blob: d1c3bade9f30dd028cfc109d7fc3f8abf2c846cf [file] [log] [blame]
Thomas Gleixner457c8992019-05-19 13:08:55 +01001// SPDX-License-Identifier: GPL-2.0-only
Linus Torvalds1da177e2005-04-16 15:20:36 -07002/*
3 * fs/libfs.c
4 * Library for filesystems writers.
5 */
6
Fabian Frederickac13a822014-06-04 16:06:27 -07007#include <linux/blkdev.h>
Paul Gortmaker630d9c42011-11-16 23:57:37 -05008#include <linux/export.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/pagemap.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090010#include <linux/slab.h>
Ingo Molnar5b825c32017-02-02 17:54:15 +010011#include <linux/cred.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/mount.h>
13#include <linux/vfs.h>
npiggin@suse.de7bb46a62010-05-27 01:05:33 +100014#include <linux/quotaops.h>
Ingo Molnar7cf34c72006-03-23 03:00:36 -080015#include <linux/mutex.h>
Al Viro87dc8002013-09-16 10:30:04 -040016#include <linux/namei.h>
Christoph Hellwig25961102007-10-21 16:42:05 -070017#include <linux/exportfs.h>
Al Virod5aacad2009-06-07 14:56:44 -040018#include <linux/writeback.h>
Al Viroff01bb42011-09-16 02:31:11 -040019#include <linux/buffer_head.h> /* sync_mapping_buffers */
David Howells31d6d5c2019-03-25 16:38:23 +000020#include <linux/fs_context.h>
21#include <linux/pseudo_fs.h>
Al Viroa3d1e7e2019-11-18 09:43:10 -050022#include <linux/fsnotify.h>
Daniel Rosenbergc8438432020-07-08 02:12:35 -070023#include <linux/unicode.h>
24#include <linux/fscrypt.h>
Ingo Molnar7cf34c72006-03-23 03:00:36 -080025
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080026#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027
Al Viroa4464db2011-07-07 15:03:58 -040028#include "internal.h"
29
David Howellsa528d352017-01-31 16:46:22 +000030int simple_getattr(const struct path *path, struct kstat *stat,
31 u32 request_mask, unsigned int query_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -070032{
David Howellsa528d352017-01-31 16:46:22 +000033 struct inode *inode = d_inode(path->dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -070034 generic_fillattr(inode, stat);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +030035 stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
Linus Torvalds1da177e2005-04-16 15:20:36 -070036 return 0;
37}
Al Viro12f38872013-09-15 21:20:49 -040038EXPORT_SYMBOL(simple_getattr);
Linus Torvalds1da177e2005-04-16 15:20:36 -070039
David Howells726c3342006-06-23 02:02:58 -070040int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070041{
David Howells726c3342006-06-23 02:02:58 -070042 buf->f_type = dentry->d_sb->s_magic;
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +030043 buf->f_bsize = PAGE_SIZE;
Linus Torvalds1da177e2005-04-16 15:20:36 -070044 buf->f_namelen = NAME_MAX;
45 return 0;
46}
Al Viro12f38872013-09-15 21:20:49 -040047EXPORT_SYMBOL(simple_statfs);
Linus Torvalds1da177e2005-04-16 15:20:36 -070048
49/*
50 * Retaining negative dentries for an in-memory filesystem just wastes
51 * memory and lookup time: arrange for them to be deleted immediately.
52 */
Al Virob26d4cd2013-10-25 18:47:37 -040053int always_delete_dentry(const struct dentry *dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -070054{
55 return 1;
56}
Al Virob26d4cd2013-10-25 18:47:37 -040057EXPORT_SYMBOL(always_delete_dentry);
58
59const struct dentry_operations simple_dentry_operations = {
60 .d_delete = always_delete_dentry,
61};
62EXPORT_SYMBOL(simple_dentry_operations);
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
64/*
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.
67 */
Al Viro00cd8dd2012-06-10 17:13:09 -040068struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -070069{
Linus Torvalds1da177e2005-04-16 15:20:36 -070070 if (dentry->d_name.len > NAME_MAX)
71 return ERR_PTR(-ENAMETOOLONG);
Al Viro74931da2013-07-14 17:43:25 +040072 if (!dentry->d_sb->s_d_op)
73 d_set_d_op(dentry, &simple_dentry_operations);
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 d_add(dentry, NULL);
75 return NULL;
76}
Al Viro12f38872013-09-15 21:20:49 -040077EXPORT_SYMBOL(simple_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -070078
Linus Torvalds1da177e2005-04-16 15:20:36 -070079int dcache_dir_open(struct inode *inode, struct file *file)
80{
Al Viroba65dc52016-06-10 11:32:47 -040081 file->private_data = d_alloc_cursor(file->f_path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -070082
83 return file->private_data ? 0 : -ENOMEM;
84}
Al Viro12f38872013-09-15 21:20:49 -040085EXPORT_SYMBOL(dcache_dir_open);
Linus Torvalds1da177e2005-04-16 15:20:36 -070086
87int dcache_dir_close(struct inode *inode, struct file *file)
88{
89 dput(file->private_data);
90 return 0;
91}
Al Viro12f38872013-09-15 21:20:49 -040092EXPORT_SYMBOL(dcache_dir_close);
Linus Torvalds1da177e2005-04-16 15:20:36 -070093
Al Viro4f42c1b2016-06-06 19:37:13 -040094/* parent is locked at least shared */
Al Virod4f4de52019-09-15 12:12:39 -040095/*
96 * Returns an element of siblings' list.
97 * We are looking for <count>th positive after <p>; if
Al Viro26b6c982019-09-20 16:32:42 -040098 * found, dentry is grabbed and returned to caller.
99 * If no such element exists, NULL is returned.
Al Virod4f4de52019-09-15 12:12:39 -0400100 */
Al Viro26b6c982019-09-20 16:32:42 -0400101static struct dentry *scan_positives(struct dentry *cursor,
Al Virod4f4de52019-09-15 12:12:39 -0400102 struct list_head *p,
103 loff_t count,
Al Viro26b6c982019-09-20 16:32:42 -0400104 struct dentry *last)
Al Viro4f42c1b2016-06-06 19:37:13 -0400105{
Al Virod4f4de52019-09-15 12:12:39 -0400106 struct dentry *dentry = cursor->d_parent, *found = NULL;
Al Viro4f42c1b2016-06-06 19:37:13 -0400107
Al Virod4f4de52019-09-15 12:12:39 -0400108 spin_lock(&dentry->d_lock);
109 while ((p = p->next) != &dentry->d_subdirs) {
Al Viro4f42c1b2016-06-06 19:37:13 -0400110 struct dentry *d = list_entry(p, struct dentry, d_child);
Al Virod4f4de52019-09-15 12:12:39 -0400111 // we must at least skip cursors, to avoid livelocks
112 if (d->d_flags & DCACHE_DENTRY_CURSOR)
113 continue;
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);
119 if (likely(found))
120 break;
121 count = 1;
122 }
123 if (need_resched()) {
124 list_move(&cursor->d_child, p);
125 p = &cursor->d_child;
126 spin_unlock(&dentry->d_lock);
127 cond_resched();
128 spin_lock(&dentry->d_lock);
Al Viro4f42c1b2016-06-06 19:37:13 -0400129 }
130 }
Al Virod4f4de52019-09-15 12:12:39 -0400131 spin_unlock(&dentry->d_lock);
Al Viro26b6c982019-09-20 16:32:42 -0400132 dput(last);
133 return found;
Al Viro4f42c1b2016-06-06 19:37:13 -0400134}
135
Andrew Morton965c8e52012-12-17 15:59:39 -0800136loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137{
Nick Piggin2fd6b7f2011-01-07 17:49:34 +1100138 struct dentry *dentry = file->f_path.dentry;
Andrew Morton965c8e52012-12-17 15:59:39 -0800139 switch (whence) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140 case 1:
141 offset += file->f_pos;
Gustavo A. R. Silvadf561f662020-08-23 17:36:59 -0500142 fallthrough;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 case 0:
144 if (offset >= 0)
145 break;
Gustavo A. R. Silvadf561f662020-08-23 17:36:59 -0500146 fallthrough;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147 default:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148 return -EINVAL;
149 }
150 if (offset != file->f_pos) {
Al Virod4f4de52019-09-15 12:12:39 -0400151 struct dentry *cursor = file->private_data;
152 struct dentry *to = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153
Al Virod4f4de52019-09-15 12:12:39 -0400154 inode_lock_shared(dentry->d_inode);
155
Al Viro26b6c982019-09-20 16:32:42 -0400156 if (offset > 2)
157 to = scan_positives(cursor, &dentry->d_subdirs,
158 offset - 2, NULL);
159 spin_lock(&dentry->d_lock);
160 if (to)
161 list_move(&cursor->d_child, &to->d_child);
162 else
Al Virod4f4de52019-09-15 12:12:39 -0400163 list_del_init(&cursor->d_child);
Al Viro26b6c982019-09-20 16:32:42 -0400164 spin_unlock(&dentry->d_lock);
Al Virod4f4de52019-09-15 12:12:39 -0400165 dput(to);
166
Al Viro26b6c982019-09-20 16:32:42 -0400167 file->f_pos = offset;
168
Al Virod4f4de52019-09-15 12:12:39 -0400169 inode_unlock_shared(dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 return offset;
172}
Al Viro12f38872013-09-15 21:20:49 -0400173EXPORT_SYMBOL(dcache_dir_lseek);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174
175/* Relationship between i_mode and the DT_xxx types */
176static inline unsigned char dt_type(struct inode *inode)
177{
178 return (inode->i_mode >> 12) & 15;
179}
180
181/*
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.
185 */
186
Al Viro5f99f4e2013-05-15 20:23:06 -0400187int dcache_readdir(struct file *file, struct dir_context *ctx)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188{
Al Viro5f99f4e2013-05-15 20:23:06 -0400189 struct dentry *dentry = file->f_path.dentry;
190 struct dentry *cursor = file->private_data;
Al Virod4f4de52019-09-15 12:12:39 -0400191 struct list_head *anchor = &dentry->d_subdirs;
192 struct dentry *next = NULL;
193 struct list_head *p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194
Al Viro5f99f4e2013-05-15 20:23:06 -0400195 if (!dir_emit_dots(file, ctx))
196 return 0;
Al Viro4f42c1b2016-06-06 19:37:13 -0400197
Al Viro5f99f4e2013-05-15 20:23:06 -0400198 if (ctx->pos == 2)
Al Virod4f4de52019-09-15 12:12:39 -0400199 p = anchor;
Al Viro26b6c982019-09-20 16:32:42 -0400200 else if (!list_empty(&cursor->d_child))
Al Virod4f4de52019-09-15 12:12:39 -0400201 p = &cursor->d_child;
Al Viro26b6c982019-09-20 16:32:42 -0400202 else
203 return 0;
Al Virod4f4de52019-09-15 12:12:39 -0400204
Al Viro26b6c982019-09-20 16:32:42 -0400205 while ((next = scan_positives(cursor, p, 1, next)) != NULL) {
Al Viro5f99f4e2013-05-15 20:23:06 -0400206 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
David Howellsdea655c2015-03-17 22:26:15 +0000207 d_inode(next)->i_ino, dt_type(d_inode(next))))
Al Viro4f42c1b2016-06-06 19:37:13 -0400208 break;
Al Viro5f99f4e2013-05-15 20:23:06 -0400209 ctx->pos++;
Al Viro26b6c982019-09-20 16:32:42 -0400210 p = &next->d_child;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 }
Al Virod4f4de52019-09-15 12:12:39 -0400212 spin_lock(&dentry->d_lock);
Al Viro26b6c982019-09-20 16:32:42 -0400213 if (next)
214 list_move_tail(&cursor->d_child, &next->d_child);
215 else
216 list_del_init(&cursor->d_child);
Al Virod4f4de52019-09-15 12:12:39 -0400217 spin_unlock(&dentry->d_lock);
218 dput(next);
219
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 return 0;
221}
Al Viro12f38872013-09-15 21:20:49 -0400222EXPORT_SYMBOL(dcache_readdir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223
224ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
225{
226 return -EISDIR;
227}
Al Viro12f38872013-09-15 21:20:49 -0400228EXPORT_SYMBOL(generic_read_dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800230const struct file_operations simple_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 .open = dcache_dir_open,
232 .release = dcache_dir_close,
233 .llseek = dcache_dir_lseek,
234 .read = generic_read_dir,
Al Viro4e829012016-04-20 19:52:15 -0400235 .iterate_shared = dcache_readdir,
Christoph Hellwig1b061d92010-05-26 17:53:41 +0200236 .fsync = noop_fsync,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237};
Al Viro12f38872013-09-15 21:20:49 -0400238EXPORT_SYMBOL(simple_dir_operations);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -0800240const struct inode_operations simple_dir_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 .lookup = simple_lookup,
242};
Al Viro12f38872013-09-15 21:20:49 -0400243EXPORT_SYMBOL(simple_dir_inode_operations);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244
Al Viroa3d1e7e2019-11-18 09:43:10 -0500245static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev)
246{
247 struct dentry *child = NULL;
248 struct list_head *p = prev ? &prev->d_child : &parent->d_subdirs;
249
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);
258 if (likely(child))
259 break;
260 }
261 }
262 spin_unlock(&parent->d_lock);
263 dput(prev);
264 return child;
265}
266
267void simple_recursive_removal(struct dentry *dentry,
268 void (*callback)(struct dentry *))
269{
270 struct dentry *this = dget(dentry);
271 while (true) {
272 struct dentry *victim = NULL, *child;
273 struct inode *inode = this->d_inode;
274
275 inode_lock(inode);
276 if (d_is_dir(this))
277 inode->i_flags |= S_DEAD;
278 while ((child = find_next_child(this, victim)) == NULL) {
279 // kill and ascend
280 // update metadata while it's still locked
281 inode->i_ctime = current_time(inode);
282 clear_nlink(inode);
283 inode_unlock(inode);
284 victim = this;
285 this = this->d_parent;
286 inode = this->d_inode;
287 inode_lock(inode);
288 if (simple_positive(victim)) {
289 d_invalidate(victim); // avoid lost mounts
290 if (d_is_dir(victim))
291 fsnotify_rmdir(inode, victim);
292 else
293 fsnotify_unlink(inode, victim);
294 if (callback)
295 callback(victim);
296 dput(victim); // unpin it
297 }
298 if (victim == dentry) {
299 inode->i_ctime = inode->i_mtime =
300 current_time(inode);
301 if (d_is_dir(dentry))
302 drop_nlink(inode);
303 inode_unlock(inode);
304 dput(dentry);
305 return;
306 }
307 }
308 inode_unlock(inode);
309 this = child;
310 }
311}
312EXPORT_SYMBOL(simple_recursive_removal);
313
Hugh Dickins759b9772007-03-05 00:30:28 -0800314static const struct super_operations simple_super_operations = {
315 .statfs = simple_statfs,
316};
317
David Howellsdb2c2462019-03-25 16:38:26 +0000318static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319{
David Howells31d6d5c2019-03-25 16:38:23 +0000320 struct pseudo_fs_context *ctx = fc->fs_private;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 struct inode *root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322
Jeff Layton89a4eb42009-08-18 14:11:08 -0700323 s->s_maxbytes = MAX_LFS_FILESIZE;
Alex Nixon3971e1a2008-07-29 22:33:03 -0700324 s->s_blocksize = PAGE_SIZE;
325 s->s_blocksize_bits = PAGE_SHIFT;
Al Viro8d9e46d2019-05-11 11:43:59 -0400326 s->s_magic = ctx->magic;
327 s->s_op = ctx->ops ?: &simple_super_operations;
328 s->s_xattr = ctx->xattr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 s->s_time_gran = 1;
330 root = new_inode(s);
331 if (!root)
David Howellsdb2c2462019-03-25 16:38:26 +0000332 return -ENOMEM;
333
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700334 /*
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).
338 */
339 root->i_ino = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
Deepa Dinamani078cd822016-09-14 07:48:04 -0700341 root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
Al Viro8d9e46d2019-05-11 11:43:59 -0400342 s->s_root = d_make_root(root);
343 if (!s->s_root)
David Howellsdb2c2462019-03-25 16:38:26 +0000344 return -ENOMEM;
Al Viro8d9e46d2019-05-11 11:43:59 -0400345 s->s_d_op = ctx->dops;
David Howells31d6d5c2019-03-25 16:38:23 +0000346 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347}
Al Viro8d9e46d2019-05-11 11:43:59 -0400348
David Howellsdb2c2462019-03-25 16:38:26 +0000349static int pseudo_fs_get_tree(struct fs_context *fc)
350{
Al Viro2ac295d2019-06-01 20:48:55 -0400351 return get_tree_nodev(fc, pseudo_fs_fill_super);
David Howells31d6d5c2019-03-25 16:38:23 +0000352}
353
354static void pseudo_fs_free(struct fs_context *fc)
355{
356 kfree(fc->fs_private);
357}
358
359static const struct fs_context_operations pseudo_fs_context_ops = {
360 .free = pseudo_fs_free,
361 .get_tree = pseudo_fs_get_tree,
362};
363
364/*
365 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
366 * will never be mountable)
367 */
368struct pseudo_fs_context *init_pseudo(struct fs_context *fc,
369 unsigned long magic)
370{
371 struct pseudo_fs_context *ctx;
372
373 ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL);
374 if (likely(ctx)) {
375 ctx->magic = magic;
376 fc->fs_private = ctx;
377 fc->ops = &pseudo_fs_context_ops;
David Howellsdb2c2462019-03-25 16:38:26 +0000378 fc->sb_flags |= SB_NOUSER;
379 fc->global = true;
David Howells31d6d5c2019-03-25 16:38:23 +0000380 }
381 return ctx;
382}
383EXPORT_SYMBOL(init_pseudo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384
Stephen Boyd20955e82012-04-05 14:25:09 -0700385int simple_open(struct inode *inode, struct file *file)
386{
387 if (inode->i_private)
388 file->private_data = inode->i_private;
389 return 0;
390}
Al Viro12f38872013-09-15 21:20:49 -0400391EXPORT_SYMBOL(simple_open);
Stephen Boyd20955e82012-04-05 14:25:09 -0700392
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
394{
David Howellsdea655c2015-03-17 22:26:15 +0000395 struct inode *inode = d_inode(old_dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396
Deepa Dinamani078cd822016-09-14 07:48:04 -0700397 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
Dave Hansend8c76e62006-09-30 23:29:04 -0700398 inc_nlink(inode);
Al Viro7de9c6ee2010-10-23 11:11:40 -0400399 ihold(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 dget(dentry);
401 d_instantiate(dentry, inode);
402 return 0;
403}
Al Viro12f38872013-09-15 21:20:49 -0400404EXPORT_SYMBOL(simple_link);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406int simple_empty(struct dentry *dentry)
407{
408 struct dentry *child;
409 int ret = 0;
410
Nick Piggin2fd6b7f2011-01-07 17:49:34 +1100411 spin_lock(&dentry->d_lock);
Al Viro946e51f2014-10-26 19:19:16 -0400412 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
Nick Pigginda502952011-01-07 17:49:33 +1100413 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
414 if (simple_positive(child)) {
415 spin_unlock(&child->d_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 goto out;
Nick Pigginda502952011-01-07 17:49:33 +1100417 }
418 spin_unlock(&child->d_lock);
419 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 ret = 1;
421out:
Nick Piggin2fd6b7f2011-01-07 17:49:34 +1100422 spin_unlock(&dentry->d_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 return ret;
424}
Al Viro12f38872013-09-15 21:20:49 -0400425EXPORT_SYMBOL(simple_empty);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426
427int simple_unlink(struct inode *dir, struct dentry *dentry)
428{
David Howellsdea655c2015-03-17 22:26:15 +0000429 struct inode *inode = d_inode(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430
Deepa Dinamani078cd822016-09-14 07:48:04 -0700431 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700432 drop_nlink(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 dput(dentry);
434 return 0;
435}
Al Viro12f38872013-09-15 21:20:49 -0400436EXPORT_SYMBOL(simple_unlink);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437
438int simple_rmdir(struct inode *dir, struct dentry *dentry)
439{
440 if (!simple_empty(dentry))
441 return -ENOTEMPTY;
442
David Howellsdea655c2015-03-17 22:26:15 +0000443 drop_nlink(d_inode(dentry));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 simple_unlink(dir, dentry);
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700445 drop_nlink(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 return 0;
447}
Al Viro12f38872013-09-15 21:20:49 -0400448EXPORT_SYMBOL(simple_rmdir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449
450int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
Miklos Szeredie0e0be82016-09-27 11:03:57 +0200451 struct inode *new_dir, struct dentry *new_dentry,
452 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453{
David Howellsdea655c2015-03-17 22:26:15 +0000454 struct inode *inode = d_inode(old_dentry);
David Howellse36cb0b2015-01-29 12:02:35 +0000455 int they_are_dirs = d_is_dir(old_dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456
Miklos Szeredie0e0be82016-09-27 11:03:57 +0200457 if (flags & ~RENAME_NOREPLACE)
458 return -EINVAL;
459
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 if (!simple_empty(new_dentry))
461 return -ENOTEMPTY;
462
David Howellsdea655c2015-03-17 22:26:15 +0000463 if (d_really_is_positive(new_dentry)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 simple_unlink(new_dir, new_dentry);
Al Viro841590c2011-07-21 15:49:09 -0400465 if (they_are_dirs) {
David Howellsdea655c2015-03-17 22:26:15 +0000466 drop_nlink(d_inode(new_dentry));
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700467 drop_nlink(old_dir);
Al Viro841590c2011-07-21 15:49:09 -0400468 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 } else if (they_are_dirs) {
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700470 drop_nlink(old_dir);
Dave Hansend8c76e62006-09-30 23:29:04 -0700471 inc_nlink(new_dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 }
473
474 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
Deepa Dinamani078cd822016-09-14 07:48:04 -0700475 new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476
477 return 0;
478}
Al Viro12f38872013-09-15 21:20:49 -0400479EXPORT_SYMBOL(simple_rename);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000481/**
Christoph Hellwigeef23802010-06-04 11:30:01 +0200482 * simple_setattr - setattr for simple filesystem
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000483 * @dentry: dentry
484 * @iattr: iattr structure
485 *
486 * Returns 0 on success, -error on failure.
487 *
Christoph Hellwigeef23802010-06-04 11:30:01 +0200488 * simple_setattr is a simple ->setattr implementation without a proper
489 * implementation of size changes.
490 *
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.
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000494 */
495int simple_setattr(struct dentry *dentry, struct iattr *iattr)
496{
David Howellsdea655c2015-03-17 22:26:15 +0000497 struct inode *inode = d_inode(dentry);
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000498 int error;
499
Jan Kara31051c82016-05-26 16:55:18 +0200500 error = setattr_prepare(dentry, iattr);
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000501 if (error)
502 return error;
503
Christoph Hellwig2c27c652010-06-04 11:30:04 +0200504 if (iattr->ia_valid & ATTR_SIZE)
505 truncate_setsize(inode, iattr->ia_size);
Christoph Hellwig6a1a90a2010-06-04 11:30:00 +0200506 setattr_copy(inode, iattr);
Christoph Hellwigeef23802010-06-04 11:30:01 +0200507 mark_inode_dirty(inode);
508 return 0;
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000509}
510EXPORT_SYMBOL(simple_setattr);
511
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512int simple_readpage(struct file *file, struct page *page)
513{
Pekka J Enbergc0d92cb2006-09-29 01:59:09 -0700514 clear_highpage(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 flush_dcache_page(page);
516 SetPageUptodate(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 unlock_page(page);
518 return 0;
519}
Al Viro12f38872013-09-15 21:20:49 -0400520EXPORT_SYMBOL(simple_readpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521
Nick Pigginafddba42007-10-16 01:25:01 -0700522int simple_write_begin(struct file *file, struct address_space *mapping,
523 loff_t pos, unsigned len, unsigned flags,
524 struct page **pagep, void **fsdata)
525{
526 struct page *page;
527 pgoff_t index;
Nick Pigginafddba42007-10-16 01:25:01 -0700528
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300529 index = pos >> PAGE_SHIFT;
Nick Pigginafddba42007-10-16 01:25:01 -0700530
Nick Piggin54566b22009-01-04 12:00:53 -0800531 page = grab_cache_page_write_begin(mapping, index, flags);
Nick Pigginafddba42007-10-16 01:25:01 -0700532 if (!page)
533 return -ENOMEM;
534
535 *pagep = page;
536
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300537 if (!PageUptodate(page) && (len != PAGE_SIZE)) {
538 unsigned from = pos & (PAGE_SIZE - 1);
Boaz Harrosh193cf4b2010-01-12 16:18:08 +0200539
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300540 zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
Boaz Harrosh193cf4b2010-01-12 16:18:08 +0200541 }
542 return 0;
Nick Pigginafddba42007-10-16 01:25:01 -0700543}
Al Viro12f38872013-09-15 21:20:49 -0400544EXPORT_SYMBOL(simple_write_begin);
Nick Pigginafddba42007-10-16 01:25:01 -0700545
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200546/**
547 * simple_write_end - .write_end helper for non-block-device FSes
Randy Dunlap8e88bfb2019-10-14 14:12:14 -0700548 * @file: See .write_end of address_space_operations
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200549 * @mapping: "
550 * @pos: "
551 * @len: "
552 * @copied: "
553 * @page: "
554 * @fsdata: "
555 *
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.
Al Viro04fff642016-08-29 22:39:56 -0400565 *
566 * Use *ONLY* with simple_readpage()
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200567 */
568int simple_write_end(struct file *file, struct address_space *mapping,
569 loff_t pos, unsigned len, unsigned copied,
570 struct page *page, void *fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571{
572 struct inode *inode = page->mapping->host;
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200573 loff_t last_pos = pos + copied;
574
575 /* zero the stale part of the page if we did a short copy */
Al Viro04fff642016-08-29 22:39:56 -0400576 if (!PageUptodate(page)) {
577 if (copied < len) {
578 unsigned from = pos & (PAGE_SIZE - 1);
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200579
Al Viro04fff642016-08-29 22:39:56 -0400580 zero_user(page, from + copied, len - copied);
581 }
Nick Piggin955eff52007-02-20 13:58:08 -0800582 SetPageUptodate(page);
Al Viro04fff642016-08-29 22:39:56 -0400583 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 /*
585 * No need to use i_size_read() here, the i_size
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800586 * cannot change under us because we hold the i_mutex.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 */
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200588 if (last_pos > inode->i_size)
589 i_size_write(inode, last_pos);
590
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 set_page_dirty(page);
Nick Pigginafddba42007-10-16 01:25:01 -0700592 unlock_page(page);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300593 put_page(page);
Nick Pigginafddba42007-10-16 01:25:01 -0700594
595 return copied;
596}
Al Viro12f38872013-09-15 21:20:49 -0400597EXPORT_SYMBOL(simple_write_end);
Nick Pigginafddba42007-10-16 01:25:01 -0700598
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700599/*
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.
603 */
Roberto Sassu7d683a02010-06-03 11:58:28 +0200604int simple_fill_super(struct super_block *s, unsigned long magic,
Eric Biggerscda37122017-03-25 21:15:37 -0700605 const struct tree_descr *files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607 struct inode *inode;
608 struct dentry *root;
609 struct dentry *dentry;
610 int i;
611
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300612 s->s_blocksize = PAGE_SIZE;
613 s->s_blocksize_bits = PAGE_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 s->s_magic = magic;
Hugh Dickins759b9772007-03-05 00:30:28 -0800615 s->s_op = &simple_super_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 s->s_time_gran = 1;
617
618 inode = new_inode(s);
619 if (!inode)
620 return -ENOMEM;
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700621 /*
622 * because the root inode is 1, the files array must not contain an
623 * entry at index 1
624 */
625 inode->i_ino = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 inode->i_mode = S_IFDIR | 0755;
Deepa Dinamani078cd822016-09-14 07:48:04 -0700627 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 inode->i_op = &simple_dir_inode_operations;
629 inode->i_fop = &simple_dir_operations;
Miklos Szeredibfe86842011-10-28 14:13:29 +0200630 set_nlink(inode, 2);
Al Viro48fde702012-01-08 22:15:13 -0500631 root = d_make_root(inode);
632 if (!root)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 for (i = 0; !files->name || files->name[0]; i++, files++) {
635 if (!files->name)
636 continue;
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700637
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__,
642 s->s_type->name);
643
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 dentry = d_alloc_name(root, files->name);
645 if (!dentry)
646 goto out;
647 inode = new_inode(s);
Konstantin Khlebnikov32096ea2011-11-01 16:12:33 +0300648 if (!inode) {
649 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 goto out;
Konstantin Khlebnikov32096ea2011-11-01 16:12:33 +0300651 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 inode->i_mode = S_IFREG | files->mode;
Deepa Dinamani078cd822016-09-14 07:48:04 -0700653 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 inode->i_fop = files->ops;
655 inode->i_ino = i;
656 d_add(dentry, inode);
657 }
658 s->s_root = root;
659 return 0;
660out:
661 d_genocide(root);
Al Viro640946f22012-04-02 19:22:25 -0400662 shrink_dcache_parent(root);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 dput(root);
664 return -ENOMEM;
665}
Al Viro12f38872013-09-15 21:20:49 -0400666EXPORT_SYMBOL(simple_fill_super);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667
668static DEFINE_SPINLOCK(pin_fs_lock);
669
Trond Myklebust1f5ce9e2006-06-09 09:34:16 -0400670int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671{
672 struct vfsmount *mnt = NULL;
673 spin_lock(&pin_fs_lock);
674 if (unlikely(!*mount)) {
675 spin_unlock(&pin_fs_lock);
Linus Torvalds1751e8a2017-11-27 13:05:09 -0800676 mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 if (IS_ERR(mnt))
678 return PTR_ERR(mnt);
679 spin_lock(&pin_fs_lock);
680 if (!*mount)
681 *mount = mnt;
682 }
683 mntget(*mount);
684 ++*count;
685 spin_unlock(&pin_fs_lock);
686 mntput(mnt);
687 return 0;
688}
Al Viro12f38872013-09-15 21:20:49 -0400689EXPORT_SYMBOL(simple_pin_fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690
691void simple_release_fs(struct vfsmount **mount, int *count)
692{
693 struct vfsmount *mnt;
694 spin_lock(&pin_fs_lock);
695 mnt = *mount;
696 if (!--*count)
697 *mount = NULL;
698 spin_unlock(&pin_fs_lock);
699 mntput(mnt);
700}
Al Viro12f38872013-09-15 21:20:49 -0400701EXPORT_SYMBOL(simple_release_fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702
Akinobu Mita6d1029b2008-07-04 09:59:51 -0700703/**
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
710 *
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.
713 *
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.
716 **/
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
718 const void *from, size_t available)
719{
720 loff_t pos = *ppos;
Steven Rostedt14be2742009-09-18 13:05:42 -0700721 size_t ret;
722
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 if (pos < 0)
724 return -EINVAL;
Steven Rostedt14be2742009-09-18 13:05:42 -0700725 if (pos >= available || !count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 return 0;
727 if (count > available - pos)
728 count = available - pos;
Steven Rostedt14be2742009-09-18 13:05:42 -0700729 ret = copy_to_user(to, from + pos, count);
730 if (ret == count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 return -EFAULT;
Steven Rostedt14be2742009-09-18 13:05:42 -0700732 count -= ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 *ppos = pos + count;
734 return count;
735}
Al Viro12f38872013-09-15 21:20:49 -0400736EXPORT_SYMBOL(simple_read_from_buffer);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737
Akinobu Mita6d1029b2008-07-04 09:59:51 -0700738/**
Jiri Slaby6a727b42010-05-01 23:51:22 +0200739 * 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
745 *
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.
748 *
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.
751 **/
752ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
753 const void __user *from, size_t count)
754{
755 loff_t pos = *ppos;
756 size_t res;
757
758 if (pos < 0)
759 return -EINVAL;
760 if (pos >= available || !count)
761 return 0;
762 if (count > available - pos)
763 count = available - pos;
764 res = copy_from_user(to + pos, from, count);
765 if (res == count)
766 return -EFAULT;
767 count -= res;
768 *ppos = pos + count;
769 return count;
770}
Al Viro12f38872013-09-15 21:20:49 -0400771EXPORT_SYMBOL(simple_write_to_buffer);
Jiri Slaby6a727b42010-05-01 23:51:22 +0200772
773/**
Akinobu Mita6d1029b2008-07-04 09:59:51 -0700774 * 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
780 *
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.
783 *
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.
786 **/
Akinobu Mita93b07112008-06-05 22:46:21 -0700787ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
788 const void *from, size_t available)
789{
790 loff_t pos = *ppos;
791
792 if (pos < 0)
793 return -EINVAL;
794 if (pos >= available)
795 return 0;
796 if (count > available - pos)
797 count = available - pos;
798 memcpy(to, from + pos, count);
799 *ppos = pos + count;
800
801 return count;
802}
Al Viro12f38872013-09-15 21:20:49 -0400803EXPORT_SYMBOL(memory_read_from_buffer);
Akinobu Mita93b07112008-06-05 22:46:21 -0700804
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805/*
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
809 * file-local buffer.
810 */
Ingo Molnar76791ab2009-03-25 16:48:35 +0100811
812void simple_transaction_set(struct file *file, size_t n)
813{
814 struct simple_transaction_argresp *ar = file->private_data;
815
816 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
817
818 /*
819 * The barrier ensures that ar->size will really remain zero until
820 * ar->data is ready for reading.
821 */
822 smp_mb();
823 ar->size = n;
824}
Al Viro12f38872013-09-15 21:20:49 -0400825EXPORT_SYMBOL(simple_transaction_set);
Ingo Molnar76791ab2009-03-25 16:48:35 +0100826
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
828{
829 struct simple_transaction_argresp *ar;
830 static DEFINE_SPINLOCK(simple_transaction_lock);
831
832 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
833 return ERR_PTR(-EFBIG);
834
835 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
836 if (!ar)
837 return ERR_PTR(-ENOMEM);
838
839 spin_lock(&simple_transaction_lock);
840
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);
846 }
847
848 file->private_data = ar;
849
850 spin_unlock(&simple_transaction_lock);
851
852 if (copy_from_user(ar->data, buf, size))
853 return ERR_PTR(-EFAULT);
854
855 return ar->data;
856}
Al Viro12f38872013-09-15 21:20:49 -0400857EXPORT_SYMBOL(simple_transaction_get);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858
859ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
860{
861 struct simple_transaction_argresp *ar = file->private_data;
862
863 if (!ar)
864 return 0;
865 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
866}
Al Viro12f38872013-09-15 21:20:49 -0400867EXPORT_SYMBOL(simple_transaction_read);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868
869int simple_transaction_release(struct inode *inode, struct file *file)
870{
871 free_page((unsigned long)file->private_data);
872 return 0;
873}
Al Viro12f38872013-09-15 21:20:49 -0400874EXPORT_SYMBOL(simple_transaction_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200876/* Simple attribute files */
877
878struct simple_attr {
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800879 int (*get)(void *, u64 *);
880 int (*set)(void *, u64);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200881 char get_buf[24]; /* enough to store a u64 and "\n\0" */
882 char set_buf[24];
883 void *data;
884 const char *fmt; /* format for read operation */
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800885 struct mutex mutex; /* protects access to these buffers */
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200886};
887
888/* simple_attr_open is called by an actual attribute open file operation
889 * to set the attribute specific access operations. */
890int simple_attr_open(struct inode *inode, struct file *file,
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800891 int (*get)(void *, u64 *), int (*set)(void *, u64),
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200892 const char *fmt)
893{
894 struct simple_attr *attr;
895
Eric Biggersa65cab72020-03-07 18:38:49 -0800896 attr = kzalloc(sizeof(*attr), GFP_KERNEL);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200897 if (!attr)
898 return -ENOMEM;
899
900 attr->get = get;
901 attr->set = set;
Theodore Ts'o8e18e292006-09-27 01:50:46 -0700902 attr->data = inode->i_private;
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200903 attr->fmt = fmt;
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800904 mutex_init(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200905
906 file->private_data = attr;
907
908 return nonseekable_open(inode, file);
909}
Al Viro12f38872013-09-15 21:20:49 -0400910EXPORT_SYMBOL_GPL(simple_attr_open);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200911
Christoph Hellwig74bedc42008-02-08 04:20:28 -0800912int simple_attr_release(struct inode *inode, struct file *file)
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200913{
914 kfree(file->private_data);
915 return 0;
916}
Al Viro12f38872013-09-15 21:20:49 -0400917EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200918
919/* read from the buffer that is filled with the get function */
920ssize_t simple_attr_read(struct file *file, char __user *buf,
921 size_t len, loff_t *ppos)
922{
923 struct simple_attr *attr;
924 size_t size;
925 ssize_t ret;
926
927 attr = file->private_data;
928
929 if (!attr->get)
930 return -EACCES;
931
Christoph Hellwig92613032008-02-08 04:20:27 -0800932 ret = mutex_lock_interruptible(&attr->mutex);
933 if (ret)
934 return ret;
935
Eric Biggersa65cab72020-03-07 18:38:49 -0800936 if (*ppos && attr->get_buf[0]) {
937 /* continued read */
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200938 size = strlen(attr->get_buf);
Eric Biggersa65cab72020-03-07 18:38:49 -0800939 } else {
940 /* first read */
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800941 u64 val;
942 ret = attr->get(attr->data, &val);
943 if (ret)
944 goto out;
945
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200946 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800947 attr->fmt, (unsigned long long)val);
948 }
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200949
950 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
Christoph Hellwig8b88b092008-02-08 04:20:26 -0800951out:
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800952 mutex_unlock(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200953 return ret;
954}
Al Viro12f38872013-09-15 21:20:49 -0400955EXPORT_SYMBOL_GPL(simple_attr_read);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200956
957/* interpret the buffer as a number to call the set function with */
958ssize_t simple_attr_write(struct file *file, const char __user *buf,
959 size_t len, loff_t *ppos)
960{
961 struct simple_attr *attr;
Yicong Yang488dac02020-11-21 22:17:19 -0800962 unsigned long long val;
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200963 size_t size;
964 ssize_t ret;
965
966 attr = file->private_data;
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200967 if (!attr->set)
968 return -EACCES;
969
Christoph Hellwig92613032008-02-08 04:20:27 -0800970 ret = mutex_lock_interruptible(&attr->mutex);
971 if (ret)
972 return ret;
973
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200974 ret = -EFAULT;
975 size = min(sizeof(attr->set_buf) - 1, len);
976 if (copy_from_user(attr->set_buf, buf, size))
977 goto out;
978
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200979 attr->set_buf[size] = '\0';
Yicong Yang488dac02020-11-21 22:17:19 -0800980 ret = kstrtoull(attr->set_buf, 0, &val);
981 if (ret)
982 goto out;
Wu Fengguang05cc0ce2009-09-18 13:06:03 -0700983 ret = attr->set(attr->data, val);
984 if (ret == 0)
985 ret = len; /* on success, claim we got the whole input */
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200986out:
Ingo Molnar7cf34c72006-03-23 03:00:36 -0800987 mutex_unlock(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200988 return ret;
989}
Al Viro12f38872013-09-15 21:20:49 -0400990EXPORT_SYMBOL_GPL(simple_attr_write);
Arnd Bergmannacaefc22005-05-18 14:40:59 +0200991
Christoph Hellwig25961102007-10-21 16:42:05 -0700992/**
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
999 *
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.
1003 */
1004struct 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))
1007{
1008 struct inode *inode = NULL;
1009
1010 if (fh_len < 2)
1011 return NULL;
1012
1013 switch (fh_type) {
1014 case FILEID_INO32_GEN:
1015 case FILEID_INO32_GEN_PARENT:
1016 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
1017 break;
1018 }
1019
Christoph Hellwig4ea3ada2008-08-11 15:48:57 +02001020 return d_obtain_alias(inode);
Christoph Hellwig25961102007-10-21 16:42:05 -07001021}
1022EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
1023
1024/**
Yanchuan Nianca186832012-09-05 16:31:29 +08001025 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
Christoph Hellwig25961102007-10-21 16:42:05 -07001026 * @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
1031 *
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.
1036 */
1037struct 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))
1040{
1041 struct inode *inode = NULL;
1042
1043 if (fh_len <= 2)
1044 return NULL;
1045
1046 switch (fh_type) {
1047 case FILEID_INO32_GEN_PARENT:
1048 inode = get_inode(sb, fid->i32.parent_ino,
1049 (fh_len > 3 ? fid->i32.parent_gen : 0));
1050 break;
1051 }
1052
Christoph Hellwig4ea3ada2008-08-11 15:48:57 +02001053 return d_obtain_alias(inode);
Christoph Hellwig25961102007-10-21 16:42:05 -07001054}
1055EXPORT_SYMBOL_GPL(generic_fh_to_parent);
1056
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001057/**
Fabian Frederickac13a822014-06-04 16:06:27 -07001058 * __generic_file_fsync - generic fsync implementation for simple filesystems
1059 *
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001060 * @file: file to synchronize
Fabian Frederickac13a822014-06-04 16:06:27 -07001061 * @start: start offset in bytes
1062 * @end: end offset in bytes (inclusive)
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001063 * @datasync: only synchronize essential metadata if true
1064 *
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.
1068 */
Fabian Frederickac13a822014-06-04 16:06:27 -07001069int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
1070 int datasync)
Al Virod5aacad2009-06-07 14:56:44 -04001071{
Christoph Hellwig7ea80852010-05-26 17:53:25 +02001072 struct inode *inode = file->f_mapping->host;
Al Virod5aacad2009-06-07 14:56:44 -04001073 int err;
1074 int ret;
1075
Jeff Layton383aa542017-07-06 07:02:29 -04001076 err = file_write_and_wait_range(file, start, end);
Josef Bacik02c24a82011-07-16 20:44:56 -04001077 if (err)
1078 return err;
1079
Al Viro59551022016-01-22 15:40:57 -05001080 inode_lock(inode);
Al Virod5aacad2009-06-07 14:56:44 -04001081 ret = sync_mapping_buffers(inode->i_mapping);
Theodore Ts'o0ae45f62015-02-02 00:37:00 -05001082 if (!(inode->i_state & I_DIRTY_ALL))
Josef Bacik02c24a82011-07-16 20:44:56 -04001083 goto out;
Al Virod5aacad2009-06-07 14:56:44 -04001084 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
Josef Bacik02c24a82011-07-16 20:44:56 -04001085 goto out;
Al Virod5aacad2009-06-07 14:56:44 -04001086
Christoph Hellwigc37650162010-10-06 10:48:20 +02001087 err = sync_inode_metadata(inode, 1);
Al Virod5aacad2009-06-07 14:56:44 -04001088 if (ret == 0)
1089 ret = err;
Fabian Frederickac13a822014-06-04 16:06:27 -07001090
Josef Bacik02c24a82011-07-16 20:44:56 -04001091out:
Al Viro59551022016-01-22 15:40:57 -05001092 inode_unlock(inode);
Jeff Layton383aa542017-07-06 07:02:29 -04001093 /* check and advance again to catch errors after syncing out buffers */
1094 err = file_check_and_advance_wb_err(file);
1095 if (ret == 0)
1096 ret = err;
1097 return ret;
Al Virod5aacad2009-06-07 14:56:44 -04001098}
Fabian Frederickac13a822014-06-04 16:06:27 -07001099EXPORT_SYMBOL(__generic_file_fsync);
1100
1101/**
1102 * generic_file_fsync - generic fsync implementation for simple filesystems
1103 * with flush
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
1108 *
1109 */
1110
1111int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1112 int datasync)
1113{
1114 struct inode *inode = file->f_mapping->host;
1115 int err;
1116
1117 err = __generic_file_fsync(file, start, end, datasync);
1118 if (err)
1119 return err;
Christoph Hellwig93985542020-05-13 14:36:00 +02001120 return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL);
Fabian Frederickac13a822014-06-04 16:06:27 -07001121}
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001122EXPORT_SYMBOL(generic_file_fsync);
1123
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07001124/**
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
1128 *
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.
1132 */
1133int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1134{
1135 u64 last_fs_block = num_blocks - 1;
Joel Beckera33f13e2010-08-16 12:10:17 -07001136 u64 last_fs_page =
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001137 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07001138
1139 if (unlikely(num_blocks == 0))
1140 return 0;
1141
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001142 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07001143 return -EINVAL;
1144
Joel Beckera33f13e2010-08-16 12:10:17 -07001145 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1146 (last_fs_page > (pgoff_t)(~0ULL))) {
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07001147 return -EFBIG;
1148 }
1149 return 0;
1150}
1151EXPORT_SYMBOL(generic_check_addressable);
1152
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001153/*
1154 * No-op implementation of ->fsync for in-memory filesystems.
1155 */
Josef Bacik02c24a82011-07-16 20:44:56 -04001156int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001157{
1158 return 0;
1159}
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001160EXPORT_SYMBOL(noop_fsync);
Al Viro87dc8002013-09-16 10:30:04 -04001161
Dan Williamsf44c7762018-03-07 15:26:44 -08001162int noop_set_page_dirty(struct page *page)
1163{
1164 /*
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.
1170 *
1171 * This callback is defined to prevent fallback to
1172 * __set_page_dirty_buffers() in set_page_dirty().
1173 */
1174 return 0;
1175}
1176EXPORT_SYMBOL_GPL(noop_set_page_dirty);
1177
1178void noop_invalidatepage(struct page *page, unsigned int offset,
1179 unsigned int length)
1180{
1181 /*
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().
1185 */
1186}
1187EXPORT_SYMBOL_GPL(noop_invalidatepage);
1188
1189ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1190{
1191 /*
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.
1196 */
1197 return -EINVAL;
1198}
1199EXPORT_SYMBOL_GPL(noop_direct_IO);
1200
Al Virofceef392015-12-29 15:58:39 -05001201/* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1202void kfree_link(void *p)
Al Viro87dc8002013-09-16 10:30:04 -04001203{
Al Virofceef392015-12-29 15:58:39 -05001204 kfree(p);
Al Viro87dc8002013-09-16 10:30:04 -04001205}
Al Virofceef392015-12-29 15:58:39 -05001206EXPORT_SYMBOL(kfree_link);
Al Viro69878432013-10-02 22:35:11 -04001207
1208/*
1209 * nop .set_page_dirty method so that people can use .page_mkwrite on
1210 * anon inodes.
1211 */
1212static int anon_set_page_dirty(struct page *page)
1213{
1214 return 0;
1215};
1216
1217/*
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.
1221 */
1222struct inode *alloc_anon_inode(struct super_block *s)
1223{
1224 static const struct address_space_operations anon_aops = {
1225 .set_page_dirty = anon_set_page_dirty,
1226 };
1227 struct inode *inode = new_inode_pseudo(s);
1228
1229 if (!inode)
1230 return ERR_PTR(-ENOMEM);
1231
1232 inode->i_ino = get_next_ino();
1233 inode->i_mapping->a_ops = &anon_aops;
1234
1235 /*
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.
1240 */
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;
Deepa Dinamani078cd822016-09-14 07:48:04 -07001246 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
Al Viro69878432013-10-02 22:35:11 -04001247 return inode;
1248}
1249EXPORT_SYMBOL(alloc_anon_inode);
Jeff Layton1c994a02014-08-27 06:49:41 -04001250
1251/**
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
Jeff Laytone6f5c782014-08-22 10:40:25 -04001256 * @priv: private data for lm_setup operation
Jeff Layton1c994a02014-08-27 06:49:41 -04001257 *
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.
1260 */
1261int
Jeff Laytone6f5c782014-08-22 10:40:25 -04001262simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1263 void **priv)
Jeff Layton1c994a02014-08-27 06:49:41 -04001264{
1265 return -EINVAL;
1266}
1267EXPORT_SYMBOL(simple_nosetlease);
Al Viro61ba64f2015-05-02 09:54:06 -04001268
Eric Biggers6ee97062019-04-11 16:16:30 -07001269/**
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
1274 *
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.
1280 *
1281 * Return: the symlink target
1282 */
Al Viro6b255392015-11-17 10:20:54 -05001283const char *simple_get_link(struct dentry *dentry, struct inode *inode,
Al Virofceef392015-12-29 15:58:39 -05001284 struct delayed_call *done)
Al Viro61ba64f2015-05-02 09:54:06 -04001285{
Al Viro6b255392015-11-17 10:20:54 -05001286 return inode->i_link;
Al Viro61ba64f2015-05-02 09:54:06 -04001287}
Al Viro6b255392015-11-17 10:20:54 -05001288EXPORT_SYMBOL(simple_get_link);
Al Viro61ba64f2015-05-02 09:54:06 -04001289
1290const struct inode_operations simple_symlink_inode_operations = {
Al Viro6b255392015-11-17 10:20:54 -05001291 .get_link = simple_get_link,
Al Viro61ba64f2015-05-02 09:54:06 -04001292};
1293EXPORT_SYMBOL(simple_symlink_inode_operations);
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001294
1295/*
1296 * Operations for a permanently empty directory.
1297 */
1298static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1299{
1300 return ERR_PTR(-ENOENT);
1301}
1302
David Howellsa528d352017-01-31 16:46:22 +00001303static int empty_dir_getattr(const struct path *path, struct kstat *stat,
1304 u32 request_mask, unsigned int query_flags)
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001305{
David Howellsa528d352017-01-31 16:46:22 +00001306 struct inode *inode = d_inode(path->dentry);
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001307 generic_fillattr(inode, stat);
1308 return 0;
1309}
1310
1311static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1312{
1313 return -EPERM;
1314}
1315
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001316static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1317{
1318 return -EOPNOTSUPP;
1319}
1320
1321static 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,
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001326 .listxattr = empty_dir_listxattr,
1327};
1328
1329static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1330{
1331 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1332 return generic_file_llseek_size(file, offset, whence, 2, 2);
1333}
1334
1335static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1336{
1337 dir_emit_dots(file, ctx);
1338 return 0;
1339}
1340
1341static const struct file_operations empty_dir_operations = {
1342 .llseek = empty_dir_llseek,
1343 .read = generic_read_dir,
Al Viroc51da202016-04-30 22:37:34 -04001344 .iterate_shared = empty_dir_readdir,
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001345 .fsync = noop_fsync,
1346};
1347
1348
1349void make_empty_dir_inode(struct inode *inode)
1350{
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;
1355 inode->i_rdev = 0;
Eric W. Biederman4b75de862015-08-12 15:00:12 -05001356 inode->i_size = 0;
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001357 inode->i_blkbits = PAGE_SHIFT;
1358 inode->i_blocks = 0;
1359
1360 inode->i_op = &empty_dir_inode_operations;
Andreas Gruenbacherf5c24432016-09-29 17:48:41 +02001361 inode->i_opflags &= ~IOP_XATTR;
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001362 inode->i_fop = &empty_dir_operations;
1363}
1364
1365bool is_empty_dir_inode(struct inode *inode)
1366{
1367 return (inode->i_fop == &empty_dir_operations) &&
1368 (inode->i_op == &empty_dir_inode_operations);
1369}
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001370
1371#ifdef CONFIG_UNICODE
1372/*
1373 * Determine if the name of a dentry should be casefolded.
1374 *
1375 * Return: if names will need casefolding
1376 */
1377static bool needs_casefold(const struct inode *dir)
1378{
1379 return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding;
1380}
1381
1382/**
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
1388 *
1389 * Return: 0 if names match, 1 if mismatch, or -ERRNO
1390 */
1391int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
1392 const char *str, const struct qstr *name)
1393{
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];
1400 int ret;
1401
1402 if (!dir || !needs_casefold(dir))
1403 goto fallback;
1404 /*
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.
1410 */
1411 if (len <= DNAME_INLINE_LEN - 1) {
1412 memcpy(strbuf, str, len);
1413 strbuf[len] = 0;
1414 qstr.name = strbuf;
1415 /* prevent compiler from optimizing out the temporary buffer */
1416 barrier();
1417 }
1418 ret = utf8_strncasecmp(um, name, &qstr);
1419 if (ret >= 0)
1420 return ret;
1421
1422 if (sb_has_strict_encoding(sb))
1423 return -EINVAL;
1424fallback:
1425 if (len != name->len)
1426 return 1;
1427 return !!memcmp(str, name->name, len);
1428}
1429EXPORT_SYMBOL(generic_ci_d_compare);
1430
1431/**
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
1435 *
1436 * Return: 0 if hash was successful or unchanged, and -EINVAL on error
1437 */
1438int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str)
1439{
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;
1443 int ret = 0;
1444
1445 if (!dir || !needs_casefold(dir))
1446 return 0;
1447
1448 ret = utf8_casefold_hash(um, dentry, str);
1449 if (ret < 0 && sb_has_strict_encoding(sb))
1450 return -EINVAL;
1451 return 0;
1452}
1453EXPORT_SYMBOL(generic_ci_d_hash);
Daniel Rosenberg608af702020-11-19 06:09:02 +00001454
1455static const struct dentry_operations generic_ci_dentry_ops = {
1456 .d_hash = generic_ci_d_hash,
1457 .d_compare = generic_ci_d_compare,
1458};
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001459#endif
Daniel Rosenberg608af702020-11-19 06:09:02 +00001460
1461#ifdef CONFIG_FS_ENCRYPTION
1462static const struct dentry_operations generic_encrypted_dentry_ops = {
1463 .d_revalidate = fscrypt_d_revalidate,
1464};
1465#endif
1466
1467#if defined(CONFIG_FS_ENCRYPTION) && defined(CONFIG_UNICODE)
1468static 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,
1472};
1473#endif
1474
1475/**
1476 * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry
1477 * @dentry: dentry to set ops on
1478 *
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).
1487 *
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.
1491 *
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.
1496 */
1497void generic_set_encrypted_ci_d_ops(struct dentry *dentry)
1498{
1499#ifdef CONFIG_FS_ENCRYPTION
1500 bool needs_encrypt_ops = dentry->d_flags & DCACHE_NOKEY_NAME;
1501#endif
1502#ifdef CONFIG_UNICODE
1503 bool needs_ci_ops = dentry->d_sb->s_encoding;
1504#endif
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);
1508 return;
1509 }
1510#endif
1511#ifdef CONFIG_FS_ENCRYPTION
1512 if (needs_encrypt_ops) {
1513 d_set_d_op(dentry, &generic_encrypted_dentry_ops);
1514 return;
1515 }
1516#endif
1517#ifdef CONFIG_UNICODE
1518 if (needs_ci_ops) {
1519 d_set_d_op(dentry, &generic_ci_dentry_ops);
1520 return;
1521 }
1522#endif
1523}
1524EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops);