1 // SPDX-License-Identifier: GPL-2.0-only
4 * Library for filesystems writers.
7 #include <linux/blkdev.h>
8 #include <linux/export.h>
9 #include <linux/pagemap.h>
10 #include <linux/slab.h>
11 #include <linux/cred.h>
12 #include <linux/mount.h>
13 #include <linux/vfs.h>
14 #include <linux/quotaops.h>
15 #include <linux/mutex.h>
16 #include <linux/namei.h>
17 #include <linux/exportfs.h>
18 #include <linux/writeback.h>
19 #include <linux/buffer_head.h> /* sync_mapping_buffers */
21 #include <linux/uaccess.h>
25 int simple_getattr(const struct path *path, struct kstat *stat,
26 u32 request_mask, unsigned int query_flags)
28 struct inode *inode = d_inode(path->dentry);
29 generic_fillattr(inode, stat);
30 stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
33 EXPORT_SYMBOL(simple_getattr);
35 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
37 buf->f_type = dentry->d_sb->s_magic;
38 buf->f_bsize = PAGE_SIZE;
39 buf->f_namelen = NAME_MAX;
42 EXPORT_SYMBOL(simple_statfs);
45 * Retaining negative dentries for an in-memory filesystem just wastes
46 * memory and lookup time: arrange for them to be deleted immediately.
48 int always_delete_dentry(const struct dentry *dentry)
52 EXPORT_SYMBOL(always_delete_dentry);
54 const struct dentry_operations simple_dentry_operations = {
55 .d_delete = always_delete_dentry,
57 EXPORT_SYMBOL(simple_dentry_operations);
60 * Lookup the data. This is trivial - if the dentry didn't already
61 * exist, we know it is negative. Set d_op to delete negative dentries.
63 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
65 if (dentry->d_name.len > NAME_MAX)
66 return ERR_PTR(-ENAMETOOLONG);
67 if (!dentry->d_sb->s_d_op)
68 d_set_d_op(dentry, &simple_dentry_operations);
72 EXPORT_SYMBOL(simple_lookup);
74 int dcache_dir_open(struct inode *inode, struct file *file)
76 file->private_data = d_alloc_cursor(file->f_path.dentry);
78 return file->private_data ? 0 : -ENOMEM;
80 EXPORT_SYMBOL(dcache_dir_open);
82 int dcache_dir_close(struct inode *inode, struct file *file)
84 dput(file->private_data);
87 EXPORT_SYMBOL(dcache_dir_close);
89 /* parent is locked at least shared */
90 static struct dentry *next_positive(struct dentry *parent,
91 struct list_head *from,
94 unsigned *seq = &parent->d_inode->i_dir_seq, n;
103 n = smp_load_acquire(seq) & ~1;
106 for (p = from->next; p != &parent->d_subdirs; p = p->next) {
107 struct dentry *d = list_entry(p, struct dentry, d_child);
108 if (!simple_positive(d)) {
118 if (unlikely(*seq != n))
124 static void move_cursor(struct dentry *cursor, struct list_head *after)
126 struct dentry *parent = cursor->d_parent;
127 unsigned n, *seq = &parent->d_inode->i_dir_seq;
128 spin_lock(&parent->d_lock);
131 if (!(n & 1) && cmpxchg(seq, n, n + 1) == n)
135 __list_del(cursor->d_child.prev, cursor->d_child.next);
137 list_add(&cursor->d_child, after);
139 list_add_tail(&cursor->d_child, &parent->d_subdirs);
140 smp_store_release(seq, n + 2);
141 spin_unlock(&parent->d_lock);
144 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
146 struct dentry *dentry = file->f_path.dentry;
149 offset += file->f_pos;
158 if (offset != file->f_pos) {
159 file->f_pos = offset;
160 if (file->f_pos >= 2) {
161 struct dentry *cursor = file->private_data;
163 loff_t n = file->f_pos - 2;
165 inode_lock_shared(dentry->d_inode);
166 to = next_positive(dentry, &dentry->d_subdirs, n);
167 move_cursor(cursor, to ? &to->d_child : NULL);
168 inode_unlock_shared(dentry->d_inode);
173 EXPORT_SYMBOL(dcache_dir_lseek);
175 /* Relationship between i_mode and the DT_xxx types */
176 static inline unsigned char dt_type(struct inode *inode)
178 return (inode->i_mode >> 12) & 15;
182 * Directory is locked and all positive dentries in it are safe, since
183 * for ramfs-type trees they can't go away without unlink() or rmdir(),
184 * both impossible due to the lock on directory.
187 int dcache_readdir(struct file *file, struct dir_context *ctx)
189 struct dentry *dentry = file->f_path.dentry;
190 struct dentry *cursor = file->private_data;
191 struct list_head *p = &cursor->d_child;
195 if (!dir_emit_dots(file, ctx))
199 p = &dentry->d_subdirs;
200 while ((next = next_positive(dentry, p, 1)) != NULL) {
201 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
202 d_inode(next)->i_ino, dt_type(d_inode(next))))
209 move_cursor(cursor, p);
212 EXPORT_SYMBOL(dcache_readdir);
214 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
218 EXPORT_SYMBOL(generic_read_dir);
220 const struct file_operations simple_dir_operations = {
221 .open = dcache_dir_open,
222 .release = dcache_dir_close,
223 .llseek = dcache_dir_lseek,
224 .read = generic_read_dir,
225 .iterate_shared = dcache_readdir,
228 EXPORT_SYMBOL(simple_dir_operations);
230 const struct inode_operations simple_dir_inode_operations = {
231 .lookup = simple_lookup,
233 EXPORT_SYMBOL(simple_dir_inode_operations);
235 static const struct super_operations simple_super_operations = {
236 .statfs = simple_statfs,
240 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
241 * will never be mountable)
243 struct dentry *mount_pseudo_xattr(struct file_system_type *fs_type, char *name,
244 const struct super_operations *ops, const struct xattr_handler **xattr,
245 const struct dentry_operations *dops, unsigned long magic)
247 struct super_block *s;
248 struct dentry *dentry;
250 struct qstr d_name = QSTR_INIT(name, strlen(name));
252 s = sget_userns(fs_type, NULL, set_anon_super, SB_KERNMOUNT|SB_NOUSER,
253 &init_user_ns, NULL);
257 s->s_maxbytes = MAX_LFS_FILESIZE;
258 s->s_blocksize = PAGE_SIZE;
259 s->s_blocksize_bits = PAGE_SHIFT;
261 s->s_op = ops ? ops : &simple_super_operations;
268 * since this is the first inode, make it number 1. New inodes created
269 * after this must take care not to collide with it (by passing
270 * max_reserved of 1 to iunique).
273 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
274 root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
275 dentry = __d_alloc(s, &d_name);
280 d_instantiate(dentry, root);
283 s->s_flags |= SB_ACTIVE;
284 return dget(s->s_root);
287 deactivate_locked_super(s);
288 return ERR_PTR(-ENOMEM);
290 EXPORT_SYMBOL(mount_pseudo_xattr);
292 int simple_open(struct inode *inode, struct file *file)
294 if (inode->i_private)
295 file->private_data = inode->i_private;
298 EXPORT_SYMBOL(simple_open);
300 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
302 struct inode *inode = d_inode(old_dentry);
304 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
308 d_instantiate(dentry, inode);
311 EXPORT_SYMBOL(simple_link);
313 int simple_empty(struct dentry *dentry)
315 struct dentry *child;
318 spin_lock(&dentry->d_lock);
319 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
320 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
321 if (simple_positive(child)) {
322 spin_unlock(&child->d_lock);
325 spin_unlock(&child->d_lock);
329 spin_unlock(&dentry->d_lock);
332 EXPORT_SYMBOL(simple_empty);
334 int simple_unlink(struct inode *dir, struct dentry *dentry)
336 struct inode *inode = d_inode(dentry);
338 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
343 EXPORT_SYMBOL(simple_unlink);
345 int simple_rmdir(struct inode *dir, struct dentry *dentry)
347 if (!simple_empty(dentry))
350 drop_nlink(d_inode(dentry));
351 simple_unlink(dir, dentry);
355 EXPORT_SYMBOL(simple_rmdir);
357 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
358 struct inode *new_dir, struct dentry *new_dentry,
361 struct inode *inode = d_inode(old_dentry);
362 int they_are_dirs = d_is_dir(old_dentry);
364 if (flags & ~RENAME_NOREPLACE)
367 if (!simple_empty(new_dentry))
370 if (d_really_is_positive(new_dentry)) {
371 simple_unlink(new_dir, new_dentry);
373 drop_nlink(d_inode(new_dentry));
376 } else if (they_are_dirs) {
381 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
382 new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
386 EXPORT_SYMBOL(simple_rename);
389 * simple_setattr - setattr for simple filesystem
391 * @iattr: iattr structure
393 * Returns 0 on success, -error on failure.
395 * simple_setattr is a simple ->setattr implementation without a proper
396 * implementation of size changes.
398 * It can either be used for in-memory filesystems or special files
399 * on simple regular filesystems. Anything that needs to change on-disk
400 * or wire state on size changes needs its own setattr method.
402 int simple_setattr(struct dentry *dentry, struct iattr *iattr)
404 struct inode *inode = d_inode(dentry);
407 error = setattr_prepare(dentry, iattr);
411 if (iattr->ia_valid & ATTR_SIZE)
412 truncate_setsize(inode, iattr->ia_size);
413 setattr_copy(inode, iattr);
414 mark_inode_dirty(inode);
417 EXPORT_SYMBOL(simple_setattr);
419 int simple_readpage(struct file *file, struct page *page)
421 clear_highpage(page);
422 flush_dcache_page(page);
423 SetPageUptodate(page);
427 EXPORT_SYMBOL(simple_readpage);
429 int simple_write_begin(struct file *file, struct address_space *mapping,
430 loff_t pos, unsigned len, unsigned flags,
431 struct page **pagep, void **fsdata)
436 index = pos >> PAGE_SHIFT;
438 page = grab_cache_page_write_begin(mapping, index, flags);
444 if (!PageUptodate(page) && (len != PAGE_SIZE)) {
445 unsigned from = pos & (PAGE_SIZE - 1);
447 zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
451 EXPORT_SYMBOL(simple_write_begin);
454 * simple_write_end - .write_end helper for non-block-device FSes
455 * @available: See .write_end of address_space_operations
464 * simple_write_end does the minimum needed for updating a page after writing is
465 * done. It has the same API signature as the .write_end of
466 * address_space_operations vector. So it can just be set onto .write_end for
467 * FSes that don't need any other processing. i_mutex is assumed to be held.
468 * Block based filesystems should use generic_write_end().
469 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
470 * is not called, so a filesystem that actually does store data in .write_inode
471 * should extend on what's done here with a call to mark_inode_dirty() in the
472 * case that i_size has changed.
474 * Use *ONLY* with simple_readpage()
476 int simple_write_end(struct file *file, struct address_space *mapping,
477 loff_t pos, unsigned len, unsigned copied,
478 struct page *page, void *fsdata)
480 struct inode *inode = page->mapping->host;
481 loff_t last_pos = pos + copied;
483 /* zero the stale part of the page if we did a short copy */
484 if (!PageUptodate(page)) {
486 unsigned from = pos & (PAGE_SIZE - 1);
488 zero_user(page, from + copied, len - copied);
490 SetPageUptodate(page);
493 * No need to use i_size_read() here, the i_size
494 * cannot change under us because we hold the i_mutex.
496 if (last_pos > inode->i_size)
497 i_size_write(inode, last_pos);
499 set_page_dirty(page);
505 EXPORT_SYMBOL(simple_write_end);
508 * the inodes created here are not hashed. If you use iunique to generate
509 * unique inode values later for this filesystem, then you must take care
510 * to pass it an appropriate max_reserved value to avoid collisions.
512 int simple_fill_super(struct super_block *s, unsigned long magic,
513 const struct tree_descr *files)
517 struct dentry *dentry;
520 s->s_blocksize = PAGE_SIZE;
521 s->s_blocksize_bits = PAGE_SHIFT;
523 s->s_op = &simple_super_operations;
526 inode = new_inode(s);
530 * because the root inode is 1, the files array must not contain an
534 inode->i_mode = S_IFDIR | 0755;
535 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
536 inode->i_op = &simple_dir_inode_operations;
537 inode->i_fop = &simple_dir_operations;
539 root = d_make_root(inode);
542 for (i = 0; !files->name || files->name[0]; i++, files++) {
546 /* warn if it tries to conflict with the root inode */
547 if (unlikely(i == 1))
548 printk(KERN_WARNING "%s: %s passed in a files array"
549 "with an index of 1!\n", __func__,
552 dentry = d_alloc_name(root, files->name);
555 inode = new_inode(s);
560 inode->i_mode = S_IFREG | files->mode;
561 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
562 inode->i_fop = files->ops;
564 d_add(dentry, inode);
570 shrink_dcache_parent(root);
574 EXPORT_SYMBOL(simple_fill_super);
576 static DEFINE_SPINLOCK(pin_fs_lock);
578 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
580 struct vfsmount *mnt = NULL;
581 spin_lock(&pin_fs_lock);
582 if (unlikely(!*mount)) {
583 spin_unlock(&pin_fs_lock);
584 mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
587 spin_lock(&pin_fs_lock);
593 spin_unlock(&pin_fs_lock);
597 EXPORT_SYMBOL(simple_pin_fs);
599 void simple_release_fs(struct vfsmount **mount, int *count)
601 struct vfsmount *mnt;
602 spin_lock(&pin_fs_lock);
606 spin_unlock(&pin_fs_lock);
609 EXPORT_SYMBOL(simple_release_fs);
612 * simple_read_from_buffer - copy data from the buffer to user space
613 * @to: the user space buffer to read to
614 * @count: the maximum number of bytes to read
615 * @ppos: the current position in the buffer
616 * @from: the buffer to read from
617 * @available: the size of the buffer
619 * The simple_read_from_buffer() function reads up to @count bytes from the
620 * buffer @from at offset @ppos into the user space address starting at @to.
622 * On success, the number of bytes read is returned and the offset @ppos is
623 * advanced by this number, or negative value is returned on error.
625 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
626 const void *from, size_t available)
633 if (pos >= available || !count)
635 if (count > available - pos)
636 count = available - pos;
637 ret = copy_to_user(to, from + pos, count);
644 EXPORT_SYMBOL(simple_read_from_buffer);
647 * simple_write_to_buffer - copy data from user space to the buffer
648 * @to: the buffer to write to
649 * @available: the size of the buffer
650 * @ppos: the current position in the buffer
651 * @from: the user space buffer to read from
652 * @count: the maximum number of bytes to read
654 * The simple_write_to_buffer() function reads up to @count bytes from the user
655 * space address starting at @from into the buffer @to at offset @ppos.
657 * On success, the number of bytes written is returned and the offset @ppos is
658 * advanced by this number, or negative value is returned on error.
660 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
661 const void __user *from, size_t count)
668 if (pos >= available || !count)
670 if (count > available - pos)
671 count = available - pos;
672 res = copy_from_user(to + pos, from, count);
679 EXPORT_SYMBOL(simple_write_to_buffer);
682 * memory_read_from_buffer - copy data from the buffer
683 * @to: the kernel space buffer to read to
684 * @count: the maximum number of bytes to read
685 * @ppos: the current position in the buffer
686 * @from: the buffer to read from
687 * @available: the size of the buffer
689 * The memory_read_from_buffer() function reads up to @count bytes from the
690 * buffer @from at offset @ppos into the kernel space address starting at @to.
692 * On success, the number of bytes read is returned and the offset @ppos is
693 * advanced by this number, or negative value is returned on error.
695 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
696 const void *from, size_t available)
702 if (pos >= available)
704 if (count > available - pos)
705 count = available - pos;
706 memcpy(to, from + pos, count);
711 EXPORT_SYMBOL(memory_read_from_buffer);
714 * Transaction based IO.
715 * The file expects a single write which triggers the transaction, and then
716 * possibly a read which collects the result - which is stored in a
720 void simple_transaction_set(struct file *file, size_t n)
722 struct simple_transaction_argresp *ar = file->private_data;
724 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
727 * The barrier ensures that ar->size will really remain zero until
728 * ar->data is ready for reading.
733 EXPORT_SYMBOL(simple_transaction_set);
735 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
737 struct simple_transaction_argresp *ar;
738 static DEFINE_SPINLOCK(simple_transaction_lock);
740 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
741 return ERR_PTR(-EFBIG);
743 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
745 return ERR_PTR(-ENOMEM);
747 spin_lock(&simple_transaction_lock);
749 /* only one write allowed per open */
750 if (file->private_data) {
751 spin_unlock(&simple_transaction_lock);
752 free_page((unsigned long)ar);
753 return ERR_PTR(-EBUSY);
756 file->private_data = ar;
758 spin_unlock(&simple_transaction_lock);
760 if (copy_from_user(ar->data, buf, size))
761 return ERR_PTR(-EFAULT);
765 EXPORT_SYMBOL(simple_transaction_get);
767 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
769 struct simple_transaction_argresp *ar = file->private_data;
773 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
775 EXPORT_SYMBOL(simple_transaction_read);
777 int simple_transaction_release(struct inode *inode, struct file *file)
779 free_page((unsigned long)file->private_data);
782 EXPORT_SYMBOL(simple_transaction_release);
784 /* Simple attribute files */
787 int (*get)(void *, u64 *);
788 int (*set)(void *, u64);
789 char get_buf[24]; /* enough to store a u64 and "\n\0" */
792 const char *fmt; /* format for read operation */
793 struct mutex mutex; /* protects access to these buffers */
796 /* simple_attr_open is called by an actual attribute open file operation
797 * to set the attribute specific access operations. */
798 int simple_attr_open(struct inode *inode, struct file *file,
799 int (*get)(void *, u64 *), int (*set)(void *, u64),
802 struct simple_attr *attr;
804 attr = kmalloc(sizeof(*attr), GFP_KERNEL);
810 attr->data = inode->i_private;
812 mutex_init(&attr->mutex);
814 file->private_data = attr;
816 return nonseekable_open(inode, file);
818 EXPORT_SYMBOL_GPL(simple_attr_open);
820 int simple_attr_release(struct inode *inode, struct file *file)
822 kfree(file->private_data);
825 EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
827 /* read from the buffer that is filled with the get function */
828 ssize_t simple_attr_read(struct file *file, char __user *buf,
829 size_t len, loff_t *ppos)
831 struct simple_attr *attr;
835 attr = file->private_data;
840 ret = mutex_lock_interruptible(&attr->mutex);
844 if (*ppos) { /* continued read */
845 size = strlen(attr->get_buf);
846 } else { /* first read */
848 ret = attr->get(attr->data, &val);
852 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
853 attr->fmt, (unsigned long long)val);
856 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
858 mutex_unlock(&attr->mutex);
861 EXPORT_SYMBOL_GPL(simple_attr_read);
863 /* interpret the buffer as a number to call the set function with */
864 ssize_t simple_attr_write(struct file *file, const char __user *buf,
865 size_t len, loff_t *ppos)
867 struct simple_attr *attr;
872 attr = file->private_data;
876 ret = mutex_lock_interruptible(&attr->mutex);
881 size = min(sizeof(attr->set_buf) - 1, len);
882 if (copy_from_user(attr->set_buf, buf, size))
885 attr->set_buf[size] = '\0';
886 val = simple_strtoll(attr->set_buf, NULL, 0);
887 ret = attr->set(attr->data, val);
889 ret = len; /* on success, claim we got the whole input */
891 mutex_unlock(&attr->mutex);
894 EXPORT_SYMBOL_GPL(simple_attr_write);
897 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
898 * @sb: filesystem to do the file handle conversion on
899 * @fid: file handle to convert
900 * @fh_len: length of the file handle in bytes
901 * @fh_type: type of file handle
902 * @get_inode: filesystem callback to retrieve inode
904 * This function decodes @fid as long as it has one of the well-known
905 * Linux filehandle types and calls @get_inode on it to retrieve the
906 * inode for the object specified in the file handle.
908 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
909 int fh_len, int fh_type, struct inode *(*get_inode)
910 (struct super_block *sb, u64 ino, u32 gen))
912 struct inode *inode = NULL;
918 case FILEID_INO32_GEN:
919 case FILEID_INO32_GEN_PARENT:
920 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
924 return d_obtain_alias(inode);
926 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
929 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
930 * @sb: filesystem to do the file handle conversion on
931 * @fid: file handle to convert
932 * @fh_len: length of the file handle in bytes
933 * @fh_type: type of file handle
934 * @get_inode: filesystem callback to retrieve inode
936 * This function decodes @fid as long as it has one of the well-known
937 * Linux filehandle types and calls @get_inode on it to retrieve the
938 * inode for the _parent_ object specified in the file handle if it
939 * is specified in the file handle, or NULL otherwise.
941 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
942 int fh_len, int fh_type, struct inode *(*get_inode)
943 (struct super_block *sb, u64 ino, u32 gen))
945 struct inode *inode = NULL;
951 case FILEID_INO32_GEN_PARENT:
952 inode = get_inode(sb, fid->i32.parent_ino,
953 (fh_len > 3 ? fid->i32.parent_gen : 0));
957 return d_obtain_alias(inode);
959 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
962 * __generic_file_fsync - generic fsync implementation for simple filesystems
964 * @file: file to synchronize
965 * @start: start offset in bytes
966 * @end: end offset in bytes (inclusive)
967 * @datasync: only synchronize essential metadata if true
969 * This is a generic implementation of the fsync method for simple
970 * filesystems which track all non-inode metadata in the buffers list
971 * hanging off the address_space structure.
973 int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
976 struct inode *inode = file->f_mapping->host;
980 err = file_write_and_wait_range(file, start, end);
985 ret = sync_mapping_buffers(inode->i_mapping);
986 if (!(inode->i_state & I_DIRTY_ALL))
988 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
991 err = sync_inode_metadata(inode, 1);
997 /* check and advance again to catch errors after syncing out buffers */
998 err = file_check_and_advance_wb_err(file);
1003 EXPORT_SYMBOL(__generic_file_fsync);
1006 * generic_file_fsync - generic fsync implementation for simple filesystems
1008 * @file: file to synchronize
1009 * @start: start offset in bytes
1010 * @end: end offset in bytes (inclusive)
1011 * @datasync: only synchronize essential metadata if true
1015 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1018 struct inode *inode = file->f_mapping->host;
1021 err = __generic_file_fsync(file, start, end, datasync);
1024 return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
1026 EXPORT_SYMBOL(generic_file_fsync);
1029 * generic_check_addressable - Check addressability of file system
1030 * @blocksize_bits: log of file system block size
1031 * @num_blocks: number of blocks in file system
1033 * Determine whether a file system with @num_blocks blocks (and a
1034 * block size of 2**@blocksize_bits) is addressable by the sector_t
1035 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1037 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1039 u64 last_fs_block = num_blocks - 1;
1041 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1043 if (unlikely(num_blocks == 0))
1046 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1049 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1050 (last_fs_page > (pgoff_t)(~0ULL))) {
1055 EXPORT_SYMBOL(generic_check_addressable);
1058 * No-op implementation of ->fsync for in-memory filesystems.
1060 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1064 EXPORT_SYMBOL(noop_fsync);
1066 int noop_set_page_dirty(struct page *page)
1069 * Unlike __set_page_dirty_no_writeback that handles dirty page
1070 * tracking in the page object, dax does all dirty tracking in
1071 * the inode address_space in response to mkwrite faults. In the
1072 * dax case we only need to worry about potentially dirty CPU
1073 * caches, not dirty page cache pages to write back.
1075 * This callback is defined to prevent fallback to
1076 * __set_page_dirty_buffers() in set_page_dirty().
1080 EXPORT_SYMBOL_GPL(noop_set_page_dirty);
1082 void noop_invalidatepage(struct page *page, unsigned int offset,
1083 unsigned int length)
1086 * There is no page cache to invalidate in the dax case, however
1087 * we need this callback defined to prevent falling back to
1088 * block_invalidatepage() in do_invalidatepage().
1091 EXPORT_SYMBOL_GPL(noop_invalidatepage);
1093 ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1096 * iomap based filesystems support direct I/O without need for
1097 * this callback. However, it still needs to be set in
1098 * inode->a_ops so that open/fcntl know that direct I/O is
1099 * generally supported.
1103 EXPORT_SYMBOL_GPL(noop_direct_IO);
1105 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1106 void kfree_link(void *p)
1110 EXPORT_SYMBOL(kfree_link);
1113 * nop .set_page_dirty method so that people can use .page_mkwrite on
1116 static int anon_set_page_dirty(struct page *page)
1122 * A single inode exists for all anon_inode files. Contrary to pipes,
1123 * anon_inode inodes have no associated per-instance data, so we need
1124 * only allocate one of them.
1126 struct inode *alloc_anon_inode(struct super_block *s)
1128 static const struct address_space_operations anon_aops = {
1129 .set_page_dirty = anon_set_page_dirty,
1131 struct inode *inode = new_inode_pseudo(s);
1134 return ERR_PTR(-ENOMEM);
1136 inode->i_ino = get_next_ino();
1137 inode->i_mapping->a_ops = &anon_aops;
1140 * Mark the inode dirty from the very beginning,
1141 * that way it will never be moved to the dirty
1142 * list because mark_inode_dirty() will think
1143 * that it already _is_ on the dirty list.
1145 inode->i_state = I_DIRTY;
1146 inode->i_mode = S_IRUSR | S_IWUSR;
1147 inode->i_uid = current_fsuid();
1148 inode->i_gid = current_fsgid();
1149 inode->i_flags |= S_PRIVATE;
1150 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1153 EXPORT_SYMBOL(alloc_anon_inode);
1156 * simple_nosetlease - generic helper for prohibiting leases
1157 * @filp: file pointer
1158 * @arg: type of lease to obtain
1159 * @flp: new lease supplied for insertion
1160 * @priv: private data for lm_setup operation
1162 * Generic helper for filesystems that do not wish to allow leases to be set.
1163 * All arguments are ignored and it just returns -EINVAL.
1166 simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1171 EXPORT_SYMBOL(simple_nosetlease);
1174 * simple_get_link - generic helper to get the target of "fast" symlinks
1175 * @dentry: not used here
1176 * @inode: the symlink inode
1177 * @done: not used here
1179 * Generic helper for filesystems to use for symlink inodes where a pointer to
1180 * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1181 * since as an optimization the path lookup code uses any non-NULL ->i_link
1182 * directly, without calling ->get_link(). But ->get_link() still must be set,
1183 * to mark the inode_operations as being for a symlink.
1185 * Return: the symlink target
1187 const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1188 struct delayed_call *done)
1190 return inode->i_link;
1192 EXPORT_SYMBOL(simple_get_link);
1194 const struct inode_operations simple_symlink_inode_operations = {
1195 .get_link = simple_get_link,
1197 EXPORT_SYMBOL(simple_symlink_inode_operations);
1200 * Operations for a permanently empty directory.
1202 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1204 return ERR_PTR(-ENOENT);
1207 static int empty_dir_getattr(const struct path *path, struct kstat *stat,
1208 u32 request_mask, unsigned int query_flags)
1210 struct inode *inode = d_inode(path->dentry);
1211 generic_fillattr(inode, stat);
1215 static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1220 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1225 static const struct inode_operations empty_dir_inode_operations = {
1226 .lookup = empty_dir_lookup,
1227 .permission = generic_permission,
1228 .setattr = empty_dir_setattr,
1229 .getattr = empty_dir_getattr,
1230 .listxattr = empty_dir_listxattr,
1233 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1235 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1236 return generic_file_llseek_size(file, offset, whence, 2, 2);
1239 static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1241 dir_emit_dots(file, ctx);
1245 static const struct file_operations empty_dir_operations = {
1246 .llseek = empty_dir_llseek,
1247 .read = generic_read_dir,
1248 .iterate_shared = empty_dir_readdir,
1249 .fsync = noop_fsync,
1253 void make_empty_dir_inode(struct inode *inode)
1255 set_nlink(inode, 2);
1256 inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1257 inode->i_uid = GLOBAL_ROOT_UID;
1258 inode->i_gid = GLOBAL_ROOT_GID;
1261 inode->i_blkbits = PAGE_SHIFT;
1262 inode->i_blocks = 0;
1264 inode->i_op = &empty_dir_inode_operations;
1265 inode->i_opflags &= ~IOP_XATTR;
1266 inode->i_fop = &empty_dir_operations;
1269 bool is_empty_dir_inode(struct inode *inode)
1271 return (inode->i_fop == &empty_dir_operations) &&
1272 (inode->i_op == &empty_dir_inode_operations);