3 * Library for filesystems writers.
6 #include <linux/module.h>
7 #include <linux/pagemap.h>
8 #include <linux/slab.h>
9 #include <linux/mount.h>
10 #include <linux/vfs.h>
11 #include <linux/mutex.h>
12 #include <linux/exportfs.h>
13 #include <linux/writeback.h>
14 #include <linux/buffer_head.h>
16 #include <asm/uaccess.h>
18 int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
21 struct inode *inode = dentry->d_inode;
22 generic_fillattr(inode, stat);
23 stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_SHIFT - 9);
27 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
29 buf->f_type = dentry->d_sb->s_magic;
30 buf->f_bsize = PAGE_CACHE_SIZE;
31 buf->f_namelen = NAME_MAX;
36 * Retaining negative dentries for an in-memory filesystem just wastes
37 * memory and lookup time: arrange for them to be deleted immediately.
39 static int simple_delete_dentry(struct dentry *dentry)
45 * Lookup the data. This is trivial - if the dentry didn't already
46 * exist, we know it is negative. Set d_op to delete negative dentries.
48 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
50 static const struct dentry_operations simple_dentry_operations = {
51 .d_delete = simple_delete_dentry,
54 if (dentry->d_name.len > NAME_MAX)
55 return ERR_PTR(-ENAMETOOLONG);
56 dentry->d_op = &simple_dentry_operations;
61 int dcache_dir_open(struct inode *inode, struct file *file)
63 static struct qstr cursor_name = {.len = 1, .name = "."};
65 file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
67 return file->private_data ? 0 : -ENOMEM;
70 int dcache_dir_close(struct inode *inode, struct file *file)
72 dput(file->private_data);
76 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
78 mutex_lock(&file->f_path.dentry->d_inode->i_mutex);
81 offset += file->f_pos;
86 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
89 if (offset != file->f_pos) {
91 if (file->f_pos >= 2) {
93 struct dentry *cursor = file->private_data;
94 loff_t n = file->f_pos - 2;
96 spin_lock(&dcache_lock);
97 list_del(&cursor->d_u.d_child);
98 p = file->f_path.dentry->d_subdirs.next;
99 while (n && p != &file->f_path.dentry->d_subdirs) {
101 next = list_entry(p, struct dentry, d_u.d_child);
102 if (!d_unhashed(next) && next->d_inode)
106 list_add_tail(&cursor->d_u.d_child, p);
107 spin_unlock(&dcache_lock);
110 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
114 /* Relationship between i_mode and the DT_xxx types */
115 static inline unsigned char dt_type(struct inode *inode)
117 return (inode->i_mode >> 12) & 15;
121 * Directory is locked and all positive dentries in it are safe, since
122 * for ramfs-type trees they can't go away without unlink() or rmdir(),
123 * both impossible due to the lock on directory.
126 int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
128 struct dentry *dentry = filp->f_path.dentry;
129 struct dentry *cursor = filp->private_data;
130 struct list_head *p, *q = &cursor->d_u.d_child;
136 ino = dentry->d_inode->i_ino;
137 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
143 ino = parent_ino(dentry);
144 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
150 spin_lock(&dcache_lock);
151 if (filp->f_pos == 2)
152 list_move(q, &dentry->d_subdirs);
154 for (p=q->next; p != &dentry->d_subdirs; p=p->next) {
156 next = list_entry(p, struct dentry, d_u.d_child);
157 if (d_unhashed(next) || !next->d_inode)
160 spin_unlock(&dcache_lock);
161 if (filldir(dirent, next->d_name.name,
162 next->d_name.len, filp->f_pos,
163 next->d_inode->i_ino,
164 dt_type(next->d_inode)) < 0)
166 spin_lock(&dcache_lock);
167 /* next is still alive */
172 spin_unlock(&dcache_lock);
177 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
182 const struct file_operations simple_dir_operations = {
183 .open = dcache_dir_open,
184 .release = dcache_dir_close,
185 .llseek = dcache_dir_lseek,
186 .read = generic_read_dir,
187 .readdir = dcache_readdir,
191 const struct inode_operations simple_dir_inode_operations = {
192 .lookup = simple_lookup,
195 static const struct super_operations simple_super_operations = {
196 .statfs = simple_statfs,
200 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
201 * will never be mountable)
203 int get_sb_pseudo(struct file_system_type *fs_type, char *name,
204 const struct super_operations *ops, unsigned long magic,
205 struct vfsmount *mnt)
207 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
208 struct dentry *dentry;
210 struct qstr d_name = {.name = name, .len = strlen(name)};
215 s->s_flags = MS_NOUSER;
216 s->s_maxbytes = MAX_LFS_FILESIZE;
217 s->s_blocksize = PAGE_SIZE;
218 s->s_blocksize_bits = PAGE_SHIFT;
220 s->s_op = ops ? ops : &simple_super_operations;
226 * since this is the first inode, make it number 1. New inodes created
227 * after this must take care not to collide with it (by passing
228 * max_reserved of 1 to iunique).
231 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
232 root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
233 dentry = d_alloc(NULL, &d_name);
239 dentry->d_parent = dentry;
240 d_instantiate(dentry, root);
242 s->s_flags |= MS_ACTIVE;
243 simple_set_mnt(mnt, s);
247 deactivate_locked_super(s);
251 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
253 struct inode *inode = old_dentry->d_inode;
255 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
257 atomic_inc(&inode->i_count);
259 d_instantiate(dentry, inode);
263 static inline int simple_positive(struct dentry *dentry)
265 return dentry->d_inode && !d_unhashed(dentry);
268 int simple_empty(struct dentry *dentry)
270 struct dentry *child;
273 spin_lock(&dcache_lock);
274 list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child)
275 if (simple_positive(child))
279 spin_unlock(&dcache_lock);
283 int simple_unlink(struct inode *dir, struct dentry *dentry)
285 struct inode *inode = dentry->d_inode;
287 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
293 int simple_rmdir(struct inode *dir, struct dentry *dentry)
295 if (!simple_empty(dentry))
298 drop_nlink(dentry->d_inode);
299 simple_unlink(dir, dentry);
304 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
305 struct inode *new_dir, struct dentry *new_dentry)
307 struct inode *inode = old_dentry->d_inode;
308 int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
310 if (!simple_empty(new_dentry))
313 if (new_dentry->d_inode) {
314 simple_unlink(new_dir, new_dentry);
317 } else if (they_are_dirs) {
322 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
323 new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
328 int simple_readpage(struct file *file, struct page *page)
330 clear_highpage(page);
331 flush_dcache_page(page);
332 SetPageUptodate(page);
337 int simple_write_begin(struct file *file, struct address_space *mapping,
338 loff_t pos, unsigned len, unsigned flags,
339 struct page **pagep, void **fsdata)
344 index = pos >> PAGE_CACHE_SHIFT;
346 page = grab_cache_page_write_begin(mapping, index, flags);
352 if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
353 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
355 zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
361 * simple_write_end - .write_end helper for non-block-device FSes
362 * @available: See .write_end of address_space_operations
371 * simple_write_end does the minimum needed for updating a page after writing is
372 * done. It has the same API signature as the .write_end of
373 * address_space_operations vector. So it can just be set onto .write_end for
374 * FSes that don't need any other processing. i_mutex is assumed to be held.
375 * Block based filesystems should use generic_write_end().
376 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
377 * is not called, so a filesystem that actually does store data in .write_inode
378 * should extend on what's done here with a call to mark_inode_dirty() in the
379 * case that i_size has changed.
381 int simple_write_end(struct file *file, struct address_space *mapping,
382 loff_t pos, unsigned len, unsigned copied,
383 struct page *page, void *fsdata)
385 struct inode *inode = page->mapping->host;
386 loff_t last_pos = pos + copied;
388 /* zero the stale part of the page if we did a short copy */
390 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
392 zero_user(page, from + copied, len - copied);
395 if (!PageUptodate(page))
396 SetPageUptodate(page);
398 * No need to use i_size_read() here, the i_size
399 * cannot change under us because we hold the i_mutex.
401 if (last_pos > inode->i_size)
402 i_size_write(inode, last_pos);
404 set_page_dirty(page);
406 page_cache_release(page);
412 * the inodes created here are not hashed. If you use iunique to generate
413 * unique inode values later for this filesystem, then you must take care
414 * to pass it an appropriate max_reserved value to avoid collisions.
416 int simple_fill_super(struct super_block *s, int magic, struct tree_descr *files)
420 struct dentry *dentry;
423 s->s_blocksize = PAGE_CACHE_SIZE;
424 s->s_blocksize_bits = PAGE_CACHE_SHIFT;
426 s->s_op = &simple_super_operations;
429 inode = new_inode(s);
433 * because the root inode is 1, the files array must not contain an
437 inode->i_mode = S_IFDIR | 0755;
438 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
439 inode->i_op = &simple_dir_inode_operations;
440 inode->i_fop = &simple_dir_operations;
442 root = d_alloc_root(inode);
447 for (i = 0; !files->name || files->name[0]; i++, files++) {
451 /* warn if it tries to conflict with the root inode */
452 if (unlikely(i == 1))
453 printk(KERN_WARNING "%s: %s passed in a files array"
454 "with an index of 1!\n", __func__,
457 dentry = d_alloc_name(root, files->name);
460 inode = new_inode(s);
463 inode->i_mode = S_IFREG | files->mode;
464 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
465 inode->i_fop = files->ops;
467 d_add(dentry, inode);
477 static DEFINE_SPINLOCK(pin_fs_lock);
479 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
481 struct vfsmount *mnt = NULL;
482 spin_lock(&pin_fs_lock);
483 if (unlikely(!*mount)) {
484 spin_unlock(&pin_fs_lock);
485 mnt = vfs_kern_mount(type, 0, type->name, NULL);
488 spin_lock(&pin_fs_lock);
494 spin_unlock(&pin_fs_lock);
499 void simple_release_fs(struct vfsmount **mount, int *count)
501 struct vfsmount *mnt;
502 spin_lock(&pin_fs_lock);
506 spin_unlock(&pin_fs_lock);
511 * simple_read_from_buffer - copy data from the buffer to user space
512 * @to: the user space buffer to read to
513 * @count: the maximum number of bytes to read
514 * @ppos: the current position in the buffer
515 * @from: the buffer to read from
516 * @available: the size of the buffer
518 * The simple_read_from_buffer() function reads up to @count bytes from the
519 * buffer @from at offset @ppos into the user space address starting at @to.
521 * On success, the number of bytes read is returned and the offset @ppos is
522 * advanced by this number, or negative value is returned on error.
524 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
525 const void *from, size_t available)
532 if (pos >= available || !count)
534 if (count > available - pos)
535 count = available - pos;
536 ret = copy_to_user(to, from + pos, count);
545 * simple_write_to_buffer - copy data from user space to the buffer
546 * @to: the buffer to write to
547 * @available: the size of the buffer
548 * @ppos: the current position in the buffer
549 * @from: the user space buffer to read from
550 * @count: the maximum number of bytes to read
552 * The simple_write_to_buffer() function reads up to @count bytes from the user
553 * space address starting at @from into the buffer @to at offset @ppos.
555 * On success, the number of bytes written is returned and the offset @ppos is
556 * advanced by this number, or negative value is returned on error.
558 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
559 const void __user *from, size_t count)
566 if (pos >= available || !count)
568 if (count > available - pos)
569 count = available - pos;
570 res = copy_from_user(to + pos, from, count);
579 * memory_read_from_buffer - copy data from the buffer
580 * @to: the kernel space buffer to read to
581 * @count: the maximum number of bytes to read
582 * @ppos: the current position in the buffer
583 * @from: the buffer to read from
584 * @available: the size of the buffer
586 * The memory_read_from_buffer() function reads up to @count bytes from the
587 * buffer @from at offset @ppos into the kernel space address starting at @to.
589 * On success, the number of bytes read is returned and the offset @ppos is
590 * advanced by this number, or negative value is returned on error.
592 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
593 const void *from, size_t available)
599 if (pos >= available)
601 if (count > available - pos)
602 count = available - pos;
603 memcpy(to, from + pos, count);
610 * Transaction based IO.
611 * The file expects a single write which triggers the transaction, and then
612 * possibly a read which collects the result - which is stored in a
616 void simple_transaction_set(struct file *file, size_t n)
618 struct simple_transaction_argresp *ar = file->private_data;
620 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
623 * The barrier ensures that ar->size will really remain zero until
624 * ar->data is ready for reading.
630 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
632 struct simple_transaction_argresp *ar;
633 static DEFINE_SPINLOCK(simple_transaction_lock);
635 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
636 return ERR_PTR(-EFBIG);
638 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
640 return ERR_PTR(-ENOMEM);
642 spin_lock(&simple_transaction_lock);
644 /* only one write allowed per open */
645 if (file->private_data) {
646 spin_unlock(&simple_transaction_lock);
647 free_page((unsigned long)ar);
648 return ERR_PTR(-EBUSY);
651 file->private_data = ar;
653 spin_unlock(&simple_transaction_lock);
655 if (copy_from_user(ar->data, buf, size))
656 return ERR_PTR(-EFAULT);
661 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
663 struct simple_transaction_argresp *ar = file->private_data;
667 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
670 int simple_transaction_release(struct inode *inode, struct file *file)
672 free_page((unsigned long)file->private_data);
676 /* Simple attribute files */
679 int (*get)(void *, u64 *);
680 int (*set)(void *, u64);
681 char get_buf[24]; /* enough to store a u64 and "\n\0" */
684 const char *fmt; /* format for read operation */
685 struct mutex mutex; /* protects access to these buffers */
688 /* simple_attr_open is called by an actual attribute open file operation
689 * to set the attribute specific access operations. */
690 int simple_attr_open(struct inode *inode, struct file *file,
691 int (*get)(void *, u64 *), int (*set)(void *, u64),
694 struct simple_attr *attr;
696 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
702 attr->data = inode->i_private;
704 mutex_init(&attr->mutex);
706 file->private_data = attr;
708 return nonseekable_open(inode, file);
711 int simple_attr_release(struct inode *inode, struct file *file)
713 kfree(file->private_data);
717 /* read from the buffer that is filled with the get function */
718 ssize_t simple_attr_read(struct file *file, char __user *buf,
719 size_t len, loff_t *ppos)
721 struct simple_attr *attr;
725 attr = file->private_data;
730 ret = mutex_lock_interruptible(&attr->mutex);
734 if (*ppos) { /* continued read */
735 size = strlen(attr->get_buf);
736 } else { /* first read */
738 ret = attr->get(attr->data, &val);
742 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
743 attr->fmt, (unsigned long long)val);
746 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
748 mutex_unlock(&attr->mutex);
752 /* interpret the buffer as a number to call the set function with */
753 ssize_t simple_attr_write(struct file *file, const char __user *buf,
754 size_t len, loff_t *ppos)
756 struct simple_attr *attr;
761 attr = file->private_data;
765 ret = mutex_lock_interruptible(&attr->mutex);
770 size = min(sizeof(attr->set_buf) - 1, len);
771 if (copy_from_user(attr->set_buf, buf, size))
774 attr->set_buf[size] = '\0';
775 val = simple_strtol(attr->set_buf, NULL, 0);
776 ret = attr->set(attr->data, val);
778 ret = len; /* on success, claim we got the whole input */
780 mutex_unlock(&attr->mutex);
785 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
786 * @sb: filesystem to do the file handle conversion on
787 * @fid: file handle to convert
788 * @fh_len: length of the file handle in bytes
789 * @fh_type: type of file handle
790 * @get_inode: filesystem callback to retrieve inode
792 * This function decodes @fid as long as it has one of the well-known
793 * Linux filehandle types and calls @get_inode on it to retrieve the
794 * inode for the object specified in the file handle.
796 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
797 int fh_len, int fh_type, struct inode *(*get_inode)
798 (struct super_block *sb, u64 ino, u32 gen))
800 struct inode *inode = NULL;
806 case FILEID_INO32_GEN:
807 case FILEID_INO32_GEN_PARENT:
808 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
812 return d_obtain_alias(inode);
814 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
817 * generic_fh_to_dentry - generic helper for the fh_to_parent export operation
818 * @sb: filesystem to do the file handle conversion on
819 * @fid: file handle to convert
820 * @fh_len: length of the file handle in bytes
821 * @fh_type: type of file handle
822 * @get_inode: filesystem callback to retrieve inode
824 * This function decodes @fid as long as it has one of the well-known
825 * Linux filehandle types and calls @get_inode on it to retrieve the
826 * inode for the _parent_ object specified in the file handle if it
827 * is specified in the file handle, or NULL otherwise.
829 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
830 int fh_len, int fh_type, struct inode *(*get_inode)
831 (struct super_block *sb, u64 ino, u32 gen))
833 struct inode *inode = NULL;
839 case FILEID_INO32_GEN_PARENT:
840 inode = get_inode(sb, fid->i32.parent_ino,
841 (fh_len > 3 ? fid->i32.parent_gen : 0));
845 return d_obtain_alias(inode);
847 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
850 * generic_file_fsync - generic fsync implementation for simple filesystems
851 * @file: file to synchronize
852 * @datasync: only synchronize essential metadata if true
854 * This is a generic implementation of the fsync method for simple
855 * filesystems which track all non-inode metadata in the buffers list
856 * hanging off the address_space structure.
858 int generic_file_fsync(struct file *file, int datasync)
860 struct writeback_control wbc = {
861 .sync_mode = WB_SYNC_ALL,
862 .nr_to_write = 0, /* metadata-only; caller takes care of data */
864 struct inode *inode = file->f_mapping->host;
868 ret = sync_mapping_buffers(inode->i_mapping);
869 if (!(inode->i_state & I_DIRTY))
871 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
874 err = sync_inode(inode, &wbc);
879 EXPORT_SYMBOL(generic_file_fsync);
882 * No-op implementation of ->fsync for in-memory filesystems.
884 int noop_fsync(struct file *file, int datasync)
889 EXPORT_SYMBOL(dcache_dir_close);
890 EXPORT_SYMBOL(dcache_dir_lseek);
891 EXPORT_SYMBOL(dcache_dir_open);
892 EXPORT_SYMBOL(dcache_readdir);
893 EXPORT_SYMBOL(generic_read_dir);
894 EXPORT_SYMBOL(get_sb_pseudo);
895 EXPORT_SYMBOL(simple_write_begin);
896 EXPORT_SYMBOL(simple_write_end);
897 EXPORT_SYMBOL(simple_dir_inode_operations);
898 EXPORT_SYMBOL(simple_dir_operations);
899 EXPORT_SYMBOL(simple_empty);
900 EXPORT_SYMBOL(simple_fill_super);
901 EXPORT_SYMBOL(simple_getattr);
902 EXPORT_SYMBOL(simple_link);
903 EXPORT_SYMBOL(simple_lookup);
904 EXPORT_SYMBOL(simple_pin_fs);
905 EXPORT_SYMBOL(simple_readpage);
906 EXPORT_SYMBOL(simple_release_fs);
907 EXPORT_SYMBOL(simple_rename);
908 EXPORT_SYMBOL(simple_rmdir);
909 EXPORT_SYMBOL(simple_statfs);
910 EXPORT_SYMBOL(noop_fsync);
911 EXPORT_SYMBOL(simple_unlink);
912 EXPORT_SYMBOL(simple_read_from_buffer);
913 EXPORT_SYMBOL(simple_write_to_buffer);
914 EXPORT_SYMBOL(memory_read_from_buffer);
915 EXPORT_SYMBOL(simple_transaction_set);
916 EXPORT_SYMBOL(simple_transaction_get);
917 EXPORT_SYMBOL(simple_transaction_read);
918 EXPORT_SYMBOL(simple_transaction_release);
919 EXPORT_SYMBOL_GPL(simple_attr_open);
920 EXPORT_SYMBOL_GPL(simple_attr_release);
921 EXPORT_SYMBOL_GPL(simple_attr_read);
922 EXPORT_SYMBOL_GPL(simple_attr_write);