mm: fix exec activate_mm vs TLB shootdown and lazy tlb switching race
[linux/fpc-iii.git] / fs / libfs.c
blobcb9310b091f5a359f2c7c9a6c398d280b5d690dd
1 /*
2 * fs/libfs.c
3 * Library for filesystems writers.
4 */
6 #include <linux/blkdev.h>
7 #include <linux/export.h>
8 #include <linux/pagemap.h>
9 #include <linux/slab.h>
10 #include <linux/cred.h>
11 #include <linux/mount.h>
12 #include <linux/vfs.h>
13 #include <linux/quotaops.h>
14 #include <linux/mutex.h>
15 #include <linux/namei.h>
16 #include <linux/exportfs.h>
17 #include <linux/writeback.h>
18 #include <linux/buffer_head.h> /* sync_mapping_buffers */
20 #include <linux/uaccess.h>
22 #include "internal.h"
24 int simple_getattr(const struct path *path, struct kstat *stat,
25 u32 request_mask, unsigned int query_flags)
27 struct inode *inode = d_inode(path->dentry);
28 generic_fillattr(inode, stat);
29 stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
30 return 0;
32 EXPORT_SYMBOL(simple_getattr);
34 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
36 buf->f_type = dentry->d_sb->s_magic;
37 buf->f_bsize = PAGE_SIZE;
38 buf->f_namelen = NAME_MAX;
39 return 0;
41 EXPORT_SYMBOL(simple_statfs);
44 * Retaining negative dentries for an in-memory filesystem just wastes
45 * memory and lookup time: arrange for them to be deleted immediately.
47 int always_delete_dentry(const struct dentry *dentry)
49 return 1;
51 EXPORT_SYMBOL(always_delete_dentry);
53 const struct dentry_operations simple_dentry_operations = {
54 .d_delete = always_delete_dentry,
56 EXPORT_SYMBOL(simple_dentry_operations);
59 * Lookup the data. This is trivial - if the dentry didn't already
60 * exist, we know it is negative. Set d_op to delete negative dentries.
62 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
64 if (dentry->d_name.len > NAME_MAX)
65 return ERR_PTR(-ENAMETOOLONG);
66 if (!dentry->d_sb->s_d_op)
67 d_set_d_op(dentry, &simple_dentry_operations);
68 d_add(dentry, NULL);
69 return NULL;
71 EXPORT_SYMBOL(simple_lookup);
73 int dcache_dir_open(struct inode *inode, struct file *file)
75 file->private_data = d_alloc_cursor(file->f_path.dentry);
77 return file->private_data ? 0 : -ENOMEM;
79 EXPORT_SYMBOL(dcache_dir_open);
81 int dcache_dir_close(struct inode *inode, struct file *file)
83 dput(file->private_data);
84 return 0;
86 EXPORT_SYMBOL(dcache_dir_close);
88 /* parent is locked at least shared */
90 * Returns an element of siblings' list.
91 * We are looking for <count>th positive after <p>; if
92 * found, dentry is grabbed and passed to caller via *<res>.
93 * If no such element exists, the anchor of list is returned
94 * and *<res> is set to NULL.
96 static struct list_head *scan_positives(struct dentry *cursor,
97 struct list_head *p,
98 loff_t count,
99 struct dentry **res)
101 struct dentry *dentry = cursor->d_parent, *found = NULL;
103 spin_lock(&dentry->d_lock);
104 while ((p = p->next) != &dentry->d_subdirs) {
105 struct dentry *d = list_entry(p, struct dentry, d_child);
106 // we must at least skip cursors, to avoid livelocks
107 if (d->d_flags & DCACHE_DENTRY_CURSOR)
108 continue;
109 if (simple_positive(d) && !--count) {
110 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
111 if (simple_positive(d))
112 found = dget_dlock(d);
113 spin_unlock(&d->d_lock);
114 if (likely(found))
115 break;
116 count = 1;
118 if (need_resched()) {
119 list_move(&cursor->d_child, p);
120 p = &cursor->d_child;
121 spin_unlock(&dentry->d_lock);
122 cond_resched();
123 spin_lock(&dentry->d_lock);
126 spin_unlock(&dentry->d_lock);
127 dput(*res);
128 *res = found;
129 return p;
132 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
134 struct dentry *dentry = file->f_path.dentry;
135 switch (whence) {
136 case 1:
137 offset += file->f_pos;
138 case 0:
139 if (offset >= 0)
140 break;
141 default:
142 return -EINVAL;
144 if (offset != file->f_pos) {
145 struct dentry *cursor = file->private_data;
146 struct dentry *to = NULL;
147 struct list_head *p;
149 file->f_pos = offset;
150 inode_lock_shared(dentry->d_inode);
152 if (file->f_pos > 2) {
153 p = scan_positives(cursor, &dentry->d_subdirs,
154 file->f_pos - 2, &to);
155 spin_lock(&dentry->d_lock);
156 list_move(&cursor->d_child, p);
157 spin_unlock(&dentry->d_lock);
158 } else {
159 spin_lock(&dentry->d_lock);
160 list_del_init(&cursor->d_child);
161 spin_unlock(&dentry->d_lock);
164 dput(to);
166 inode_unlock_shared(dentry->d_inode);
168 return offset;
170 EXPORT_SYMBOL(dcache_dir_lseek);
172 /* Relationship between i_mode and the DT_xxx types */
173 static inline unsigned char dt_type(struct inode *inode)
175 return (inode->i_mode >> 12) & 15;
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.
184 int dcache_readdir(struct file *file, struct dir_context *ctx)
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;
192 if (!dir_emit_dots(file, ctx))
193 return 0;
195 if (ctx->pos == 2)
196 p = anchor;
197 else
198 p = &cursor->d_child;
200 while ((p = scan_positives(cursor, p, 1, &next)) != anchor) {
201 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
202 d_inode(next)->i_ino, dt_type(d_inode(next))))
203 break;
204 ctx->pos++;
206 spin_lock(&dentry->d_lock);
207 list_move_tail(&cursor->d_child, p);
208 spin_unlock(&dentry->d_lock);
209 dput(next);
211 return 0;
213 EXPORT_SYMBOL(dcache_readdir);
215 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
217 return -EISDIR;
219 EXPORT_SYMBOL(generic_read_dir);
221 const struct file_operations simple_dir_operations = {
222 .open = dcache_dir_open,
223 .release = dcache_dir_close,
224 .llseek = dcache_dir_lseek,
225 .read = generic_read_dir,
226 .iterate_shared = dcache_readdir,
227 .fsync = noop_fsync,
229 EXPORT_SYMBOL(simple_dir_operations);
231 const struct inode_operations simple_dir_inode_operations = {
232 .lookup = simple_lookup,
234 EXPORT_SYMBOL(simple_dir_inode_operations);
236 static const struct super_operations simple_super_operations = {
237 .statfs = simple_statfs,
241 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
242 * will never be mountable)
244 struct dentry *mount_pseudo_xattr(struct file_system_type *fs_type, char *name,
245 const struct super_operations *ops, const struct xattr_handler **xattr,
246 const struct dentry_operations *dops, unsigned long magic)
248 struct super_block *s;
249 struct dentry *dentry;
250 struct inode *root;
251 struct qstr d_name = QSTR_INIT(name, strlen(name));
253 s = sget_userns(fs_type, NULL, set_anon_super, MS_KERNMOUNT|MS_NOUSER,
254 &init_user_ns, NULL);
255 if (IS_ERR(s))
256 return ERR_CAST(s);
258 s->s_maxbytes = MAX_LFS_FILESIZE;
259 s->s_blocksize = PAGE_SIZE;
260 s->s_blocksize_bits = PAGE_SHIFT;
261 s->s_magic = magic;
262 s->s_op = ops ? ops : &simple_super_operations;
263 s->s_xattr = xattr;
264 s->s_time_gran = 1;
265 root = new_inode(s);
266 if (!root)
267 goto Enomem;
269 * since this is the first inode, make it number 1. New inodes created
270 * after this must take care not to collide with it (by passing
271 * max_reserved of 1 to iunique).
273 root->i_ino = 1;
274 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
275 root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
276 dentry = __d_alloc(s, &d_name);
277 if (!dentry) {
278 iput(root);
279 goto Enomem;
281 d_instantiate(dentry, root);
282 s->s_root = dentry;
283 s->s_d_op = dops;
284 s->s_flags |= MS_ACTIVE;
285 return dget(s->s_root);
287 Enomem:
288 deactivate_locked_super(s);
289 return ERR_PTR(-ENOMEM);
291 EXPORT_SYMBOL(mount_pseudo_xattr);
293 int simple_open(struct inode *inode, struct file *file)
295 if (inode->i_private)
296 file->private_data = inode->i_private;
297 return 0;
299 EXPORT_SYMBOL(simple_open);
301 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
303 struct inode *inode = d_inode(old_dentry);
305 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
306 inc_nlink(inode);
307 ihold(inode);
308 dget(dentry);
309 d_instantiate(dentry, inode);
310 return 0;
312 EXPORT_SYMBOL(simple_link);
314 int simple_empty(struct dentry *dentry)
316 struct dentry *child;
317 int ret = 0;
319 spin_lock(&dentry->d_lock);
320 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
321 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
322 if (simple_positive(child)) {
323 spin_unlock(&child->d_lock);
324 goto out;
326 spin_unlock(&child->d_lock);
328 ret = 1;
329 out:
330 spin_unlock(&dentry->d_lock);
331 return ret;
333 EXPORT_SYMBOL(simple_empty);
335 int simple_unlink(struct inode *dir, struct dentry *dentry)
337 struct inode *inode = d_inode(dentry);
339 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
340 drop_nlink(inode);
341 dput(dentry);
342 return 0;
344 EXPORT_SYMBOL(simple_unlink);
346 int simple_rmdir(struct inode *dir, struct dentry *dentry)
348 if (!simple_empty(dentry))
349 return -ENOTEMPTY;
351 drop_nlink(d_inode(dentry));
352 simple_unlink(dir, dentry);
353 drop_nlink(dir);
354 return 0;
356 EXPORT_SYMBOL(simple_rmdir);
358 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
359 struct inode *new_dir, struct dentry *new_dentry,
360 unsigned int flags)
362 struct inode *inode = d_inode(old_dentry);
363 int they_are_dirs = d_is_dir(old_dentry);
365 if (flags & ~RENAME_NOREPLACE)
366 return -EINVAL;
368 if (!simple_empty(new_dentry))
369 return -ENOTEMPTY;
371 if (d_really_is_positive(new_dentry)) {
372 simple_unlink(new_dir, new_dentry);
373 if (they_are_dirs) {
374 drop_nlink(d_inode(new_dentry));
375 drop_nlink(old_dir);
377 } else if (they_are_dirs) {
378 drop_nlink(old_dir);
379 inc_nlink(new_dir);
382 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
383 new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
385 return 0;
387 EXPORT_SYMBOL(simple_rename);
390 * simple_setattr - setattr for simple filesystem
391 * @dentry: dentry
392 * @iattr: iattr structure
394 * Returns 0 on success, -error on failure.
396 * simple_setattr is a simple ->setattr implementation without a proper
397 * implementation of size changes.
399 * It can either be used for in-memory filesystems or special files
400 * on simple regular filesystems. Anything that needs to change on-disk
401 * or wire state on size changes needs its own setattr method.
403 int simple_setattr(struct dentry *dentry, struct iattr *iattr)
405 struct inode *inode = d_inode(dentry);
406 int error;
408 error = setattr_prepare(dentry, iattr);
409 if (error)
410 return error;
412 if (iattr->ia_valid & ATTR_SIZE)
413 truncate_setsize(inode, iattr->ia_size);
414 setattr_copy(inode, iattr);
415 mark_inode_dirty(inode);
416 return 0;
418 EXPORT_SYMBOL(simple_setattr);
420 int simple_readpage(struct file *file, struct page *page)
422 clear_highpage(page);
423 flush_dcache_page(page);
424 SetPageUptodate(page);
425 unlock_page(page);
426 return 0;
428 EXPORT_SYMBOL(simple_readpage);
430 int simple_write_begin(struct file *file, struct address_space *mapping,
431 loff_t pos, unsigned len, unsigned flags,
432 struct page **pagep, void **fsdata)
434 struct page *page;
435 pgoff_t index;
437 index = pos >> PAGE_SHIFT;
439 page = grab_cache_page_write_begin(mapping, index, flags);
440 if (!page)
441 return -ENOMEM;
443 *pagep = page;
445 if (!PageUptodate(page) && (len != PAGE_SIZE)) {
446 unsigned from = pos & (PAGE_SIZE - 1);
448 zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
450 return 0;
452 EXPORT_SYMBOL(simple_write_begin);
455 * simple_write_end - .write_end helper for non-block-device FSes
456 * @available: See .write_end of address_space_operations
457 * @file: "
458 * @mapping: "
459 * @pos: "
460 * @len: "
461 * @copied: "
462 * @page: "
463 * @fsdata: "
465 * simple_write_end does the minimum needed for updating a page after writing is
466 * done. It has the same API signature as the .write_end of
467 * address_space_operations vector. So it can just be set onto .write_end for
468 * FSes that don't need any other processing. i_mutex is assumed to be held.
469 * Block based filesystems should use generic_write_end().
470 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
471 * is not called, so a filesystem that actually does store data in .write_inode
472 * should extend on what's done here with a call to mark_inode_dirty() in the
473 * case that i_size has changed.
475 * Use *ONLY* with simple_readpage()
477 int simple_write_end(struct file *file, struct address_space *mapping,
478 loff_t pos, unsigned len, unsigned copied,
479 struct page *page, void *fsdata)
481 struct inode *inode = page->mapping->host;
482 loff_t last_pos = pos + copied;
484 /* zero the stale part of the page if we did a short copy */
485 if (!PageUptodate(page)) {
486 if (copied < len) {
487 unsigned from = pos & (PAGE_SIZE - 1);
489 zero_user(page, from + copied, len - copied);
491 SetPageUptodate(page);
494 * No need to use i_size_read() here, the i_size
495 * cannot change under us because we hold the i_mutex.
497 if (last_pos > inode->i_size)
498 i_size_write(inode, last_pos);
500 set_page_dirty(page);
501 unlock_page(page);
502 put_page(page);
504 return copied;
506 EXPORT_SYMBOL(simple_write_end);
509 * the inodes created here are not hashed. If you use iunique to generate
510 * unique inode values later for this filesystem, then you must take care
511 * to pass it an appropriate max_reserved value to avoid collisions.
513 int simple_fill_super(struct super_block *s, unsigned long magic,
514 const struct tree_descr *files)
516 struct inode *inode;
517 struct dentry *root;
518 struct dentry *dentry;
519 int i;
521 s->s_blocksize = PAGE_SIZE;
522 s->s_blocksize_bits = PAGE_SHIFT;
523 s->s_magic = magic;
524 s->s_op = &simple_super_operations;
525 s->s_time_gran = 1;
527 inode = new_inode(s);
528 if (!inode)
529 return -ENOMEM;
531 * because the root inode is 1, the files array must not contain an
532 * entry at index 1
534 inode->i_ino = 1;
535 inode->i_mode = S_IFDIR | 0755;
536 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
537 inode->i_op = &simple_dir_inode_operations;
538 inode->i_fop = &simple_dir_operations;
539 set_nlink(inode, 2);
540 root = d_make_root(inode);
541 if (!root)
542 return -ENOMEM;
543 for (i = 0; !files->name || files->name[0]; i++, files++) {
544 if (!files->name)
545 continue;
547 /* warn if it tries to conflict with the root inode */
548 if (unlikely(i == 1))
549 printk(KERN_WARNING "%s: %s passed in a files array"
550 "with an index of 1!\n", __func__,
551 s->s_type->name);
553 dentry = d_alloc_name(root, files->name);
554 if (!dentry)
555 goto out;
556 inode = new_inode(s);
557 if (!inode) {
558 dput(dentry);
559 goto out;
561 inode->i_mode = S_IFREG | files->mode;
562 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
563 inode->i_fop = files->ops;
564 inode->i_ino = i;
565 d_add(dentry, inode);
567 s->s_root = root;
568 return 0;
569 out:
570 d_genocide(root);
571 shrink_dcache_parent(root);
572 dput(root);
573 return -ENOMEM;
575 EXPORT_SYMBOL(simple_fill_super);
577 static DEFINE_SPINLOCK(pin_fs_lock);
579 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
581 struct vfsmount *mnt = NULL;
582 spin_lock(&pin_fs_lock);
583 if (unlikely(!*mount)) {
584 spin_unlock(&pin_fs_lock);
585 mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, NULL);
586 if (IS_ERR(mnt))
587 return PTR_ERR(mnt);
588 spin_lock(&pin_fs_lock);
589 if (!*mount)
590 *mount = mnt;
592 mntget(*mount);
593 ++*count;
594 spin_unlock(&pin_fs_lock);
595 mntput(mnt);
596 return 0;
598 EXPORT_SYMBOL(simple_pin_fs);
600 void simple_release_fs(struct vfsmount **mount, int *count)
602 struct vfsmount *mnt;
603 spin_lock(&pin_fs_lock);
604 mnt = *mount;
605 if (!--*count)
606 *mount = NULL;
607 spin_unlock(&pin_fs_lock);
608 mntput(mnt);
610 EXPORT_SYMBOL(simple_release_fs);
613 * simple_read_from_buffer - copy data from the buffer to user space
614 * @to: the user space buffer to read to
615 * @count: the maximum number of bytes to read
616 * @ppos: the current position in the buffer
617 * @from: the buffer to read from
618 * @available: the size of the buffer
620 * The simple_read_from_buffer() function reads up to @count bytes from the
621 * buffer @from at offset @ppos into the user space address starting at @to.
623 * On success, the number of bytes read is returned and the offset @ppos is
624 * advanced by this number, or negative value is returned on error.
626 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
627 const void *from, size_t available)
629 loff_t pos = *ppos;
630 size_t ret;
632 if (pos < 0)
633 return -EINVAL;
634 if (pos >= available || !count)
635 return 0;
636 if (count > available - pos)
637 count = available - pos;
638 ret = copy_to_user(to, from + pos, count);
639 if (ret == count)
640 return -EFAULT;
641 count -= ret;
642 *ppos = pos + count;
643 return count;
645 EXPORT_SYMBOL(simple_read_from_buffer);
648 * simple_write_to_buffer - copy data from user space to the buffer
649 * @to: the buffer to write to
650 * @available: the size of the buffer
651 * @ppos: the current position in the buffer
652 * @from: the user space buffer to read from
653 * @count: the maximum number of bytes to read
655 * The simple_write_to_buffer() function reads up to @count bytes from the user
656 * space address starting at @from into the buffer @to at offset @ppos.
658 * On success, the number of bytes written is returned and the offset @ppos is
659 * advanced by this number, or negative value is returned on error.
661 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
662 const void __user *from, size_t count)
664 loff_t pos = *ppos;
665 size_t res;
667 if (pos < 0)
668 return -EINVAL;
669 if (pos >= available || !count)
670 return 0;
671 if (count > available - pos)
672 count = available - pos;
673 res = copy_from_user(to + pos, from, count);
674 if (res == count)
675 return -EFAULT;
676 count -= res;
677 *ppos = pos + count;
678 return count;
680 EXPORT_SYMBOL(simple_write_to_buffer);
683 * memory_read_from_buffer - copy data from the buffer
684 * @to: the kernel space buffer to read to
685 * @count: the maximum number of bytes to read
686 * @ppos: the current position in the buffer
687 * @from: the buffer to read from
688 * @available: the size of the buffer
690 * The memory_read_from_buffer() function reads up to @count bytes from the
691 * buffer @from at offset @ppos into the kernel space address starting at @to.
693 * On success, the number of bytes read is returned and the offset @ppos is
694 * advanced by this number, or negative value is returned on error.
696 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
697 const void *from, size_t available)
699 loff_t pos = *ppos;
701 if (pos < 0)
702 return -EINVAL;
703 if (pos >= available)
704 return 0;
705 if (count > available - pos)
706 count = available - pos;
707 memcpy(to, from + pos, count);
708 *ppos = pos + count;
710 return count;
712 EXPORT_SYMBOL(memory_read_from_buffer);
715 * Transaction based IO.
716 * The file expects a single write which triggers the transaction, and then
717 * possibly a read which collects the result - which is stored in a
718 * file-local buffer.
721 void simple_transaction_set(struct file *file, size_t n)
723 struct simple_transaction_argresp *ar = file->private_data;
725 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
728 * The barrier ensures that ar->size will really remain zero until
729 * ar->data is ready for reading.
731 smp_mb();
732 ar->size = n;
734 EXPORT_SYMBOL(simple_transaction_set);
736 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
738 struct simple_transaction_argresp *ar;
739 static DEFINE_SPINLOCK(simple_transaction_lock);
741 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
742 return ERR_PTR(-EFBIG);
744 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
745 if (!ar)
746 return ERR_PTR(-ENOMEM);
748 spin_lock(&simple_transaction_lock);
750 /* only one write allowed per open */
751 if (file->private_data) {
752 spin_unlock(&simple_transaction_lock);
753 free_page((unsigned long)ar);
754 return ERR_PTR(-EBUSY);
757 file->private_data = ar;
759 spin_unlock(&simple_transaction_lock);
761 if (copy_from_user(ar->data, buf, size))
762 return ERR_PTR(-EFAULT);
764 return ar->data;
766 EXPORT_SYMBOL(simple_transaction_get);
768 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
770 struct simple_transaction_argresp *ar = file->private_data;
772 if (!ar)
773 return 0;
774 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
776 EXPORT_SYMBOL(simple_transaction_read);
778 int simple_transaction_release(struct inode *inode, struct file *file)
780 free_page((unsigned long)file->private_data);
781 return 0;
783 EXPORT_SYMBOL(simple_transaction_release);
785 /* Simple attribute files */
787 struct simple_attr {
788 int (*get)(void *, u64 *);
789 int (*set)(void *, u64);
790 char get_buf[24]; /* enough to store a u64 and "\n\0" */
791 char set_buf[24];
792 void *data;
793 const char *fmt; /* format for read operation */
794 struct mutex mutex; /* protects access to these buffers */
797 /* simple_attr_open is called by an actual attribute open file operation
798 * to set the attribute specific access operations. */
799 int simple_attr_open(struct inode *inode, struct file *file,
800 int (*get)(void *, u64 *), int (*set)(void *, u64),
801 const char *fmt)
803 struct simple_attr *attr;
805 attr = kzalloc(sizeof(*attr), GFP_KERNEL);
806 if (!attr)
807 return -ENOMEM;
809 attr->get = get;
810 attr->set = set;
811 attr->data = inode->i_private;
812 attr->fmt = fmt;
813 mutex_init(&attr->mutex);
815 file->private_data = attr;
817 return nonseekable_open(inode, file);
819 EXPORT_SYMBOL_GPL(simple_attr_open);
821 int simple_attr_release(struct inode *inode, struct file *file)
823 kfree(file->private_data);
824 return 0;
826 EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
828 /* read from the buffer that is filled with the get function */
829 ssize_t simple_attr_read(struct file *file, char __user *buf,
830 size_t len, loff_t *ppos)
832 struct simple_attr *attr;
833 size_t size;
834 ssize_t ret;
836 attr = file->private_data;
838 if (!attr->get)
839 return -EACCES;
841 ret = mutex_lock_interruptible(&attr->mutex);
842 if (ret)
843 return ret;
845 if (*ppos && attr->get_buf[0]) {
846 /* continued read */
847 size = strlen(attr->get_buf);
848 } else {
849 /* first read */
850 u64 val;
851 ret = attr->get(attr->data, &val);
852 if (ret)
853 goto out;
855 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
856 attr->fmt, (unsigned long long)val);
859 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
860 out:
861 mutex_unlock(&attr->mutex);
862 return ret;
864 EXPORT_SYMBOL_GPL(simple_attr_read);
866 /* interpret the buffer as a number to call the set function with */
867 ssize_t simple_attr_write(struct file *file, const char __user *buf,
868 size_t len, loff_t *ppos)
870 struct simple_attr *attr;
871 u64 val;
872 size_t size;
873 ssize_t ret;
875 attr = file->private_data;
876 if (!attr->set)
877 return -EACCES;
879 ret = mutex_lock_interruptible(&attr->mutex);
880 if (ret)
881 return ret;
883 ret = -EFAULT;
884 size = min(sizeof(attr->set_buf) - 1, len);
885 if (copy_from_user(attr->set_buf, buf, size))
886 goto out;
888 attr->set_buf[size] = '\0';
889 val = simple_strtoll(attr->set_buf, NULL, 0);
890 ret = attr->set(attr->data, val);
891 if (ret == 0)
892 ret = len; /* on success, claim we got the whole input */
893 out:
894 mutex_unlock(&attr->mutex);
895 return ret;
897 EXPORT_SYMBOL_GPL(simple_attr_write);
900 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
901 * @sb: filesystem to do the file handle conversion on
902 * @fid: file handle to convert
903 * @fh_len: length of the file handle in bytes
904 * @fh_type: type of file handle
905 * @get_inode: filesystem callback to retrieve inode
907 * This function decodes @fid as long as it has one of the well-known
908 * Linux filehandle types and calls @get_inode on it to retrieve the
909 * inode for the object specified in the file handle.
911 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
912 int fh_len, int fh_type, struct inode *(*get_inode)
913 (struct super_block *sb, u64 ino, u32 gen))
915 struct inode *inode = NULL;
917 if (fh_len < 2)
918 return NULL;
920 switch (fh_type) {
921 case FILEID_INO32_GEN:
922 case FILEID_INO32_GEN_PARENT:
923 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
924 break;
927 return d_obtain_alias(inode);
929 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
932 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
933 * @sb: filesystem to do the file handle conversion on
934 * @fid: file handle to convert
935 * @fh_len: length of the file handle in bytes
936 * @fh_type: type of file handle
937 * @get_inode: filesystem callback to retrieve inode
939 * This function decodes @fid as long as it has one of the well-known
940 * Linux filehandle types and calls @get_inode on it to retrieve the
941 * inode for the _parent_ object specified in the file handle if it
942 * is specified in the file handle, or NULL otherwise.
944 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
945 int fh_len, int fh_type, struct inode *(*get_inode)
946 (struct super_block *sb, u64 ino, u32 gen))
948 struct inode *inode = NULL;
950 if (fh_len <= 2)
951 return NULL;
953 switch (fh_type) {
954 case FILEID_INO32_GEN_PARENT:
955 inode = get_inode(sb, fid->i32.parent_ino,
956 (fh_len > 3 ? fid->i32.parent_gen : 0));
957 break;
960 return d_obtain_alias(inode);
962 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
965 * __generic_file_fsync - generic fsync implementation for simple filesystems
967 * @file: file to synchronize
968 * @start: start offset in bytes
969 * @end: end offset in bytes (inclusive)
970 * @datasync: only synchronize essential metadata if true
972 * This is a generic implementation of the fsync method for simple
973 * filesystems which track all non-inode metadata in the buffers list
974 * hanging off the address_space structure.
976 int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
977 int datasync)
979 struct inode *inode = file->f_mapping->host;
980 int err;
981 int ret;
983 err = file_write_and_wait_range(file, start, end);
984 if (err)
985 return err;
987 inode_lock(inode);
988 ret = sync_mapping_buffers(inode->i_mapping);
989 if (!(inode->i_state & I_DIRTY_ALL))
990 goto out;
991 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
992 goto out;
994 err = sync_inode_metadata(inode, 1);
995 if (ret == 0)
996 ret = err;
998 out:
999 inode_unlock(inode);
1000 /* check and advance again to catch errors after syncing out buffers */
1001 err = file_check_and_advance_wb_err(file);
1002 if (ret == 0)
1003 ret = err;
1004 return ret;
1006 EXPORT_SYMBOL(__generic_file_fsync);
1009 * generic_file_fsync - generic fsync implementation for simple filesystems
1010 * with flush
1011 * @file: file to synchronize
1012 * @start: start offset in bytes
1013 * @end: end offset in bytes (inclusive)
1014 * @datasync: only synchronize essential metadata if true
1018 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1019 int datasync)
1021 struct inode *inode = file->f_mapping->host;
1022 int err;
1024 err = __generic_file_fsync(file, start, end, datasync);
1025 if (err)
1026 return err;
1027 return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
1029 EXPORT_SYMBOL(generic_file_fsync);
1032 * generic_check_addressable - Check addressability of file system
1033 * @blocksize_bits: log of file system block size
1034 * @num_blocks: number of blocks in file system
1036 * Determine whether a file system with @num_blocks blocks (and a
1037 * block size of 2**@blocksize_bits) is addressable by the sector_t
1038 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1040 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1042 u64 last_fs_block = num_blocks - 1;
1043 u64 last_fs_page =
1044 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1046 if (unlikely(num_blocks == 0))
1047 return 0;
1049 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1050 return -EINVAL;
1052 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1053 (last_fs_page > (pgoff_t)(~0ULL))) {
1054 return -EFBIG;
1056 return 0;
1058 EXPORT_SYMBOL(generic_check_addressable);
1061 * No-op implementation of ->fsync for in-memory filesystems.
1063 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1065 return 0;
1067 EXPORT_SYMBOL(noop_fsync);
1069 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1070 void kfree_link(void *p)
1072 kfree(p);
1074 EXPORT_SYMBOL(kfree_link);
1077 * nop .set_page_dirty method so that people can use .page_mkwrite on
1078 * anon inodes.
1080 static int anon_set_page_dirty(struct page *page)
1082 return 0;
1086 * A single inode exists for all anon_inode files. Contrary to pipes,
1087 * anon_inode inodes have no associated per-instance data, so we need
1088 * only allocate one of them.
1090 struct inode *alloc_anon_inode(struct super_block *s)
1092 static const struct address_space_operations anon_aops = {
1093 .set_page_dirty = anon_set_page_dirty,
1095 struct inode *inode = new_inode_pseudo(s);
1097 if (!inode)
1098 return ERR_PTR(-ENOMEM);
1100 inode->i_ino = get_next_ino();
1101 inode->i_mapping->a_ops = &anon_aops;
1104 * Mark the inode dirty from the very beginning,
1105 * that way it will never be moved to the dirty
1106 * list because mark_inode_dirty() will think
1107 * that it already _is_ on the dirty list.
1109 inode->i_state = I_DIRTY;
1110 inode->i_mode = S_IRUSR | S_IWUSR;
1111 inode->i_uid = current_fsuid();
1112 inode->i_gid = current_fsgid();
1113 inode->i_flags |= S_PRIVATE;
1114 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1115 return inode;
1117 EXPORT_SYMBOL(alloc_anon_inode);
1120 * simple_nosetlease - generic helper for prohibiting leases
1121 * @filp: file pointer
1122 * @arg: type of lease to obtain
1123 * @flp: new lease supplied for insertion
1124 * @priv: private data for lm_setup operation
1126 * Generic helper for filesystems that do not wish to allow leases to be set.
1127 * All arguments are ignored and it just returns -EINVAL.
1130 simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1131 void **priv)
1133 return -EINVAL;
1135 EXPORT_SYMBOL(simple_nosetlease);
1137 const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1138 struct delayed_call *done)
1140 return inode->i_link;
1142 EXPORT_SYMBOL(simple_get_link);
1144 const struct inode_operations simple_symlink_inode_operations = {
1145 .get_link = simple_get_link,
1147 EXPORT_SYMBOL(simple_symlink_inode_operations);
1150 * Operations for a permanently empty directory.
1152 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1154 return ERR_PTR(-ENOENT);
1157 static int empty_dir_getattr(const struct path *path, struct kstat *stat,
1158 u32 request_mask, unsigned int query_flags)
1160 struct inode *inode = d_inode(path->dentry);
1161 generic_fillattr(inode, stat);
1162 return 0;
1165 static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1167 return -EPERM;
1170 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1172 return -EOPNOTSUPP;
1175 static const struct inode_operations empty_dir_inode_operations = {
1176 .lookup = empty_dir_lookup,
1177 .permission = generic_permission,
1178 .setattr = empty_dir_setattr,
1179 .getattr = empty_dir_getattr,
1180 .listxattr = empty_dir_listxattr,
1183 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1185 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1186 return generic_file_llseek_size(file, offset, whence, 2, 2);
1189 static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1191 dir_emit_dots(file, ctx);
1192 return 0;
1195 static const struct file_operations empty_dir_operations = {
1196 .llseek = empty_dir_llseek,
1197 .read = generic_read_dir,
1198 .iterate_shared = empty_dir_readdir,
1199 .fsync = noop_fsync,
1203 void make_empty_dir_inode(struct inode *inode)
1205 set_nlink(inode, 2);
1206 inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1207 inode->i_uid = GLOBAL_ROOT_UID;
1208 inode->i_gid = GLOBAL_ROOT_GID;
1209 inode->i_rdev = 0;
1210 inode->i_size = 0;
1211 inode->i_blkbits = PAGE_SHIFT;
1212 inode->i_blocks = 0;
1214 inode->i_op = &empty_dir_inode_operations;
1215 inode->i_opflags &= ~IOP_XATTR;
1216 inode->i_fop = &empty_dir_operations;
1219 bool is_empty_dir_inode(struct inode *inode)
1221 return (inode->i_fop == &empty_dir_operations) &&
1222 (inode->i_op == &empty_dir_inode_operations);