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/quotaops.h>
12 #include <linux/mutex.h>
13 #include <linux/exportfs.h>
14 #include <linux/writeback.h>
15 #include <linux/buffer_head.h>
17 #include <asm/uaccess.h>
19 int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
22 struct inode *inode = dentry->d_inode;
23 generic_fillattr(inode, stat);
24 stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_SHIFT - 9);
28 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
30 buf->f_type = dentry->d_sb->s_magic;
31 buf->f_bsize = PAGE_CACHE_SIZE;
32 buf->f_namelen = NAME_MAX;
37 * Retaining negative dentries for an in-memory filesystem just wastes
38 * memory and lookup time: arrange for them to be deleted immediately.
40 static int simple_delete_dentry(struct dentry *dentry)
46 * Lookup the data. This is trivial - if the dentry didn't already
47 * exist, we know it is negative. Set d_op to delete negative dentries.
49 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
51 static const struct dentry_operations simple_dentry_operations = {
52 .d_delete = simple_delete_dentry,
55 if (dentry->d_name.len > NAME_MAX)
56 return ERR_PTR(-ENAMETOOLONG);
57 dentry->d_op = &simple_dentry_operations;
62 int dcache_dir_open(struct inode *inode, struct file *file)
64 static struct qstr cursor_name = {.len = 1, .name = "."};
66 file->private_data = d_alloc(file->f_path.dentry, &cursor_name);
68 return file->private_data ? 0 : -ENOMEM;
71 int dcache_dir_close(struct inode *inode, struct file *file)
73 dput(file->private_data);
77 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
79 mutex_lock(&file->f_path.dentry->d_inode->i_mutex);
82 offset += file->f_pos;
87 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
90 if (offset != file->f_pos) {
92 if (file->f_pos >= 2) {
94 struct dentry *cursor = file->private_data;
95 loff_t n = file->f_pos - 2;
97 spin_lock(&dcache_lock);
98 list_del(&cursor->d_u.d_child);
99 p = file->f_path.dentry->d_subdirs.next;
100 while (n && p != &file->f_path.dentry->d_subdirs) {
102 next = list_entry(p, struct dentry, d_u.d_child);
103 if (!d_unhashed(next) && next->d_inode)
107 list_add_tail(&cursor->d_u.d_child, p);
108 spin_unlock(&dcache_lock);
111 mutex_unlock(&file->f_path.dentry->d_inode->i_mutex);
115 /* Relationship between i_mode and the DT_xxx types */
116 static inline unsigned char dt_type(struct inode *inode)
118 return (inode->i_mode >> 12) & 15;
122 * Directory is locked and all positive dentries in it are safe, since
123 * for ramfs-type trees they can't go away without unlink() or rmdir(),
124 * both impossible due to the lock on directory.
127 int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
129 struct dentry *dentry = filp->f_path.dentry;
130 struct dentry *cursor = filp->private_data;
131 struct list_head *p, *q = &cursor->d_u.d_child;
137 ino = dentry->d_inode->i_ino;
138 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
144 ino = parent_ino(dentry);
145 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
151 spin_lock(&dcache_lock);
152 if (filp->f_pos == 2)
153 list_move(q, &dentry->d_subdirs);
155 for (p=q->next; p != &dentry->d_subdirs; p=p->next) {
157 next = list_entry(p, struct dentry, d_u.d_child);
158 if (d_unhashed(next) || !next->d_inode)
161 spin_unlock(&dcache_lock);
162 if (filldir(dirent, next->d_name.name,
163 next->d_name.len, filp->f_pos,
164 next->d_inode->i_ino,
165 dt_type(next->d_inode)) < 0)
167 spin_lock(&dcache_lock);
168 /* next is still alive */
173 spin_unlock(&dcache_lock);
178 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
183 const struct file_operations simple_dir_operations = {
184 .open = dcache_dir_open,
185 .release = dcache_dir_close,
186 .llseek = dcache_dir_lseek,
187 .read = generic_read_dir,
188 .readdir = dcache_readdir,
192 const struct inode_operations simple_dir_inode_operations = {
193 .lookup = simple_lookup,
196 static const struct super_operations simple_super_operations = {
197 .statfs = simple_statfs,
201 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
202 * will never be mountable)
204 int get_sb_pseudo(struct file_system_type *fs_type, char *name,
205 const struct super_operations *ops, unsigned long magic,
206 struct vfsmount *mnt)
208 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
209 struct dentry *dentry;
211 struct qstr d_name = {.name = name, .len = strlen(name)};
216 s->s_flags = MS_NOUSER;
217 s->s_maxbytes = MAX_LFS_FILESIZE;
218 s->s_blocksize = PAGE_SIZE;
219 s->s_blocksize_bits = PAGE_SHIFT;
221 s->s_op = ops ? ops : &simple_super_operations;
227 * since this is the first inode, make it number 1. New inodes created
228 * after this must take care not to collide with it (by passing
229 * max_reserved of 1 to iunique).
232 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
233 root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
234 dentry = d_alloc(NULL, &d_name);
240 dentry->d_parent = dentry;
241 d_instantiate(dentry, root);
243 s->s_flags |= MS_ACTIVE;
244 simple_set_mnt(mnt, s);
248 deactivate_locked_super(s);
252 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
254 struct inode *inode = old_dentry->d_inode;
256 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
258 atomic_inc(&inode->i_count);
260 d_instantiate(dentry, inode);
264 static inline int simple_positive(struct dentry *dentry)
266 return dentry->d_inode && !d_unhashed(dentry);
269 int simple_empty(struct dentry *dentry)
271 struct dentry *child;
274 spin_lock(&dcache_lock);
275 list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child)
276 if (simple_positive(child))
280 spin_unlock(&dcache_lock);
284 int simple_unlink(struct inode *dir, struct dentry *dentry)
286 struct inode *inode = dentry->d_inode;
288 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
294 int simple_rmdir(struct inode *dir, struct dentry *dentry)
296 if (!simple_empty(dentry))
299 drop_nlink(dentry->d_inode);
300 simple_unlink(dir, dentry);
305 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
306 struct inode *new_dir, struct dentry *new_dentry)
308 struct inode *inode = old_dentry->d_inode;
309 int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
311 if (!simple_empty(new_dentry))
314 if (new_dentry->d_inode) {
315 simple_unlink(new_dir, new_dentry);
318 } else if (they_are_dirs) {
323 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
324 new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
330 * simple_setattr - setattr for simple filesystem
332 * @iattr: iattr structure
334 * Returns 0 on success, -error on failure.
336 * simple_setattr is a simple ->setattr implementation without a proper
337 * implementation of size changes.
339 * It can either be used for in-memory filesystems or special files
340 * on simple regular filesystems. Anything that needs to change on-disk
341 * or wire state on size changes needs its own setattr method.
343 int simple_setattr(struct dentry *dentry, struct iattr *iattr)
345 struct inode *inode = dentry->d_inode;
348 WARN_ON_ONCE(inode->i_op->truncate);
350 error = inode_change_ok(inode, iattr);
354 if (iattr->ia_valid & ATTR_SIZE)
355 truncate_setsize(inode, iattr->ia_size);
356 setattr_copy(inode, iattr);
357 mark_inode_dirty(inode);
360 EXPORT_SYMBOL(simple_setattr);
362 int simple_readpage(struct file *file, struct page *page)
364 clear_highpage(page);
365 flush_dcache_page(page);
366 SetPageUptodate(page);
371 int simple_write_begin(struct file *file, struct address_space *mapping,
372 loff_t pos, unsigned len, unsigned flags,
373 struct page **pagep, void **fsdata)
378 index = pos >> PAGE_CACHE_SHIFT;
380 page = grab_cache_page_write_begin(mapping, index, flags);
386 if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) {
387 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
389 zero_user_segments(page, 0, from, from + len, PAGE_CACHE_SIZE);
395 * simple_write_end - .write_end helper for non-block-device FSes
396 * @available: See .write_end of address_space_operations
405 * simple_write_end does the minimum needed for updating a page after writing is
406 * done. It has the same API signature as the .write_end of
407 * address_space_operations vector. So it can just be set onto .write_end for
408 * FSes that don't need any other processing. i_mutex is assumed to be held.
409 * Block based filesystems should use generic_write_end().
410 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
411 * is not called, so a filesystem that actually does store data in .write_inode
412 * should extend on what's done here with a call to mark_inode_dirty() in the
413 * case that i_size has changed.
415 int simple_write_end(struct file *file, struct address_space *mapping,
416 loff_t pos, unsigned len, unsigned copied,
417 struct page *page, void *fsdata)
419 struct inode *inode = page->mapping->host;
420 loff_t last_pos = pos + copied;
422 /* zero the stale part of the page if we did a short copy */
424 unsigned from = pos & (PAGE_CACHE_SIZE - 1);
426 zero_user(page, from + copied, len - copied);
429 if (!PageUptodate(page))
430 SetPageUptodate(page);
432 * No need to use i_size_read() here, the i_size
433 * cannot change under us because we hold the i_mutex.
435 if (last_pos > inode->i_size)
436 i_size_write(inode, last_pos);
438 set_page_dirty(page);
440 page_cache_release(page);
446 * the inodes created here are not hashed. If you use iunique to generate
447 * unique inode values later for this filesystem, then you must take care
448 * to pass it an appropriate max_reserved value to avoid collisions.
450 int simple_fill_super(struct super_block *s, unsigned long magic,
451 struct tree_descr *files)
455 struct dentry *dentry;
458 s->s_blocksize = PAGE_CACHE_SIZE;
459 s->s_blocksize_bits = PAGE_CACHE_SHIFT;
461 s->s_op = &simple_super_operations;
464 inode = new_inode(s);
468 * because the root inode is 1, the files array must not contain an
472 inode->i_mode = S_IFDIR | 0755;
473 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
474 inode->i_op = &simple_dir_inode_operations;
475 inode->i_fop = &simple_dir_operations;
477 root = d_alloc_root(inode);
482 for (i = 0; !files->name || files->name[0]; i++, files++) {
486 /* warn if it tries to conflict with the root inode */
487 if (unlikely(i == 1))
488 printk(KERN_WARNING "%s: %s passed in a files array"
489 "with an index of 1!\n", __func__,
492 dentry = d_alloc_name(root, files->name);
495 inode = new_inode(s);
498 inode->i_mode = S_IFREG | files->mode;
499 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
500 inode->i_fop = files->ops;
502 d_add(dentry, inode);
512 static DEFINE_SPINLOCK(pin_fs_lock);
514 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
516 struct vfsmount *mnt = NULL;
517 spin_lock(&pin_fs_lock);
518 if (unlikely(!*mount)) {
519 spin_unlock(&pin_fs_lock);
520 mnt = vfs_kern_mount(type, 0, type->name, NULL);
523 spin_lock(&pin_fs_lock);
529 spin_unlock(&pin_fs_lock);
534 void simple_release_fs(struct vfsmount **mount, int *count)
536 struct vfsmount *mnt;
537 spin_lock(&pin_fs_lock);
541 spin_unlock(&pin_fs_lock);
546 * simple_read_from_buffer - copy data from the buffer to user space
547 * @to: the user space buffer to read to
548 * @count: the maximum number of bytes to read
549 * @ppos: the current position in the buffer
550 * @from: the buffer to read from
551 * @available: the size of the buffer
553 * The simple_read_from_buffer() function reads up to @count bytes from the
554 * buffer @from at offset @ppos into the user space address starting at @to.
556 * On success, the number of bytes read is returned and the offset @ppos is
557 * advanced by this number, or negative value is returned on error.
559 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
560 const void *from, size_t available)
567 if (pos >= available || !count)
569 if (count > available - pos)
570 count = available - pos;
571 ret = copy_to_user(to, from + pos, count);
580 * simple_write_to_buffer - copy data from user space to the buffer
581 * @to: the buffer to write to
582 * @available: the size of the buffer
583 * @ppos: the current position in the buffer
584 * @from: the user space buffer to read from
585 * @count: the maximum number of bytes to read
587 * The simple_write_to_buffer() function reads up to @count bytes from the user
588 * space address starting at @from into the buffer @to at offset @ppos.
590 * On success, the number of bytes written is returned and the offset @ppos is
591 * advanced by this number, or negative value is returned on error.
593 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
594 const void __user *from, size_t count)
601 if (pos >= available || !count)
603 if (count > available - pos)
604 count = available - pos;
605 res = copy_from_user(to + pos, from, count);
614 * memory_read_from_buffer - copy data from the buffer
615 * @to: the kernel space buffer to read to
616 * @count: the maximum number of bytes to read
617 * @ppos: the current position in the buffer
618 * @from: the buffer to read from
619 * @available: the size of the buffer
621 * The memory_read_from_buffer() function reads up to @count bytes from the
622 * buffer @from at offset @ppos into the kernel space address starting at @to.
624 * On success, the number of bytes read is returned and the offset @ppos is
625 * advanced by this number, or negative value is returned on error.
627 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
628 const void *from, size_t available)
634 if (pos >= available)
636 if (count > available - pos)
637 count = available - pos;
638 memcpy(to, from + pos, count);
645 * Transaction based IO.
646 * The file expects a single write which triggers the transaction, and then
647 * possibly a read which collects the result - which is stored in a
651 void simple_transaction_set(struct file *file, size_t n)
653 struct simple_transaction_argresp *ar = file->private_data;
655 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
658 * The barrier ensures that ar->size will really remain zero until
659 * ar->data is ready for reading.
665 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
667 struct simple_transaction_argresp *ar;
668 static DEFINE_SPINLOCK(simple_transaction_lock);
670 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
671 return ERR_PTR(-EFBIG);
673 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
675 return ERR_PTR(-ENOMEM);
677 spin_lock(&simple_transaction_lock);
679 /* only one write allowed per open */
680 if (file->private_data) {
681 spin_unlock(&simple_transaction_lock);
682 free_page((unsigned long)ar);
683 return ERR_PTR(-EBUSY);
686 file->private_data = ar;
688 spin_unlock(&simple_transaction_lock);
690 if (copy_from_user(ar->data, buf, size))
691 return ERR_PTR(-EFAULT);
696 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
698 struct simple_transaction_argresp *ar = file->private_data;
702 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
705 int simple_transaction_release(struct inode *inode, struct file *file)
707 free_page((unsigned long)file->private_data);
711 /* Simple attribute files */
714 int (*get)(void *, u64 *);
715 int (*set)(void *, u64);
716 char get_buf[24]; /* enough to store a u64 and "\n\0" */
719 const char *fmt; /* format for read operation */
720 struct mutex mutex; /* protects access to these buffers */
723 /* simple_attr_open is called by an actual attribute open file operation
724 * to set the attribute specific access operations. */
725 int simple_attr_open(struct inode *inode, struct file *file,
726 int (*get)(void *, u64 *), int (*set)(void *, u64),
729 struct simple_attr *attr;
731 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
737 attr->data = inode->i_private;
739 mutex_init(&attr->mutex);
741 file->private_data = attr;
743 return nonseekable_open(inode, file);
746 int simple_attr_release(struct inode *inode, struct file *file)
748 kfree(file->private_data);
752 /* read from the buffer that is filled with the get function */
753 ssize_t simple_attr_read(struct file *file, char __user *buf,
754 size_t len, loff_t *ppos)
756 struct simple_attr *attr;
760 attr = file->private_data;
765 ret = mutex_lock_interruptible(&attr->mutex);
769 if (*ppos) { /* continued read */
770 size = strlen(attr->get_buf);
771 } else { /* first read */
773 ret = attr->get(attr->data, &val);
777 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
778 attr->fmt, (unsigned long long)val);
781 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
783 mutex_unlock(&attr->mutex);
787 /* interpret the buffer as a number to call the set function with */
788 ssize_t simple_attr_write(struct file *file, const char __user *buf,
789 size_t len, loff_t *ppos)
791 struct simple_attr *attr;
796 attr = file->private_data;
800 ret = mutex_lock_interruptible(&attr->mutex);
805 size = min(sizeof(attr->set_buf) - 1, len);
806 if (copy_from_user(attr->set_buf, buf, size))
809 attr->set_buf[size] = '\0';
810 val = simple_strtol(attr->set_buf, NULL, 0);
811 ret = attr->set(attr->data, val);
813 ret = len; /* on success, claim we got the whole input */
815 mutex_unlock(&attr->mutex);
820 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
821 * @sb: filesystem to do the file handle conversion on
822 * @fid: file handle to convert
823 * @fh_len: length of the file handle in bytes
824 * @fh_type: type of file handle
825 * @get_inode: filesystem callback to retrieve inode
827 * This function decodes @fid as long as it has one of the well-known
828 * Linux filehandle types and calls @get_inode on it to retrieve the
829 * inode for the object specified in the file handle.
831 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
832 int fh_len, int fh_type, struct inode *(*get_inode)
833 (struct super_block *sb, u64 ino, u32 gen))
835 struct inode *inode = NULL;
841 case FILEID_INO32_GEN:
842 case FILEID_INO32_GEN_PARENT:
843 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
847 return d_obtain_alias(inode);
849 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
852 * generic_fh_to_dentry - generic helper for the fh_to_parent export operation
853 * @sb: filesystem to do the file handle conversion on
854 * @fid: file handle to convert
855 * @fh_len: length of the file handle in bytes
856 * @fh_type: type of file handle
857 * @get_inode: filesystem callback to retrieve inode
859 * This function decodes @fid as long as it has one of the well-known
860 * Linux filehandle types and calls @get_inode on it to retrieve the
861 * inode for the _parent_ object specified in the file handle if it
862 * is specified in the file handle, or NULL otherwise.
864 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
865 int fh_len, int fh_type, struct inode *(*get_inode)
866 (struct super_block *sb, u64 ino, u32 gen))
868 struct inode *inode = NULL;
874 case FILEID_INO32_GEN_PARENT:
875 inode = get_inode(sb, fid->i32.parent_ino,
876 (fh_len > 3 ? fid->i32.parent_gen : 0));
880 return d_obtain_alias(inode);
882 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
885 * generic_file_fsync - generic fsync implementation for simple filesystems
886 * @file: file to synchronize
887 * @datasync: only synchronize essential metadata if true
889 * This is a generic implementation of the fsync method for simple
890 * filesystems which track all non-inode metadata in the buffers list
891 * hanging off the address_space structure.
893 int generic_file_fsync(struct file *file, int datasync)
895 struct writeback_control wbc = {
896 .sync_mode = WB_SYNC_ALL,
897 .nr_to_write = 0, /* metadata-only; caller takes care of data */
899 struct inode *inode = file->f_mapping->host;
903 ret = sync_mapping_buffers(inode->i_mapping);
904 if (!(inode->i_state & I_DIRTY))
906 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
909 err = sync_inode(inode, &wbc);
914 EXPORT_SYMBOL(generic_file_fsync);
917 * No-op implementation of ->fsync for in-memory filesystems.
919 int noop_fsync(struct file *file, int datasync)
924 EXPORT_SYMBOL(dcache_dir_close);
925 EXPORT_SYMBOL(dcache_dir_lseek);
926 EXPORT_SYMBOL(dcache_dir_open);
927 EXPORT_SYMBOL(dcache_readdir);
928 EXPORT_SYMBOL(generic_read_dir);
929 EXPORT_SYMBOL(get_sb_pseudo);
930 EXPORT_SYMBOL(simple_write_begin);
931 EXPORT_SYMBOL(simple_write_end);
932 EXPORT_SYMBOL(simple_dir_inode_operations);
933 EXPORT_SYMBOL(simple_dir_operations);
934 EXPORT_SYMBOL(simple_empty);
935 EXPORT_SYMBOL(simple_fill_super);
936 EXPORT_SYMBOL(simple_getattr);
937 EXPORT_SYMBOL(simple_link);
938 EXPORT_SYMBOL(simple_lookup);
939 EXPORT_SYMBOL(simple_pin_fs);
940 EXPORT_SYMBOL(simple_readpage);
941 EXPORT_SYMBOL(simple_release_fs);
942 EXPORT_SYMBOL(simple_rename);
943 EXPORT_SYMBOL(simple_rmdir);
944 EXPORT_SYMBOL(simple_statfs);
945 EXPORT_SYMBOL(noop_fsync);
946 EXPORT_SYMBOL(simple_unlink);
947 EXPORT_SYMBOL(simple_read_from_buffer);
948 EXPORT_SYMBOL(simple_write_to_buffer);
949 EXPORT_SYMBOL(memory_read_from_buffer);
950 EXPORT_SYMBOL(simple_transaction_set);
951 EXPORT_SYMBOL(simple_transaction_get);
952 EXPORT_SYMBOL(simple_transaction_read);
953 EXPORT_SYMBOL(simple_transaction_release);
954 EXPORT_SYMBOL_GPL(simple_attr_open);
955 EXPORT_SYMBOL_GPL(simple_attr_release);
956 EXPORT_SYMBOL_GPL(simple_attr_read);
957 EXPORT_SYMBOL_GPL(simple_attr_write);