334 
335 	inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
336 	drop_nlink(inode);
337 	dput(dentry);
338 	return 0;
339 }
340 EXPORT_SYMBOL(simple_unlink);
341 
342 int simple_rmdir(struct inode *dir, struct dentry *dentry)
343 {
344 	if (!simple_empty(dentry))
345 		return -ENOTEMPTY;
346 
347 	drop_nlink(d_inode(dentry));
348 	simple_unlink(dir, dentry);
349 	drop_nlink(dir);
350 	return 0;
351 }
352 EXPORT_SYMBOL(simple_rmdir);
353 
354 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
355 		  struct inode *new_dir, struct dentry *new_dentry,
356 		  unsigned int flags)
357 {
358 	struct inode *inode = d_inode(old_dentry);
359 	int they_are_dirs = d_is_dir(old_dentry);
360 
361 	if (flags & ~RENAME_NOREPLACE)
362 		return -EINVAL;
363 
364 	if (!simple_empty(new_dentry))
365 		return -ENOTEMPTY;
366 
367 	if (d_really_is_positive(new_dentry)) {
368 		simple_unlink(new_dir, new_dentry);
369 		if (they_are_dirs) {
370 			drop_nlink(d_inode(new_dentry));
371 			drop_nlink(old_dir);
372 		}
373 	} else if (they_are_dirs) {
374 		drop_nlink(old_dir);
375 		inc_nlink(new_dir);
376 	}
377 
378 	old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
379 		new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
380 
381 	return 0;
382 }
383 EXPORT_SYMBOL(simple_rename);
384 
385 /**
386  * simple_setattr - setattr for simple filesystem
387  * @dentry: dentry
388  * @iattr: iattr structure
389  *
390  * Returns 0 on success, -error on failure.
391  *
392  * simple_setattr is a simple ->setattr implementation without a proper
393  * implementation of size changes.
394  *
395  * It can either be used for in-memory filesystems or special files
396  * on simple regular filesystems.  Anything that needs to change on-disk
397  * or wire state on size changes needs its own setattr method.
398  */
399 int simple_setattr(struct dentry *dentry, struct iattr *iattr)
400 {
401 	struct inode *inode = d_inode(dentry);
402 	int error;
403 
404 	error = setattr_prepare(dentry, iattr);
405 	if (error)
406 		return error;
407 
408 	if (iattr->ia_valid & ATTR_SIZE)
409 		truncate_setsize(inode, iattr->ia_size);
410 	setattr_copy(inode, iattr);
411 	mark_inode_dirty(inode);
412 	return 0;
413 }
414 EXPORT_SYMBOL(simple_setattr);
415 
416 int simple_readpage(struct file *file, struct page *page)
417 {
418 	clear_highpage(page);
419 	flush_dcache_page(page);
420 	SetPageUptodate(page);
421 	unlock_page(page);
422 	return 0;
423 }
424 EXPORT_SYMBOL(simple_readpage);
425 
426 int simple_write_begin(struct file *file, struct address_space *mapping,
427 			loff_t pos, unsigned len, unsigned flags,
428 			struct page **pagep, void **fsdata)
429 {
430 	struct page *page;
431 	pgoff_t index;
432 
433 	index = pos >> PAGE_SHIFT;
434 
435 	page = grab_cache_page_write_begin(mapping, index, flags);
436 	if (!page)
437 		return -ENOMEM;
438 
439 	*pagep = page;
440 
441 	if (!PageUptodate(page) && (len != PAGE_SIZE)) {
442 		unsigned from = pos & (PAGE_SIZE - 1);
443 
444 		zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
445 	}
446 	return 0;
447 }
448 EXPORT_SYMBOL(simple_write_begin);
449 
450 /**
451  * simple_write_end - .write_end helper for non-block-device FSes
452  * @available: See .write_end of address_space_operations
453  * @file: 		"
454  * @mapping: 		"
455  * @pos: 		"
456  * @len: 		"
457  * @copied: 		"
458  * @page: 		"
459  * @fsdata: 		"
460  *
461  * simple_write_end does the minimum needed for updating a page after writing is
462  * done. It has the same API signature as the .write_end of
463  * address_space_operations vector. So it can just be set onto .write_end for
464  * FSes that don't need any other processing. i_mutex is assumed to be held.
465  * Block based filesystems should use generic_write_end().
466  * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
467  * is not called, so a filesystem that actually does store data in .write_inode
468  * should extend on what's done here with a call to mark_inode_dirty() in the
469  * case that i_size has changed.
470  *
471  * Use *ONLY* with simple_readpage()
472  */
473 int simple_write_end(struct file *file, struct address_space *mapping,
474 			loff_t pos, unsigned len, unsigned copied,
475 			struct page *page, void *fsdata)
476 {
477 	struct inode *inode = page->mapping->host;
478 	loff_t last_pos = pos + copied;
479 
480 	/* zero the stale part of the page if we did a short copy */
481 	if (!PageUptodate(page)) {
482 		if (copied < len) {
483 			unsigned from = pos & (PAGE_SIZE - 1);
484 
485 			zero_user(page, from + copied, len - copied);
486 		}
487 		SetPageUptodate(page);
488 	}
489 	/*
490 	 * No need to use i_size_read() here, the i_size
491 	 * cannot change under us because we hold the i_mutex.
492 	 */
493 	if (last_pos > inode->i_size)
494 		i_size_write(inode, last_pos);
495 
496 	set_page_dirty(page);
497 	unlock_page(page);
498 	put_page(page);
499 
500 	return copied;
501 }
502 EXPORT_SYMBOL(simple_write_end);
503 
504 /*
505  * the inodes created here are not hashed. If you use iunique to generate
506  * unique inode values later for this filesystem, then you must take care
507  * to pass it an appropriate max_reserved value to avoid collisions.
508  */
509 int simple_fill_super(struct super_block *s, unsigned long magic,
510 		      const struct tree_descr *files)
511 {
512 	struct inode *inode;
513 	struct dentry *root;
514 	struct dentry *dentry;
515 	int i;
516 
517 	s->s_blocksize = PAGE_SIZE;
518 	s->s_blocksize_bits = PAGE_SHIFT;
519 	s->s_magic = magic;
520 	s->s_op = &simple_super_operations;
521 	s->s_time_gran = 1;
522 
523 	inode = new_inode(s);
524 	if (!inode)
525 		return -ENOMEM;
526 	/*
527 	 * because the root inode is 1, the files array must not contain an
528 	 * entry at index 1
529 	 */
530 	inode->i_ino = 1;
531 	inode->i_mode = S_IFDIR | 0755;
532 	inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
533 	inode->i_op = &simple_dir_inode_operations;
534 	inode->i_fop = &simple_dir_operations;
535 	set_nlink(inode, 2);
536 	root = d_make_root(inode);
537 	if (!root)
538 		return -ENOMEM;
539 	for (i = 0; !files->name || files->name[0]; i++, files++) {
540 		if (!files->name)
541 			continue;
542 
543 		/* warn if it tries to conflict with the root inode */
544 		if (unlikely(i == 1))
545 			printk(KERN_WARNING "%s: %s passed in a files array"
546 				"with an index of 1!\n", __func__,
547 				s->s_type->name);
548 
549 		dentry = d_alloc_name(root, files->name);
550 		if (!dentry)
551 			goto out;
552 		inode = new_inode(s);
553 		if (!inode) {
554 			dput(dentry);
555 			goto out;
556 		}
557 		inode->i_mode = S_IFREG | files->mode;
558 		inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
559 		inode->i_fop = files->ops;
560 		inode->i_ino = i;
561 		d_add(dentry, inode);
562 	}
563 	s->s_root = root;
564 	return 0;
565 out:
566 	d_genocide(root);
567 	shrink_dcache_parent(root);
568 	dput(root);
569 	return -ENOMEM;
570 }
571 EXPORT_SYMBOL(simple_fill_super);
572 
573 static DEFINE_SPINLOCK(pin_fs_lock);
574 
575 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
576 {
577 	struct vfsmount *mnt = NULL;
578 	spin_lock(&pin_fs_lock);
579 	if (unlikely(!*mount)) {
580 		spin_unlock(&pin_fs_lock);
581 		mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, NULL);
582 		if (IS_ERR(mnt))
583 			return PTR_ERR(mnt);
584 		spin_lock(&pin_fs_lock);
585 		if (!*mount)
586 			*mount = mnt;
587 	}
588 	mntget(*mount);
589 	++*count;
590 	spin_unlock(&pin_fs_lock);
591 	mntput(mnt);
592 	return 0;
593 }
594 EXPORT_SYMBOL(simple_pin_fs);
595 
596 void simple_release_fs(struct vfsmount **mount, int *count)
597 {
598 	struct vfsmount *mnt;
599 	spin_lock(&pin_fs_lock);
600 	mnt = *mount;
601 	if (!--*count)
602 		*mount = NULL;
603 	spin_unlock(&pin_fs_lock);
604 	mntput(mnt);
605 }
606 EXPORT_SYMBOL(simple_release_fs);
607 
608 /**
609  * simple_read_from_buffer - copy data from the buffer to user space
610  * @to: the user space buffer to read to
611  * @count: the maximum number of bytes to read
612  * @ppos: the current position in the buffer
613  * @from: the buffer to read from
614  * @available: the size of the buffer
615  *
616  * The simple_read_from_buffer() function reads up to @count bytes from the
617  * buffer @from at offset @ppos into the user space address starting at @to.
618  *
619  * On success, the number of bytes read is returned and the offset @ppos is
620  * advanced by this number, or negative value is returned on error.
621  **/
622 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
623 				const void *from, size_t available)
624 {
625 	loff_t pos = *ppos;
626 	size_t ret;
627 
628 	if (pos < 0)
629 		return -EINVAL;
630 	if (pos >= available || !count)
631 		return 0;
632 	if (count > available - pos)
633 		count = available - pos;
634 	ret = copy_to_user(to, from + pos, count);
635 	if (ret == count)
636 		return -EFAULT;
637 	count -= ret;
638 	*ppos = pos + count;
639 	return count;
640 }
641 EXPORT_SYMBOL(simple_read_from_buffer);
642 
643 /**
644  * simple_write_to_buffer - copy data from user space to the buffer
645  * @to: the buffer to write to
646  * @available: the size of the buffer
647  * @ppos: the current position in the buffer
648  * @from: the user space buffer to read from
649  * @count: the maximum number of bytes to read
650  *
651  * The simple_write_to_buffer() function reads up to @count bytes from the user
652  * space address starting at @from into the buffer @to at offset @ppos.
653  *
654  * On success, the number of bytes written is returned and the offset @ppos is
655  * advanced by this number, or negative value is returned on error.
656  **/
657 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
658 		const void __user *from, size_t count)
659 {
660 	loff_t pos = *ppos;
661 	size_t res;
662 
663 	if (pos < 0)
664 		return -EINVAL;
665 	if (pos >= available || !count)
666 		return 0;
667 	if (count > available - pos)
668 		count = available - pos;
669 	res = copy_from_user(to + pos, from, count);
670 	if (res == count)
671 		return -EFAULT;
672 	count -= res;
673 	*ppos = pos + count;
674 	return count;
675 }
676 EXPORT_SYMBOL(simple_write_to_buffer);
677 
678 /**
679  * memory_read_from_buffer - copy data from the buffer
680  * @to: the kernel space buffer to read to
681  * @count: the maximum number of bytes to read
682  * @ppos: the current position in the buffer
683  * @from: the buffer to read from
684  * @available: the size of the buffer
685  *
686  * The memory_read_from_buffer() function reads up to @count bytes from the
687  * buffer @from at offset @ppos into the kernel space address starting at @to.
688  *
689  * On success, the number of bytes read is returned and the offset @ppos is
690  * advanced by this number, or negative value is returned on error.
691  **/
692 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
693 				const void *from, size_t available)
694 {
695 	loff_t pos = *ppos;
696 
697 	if (pos < 0)
698 		return -EINVAL;
699 	if (pos >= available)
700 		return 0;
701 	if (count > available - pos)
702 		count = available - pos;
703 	memcpy(to, from + pos, count);
704 	*ppos = pos + count;
705 
706 	return count;
707 }
708 EXPORT_SYMBOL(memory_read_from_buffer);
709 
710 /*
711  * Transaction based IO.
712  * The file expects a single write which triggers the transaction, and then
713  * possibly a read which collects the result - which is stored in a
714  * file-local buffer.
715  */
716 
717 void simple_transaction_set(struct file *file, size_t n)
718 {
719 	struct simple_transaction_argresp *ar = file->private_data;
720 
721 	BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
722 
723 	/*
724 	 * The barrier ensures that ar->size will really remain zero until
725 	 * ar->data is ready for reading.
726 	 */
727 	smp_mb();
728 	ar->size = n;
729 }
730 EXPORT_SYMBOL(simple_transaction_set);
731 
732 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
733 {
734 	struct simple_transaction_argresp *ar;
735 	static DEFINE_SPINLOCK(simple_transaction_lock);
736 
737 	if (size > SIMPLE_TRANSACTION_LIMIT - 1)
738 		return ERR_PTR(-EFBIG);
739 
740 	ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
741 	if (!ar)
742 		return ERR_PTR(-ENOMEM);
743 
744 	spin_lock(&simple_transaction_lock);
745 
746 	/* only one write allowed per open */
747 	if (file->private_data) {
748 		spin_unlock(&simple_transaction_lock);
749 		free_page((unsigned long)ar);
750 		return ERR_PTR(-EBUSY);
751 	}
752 
753 	file->private_data = ar;
754 
755 	spin_unlock(&simple_transaction_lock);
756 
757 	if (copy_from_user(ar->data, buf, size))
758 		return ERR_PTR(-EFAULT);
759 
760 	return ar->data;
761 }
762 EXPORT_SYMBOL(simple_transaction_get);
763 
764 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
765 {
766 	struct simple_transaction_argresp *ar = file->private_data;
767 
768 	if (!ar)
769 		return 0;
770 	return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
771 }
772 EXPORT_SYMBOL(simple_transaction_read);
773 
774 int simple_transaction_release(struct inode *inode, struct file *file)
775 {
776 	free_page((unsigned long)file->private_data);
777 	return 0;
778 }
779 EXPORT_SYMBOL(simple_transaction_release);
780 
781 /* Simple attribute files */
782 
783 struct simple_attr {
784 	int (*get)(void *, u64 *);
785 	int (*set)(void *, u64);
786 	char get_buf[24];	/* enough to store a u64 and "\n\0" */
787 	char set_buf[24];
788 	void *data;
789 	const char *fmt;	/* format for read operation */
790 	struct mutex mutex;	/* protects access to these buffers */
791 };
792 
793 /* simple_attr_open is called by an actual attribute open file operation
794  * to set the attribute specific access operations. */
795 int simple_attr_open(struct inode *inode, struct file *file,
796 		     int (*get)(void *, u64 *), int (*set)(void *, u64),
797 		     const char *fmt)
798 {
799 	struct simple_attr *attr;
800 
801 	attr = kmalloc(sizeof(*attr), GFP_KERNEL);
802 	if (!attr)
803 		return -ENOMEM;
804 
805 	attr->get = get;
806 	attr->set = set;
807 	attr->data = inode->i_private;
808 	attr->fmt = fmt;
809 	mutex_init(&attr->mutex);
810 
811 	file->private_data = attr;
812 
813 	return nonseekable_open(inode, file);
814 }
815 EXPORT_SYMBOL_GPL(simple_attr_open);
816 
817 int simple_attr_release(struct inode *inode, struct file *file)
818 {
819 	kfree(file->private_data);
820 	return 0;
821 }
822 EXPORT_SYMBOL_GPL(simple_attr_release);	/* GPL-only?  This?  Really? */
823 
824 /* read from the buffer that is filled with the get function */
825 ssize_t simple_attr_read(struct file *file, char __user *buf,
826 			 size_t len, loff_t *ppos)
827 {
828 	struct simple_attr *attr;
829 	size_t size;
830 	ssize_t ret;
831 
832 	attr = file->private_data;
833 
834 	if (!attr->get)
835 		return -EACCES;
836 
837 	ret = mutex_lock_interruptible(&attr->mutex);
838 	if (ret)
839 		return ret;
840 
841 	if (*ppos) {		/* continued read */
842 		size = strlen(attr->get_buf);
843 	} else {		/* first read */
844 		u64 val;
845 		ret = attr->get(attr->data, &val);
846 		if (ret)
847 			goto out;
848 
849 		size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
850 				 attr->fmt, (unsigned long long)val);
851 	}
852 
853 	ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
854 out:
855 	mutex_unlock(&attr->mutex);
856 	return ret;
857 }
858 EXPORT_SYMBOL_GPL(simple_attr_read);
859 
860 /* interpret the buffer as a number to call the set function with */
861 ssize_t simple_attr_write(struct file *file, const char __user *buf,
862 			  size_t len, loff_t *ppos)
863 {
864 	struct simple_attr *attr;
865 	u64 val;
866 	size_t size;
867 	ssize_t ret;
868 
869 	attr = file->private_data;
870 	if (!attr->set)
871 		return -EACCES;
872 
873 	ret = mutex_lock_interruptible(&attr->mutex);
874 	if (ret)
875 		return ret;
876 
877 	ret = -EFAULT;
878 	size = min(sizeof(attr->set_buf) - 1, len);
879 	if (copy_from_user(attr->set_buf, buf, size))
880 		goto out;
881 
882 	attr->set_buf[size] = '\0';
883 	val = simple_strtoll(attr->set_buf, NULL, 0);
884 	ret = attr->set(attr->data, val);
885 	if (ret == 0)
886 		ret = len; /* on success, claim we got the whole input */
887 out:
888 	mutex_unlock(&attr->mutex);
889 	return ret;
890 }
891 EXPORT_SYMBOL_GPL(simple_attr_write);
892 
893 /**
894  * generic_fh_to_dentry - generic helper for the fh_to_dentry 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 object specified in the file handle.
904  */
905 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
906 		int fh_len, int fh_type, struct inode *(*get_inode)
907 			(struct super_block *sb, u64 ino, u32 gen))
908 {
909 	struct inode *inode = NULL;
910 
911 	if (fh_len < 2)
912 		return NULL;
913 
914 	switch (fh_type) {
915 	case FILEID_INO32_GEN:
916 	case FILEID_INO32_GEN_PARENT:
917 		inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
918 		break;
919 	}
920 
921 	return d_obtain_alias(inode);
922 }
923 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
924 
925 /**
926  * generic_fh_to_parent - generic helper for the fh_to_parent export operation
927  * @sb:		filesystem to do the file handle conversion on
928  * @fid:	file handle to convert
929  * @fh_len:	length of the file handle in bytes
930  * @fh_type:	type of file handle
931  * @get_inode:	filesystem callback to retrieve inode
932  *
933  * This function decodes @fid as long as it has one of the well-known
934  * Linux filehandle types and calls @get_inode on it to retrieve the
935  * inode for the _parent_ object specified in the file handle if it
936  * is specified in the file handle, or NULL otherwise.
937  */
938 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
939 		int fh_len, int fh_type, struct inode *(*get_inode)
940 			(struct super_block *sb, u64 ino, u32 gen))
941 {
942 	struct inode *inode = NULL;
943 
944 	if (fh_len <= 2)
945 		return NULL;
946 
947 	switch (fh_type) {
948 	case FILEID_INO32_GEN_PARENT:
949 		inode = get_inode(sb, fid->i32.parent_ino,
950 				  (fh_len > 3 ? fid->i32.parent_gen : 0));
951 		break;
952 	}
953 
954 	return d_obtain_alias(inode);
955 }
956 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
957 
958 /**
959  * __generic_file_fsync - generic fsync implementation for simple filesystems
960  *
961  * @file:	file to synchronize
962  * @start:	start offset in bytes
963  * @end:	end offset in bytes (inclusive)
964  * @datasync:	only synchronize essential metadata if true
965  *
966  * This is a generic implementation of the fsync method for simple
967  * filesystems which track all non-inode metadata in the buffers list
968  * hanging off the address_space structure.
969  */
970 int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
971 				 int datasync)
972 {
973 	struct inode *inode = file->f_mapping->host;
974 	int err;
975 	int ret;
976 
977 	err = file_write_and_wait_range(file, start, end);
978 	if (err)
979 		return err;
980 
981 	inode_lock(inode);
982 	ret = sync_mapping_buffers(inode->i_mapping);
983 	if (!(inode->i_state & I_DIRTY_ALL))
984 		goto out;
985 	if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
986 		goto out;
987 
988 	err = sync_inode_metadata(inode, 1);
989 	if (ret == 0)
990 		ret = err;
991 
992 out:
993 	inode_unlock(inode);
994 	/* check and advance again to catch errors after syncing out buffers */
995 	err = file_check_and_advance_wb_err(file);
996 	if (ret == 0)
997 		ret = err;
998 	return ret;
999 }
1000 EXPORT_SYMBOL(__generic_file_fsync);
1001 
1002 /**
1003  * generic_file_fsync - generic fsync implementation for simple filesystems
1004  *			with flush
1005  * @file:	file to synchronize
1006  * @start:	start offset in bytes
1007  * @end:	end offset in bytes (inclusive)
1008  * @datasync:	only synchronize essential metadata if true
1009  *
1010  */
1011 
1012 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1013 		       int datasync)
1014 {
1015 	struct inode *inode = file->f_mapping->host;
1016 	int err;
1017 
1018 	err = __generic_file_fsync(file, start, end, datasync);
1019 	if (err)
1020 		return err;
1021 	return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
1022 }
1023 EXPORT_SYMBOL(generic_file_fsync);
1024 
1025 /**
1026  * generic_check_addressable - Check addressability of file system
1027  * @blocksize_bits:	log of file system block size
1028  * @num_blocks:		number of blocks in file system
1029  *
1030  * Determine whether a file system with @num_blocks blocks (and a
1031  * block size of 2**@blocksize_bits) is addressable by the sector_t
1032  * and page cache of the system.  Return 0 if so and -EFBIG otherwise.
1033  */
1034 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1035 {
1036 	u64 last_fs_block = num_blocks - 1;
1037 	u64 last_fs_page =
1038 		last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1039 
1040 	if (unlikely(num_blocks == 0))
1041 		return 0;
1042 
1043 	if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1044 		return -EINVAL;
1045 
1046 	if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1047 	    (last_fs_page > (pgoff_t)(~0ULL))) {
1048 		return -EFBIG;
1049 	}
1050 	return 0;
1051 }
1052 EXPORT_SYMBOL(generic_check_addressable);
1053 
1054 /*
1055  * No-op implementation of ->fsync for in-memory filesystems.
1056  */
1057 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1058 {
1059 	return 0;
1060 }
1061 EXPORT_SYMBOL(noop_fsync);
1062 
1063 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1064 void kfree_link(void *p)
1065 {
1066 	kfree(p);
1067 }
1068 EXPORT_SYMBOL(kfree_link);
1069 
1070 /*
1071  * nop .set_page_dirty method so that people can use .page_mkwrite on
1072  * anon inodes.
1073  */
1074 static int anon_set_page_dirty(struct page *page)
1075 {
1076 	return 0;
1077 };
1078 
1079 /*
1080  * A single inode exists for all anon_inode files. Contrary to pipes,
1081  * anon_inode inodes have no associated per-instance data, so we need
1082  * only allocate one of them.
1083  */
1084 struct inode *alloc_anon_inode(struct super_block *s)
1085 {
1086 	static const struct address_space_operations anon_aops = {
1087 		.set_page_dirty = anon_set_page_dirty,
1088 	};
1089 	struct inode *inode = new_inode_pseudo(s);
1090 
1091 	if (!inode)
1092 		return ERR_PTR(-ENOMEM);
1093 
1094 	inode->i_ino = get_next_ino();
1095 	inode->i_mapping->a_ops = &anon_aops;
1096 
1097 	/*
1098 	 * Mark the inode dirty from the very beginning,
1099 	 * that way it will never be moved to the dirty
1100 	 * list because mark_inode_dirty() will think
1101 	 * that it already _is_ on the dirty list.
1102 	 */
1103 	inode->i_state = I_DIRTY;
1104 	inode->i_mode = S_IRUSR | S_IWUSR;
1105 	inode->i_uid = current_fsuid();
1106 	inode->i_gid = current_fsgid();
1107 	inode->i_flags |= S_PRIVATE;
1108 	inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1109 	return inode;
1110 }
1111 EXPORT_SYMBOL(alloc_anon_inode);
1112 
1113 /**
1114  * simple_nosetlease - generic helper for prohibiting leases
1115  * @filp: file pointer
1116  * @arg: type of lease to obtain
1117  * @flp: new lease supplied for insertion
1118  * @priv: private data for lm_setup operation
1119  *
1120  * Generic helper for filesystems that do not wish to allow leases to be set.
1121  * All arguments are ignored and it just returns -EINVAL.
1122  */
1123 int
1124 simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1125 		  void **priv)
1126 {
1127 	return -EINVAL;
1128 }
1129 EXPORT_SYMBOL(simple_nosetlease);
1130 
1131 const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1132 			    struct delayed_call *done)
1133 {
1134 	return inode->i_link;
1135 }
1136 EXPORT_SYMBOL(simple_get_link);
1137 
1138 const struct inode_operations simple_symlink_inode_operations = {
1139 	.get_link = simple_get_link,
1140 };
1141 EXPORT_SYMBOL(simple_symlink_inode_operations);
1142 
1143 /*
1144  * Operations for a permanently empty directory.
1145  */
1146 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1147 {
1148 	return ERR_PTR(-ENOENT);
1149 }
1150 
1151 static int empty_dir_getattr(const struct path *path, struct kstat *stat,
1152 			     u32 request_mask, unsigned int query_flags)
1153 {
1154 	struct inode *inode = d_inode(path->dentry);
1155 	generic_fillattr(inode, stat);
1156 	return 0;
1157 }
1158 
1159 static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1160 {
1161 	return -EPERM;
1162 }
1163 
1164 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1165 {
1166 	return -EOPNOTSUPP;
1167 }
1168 
1169 static const struct inode_operations empty_dir_inode_operations = {
1170 	.lookup		= empty_dir_lookup,
1171 	.permission	= generic_permission,
1172 	.setattr	= empty_dir_setattr,
1173 	.getattr	= empty_dir_getattr,
1174 	.listxattr	= empty_dir_listxattr,
1175 };
1176 
1177 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1178 {
1179 	/* An empty directory has two entries . and .. at offsets 0 and 1 */
1180 	return generic_file_llseek_size(file, offset, whence, 2, 2);
1181 }
1182 
1183 static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1184 {
1185 	dir_emit_dots(file, ctx);
1186 	return 0;
1187 }
1188 
1189 static const struct file_operations empty_dir_operations = {
1190 	.llseek		= empty_dir_llseek,
1191 	.read		= generic_read_dir,
1192 	.iterate_shared	= empty_dir_readdir,
1193 	.fsync		= noop_fsync,
1194 };
1195 
1196 
1197 void make_empty_dir_inode(struct inode *inode)
1198 {
1199 	set_nlink(inode, 2);
1200 	inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1201 	inode->i_uid = GLOBAL_ROOT_UID;
1202 	inode->i_gid = GLOBAL_ROOT_GID;
1203 	inode->i_rdev = 0;
1204 	inode->i_size = 0;
1205 	inode->i_blkbits = PAGE_SHIFT;
1206 	inode->i_blocks = 0;
1207 
1208 	inode->i_op = &empty_dir_inode_operations;
1209 	inode->i_opflags &= ~IOP_XATTR;
1210 	inode->i_fop = &empty_dir_operations;
1211 }
1212 
1213 bool is_empty_dir_inode(struct inode *inode)
1214 {
1215 	return (inode->i_fop == &empty_dir_operations) &&
1216 		(inode->i_op == &empty_dir_inode_operations);
1217 }
1218