3 * Library for filesystems writers.
6 #include <linux/module.h>
7 #include <linux/pagemap.h>
8 #include <linux/mount.h>
10 #include <linux/mutex.h>
11 #include <linux/exportfs.h>
12 #include <linux/writeback.h>
13 #include <linux/buffer_head.h>
15 #include <asm/uaccess.h>
17 int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
20 struct inode *inode = dentry->d_inode;
21 generic_fillattr(inode, stat);
22 stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_SHIFT - 9);
26 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
28 buf->f_type = dentry->d_sb->s_magic;
29 buf->f_bsize = PAGE_CACHE_SIZE;
30 buf->f_namelen = NAME_MAX;
35 * Retaining negative dentries for an in-memory filesystem just wastes
36 * memory and lookup time: arrange for them to be deleted immediately.
38 static int simple_delete_dentry(struct dentry *dentry)
44 * Lookup the data. This is trivial - if the dentry didn't already
45 * exist, we know it is negative. Set d_op to delete negative dentries.
47 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
49 static const struct dentry_operations simple_dentry_operations = {
50 .d_delete = simple_delete_dentry,
53 if (dentry->d_name.len > NAME_MAX)
54 return ERR_PTR(-ENAMETOOLONG);
55 dentry->d_op = &simple_dentry_operations;
60 int simple_sync_file(struct file * file, struct dentry *dentry, int datasync)
65 int dcache_dir_open(struct inode *inode, struct file *file)
67 static struct qstr cursor_name = {.len = 1, .name = "."};
69 file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
71 return file->private_data ? 0 : -ENOMEM;
74 int dcache_dir_close(struct inode *inode, struct file *file)
76 dput(file->private_data);
80 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
82 mutex_lock(&file->f_path.dentry->d_inode->i_mutex);
85 offset += file->f_pos;
90 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
93 if (offset != file->f_pos) {
95 if (file->f_pos >= 2) {
97 struct dentry *cursor = file->private_data;
98 loff_t n = file->f_pos - 2;
100 spin_lock(&dcache_lock);
101 list_del(&cursor->d_u.d_child);
102 p = file->f_path.dentry->d_subdirs.next;
103 while (n && p != &file->f_path.dentry->d_subdirs) {
105 next = list_entry(p, struct dentry, d_u.d_child);
106 if (!d_unhashed(next) && next->d_inode)
110 list_add_tail(&cursor->d_u.d_child, p);
111 spin_unlock(&dcache_lock);
114 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
118 /* Relationship between i_mode and the DT_xxx types */
119 static inline unsigned char dt_type(struct inode *inode)
121 return (inode->i_mode >> 12) & 15;
125 * Directory is locked and all positive dentries in it are safe, since
126 * for ramfs-type trees they can't go away without unlink() or rmdir(),
127 * both impossible due to the lock on directory.
130 int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
132 struct dentry *dentry = filp->f_path.dentry;
133 struct dentry *cursor = filp->private_data;
134 struct list_head *p, *q = &cursor->d_u.d_child;
140 ino = dentry->d_inode->i_ino;
141 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
147 ino = parent_ino(dentry);
148 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
154 spin_lock(&dcache_lock);
155 if (filp->f_pos == 2)
156 list_move(q, &dentry->d_subdirs);
158 for (p=q->next; p != &dentry->d_subdirs; p=p->next) {
160 next = list_entry(p, struct dentry, d_u.d_child);
161 if (d_unhashed(next) || !next->d_inode)
164 spin_unlock(&dcache_lock);
165 if (filldir(dirent, next->d_name.name,
166 next->d_name.len, filp->f_pos,
167 next->d_inode->i_ino,
168 dt_type(next->d_inode)) < 0)
170 spin_lock(&dcache_lock);
171 /* next is still alive */
176 spin_unlock(&dcache_lock);
181 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
186 const struct file_operations simple_dir_operations = {
187 .open = dcache_dir_open,
188 .release = dcache_dir_close,
189 .llseek = dcache_dir_lseek,
190 .read = generic_read_dir,
191 .readdir = dcache_readdir,
192 .fsync = simple_sync_file,
195 const struct inode_operations simple_dir_inode_operations = {
196 .lookup = simple_lookup,
199 static const struct super_operations simple_super_operations = {
200 .statfs = simple_statfs,
204 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
205 * will never be mountable)
207 int get_sb_pseudo(struct file_system_type *fs_type, char *name,
208 const struct super_operations *ops, unsigned long magic,
209 struct vfsmount *mnt)
211 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
212 struct dentry *dentry;
214 struct qstr d_name = {.name = name, .len = strlen(name)};
219 s->s_flags = MS_NOUSER;
220 s->s_maxbytes = MAX_LFS_FILESIZE;
221 s->s_blocksize = PAGE_SIZE;
222 s->s_blocksize_bits = PAGE_SHIFT;
224 s->s_op = ops ? ops : &simple_super_operations;
230 * since this is the first inode, make it number 1. New inodes created
231 * after this must take care not to collide with it (by passing
232 * max_reserved of 1 to iunique).
235 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
236 root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
237 dentry = d_alloc(NULL, &d_name);
243 dentry->d_parent = dentry;
244 d_instantiate(dentry, root);
246 s->s_flags |= MS_ACTIVE;
247 simple_set_mnt(mnt, s);
251 deactivate_locked_super(s);
255 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
257 struct inode *inode = old_dentry->d_inode;
259 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
261 atomic_inc(&inode->i_count);
263 d_instantiate(dentry, inode);
267 static inline int simple_positive(struct dentry *dentry)
269 return dentry->d_inode && !d_unhashed(dentry);
272 int simple_empty(struct dentry *dentry)
274 struct dentry *child;
277 spin_lock(&dcache_lock);
278 list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child)
279 if (simple_positive(child))
283 spin_unlock(&dcache_lock);
287 int simple_unlink(struct inode *dir, struct dentry *dentry)
289 struct inode *inode = dentry->d_inode;
291 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
297 int simple_rmdir(struct inode *dir, struct dentry *dentry)
299 if (!simple_empty(dentry))
302 drop_nlink(dentry->d_inode);
303 simple_unlink(dir, dentry);
308 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
309 struct inode *new_dir, struct dentry *new_dentry)
311 struct inode *inode = old_dentry->d_inode;
312 int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
314 if (!simple_empty(new_dentry))
317 if (new_dentry->d_inode) {
318 simple_unlink(new_dir, new_dentry);
321 } else if (they_are_dirs) {
326 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
327 new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
332 int simple_readpage(struct file *file, struct page *page)
334 clear_highpage(page);
335 flush_dcache_page(page);
336 SetPageUptodate(page);
341 int simple_prepare_write(struct file *file, struct page *page,
342 unsigned from, unsigned to)
344 if (!PageUptodate(page)) {
345 if (to - from != PAGE_CACHE_SIZE)
346 zero_user_segments(page,
348 to, PAGE_CACHE_SIZE);
353 int simple_write_begin(struct file *file, struct address_space *mapping,
354 loff_t pos, unsigned len, unsigned flags,
355 struct page **pagep, void **fsdata)
361 index = pos >> PAGE_CACHE_SHIFT;
362 from = pos & (PAGE_CACHE_SIZE - 1);
364 page = grab_cache_page_write_begin(mapping, index, flags);
370 return simple_prepare_write(file, page, from, from+len);
374 * simple_write_end - .write_end helper for non-block-device FSes
375 * @available: See .write_end of address_space_operations
384 * simple_write_end does the minimum needed for updating a page after writing is
385 * done. It has the same API signature as the .write_end of
386 * address_space_operations vector. So it can just be set onto .write_end for
387 * FSes that don't need any other processing. i_mutex is assumed to be held.
388 * Block based filesystems should use generic_write_end().
389 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
390 * is not called, so a filesystem that actually does store data in .write_inode
391 * should extend on what's done here with a call to mark_inode_dirty() in the
392 * case that i_size has changed.
394 int simple_write_end(struct file *file, struct address_space *mapping,
395 loff_t pos, unsigned len, unsigned copied,
396 struct page *page, void *fsdata)
398 struct inode *inode = page->mapping->host;
399 loff_t last_pos = pos + copied;
401 /* zero the stale part of the page if we did a short copy */
403 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
405 zero_user(page, from + copied, len - copied);
408 if (!PageUptodate(page))
409 SetPageUptodate(page);
411 * No need to use i_size_read() here, the i_size
412 * cannot change under us because we hold the i_mutex.
414 if (last_pos > inode->i_size)
415 i_size_write(inode, last_pos);
417 set_page_dirty(page);
419 page_cache_release(page);
425 * the inodes created here are not hashed. If you use iunique to generate
426 * unique inode values later for this filesystem, then you must take care
427 * to pass it an appropriate max_reserved value to avoid collisions.
429 int simple_fill_super(struct super_block *s, int magic, struct tree_descr *files)
433 struct dentry *dentry;
436 s->s_blocksize = PAGE_CACHE_SIZE;
437 s->s_blocksize_bits = PAGE_CACHE_SHIFT;
439 s->s_op = &simple_super_operations;
442 inode = new_inode(s);
446 * because the root inode is 1, the files array must not contain an
450 inode->i_mode = S_IFDIR | 0755;
451 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
452 inode->i_op = &simple_dir_inode_operations;
453 inode->i_fop = &simple_dir_operations;
455 root = d_alloc_root(inode);
460 for (i = 0; !files->name || files->name[0]; i++, files++) {
464 /* warn if it tries to conflict with the root inode */
465 if (unlikely(i == 1))
466 printk(KERN_WARNING "%s: %s passed in a files array"
467 "with an index of 1!\n", __func__,
470 dentry = d_alloc_name(root, files->name);
473 inode = new_inode(s);
476 inode->i_mode = S_IFREG | files->mode;
477 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
478 inode->i_fop = files->ops;
480 d_add(dentry, inode);
490 static DEFINE_SPINLOCK(pin_fs_lock);
492 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
494 struct vfsmount *mnt = NULL;
495 spin_lock(&pin_fs_lock);
496 if (unlikely(!*mount)) {
497 spin_unlock(&pin_fs_lock);
498 mnt = vfs_kern_mount(type, 0, type->name, NULL);
501 spin_lock(&pin_fs_lock);
507 spin_unlock(&pin_fs_lock);
512 void simple_release_fs(struct vfsmount **mount, int *count)
514 struct vfsmount *mnt;
515 spin_lock(&pin_fs_lock);
519 spin_unlock(&pin_fs_lock);
524 * simple_read_from_buffer - copy data from the buffer to user space
525 * @to: the user space buffer to read to
526 * @count: the maximum number of bytes to read
527 * @ppos: the current position in the buffer
528 * @from: the buffer to read from
529 * @available: the size of the buffer
531 * The simple_read_from_buffer() function reads up to @count bytes from the
532 * buffer @from at offset @ppos into the user space address starting at @to.
534 * On success, the number of bytes read is returned and the offset @ppos is
535 * advanced by this number, or negative value is returned on error.
537 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
538 const void *from, size_t available)
545 if (pos >= available || !count)
547 if (count > available - pos)
548 count = available - pos;
549 ret = copy_to_user(to, from + pos, count);
558 * memory_read_from_buffer - copy data from the buffer
559 * @to: the kernel space buffer to read to
560 * @count: the maximum number of bytes to read
561 * @ppos: the current position in the buffer
562 * @from: the buffer to read from
563 * @available: the size of the buffer
565 * The memory_read_from_buffer() function reads up to @count bytes from the
566 * buffer @from at offset @ppos into the kernel space address starting at @to.
568 * On success, the number of bytes read is returned and the offset @ppos is
569 * advanced by this number, or negative value is returned on error.
571 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
572 const void *from, size_t available)
578 if (pos >= available)
580 if (count > available - pos)
581 count = available - pos;
582 memcpy(to, from + pos, count);
589 * Transaction based IO.
590 * The file expects a single write which triggers the transaction, and then
591 * possibly a read which collects the result - which is stored in a
595 void simple_transaction_set(struct file *file, size_t n)
597 struct simple_transaction_argresp *ar = file->private_data;
599 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
602 * The barrier ensures that ar->size will really remain zero until
603 * ar->data is ready for reading.
609 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
611 struct simple_transaction_argresp *ar;
612 static DEFINE_SPINLOCK(simple_transaction_lock);
614 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
615 return ERR_PTR(-EFBIG);
617 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
619 return ERR_PTR(-ENOMEM);
621 spin_lock(&simple_transaction_lock);
623 /* only one write allowed per open */
624 if (file->private_data) {
625 spin_unlock(&simple_transaction_lock);
626 free_page((unsigned long)ar);
627 return ERR_PTR(-EBUSY);
630 file->private_data = ar;
632 spin_unlock(&simple_transaction_lock);
634 if (copy_from_user(ar->data, buf, size))
635 return ERR_PTR(-EFAULT);
640 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
642 struct simple_transaction_argresp *ar = file->private_data;
646 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
649 int simple_transaction_release(struct inode *inode, struct file *file)
651 free_page((unsigned long)file->private_data);
655 /* Simple attribute files */
658 int (*get)(void *, u64 *);
659 int (*set)(void *, u64);
660 char get_buf[24]; /* enough to store a u64 and "\n\0" */
663 const char *fmt; /* format for read operation */
664 struct mutex mutex; /* protects access to these buffers */
667 /* simple_attr_open is called by an actual attribute open file operation
668 * to set the attribute specific access operations. */
669 int simple_attr_open(struct inode *inode, struct file *file,
670 int (*get)(void *, u64 *), int (*set)(void *, u64),
673 struct simple_attr *attr;
675 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
681 attr->data = inode->i_private;
683 mutex_init(&attr->mutex);
685 file->private_data = attr;
687 return nonseekable_open(inode, file);
690 int simple_attr_release(struct inode *inode, struct file *file)
692 kfree(file->private_data);
696 /* read from the buffer that is filled with the get function */
697 ssize_t simple_attr_read(struct file *file, char __user *buf,
698 size_t len, loff_t *ppos)
700 struct simple_attr *attr;
704 attr = file->private_data;
709 ret = mutex_lock_interruptible(&attr->mutex);
713 if (*ppos) { /* continued read */
714 size = strlen(attr->get_buf);
715 } else { /* first read */
717 ret = attr->get(attr->data, &val);
721 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
722 attr->fmt, (unsigned long long)val);
725 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
727 mutex_unlock(&attr->mutex);
731 /* interpret the buffer as a number to call the set function with */
732 ssize_t simple_attr_write(struct file *file, const char __user *buf,
733 size_t len, loff_t *ppos)
735 struct simple_attr *attr;
740 attr = file->private_data;
744 ret = mutex_lock_interruptible(&attr->mutex);
749 size = min(sizeof(attr->set_buf) - 1, len);
750 if (copy_from_user(attr->set_buf, buf, size))
753 attr->set_buf[size] = '\0';
754 val = simple_strtol(attr->set_buf, NULL, 0);
755 ret = attr->set(attr->data, val);
757 ret = len; /* on success, claim we got the whole input */
759 mutex_unlock(&attr->mutex);
764 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
765 * @sb: filesystem to do the file handle conversion on
766 * @fid: file handle to convert
767 * @fh_len: length of the file handle in bytes
768 * @fh_type: type of file handle
769 * @get_inode: filesystem callback to retrieve inode
771 * This function decodes @fid as long as it has one of the well-known
772 * Linux filehandle types and calls @get_inode on it to retrieve the
773 * inode for the object specified in the file handle.
775 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
776 int fh_len, int fh_type, struct inode *(*get_inode)
777 (struct super_block *sb, u64 ino, u32 gen))
779 struct inode *inode = NULL;
785 case FILEID_INO32_GEN:
786 case FILEID_INO32_GEN_PARENT:
787 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
791 return d_obtain_alias(inode);
793 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
796 * generic_fh_to_dentry - generic helper for the fh_to_parent export operation
797 * @sb: filesystem to do the file handle conversion on
798 * @fid: file handle to convert
799 * @fh_len: length of the file handle in bytes
800 * @fh_type: type of file handle
801 * @get_inode: filesystem callback to retrieve inode
803 * This function decodes @fid as long as it has one of the well-known
804 * Linux filehandle types and calls @get_inode on it to retrieve the
805 * inode for the _parent_ object specified in the file handle if it
806 * is specified in the file handle, or NULL otherwise.
808 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
809 int fh_len, int fh_type, struct inode *(*get_inode)
810 (struct super_block *sb, u64 ino, u32 gen))
812 struct inode *inode = NULL;
818 case FILEID_INO32_GEN_PARENT:
819 inode = get_inode(sb, fid->i32.parent_ino,
820 (fh_len > 3 ? fid->i32.parent_gen : 0));
824 return d_obtain_alias(inode);
826 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
828 int simple_fsync(struct file *file, struct dentry *dentry, int datasync)
830 struct writeback_control wbc = {
831 .sync_mode = WB_SYNC_ALL,
832 .nr_to_write = 0, /* metadata-only; caller takes care of data */
834 struct inode *inode = dentry->d_inode;
838 ret = sync_mapping_buffers(inode->i_mapping);
839 if (!(inode->i_state & I_DIRTY))
841 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
844 err = sync_inode(inode, &wbc);
849 EXPORT_SYMBOL(simple_fsync);
851 EXPORT_SYMBOL(dcache_dir_close);
852 EXPORT_SYMBOL(dcache_dir_lseek);
853 EXPORT_SYMBOL(dcache_dir_open);
854 EXPORT_SYMBOL(dcache_readdir);
855 EXPORT_SYMBOL(generic_read_dir);
856 EXPORT_SYMBOL(get_sb_pseudo);
857 EXPORT_SYMBOL(simple_write_begin);
858 EXPORT_SYMBOL(simple_write_end);
859 EXPORT_SYMBOL(simple_dir_inode_operations);
860 EXPORT_SYMBOL(simple_dir_operations);
861 EXPORT_SYMBOL(simple_empty);
862 EXPORT_SYMBOL(simple_fill_super);
863 EXPORT_SYMBOL(simple_getattr);
864 EXPORT_SYMBOL(simple_link);
865 EXPORT_SYMBOL(simple_lookup);
866 EXPORT_SYMBOL(simple_pin_fs);
867 EXPORT_UNUSED_SYMBOL(simple_prepare_write);
868 EXPORT_SYMBOL(simple_readpage);
869 EXPORT_SYMBOL(simple_release_fs);
870 EXPORT_SYMBOL(simple_rename);
871 EXPORT_SYMBOL(simple_rmdir);
872 EXPORT_SYMBOL(simple_statfs);
873 EXPORT_SYMBOL(simple_sync_file);
874 EXPORT_SYMBOL(simple_unlink);
875 EXPORT_SYMBOL(simple_read_from_buffer);
876 EXPORT_SYMBOL(memory_read_from_buffer);
877 EXPORT_SYMBOL(simple_transaction_set);
878 EXPORT_SYMBOL(simple_transaction_get);
879 EXPORT_SYMBOL(simple_transaction_read);
880 EXPORT_SYMBOL(simple_transaction_release);
881 EXPORT_SYMBOL_GPL(simple_attr_open);
882 EXPORT_SYMBOL_GPL(simple_attr_release);
883 EXPORT_SYMBOL_GPL(simple_attr_read);
884 EXPORT_SYMBOL_GPL(simple_attr_write);