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 */
89 static struct dentry *next_positive(struct dentry *parent,
90 struct list_head *from,
93 unsigned *seq = &parent->d_inode->i_dir_seq, n;
102 n = smp_load_acquire(seq) & ~1;
105 for (p = from->next; p != &parent->d_subdirs; p = p->next) {
106 struct dentry *d = list_entry(p, struct dentry, d_child);
107 if (!simple_positive(d)) {
117 if (unlikely(*seq != n))
123 static void move_cursor(struct dentry *cursor, struct list_head *after)
125 struct dentry *parent = cursor->d_parent;
126 unsigned n, *seq = &parent->d_inode->i_dir_seq;
127 spin_lock(&parent->d_lock);
130 if (!(n & 1) && cmpxchg(seq, n, n + 1) == n)
134 __list_del(cursor->d_child.prev, cursor->d_child.next);
136 list_add(&cursor->d_child, after);
138 list_add_tail(&cursor->d_child, &parent->d_subdirs);
139 smp_store_release(seq, n + 2);
140 spin_unlock(&parent->d_lock);
143 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
145 struct dentry *dentry = file->f_path.dentry;
148 offset += file->f_pos;
157 if (offset != file->f_pos) {
158 file->f_pos = offset;
159 if (file->f_pos >= 2) {
160 struct dentry *cursor = file->private_data;
162 loff_t n = file->f_pos - 2;
164 inode_lock_shared(dentry->d_inode);
165 to = next_positive(dentry, &dentry->d_subdirs, n);
166 move_cursor(cursor, to ? &to->d_child : NULL);
167 inode_unlock_shared(dentry->d_inode);
172 EXPORT_SYMBOL(dcache_dir_lseek);
174 /* Relationship between i_mode and the DT_xxx types */
175 static inline unsigned char dt_type(struct inode *inode)
177 return (inode->i_mode >> 12) & 15;
181 * Directory is locked and all positive dentries in it are safe, since
182 * for ramfs-type trees they can't go away without unlink() or rmdir(),
183 * both impossible due to the lock on directory.
186 int dcache_readdir(struct file *file, struct dir_context *ctx)
188 struct dentry *dentry = file->f_path.dentry;
189 struct dentry *cursor = file->private_data;
190 struct list_head *p = &cursor->d_child;
194 if (!dir_emit_dots(file, ctx))
198 p = &dentry->d_subdirs;
199 while ((next = next_positive(dentry, p, 1)) != NULL) {
200 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
201 d_inode(next)->i_ino, dt_type(d_inode(next))))
208 move_cursor(cursor, p);
211 EXPORT_SYMBOL(dcache_readdir);
213 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
217 EXPORT_SYMBOL(generic_read_dir);
219 const struct file_operations simple_dir_operations = {
220 .open = dcache_dir_open,
221 .release = dcache_dir_close,
222 .llseek = dcache_dir_lseek,
223 .read = generic_read_dir,
224 .iterate_shared = dcache_readdir,
227 EXPORT_SYMBOL(simple_dir_operations);
229 const struct inode_operations simple_dir_inode_operations = {
230 .lookup = simple_lookup,
232 EXPORT_SYMBOL(simple_dir_inode_operations);
234 static const struct super_operations simple_super_operations = {
235 .statfs = simple_statfs,
239 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
240 * will never be mountable)
242 struct dentry *mount_pseudo_xattr(struct file_system_type *fs_type, char *name,
243 const struct super_operations *ops, const struct xattr_handler **xattr,
244 const struct dentry_operations *dops, unsigned long magic)
246 struct super_block *s;
247 struct dentry *dentry;
249 struct qstr d_name = QSTR_INIT(name, strlen(name));
251 s = sget_userns(fs_type, NULL, set_anon_super, SB_KERNMOUNT|SB_NOUSER,
252 &init_user_ns, NULL);
256 s->s_maxbytes = MAX_LFS_FILESIZE;
257 s->s_blocksize = PAGE_SIZE;
258 s->s_blocksize_bits = PAGE_SHIFT;
260 s->s_op = ops ? ops : &simple_super_operations;
267 * since this is the first inode, make it number 1. New inodes created
268 * after this must take care not to collide with it (by passing
269 * max_reserved of 1 to iunique).
272 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
273 root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
274 dentry = __d_alloc(s, &d_name);
279 d_instantiate(dentry, root);
282 s->s_flags |= SB_ACTIVE;
283 return dget(s->s_root);
286 deactivate_locked_super(s);
287 return ERR_PTR(-ENOMEM);
289 EXPORT_SYMBOL(mount_pseudo_xattr);
291 int simple_open(struct inode *inode, struct file *file)
293 if (inode->i_private)
294 file->private_data = inode->i_private;
297 EXPORT_SYMBOL(simple_open);
299 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
301 struct inode *inode = d_inode(old_dentry);
303 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
307 d_instantiate(dentry, inode);
310 EXPORT_SYMBOL(simple_link);
312 int simple_empty(struct dentry *dentry)
314 struct dentry *child;
317 spin_lock(&dentry->d_lock);
318 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
319 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
320 if (simple_positive(child)) {
321 spin_unlock(&child->d_lock);
324 spin_unlock(&child->d_lock);
328 spin_unlock(&dentry->d_lock);
331 EXPORT_SYMBOL(simple_empty);
333 int simple_unlink(struct inode *dir, struct dentry *dentry)
335 struct inode *inode = d_inode(dentry);
337 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
342 EXPORT_SYMBOL(simple_unlink);
344 int simple_rmdir(struct inode *dir, struct dentry *dentry)
346 if (!simple_empty(dentry))
349 drop_nlink(d_inode(dentry));
350 simple_unlink(dir, dentry);
354 EXPORT_SYMBOL(simple_rmdir);
356 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
357 struct inode *new_dir, struct dentry *new_dentry,
360 struct inode *inode = d_inode(old_dentry);
361 int they_are_dirs = d_is_dir(old_dentry);
363 if (flags & ~RENAME_NOREPLACE)
366 if (!simple_empty(new_dentry))
369 if (d_really_is_positive(new_dentry)) {
370 simple_unlink(new_dir, new_dentry);
372 drop_nlink(d_inode(new_dentry));
375 } else if (they_are_dirs) {
380 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
381 new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
385 EXPORT_SYMBOL(simple_rename);
388 * simple_setattr - setattr for simple filesystem
390 * @iattr: iattr structure
392 * Returns 0 on success, -error on failure.
394 * simple_setattr is a simple ->setattr implementation without a proper
395 * implementation of size changes.
397 * It can either be used for in-memory filesystems or special files
398 * on simple regular filesystems. Anything that needs to change on-disk
399 * or wire state on size changes needs its own setattr method.
401 int simple_setattr(struct dentry *dentry, struct iattr *iattr)
403 struct inode *inode = d_inode(dentry);
406 error = setattr_prepare(dentry, iattr);
410 if (iattr->ia_valid & ATTR_SIZE)
411 truncate_setsize(inode, iattr->ia_size);
412 setattr_copy(inode, iattr);
413 mark_inode_dirty(inode);
416 EXPORT_SYMBOL(simple_setattr);
418 int simple_readpage(struct file *file, struct page *page)
420 clear_highpage(page);
421 flush_dcache_page(page);
422 SetPageUptodate(page);
426 EXPORT_SYMBOL(simple_readpage);
428 int simple_write_begin(struct file *file, struct address_space *mapping,
429 loff_t pos, unsigned len, unsigned flags,
430 struct page **pagep, void **fsdata)
435 index = pos >> PAGE_SHIFT;
437 page = grab_cache_page_write_begin(mapping, index, flags);
443 if (!PageUptodate(page) && (len != PAGE_SIZE)) {
444 unsigned from = pos & (PAGE_SIZE - 1);
446 zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
450 EXPORT_SYMBOL(simple_write_begin);
453 * simple_write_end - .write_end helper for non-block-device FSes
454 * @available: See .write_end of address_space_operations
463 * simple_write_end does the minimum needed for updating a page after writing is
464 * done. It has the same API signature as the .write_end of
465 * address_space_operations vector. So it can just be set onto .write_end for
466 * FSes that don't need any other processing. i_mutex is assumed to be held.
467 * Block based filesystems should use generic_write_end().
468 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
469 * is not called, so a filesystem that actually does store data in .write_inode
470 * should extend on what's done here with a call to mark_inode_dirty() in the
471 * case that i_size has changed.
473 * Use *ONLY* with simple_readpage()
475 int simple_write_end(struct file *file, struct address_space *mapping,
476 loff_t pos, unsigned len, unsigned copied,
477 struct page *page, void *fsdata)
479 struct inode *inode = page->mapping->host;
480 loff_t last_pos = pos + copied;
482 /* zero the stale part of the page if we did a short copy */
483 if (!PageUptodate(page)) {
485 unsigned from = pos & (PAGE_SIZE - 1);
487 zero_user(page, from + copied, len - copied);
489 SetPageUptodate(page);
492 * No need to use i_size_read() here, the i_size
493 * cannot change under us because we hold the i_mutex.
495 if (last_pos > inode->i_size)
496 i_size_write(inode, last_pos);
498 set_page_dirty(page);
504 EXPORT_SYMBOL(simple_write_end);
507 * the inodes created here are not hashed. If you use iunique to generate
508 * unique inode values later for this filesystem, then you must take care
509 * to pass it an appropriate max_reserved value to avoid collisions.
511 int simple_fill_super(struct super_block *s, unsigned long magic,
512 const struct tree_descr *files)
516 struct dentry *dentry;
519 s->s_blocksize = PAGE_SIZE;
520 s->s_blocksize_bits = PAGE_SHIFT;
522 s->s_op = &simple_super_operations;
525 inode = new_inode(s);
529 * because the root inode is 1, the files array must not contain an
533 inode->i_mode = S_IFDIR | 0755;
534 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
535 inode->i_op = &simple_dir_inode_operations;
536 inode->i_fop = &simple_dir_operations;
538 root = d_make_root(inode);
541 for (i = 0; !files->name || files->name[0]; i++, files++) {
545 /* warn if it tries to conflict with the root inode */
546 if (unlikely(i == 1))
547 printk(KERN_WARNING "%s: %s passed in a files array"
548 "with an index of 1!\n", __func__,
551 dentry = d_alloc_name(root, files->name);
554 inode = new_inode(s);
559 inode->i_mode = S_IFREG | files->mode;
560 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
561 inode->i_fop = files->ops;
563 d_add(dentry, inode);
569 shrink_dcache_parent(root);
573 EXPORT_SYMBOL(simple_fill_super);
575 static DEFINE_SPINLOCK(pin_fs_lock);
577 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
579 struct vfsmount *mnt = NULL;
580 spin_lock(&pin_fs_lock);
581 if (unlikely(!*mount)) {
582 spin_unlock(&pin_fs_lock);
583 mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
586 spin_lock(&pin_fs_lock);
592 spin_unlock(&pin_fs_lock);
596 EXPORT_SYMBOL(simple_pin_fs);
598 void simple_release_fs(struct vfsmount **mount, int *count)
600 struct vfsmount *mnt;
601 spin_lock(&pin_fs_lock);
605 spin_unlock(&pin_fs_lock);
608 EXPORT_SYMBOL(simple_release_fs);
611 * simple_read_from_buffer - copy data from the buffer to user space
612 * @to: the user space buffer to read to
613 * @count: the maximum number of bytes to read
614 * @ppos: the current position in the buffer
615 * @from: the buffer to read from
616 * @available: the size of the buffer
618 * The simple_read_from_buffer() function reads up to @count bytes from the
619 * buffer @from at offset @ppos into the user space address starting at @to.
621 * On success, the number of bytes read is returned and the offset @ppos is
622 * advanced by this number, or negative value is returned on error.
624 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
625 const void *from, size_t available)
632 if (pos >= available || !count)
634 if (count > available - pos)
635 count = available - pos;
636 ret = copy_to_user(to, from + pos, count);
643 EXPORT_SYMBOL(simple_read_from_buffer);
646 * simple_write_to_buffer - copy data from user space to the buffer
647 * @to: the buffer to write to
648 * @available: the size of the buffer
649 * @ppos: the current position in the buffer
650 * @from: the user space buffer to read from
651 * @count: the maximum number of bytes to read
653 * The simple_write_to_buffer() function reads up to @count bytes from the user
654 * space address starting at @from into the buffer @to at offset @ppos.
656 * On success, the number of bytes written is returned and the offset @ppos is
657 * advanced by this number, or negative value is returned on error.
659 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
660 const void __user *from, size_t count)
667 if (pos >= available || !count)
669 if (count > available - pos)
670 count = available - pos;
671 res = copy_from_user(to + pos, from, count);
678 EXPORT_SYMBOL(simple_write_to_buffer);
681 * memory_read_from_buffer - copy data from the buffer
682 * @to: the kernel space buffer to read to
683 * @count: the maximum number of bytes to read
684 * @ppos: the current position in the buffer
685 * @from: the buffer to read from
686 * @available: the size of the buffer
688 * The memory_read_from_buffer() function reads up to @count bytes from the
689 * buffer @from at offset @ppos into the kernel space address starting at @to.
691 * On success, the number of bytes read is returned and the offset @ppos is
692 * advanced by this number, or negative value is returned on error.
694 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
695 const void *from, size_t available)
701 if (pos >= available)
703 if (count > available - pos)
704 count = available - pos;
705 memcpy(to, from + pos, count);
710 EXPORT_SYMBOL(memory_read_from_buffer);
713 * Transaction based IO.
714 * The file expects a single write which triggers the transaction, and then
715 * possibly a read which collects the result - which is stored in a
719 void simple_transaction_set(struct file *file, size_t n)
721 struct simple_transaction_argresp *ar = file->private_data;
723 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
726 * The barrier ensures that ar->size will really remain zero until
727 * ar->data is ready for reading.
732 EXPORT_SYMBOL(simple_transaction_set);
734 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
736 struct simple_transaction_argresp *ar;
737 static DEFINE_SPINLOCK(simple_transaction_lock);
739 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
740 return ERR_PTR(-EFBIG);
742 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
744 return ERR_PTR(-ENOMEM);
746 spin_lock(&simple_transaction_lock);
748 /* only one write allowed per open */
749 if (file->private_data) {
750 spin_unlock(&simple_transaction_lock);
751 free_page((unsigned long)ar);
752 return ERR_PTR(-EBUSY);
755 file->private_data = ar;
757 spin_unlock(&simple_transaction_lock);
759 if (copy_from_user(ar->data, buf, size))
760 return ERR_PTR(-EFAULT);
764 EXPORT_SYMBOL(simple_transaction_get);
766 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
768 struct simple_transaction_argresp *ar = file->private_data;
772 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
774 EXPORT_SYMBOL(simple_transaction_read);
776 int simple_transaction_release(struct inode *inode, struct file *file)
778 free_page((unsigned long)file->private_data);
781 EXPORT_SYMBOL(simple_transaction_release);
783 /* Simple attribute files */
786 int (*get)(void *, u64 *);
787 int (*set)(void *, u64);
788 char get_buf[24]; /* enough to store a u64 and "\n\0" */
791 const char *fmt; /* format for read operation */
792 struct mutex mutex; /* protects access to these buffers */
795 /* simple_attr_open is called by an actual attribute open file operation
796 * to set the attribute specific access operations. */
797 int simple_attr_open(struct inode *inode, struct file *file,
798 int (*get)(void *, u64 *), int (*set)(void *, u64),
801 struct simple_attr *attr;
803 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
809 attr->data = inode->i_private;
811 mutex_init(&attr->mutex);
813 file->private_data = attr;
815 return nonseekable_open(inode, file);
817 EXPORT_SYMBOL_GPL(simple_attr_open);
819 int simple_attr_release(struct inode *inode, struct file *file)
821 kfree(file->private_data);
824 EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
826 /* read from the buffer that is filled with the get function */
827 ssize_t simple_attr_read(struct file *file, char __user *buf,
828 size_t len, loff_t *ppos)
830 struct simple_attr *attr;
834 attr = file->private_data;
839 ret = mutex_lock_interruptible(&attr->mutex);
843 if (*ppos) { /* continued read */
844 size = strlen(attr->get_buf);
845 } else { /* first read */
847 ret = attr->get(attr->data, &val);
851 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
852 attr->fmt, (unsigned long long)val);
855 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
857 mutex_unlock(&attr->mutex);
860 EXPORT_SYMBOL_GPL(simple_attr_read);
862 /* interpret the buffer as a number to call the set function with */
863 ssize_t simple_attr_write(struct file *file, const char __user *buf,
864 size_t len, loff_t *ppos)
866 struct simple_attr *attr;
871 attr = file->private_data;
875 ret = mutex_lock_interruptible(&attr->mutex);
880 size = min(sizeof(attr->set_buf) - 1, len);
881 if (copy_from_user(attr->set_buf, buf, size))
884 attr->set_buf[size] = '\0';
885 val = simple_strtoll(attr->set_buf, NULL, 0);
886 ret = attr->set(attr->data, val);
888 ret = len; /* on success, claim we got the whole input */
890 mutex_unlock(&attr->mutex);
893 EXPORT_SYMBOL_GPL(simple_attr_write);
896 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
897 * @sb: filesystem to do the file handle conversion on
898 * @fid: file handle to convert
899 * @fh_len: length of the file handle in bytes
900 * @fh_type: type of file handle
901 * @get_inode: filesystem callback to retrieve inode
903 * This function decodes @fid as long as it has one of the well-known
904 * Linux filehandle types and calls @get_inode on it to retrieve the
905 * inode for the object specified in the file handle.
907 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
908 int fh_len, int fh_type, struct inode *(*get_inode)
909 (struct super_block *sb, u64 ino, u32 gen))
911 struct inode *inode = NULL;
917 case FILEID_INO32_GEN:
918 case FILEID_INO32_GEN_PARENT:
919 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
923 return d_obtain_alias(inode);
925 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
928 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
929 * @sb: filesystem to do the file handle conversion on
930 * @fid: file handle to convert
931 * @fh_len: length of the file handle in bytes
932 * @fh_type: type of file handle
933 * @get_inode: filesystem callback to retrieve inode
935 * This function decodes @fid as long as it has one of the well-known
936 * Linux filehandle types and calls @get_inode on it to retrieve the
937 * inode for the _parent_ object specified in the file handle if it
938 * is specified in the file handle, or NULL otherwise.
940 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
941 int fh_len, int fh_type, struct inode *(*get_inode)
942 (struct super_block *sb, u64 ino, u32 gen))
944 struct inode *inode = NULL;
950 case FILEID_INO32_GEN_PARENT:
951 inode = get_inode(sb, fid->i32.parent_ino,
952 (fh_len > 3 ? fid->i32.parent_gen : 0));
956 return d_obtain_alias(inode);
958 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
961 * __generic_file_fsync - generic fsync implementation for simple filesystems
963 * @file: file to synchronize
964 * @start: start offset in bytes
965 * @end: end offset in bytes (inclusive)
966 * @datasync: only synchronize essential metadata if true
968 * This is a generic implementation of the fsync method for simple
969 * filesystems which track all non-inode metadata in the buffers list
970 * hanging off the address_space structure.
972 int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
975 struct inode *inode = file->f_mapping->host;
979 err = file_write_and_wait_range(file, start, end);
984 ret = sync_mapping_buffers(inode->i_mapping);
985 if (!(inode->i_state & I_DIRTY_ALL))
987 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
990 err = sync_inode_metadata(inode, 1);
996 /* check and advance again to catch errors after syncing out buffers */
997 err = file_check_and_advance_wb_err(file);
1002 EXPORT_SYMBOL(__generic_file_fsync);
1005 * generic_file_fsync - generic fsync implementation for simple filesystems
1007 * @file: file to synchronize
1008 * @start: start offset in bytes
1009 * @end: end offset in bytes (inclusive)
1010 * @datasync: only synchronize essential metadata if true
1014 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1017 struct inode *inode = file->f_mapping->host;
1020 err = __generic_file_fsync(file, start, end, datasync);
1023 return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
1025 EXPORT_SYMBOL(generic_file_fsync);
1028 * generic_check_addressable - Check addressability of file system
1029 * @blocksize_bits: log of file system block size
1030 * @num_blocks: number of blocks in file system
1032 * Determine whether a file system with @num_blocks blocks (and a
1033 * block size of 2**@blocksize_bits) is addressable by the sector_t
1034 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1036 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1038 u64 last_fs_block = num_blocks - 1;
1040 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1042 if (unlikely(num_blocks == 0))
1045 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1048 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1049 (last_fs_page > (pgoff_t)(~0ULL))) {
1054 EXPORT_SYMBOL(generic_check_addressable);
1057 * No-op implementation of ->fsync for in-memory filesystems.
1059 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1063 EXPORT_SYMBOL(noop_fsync);
1065 int noop_set_page_dirty(struct page *page)
1068 * Unlike __set_page_dirty_no_writeback that handles dirty page
1069 * tracking in the page object, dax does all dirty tracking in
1070 * the inode address_space in response to mkwrite faults. In the
1071 * dax case we only need to worry about potentially dirty CPU
1072 * caches, not dirty page cache pages to write back.
1074 * This callback is defined to prevent fallback to
1075 * __set_page_dirty_buffers() in set_page_dirty().
1079 EXPORT_SYMBOL_GPL(noop_set_page_dirty);
1081 void noop_invalidatepage(struct page *page, unsigned int offset,
1082 unsigned int length)
1085 * There is no page cache to invalidate in the dax case, however
1086 * we need this callback defined to prevent falling back to
1087 * block_invalidatepage() in do_invalidatepage().
1090 EXPORT_SYMBOL_GPL(noop_invalidatepage);
1092 ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1095 * iomap based filesystems support direct I/O without need for
1096 * this callback. However, it still needs to be set in
1097 * inode->a_ops so that open/fcntl know that direct I/O is
1098 * generally supported.
1102 EXPORT_SYMBOL_GPL(noop_direct_IO);
1104 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1105 void kfree_link(void *p)
1109 EXPORT_SYMBOL(kfree_link);
1112 * nop .set_page_dirty method so that people can use .page_mkwrite on
1115 static int anon_set_page_dirty(struct page *page)
1121 * A single inode exists for all anon_inode files. Contrary to pipes,
1122 * anon_inode inodes have no associated per-instance data, so we need
1123 * only allocate one of them.
1125 struct inode *alloc_anon_inode(struct super_block *s)
1127 static const struct address_space_operations anon_aops = {
1128 .set_page_dirty = anon_set_page_dirty,
1130 struct inode *inode = new_inode_pseudo(s);
1133 return ERR_PTR(-ENOMEM);
1135 inode->i_ino = get_next_ino();
1136 inode->i_mapping->a_ops = &anon_aops;
1139 * Mark the inode dirty from the very beginning,
1140 * that way it will never be moved to the dirty
1141 * list because mark_inode_dirty() will think
1142 * that it already _is_ on the dirty list.
1144 inode->i_state = I_DIRTY;
1145 inode->i_mode = S_IRUSR | S_IWUSR;
1146 inode->i_uid = current_fsuid();
1147 inode->i_gid = current_fsgid();
1148 inode->i_flags |= S_PRIVATE;
1149 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1152 EXPORT_SYMBOL(alloc_anon_inode);
1155 * simple_nosetlease - generic helper for prohibiting leases
1156 * @filp: file pointer
1157 * @arg: type of lease to obtain
1158 * @flp: new lease supplied for insertion
1159 * @priv: private data for lm_setup operation
1161 * Generic helper for filesystems that do not wish to allow leases to be set.
1162 * All arguments are ignored and it just returns -EINVAL.
1165 simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1170 EXPORT_SYMBOL(simple_nosetlease);
1173 * simple_get_link - generic helper to get the target of "fast" symlinks
1174 * @dentry: not used here
1175 * @inode: the symlink inode
1176 * @done: not used here
1178 * Generic helper for filesystems to use for symlink inodes where a pointer to
1179 * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1180 * since as an optimization the path lookup code uses any non-NULL ->i_link
1181 * directly, without calling ->get_link(). But ->get_link() still must be set,
1182 * to mark the inode_operations as being for a symlink.
1184 * Return: the symlink target
1186 const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1187 struct delayed_call *done)
1189 return inode->i_link;
1191 EXPORT_SYMBOL(simple_get_link);
1193 const struct inode_operations simple_symlink_inode_operations = {
1194 .get_link = simple_get_link,
1196 EXPORT_SYMBOL(simple_symlink_inode_operations);
1199 * Operations for a permanently empty directory.
1201 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1203 return ERR_PTR(-ENOENT);
1206 static int empty_dir_getattr(const struct path *path, struct kstat *stat,
1207 u32 request_mask, unsigned int query_flags)
1209 struct inode *inode = d_inode(path->dentry);
1210 generic_fillattr(inode, stat);
1214 static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1219 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1224 static const struct inode_operations empty_dir_inode_operations = {
1225 .lookup = empty_dir_lookup,
1226 .permission = generic_permission,
1227 .setattr = empty_dir_setattr,
1228 .getattr = empty_dir_getattr,
1229 .listxattr = empty_dir_listxattr,
1232 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1234 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1235 return generic_file_llseek_size(file, offset, whence, 2, 2);
1238 static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1240 dir_emit_dots(file, ctx);
1244 static const struct file_operations empty_dir_operations = {
1245 .llseek = empty_dir_llseek,
1246 .read = generic_read_dir,
1247 .iterate_shared = empty_dir_readdir,
1248 .fsync = noop_fsync,
1252 void make_empty_dir_inode(struct inode *inode)
1254 set_nlink(inode, 2);
1255 inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1256 inode->i_uid = GLOBAL_ROOT_UID;
1257 inode->i_gid = GLOBAL_ROOT_GID;
1260 inode->i_blkbits = PAGE_SHIFT;
1261 inode->i_blocks = 0;
1263 inode->i_op = &empty_dir_inode_operations;
1264 inode->i_opflags &= ~IOP_XATTR;
1265 inode->i_fop = &empty_dir_operations;
1268 bool is_empty_dir_inode(struct inode *inode)
1270 return (inode->i_fop == &empty_dir_operations) &&
1271 (inode->i_op == &empty_dir_inode_operations);