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 */
20 #include <linux/fs_context.h>
21 #include <linux/pseudo_fs.h>
22 #include <linux/fsnotify.h>
23 #include <linux/unicode.h>
24 #include <linux/fscrypt.h>
26 #include <linux/uaccess.h>
30 int simple_getattr(const struct path *path, struct kstat *stat,
31 u32 request_mask, unsigned int query_flags)
33 struct inode *inode = d_inode(path->dentry);
34 generic_fillattr(inode, stat);
35 stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
38 EXPORT_SYMBOL(simple_getattr);
40 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
42 buf->f_type = dentry->d_sb->s_magic;
43 buf->f_bsize = PAGE_SIZE;
44 buf->f_namelen = NAME_MAX;
47 EXPORT_SYMBOL(simple_statfs);
50 * Retaining negative dentries for an in-memory filesystem just wastes
51 * memory and lookup time: arrange for them to be deleted immediately.
53 int always_delete_dentry(const struct dentry *dentry)
57 EXPORT_SYMBOL(always_delete_dentry);
59 const struct dentry_operations simple_dentry_operations = {
60 .d_delete = always_delete_dentry,
62 EXPORT_SYMBOL(simple_dentry_operations);
65 * Lookup the data. This is trivial - if the dentry didn't already
66 * exist, we know it is negative. Set d_op to delete negative dentries.
68 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
70 if (dentry->d_name.len > NAME_MAX)
71 return ERR_PTR(-ENAMETOOLONG);
72 if (!dentry->d_sb->s_d_op)
73 d_set_d_op(dentry, &simple_dentry_operations);
77 EXPORT_SYMBOL(simple_lookup);
79 int dcache_dir_open(struct inode *inode, struct file *file)
81 file->private_data = d_alloc_cursor(file->f_path.dentry);
83 return file->private_data ? 0 : -ENOMEM;
85 EXPORT_SYMBOL(dcache_dir_open);
87 int dcache_dir_close(struct inode *inode, struct file *file)
89 dput(file->private_data);
92 EXPORT_SYMBOL(dcache_dir_close);
94 /* parent is locked at least shared */
96 * Returns an element of siblings' list.
97 * We are looking for <count>th positive after <p>; if
98 * found, dentry is grabbed and returned to caller.
99 * If no such element exists, NULL is returned.
101 static struct dentry *scan_positives(struct dentry *cursor,
106 struct dentry *dentry = cursor->d_parent, *found = NULL;
108 spin_lock(&dentry->d_lock);
109 while ((p = p->next) != &dentry->d_subdirs) {
110 struct dentry *d = list_entry(p, struct dentry, d_child);
111 // we must at least skip cursors, to avoid livelocks
112 if (d->d_flags & DCACHE_DENTRY_CURSOR)
114 if (simple_positive(d) && !--count) {
115 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
116 if (simple_positive(d))
117 found = dget_dlock(d);
118 spin_unlock(&d->d_lock);
123 if (need_resched()) {
124 list_move(&cursor->d_child, p);
125 p = &cursor->d_child;
126 spin_unlock(&dentry->d_lock);
128 spin_lock(&dentry->d_lock);
131 spin_unlock(&dentry->d_lock);
136 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
138 struct dentry *dentry = file->f_path.dentry;
141 offset += file->f_pos;
150 if (offset != file->f_pos) {
151 struct dentry *cursor = file->private_data;
152 struct dentry *to = NULL;
154 inode_lock_shared(dentry->d_inode);
157 to = scan_positives(cursor, &dentry->d_subdirs,
159 spin_lock(&dentry->d_lock);
161 list_move(&cursor->d_child, &to->d_child);
163 list_del_init(&cursor->d_child);
164 spin_unlock(&dentry->d_lock);
167 file->f_pos = offset;
169 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 *anchor = &dentry->d_subdirs;
192 struct dentry *next = NULL;
195 if (!dir_emit_dots(file, ctx))
200 else if (!list_empty(&cursor->d_child))
201 p = &cursor->d_child;
205 while ((next = scan_positives(cursor, p, 1, next)) != NULL) {
206 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
207 d_inode(next)->i_ino, dt_type(d_inode(next))))
212 spin_lock(&dentry->d_lock);
214 list_move_tail(&cursor->d_child, &next->d_child);
216 list_del_init(&cursor->d_child);
217 spin_unlock(&dentry->d_lock);
222 EXPORT_SYMBOL(dcache_readdir);
224 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
228 EXPORT_SYMBOL(generic_read_dir);
230 const struct file_operations simple_dir_operations = {
231 .open = dcache_dir_open,
232 .release = dcache_dir_close,
233 .llseek = dcache_dir_lseek,
234 .read = generic_read_dir,
235 .iterate_shared = dcache_readdir,
238 EXPORT_SYMBOL(simple_dir_operations);
240 const struct inode_operations simple_dir_inode_operations = {
241 .lookup = simple_lookup,
243 EXPORT_SYMBOL(simple_dir_inode_operations);
245 static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev)
247 struct dentry *child = NULL;
248 struct list_head *p = prev ? &prev->d_child : &parent->d_subdirs;
250 spin_lock(&parent->d_lock);
251 while ((p = p->next) != &parent->d_subdirs) {
252 struct dentry *d = container_of(p, struct dentry, d_child);
253 if (simple_positive(d)) {
254 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
255 if (simple_positive(d))
256 child = dget_dlock(d);
257 spin_unlock(&d->d_lock);
262 spin_unlock(&parent->d_lock);
267 void simple_recursive_removal(struct dentry *dentry,
268 void (*callback)(struct dentry *))
270 struct dentry *this = dget(dentry);
272 struct dentry *victim = NULL, *child;
273 struct inode *inode = this->d_inode;
277 inode->i_flags |= S_DEAD;
278 while ((child = find_next_child(this, victim)) == NULL) {
280 // update metadata while it's still locked
281 inode->i_ctime = current_time(inode);
285 this = this->d_parent;
286 inode = this->d_inode;
288 if (simple_positive(victim)) {
289 d_invalidate(victim); // avoid lost mounts
290 if (d_is_dir(victim))
291 fsnotify_rmdir(inode, victim);
293 fsnotify_unlink(inode, victim);
296 dput(victim); // unpin it
298 if (victim == dentry) {
299 inode->i_ctime = inode->i_mtime =
301 if (d_is_dir(dentry))
312 EXPORT_SYMBOL(simple_recursive_removal);
314 static const struct super_operations simple_super_operations = {
315 .statfs = simple_statfs,
318 static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc)
320 struct pseudo_fs_context *ctx = fc->fs_private;
323 s->s_maxbytes = MAX_LFS_FILESIZE;
324 s->s_blocksize = PAGE_SIZE;
325 s->s_blocksize_bits = PAGE_SHIFT;
326 s->s_magic = ctx->magic;
327 s->s_op = ctx->ops ?: &simple_super_operations;
328 s->s_xattr = ctx->xattr;
335 * since this is the first inode, make it number 1. New inodes created
336 * after this must take care not to collide with it (by passing
337 * max_reserved of 1 to iunique).
340 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
341 root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
342 s->s_root = d_make_root(root);
345 s->s_d_op = ctx->dops;
349 static int pseudo_fs_get_tree(struct fs_context *fc)
351 return get_tree_nodev(fc, pseudo_fs_fill_super);
354 static void pseudo_fs_free(struct fs_context *fc)
356 kfree(fc->fs_private);
359 static const struct fs_context_operations pseudo_fs_context_ops = {
360 .free = pseudo_fs_free,
361 .get_tree = pseudo_fs_get_tree,
365 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
366 * will never be mountable)
368 struct pseudo_fs_context *init_pseudo(struct fs_context *fc,
371 struct pseudo_fs_context *ctx;
373 ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL);
376 fc->fs_private = ctx;
377 fc->ops = &pseudo_fs_context_ops;
378 fc->sb_flags |= SB_NOUSER;
383 EXPORT_SYMBOL(init_pseudo);
385 int simple_open(struct inode *inode, struct file *file)
387 if (inode->i_private)
388 file->private_data = inode->i_private;
391 EXPORT_SYMBOL(simple_open);
393 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
395 struct inode *inode = d_inode(old_dentry);
397 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
401 d_instantiate(dentry, inode);
404 EXPORT_SYMBOL(simple_link);
406 int simple_empty(struct dentry *dentry)
408 struct dentry *child;
411 spin_lock(&dentry->d_lock);
412 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
413 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
414 if (simple_positive(child)) {
415 spin_unlock(&child->d_lock);
418 spin_unlock(&child->d_lock);
422 spin_unlock(&dentry->d_lock);
425 EXPORT_SYMBOL(simple_empty);
427 int simple_unlink(struct inode *dir, struct dentry *dentry)
429 struct inode *inode = d_inode(dentry);
431 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
436 EXPORT_SYMBOL(simple_unlink);
438 int simple_rmdir(struct inode *dir, struct dentry *dentry)
440 if (!simple_empty(dentry))
443 drop_nlink(d_inode(dentry));
444 simple_unlink(dir, dentry);
448 EXPORT_SYMBOL(simple_rmdir);
450 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
451 struct inode *new_dir, struct dentry *new_dentry,
454 struct inode *inode = d_inode(old_dentry);
455 int they_are_dirs = d_is_dir(old_dentry);
457 if (flags & ~RENAME_NOREPLACE)
460 if (!simple_empty(new_dentry))
463 if (d_really_is_positive(new_dentry)) {
464 simple_unlink(new_dir, new_dentry);
466 drop_nlink(d_inode(new_dentry));
469 } else if (they_are_dirs) {
474 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
475 new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
479 EXPORT_SYMBOL(simple_rename);
482 * simple_setattr - setattr for simple filesystem
484 * @iattr: iattr structure
486 * Returns 0 on success, -error on failure.
488 * simple_setattr is a simple ->setattr implementation without a proper
489 * implementation of size changes.
491 * It can either be used for in-memory filesystems or special files
492 * on simple regular filesystems. Anything that needs to change on-disk
493 * or wire state on size changes needs its own setattr method.
495 int simple_setattr(struct dentry *dentry, struct iattr *iattr)
497 struct inode *inode = d_inode(dentry);
500 error = setattr_prepare(dentry, iattr);
504 if (iattr->ia_valid & ATTR_SIZE)
505 truncate_setsize(inode, iattr->ia_size);
506 setattr_copy(inode, iattr);
507 mark_inode_dirty(inode);
510 EXPORT_SYMBOL(simple_setattr);
512 int simple_readpage(struct file *file, struct page *page)
514 clear_highpage(page);
515 flush_dcache_page(page);
516 SetPageUptodate(page);
520 EXPORT_SYMBOL(simple_readpage);
522 int simple_write_begin(struct file *file, struct address_space *mapping,
523 loff_t pos, unsigned len, unsigned flags,
524 struct page **pagep, void **fsdata)
529 index = pos >> PAGE_SHIFT;
531 page = grab_cache_page_write_begin(mapping, index, flags);
537 if (!PageUptodate(page) && (len != PAGE_SIZE)) {
538 unsigned from = pos & (PAGE_SIZE - 1);
540 zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
544 EXPORT_SYMBOL(simple_write_begin);
547 * simple_write_end - .write_end helper for non-block-device FSes
548 * @file: See .write_end of address_space_operations
556 * simple_write_end does the minimum needed for updating a page after writing is
557 * done. It has the same API signature as the .write_end of
558 * address_space_operations vector. So it can just be set onto .write_end for
559 * FSes that don't need any other processing. i_mutex is assumed to be held.
560 * Block based filesystems should use generic_write_end().
561 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
562 * is not called, so a filesystem that actually does store data in .write_inode
563 * should extend on what's done here with a call to mark_inode_dirty() in the
564 * case that i_size has changed.
566 * Use *ONLY* with simple_readpage()
568 int simple_write_end(struct file *file, struct address_space *mapping,
569 loff_t pos, unsigned len, unsigned copied,
570 struct page *page, void *fsdata)
572 struct inode *inode = page->mapping->host;
573 loff_t last_pos = pos + copied;
575 /* zero the stale part of the page if we did a short copy */
576 if (!PageUptodate(page)) {
578 unsigned from = pos & (PAGE_SIZE - 1);
580 zero_user(page, from + copied, len - copied);
582 SetPageUptodate(page);
585 * No need to use i_size_read() here, the i_size
586 * cannot change under us because we hold the i_mutex.
588 if (last_pos > inode->i_size)
589 i_size_write(inode, last_pos);
591 set_page_dirty(page);
597 EXPORT_SYMBOL(simple_write_end);
600 * the inodes created here are not hashed. If you use iunique to generate
601 * unique inode values later for this filesystem, then you must take care
602 * to pass it an appropriate max_reserved value to avoid collisions.
604 int simple_fill_super(struct super_block *s, unsigned long magic,
605 const struct tree_descr *files)
609 struct dentry *dentry;
612 s->s_blocksize = PAGE_SIZE;
613 s->s_blocksize_bits = PAGE_SHIFT;
615 s->s_op = &simple_super_operations;
618 inode = new_inode(s);
622 * because the root inode is 1, the files array must not contain an
626 inode->i_mode = S_IFDIR | 0755;
627 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
628 inode->i_op = &simple_dir_inode_operations;
629 inode->i_fop = &simple_dir_operations;
631 root = d_make_root(inode);
634 for (i = 0; !files->name || files->name[0]; i++, files++) {
638 /* warn if it tries to conflict with the root inode */
639 if (unlikely(i == 1))
640 printk(KERN_WARNING "%s: %s passed in a files array"
641 "with an index of 1!\n", __func__,
644 dentry = d_alloc_name(root, files->name);
647 inode = new_inode(s);
652 inode->i_mode = S_IFREG | files->mode;
653 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
654 inode->i_fop = files->ops;
656 d_add(dentry, inode);
662 shrink_dcache_parent(root);
666 EXPORT_SYMBOL(simple_fill_super);
668 static DEFINE_SPINLOCK(pin_fs_lock);
670 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
672 struct vfsmount *mnt = NULL;
673 spin_lock(&pin_fs_lock);
674 if (unlikely(!*mount)) {
675 spin_unlock(&pin_fs_lock);
676 mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
679 spin_lock(&pin_fs_lock);
685 spin_unlock(&pin_fs_lock);
689 EXPORT_SYMBOL(simple_pin_fs);
691 void simple_release_fs(struct vfsmount **mount, int *count)
693 struct vfsmount *mnt;
694 spin_lock(&pin_fs_lock);
698 spin_unlock(&pin_fs_lock);
701 EXPORT_SYMBOL(simple_release_fs);
704 * simple_read_from_buffer - copy data from the buffer to user space
705 * @to: the user space buffer to read to
706 * @count: the maximum number of bytes to read
707 * @ppos: the current position in the buffer
708 * @from: the buffer to read from
709 * @available: the size of the buffer
711 * The simple_read_from_buffer() function reads up to @count bytes from the
712 * buffer @from at offset @ppos into the user space address starting at @to.
714 * On success, the number of bytes read is returned and the offset @ppos is
715 * advanced by this number, or negative value is returned on error.
717 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
718 const void *from, size_t available)
725 if (pos >= available || !count)
727 if (count > available - pos)
728 count = available - pos;
729 ret = copy_to_user(to, from + pos, count);
736 EXPORT_SYMBOL(simple_read_from_buffer);
739 * simple_write_to_buffer - copy data from user space to the buffer
740 * @to: the buffer to write to
741 * @available: the size of the buffer
742 * @ppos: the current position in the buffer
743 * @from: the user space buffer to read from
744 * @count: the maximum number of bytes to read
746 * The simple_write_to_buffer() function reads up to @count bytes from the user
747 * space address starting at @from into the buffer @to at offset @ppos.
749 * On success, the number of bytes written is returned and the offset @ppos is
750 * advanced by this number, or negative value is returned on error.
752 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
753 const void __user *from, size_t count)
760 if (pos >= available || !count)
762 if (count > available - pos)
763 count = available - pos;
764 res = copy_from_user(to + pos, from, count);
771 EXPORT_SYMBOL(simple_write_to_buffer);
774 * memory_read_from_buffer - copy data from the buffer
775 * @to: the kernel space buffer to read to
776 * @count: the maximum number of bytes to read
777 * @ppos: the current position in the buffer
778 * @from: the buffer to read from
779 * @available: the size of the buffer
781 * The memory_read_from_buffer() function reads up to @count bytes from the
782 * buffer @from at offset @ppos into the kernel space address starting at @to.
784 * On success, the number of bytes read is returned and the offset @ppos is
785 * advanced by this number, or negative value is returned on error.
787 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
788 const void *from, size_t available)
794 if (pos >= available)
796 if (count > available - pos)
797 count = available - pos;
798 memcpy(to, from + pos, count);
803 EXPORT_SYMBOL(memory_read_from_buffer);
806 * Transaction based IO.
807 * The file expects a single write which triggers the transaction, and then
808 * possibly a read which collects the result - which is stored in a
812 void simple_transaction_set(struct file *file, size_t n)
814 struct simple_transaction_argresp *ar = file->private_data;
816 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
819 * The barrier ensures that ar->size will really remain zero until
820 * ar->data is ready for reading.
825 EXPORT_SYMBOL(simple_transaction_set);
827 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
829 struct simple_transaction_argresp *ar;
830 static DEFINE_SPINLOCK(simple_transaction_lock);
832 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
833 return ERR_PTR(-EFBIG);
835 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
837 return ERR_PTR(-ENOMEM);
839 spin_lock(&simple_transaction_lock);
841 /* only one write allowed per open */
842 if (file->private_data) {
843 spin_unlock(&simple_transaction_lock);
844 free_page((unsigned long)ar);
845 return ERR_PTR(-EBUSY);
848 file->private_data = ar;
850 spin_unlock(&simple_transaction_lock);
852 if (copy_from_user(ar->data, buf, size))
853 return ERR_PTR(-EFAULT);
857 EXPORT_SYMBOL(simple_transaction_get);
859 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
861 struct simple_transaction_argresp *ar = file->private_data;
865 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
867 EXPORT_SYMBOL(simple_transaction_read);
869 int simple_transaction_release(struct inode *inode, struct file *file)
871 free_page((unsigned long)file->private_data);
874 EXPORT_SYMBOL(simple_transaction_release);
876 /* Simple attribute files */
879 int (*get)(void *, u64 *);
880 int (*set)(void *, u64);
881 char get_buf[24]; /* enough to store a u64 and "\n\0" */
884 const char *fmt; /* format for read operation */
885 struct mutex mutex; /* protects access to these buffers */
888 /* simple_attr_open is called by an actual attribute open file operation
889 * to set the attribute specific access operations. */
890 int simple_attr_open(struct inode *inode, struct file *file,
891 int (*get)(void *, u64 *), int (*set)(void *, u64),
894 struct simple_attr *attr;
896 attr = kzalloc(sizeof(*attr), GFP_KERNEL);
902 attr->data = inode->i_private;
904 mutex_init(&attr->mutex);
906 file->private_data = attr;
908 return nonseekable_open(inode, file);
910 EXPORT_SYMBOL_GPL(simple_attr_open);
912 int simple_attr_release(struct inode *inode, struct file *file)
914 kfree(file->private_data);
917 EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
919 /* read from the buffer that is filled with the get function */
920 ssize_t simple_attr_read(struct file *file, char __user *buf,
921 size_t len, loff_t *ppos)
923 struct simple_attr *attr;
927 attr = file->private_data;
932 ret = mutex_lock_interruptible(&attr->mutex);
936 if (*ppos && attr->get_buf[0]) {
938 size = strlen(attr->get_buf);
942 ret = attr->get(attr->data, &val);
946 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
947 attr->fmt, (unsigned long long)val);
950 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
952 mutex_unlock(&attr->mutex);
955 EXPORT_SYMBOL_GPL(simple_attr_read);
957 /* interpret the buffer as a number to call the set function with */
958 ssize_t simple_attr_write(struct file *file, const char __user *buf,
959 size_t len, loff_t *ppos)
961 struct simple_attr *attr;
962 unsigned long long val;
966 attr = file->private_data;
970 ret = mutex_lock_interruptible(&attr->mutex);
975 size = min(sizeof(attr->set_buf) - 1, len);
976 if (copy_from_user(attr->set_buf, buf, size))
979 attr->set_buf[size] = '\0';
980 ret = kstrtoull(attr->set_buf, 0, &val);
983 ret = attr->set(attr->data, val);
985 ret = len; /* on success, claim we got the whole input */
987 mutex_unlock(&attr->mutex);
990 EXPORT_SYMBOL_GPL(simple_attr_write);
993 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
994 * @sb: filesystem to do the file handle conversion on
995 * @fid: file handle to convert
996 * @fh_len: length of the file handle in bytes
997 * @fh_type: type of file handle
998 * @get_inode: filesystem callback to retrieve inode
1000 * This function decodes @fid as long as it has one of the well-known
1001 * Linux filehandle types and calls @get_inode on it to retrieve the
1002 * inode for the object specified in the file handle.
1004 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
1005 int fh_len, int fh_type, struct inode *(*get_inode)
1006 (struct super_block *sb, u64 ino, u32 gen))
1008 struct inode *inode = NULL;
1014 case FILEID_INO32_GEN:
1015 case FILEID_INO32_GEN_PARENT:
1016 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
1020 return d_obtain_alias(inode);
1022 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
1025 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
1026 * @sb: filesystem to do the file handle conversion on
1027 * @fid: file handle to convert
1028 * @fh_len: length of the file handle in bytes
1029 * @fh_type: type of file handle
1030 * @get_inode: filesystem callback to retrieve inode
1032 * This function decodes @fid as long as it has one of the well-known
1033 * Linux filehandle types and calls @get_inode on it to retrieve the
1034 * inode for the _parent_ object specified in the file handle if it
1035 * is specified in the file handle, or NULL otherwise.
1037 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
1038 int fh_len, int fh_type, struct inode *(*get_inode)
1039 (struct super_block *sb, u64 ino, u32 gen))
1041 struct inode *inode = NULL;
1047 case FILEID_INO32_GEN_PARENT:
1048 inode = get_inode(sb, fid->i32.parent_ino,
1049 (fh_len > 3 ? fid->i32.parent_gen : 0));
1053 return d_obtain_alias(inode);
1055 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
1058 * __generic_file_fsync - generic fsync implementation for simple filesystems
1060 * @file: file to synchronize
1061 * @start: start offset in bytes
1062 * @end: end offset in bytes (inclusive)
1063 * @datasync: only synchronize essential metadata if true
1065 * This is a generic implementation of the fsync method for simple
1066 * filesystems which track all non-inode metadata in the buffers list
1067 * hanging off the address_space structure.
1069 int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
1072 struct inode *inode = file->f_mapping->host;
1076 err = file_write_and_wait_range(file, start, end);
1081 ret = sync_mapping_buffers(inode->i_mapping);
1082 if (!(inode->i_state & I_DIRTY_ALL))
1084 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
1087 err = sync_inode_metadata(inode, 1);
1092 inode_unlock(inode);
1093 /* check and advance again to catch errors after syncing out buffers */
1094 err = file_check_and_advance_wb_err(file);
1099 EXPORT_SYMBOL(__generic_file_fsync);
1102 * generic_file_fsync - generic fsync implementation for simple filesystems
1104 * @file: file to synchronize
1105 * @start: start offset in bytes
1106 * @end: end offset in bytes (inclusive)
1107 * @datasync: only synchronize essential metadata if true
1111 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1114 struct inode *inode = file->f_mapping->host;
1117 err = __generic_file_fsync(file, start, end, datasync);
1120 return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL);
1122 EXPORT_SYMBOL(generic_file_fsync);
1125 * generic_check_addressable - Check addressability of file system
1126 * @blocksize_bits: log of file system block size
1127 * @num_blocks: number of blocks in file system
1129 * Determine whether a file system with @num_blocks blocks (and a
1130 * block size of 2**@blocksize_bits) is addressable by the sector_t
1131 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1133 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1135 u64 last_fs_block = num_blocks - 1;
1137 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1139 if (unlikely(num_blocks == 0))
1142 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1145 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1146 (last_fs_page > (pgoff_t)(~0ULL))) {
1151 EXPORT_SYMBOL(generic_check_addressable);
1154 * No-op implementation of ->fsync for in-memory filesystems.
1156 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1160 EXPORT_SYMBOL(noop_fsync);
1162 int noop_set_page_dirty(struct page *page)
1165 * Unlike __set_page_dirty_no_writeback that handles dirty page
1166 * tracking in the page object, dax does all dirty tracking in
1167 * the inode address_space in response to mkwrite faults. In the
1168 * dax case we only need to worry about potentially dirty CPU
1169 * caches, not dirty page cache pages to write back.
1171 * This callback is defined to prevent fallback to
1172 * __set_page_dirty_buffers() in set_page_dirty().
1176 EXPORT_SYMBOL_GPL(noop_set_page_dirty);
1178 void noop_invalidatepage(struct page *page, unsigned int offset,
1179 unsigned int length)
1182 * There is no page cache to invalidate in the dax case, however
1183 * we need this callback defined to prevent falling back to
1184 * block_invalidatepage() in do_invalidatepage().
1187 EXPORT_SYMBOL_GPL(noop_invalidatepage);
1189 ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1192 * iomap based filesystems support direct I/O without need for
1193 * this callback. However, it still needs to be set in
1194 * inode->a_ops so that open/fcntl know that direct I/O is
1195 * generally supported.
1199 EXPORT_SYMBOL_GPL(noop_direct_IO);
1201 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1202 void kfree_link(void *p)
1206 EXPORT_SYMBOL(kfree_link);
1209 * nop .set_page_dirty method so that people can use .page_mkwrite on
1212 static int anon_set_page_dirty(struct page *page)
1218 * A single inode exists for all anon_inode files. Contrary to pipes,
1219 * anon_inode inodes have no associated per-instance data, so we need
1220 * only allocate one of them.
1222 struct inode *alloc_anon_inode(struct super_block *s)
1224 static const struct address_space_operations anon_aops = {
1225 .set_page_dirty = anon_set_page_dirty,
1227 struct inode *inode = new_inode_pseudo(s);
1230 return ERR_PTR(-ENOMEM);
1232 inode->i_ino = get_next_ino();
1233 inode->i_mapping->a_ops = &anon_aops;
1236 * Mark the inode dirty from the very beginning,
1237 * that way it will never be moved to the dirty
1238 * list because mark_inode_dirty() will think
1239 * that it already _is_ on the dirty list.
1241 inode->i_state = I_DIRTY;
1242 inode->i_mode = S_IRUSR | S_IWUSR;
1243 inode->i_uid = current_fsuid();
1244 inode->i_gid = current_fsgid();
1245 inode->i_flags |= S_PRIVATE;
1246 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1249 EXPORT_SYMBOL(alloc_anon_inode);
1252 * simple_nosetlease - generic helper for prohibiting leases
1253 * @filp: file pointer
1254 * @arg: type of lease to obtain
1255 * @flp: new lease supplied for insertion
1256 * @priv: private data for lm_setup operation
1258 * Generic helper for filesystems that do not wish to allow leases to be set.
1259 * All arguments are ignored and it just returns -EINVAL.
1262 simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1267 EXPORT_SYMBOL(simple_nosetlease);
1270 * simple_get_link - generic helper to get the target of "fast" symlinks
1271 * @dentry: not used here
1272 * @inode: the symlink inode
1273 * @done: not used here
1275 * Generic helper for filesystems to use for symlink inodes where a pointer to
1276 * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1277 * since as an optimization the path lookup code uses any non-NULL ->i_link
1278 * directly, without calling ->get_link(). But ->get_link() still must be set,
1279 * to mark the inode_operations as being for a symlink.
1281 * Return: the symlink target
1283 const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1284 struct delayed_call *done)
1286 return inode->i_link;
1288 EXPORT_SYMBOL(simple_get_link);
1290 const struct inode_operations simple_symlink_inode_operations = {
1291 .get_link = simple_get_link,
1293 EXPORT_SYMBOL(simple_symlink_inode_operations);
1296 * Operations for a permanently empty directory.
1298 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1300 return ERR_PTR(-ENOENT);
1303 static int empty_dir_getattr(const struct path *path, struct kstat *stat,
1304 u32 request_mask, unsigned int query_flags)
1306 struct inode *inode = d_inode(path->dentry);
1307 generic_fillattr(inode, stat);
1311 static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1316 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1321 static const struct inode_operations empty_dir_inode_operations = {
1322 .lookup = empty_dir_lookup,
1323 .permission = generic_permission,
1324 .setattr = empty_dir_setattr,
1325 .getattr = empty_dir_getattr,
1326 .listxattr = empty_dir_listxattr,
1329 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1331 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1332 return generic_file_llseek_size(file, offset, whence, 2, 2);
1335 static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1337 dir_emit_dots(file, ctx);
1341 static const struct file_operations empty_dir_operations = {
1342 .llseek = empty_dir_llseek,
1343 .read = generic_read_dir,
1344 .iterate_shared = empty_dir_readdir,
1345 .fsync = noop_fsync,
1349 void make_empty_dir_inode(struct inode *inode)
1351 set_nlink(inode, 2);
1352 inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1353 inode->i_uid = GLOBAL_ROOT_UID;
1354 inode->i_gid = GLOBAL_ROOT_GID;
1357 inode->i_blkbits = PAGE_SHIFT;
1358 inode->i_blocks = 0;
1360 inode->i_op = &empty_dir_inode_operations;
1361 inode->i_opflags &= ~IOP_XATTR;
1362 inode->i_fop = &empty_dir_operations;
1365 bool is_empty_dir_inode(struct inode *inode)
1367 return (inode->i_fop == &empty_dir_operations) &&
1368 (inode->i_op == &empty_dir_inode_operations);
1371 #ifdef CONFIG_UNICODE
1373 * Determine if the name of a dentry should be casefolded.
1375 * Return: if names will need casefolding
1377 static bool needs_casefold(const struct inode *dir)
1379 return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding;
1383 * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
1384 * @dentry: dentry whose name we are checking against
1385 * @len: len of name of dentry
1386 * @str: str pointer to name of dentry
1387 * @name: Name to compare against
1389 * Return: 0 if names match, 1 if mismatch, or -ERRNO
1391 int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
1392 const char *str, const struct qstr *name)
1394 const struct dentry *parent = READ_ONCE(dentry->d_parent);
1395 const struct inode *dir = READ_ONCE(parent->d_inode);
1396 const struct super_block *sb = dentry->d_sb;
1397 const struct unicode_map *um = sb->s_encoding;
1398 struct qstr qstr = QSTR_INIT(str, len);
1399 char strbuf[DNAME_INLINE_LEN];
1402 if (!dir || !needs_casefold(dir))
1405 * If the dentry name is stored in-line, then it may be concurrently
1406 * modified by a rename. If this happens, the VFS will eventually retry
1407 * the lookup, so it doesn't matter what ->d_compare() returns.
1408 * However, it's unsafe to call utf8_strncasecmp() with an unstable
1409 * string. Therefore, we have to copy the name into a temporary buffer.
1411 if (len <= DNAME_INLINE_LEN - 1) {
1412 memcpy(strbuf, str, len);
1415 /* prevent compiler from optimizing out the temporary buffer */
1418 ret = utf8_strncasecmp(um, name, &qstr);
1422 if (sb_has_strict_encoding(sb))
1425 if (len != name->len)
1427 return !!memcmp(str, name->name, len);
1429 EXPORT_SYMBOL(generic_ci_d_compare);
1432 * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
1433 * @dentry: dentry of the parent directory
1434 * @str: qstr of name whose hash we should fill in
1436 * Return: 0 if hash was successful or unchanged, and -EINVAL on error
1438 int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str)
1440 const struct inode *dir = READ_ONCE(dentry->d_inode);
1441 struct super_block *sb = dentry->d_sb;
1442 const struct unicode_map *um = sb->s_encoding;
1445 if (!dir || !needs_casefold(dir))
1448 ret = utf8_casefold_hash(um, dentry, str);
1449 if (ret < 0 && sb_has_strict_encoding(sb))
1453 EXPORT_SYMBOL(generic_ci_d_hash);
1455 static const struct dentry_operations generic_ci_dentry_ops = {
1456 .d_hash = generic_ci_d_hash,
1457 .d_compare = generic_ci_d_compare,
1461 #ifdef CONFIG_FS_ENCRYPTION
1462 static const struct dentry_operations generic_encrypted_dentry_ops = {
1463 .d_revalidate = fscrypt_d_revalidate,
1467 #if defined(CONFIG_FS_ENCRYPTION) && defined(CONFIG_UNICODE)
1468 static const struct dentry_operations generic_encrypted_ci_dentry_ops = {
1469 .d_hash = generic_ci_d_hash,
1470 .d_compare = generic_ci_d_compare,
1471 .d_revalidate = fscrypt_d_revalidate,
1476 * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry
1477 * @dentry: dentry to set ops on
1479 * Casefolded directories need d_hash and d_compare set, so that the dentries
1480 * contained in them are handled case-insensitively. Note that these operations
1481 * are needed on the parent directory rather than on the dentries in it, and
1482 * while the casefolding flag can be toggled on and off on an empty directory,
1483 * dentry_operations can't be changed later. As a result, if the filesystem has
1484 * casefolding support enabled at all, we have to give all dentries the
1485 * casefolding operations even if their inode doesn't have the casefolding flag
1486 * currently (and thus the casefolding ops would be no-ops for now).
1488 * Encryption works differently in that the only dentry operation it needs is
1489 * d_revalidate, which it only needs on dentries that have the no-key name flag.
1490 * The no-key flag can't be set "later", so we don't have to worry about that.
1492 * Finally, to maximize compatibility with overlayfs (which isn't compatible
1493 * with certain dentry operations) and to avoid taking an unnecessary
1494 * performance hit, we use custom dentry_operations for each possible
1495 * combination rather than always installing all operations.
1497 void generic_set_encrypted_ci_d_ops(struct dentry *dentry)
1499 #ifdef CONFIG_FS_ENCRYPTION
1500 bool needs_encrypt_ops = dentry->d_flags & DCACHE_NOKEY_NAME;
1502 #ifdef CONFIG_UNICODE
1503 bool needs_ci_ops = dentry->d_sb->s_encoding;
1505 #if defined(CONFIG_FS_ENCRYPTION) && defined(CONFIG_UNICODE)
1506 if (needs_encrypt_ops && needs_ci_ops) {
1507 d_set_d_op(dentry, &generic_encrypted_ci_dentry_ops);
1511 #ifdef CONFIG_FS_ENCRYPTION
1512 if (needs_encrypt_ops) {
1513 d_set_d_op(dentry, &generic_encrypted_dentry_ops);
1517 #ifdef CONFIG_UNICODE
1519 d_set_d_op(dentry, &generic_ci_dentry_ops);
1524 EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops);