3 * Library for filesystems writers.
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>
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);
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;
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)
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);
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);
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,
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)
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);
118 if (need_resched()) {
119 list_move(&cursor->d_child, p);
120 p = &cursor->d_child;
121 spin_unlock(&dentry->d_lock);
123 spin_lock(&dentry->d_lock);
126 spin_unlock(&dentry->d_lock);
132 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
134 struct dentry *dentry = file->f_path.dentry;
137 offset += file->f_pos;
144 if (offset != file->f_pos) {
145 struct dentry *cursor = file->private_data;
146 struct dentry *to = NULL;
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);
159 spin_lock(&dentry->d_lock);
160 list_del_init(&cursor->d_child);
161 spin_unlock(&dentry->d_lock);
166 inode_unlock_shared(dentry->d_inode);
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;
192 if (!dir_emit_dots(file, ctx))
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))))
206 spin_lock(&dentry->d_lock);
207 list_move_tail(&cursor->d_child, p);
208 spin_unlock(&dentry->d_lock);
213 EXPORT_SYMBOL(dcache_readdir);
215 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
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,
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;
251 struct qstr d_name = QSTR_INIT(name, strlen(name));
253 s = sget_userns(fs_type, NULL, set_anon_super, SB_KERNMOUNT|SB_NOUSER,
254 &init_user_ns, NULL);
258 s->s_maxbytes = MAX_LFS_FILESIZE;
259 s->s_blocksize = PAGE_SIZE;
260 s->s_blocksize_bits = PAGE_SHIFT;
262 s->s_op = ops ? ops : &simple_super_operations;
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).
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);
281 d_instantiate(dentry, root);
284 s->s_flags |= SB_ACTIVE;
285 return dget(s->s_root);
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;
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);
309 d_instantiate(dentry, inode);
312 EXPORT_SYMBOL(simple_link);
314 int simple_empty(struct dentry *dentry)
316 struct dentry *child;
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);
326 spin_unlock(&child->d_lock);
330 spin_unlock(&dentry->d_lock);
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);
344 EXPORT_SYMBOL(simple_unlink);
346 int simple_rmdir(struct inode *dir, struct dentry *dentry)
348 if (!simple_empty(dentry))
351 drop_nlink(d_inode(dentry));
352 simple_unlink(dir, dentry);
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,
362 struct inode *inode = d_inode(old_dentry);
363 int they_are_dirs = d_is_dir(old_dentry);
365 if (flags & ~RENAME_NOREPLACE)
368 if (!simple_empty(new_dentry))
371 if (d_really_is_positive(new_dentry)) {
372 simple_unlink(new_dir, new_dentry);
374 drop_nlink(d_inode(new_dentry));
377 } else if (they_are_dirs) {
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);
387 EXPORT_SYMBOL(simple_rename);
390 * simple_setattr - setattr for simple filesystem
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);
408 error = setattr_prepare(dentry, iattr);
412 if (iattr->ia_valid & ATTR_SIZE)
413 truncate_setsize(inode, iattr->ia_size);
414 setattr_copy(inode, iattr);
415 mark_inode_dirty(inode);
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);
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)
437 index = pos >> PAGE_SHIFT;
439 page = grab_cache_page_write_begin(mapping, index, flags);
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);
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
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)) {
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);
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)
518 struct dentry *dentry;
521 s->s_blocksize = PAGE_SIZE;
522 s->s_blocksize_bits = PAGE_SHIFT;
524 s->s_op = &simple_super_operations;
527 inode = new_inode(s);
531 * because the root inode is 1, the files array must not contain an
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;
540 root = d_make_root(inode);
543 for (i = 0; !files->name || files->name[0]; i++, files++) {
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__,
553 dentry = d_alloc_name(root, files->name);
556 inode = new_inode(s);
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;
565 d_add(dentry, inode);
571 shrink_dcache_parent(root);
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, SB_KERNMOUNT, type->name, NULL);
588 spin_lock(&pin_fs_lock);
594 spin_unlock(&pin_fs_lock);
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);
607 spin_unlock(&pin_fs_lock);
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)
634 if (pos >= available || !count)
636 if (count > available - pos)
637 count = available - pos;
638 ret = copy_to_user(to, from + pos, 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)
669 if (pos >= available || !count)
671 if (count > available - pos)
672 count = available - pos;
673 res = copy_from_user(to + pos, from, 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)
703 if (pos >= available)
705 if (count > available - pos)
706 count = available - pos;
707 memcpy(to, from + pos, 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
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.
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);
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);
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;
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);
783 EXPORT_SYMBOL(simple_transaction_release);
785 /* Simple attribute files */
788 int (*get)(void *, u64 *);
789 int (*set)(void *, u64);
790 char get_buf[24]; /* enough to store a u64 and "\n\0" */
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),
803 struct simple_attr *attr;
805 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
811 attr->data = inode->i_private;
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);
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;
836 attr = file->private_data;
841 ret = mutex_lock_interruptible(&attr->mutex);
845 if (*ppos) { /* continued read */
846 size = strlen(attr->get_buf);
847 } else { /* first read */
849 ret = attr->get(attr->data, &val);
853 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
854 attr->fmt, (unsigned long long)val);
857 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
859 mutex_unlock(&attr->mutex);
862 EXPORT_SYMBOL_GPL(simple_attr_read);
864 /* interpret the buffer as a number to call the set function with */
865 ssize_t simple_attr_write(struct file *file, const char __user *buf,
866 size_t len, loff_t *ppos)
868 struct simple_attr *attr;
873 attr = file->private_data;
877 ret = mutex_lock_interruptible(&attr->mutex);
882 size = min(sizeof(attr->set_buf) - 1, len);
883 if (copy_from_user(attr->set_buf, buf, size))
886 attr->set_buf[size] = '\0';
887 val = simple_strtoll(attr->set_buf, NULL, 0);
888 ret = attr->set(attr->data, val);
890 ret = len; /* on success, claim we got the whole input */
892 mutex_unlock(&attr->mutex);
895 EXPORT_SYMBOL_GPL(simple_attr_write);
898 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
899 * @sb: filesystem to do the file handle conversion on
900 * @fid: file handle to convert
901 * @fh_len: length of the file handle in bytes
902 * @fh_type: type of file handle
903 * @get_inode: filesystem callback to retrieve inode
905 * This function decodes @fid as long as it has one of the well-known
906 * Linux filehandle types and calls @get_inode on it to retrieve the
907 * inode for the object specified in the file handle.
909 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
910 int fh_len, int fh_type, struct inode *(*get_inode)
911 (struct super_block *sb, u64 ino, u32 gen))
913 struct inode *inode = NULL;
919 case FILEID_INO32_GEN:
920 case FILEID_INO32_GEN_PARENT:
921 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
925 return d_obtain_alias(inode);
927 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
930 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
931 * @sb: filesystem to do the file handle conversion on
932 * @fid: file handle to convert
933 * @fh_len: length of the file handle in bytes
934 * @fh_type: type of file handle
935 * @get_inode: filesystem callback to retrieve inode
937 * This function decodes @fid as long as it has one of the well-known
938 * Linux filehandle types and calls @get_inode on it to retrieve the
939 * inode for the _parent_ object specified in the file handle if it
940 * is specified in the file handle, or NULL otherwise.
942 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
943 int fh_len, int fh_type, struct inode *(*get_inode)
944 (struct super_block *sb, u64 ino, u32 gen))
946 struct inode *inode = NULL;
952 case FILEID_INO32_GEN_PARENT:
953 inode = get_inode(sb, fid->i32.parent_ino,
954 (fh_len > 3 ? fid->i32.parent_gen : 0));
958 return d_obtain_alias(inode);
960 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
963 * __generic_file_fsync - generic fsync implementation for simple filesystems
965 * @file: file to synchronize
966 * @start: start offset in bytes
967 * @end: end offset in bytes (inclusive)
968 * @datasync: only synchronize essential metadata if true
970 * This is a generic implementation of the fsync method for simple
971 * filesystems which track all non-inode metadata in the buffers list
972 * hanging off the address_space structure.
974 int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
977 struct inode *inode = file->f_mapping->host;
981 err = file_write_and_wait_range(file, start, end);
986 ret = sync_mapping_buffers(inode->i_mapping);
987 if (!(inode->i_state & I_DIRTY_ALL))
989 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
992 err = sync_inode_metadata(inode, 1);
998 /* check and advance again to catch errors after syncing out buffers */
999 err = file_check_and_advance_wb_err(file);
1004 EXPORT_SYMBOL(__generic_file_fsync);
1007 * generic_file_fsync - generic fsync implementation for simple filesystems
1009 * @file: file to synchronize
1010 * @start: start offset in bytes
1011 * @end: end offset in bytes (inclusive)
1012 * @datasync: only synchronize essential metadata if true
1016 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1019 struct inode *inode = file->f_mapping->host;
1022 err = __generic_file_fsync(file, start, end, datasync);
1025 return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
1027 EXPORT_SYMBOL(generic_file_fsync);
1030 * generic_check_addressable - Check addressability of file system
1031 * @blocksize_bits: log of file system block size
1032 * @num_blocks: number of blocks in file system
1034 * Determine whether a file system with @num_blocks blocks (and a
1035 * block size of 2**@blocksize_bits) is addressable by the sector_t
1036 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1038 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1040 u64 last_fs_block = num_blocks - 1;
1042 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1044 if (unlikely(num_blocks == 0))
1047 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1050 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1051 (last_fs_page > (pgoff_t)(~0ULL))) {
1056 EXPORT_SYMBOL(generic_check_addressable);
1059 * No-op implementation of ->fsync for in-memory filesystems.
1061 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1065 EXPORT_SYMBOL(noop_fsync);
1067 int noop_set_page_dirty(struct page *page)
1070 * Unlike __set_page_dirty_no_writeback that handles dirty page
1071 * tracking in the page object, dax does all dirty tracking in
1072 * the inode address_space in response to mkwrite faults. In the
1073 * dax case we only need to worry about potentially dirty CPU
1074 * caches, not dirty page cache pages to write back.
1076 * This callback is defined to prevent fallback to
1077 * __set_page_dirty_buffers() in set_page_dirty().
1081 EXPORT_SYMBOL_GPL(noop_set_page_dirty);
1083 void noop_invalidatepage(struct page *page, unsigned int offset,
1084 unsigned int length)
1087 * There is no page cache to invalidate in the dax case, however
1088 * we need this callback defined to prevent falling back to
1089 * block_invalidatepage() in do_invalidatepage().
1092 EXPORT_SYMBOL_GPL(noop_invalidatepage);
1094 ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1097 * iomap based filesystems support direct I/O without need for
1098 * this callback. However, it still needs to be set in
1099 * inode->a_ops so that open/fcntl know that direct I/O is
1100 * generally supported.
1104 EXPORT_SYMBOL_GPL(noop_direct_IO);
1106 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1107 void kfree_link(void *p)
1111 EXPORT_SYMBOL(kfree_link);
1114 * nop .set_page_dirty method so that people can use .page_mkwrite on
1117 static int anon_set_page_dirty(struct page *page)
1123 * A single inode exists for all anon_inode files. Contrary to pipes,
1124 * anon_inode inodes have no associated per-instance data, so we need
1125 * only allocate one of them.
1127 struct inode *alloc_anon_inode(struct super_block *s)
1129 static const struct address_space_operations anon_aops = {
1130 .set_page_dirty = anon_set_page_dirty,
1132 struct inode *inode = new_inode_pseudo(s);
1135 return ERR_PTR(-ENOMEM);
1137 inode->i_ino = get_next_ino();
1138 inode->i_mapping->a_ops = &anon_aops;
1141 * Mark the inode dirty from the very beginning,
1142 * that way it will never be moved to the dirty
1143 * list because mark_inode_dirty() will think
1144 * that it already _is_ on the dirty list.
1146 inode->i_state = I_DIRTY;
1147 inode->i_mode = S_IRUSR | S_IWUSR;
1148 inode->i_uid = current_fsuid();
1149 inode->i_gid = current_fsgid();
1150 inode->i_flags |= S_PRIVATE;
1151 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1154 EXPORT_SYMBOL(alloc_anon_inode);
1157 * simple_nosetlease - generic helper for prohibiting leases
1158 * @filp: file pointer
1159 * @arg: type of lease to obtain
1160 * @flp: new lease supplied for insertion
1161 * @priv: private data for lm_setup operation
1163 * Generic helper for filesystems that do not wish to allow leases to be set.
1164 * All arguments are ignored and it just returns -EINVAL.
1167 simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1172 EXPORT_SYMBOL(simple_nosetlease);
1174 const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1175 struct delayed_call *done)
1177 return inode->i_link;
1179 EXPORT_SYMBOL(simple_get_link);
1181 const struct inode_operations simple_symlink_inode_operations = {
1182 .get_link = simple_get_link,
1184 EXPORT_SYMBOL(simple_symlink_inode_operations);
1187 * Operations for a permanently empty directory.
1189 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1191 return ERR_PTR(-ENOENT);
1194 static int empty_dir_getattr(const struct path *path, struct kstat *stat,
1195 u32 request_mask, unsigned int query_flags)
1197 struct inode *inode = d_inode(path->dentry);
1198 generic_fillattr(inode, stat);
1202 static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1207 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1212 static const struct inode_operations empty_dir_inode_operations = {
1213 .lookup = empty_dir_lookup,
1214 .permission = generic_permission,
1215 .setattr = empty_dir_setattr,
1216 .getattr = empty_dir_getattr,
1217 .listxattr = empty_dir_listxattr,
1220 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1222 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1223 return generic_file_llseek_size(file, offset, whence, 2, 2);
1226 static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1228 dir_emit_dots(file, ctx);
1232 static const struct file_operations empty_dir_operations = {
1233 .llseek = empty_dir_llseek,
1234 .read = generic_read_dir,
1235 .iterate_shared = empty_dir_readdir,
1236 .fsync = noop_fsync,
1240 void make_empty_dir_inode(struct inode *inode)
1242 set_nlink(inode, 2);
1243 inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1244 inode->i_uid = GLOBAL_ROOT_UID;
1245 inode->i_gid = GLOBAL_ROOT_GID;
1248 inode->i_blkbits = PAGE_SHIFT;
1249 inode->i_blocks = 0;
1251 inode->i_op = &empty_dir_inode_operations;
1252 inode->i_opflags &= ~IOP_XATTR;
1253 inode->i_fop = &empty_dir_operations;
1256 bool is_empty_dir_inode(struct inode *inode)
1258 return (inode->i_fop == &empty_dir_operations) &&
1259 (inode->i_op == &empty_dir_inode_operations);