Loading...
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * fs/libfs.c
4 * Library for filesystems writers.
5 */
6
7#include <linux/blkdev.h>
8#include <linux/export.h>
9#include <linux/pagemap.h>
10#include <linux/slab.h>
11#include <linux/cred.h>
12#include <linux/mount.h>
13#include <linux/vfs.h>
14#include <linux/quotaops.h>
15#include <linux/mutex.h>
16#include <linux/namei.h>
17#include <linux/exportfs.h>
18#include <linux/writeback.h>
19#include <linux/buffer_head.h> /* sync_mapping_buffers */
20#include <linux/fs_context.h>
21#include <linux/pseudo_fs.h>
22
23#include <linux/uaccess.h>
24
25#include "internal.h"
26
27int simple_getattr(const struct path *path, struct kstat *stat,
28 u32 request_mask, unsigned int query_flags)
29{
30 struct inode *inode = d_inode(path->dentry);
31 generic_fillattr(inode, stat);
32 stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
33 return 0;
34}
35EXPORT_SYMBOL(simple_getattr);
36
37int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
38{
39 buf->f_type = dentry->d_sb->s_magic;
40 buf->f_bsize = PAGE_SIZE;
41 buf->f_namelen = NAME_MAX;
42 return 0;
43}
44EXPORT_SYMBOL(simple_statfs);
45
46/*
47 * Retaining negative dentries for an in-memory filesystem just wastes
48 * memory and lookup time: arrange for them to be deleted immediately.
49 */
50int always_delete_dentry(const struct dentry *dentry)
51{
52 return 1;
53}
54EXPORT_SYMBOL(always_delete_dentry);
55
56const struct dentry_operations simple_dentry_operations = {
57 .d_delete = always_delete_dentry,
58};
59EXPORT_SYMBOL(simple_dentry_operations);
60
61/*
62 * Lookup the data. This is trivial - if the dentry didn't already
63 * exist, we know it is negative. Set d_op to delete negative dentries.
64 */
65struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
66{
67 if (dentry->d_name.len > NAME_MAX)
68 return ERR_PTR(-ENAMETOOLONG);
69 if (!dentry->d_sb->s_d_op)
70 d_set_d_op(dentry, &simple_dentry_operations);
71 d_add(dentry, NULL);
72 return NULL;
73}
74EXPORT_SYMBOL(simple_lookup);
75
76int dcache_dir_open(struct inode *inode, struct file *file)
77{
78 file->private_data = d_alloc_cursor(file->f_path.dentry);
79
80 return file->private_data ? 0 : -ENOMEM;
81}
82EXPORT_SYMBOL(dcache_dir_open);
83
84int dcache_dir_close(struct inode *inode, struct file *file)
85{
86 dput(file->private_data);
87 return 0;
88}
89EXPORT_SYMBOL(dcache_dir_close);
90
91/* parent is locked at least shared */
92/*
93 * Returns an element of siblings' list.
94 * We are looking for <count>th positive after <p>; if
95 * found, dentry is grabbed and returned to caller.
96 * If no such element exists, NULL is returned.
97 */
98static struct dentry *scan_positives(struct dentry *cursor,
99 struct list_head *p,
100 loff_t count,
101 struct dentry *last)
102{
103 struct dentry *dentry = cursor->d_parent, *found = NULL;
104
105 spin_lock(&dentry->d_lock);
106 while ((p = p->next) != &dentry->d_subdirs) {
107 struct dentry *d = list_entry(p, struct dentry, d_child);
108 // we must at least skip cursors, to avoid livelocks
109 if (d->d_flags & DCACHE_DENTRY_CURSOR)
110 continue;
111 if (simple_positive(d) && !--count) {
112 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
113 if (simple_positive(d))
114 found = dget_dlock(d);
115 spin_unlock(&d->d_lock);
116 if (likely(found))
117 break;
118 count = 1;
119 }
120 if (need_resched()) {
121 list_move(&cursor->d_child, p);
122 p = &cursor->d_child;
123 spin_unlock(&dentry->d_lock);
124 cond_resched();
125 spin_lock(&dentry->d_lock);
126 }
127 }
128 spin_unlock(&dentry->d_lock);
129 dput(last);
130 return found;
131}
132
133loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
134{
135 struct dentry *dentry = file->f_path.dentry;
136 switch (whence) {
137 case 1:
138 offset += file->f_pos;
139 /* fall through */
140 case 0:
141 if (offset >= 0)
142 break;
143 /* fall through */
144 default:
145 return -EINVAL;
146 }
147 if (offset != file->f_pos) {
148 struct dentry *cursor = file->private_data;
149 struct dentry *to = NULL;
150
151 inode_lock_shared(dentry->d_inode);
152
153 if (offset > 2)
154 to = scan_positives(cursor, &dentry->d_subdirs,
155 offset - 2, NULL);
156 spin_lock(&dentry->d_lock);
157 if (to)
158 list_move(&cursor->d_child, &to->d_child);
159 else
160 list_del_init(&cursor->d_child);
161 spin_unlock(&dentry->d_lock);
162 dput(to);
163
164 file->f_pos = offset;
165
166 inode_unlock_shared(dentry->d_inode);
167 }
168 return offset;
169}
170EXPORT_SYMBOL(dcache_dir_lseek);
171
172/* Relationship between i_mode and the DT_xxx types */
173static 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
184int 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 *anchor = &dentry->d_subdirs;
189 struct dentry *next = NULL;
190 struct list_head *p;
191
192 if (!dir_emit_dots(file, ctx))
193 return 0;
194
195 if (ctx->pos == 2)
196 p = anchor;
197 else if (!list_empty(&cursor->d_child))
198 p = &cursor->d_child;
199 else
200 return 0;
201
202 while ((next = scan_positives(cursor, p, 1, next)) != NULL) {
203 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
204 d_inode(next)->i_ino, dt_type(d_inode(next))))
205 break;
206 ctx->pos++;
207 p = &next->d_child;
208 }
209 spin_lock(&dentry->d_lock);
210 if (next)
211 list_move_tail(&cursor->d_child, &next->d_child);
212 else
213 list_del_init(&cursor->d_child);
214 spin_unlock(&dentry->d_lock);
215 dput(next);
216
217 return 0;
218}
219EXPORT_SYMBOL(dcache_readdir);
220
221ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
222{
223 return -EISDIR;
224}
225EXPORT_SYMBOL(generic_read_dir);
226
227const struct file_operations simple_dir_operations = {
228 .open = dcache_dir_open,
229 .release = dcache_dir_close,
230 .llseek = dcache_dir_lseek,
231 .read = generic_read_dir,
232 .iterate_shared = dcache_readdir,
233 .fsync = noop_fsync,
234};
235EXPORT_SYMBOL(simple_dir_operations);
236
237const struct inode_operations simple_dir_inode_operations = {
238 .lookup = simple_lookup,
239};
240EXPORT_SYMBOL(simple_dir_inode_operations);
241
242static const struct super_operations simple_super_operations = {
243 .statfs = simple_statfs,
244};
245
246static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc)
247{
248 struct pseudo_fs_context *ctx = fc->fs_private;
249 struct inode *root;
250
251 s->s_maxbytes = MAX_LFS_FILESIZE;
252 s->s_blocksize = PAGE_SIZE;
253 s->s_blocksize_bits = PAGE_SHIFT;
254 s->s_magic = ctx->magic;
255 s->s_op = ctx->ops ?: &simple_super_operations;
256 s->s_xattr = ctx->xattr;
257 s->s_time_gran = 1;
258 root = new_inode(s);
259 if (!root)
260 return -ENOMEM;
261
262 /*
263 * since this is the first inode, make it number 1. New inodes created
264 * after this must take care not to collide with it (by passing
265 * max_reserved of 1 to iunique).
266 */
267 root->i_ino = 1;
268 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
269 root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
270 s->s_root = d_make_root(root);
271 if (!s->s_root)
272 return -ENOMEM;
273 s->s_d_op = ctx->dops;
274 return 0;
275}
276
277static int pseudo_fs_get_tree(struct fs_context *fc)
278{
279 return get_tree_nodev(fc, pseudo_fs_fill_super);
280}
281
282static void pseudo_fs_free(struct fs_context *fc)
283{
284 kfree(fc->fs_private);
285}
286
287static const struct fs_context_operations pseudo_fs_context_ops = {
288 .free = pseudo_fs_free,
289 .get_tree = pseudo_fs_get_tree,
290};
291
292/*
293 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
294 * will never be mountable)
295 */
296struct pseudo_fs_context *init_pseudo(struct fs_context *fc,
297 unsigned long magic)
298{
299 struct pseudo_fs_context *ctx;
300
301 ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL);
302 if (likely(ctx)) {
303 ctx->magic = magic;
304 fc->fs_private = ctx;
305 fc->ops = &pseudo_fs_context_ops;
306 fc->sb_flags |= SB_NOUSER;
307 fc->global = true;
308 }
309 return ctx;
310}
311EXPORT_SYMBOL(init_pseudo);
312
313int simple_open(struct inode *inode, struct file *file)
314{
315 if (inode->i_private)
316 file->private_data = inode->i_private;
317 return 0;
318}
319EXPORT_SYMBOL(simple_open);
320
321int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
322{
323 struct inode *inode = d_inode(old_dentry);
324
325 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
326 inc_nlink(inode);
327 ihold(inode);
328 dget(dentry);
329 d_instantiate(dentry, inode);
330 return 0;
331}
332EXPORT_SYMBOL(simple_link);
333
334int simple_empty(struct dentry *dentry)
335{
336 struct dentry *child;
337 int ret = 0;
338
339 spin_lock(&dentry->d_lock);
340 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
341 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
342 if (simple_positive(child)) {
343 spin_unlock(&child->d_lock);
344 goto out;
345 }
346 spin_unlock(&child->d_lock);
347 }
348 ret = 1;
349out:
350 spin_unlock(&dentry->d_lock);
351 return ret;
352}
353EXPORT_SYMBOL(simple_empty);
354
355int simple_unlink(struct inode *dir, struct dentry *dentry)
356{
357 struct inode *inode = d_inode(dentry);
358
359 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
360 drop_nlink(inode);
361 dput(dentry);
362 return 0;
363}
364EXPORT_SYMBOL(simple_unlink);
365
366int simple_rmdir(struct inode *dir, struct dentry *dentry)
367{
368 if (!simple_empty(dentry))
369 return -ENOTEMPTY;
370
371 drop_nlink(d_inode(dentry));
372 simple_unlink(dir, dentry);
373 drop_nlink(dir);
374 return 0;
375}
376EXPORT_SYMBOL(simple_rmdir);
377
378int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
379 struct inode *new_dir, struct dentry *new_dentry,
380 unsigned int flags)
381{
382 struct inode *inode = d_inode(old_dentry);
383 int they_are_dirs = d_is_dir(old_dentry);
384
385 if (flags & ~RENAME_NOREPLACE)
386 return -EINVAL;
387
388 if (!simple_empty(new_dentry))
389 return -ENOTEMPTY;
390
391 if (d_really_is_positive(new_dentry)) {
392 simple_unlink(new_dir, new_dentry);
393 if (they_are_dirs) {
394 drop_nlink(d_inode(new_dentry));
395 drop_nlink(old_dir);
396 }
397 } else if (they_are_dirs) {
398 drop_nlink(old_dir);
399 inc_nlink(new_dir);
400 }
401
402 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
403 new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
404
405 return 0;
406}
407EXPORT_SYMBOL(simple_rename);
408
409/**
410 * simple_setattr - setattr for simple filesystem
411 * @dentry: dentry
412 * @iattr: iattr structure
413 *
414 * Returns 0 on success, -error on failure.
415 *
416 * simple_setattr is a simple ->setattr implementation without a proper
417 * implementation of size changes.
418 *
419 * It can either be used for in-memory filesystems or special files
420 * on simple regular filesystems. Anything that needs to change on-disk
421 * or wire state on size changes needs its own setattr method.
422 */
423int simple_setattr(struct dentry *dentry, struct iattr *iattr)
424{
425 struct inode *inode = d_inode(dentry);
426 int error;
427
428 error = setattr_prepare(dentry, iattr);
429 if (error)
430 return error;
431
432 if (iattr->ia_valid & ATTR_SIZE)
433 truncate_setsize(inode, iattr->ia_size);
434 setattr_copy(inode, iattr);
435 mark_inode_dirty(inode);
436 return 0;
437}
438EXPORT_SYMBOL(simple_setattr);
439
440int simple_readpage(struct file *file, struct page *page)
441{
442 clear_highpage(page);
443 flush_dcache_page(page);
444 SetPageUptodate(page);
445 unlock_page(page);
446 return 0;
447}
448EXPORT_SYMBOL(simple_readpage);
449
450int simple_write_begin(struct file *file, struct address_space *mapping,
451 loff_t pos, unsigned len, unsigned flags,
452 struct page **pagep, void **fsdata)
453{
454 struct page *page;
455 pgoff_t index;
456
457 index = pos >> PAGE_SHIFT;
458
459 page = grab_cache_page_write_begin(mapping, index, flags);
460 if (!page)
461 return -ENOMEM;
462
463 *pagep = page;
464
465 if (!PageUptodate(page) && (len != PAGE_SIZE)) {
466 unsigned from = pos & (PAGE_SIZE - 1);
467
468 zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
469 }
470 return 0;
471}
472EXPORT_SYMBOL(simple_write_begin);
473
474/**
475 * simple_write_end - .write_end helper for non-block-device FSes
476 * @file: See .write_end of address_space_operations
477 * @mapping: "
478 * @pos: "
479 * @len: "
480 * @copied: "
481 * @page: "
482 * @fsdata: "
483 *
484 * simple_write_end does the minimum needed for updating a page after writing is
485 * done. It has the same API signature as the .write_end of
486 * address_space_operations vector. So it can just be set onto .write_end for
487 * FSes that don't need any other processing. i_mutex is assumed to be held.
488 * Block based filesystems should use generic_write_end().
489 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
490 * is not called, so a filesystem that actually does store data in .write_inode
491 * should extend on what's done here with a call to mark_inode_dirty() in the
492 * case that i_size has changed.
493 *
494 * Use *ONLY* with simple_readpage()
495 */
496int simple_write_end(struct file *file, struct address_space *mapping,
497 loff_t pos, unsigned len, unsigned copied,
498 struct page *page, void *fsdata)
499{
500 struct inode *inode = page->mapping->host;
501 loff_t last_pos = pos + copied;
502
503 /* zero the stale part of the page if we did a short copy */
504 if (!PageUptodate(page)) {
505 if (copied < len) {
506 unsigned from = pos & (PAGE_SIZE - 1);
507
508 zero_user(page, from + copied, len - copied);
509 }
510 SetPageUptodate(page);
511 }
512 /*
513 * No need to use i_size_read() here, the i_size
514 * cannot change under us because we hold the i_mutex.
515 */
516 if (last_pos > inode->i_size)
517 i_size_write(inode, last_pos);
518
519 set_page_dirty(page);
520 unlock_page(page);
521 put_page(page);
522
523 return copied;
524}
525EXPORT_SYMBOL(simple_write_end);
526
527/*
528 * the inodes created here are not hashed. If you use iunique to generate
529 * unique inode values later for this filesystem, then you must take care
530 * to pass it an appropriate max_reserved value to avoid collisions.
531 */
532int simple_fill_super(struct super_block *s, unsigned long magic,
533 const struct tree_descr *files)
534{
535 struct inode *inode;
536 struct dentry *root;
537 struct dentry *dentry;
538 int i;
539
540 s->s_blocksize = PAGE_SIZE;
541 s->s_blocksize_bits = PAGE_SHIFT;
542 s->s_magic = magic;
543 s->s_op = &simple_super_operations;
544 s->s_time_gran = 1;
545
546 inode = new_inode(s);
547 if (!inode)
548 return -ENOMEM;
549 /*
550 * because the root inode is 1, the files array must not contain an
551 * entry at index 1
552 */
553 inode->i_ino = 1;
554 inode->i_mode = S_IFDIR | 0755;
555 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
556 inode->i_op = &simple_dir_inode_operations;
557 inode->i_fop = &simple_dir_operations;
558 set_nlink(inode, 2);
559 root = d_make_root(inode);
560 if (!root)
561 return -ENOMEM;
562 for (i = 0; !files->name || files->name[0]; i++, files++) {
563 if (!files->name)
564 continue;
565
566 /* warn if it tries to conflict with the root inode */
567 if (unlikely(i == 1))
568 printk(KERN_WARNING "%s: %s passed in a files array"
569 "with an index of 1!\n", __func__,
570 s->s_type->name);
571
572 dentry = d_alloc_name(root, files->name);
573 if (!dentry)
574 goto out;
575 inode = new_inode(s);
576 if (!inode) {
577 dput(dentry);
578 goto out;
579 }
580 inode->i_mode = S_IFREG | files->mode;
581 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
582 inode->i_fop = files->ops;
583 inode->i_ino = i;
584 d_add(dentry, inode);
585 }
586 s->s_root = root;
587 return 0;
588out:
589 d_genocide(root);
590 shrink_dcache_parent(root);
591 dput(root);
592 return -ENOMEM;
593}
594EXPORT_SYMBOL(simple_fill_super);
595
596static DEFINE_SPINLOCK(pin_fs_lock);
597
598int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
599{
600 struct vfsmount *mnt = NULL;
601 spin_lock(&pin_fs_lock);
602 if (unlikely(!*mount)) {
603 spin_unlock(&pin_fs_lock);
604 mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
605 if (IS_ERR(mnt))
606 return PTR_ERR(mnt);
607 spin_lock(&pin_fs_lock);
608 if (!*mount)
609 *mount = mnt;
610 }
611 mntget(*mount);
612 ++*count;
613 spin_unlock(&pin_fs_lock);
614 mntput(mnt);
615 return 0;
616}
617EXPORT_SYMBOL(simple_pin_fs);
618
619void simple_release_fs(struct vfsmount **mount, int *count)
620{
621 struct vfsmount *mnt;
622 spin_lock(&pin_fs_lock);
623 mnt = *mount;
624 if (!--*count)
625 *mount = NULL;
626 spin_unlock(&pin_fs_lock);
627 mntput(mnt);
628}
629EXPORT_SYMBOL(simple_release_fs);
630
631/**
632 * simple_read_from_buffer - copy data from the buffer to user space
633 * @to: the user space buffer to read to
634 * @count: the maximum number of bytes to read
635 * @ppos: the current position in the buffer
636 * @from: the buffer to read from
637 * @available: the size of the buffer
638 *
639 * The simple_read_from_buffer() function reads up to @count bytes from the
640 * buffer @from at offset @ppos into the user space address starting at @to.
641 *
642 * On success, the number of bytes read is returned and the offset @ppos is
643 * advanced by this number, or negative value is returned on error.
644 **/
645ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
646 const void *from, size_t available)
647{
648 loff_t pos = *ppos;
649 size_t ret;
650
651 if (pos < 0)
652 return -EINVAL;
653 if (pos >= available || !count)
654 return 0;
655 if (count > available - pos)
656 count = available - pos;
657 ret = copy_to_user(to, from + pos, count);
658 if (ret == count)
659 return -EFAULT;
660 count -= ret;
661 *ppos = pos + count;
662 return count;
663}
664EXPORT_SYMBOL(simple_read_from_buffer);
665
666/**
667 * simple_write_to_buffer - copy data from user space to the buffer
668 * @to: the buffer to write to
669 * @available: the size of the buffer
670 * @ppos: the current position in the buffer
671 * @from: the user space buffer to read from
672 * @count: the maximum number of bytes to read
673 *
674 * The simple_write_to_buffer() function reads up to @count bytes from the user
675 * space address starting at @from into the buffer @to at offset @ppos.
676 *
677 * On success, the number of bytes written is returned and the offset @ppos is
678 * advanced by this number, or negative value is returned on error.
679 **/
680ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
681 const void __user *from, size_t count)
682{
683 loff_t pos = *ppos;
684 size_t res;
685
686 if (pos < 0)
687 return -EINVAL;
688 if (pos >= available || !count)
689 return 0;
690 if (count > available - pos)
691 count = available - pos;
692 res = copy_from_user(to + pos, from, count);
693 if (res == count)
694 return -EFAULT;
695 count -= res;
696 *ppos = pos + count;
697 return count;
698}
699EXPORT_SYMBOL(simple_write_to_buffer);
700
701/**
702 * memory_read_from_buffer - copy data from the buffer
703 * @to: the kernel space buffer to read to
704 * @count: the maximum number of bytes to read
705 * @ppos: the current position in the buffer
706 * @from: the buffer to read from
707 * @available: the size of the buffer
708 *
709 * The memory_read_from_buffer() function reads up to @count bytes from the
710 * buffer @from at offset @ppos into the kernel space address starting at @to.
711 *
712 * On success, the number of bytes read is returned and the offset @ppos is
713 * advanced by this number, or negative value is returned on error.
714 **/
715ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
716 const void *from, size_t available)
717{
718 loff_t pos = *ppos;
719
720 if (pos < 0)
721 return -EINVAL;
722 if (pos >= available)
723 return 0;
724 if (count > available - pos)
725 count = available - pos;
726 memcpy(to, from + pos, count);
727 *ppos = pos + count;
728
729 return count;
730}
731EXPORT_SYMBOL(memory_read_from_buffer);
732
733/*
734 * Transaction based IO.
735 * The file expects a single write which triggers the transaction, and then
736 * possibly a read which collects the result - which is stored in a
737 * file-local buffer.
738 */
739
740void simple_transaction_set(struct file *file, size_t n)
741{
742 struct simple_transaction_argresp *ar = file->private_data;
743
744 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
745
746 /*
747 * The barrier ensures that ar->size will really remain zero until
748 * ar->data is ready for reading.
749 */
750 smp_mb();
751 ar->size = n;
752}
753EXPORT_SYMBOL(simple_transaction_set);
754
755char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
756{
757 struct simple_transaction_argresp *ar;
758 static DEFINE_SPINLOCK(simple_transaction_lock);
759
760 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
761 return ERR_PTR(-EFBIG);
762
763 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
764 if (!ar)
765 return ERR_PTR(-ENOMEM);
766
767 spin_lock(&simple_transaction_lock);
768
769 /* only one write allowed per open */
770 if (file->private_data) {
771 spin_unlock(&simple_transaction_lock);
772 free_page((unsigned long)ar);
773 return ERR_PTR(-EBUSY);
774 }
775
776 file->private_data = ar;
777
778 spin_unlock(&simple_transaction_lock);
779
780 if (copy_from_user(ar->data, buf, size))
781 return ERR_PTR(-EFAULT);
782
783 return ar->data;
784}
785EXPORT_SYMBOL(simple_transaction_get);
786
787ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
788{
789 struct simple_transaction_argresp *ar = file->private_data;
790
791 if (!ar)
792 return 0;
793 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
794}
795EXPORT_SYMBOL(simple_transaction_read);
796
797int simple_transaction_release(struct inode *inode, struct file *file)
798{
799 free_page((unsigned long)file->private_data);
800 return 0;
801}
802EXPORT_SYMBOL(simple_transaction_release);
803
804/* Simple attribute files */
805
806struct simple_attr {
807 int (*get)(void *, u64 *);
808 int (*set)(void *, u64);
809 char get_buf[24]; /* enough to store a u64 and "\n\0" */
810 char set_buf[24];
811 void *data;
812 const char *fmt; /* format for read operation */
813 struct mutex mutex; /* protects access to these buffers */
814};
815
816/* simple_attr_open is called by an actual attribute open file operation
817 * to set the attribute specific access operations. */
818int simple_attr_open(struct inode *inode, struct file *file,
819 int (*get)(void *, u64 *), int (*set)(void *, u64),
820 const char *fmt)
821{
822 struct simple_attr *attr;
823
824 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
825 if (!attr)
826 return -ENOMEM;
827
828 attr->get = get;
829 attr->set = set;
830 attr->data = inode->i_private;
831 attr->fmt = fmt;
832 mutex_init(&attr->mutex);
833
834 file->private_data = attr;
835
836 return nonseekable_open(inode, file);
837}
838EXPORT_SYMBOL_GPL(simple_attr_open);
839
840int simple_attr_release(struct inode *inode, struct file *file)
841{
842 kfree(file->private_data);
843 return 0;
844}
845EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
846
847/* read from the buffer that is filled with the get function */
848ssize_t simple_attr_read(struct file *file, char __user *buf,
849 size_t len, loff_t *ppos)
850{
851 struct simple_attr *attr;
852 size_t size;
853 ssize_t ret;
854
855 attr = file->private_data;
856
857 if (!attr->get)
858 return -EACCES;
859
860 ret = mutex_lock_interruptible(&attr->mutex);
861 if (ret)
862 return ret;
863
864 if (*ppos) { /* continued read */
865 size = strlen(attr->get_buf);
866 } else { /* first read */
867 u64 val;
868 ret = attr->get(attr->data, &val);
869 if (ret)
870 goto out;
871
872 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
873 attr->fmt, (unsigned long long)val);
874 }
875
876 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
877out:
878 mutex_unlock(&attr->mutex);
879 return ret;
880}
881EXPORT_SYMBOL_GPL(simple_attr_read);
882
883/* interpret the buffer as a number to call the set function with */
884ssize_t simple_attr_write(struct file *file, const char __user *buf,
885 size_t len, loff_t *ppos)
886{
887 struct simple_attr *attr;
888 u64 val;
889 size_t size;
890 ssize_t ret;
891
892 attr = file->private_data;
893 if (!attr->set)
894 return -EACCES;
895
896 ret = mutex_lock_interruptible(&attr->mutex);
897 if (ret)
898 return ret;
899
900 ret = -EFAULT;
901 size = min(sizeof(attr->set_buf) - 1, len);
902 if (copy_from_user(attr->set_buf, buf, size))
903 goto out;
904
905 attr->set_buf[size] = '\0';
906 val = simple_strtoll(attr->set_buf, NULL, 0);
907 ret = attr->set(attr->data, val);
908 if (ret == 0)
909 ret = len; /* on success, claim we got the whole input */
910out:
911 mutex_unlock(&attr->mutex);
912 return ret;
913}
914EXPORT_SYMBOL_GPL(simple_attr_write);
915
916/**
917 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
918 * @sb: filesystem to do the file handle conversion on
919 * @fid: file handle to convert
920 * @fh_len: length of the file handle in bytes
921 * @fh_type: type of file handle
922 * @get_inode: filesystem callback to retrieve inode
923 *
924 * This function decodes @fid as long as it has one of the well-known
925 * Linux filehandle types and calls @get_inode on it to retrieve the
926 * inode for the object specified in the file handle.
927 */
928struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
929 int fh_len, int fh_type, struct inode *(*get_inode)
930 (struct super_block *sb, u64 ino, u32 gen))
931{
932 struct inode *inode = NULL;
933
934 if (fh_len < 2)
935 return NULL;
936
937 switch (fh_type) {
938 case FILEID_INO32_GEN:
939 case FILEID_INO32_GEN_PARENT:
940 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
941 break;
942 }
943
944 return d_obtain_alias(inode);
945}
946EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
947
948/**
949 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
950 * @sb: filesystem to do the file handle conversion on
951 * @fid: file handle to convert
952 * @fh_len: length of the file handle in bytes
953 * @fh_type: type of file handle
954 * @get_inode: filesystem callback to retrieve inode
955 *
956 * This function decodes @fid as long as it has one of the well-known
957 * Linux filehandle types and calls @get_inode on it to retrieve the
958 * inode for the _parent_ object specified in the file handle if it
959 * is specified in the file handle, or NULL otherwise.
960 */
961struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
962 int fh_len, int fh_type, struct inode *(*get_inode)
963 (struct super_block *sb, u64 ino, u32 gen))
964{
965 struct inode *inode = NULL;
966
967 if (fh_len <= 2)
968 return NULL;
969
970 switch (fh_type) {
971 case FILEID_INO32_GEN_PARENT:
972 inode = get_inode(sb, fid->i32.parent_ino,
973 (fh_len > 3 ? fid->i32.parent_gen : 0));
974 break;
975 }
976
977 return d_obtain_alias(inode);
978}
979EXPORT_SYMBOL_GPL(generic_fh_to_parent);
980
981/**
982 * __generic_file_fsync - generic fsync implementation for simple filesystems
983 *
984 * @file: file to synchronize
985 * @start: start offset in bytes
986 * @end: end offset in bytes (inclusive)
987 * @datasync: only synchronize essential metadata if true
988 *
989 * This is a generic implementation of the fsync method for simple
990 * filesystems which track all non-inode metadata in the buffers list
991 * hanging off the address_space structure.
992 */
993int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
994 int datasync)
995{
996 struct inode *inode = file->f_mapping->host;
997 int err;
998 int ret;
999
1000 err = file_write_and_wait_range(file, start, end);
1001 if (err)
1002 return err;
1003
1004 inode_lock(inode);
1005 ret = sync_mapping_buffers(inode->i_mapping);
1006 if (!(inode->i_state & I_DIRTY_ALL))
1007 goto out;
1008 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
1009 goto out;
1010
1011 err = sync_inode_metadata(inode, 1);
1012 if (ret == 0)
1013 ret = err;
1014
1015out:
1016 inode_unlock(inode);
1017 /* check and advance again to catch errors after syncing out buffers */
1018 err = file_check_and_advance_wb_err(file);
1019 if (ret == 0)
1020 ret = err;
1021 return ret;
1022}
1023EXPORT_SYMBOL(__generic_file_fsync);
1024
1025/**
1026 * generic_file_fsync - generic fsync implementation for simple filesystems
1027 * with flush
1028 * @file: file to synchronize
1029 * @start: start offset in bytes
1030 * @end: end offset in bytes (inclusive)
1031 * @datasync: only synchronize essential metadata if true
1032 *
1033 */
1034
1035int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1036 int datasync)
1037{
1038 struct inode *inode = file->f_mapping->host;
1039 int err;
1040
1041 err = __generic_file_fsync(file, start, end, datasync);
1042 if (err)
1043 return err;
1044 return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
1045}
1046EXPORT_SYMBOL(generic_file_fsync);
1047
1048/**
1049 * generic_check_addressable - Check addressability of file system
1050 * @blocksize_bits: log of file system block size
1051 * @num_blocks: number of blocks in file system
1052 *
1053 * Determine whether a file system with @num_blocks blocks (and a
1054 * block size of 2**@blocksize_bits) is addressable by the sector_t
1055 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1056 */
1057int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1058{
1059 u64 last_fs_block = num_blocks - 1;
1060 u64 last_fs_page =
1061 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1062
1063 if (unlikely(num_blocks == 0))
1064 return 0;
1065
1066 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1067 return -EINVAL;
1068
1069 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1070 (last_fs_page > (pgoff_t)(~0ULL))) {
1071 return -EFBIG;
1072 }
1073 return 0;
1074}
1075EXPORT_SYMBOL(generic_check_addressable);
1076
1077/*
1078 * No-op implementation of ->fsync for in-memory filesystems.
1079 */
1080int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1081{
1082 return 0;
1083}
1084EXPORT_SYMBOL(noop_fsync);
1085
1086int noop_set_page_dirty(struct page *page)
1087{
1088 /*
1089 * Unlike __set_page_dirty_no_writeback that handles dirty page
1090 * tracking in the page object, dax does all dirty tracking in
1091 * the inode address_space in response to mkwrite faults. In the
1092 * dax case we only need to worry about potentially dirty CPU
1093 * caches, not dirty page cache pages to write back.
1094 *
1095 * This callback is defined to prevent fallback to
1096 * __set_page_dirty_buffers() in set_page_dirty().
1097 */
1098 return 0;
1099}
1100EXPORT_SYMBOL_GPL(noop_set_page_dirty);
1101
1102void noop_invalidatepage(struct page *page, unsigned int offset,
1103 unsigned int length)
1104{
1105 /*
1106 * There is no page cache to invalidate in the dax case, however
1107 * we need this callback defined to prevent falling back to
1108 * block_invalidatepage() in do_invalidatepage().
1109 */
1110}
1111EXPORT_SYMBOL_GPL(noop_invalidatepage);
1112
1113ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1114{
1115 /*
1116 * iomap based filesystems support direct I/O without need for
1117 * this callback. However, it still needs to be set in
1118 * inode->a_ops so that open/fcntl know that direct I/O is
1119 * generally supported.
1120 */
1121 return -EINVAL;
1122}
1123EXPORT_SYMBOL_GPL(noop_direct_IO);
1124
1125/* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1126void kfree_link(void *p)
1127{
1128 kfree(p);
1129}
1130EXPORT_SYMBOL(kfree_link);
1131
1132/*
1133 * nop .set_page_dirty method so that people can use .page_mkwrite on
1134 * anon inodes.
1135 */
1136static int anon_set_page_dirty(struct page *page)
1137{
1138 return 0;
1139};
1140
1141/*
1142 * A single inode exists for all anon_inode files. Contrary to pipes,
1143 * anon_inode inodes have no associated per-instance data, so we need
1144 * only allocate one of them.
1145 */
1146struct inode *alloc_anon_inode(struct super_block *s)
1147{
1148 static const struct address_space_operations anon_aops = {
1149 .set_page_dirty = anon_set_page_dirty,
1150 };
1151 struct inode *inode = new_inode_pseudo(s);
1152
1153 if (!inode)
1154 return ERR_PTR(-ENOMEM);
1155
1156 inode->i_ino = get_next_ino();
1157 inode->i_mapping->a_ops = &anon_aops;
1158
1159 /*
1160 * Mark the inode dirty from the very beginning,
1161 * that way it will never be moved to the dirty
1162 * list because mark_inode_dirty() will think
1163 * that it already _is_ on the dirty list.
1164 */
1165 inode->i_state = I_DIRTY;
1166 inode->i_mode = S_IRUSR | S_IWUSR;
1167 inode->i_uid = current_fsuid();
1168 inode->i_gid = current_fsgid();
1169 inode->i_flags |= S_PRIVATE;
1170 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1171 return inode;
1172}
1173EXPORT_SYMBOL(alloc_anon_inode);
1174
1175/**
1176 * simple_nosetlease - generic helper for prohibiting leases
1177 * @filp: file pointer
1178 * @arg: type of lease to obtain
1179 * @flp: new lease supplied for insertion
1180 * @priv: private data for lm_setup operation
1181 *
1182 * Generic helper for filesystems that do not wish to allow leases to be set.
1183 * All arguments are ignored and it just returns -EINVAL.
1184 */
1185int
1186simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1187 void **priv)
1188{
1189 return -EINVAL;
1190}
1191EXPORT_SYMBOL(simple_nosetlease);
1192
1193/**
1194 * simple_get_link - generic helper to get the target of "fast" symlinks
1195 * @dentry: not used here
1196 * @inode: the symlink inode
1197 * @done: not used here
1198 *
1199 * Generic helper for filesystems to use for symlink inodes where a pointer to
1200 * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1201 * since as an optimization the path lookup code uses any non-NULL ->i_link
1202 * directly, without calling ->get_link(). But ->get_link() still must be set,
1203 * to mark the inode_operations as being for a symlink.
1204 *
1205 * Return: the symlink target
1206 */
1207const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1208 struct delayed_call *done)
1209{
1210 return inode->i_link;
1211}
1212EXPORT_SYMBOL(simple_get_link);
1213
1214const struct inode_operations simple_symlink_inode_operations = {
1215 .get_link = simple_get_link,
1216};
1217EXPORT_SYMBOL(simple_symlink_inode_operations);
1218
1219/*
1220 * Operations for a permanently empty directory.
1221 */
1222static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1223{
1224 return ERR_PTR(-ENOENT);
1225}
1226
1227static int empty_dir_getattr(const struct path *path, struct kstat *stat,
1228 u32 request_mask, unsigned int query_flags)
1229{
1230 struct inode *inode = d_inode(path->dentry);
1231 generic_fillattr(inode, stat);
1232 return 0;
1233}
1234
1235static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1236{
1237 return -EPERM;
1238}
1239
1240static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1241{
1242 return -EOPNOTSUPP;
1243}
1244
1245static const struct inode_operations empty_dir_inode_operations = {
1246 .lookup = empty_dir_lookup,
1247 .permission = generic_permission,
1248 .setattr = empty_dir_setattr,
1249 .getattr = empty_dir_getattr,
1250 .listxattr = empty_dir_listxattr,
1251};
1252
1253static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1254{
1255 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1256 return generic_file_llseek_size(file, offset, whence, 2, 2);
1257}
1258
1259static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1260{
1261 dir_emit_dots(file, ctx);
1262 return 0;
1263}
1264
1265static const struct file_operations empty_dir_operations = {
1266 .llseek = empty_dir_llseek,
1267 .read = generic_read_dir,
1268 .iterate_shared = empty_dir_readdir,
1269 .fsync = noop_fsync,
1270};
1271
1272
1273void make_empty_dir_inode(struct inode *inode)
1274{
1275 set_nlink(inode, 2);
1276 inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1277 inode->i_uid = GLOBAL_ROOT_UID;
1278 inode->i_gid = GLOBAL_ROOT_GID;
1279 inode->i_rdev = 0;
1280 inode->i_size = 0;
1281 inode->i_blkbits = PAGE_SHIFT;
1282 inode->i_blocks = 0;
1283
1284 inode->i_op = &empty_dir_inode_operations;
1285 inode->i_opflags &= ~IOP_XATTR;
1286 inode->i_fop = &empty_dir_operations;
1287}
1288
1289bool is_empty_dir_inode(struct inode *inode)
1290{
1291 return (inode->i_fop == &empty_dir_operations) &&
1292 (inode->i_op == &empty_dir_inode_operations);
1293}
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 <linux/uaccess.h>
20
21#include "internal.h"
22
23int 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}
31EXPORT_SYMBOL(simple_getattr);
32
33int 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}
40EXPORT_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 */
46int always_delete_dentry(const struct dentry *dentry)
47{
48 return 1;
49}
50EXPORT_SYMBOL(always_delete_dentry);
51
52const struct dentry_operations simple_dentry_operations = {
53 .d_delete = always_delete_dentry,
54};
55EXPORT_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 */
61struct 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}
70EXPORT_SYMBOL(simple_lookup);
71
72int 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}
78EXPORT_SYMBOL(dcache_dir_open);
79
80int dcache_dir_close(struct inode *inode, struct file *file)
81{
82 dput(file->private_data);
83 return 0;
84}
85EXPORT_SYMBOL(dcache_dir_close);
86
87/* parent is locked at least shared */
88static 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
98retry:
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
122static 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
142loff_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}
169EXPORT_SYMBOL(dcache_dir_lseek);
170
171/* Relationship between i_mode and the DT_xxx types */
172static 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
183int 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}
208EXPORT_SYMBOL(dcache_readdir);
209
210ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
211{
212 return -EISDIR;
213}
214EXPORT_SYMBOL(generic_read_dir);
215
216const 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};
224EXPORT_SYMBOL(simple_dir_operations);
225
226const struct inode_operations simple_dir_inode_operations = {
227 .lookup = simple_lookup,
228};
229EXPORT_SYMBOL(simple_dir_inode_operations);
230
231static 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 */
239struct 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
282Enomem:
283 deactivate_locked_super(s);
284 return ERR_PTR(-ENOMEM);
285}
286EXPORT_SYMBOL(mount_pseudo_xattr);
287
288int 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}
294EXPORT_SYMBOL(simple_open);
295
296int 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}
307EXPORT_SYMBOL(simple_link);
308
309int 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;
324out:
325 spin_unlock(&dentry->d_lock);
326 return ret;
327}
328EXPORT_SYMBOL(simple_empty);
329
330int 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}
339EXPORT_SYMBOL(simple_unlink);
340
341int 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}
351EXPORT_SYMBOL(simple_rmdir);
352
353int 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}
382EXPORT_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 */
398int 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}
413EXPORT_SYMBOL(simple_setattr);
414
415int 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}
423EXPORT_SYMBOL(simple_readpage);
424
425int 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}
447EXPORT_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 * Use *ONLY* with simple_readpage()
471 */
472int simple_write_end(struct file *file, struct address_space *mapping,
473 loff_t pos, unsigned len, unsigned copied,
474 struct page *page, void *fsdata)
475{
476 struct inode *inode = page->mapping->host;
477 loff_t last_pos = pos + copied;
478
479 /* zero the stale part of the page if we did a short copy */
480 if (!PageUptodate(page)) {
481 if (copied < len) {
482 unsigned from = pos & (PAGE_SIZE - 1);
483
484 zero_user(page, from + copied, len - copied);
485 }
486 SetPageUptodate(page);
487 }
488 /*
489 * No need to use i_size_read() here, the i_size
490 * cannot change under us because we hold the i_mutex.
491 */
492 if (last_pos > inode->i_size)
493 i_size_write(inode, last_pos);
494
495 set_page_dirty(page);
496 unlock_page(page);
497 put_page(page);
498
499 return copied;
500}
501EXPORT_SYMBOL(simple_write_end);
502
503/*
504 * the inodes created here are not hashed. If you use iunique to generate
505 * unique inode values later for this filesystem, then you must take care
506 * to pass it an appropriate max_reserved value to avoid collisions.
507 */
508int simple_fill_super(struct super_block *s, unsigned long magic,
509 struct tree_descr *files)
510{
511 struct inode *inode;
512 struct dentry *root;
513 struct dentry *dentry;
514 int i;
515
516 s->s_blocksize = PAGE_SIZE;
517 s->s_blocksize_bits = PAGE_SHIFT;
518 s->s_magic = magic;
519 s->s_op = &simple_super_operations;
520 s->s_time_gran = 1;
521
522 inode = new_inode(s);
523 if (!inode)
524 return -ENOMEM;
525 /*
526 * because the root inode is 1, the files array must not contain an
527 * entry at index 1
528 */
529 inode->i_ino = 1;
530 inode->i_mode = S_IFDIR | 0755;
531 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
532 inode->i_op = &simple_dir_inode_operations;
533 inode->i_fop = &simple_dir_operations;
534 set_nlink(inode, 2);
535 root = d_make_root(inode);
536 if (!root)
537 return -ENOMEM;
538 for (i = 0; !files->name || files->name[0]; i++, files++) {
539 if (!files->name)
540 continue;
541
542 /* warn if it tries to conflict with the root inode */
543 if (unlikely(i == 1))
544 printk(KERN_WARNING "%s: %s passed in a files array"
545 "with an index of 1!\n", __func__,
546 s->s_type->name);
547
548 dentry = d_alloc_name(root, files->name);
549 if (!dentry)
550 goto out;
551 inode = new_inode(s);
552 if (!inode) {
553 dput(dentry);
554 goto out;
555 }
556 inode->i_mode = S_IFREG | files->mode;
557 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
558 inode->i_fop = files->ops;
559 inode->i_ino = i;
560 d_add(dentry, inode);
561 }
562 s->s_root = root;
563 return 0;
564out:
565 d_genocide(root);
566 shrink_dcache_parent(root);
567 dput(root);
568 return -ENOMEM;
569}
570EXPORT_SYMBOL(simple_fill_super);
571
572static DEFINE_SPINLOCK(pin_fs_lock);
573
574int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
575{
576 struct vfsmount *mnt = NULL;
577 spin_lock(&pin_fs_lock);
578 if (unlikely(!*mount)) {
579 spin_unlock(&pin_fs_lock);
580 mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, NULL);
581 if (IS_ERR(mnt))
582 return PTR_ERR(mnt);
583 spin_lock(&pin_fs_lock);
584 if (!*mount)
585 *mount = mnt;
586 }
587 mntget(*mount);
588 ++*count;
589 spin_unlock(&pin_fs_lock);
590 mntput(mnt);
591 return 0;
592}
593EXPORT_SYMBOL(simple_pin_fs);
594
595void simple_release_fs(struct vfsmount **mount, int *count)
596{
597 struct vfsmount *mnt;
598 spin_lock(&pin_fs_lock);
599 mnt = *mount;
600 if (!--*count)
601 *mount = NULL;
602 spin_unlock(&pin_fs_lock);
603 mntput(mnt);
604}
605EXPORT_SYMBOL(simple_release_fs);
606
607/**
608 * simple_read_from_buffer - copy data from the buffer to user space
609 * @to: the user space buffer to read to
610 * @count: the maximum number of bytes to read
611 * @ppos: the current position in the buffer
612 * @from: the buffer to read from
613 * @available: the size of the buffer
614 *
615 * The simple_read_from_buffer() function reads up to @count bytes from the
616 * buffer @from at offset @ppos into the user space address starting at @to.
617 *
618 * On success, the number of bytes read is returned and the offset @ppos is
619 * advanced by this number, or negative value is returned on error.
620 **/
621ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
622 const void *from, size_t available)
623{
624 loff_t pos = *ppos;
625 size_t ret;
626
627 if (pos < 0)
628 return -EINVAL;
629 if (pos >= available || !count)
630 return 0;
631 if (count > available - pos)
632 count = available - pos;
633 ret = copy_to_user(to, from + pos, count);
634 if (ret == count)
635 return -EFAULT;
636 count -= ret;
637 *ppos = pos + count;
638 return count;
639}
640EXPORT_SYMBOL(simple_read_from_buffer);
641
642/**
643 * simple_write_to_buffer - copy data from user space to the buffer
644 * @to: the buffer to write to
645 * @available: the size of the buffer
646 * @ppos: the current position in the buffer
647 * @from: the user space buffer to read from
648 * @count: the maximum number of bytes to read
649 *
650 * The simple_write_to_buffer() function reads up to @count bytes from the user
651 * space address starting at @from into the buffer @to at offset @ppos.
652 *
653 * On success, the number of bytes written is returned and the offset @ppos is
654 * advanced by this number, or negative value is returned on error.
655 **/
656ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
657 const void __user *from, size_t count)
658{
659 loff_t pos = *ppos;
660 size_t res;
661
662 if (pos < 0)
663 return -EINVAL;
664 if (pos >= available || !count)
665 return 0;
666 if (count > available - pos)
667 count = available - pos;
668 res = copy_from_user(to + pos, from, count);
669 if (res == count)
670 return -EFAULT;
671 count -= res;
672 *ppos = pos + count;
673 return count;
674}
675EXPORT_SYMBOL(simple_write_to_buffer);
676
677/**
678 * memory_read_from_buffer - copy data from the buffer
679 * @to: the kernel space buffer to read to
680 * @count: the maximum number of bytes to read
681 * @ppos: the current position in the buffer
682 * @from: the buffer to read from
683 * @available: the size of the buffer
684 *
685 * The memory_read_from_buffer() function reads up to @count bytes from the
686 * buffer @from at offset @ppos into the kernel space address starting at @to.
687 *
688 * On success, the number of bytes read is returned and the offset @ppos is
689 * advanced by this number, or negative value is returned on error.
690 **/
691ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
692 const void *from, size_t available)
693{
694 loff_t pos = *ppos;
695
696 if (pos < 0)
697 return -EINVAL;
698 if (pos >= available)
699 return 0;
700 if (count > available - pos)
701 count = available - pos;
702 memcpy(to, from + pos, count);
703 *ppos = pos + count;
704
705 return count;
706}
707EXPORT_SYMBOL(memory_read_from_buffer);
708
709/*
710 * Transaction based IO.
711 * The file expects a single write which triggers the transaction, and then
712 * possibly a read which collects the result - which is stored in a
713 * file-local buffer.
714 */
715
716void simple_transaction_set(struct file *file, size_t n)
717{
718 struct simple_transaction_argresp *ar = file->private_data;
719
720 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
721
722 /*
723 * The barrier ensures that ar->size will really remain zero until
724 * ar->data is ready for reading.
725 */
726 smp_mb();
727 ar->size = n;
728}
729EXPORT_SYMBOL(simple_transaction_set);
730
731char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
732{
733 struct simple_transaction_argresp *ar;
734 static DEFINE_SPINLOCK(simple_transaction_lock);
735
736 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
737 return ERR_PTR(-EFBIG);
738
739 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
740 if (!ar)
741 return ERR_PTR(-ENOMEM);
742
743 spin_lock(&simple_transaction_lock);
744
745 /* only one write allowed per open */
746 if (file->private_data) {
747 spin_unlock(&simple_transaction_lock);
748 free_page((unsigned long)ar);
749 return ERR_PTR(-EBUSY);
750 }
751
752 file->private_data = ar;
753
754 spin_unlock(&simple_transaction_lock);
755
756 if (copy_from_user(ar->data, buf, size))
757 return ERR_PTR(-EFAULT);
758
759 return ar->data;
760}
761EXPORT_SYMBOL(simple_transaction_get);
762
763ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
764{
765 struct simple_transaction_argresp *ar = file->private_data;
766
767 if (!ar)
768 return 0;
769 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
770}
771EXPORT_SYMBOL(simple_transaction_read);
772
773int simple_transaction_release(struct inode *inode, struct file *file)
774{
775 free_page((unsigned long)file->private_data);
776 return 0;
777}
778EXPORT_SYMBOL(simple_transaction_release);
779
780/* Simple attribute files */
781
782struct simple_attr {
783 int (*get)(void *, u64 *);
784 int (*set)(void *, u64);
785 char get_buf[24]; /* enough to store a u64 and "\n\0" */
786 char set_buf[24];
787 void *data;
788 const char *fmt; /* format for read operation */
789 struct mutex mutex; /* protects access to these buffers */
790};
791
792/* simple_attr_open is called by an actual attribute open file operation
793 * to set the attribute specific access operations. */
794int simple_attr_open(struct inode *inode, struct file *file,
795 int (*get)(void *, u64 *), int (*set)(void *, u64),
796 const char *fmt)
797{
798 struct simple_attr *attr;
799
800 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
801 if (!attr)
802 return -ENOMEM;
803
804 attr->get = get;
805 attr->set = set;
806 attr->data = inode->i_private;
807 attr->fmt = fmt;
808 mutex_init(&attr->mutex);
809
810 file->private_data = attr;
811
812 return nonseekable_open(inode, file);
813}
814EXPORT_SYMBOL_GPL(simple_attr_open);
815
816int simple_attr_release(struct inode *inode, struct file *file)
817{
818 kfree(file->private_data);
819 return 0;
820}
821EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
822
823/* read from the buffer that is filled with the get function */
824ssize_t simple_attr_read(struct file *file, char __user *buf,
825 size_t len, loff_t *ppos)
826{
827 struct simple_attr *attr;
828 size_t size;
829 ssize_t ret;
830
831 attr = file->private_data;
832
833 if (!attr->get)
834 return -EACCES;
835
836 ret = mutex_lock_interruptible(&attr->mutex);
837 if (ret)
838 return ret;
839
840 if (*ppos) { /* continued read */
841 size = strlen(attr->get_buf);
842 } else { /* first read */
843 u64 val;
844 ret = attr->get(attr->data, &val);
845 if (ret)
846 goto out;
847
848 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
849 attr->fmt, (unsigned long long)val);
850 }
851
852 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
853out:
854 mutex_unlock(&attr->mutex);
855 return ret;
856}
857EXPORT_SYMBOL_GPL(simple_attr_read);
858
859/* interpret the buffer as a number to call the set function with */
860ssize_t simple_attr_write(struct file *file, const char __user *buf,
861 size_t len, loff_t *ppos)
862{
863 struct simple_attr *attr;
864 u64 val;
865 size_t size;
866 ssize_t ret;
867
868 attr = file->private_data;
869 if (!attr->set)
870 return -EACCES;
871
872 ret = mutex_lock_interruptible(&attr->mutex);
873 if (ret)
874 return ret;
875
876 ret = -EFAULT;
877 size = min(sizeof(attr->set_buf) - 1, len);
878 if (copy_from_user(attr->set_buf, buf, size))
879 goto out;
880
881 attr->set_buf[size] = '\0';
882 val = simple_strtoll(attr->set_buf, NULL, 0);
883 ret = attr->set(attr->data, val);
884 if (ret == 0)
885 ret = len; /* on success, claim we got the whole input */
886out:
887 mutex_unlock(&attr->mutex);
888 return ret;
889}
890EXPORT_SYMBOL_GPL(simple_attr_write);
891
892/**
893 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
894 * @sb: filesystem to do the file handle conversion on
895 * @fid: file handle to convert
896 * @fh_len: length of the file handle in bytes
897 * @fh_type: type of file handle
898 * @get_inode: filesystem callback to retrieve inode
899 *
900 * This function decodes @fid as long as it has one of the well-known
901 * Linux filehandle types and calls @get_inode on it to retrieve the
902 * inode for the object specified in the file handle.
903 */
904struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
905 int fh_len, int fh_type, struct inode *(*get_inode)
906 (struct super_block *sb, u64 ino, u32 gen))
907{
908 struct inode *inode = NULL;
909
910 if (fh_len < 2)
911 return NULL;
912
913 switch (fh_type) {
914 case FILEID_INO32_GEN:
915 case FILEID_INO32_GEN_PARENT:
916 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
917 break;
918 }
919
920 return d_obtain_alias(inode);
921}
922EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
923
924/**
925 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
926 * @sb: filesystem to do the file handle conversion on
927 * @fid: file handle to convert
928 * @fh_len: length of the file handle in bytes
929 * @fh_type: type of file handle
930 * @get_inode: filesystem callback to retrieve inode
931 *
932 * This function decodes @fid as long as it has one of the well-known
933 * Linux filehandle types and calls @get_inode on it to retrieve the
934 * inode for the _parent_ object specified in the file handle if it
935 * is specified in the file handle, or NULL otherwise.
936 */
937struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
938 int fh_len, int fh_type, struct inode *(*get_inode)
939 (struct super_block *sb, u64 ino, u32 gen))
940{
941 struct inode *inode = NULL;
942
943 if (fh_len <= 2)
944 return NULL;
945
946 switch (fh_type) {
947 case FILEID_INO32_GEN_PARENT:
948 inode = get_inode(sb, fid->i32.parent_ino,
949 (fh_len > 3 ? fid->i32.parent_gen : 0));
950 break;
951 }
952
953 return d_obtain_alias(inode);
954}
955EXPORT_SYMBOL_GPL(generic_fh_to_parent);
956
957/**
958 * __generic_file_fsync - generic fsync implementation for simple filesystems
959 *
960 * @file: file to synchronize
961 * @start: start offset in bytes
962 * @end: end offset in bytes (inclusive)
963 * @datasync: only synchronize essential metadata if true
964 *
965 * This is a generic implementation of the fsync method for simple
966 * filesystems which track all non-inode metadata in the buffers list
967 * hanging off the address_space structure.
968 */
969int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
970 int datasync)
971{
972 struct inode *inode = file->f_mapping->host;
973 int err;
974 int ret;
975
976 err = filemap_write_and_wait_range(inode->i_mapping, start, end);
977 if (err)
978 return err;
979
980 inode_lock(inode);
981 ret = sync_mapping_buffers(inode->i_mapping);
982 if (!(inode->i_state & I_DIRTY_ALL))
983 goto out;
984 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
985 goto out;
986
987 err = sync_inode_metadata(inode, 1);
988 if (ret == 0)
989 ret = err;
990
991out:
992 inode_unlock(inode);
993 return ret;
994}
995EXPORT_SYMBOL(__generic_file_fsync);
996
997/**
998 * generic_file_fsync - generic fsync implementation for simple filesystems
999 * with flush
1000 * @file: file to synchronize
1001 * @start: start offset in bytes
1002 * @end: end offset in bytes (inclusive)
1003 * @datasync: only synchronize essential metadata if true
1004 *
1005 */
1006
1007int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1008 int datasync)
1009{
1010 struct inode *inode = file->f_mapping->host;
1011 int err;
1012
1013 err = __generic_file_fsync(file, start, end, datasync);
1014 if (err)
1015 return err;
1016 return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
1017}
1018EXPORT_SYMBOL(generic_file_fsync);
1019
1020/**
1021 * generic_check_addressable - Check addressability of file system
1022 * @blocksize_bits: log of file system block size
1023 * @num_blocks: number of blocks in file system
1024 *
1025 * Determine whether a file system with @num_blocks blocks (and a
1026 * block size of 2**@blocksize_bits) is addressable by the sector_t
1027 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1028 */
1029int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1030{
1031 u64 last_fs_block = num_blocks - 1;
1032 u64 last_fs_page =
1033 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1034
1035 if (unlikely(num_blocks == 0))
1036 return 0;
1037
1038 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1039 return -EINVAL;
1040
1041 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1042 (last_fs_page > (pgoff_t)(~0ULL))) {
1043 return -EFBIG;
1044 }
1045 return 0;
1046}
1047EXPORT_SYMBOL(generic_check_addressable);
1048
1049/*
1050 * No-op implementation of ->fsync for in-memory filesystems.
1051 */
1052int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1053{
1054 return 0;
1055}
1056EXPORT_SYMBOL(noop_fsync);
1057
1058/* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1059void kfree_link(void *p)
1060{
1061 kfree(p);
1062}
1063EXPORT_SYMBOL(kfree_link);
1064
1065/*
1066 * nop .set_page_dirty method so that people can use .page_mkwrite on
1067 * anon inodes.
1068 */
1069static int anon_set_page_dirty(struct page *page)
1070{
1071 return 0;
1072};
1073
1074/*
1075 * A single inode exists for all anon_inode files. Contrary to pipes,
1076 * anon_inode inodes have no associated per-instance data, so we need
1077 * only allocate one of them.
1078 */
1079struct inode *alloc_anon_inode(struct super_block *s)
1080{
1081 static const struct address_space_operations anon_aops = {
1082 .set_page_dirty = anon_set_page_dirty,
1083 };
1084 struct inode *inode = new_inode_pseudo(s);
1085
1086 if (!inode)
1087 return ERR_PTR(-ENOMEM);
1088
1089 inode->i_ino = get_next_ino();
1090 inode->i_mapping->a_ops = &anon_aops;
1091
1092 /*
1093 * Mark the inode dirty from the very beginning,
1094 * that way it will never be moved to the dirty
1095 * list because mark_inode_dirty() will think
1096 * that it already _is_ on the dirty list.
1097 */
1098 inode->i_state = I_DIRTY;
1099 inode->i_mode = S_IRUSR | S_IWUSR;
1100 inode->i_uid = current_fsuid();
1101 inode->i_gid = current_fsgid();
1102 inode->i_flags |= S_PRIVATE;
1103 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1104 return inode;
1105}
1106EXPORT_SYMBOL(alloc_anon_inode);
1107
1108/**
1109 * simple_nosetlease - generic helper for prohibiting leases
1110 * @filp: file pointer
1111 * @arg: type of lease to obtain
1112 * @flp: new lease supplied for insertion
1113 * @priv: private data for lm_setup operation
1114 *
1115 * Generic helper for filesystems that do not wish to allow leases to be set.
1116 * All arguments are ignored and it just returns -EINVAL.
1117 */
1118int
1119simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1120 void **priv)
1121{
1122 return -EINVAL;
1123}
1124EXPORT_SYMBOL(simple_nosetlease);
1125
1126const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1127 struct delayed_call *done)
1128{
1129 return inode->i_link;
1130}
1131EXPORT_SYMBOL(simple_get_link);
1132
1133const struct inode_operations simple_symlink_inode_operations = {
1134 .get_link = simple_get_link,
1135};
1136EXPORT_SYMBOL(simple_symlink_inode_operations);
1137
1138/*
1139 * Operations for a permanently empty directory.
1140 */
1141static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1142{
1143 return ERR_PTR(-ENOENT);
1144}
1145
1146static int empty_dir_getattr(struct vfsmount *mnt, struct dentry *dentry,
1147 struct kstat *stat)
1148{
1149 struct inode *inode = d_inode(dentry);
1150 generic_fillattr(inode, stat);
1151 return 0;
1152}
1153
1154static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1155{
1156 return -EPERM;
1157}
1158
1159static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1160{
1161 return -EOPNOTSUPP;
1162}
1163
1164static const struct inode_operations empty_dir_inode_operations = {
1165 .lookup = empty_dir_lookup,
1166 .permission = generic_permission,
1167 .setattr = empty_dir_setattr,
1168 .getattr = empty_dir_getattr,
1169 .listxattr = empty_dir_listxattr,
1170};
1171
1172static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1173{
1174 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1175 return generic_file_llseek_size(file, offset, whence, 2, 2);
1176}
1177
1178static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1179{
1180 dir_emit_dots(file, ctx);
1181 return 0;
1182}
1183
1184static const struct file_operations empty_dir_operations = {
1185 .llseek = empty_dir_llseek,
1186 .read = generic_read_dir,
1187 .iterate_shared = empty_dir_readdir,
1188 .fsync = noop_fsync,
1189};
1190
1191
1192void make_empty_dir_inode(struct inode *inode)
1193{
1194 set_nlink(inode, 2);
1195 inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1196 inode->i_uid = GLOBAL_ROOT_UID;
1197 inode->i_gid = GLOBAL_ROOT_GID;
1198 inode->i_rdev = 0;
1199 inode->i_size = 0;
1200 inode->i_blkbits = PAGE_SHIFT;
1201 inode->i_blocks = 0;
1202
1203 inode->i_op = &empty_dir_inode_operations;
1204 inode->i_opflags &= ~IOP_XATTR;
1205 inode->i_fop = &empty_dir_operations;
1206}
1207
1208bool is_empty_dir_inode(struct inode *inode)
1209{
1210 return (inode->i_fop == &empty_dir_operations) &&
1211 (inode->i_op == &empty_dir_inode_operations);
1212}