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#include <linux/fsnotify.h>
23
24#include <linux/uaccess.h>
25
26#include "internal.h"
27
28int simple_getattr(const struct path *path, struct kstat *stat,
29 u32 request_mask, unsigned int query_flags)
30{
31 struct inode *inode = d_inode(path->dentry);
32 generic_fillattr(inode, stat);
33 stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
34 return 0;
35}
36EXPORT_SYMBOL(simple_getattr);
37
38int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
39{
40 buf->f_type = dentry->d_sb->s_magic;
41 buf->f_bsize = PAGE_SIZE;
42 buf->f_namelen = NAME_MAX;
43 return 0;
44}
45EXPORT_SYMBOL(simple_statfs);
46
47/*
48 * Retaining negative dentries for an in-memory filesystem just wastes
49 * memory and lookup time: arrange for them to be deleted immediately.
50 */
51int always_delete_dentry(const struct dentry *dentry)
52{
53 return 1;
54}
55EXPORT_SYMBOL(always_delete_dentry);
56
57const struct dentry_operations simple_dentry_operations = {
58 .d_delete = always_delete_dentry,
59};
60EXPORT_SYMBOL(simple_dentry_operations);
61
62/*
63 * Lookup the data. This is trivial - if the dentry didn't already
64 * exist, we know it is negative. Set d_op to delete negative dentries.
65 */
66struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
67{
68 if (dentry->d_name.len > NAME_MAX)
69 return ERR_PTR(-ENAMETOOLONG);
70 if (!dentry->d_sb->s_d_op)
71 d_set_d_op(dentry, &simple_dentry_operations);
72 d_add(dentry, NULL);
73 return NULL;
74}
75EXPORT_SYMBOL(simple_lookup);
76
77int dcache_dir_open(struct inode *inode, struct file *file)
78{
79 file->private_data = d_alloc_cursor(file->f_path.dentry);
80
81 return file->private_data ? 0 : -ENOMEM;
82}
83EXPORT_SYMBOL(dcache_dir_open);
84
85int dcache_dir_close(struct inode *inode, struct file *file)
86{
87 dput(file->private_data);
88 return 0;
89}
90EXPORT_SYMBOL(dcache_dir_close);
91
92/* parent is locked at least shared */
93/*
94 * Returns an element of siblings' list.
95 * We are looking for <count>th positive after <p>; if
96 * found, dentry is grabbed and returned to caller.
97 * If no such element exists, NULL is returned.
98 */
99static struct dentry *scan_positives(struct dentry *cursor,
100 struct list_head *p,
101 loff_t count,
102 struct dentry *last)
103{
104 struct dentry *dentry = cursor->d_parent, *found = NULL;
105
106 spin_lock(&dentry->d_lock);
107 while ((p = p->next) != &dentry->d_subdirs) {
108 struct dentry *d = list_entry(p, struct dentry, d_child);
109 // we must at least skip cursors, to avoid livelocks
110 if (d->d_flags & DCACHE_DENTRY_CURSOR)
111 continue;
112 if (simple_positive(d) && !--count) {
113 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
114 if (simple_positive(d))
115 found = dget_dlock(d);
116 spin_unlock(&d->d_lock);
117 if (likely(found))
118 break;
119 count = 1;
120 }
121 if (need_resched()) {
122 list_move(&cursor->d_child, p);
123 p = &cursor->d_child;
124 spin_unlock(&dentry->d_lock);
125 cond_resched();
126 spin_lock(&dentry->d_lock);
127 }
128 }
129 spin_unlock(&dentry->d_lock);
130 dput(last);
131 return found;
132}
133
134loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
135{
136 struct dentry *dentry = file->f_path.dentry;
137 switch (whence) {
138 case 1:
139 offset += file->f_pos;
140 fallthrough;
141 case 0:
142 if (offset >= 0)
143 break;
144 fallthrough;
145 default:
146 return -EINVAL;
147 }
148 if (offset != file->f_pos) {
149 struct dentry *cursor = file->private_data;
150 struct dentry *to = NULL;
151
152 inode_lock_shared(dentry->d_inode);
153
154 if (offset > 2)
155 to = scan_positives(cursor, &dentry->d_subdirs,
156 offset - 2, NULL);
157 spin_lock(&dentry->d_lock);
158 if (to)
159 list_move(&cursor->d_child, &to->d_child);
160 else
161 list_del_init(&cursor->d_child);
162 spin_unlock(&dentry->d_lock);
163 dput(to);
164
165 file->f_pos = offset;
166
167 inode_unlock_shared(dentry->d_inode);
168 }
169 return offset;
170}
171EXPORT_SYMBOL(dcache_dir_lseek);
172
173/* Relationship between i_mode and the DT_xxx types */
174static inline unsigned char dt_type(struct inode *inode)
175{
176 return (inode->i_mode >> 12) & 15;
177}
178
179/*
180 * Directory is locked and all positive dentries in it are safe, since
181 * for ramfs-type trees they can't go away without unlink() or rmdir(),
182 * both impossible due to the lock on directory.
183 */
184
185int dcache_readdir(struct file *file, struct dir_context *ctx)
186{
187 struct dentry *dentry = file->f_path.dentry;
188 struct dentry *cursor = file->private_data;
189 struct list_head *anchor = &dentry->d_subdirs;
190 struct dentry *next = NULL;
191 struct list_head *p;
192
193 if (!dir_emit_dots(file, ctx))
194 return 0;
195
196 if (ctx->pos == 2)
197 p = anchor;
198 else if (!list_empty(&cursor->d_child))
199 p = &cursor->d_child;
200 else
201 return 0;
202
203 while ((next = scan_positives(cursor, p, 1, next)) != NULL) {
204 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
205 d_inode(next)->i_ino, dt_type(d_inode(next))))
206 break;
207 ctx->pos++;
208 p = &next->d_child;
209 }
210 spin_lock(&dentry->d_lock);
211 if (next)
212 list_move_tail(&cursor->d_child, &next->d_child);
213 else
214 list_del_init(&cursor->d_child);
215 spin_unlock(&dentry->d_lock);
216 dput(next);
217
218 return 0;
219}
220EXPORT_SYMBOL(dcache_readdir);
221
222ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
223{
224 return -EISDIR;
225}
226EXPORT_SYMBOL(generic_read_dir);
227
228const struct file_operations simple_dir_operations = {
229 .open = dcache_dir_open,
230 .release = dcache_dir_close,
231 .llseek = dcache_dir_lseek,
232 .read = generic_read_dir,
233 .iterate_shared = dcache_readdir,
234 .fsync = noop_fsync,
235};
236EXPORT_SYMBOL(simple_dir_operations);
237
238const struct inode_operations simple_dir_inode_operations = {
239 .lookup = simple_lookup,
240};
241EXPORT_SYMBOL(simple_dir_inode_operations);
242
243static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev)
244{
245 struct dentry *child = NULL;
246 struct list_head *p = prev ? &prev->d_child : &parent->d_subdirs;
247
248 spin_lock(&parent->d_lock);
249 while ((p = p->next) != &parent->d_subdirs) {
250 struct dentry *d = container_of(p, struct dentry, d_child);
251 if (simple_positive(d)) {
252 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
253 if (simple_positive(d))
254 child = dget_dlock(d);
255 spin_unlock(&d->d_lock);
256 if (likely(child))
257 break;
258 }
259 }
260 spin_unlock(&parent->d_lock);
261 dput(prev);
262 return child;
263}
264
265void simple_recursive_removal(struct dentry *dentry,
266 void (*callback)(struct dentry *))
267{
268 struct dentry *this = dget(dentry);
269 while (true) {
270 struct dentry *victim = NULL, *child;
271 struct inode *inode = this->d_inode;
272
273 inode_lock(inode);
274 if (d_is_dir(this))
275 inode->i_flags |= S_DEAD;
276 while ((child = find_next_child(this, victim)) == NULL) {
277 // kill and ascend
278 // update metadata while it's still locked
279 inode->i_ctime = current_time(inode);
280 clear_nlink(inode);
281 inode_unlock(inode);
282 victim = this;
283 this = this->d_parent;
284 inode = this->d_inode;
285 inode_lock(inode);
286 if (simple_positive(victim)) {
287 d_invalidate(victim); // avoid lost mounts
288 if (d_is_dir(victim))
289 fsnotify_rmdir(inode, victim);
290 else
291 fsnotify_unlink(inode, victim);
292 if (callback)
293 callback(victim);
294 dput(victim); // unpin it
295 }
296 if (victim == dentry) {
297 inode->i_ctime = inode->i_mtime =
298 current_time(inode);
299 if (d_is_dir(dentry))
300 drop_nlink(inode);
301 inode_unlock(inode);
302 dput(dentry);
303 return;
304 }
305 }
306 inode_unlock(inode);
307 this = child;
308 }
309}
310EXPORT_SYMBOL(simple_recursive_removal);
311
312static const struct super_operations simple_super_operations = {
313 .statfs = simple_statfs,
314};
315
316static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc)
317{
318 struct pseudo_fs_context *ctx = fc->fs_private;
319 struct inode *root;
320
321 s->s_maxbytes = MAX_LFS_FILESIZE;
322 s->s_blocksize = PAGE_SIZE;
323 s->s_blocksize_bits = PAGE_SHIFT;
324 s->s_magic = ctx->magic;
325 s->s_op = ctx->ops ?: &simple_super_operations;
326 s->s_xattr = ctx->xattr;
327 s->s_time_gran = 1;
328 root = new_inode(s);
329 if (!root)
330 return -ENOMEM;
331
332 /*
333 * since this is the first inode, make it number 1. New inodes created
334 * after this must take care not to collide with it (by passing
335 * max_reserved of 1 to iunique).
336 */
337 root->i_ino = 1;
338 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
339 root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
340 s->s_root = d_make_root(root);
341 if (!s->s_root)
342 return -ENOMEM;
343 s->s_d_op = ctx->dops;
344 return 0;
345}
346
347static int pseudo_fs_get_tree(struct fs_context *fc)
348{
349 return get_tree_nodev(fc, pseudo_fs_fill_super);
350}
351
352static void pseudo_fs_free(struct fs_context *fc)
353{
354 kfree(fc->fs_private);
355}
356
357static const struct fs_context_operations pseudo_fs_context_ops = {
358 .free = pseudo_fs_free,
359 .get_tree = pseudo_fs_get_tree,
360};
361
362/*
363 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
364 * will never be mountable)
365 */
366struct pseudo_fs_context *init_pseudo(struct fs_context *fc,
367 unsigned long magic)
368{
369 struct pseudo_fs_context *ctx;
370
371 ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL);
372 if (likely(ctx)) {
373 ctx->magic = magic;
374 fc->fs_private = ctx;
375 fc->ops = &pseudo_fs_context_ops;
376 fc->sb_flags |= SB_NOUSER;
377 fc->global = true;
378 }
379 return ctx;
380}
381EXPORT_SYMBOL(init_pseudo);
382
383int simple_open(struct inode *inode, struct file *file)
384{
385 if (inode->i_private)
386 file->private_data = inode->i_private;
387 return 0;
388}
389EXPORT_SYMBOL(simple_open);
390
391int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
392{
393 struct inode *inode = d_inode(old_dentry);
394
395 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
396 inc_nlink(inode);
397 ihold(inode);
398 dget(dentry);
399 d_instantiate(dentry, inode);
400 return 0;
401}
402EXPORT_SYMBOL(simple_link);
403
404int simple_empty(struct dentry *dentry)
405{
406 struct dentry *child;
407 int ret = 0;
408
409 spin_lock(&dentry->d_lock);
410 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
411 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
412 if (simple_positive(child)) {
413 spin_unlock(&child->d_lock);
414 goto out;
415 }
416 spin_unlock(&child->d_lock);
417 }
418 ret = 1;
419out:
420 spin_unlock(&dentry->d_lock);
421 return ret;
422}
423EXPORT_SYMBOL(simple_empty);
424
425int simple_unlink(struct inode *dir, struct dentry *dentry)
426{
427 struct inode *inode = d_inode(dentry);
428
429 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
430 drop_nlink(inode);
431 dput(dentry);
432 return 0;
433}
434EXPORT_SYMBOL(simple_unlink);
435
436int simple_rmdir(struct inode *dir, struct dentry *dentry)
437{
438 if (!simple_empty(dentry))
439 return -ENOTEMPTY;
440
441 drop_nlink(d_inode(dentry));
442 simple_unlink(dir, dentry);
443 drop_nlink(dir);
444 return 0;
445}
446EXPORT_SYMBOL(simple_rmdir);
447
448int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
449 struct inode *new_dir, struct dentry *new_dentry,
450 unsigned int flags)
451{
452 struct inode *inode = d_inode(old_dentry);
453 int they_are_dirs = d_is_dir(old_dentry);
454
455 if (flags & ~RENAME_NOREPLACE)
456 return -EINVAL;
457
458 if (!simple_empty(new_dentry))
459 return -ENOTEMPTY;
460
461 if (d_really_is_positive(new_dentry)) {
462 simple_unlink(new_dir, new_dentry);
463 if (they_are_dirs) {
464 drop_nlink(d_inode(new_dentry));
465 drop_nlink(old_dir);
466 }
467 } else if (they_are_dirs) {
468 drop_nlink(old_dir);
469 inc_nlink(new_dir);
470 }
471
472 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
473 new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
474
475 return 0;
476}
477EXPORT_SYMBOL(simple_rename);
478
479/**
480 * simple_setattr - setattr for simple filesystem
481 * @dentry: dentry
482 * @iattr: iattr structure
483 *
484 * Returns 0 on success, -error on failure.
485 *
486 * simple_setattr is a simple ->setattr implementation without a proper
487 * implementation of size changes.
488 *
489 * It can either be used for in-memory filesystems or special files
490 * on simple regular filesystems. Anything that needs to change on-disk
491 * or wire state on size changes needs its own setattr method.
492 */
493int simple_setattr(struct dentry *dentry, struct iattr *iattr)
494{
495 struct inode *inode = d_inode(dentry);
496 int error;
497
498 error = setattr_prepare(dentry, iattr);
499 if (error)
500 return error;
501
502 if (iattr->ia_valid & ATTR_SIZE)
503 truncate_setsize(inode, iattr->ia_size);
504 setattr_copy(inode, iattr);
505 mark_inode_dirty(inode);
506 return 0;
507}
508EXPORT_SYMBOL(simple_setattr);
509
510int simple_readpage(struct file *file, struct page *page)
511{
512 clear_highpage(page);
513 flush_dcache_page(page);
514 SetPageUptodate(page);
515 unlock_page(page);
516 return 0;
517}
518EXPORT_SYMBOL(simple_readpage);
519
520int simple_write_begin(struct file *file, struct address_space *mapping,
521 loff_t pos, unsigned len, unsigned flags,
522 struct page **pagep, void **fsdata)
523{
524 struct page *page;
525 pgoff_t index;
526
527 index = pos >> PAGE_SHIFT;
528
529 page = grab_cache_page_write_begin(mapping, index, flags);
530 if (!page)
531 return -ENOMEM;
532
533 *pagep = page;
534
535 if (!PageUptodate(page) && (len != PAGE_SIZE)) {
536 unsigned from = pos & (PAGE_SIZE - 1);
537
538 zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
539 }
540 return 0;
541}
542EXPORT_SYMBOL(simple_write_begin);
543
544/**
545 * simple_write_end - .write_end helper for non-block-device FSes
546 * @file: See .write_end of address_space_operations
547 * @mapping: "
548 * @pos: "
549 * @len: "
550 * @copied: "
551 * @page: "
552 * @fsdata: "
553 *
554 * simple_write_end does the minimum needed for updating a page after writing is
555 * done. It has the same API signature as the .write_end of
556 * address_space_operations vector. So it can just be set onto .write_end for
557 * FSes that don't need any other processing. i_mutex is assumed to be held.
558 * Block based filesystems should use generic_write_end().
559 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
560 * is not called, so a filesystem that actually does store data in .write_inode
561 * should extend on what's done here with a call to mark_inode_dirty() in the
562 * case that i_size has changed.
563 *
564 * Use *ONLY* with simple_readpage()
565 */
566int simple_write_end(struct file *file, struct address_space *mapping,
567 loff_t pos, unsigned len, unsigned copied,
568 struct page *page, void *fsdata)
569{
570 struct inode *inode = page->mapping->host;
571 loff_t last_pos = pos + copied;
572
573 /* zero the stale part of the page if we did a short copy */
574 if (!PageUptodate(page)) {
575 if (copied < len) {
576 unsigned from = pos & (PAGE_SIZE - 1);
577
578 zero_user(page, from + copied, len - copied);
579 }
580 SetPageUptodate(page);
581 }
582 /*
583 * No need to use i_size_read() here, the i_size
584 * cannot change under us because we hold the i_mutex.
585 */
586 if (last_pos > inode->i_size)
587 i_size_write(inode, last_pos);
588
589 set_page_dirty(page);
590 unlock_page(page);
591 put_page(page);
592
593 return copied;
594}
595EXPORT_SYMBOL(simple_write_end);
596
597/*
598 * the inodes created here are not hashed. If you use iunique to generate
599 * unique inode values later for this filesystem, then you must take care
600 * to pass it an appropriate max_reserved value to avoid collisions.
601 */
602int simple_fill_super(struct super_block *s, unsigned long magic,
603 const struct tree_descr *files)
604{
605 struct inode *inode;
606 struct dentry *root;
607 struct dentry *dentry;
608 int i;
609
610 s->s_blocksize = PAGE_SIZE;
611 s->s_blocksize_bits = PAGE_SHIFT;
612 s->s_magic = magic;
613 s->s_op = &simple_super_operations;
614 s->s_time_gran = 1;
615
616 inode = new_inode(s);
617 if (!inode)
618 return -ENOMEM;
619 /*
620 * because the root inode is 1, the files array must not contain an
621 * entry at index 1
622 */
623 inode->i_ino = 1;
624 inode->i_mode = S_IFDIR | 0755;
625 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
626 inode->i_op = &simple_dir_inode_operations;
627 inode->i_fop = &simple_dir_operations;
628 set_nlink(inode, 2);
629 root = d_make_root(inode);
630 if (!root)
631 return -ENOMEM;
632 for (i = 0; !files->name || files->name[0]; i++, files++) {
633 if (!files->name)
634 continue;
635
636 /* warn if it tries to conflict with the root inode */
637 if (unlikely(i == 1))
638 printk(KERN_WARNING "%s: %s passed in a files array"
639 "with an index of 1!\n", __func__,
640 s->s_type->name);
641
642 dentry = d_alloc_name(root, files->name);
643 if (!dentry)
644 goto out;
645 inode = new_inode(s);
646 if (!inode) {
647 dput(dentry);
648 goto out;
649 }
650 inode->i_mode = S_IFREG | files->mode;
651 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
652 inode->i_fop = files->ops;
653 inode->i_ino = i;
654 d_add(dentry, inode);
655 }
656 s->s_root = root;
657 return 0;
658out:
659 d_genocide(root);
660 shrink_dcache_parent(root);
661 dput(root);
662 return -ENOMEM;
663}
664EXPORT_SYMBOL(simple_fill_super);
665
666static DEFINE_SPINLOCK(pin_fs_lock);
667
668int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
669{
670 struct vfsmount *mnt = NULL;
671 spin_lock(&pin_fs_lock);
672 if (unlikely(!*mount)) {
673 spin_unlock(&pin_fs_lock);
674 mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
675 if (IS_ERR(mnt))
676 return PTR_ERR(mnt);
677 spin_lock(&pin_fs_lock);
678 if (!*mount)
679 *mount = mnt;
680 }
681 mntget(*mount);
682 ++*count;
683 spin_unlock(&pin_fs_lock);
684 mntput(mnt);
685 return 0;
686}
687EXPORT_SYMBOL(simple_pin_fs);
688
689void simple_release_fs(struct vfsmount **mount, int *count)
690{
691 struct vfsmount *mnt;
692 spin_lock(&pin_fs_lock);
693 mnt = *mount;
694 if (!--*count)
695 *mount = NULL;
696 spin_unlock(&pin_fs_lock);
697 mntput(mnt);
698}
699EXPORT_SYMBOL(simple_release_fs);
700
701/**
702 * simple_read_from_buffer - copy data from the buffer to user space
703 * @to: the user 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 simple_read_from_buffer() function reads up to @count bytes from the
710 * buffer @from at offset @ppos into the user 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 simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
716 const void *from, size_t available)
717{
718 loff_t pos = *ppos;
719 size_t ret;
720
721 if (pos < 0)
722 return -EINVAL;
723 if (pos >= available || !count)
724 return 0;
725 if (count > available - pos)
726 count = available - pos;
727 ret = copy_to_user(to, from + pos, count);
728 if (ret == count)
729 return -EFAULT;
730 count -= ret;
731 *ppos = pos + count;
732 return count;
733}
734EXPORT_SYMBOL(simple_read_from_buffer);
735
736/**
737 * simple_write_to_buffer - copy data from user space to the buffer
738 * @to: the buffer to write to
739 * @available: the size of the buffer
740 * @ppos: the current position in the buffer
741 * @from: the user space buffer to read from
742 * @count: the maximum number of bytes to read
743 *
744 * The simple_write_to_buffer() function reads up to @count bytes from the user
745 * space address starting at @from into the buffer @to at offset @ppos.
746 *
747 * On success, the number of bytes written is returned and the offset @ppos is
748 * advanced by this number, or negative value is returned on error.
749 **/
750ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
751 const void __user *from, size_t count)
752{
753 loff_t pos = *ppos;
754 size_t res;
755
756 if (pos < 0)
757 return -EINVAL;
758 if (pos >= available || !count)
759 return 0;
760 if (count > available - pos)
761 count = available - pos;
762 res = copy_from_user(to + pos, from, count);
763 if (res == count)
764 return -EFAULT;
765 count -= res;
766 *ppos = pos + count;
767 return count;
768}
769EXPORT_SYMBOL(simple_write_to_buffer);
770
771/**
772 * memory_read_from_buffer - copy data from the buffer
773 * @to: the kernel space buffer to read to
774 * @count: the maximum number of bytes to read
775 * @ppos: the current position in the buffer
776 * @from: the buffer to read from
777 * @available: the size of the buffer
778 *
779 * The memory_read_from_buffer() function reads up to @count bytes from the
780 * buffer @from at offset @ppos into the kernel space address starting at @to.
781 *
782 * On success, the number of bytes read is returned and the offset @ppos is
783 * advanced by this number, or negative value is returned on error.
784 **/
785ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
786 const void *from, size_t available)
787{
788 loff_t pos = *ppos;
789
790 if (pos < 0)
791 return -EINVAL;
792 if (pos >= available)
793 return 0;
794 if (count > available - pos)
795 count = available - pos;
796 memcpy(to, from + pos, count);
797 *ppos = pos + count;
798
799 return count;
800}
801EXPORT_SYMBOL(memory_read_from_buffer);
802
803/*
804 * Transaction based IO.
805 * The file expects a single write which triggers the transaction, and then
806 * possibly a read which collects the result - which is stored in a
807 * file-local buffer.
808 */
809
810void simple_transaction_set(struct file *file, size_t n)
811{
812 struct simple_transaction_argresp *ar = file->private_data;
813
814 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
815
816 /*
817 * The barrier ensures that ar->size will really remain zero until
818 * ar->data is ready for reading.
819 */
820 smp_mb();
821 ar->size = n;
822}
823EXPORT_SYMBOL(simple_transaction_set);
824
825char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
826{
827 struct simple_transaction_argresp *ar;
828 static DEFINE_SPINLOCK(simple_transaction_lock);
829
830 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
831 return ERR_PTR(-EFBIG);
832
833 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
834 if (!ar)
835 return ERR_PTR(-ENOMEM);
836
837 spin_lock(&simple_transaction_lock);
838
839 /* only one write allowed per open */
840 if (file->private_data) {
841 spin_unlock(&simple_transaction_lock);
842 free_page((unsigned long)ar);
843 return ERR_PTR(-EBUSY);
844 }
845
846 file->private_data = ar;
847
848 spin_unlock(&simple_transaction_lock);
849
850 if (copy_from_user(ar->data, buf, size))
851 return ERR_PTR(-EFAULT);
852
853 return ar->data;
854}
855EXPORT_SYMBOL(simple_transaction_get);
856
857ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
858{
859 struct simple_transaction_argresp *ar = file->private_data;
860
861 if (!ar)
862 return 0;
863 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
864}
865EXPORT_SYMBOL(simple_transaction_read);
866
867int simple_transaction_release(struct inode *inode, struct file *file)
868{
869 free_page((unsigned long)file->private_data);
870 return 0;
871}
872EXPORT_SYMBOL(simple_transaction_release);
873
874/* Simple attribute files */
875
876struct simple_attr {
877 int (*get)(void *, u64 *);
878 int (*set)(void *, u64);
879 char get_buf[24]; /* enough to store a u64 and "\n\0" */
880 char set_buf[24];
881 void *data;
882 const char *fmt; /* format for read operation */
883 struct mutex mutex; /* protects access to these buffers */
884};
885
886/* simple_attr_open is called by an actual attribute open file operation
887 * to set the attribute specific access operations. */
888int simple_attr_open(struct inode *inode, struct file *file,
889 int (*get)(void *, u64 *), int (*set)(void *, u64),
890 const char *fmt)
891{
892 struct simple_attr *attr;
893
894 attr = kzalloc(sizeof(*attr), GFP_KERNEL);
895 if (!attr)
896 return -ENOMEM;
897
898 attr->get = get;
899 attr->set = set;
900 attr->data = inode->i_private;
901 attr->fmt = fmt;
902 mutex_init(&attr->mutex);
903
904 file->private_data = attr;
905
906 return nonseekable_open(inode, file);
907}
908EXPORT_SYMBOL_GPL(simple_attr_open);
909
910int simple_attr_release(struct inode *inode, struct file *file)
911{
912 kfree(file->private_data);
913 return 0;
914}
915EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
916
917/* read from the buffer that is filled with the get function */
918ssize_t simple_attr_read(struct file *file, char __user *buf,
919 size_t len, loff_t *ppos)
920{
921 struct simple_attr *attr;
922 size_t size;
923 ssize_t ret;
924
925 attr = file->private_data;
926
927 if (!attr->get)
928 return -EACCES;
929
930 ret = mutex_lock_interruptible(&attr->mutex);
931 if (ret)
932 return ret;
933
934 if (*ppos && attr->get_buf[0]) {
935 /* continued read */
936 size = strlen(attr->get_buf);
937 } else {
938 /* first read */
939 u64 val;
940 ret = attr->get(attr->data, &val);
941 if (ret)
942 goto out;
943
944 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
945 attr->fmt, (unsigned long long)val);
946 }
947
948 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
949out:
950 mutex_unlock(&attr->mutex);
951 return ret;
952}
953EXPORT_SYMBOL_GPL(simple_attr_read);
954
955/* interpret the buffer as a number to call the set function with */
956ssize_t simple_attr_write(struct file *file, const char __user *buf,
957 size_t len, loff_t *ppos)
958{
959 struct simple_attr *attr;
960 u64 val;
961 size_t size;
962 ssize_t ret;
963
964 attr = file->private_data;
965 if (!attr->set)
966 return -EACCES;
967
968 ret = mutex_lock_interruptible(&attr->mutex);
969 if (ret)
970 return ret;
971
972 ret = -EFAULT;
973 size = min(sizeof(attr->set_buf) - 1, len);
974 if (copy_from_user(attr->set_buf, buf, size))
975 goto out;
976
977 attr->set_buf[size] = '\0';
978 val = simple_strtoll(attr->set_buf, NULL, 0);
979 ret = attr->set(attr->data, val);
980 if (ret == 0)
981 ret = len; /* on success, claim we got the whole input */
982out:
983 mutex_unlock(&attr->mutex);
984 return ret;
985}
986EXPORT_SYMBOL_GPL(simple_attr_write);
987
988/**
989 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
990 * @sb: filesystem to do the file handle conversion on
991 * @fid: file handle to convert
992 * @fh_len: length of the file handle in bytes
993 * @fh_type: type of file handle
994 * @get_inode: filesystem callback to retrieve inode
995 *
996 * This function decodes @fid as long as it has one of the well-known
997 * Linux filehandle types and calls @get_inode on it to retrieve the
998 * inode for the object specified in the file handle.
999 */
1000struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
1001 int fh_len, int fh_type, struct inode *(*get_inode)
1002 (struct super_block *sb, u64 ino, u32 gen))
1003{
1004 struct inode *inode = NULL;
1005
1006 if (fh_len < 2)
1007 return NULL;
1008
1009 switch (fh_type) {
1010 case FILEID_INO32_GEN:
1011 case FILEID_INO32_GEN_PARENT:
1012 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
1013 break;
1014 }
1015
1016 return d_obtain_alias(inode);
1017}
1018EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
1019
1020/**
1021 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
1022 * @sb: filesystem to do the file handle conversion on
1023 * @fid: file handle to convert
1024 * @fh_len: length of the file handle in bytes
1025 * @fh_type: type of file handle
1026 * @get_inode: filesystem callback to retrieve inode
1027 *
1028 * This function decodes @fid as long as it has one of the well-known
1029 * Linux filehandle types and calls @get_inode on it to retrieve the
1030 * inode for the _parent_ object specified in the file handle if it
1031 * is specified in the file handle, or NULL otherwise.
1032 */
1033struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
1034 int fh_len, int fh_type, struct inode *(*get_inode)
1035 (struct super_block *sb, u64 ino, u32 gen))
1036{
1037 struct inode *inode = NULL;
1038
1039 if (fh_len <= 2)
1040 return NULL;
1041
1042 switch (fh_type) {
1043 case FILEID_INO32_GEN_PARENT:
1044 inode = get_inode(sb, fid->i32.parent_ino,
1045 (fh_len > 3 ? fid->i32.parent_gen : 0));
1046 break;
1047 }
1048
1049 return d_obtain_alias(inode);
1050}
1051EXPORT_SYMBOL_GPL(generic_fh_to_parent);
1052
1053/**
1054 * __generic_file_fsync - generic fsync implementation for simple filesystems
1055 *
1056 * @file: file to synchronize
1057 * @start: start offset in bytes
1058 * @end: end offset in bytes (inclusive)
1059 * @datasync: only synchronize essential metadata if true
1060 *
1061 * This is a generic implementation of the fsync method for simple
1062 * filesystems which track all non-inode metadata in the buffers list
1063 * hanging off the address_space structure.
1064 */
1065int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
1066 int datasync)
1067{
1068 struct inode *inode = file->f_mapping->host;
1069 int err;
1070 int ret;
1071
1072 err = file_write_and_wait_range(file, start, end);
1073 if (err)
1074 return err;
1075
1076 inode_lock(inode);
1077 ret = sync_mapping_buffers(inode->i_mapping);
1078 if (!(inode->i_state & I_DIRTY_ALL))
1079 goto out;
1080 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
1081 goto out;
1082
1083 err = sync_inode_metadata(inode, 1);
1084 if (ret == 0)
1085 ret = err;
1086
1087out:
1088 inode_unlock(inode);
1089 /* check and advance again to catch errors after syncing out buffers */
1090 err = file_check_and_advance_wb_err(file);
1091 if (ret == 0)
1092 ret = err;
1093 return ret;
1094}
1095EXPORT_SYMBOL(__generic_file_fsync);
1096
1097/**
1098 * generic_file_fsync - generic fsync implementation for simple filesystems
1099 * with flush
1100 * @file: file to synchronize
1101 * @start: start offset in bytes
1102 * @end: end offset in bytes (inclusive)
1103 * @datasync: only synchronize essential metadata if true
1104 *
1105 */
1106
1107int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1108 int datasync)
1109{
1110 struct inode *inode = file->f_mapping->host;
1111 int err;
1112
1113 err = __generic_file_fsync(file, start, end, datasync);
1114 if (err)
1115 return err;
1116 return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL);
1117}
1118EXPORT_SYMBOL(generic_file_fsync);
1119
1120/**
1121 * generic_check_addressable - Check addressability of file system
1122 * @blocksize_bits: log of file system block size
1123 * @num_blocks: number of blocks in file system
1124 *
1125 * Determine whether a file system with @num_blocks blocks (and a
1126 * block size of 2**@blocksize_bits) is addressable by the sector_t
1127 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1128 */
1129int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1130{
1131 u64 last_fs_block = num_blocks - 1;
1132 u64 last_fs_page =
1133 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1134
1135 if (unlikely(num_blocks == 0))
1136 return 0;
1137
1138 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1139 return -EINVAL;
1140
1141 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1142 (last_fs_page > (pgoff_t)(~0ULL))) {
1143 return -EFBIG;
1144 }
1145 return 0;
1146}
1147EXPORT_SYMBOL(generic_check_addressable);
1148
1149/*
1150 * No-op implementation of ->fsync for in-memory filesystems.
1151 */
1152int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1153{
1154 return 0;
1155}
1156EXPORT_SYMBOL(noop_fsync);
1157
1158int noop_set_page_dirty(struct page *page)
1159{
1160 /*
1161 * Unlike __set_page_dirty_no_writeback that handles dirty page
1162 * tracking in the page object, dax does all dirty tracking in
1163 * the inode address_space in response to mkwrite faults. In the
1164 * dax case we only need to worry about potentially dirty CPU
1165 * caches, not dirty page cache pages to write back.
1166 *
1167 * This callback is defined to prevent fallback to
1168 * __set_page_dirty_buffers() in set_page_dirty().
1169 */
1170 return 0;
1171}
1172EXPORT_SYMBOL_GPL(noop_set_page_dirty);
1173
1174void noop_invalidatepage(struct page *page, unsigned int offset,
1175 unsigned int length)
1176{
1177 /*
1178 * There is no page cache to invalidate in the dax case, however
1179 * we need this callback defined to prevent falling back to
1180 * block_invalidatepage() in do_invalidatepage().
1181 */
1182}
1183EXPORT_SYMBOL_GPL(noop_invalidatepage);
1184
1185ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1186{
1187 /*
1188 * iomap based filesystems support direct I/O without need for
1189 * this callback. However, it still needs to be set in
1190 * inode->a_ops so that open/fcntl know that direct I/O is
1191 * generally supported.
1192 */
1193 return -EINVAL;
1194}
1195EXPORT_SYMBOL_GPL(noop_direct_IO);
1196
1197/* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1198void kfree_link(void *p)
1199{
1200 kfree(p);
1201}
1202EXPORT_SYMBOL(kfree_link);
1203
1204/*
1205 * nop .set_page_dirty method so that people can use .page_mkwrite on
1206 * anon inodes.
1207 */
1208static int anon_set_page_dirty(struct page *page)
1209{
1210 return 0;
1211};
1212
1213/*
1214 * A single inode exists for all anon_inode files. Contrary to pipes,
1215 * anon_inode inodes have no associated per-instance data, so we need
1216 * only allocate one of them.
1217 */
1218struct inode *alloc_anon_inode(struct super_block *s)
1219{
1220 static const struct address_space_operations anon_aops = {
1221 .set_page_dirty = anon_set_page_dirty,
1222 };
1223 struct inode *inode = new_inode_pseudo(s);
1224
1225 if (!inode)
1226 return ERR_PTR(-ENOMEM);
1227
1228 inode->i_ino = get_next_ino();
1229 inode->i_mapping->a_ops = &anon_aops;
1230
1231 /*
1232 * Mark the inode dirty from the very beginning,
1233 * that way it will never be moved to the dirty
1234 * list because mark_inode_dirty() will think
1235 * that it already _is_ on the dirty list.
1236 */
1237 inode->i_state = I_DIRTY;
1238 inode->i_mode = S_IRUSR | S_IWUSR;
1239 inode->i_uid = current_fsuid();
1240 inode->i_gid = current_fsgid();
1241 inode->i_flags |= S_PRIVATE;
1242 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1243 return inode;
1244}
1245EXPORT_SYMBOL(alloc_anon_inode);
1246
1247/**
1248 * simple_nosetlease - generic helper for prohibiting leases
1249 * @filp: file pointer
1250 * @arg: type of lease to obtain
1251 * @flp: new lease supplied for insertion
1252 * @priv: private data for lm_setup operation
1253 *
1254 * Generic helper for filesystems that do not wish to allow leases to be set.
1255 * All arguments are ignored and it just returns -EINVAL.
1256 */
1257int
1258simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1259 void **priv)
1260{
1261 return -EINVAL;
1262}
1263EXPORT_SYMBOL(simple_nosetlease);
1264
1265/**
1266 * simple_get_link - generic helper to get the target of "fast" symlinks
1267 * @dentry: not used here
1268 * @inode: the symlink inode
1269 * @done: not used here
1270 *
1271 * Generic helper for filesystems to use for symlink inodes where a pointer to
1272 * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1273 * since as an optimization the path lookup code uses any non-NULL ->i_link
1274 * directly, without calling ->get_link(). But ->get_link() still must be set,
1275 * to mark the inode_operations as being for a symlink.
1276 *
1277 * Return: the symlink target
1278 */
1279const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1280 struct delayed_call *done)
1281{
1282 return inode->i_link;
1283}
1284EXPORT_SYMBOL(simple_get_link);
1285
1286const struct inode_operations simple_symlink_inode_operations = {
1287 .get_link = simple_get_link,
1288};
1289EXPORT_SYMBOL(simple_symlink_inode_operations);
1290
1291/*
1292 * Operations for a permanently empty directory.
1293 */
1294static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1295{
1296 return ERR_PTR(-ENOENT);
1297}
1298
1299static int empty_dir_getattr(const struct path *path, struct kstat *stat,
1300 u32 request_mask, unsigned int query_flags)
1301{
1302 struct inode *inode = d_inode(path->dentry);
1303 generic_fillattr(inode, stat);
1304 return 0;
1305}
1306
1307static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1308{
1309 return -EPERM;
1310}
1311
1312static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1313{
1314 return -EOPNOTSUPP;
1315}
1316
1317static const struct inode_operations empty_dir_inode_operations = {
1318 .lookup = empty_dir_lookup,
1319 .permission = generic_permission,
1320 .setattr = empty_dir_setattr,
1321 .getattr = empty_dir_getattr,
1322 .listxattr = empty_dir_listxattr,
1323};
1324
1325static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1326{
1327 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1328 return generic_file_llseek_size(file, offset, whence, 2, 2);
1329}
1330
1331static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1332{
1333 dir_emit_dots(file, ctx);
1334 return 0;
1335}
1336
1337static const struct file_operations empty_dir_operations = {
1338 .llseek = empty_dir_llseek,
1339 .read = generic_read_dir,
1340 .iterate_shared = empty_dir_readdir,
1341 .fsync = noop_fsync,
1342};
1343
1344
1345void make_empty_dir_inode(struct inode *inode)
1346{
1347 set_nlink(inode, 2);
1348 inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1349 inode->i_uid = GLOBAL_ROOT_UID;
1350 inode->i_gid = GLOBAL_ROOT_GID;
1351 inode->i_rdev = 0;
1352 inode->i_size = 0;
1353 inode->i_blkbits = PAGE_SHIFT;
1354 inode->i_blocks = 0;
1355
1356 inode->i_op = &empty_dir_inode_operations;
1357 inode->i_opflags &= ~IOP_XATTR;
1358 inode->i_fop = &empty_dir_operations;
1359}
1360
1361bool is_empty_dir_inode(struct inode *inode)
1362{
1363 return (inode->i_fop == &empty_dir_operations) &&
1364 (inode->i_op == &empty_dir_inode_operations);
1365}
1/*
2 * fs/libfs.c
3 * Library for filesystems writers.
4 */
5
6#include <linux/export.h>
7#include <linux/pagemap.h>
8#include <linux/slab.h>
9#include <linux/mount.h>
10#include <linux/vfs.h>
11#include <linux/quotaops.h>
12#include <linux/mutex.h>
13#include <linux/namei.h>
14#include <linux/exportfs.h>
15#include <linux/writeback.h>
16#include <linux/buffer_head.h> /* sync_mapping_buffers */
17
18#include <asm/uaccess.h>
19
20#include "internal.h"
21
22static inline int simple_positive(struct dentry *dentry)
23{
24 return dentry->d_inode && !d_unhashed(dentry);
25}
26
27int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
28 struct kstat *stat)
29{
30 struct inode *inode = dentry->d_inode;
31 generic_fillattr(inode, stat);
32 stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_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_CACHE_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 static struct qstr cursor_name = QSTR_INIT(".", 1);
79
80 file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
81
82 return file->private_data ? 0 : -ENOMEM;
83}
84EXPORT_SYMBOL(dcache_dir_open);
85
86int dcache_dir_close(struct inode *inode, struct file *file)
87{
88 dput(file->private_data);
89 return 0;
90}
91EXPORT_SYMBOL(dcache_dir_close);
92
93loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
94{
95 struct dentry *dentry = file->f_path.dentry;
96 mutex_lock(&dentry->d_inode->i_mutex);
97 switch (whence) {
98 case 1:
99 offset += file->f_pos;
100 case 0:
101 if (offset >= 0)
102 break;
103 default:
104 mutex_unlock(&dentry->d_inode->i_mutex);
105 return -EINVAL;
106 }
107 if (offset != file->f_pos) {
108 file->f_pos = offset;
109 if (file->f_pos >= 2) {
110 struct list_head *p;
111 struct dentry *cursor = file->private_data;
112 loff_t n = file->f_pos - 2;
113
114 spin_lock(&dentry->d_lock);
115 /* d_lock not required for cursor */
116 list_del(&cursor->d_u.d_child);
117 p = dentry->d_subdirs.next;
118 while (n && p != &dentry->d_subdirs) {
119 struct dentry *next;
120 next = list_entry(p, struct dentry, d_u.d_child);
121 spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
122 if (simple_positive(next))
123 n--;
124 spin_unlock(&next->d_lock);
125 p = p->next;
126 }
127 list_add_tail(&cursor->d_u.d_child, p);
128 spin_unlock(&dentry->d_lock);
129 }
130 }
131 mutex_unlock(&dentry->d_inode->i_mutex);
132 return offset;
133}
134EXPORT_SYMBOL(dcache_dir_lseek);
135
136/* Relationship between i_mode and the DT_xxx types */
137static inline unsigned char dt_type(struct inode *inode)
138{
139 return (inode->i_mode >> 12) & 15;
140}
141
142/*
143 * Directory is locked and all positive dentries in it are safe, since
144 * for ramfs-type trees they can't go away without unlink() or rmdir(),
145 * both impossible due to the lock on directory.
146 */
147
148int dcache_readdir(struct file *file, struct dir_context *ctx)
149{
150 struct dentry *dentry = file->f_path.dentry;
151 struct dentry *cursor = file->private_data;
152 struct list_head *p, *q = &cursor->d_u.d_child;
153
154 if (!dir_emit_dots(file, ctx))
155 return 0;
156 spin_lock(&dentry->d_lock);
157 if (ctx->pos == 2)
158 list_move(q, &dentry->d_subdirs);
159
160 for (p = q->next; p != &dentry->d_subdirs; p = p->next) {
161 struct dentry *next = list_entry(p, struct dentry, d_u.d_child);
162 spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
163 if (!simple_positive(next)) {
164 spin_unlock(&next->d_lock);
165 continue;
166 }
167
168 spin_unlock(&next->d_lock);
169 spin_unlock(&dentry->d_lock);
170 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
171 next->d_inode->i_ino, dt_type(next->d_inode)))
172 return 0;
173 spin_lock(&dentry->d_lock);
174 spin_lock_nested(&next->d_lock, DENTRY_D_LOCK_NESTED);
175 /* next is still alive */
176 list_move(q, p);
177 spin_unlock(&next->d_lock);
178 p = q;
179 ctx->pos++;
180 }
181 spin_unlock(&dentry->d_lock);
182 return 0;
183}
184EXPORT_SYMBOL(dcache_readdir);
185
186ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
187{
188 return -EISDIR;
189}
190EXPORT_SYMBOL(generic_read_dir);
191
192const struct file_operations simple_dir_operations = {
193 .open = dcache_dir_open,
194 .release = dcache_dir_close,
195 .llseek = dcache_dir_lseek,
196 .read = generic_read_dir,
197 .iterate = dcache_readdir,
198 .fsync = noop_fsync,
199};
200EXPORT_SYMBOL(simple_dir_operations);
201
202const struct inode_operations simple_dir_inode_operations = {
203 .lookup = simple_lookup,
204};
205EXPORT_SYMBOL(simple_dir_inode_operations);
206
207static const struct super_operations simple_super_operations = {
208 .statfs = simple_statfs,
209};
210
211/*
212 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
213 * will never be mountable)
214 */
215struct dentry *mount_pseudo(struct file_system_type *fs_type, char *name,
216 const struct super_operations *ops,
217 const struct dentry_operations *dops, unsigned long magic)
218{
219 struct super_block *s;
220 struct dentry *dentry;
221 struct inode *root;
222 struct qstr d_name = QSTR_INIT(name, strlen(name));
223
224 s = sget(fs_type, NULL, set_anon_super, MS_NOUSER, NULL);
225 if (IS_ERR(s))
226 return ERR_CAST(s);
227
228 s->s_maxbytes = MAX_LFS_FILESIZE;
229 s->s_blocksize = PAGE_SIZE;
230 s->s_blocksize_bits = PAGE_SHIFT;
231 s->s_magic = magic;
232 s->s_op = ops ? ops : &simple_super_operations;
233 s->s_time_gran = 1;
234 root = new_inode(s);
235 if (!root)
236 goto Enomem;
237 /*
238 * since this is the first inode, make it number 1. New inodes created
239 * after this must take care not to collide with it (by passing
240 * max_reserved of 1 to iunique).
241 */
242 root->i_ino = 1;
243 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
244 root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
245 dentry = __d_alloc(s, &d_name);
246 if (!dentry) {
247 iput(root);
248 goto Enomem;
249 }
250 d_instantiate(dentry, root);
251 s->s_root = dentry;
252 s->s_d_op = dops;
253 s->s_flags |= MS_ACTIVE;
254 return dget(s->s_root);
255
256Enomem:
257 deactivate_locked_super(s);
258 return ERR_PTR(-ENOMEM);
259}
260EXPORT_SYMBOL(mount_pseudo);
261
262int simple_open(struct inode *inode, struct file *file)
263{
264 if (inode->i_private)
265 file->private_data = inode->i_private;
266 return 0;
267}
268EXPORT_SYMBOL(simple_open);
269
270int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
271{
272 struct inode *inode = old_dentry->d_inode;
273
274 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
275 inc_nlink(inode);
276 ihold(inode);
277 dget(dentry);
278 d_instantiate(dentry, inode);
279 return 0;
280}
281EXPORT_SYMBOL(simple_link);
282
283int simple_empty(struct dentry *dentry)
284{
285 struct dentry *child;
286 int ret = 0;
287
288 spin_lock(&dentry->d_lock);
289 list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child) {
290 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
291 if (simple_positive(child)) {
292 spin_unlock(&child->d_lock);
293 goto out;
294 }
295 spin_unlock(&child->d_lock);
296 }
297 ret = 1;
298out:
299 spin_unlock(&dentry->d_lock);
300 return ret;
301}
302EXPORT_SYMBOL(simple_empty);
303
304int simple_unlink(struct inode *dir, struct dentry *dentry)
305{
306 struct inode *inode = dentry->d_inode;
307
308 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
309 drop_nlink(inode);
310 dput(dentry);
311 return 0;
312}
313EXPORT_SYMBOL(simple_unlink);
314
315int simple_rmdir(struct inode *dir, struct dentry *dentry)
316{
317 if (!simple_empty(dentry))
318 return -ENOTEMPTY;
319
320 drop_nlink(dentry->d_inode);
321 simple_unlink(dir, dentry);
322 drop_nlink(dir);
323 return 0;
324}
325EXPORT_SYMBOL(simple_rmdir);
326
327int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
328 struct inode *new_dir, struct dentry *new_dentry)
329{
330 struct inode *inode = old_dentry->d_inode;
331 int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
332
333 if (!simple_empty(new_dentry))
334 return -ENOTEMPTY;
335
336 if (new_dentry->d_inode) {
337 simple_unlink(new_dir, new_dentry);
338 if (they_are_dirs) {
339 drop_nlink(new_dentry->d_inode);
340 drop_nlink(old_dir);
341 }
342 } else if (they_are_dirs) {
343 drop_nlink(old_dir);
344 inc_nlink(new_dir);
345 }
346
347 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
348 new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
349
350 return 0;
351}
352EXPORT_SYMBOL(simple_rename);
353
354/**
355 * simple_setattr - setattr for simple filesystem
356 * @dentry: dentry
357 * @iattr: iattr structure
358 *
359 * Returns 0 on success, -error on failure.
360 *
361 * simple_setattr is a simple ->setattr implementation without a proper
362 * implementation of size changes.
363 *
364 * It can either be used for in-memory filesystems or special files
365 * on simple regular filesystems. Anything that needs to change on-disk
366 * or wire state on size changes needs its own setattr method.
367 */
368int simple_setattr(struct dentry *dentry, struct iattr *iattr)
369{
370 struct inode *inode = dentry->d_inode;
371 int error;
372
373 error = inode_change_ok(inode, iattr);
374 if (error)
375 return error;
376
377 if (iattr->ia_valid & ATTR_SIZE)
378 truncate_setsize(inode, iattr->ia_size);
379 setattr_copy(inode, iattr);
380 mark_inode_dirty(inode);
381 return 0;
382}
383EXPORT_SYMBOL(simple_setattr);
384
385int simple_readpage(struct file *file, struct page *page)
386{
387 clear_highpage(page);
388 flush_dcache_page(page);
389 SetPageUptodate(page);
390 unlock_page(page);
391 return 0;
392}
393EXPORT_SYMBOL(simple_readpage);
394
395int simple_write_begin(struct file *file, struct address_space *mapping,
396 loff_t pos, unsigned len, unsigned flags,
397 struct page **pagep, void **fsdata)
398{
399 struct page *page;
400 pgoff_t index;
401
402 index = pos >> PAGE_CACHE_SHIFT;
403
404 page = grab_cache_page_write_begin(mapping, index, flags);
405 if (!page)
406 return -ENOMEM;
407
408 *pagep = page;
409
410 if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
411 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
412
413 zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
414 }
415 return 0;
416}
417EXPORT_SYMBOL(simple_write_begin);
418
419/**
420 * simple_write_end - .write_end helper for non-block-device FSes
421 * @available: See .write_end of address_space_operations
422 * @file: "
423 * @mapping: "
424 * @pos: "
425 * @len: "
426 * @copied: "
427 * @page: "
428 * @fsdata: "
429 *
430 * simple_write_end does the minimum needed for updating a page after writing is
431 * done. It has the same API signature as the .write_end of
432 * address_space_operations vector. So it can just be set onto .write_end for
433 * FSes that don't need any other processing. i_mutex is assumed to be held.
434 * Block based filesystems should use generic_write_end().
435 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
436 * is not called, so a filesystem that actually does store data in .write_inode
437 * should extend on what's done here with a call to mark_inode_dirty() in the
438 * case that i_size has changed.
439 */
440int simple_write_end(struct file *file, struct address_space *mapping,
441 loff_t pos, unsigned len, unsigned copied,
442 struct page *page, void *fsdata)
443{
444 struct inode *inode = page->mapping->host;
445 loff_t last_pos = pos + copied;
446
447 /* zero the stale part of the page if we did a short copy */
448 if (copied < len) {
449 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
450
451 zero_user(page, from + copied, len - copied);
452 }
453
454 if (!PageUptodate(page))
455 SetPageUptodate(page);
456 /*
457 * No need to use i_size_read() here, the i_size
458 * cannot change under us because we hold the i_mutex.
459 */
460 if (last_pos > inode->i_size)
461 i_size_write(inode, last_pos);
462
463 set_page_dirty(page);
464 unlock_page(page);
465 page_cache_release(page);
466
467 return copied;
468}
469EXPORT_SYMBOL(simple_write_end);
470
471/*
472 * the inodes created here are not hashed. If you use iunique to generate
473 * unique inode values later for this filesystem, then you must take care
474 * to pass it an appropriate max_reserved value to avoid collisions.
475 */
476int simple_fill_super(struct super_block *s, unsigned long magic,
477 struct tree_descr *files)
478{
479 struct inode *inode;
480 struct dentry *root;
481 struct dentry *dentry;
482 int i;
483
484 s->s_blocksize = PAGE_CACHE_SIZE;
485 s->s_blocksize_bits = PAGE_CACHE_SHIFT;
486 s->s_magic = magic;
487 s->s_op = &simple_super_operations;
488 s->s_time_gran = 1;
489
490 inode = new_inode(s);
491 if (!inode)
492 return -ENOMEM;
493 /*
494 * because the root inode is 1, the files array must not contain an
495 * entry at index 1
496 */
497 inode->i_ino = 1;
498 inode->i_mode = S_IFDIR | 0755;
499 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
500 inode->i_op = &simple_dir_inode_operations;
501 inode->i_fop = &simple_dir_operations;
502 set_nlink(inode, 2);
503 root = d_make_root(inode);
504 if (!root)
505 return -ENOMEM;
506 for (i = 0; !files->name || files->name[0]; i++, files++) {
507 if (!files->name)
508 continue;
509
510 /* warn if it tries to conflict with the root inode */
511 if (unlikely(i == 1))
512 printk(KERN_WARNING "%s: %s passed in a files array"
513 "with an index of 1!\n", __func__,
514 s->s_type->name);
515
516 dentry = d_alloc_name(root, files->name);
517 if (!dentry)
518 goto out;
519 inode = new_inode(s);
520 if (!inode) {
521 dput(dentry);
522 goto out;
523 }
524 inode->i_mode = S_IFREG | files->mode;
525 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
526 inode->i_fop = files->ops;
527 inode->i_ino = i;
528 d_add(dentry, inode);
529 }
530 s->s_root = root;
531 return 0;
532out:
533 d_genocide(root);
534 shrink_dcache_parent(root);
535 dput(root);
536 return -ENOMEM;
537}
538EXPORT_SYMBOL(simple_fill_super);
539
540static DEFINE_SPINLOCK(pin_fs_lock);
541
542int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
543{
544 struct vfsmount *mnt = NULL;
545 spin_lock(&pin_fs_lock);
546 if (unlikely(!*mount)) {
547 spin_unlock(&pin_fs_lock);
548 mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, NULL);
549 if (IS_ERR(mnt))
550 return PTR_ERR(mnt);
551 spin_lock(&pin_fs_lock);
552 if (!*mount)
553 *mount = mnt;
554 }
555 mntget(*mount);
556 ++*count;
557 spin_unlock(&pin_fs_lock);
558 mntput(mnt);
559 return 0;
560}
561EXPORT_SYMBOL(simple_pin_fs);
562
563void simple_release_fs(struct vfsmount **mount, int *count)
564{
565 struct vfsmount *mnt;
566 spin_lock(&pin_fs_lock);
567 mnt = *mount;
568 if (!--*count)
569 *mount = NULL;
570 spin_unlock(&pin_fs_lock);
571 mntput(mnt);
572}
573EXPORT_SYMBOL(simple_release_fs);
574
575/**
576 * simple_read_from_buffer - copy data from the buffer to user space
577 * @to: the user space buffer to read to
578 * @count: the maximum number of bytes to read
579 * @ppos: the current position in the buffer
580 * @from: the buffer to read from
581 * @available: the size of the buffer
582 *
583 * The simple_read_from_buffer() function reads up to @count bytes from the
584 * buffer @from at offset @ppos into the user space address starting at @to.
585 *
586 * On success, the number of bytes read is returned and the offset @ppos is
587 * advanced by this number, or negative value is returned on error.
588 **/
589ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
590 const void *from, size_t available)
591{
592 loff_t pos = *ppos;
593 size_t ret;
594
595 if (pos < 0)
596 return -EINVAL;
597 if (pos >= available || !count)
598 return 0;
599 if (count > available - pos)
600 count = available - pos;
601 ret = copy_to_user(to, from + pos, count);
602 if (ret == count)
603 return -EFAULT;
604 count -= ret;
605 *ppos = pos + count;
606 return count;
607}
608EXPORT_SYMBOL(simple_read_from_buffer);
609
610/**
611 * simple_write_to_buffer - copy data from user space to the buffer
612 * @to: the buffer to write to
613 * @available: the size of the buffer
614 * @ppos: the current position in the buffer
615 * @from: the user space buffer to read from
616 * @count: the maximum number of bytes to read
617 *
618 * The simple_write_to_buffer() function reads up to @count bytes from the user
619 * space address starting at @from into the buffer @to at offset @ppos.
620 *
621 * On success, the number of bytes written is returned and the offset @ppos is
622 * advanced by this number, or negative value is returned on error.
623 **/
624ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
625 const void __user *from, size_t count)
626{
627 loff_t pos = *ppos;
628 size_t res;
629
630 if (pos < 0)
631 return -EINVAL;
632 if (pos >= available || !count)
633 return 0;
634 if (count > available - pos)
635 count = available - pos;
636 res = copy_from_user(to + pos, from, count);
637 if (res == count)
638 return -EFAULT;
639 count -= res;
640 *ppos = pos + count;
641 return count;
642}
643EXPORT_SYMBOL(simple_write_to_buffer);
644
645/**
646 * memory_read_from_buffer - copy data from the buffer
647 * @to: the kernel space buffer to read to
648 * @count: the maximum number of bytes to read
649 * @ppos: the current position in the buffer
650 * @from: the buffer to read from
651 * @available: the size of the buffer
652 *
653 * The memory_read_from_buffer() function reads up to @count bytes from the
654 * buffer @from at offset @ppos into the kernel space address starting at @to.
655 *
656 * On success, the number of bytes read is returned and the offset @ppos is
657 * advanced by this number, or negative value is returned on error.
658 **/
659ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
660 const void *from, size_t available)
661{
662 loff_t pos = *ppos;
663
664 if (pos < 0)
665 return -EINVAL;
666 if (pos >= available)
667 return 0;
668 if (count > available - pos)
669 count = available - pos;
670 memcpy(to, from + pos, count);
671 *ppos = pos + count;
672
673 return count;
674}
675EXPORT_SYMBOL(memory_read_from_buffer);
676
677/*
678 * Transaction based IO.
679 * The file expects a single write which triggers the transaction, and then
680 * possibly a read which collects the result - which is stored in a
681 * file-local buffer.
682 */
683
684void simple_transaction_set(struct file *file, size_t n)
685{
686 struct simple_transaction_argresp *ar = file->private_data;
687
688 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
689
690 /*
691 * The barrier ensures that ar->size will really remain zero until
692 * ar->data is ready for reading.
693 */
694 smp_mb();
695 ar->size = n;
696}
697EXPORT_SYMBOL(simple_transaction_set);
698
699char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
700{
701 struct simple_transaction_argresp *ar;
702 static DEFINE_SPINLOCK(simple_transaction_lock);
703
704 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
705 return ERR_PTR(-EFBIG);
706
707 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
708 if (!ar)
709 return ERR_PTR(-ENOMEM);
710
711 spin_lock(&simple_transaction_lock);
712
713 /* only one write allowed per open */
714 if (file->private_data) {
715 spin_unlock(&simple_transaction_lock);
716 free_page((unsigned long)ar);
717 return ERR_PTR(-EBUSY);
718 }
719
720 file->private_data = ar;
721
722 spin_unlock(&simple_transaction_lock);
723
724 if (copy_from_user(ar->data, buf, size))
725 return ERR_PTR(-EFAULT);
726
727 return ar->data;
728}
729EXPORT_SYMBOL(simple_transaction_get);
730
731ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
732{
733 struct simple_transaction_argresp *ar = file->private_data;
734
735 if (!ar)
736 return 0;
737 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
738}
739EXPORT_SYMBOL(simple_transaction_read);
740
741int simple_transaction_release(struct inode *inode, struct file *file)
742{
743 free_page((unsigned long)file->private_data);
744 return 0;
745}
746EXPORT_SYMBOL(simple_transaction_release);
747
748/* Simple attribute files */
749
750struct simple_attr {
751 int (*get)(void *, u64 *);
752 int (*set)(void *, u64);
753 char get_buf[24]; /* enough to store a u64 and "\n\0" */
754 char set_buf[24];
755 void *data;
756 const char *fmt; /* format for read operation */
757 struct mutex mutex; /* protects access to these buffers */
758};
759
760/* simple_attr_open is called by an actual attribute open file operation
761 * to set the attribute specific access operations. */
762int simple_attr_open(struct inode *inode, struct file *file,
763 int (*get)(void *, u64 *), int (*set)(void *, u64),
764 const char *fmt)
765{
766 struct simple_attr *attr;
767
768 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
769 if (!attr)
770 return -ENOMEM;
771
772 attr->get = get;
773 attr->set = set;
774 attr->data = inode->i_private;
775 attr->fmt = fmt;
776 mutex_init(&attr->mutex);
777
778 file->private_data = attr;
779
780 return nonseekable_open(inode, file);
781}
782EXPORT_SYMBOL_GPL(simple_attr_open);
783
784int simple_attr_release(struct inode *inode, struct file *file)
785{
786 kfree(file->private_data);
787 return 0;
788}
789EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
790
791/* read from the buffer that is filled with the get function */
792ssize_t simple_attr_read(struct file *file, char __user *buf,
793 size_t len, loff_t *ppos)
794{
795 struct simple_attr *attr;
796 size_t size;
797 ssize_t ret;
798
799 attr = file->private_data;
800
801 if (!attr->get)
802 return -EACCES;
803
804 ret = mutex_lock_interruptible(&attr->mutex);
805 if (ret)
806 return ret;
807
808 if (*ppos) { /* continued read */
809 size = strlen(attr->get_buf);
810 } else { /* first read */
811 u64 val;
812 ret = attr->get(attr->data, &val);
813 if (ret)
814 goto out;
815
816 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
817 attr->fmt, (unsigned long long)val);
818 }
819
820 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
821out:
822 mutex_unlock(&attr->mutex);
823 return ret;
824}
825EXPORT_SYMBOL_GPL(simple_attr_read);
826
827/* interpret the buffer as a number to call the set function with */
828ssize_t simple_attr_write(struct file *file, const char __user *buf,
829 size_t len, loff_t *ppos)
830{
831 struct simple_attr *attr;
832 u64 val;
833 size_t size;
834 ssize_t ret;
835
836 attr = file->private_data;
837 if (!attr->set)
838 return -EACCES;
839
840 ret = mutex_lock_interruptible(&attr->mutex);
841 if (ret)
842 return ret;
843
844 ret = -EFAULT;
845 size = min(sizeof(attr->set_buf) - 1, len);
846 if (copy_from_user(attr->set_buf, buf, size))
847 goto out;
848
849 attr->set_buf[size] = '\0';
850 val = simple_strtoll(attr->set_buf, NULL, 0);
851 ret = attr->set(attr->data, val);
852 if (ret == 0)
853 ret = len; /* on success, claim we got the whole input */
854out:
855 mutex_unlock(&attr->mutex);
856 return ret;
857}
858EXPORT_SYMBOL_GPL(simple_attr_write);
859
860/**
861 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
862 * @sb: filesystem to do the file handle conversion on
863 * @fid: file handle to convert
864 * @fh_len: length of the file handle in bytes
865 * @fh_type: type of file handle
866 * @get_inode: filesystem callback to retrieve inode
867 *
868 * This function decodes @fid as long as it has one of the well-known
869 * Linux filehandle types and calls @get_inode on it to retrieve the
870 * inode for the object specified in the file handle.
871 */
872struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
873 int fh_len, int fh_type, struct inode *(*get_inode)
874 (struct super_block *sb, u64 ino, u32 gen))
875{
876 struct inode *inode = NULL;
877
878 if (fh_len < 2)
879 return NULL;
880
881 switch (fh_type) {
882 case FILEID_INO32_GEN:
883 case FILEID_INO32_GEN_PARENT:
884 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
885 break;
886 }
887
888 return d_obtain_alias(inode);
889}
890EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
891
892/**
893 * generic_fh_to_parent - generic helper for the fh_to_parent 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 _parent_ object specified in the file handle if it
903 * is specified in the file handle, or NULL otherwise.
904 */
905struct dentry *generic_fh_to_parent(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_PARENT:
916 inode = get_inode(sb, fid->i32.parent_ino,
917 (fh_len > 3 ? fid->i32.parent_gen : 0));
918 break;
919 }
920
921 return d_obtain_alias(inode);
922}
923EXPORT_SYMBOL_GPL(generic_fh_to_parent);
924
925/**
926 * generic_file_fsync - generic fsync implementation for simple filesystems
927 * @file: file to synchronize
928 * @datasync: only synchronize essential metadata if true
929 *
930 * This is a generic implementation of the fsync method for simple
931 * filesystems which track all non-inode metadata in the buffers list
932 * hanging off the address_space structure.
933 */
934int generic_file_fsync(struct file *file, loff_t start, loff_t end,
935 int datasync)
936{
937 struct inode *inode = file->f_mapping->host;
938 int err;
939 int ret;
940
941 err = filemap_write_and_wait_range(inode->i_mapping, start, end);
942 if (err)
943 return err;
944
945 mutex_lock(&inode->i_mutex);
946 ret = sync_mapping_buffers(inode->i_mapping);
947 if (!(inode->i_state & I_DIRTY))
948 goto out;
949 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
950 goto out;
951
952 err = sync_inode_metadata(inode, 1);
953 if (ret == 0)
954 ret = err;
955out:
956 mutex_unlock(&inode->i_mutex);
957 return ret;
958}
959EXPORT_SYMBOL(generic_file_fsync);
960
961/**
962 * generic_check_addressable - Check addressability of file system
963 * @blocksize_bits: log of file system block size
964 * @num_blocks: number of blocks in file system
965 *
966 * Determine whether a file system with @num_blocks blocks (and a
967 * block size of 2**@blocksize_bits) is addressable by the sector_t
968 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
969 */
970int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
971{
972 u64 last_fs_block = num_blocks - 1;
973 u64 last_fs_page =
974 last_fs_block >> (PAGE_CACHE_SHIFT - blocksize_bits);
975
976 if (unlikely(num_blocks == 0))
977 return 0;
978
979 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_CACHE_SHIFT))
980 return -EINVAL;
981
982 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
983 (last_fs_page > (pgoff_t)(~0ULL))) {
984 return -EFBIG;
985 }
986 return 0;
987}
988EXPORT_SYMBOL(generic_check_addressable);
989
990/*
991 * No-op implementation of ->fsync for in-memory filesystems.
992 */
993int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
994{
995 return 0;
996}
997EXPORT_SYMBOL(noop_fsync);
998
999void kfree_put_link(struct dentry *dentry, struct nameidata *nd,
1000 void *cookie)
1001{
1002 char *s = nd_get_link(nd);
1003 if (!IS_ERR(s))
1004 kfree(s);
1005}
1006EXPORT_SYMBOL(kfree_put_link);
1007
1008/*
1009 * nop .set_page_dirty method so that people can use .page_mkwrite on
1010 * anon inodes.
1011 */
1012static int anon_set_page_dirty(struct page *page)
1013{
1014 return 0;
1015};
1016
1017/*
1018 * A single inode exists for all anon_inode files. Contrary to pipes,
1019 * anon_inode inodes have no associated per-instance data, so we need
1020 * only allocate one of them.
1021 */
1022struct inode *alloc_anon_inode(struct super_block *s)
1023{
1024 static const struct address_space_operations anon_aops = {
1025 .set_page_dirty = anon_set_page_dirty,
1026 };
1027 struct inode *inode = new_inode_pseudo(s);
1028
1029 if (!inode)
1030 return ERR_PTR(-ENOMEM);
1031
1032 inode->i_ino = get_next_ino();
1033 inode->i_mapping->a_ops = &anon_aops;
1034
1035 /*
1036 * Mark the inode dirty from the very beginning,
1037 * that way it will never be moved to the dirty
1038 * list because mark_inode_dirty() will think
1039 * that it already _is_ on the dirty list.
1040 */
1041 inode->i_state = I_DIRTY;
1042 inode->i_mode = S_IRUSR | S_IWUSR;
1043 inode->i_uid = current_fsuid();
1044 inode->i_gid = current_fsgid();
1045 inode->i_flags |= S_PRIVATE;
1046 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1047 return inode;
1048}
1049EXPORT_SYMBOL(alloc_anon_inode);