1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *	fs/libfs.c
4  *	Library for filesystems writers.
5  */
6 
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/iversion.h>
19 #include <linux/writeback.h>
20 #include <linux/buffer_head.h> /* sync_mapping_buffers */
21 #include <linux/fs_context.h>
22 #include <linux/pseudo_fs.h>
23 #include <linux/fsnotify.h>
24 #include <linux/unicode.h>
25 #include <linux/fscrypt.h>
26 
27 #include <linux/uaccess.h>
28 
29 #include "internal.h"
30 
simple_getattr(struct mnt_idmap * idmap,const struct path * path,struct kstat * stat,u32 request_mask,unsigned int query_flags)31 int simple_getattr(struct mnt_idmap *idmap, const struct path *path,
32 		   struct kstat *stat, u32 request_mask,
33 		   unsigned int query_flags)
34 {
35 	struct inode *inode = d_inode(path->dentry);
36 	generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
37 	stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
38 	return 0;
39 }
40 EXPORT_SYMBOL(simple_getattr);
41 
simple_statfs(struct dentry * dentry,struct kstatfs * buf)42 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
43 {
44 	buf->f_type = dentry->d_sb->s_magic;
45 	buf->f_bsize = PAGE_SIZE;
46 	buf->f_namelen = NAME_MAX;
47 	return 0;
48 }
49 EXPORT_SYMBOL(simple_statfs);
50 
51 /*
52  * Retaining negative dentries for an in-memory filesystem just wastes
53  * memory and lookup time: arrange for them to be deleted immediately.
54  */
always_delete_dentry(const struct dentry * dentry)55 int always_delete_dentry(const struct dentry *dentry)
56 {
57 	return 1;
58 }
59 EXPORT_SYMBOL(always_delete_dentry);
60 
61 const struct dentry_operations simple_dentry_operations = {
62 	.d_delete = always_delete_dentry,
63 };
64 EXPORT_SYMBOL(simple_dentry_operations);
65 
66 /*
67  * Lookup the data. This is trivial - if the dentry didn't already
68  * exist, we know it is negative.  Set d_op to delete negative dentries.
69  */
simple_lookup(struct inode * dir,struct dentry * dentry,unsigned int flags)70 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
71 {
72 	if (dentry->d_name.len > NAME_MAX)
73 		return ERR_PTR(-ENAMETOOLONG);
74 	if (!dentry->d_sb->s_d_op)
75 		d_set_d_op(dentry, &simple_dentry_operations);
76 	d_add(dentry, NULL);
77 	return NULL;
78 }
79 EXPORT_SYMBOL(simple_lookup);
80 
dcache_dir_open(struct inode * inode,struct file * file)81 int dcache_dir_open(struct inode *inode, struct file *file)
82 {
83 	file->private_data = d_alloc_cursor(file->f_path.dentry);
84 
85 	return file->private_data ? 0 : -ENOMEM;
86 }
87 EXPORT_SYMBOL(dcache_dir_open);
88 
dcache_dir_close(struct inode * inode,struct file * file)89 int dcache_dir_close(struct inode *inode, struct file *file)
90 {
91 	dput(file->private_data);
92 	return 0;
93 }
94 EXPORT_SYMBOL(dcache_dir_close);
95 
96 /* parent is locked at least shared */
97 /*
98  * Returns an element of siblings' list.
99  * We are looking for <count>th positive after <p>; if
100  * found, dentry is grabbed and returned to caller.
101  * If no such element exists, NULL is returned.
102  */
scan_positives(struct dentry * cursor,struct list_head * p,loff_t count,struct dentry * last)103 static struct dentry *scan_positives(struct dentry *cursor,
104 					struct list_head *p,
105 					loff_t count,
106 					struct dentry *last)
107 {
108 	struct dentry *dentry = cursor->d_parent, *found = NULL;
109 
110 	spin_lock(&dentry->d_lock);
111 	while ((p = p->next) != &dentry->d_subdirs) {
112 		struct dentry *d = list_entry(p, struct dentry, d_child);
113 		// we must at least skip cursors, to avoid livelocks
114 		if (d->d_flags & DCACHE_DENTRY_CURSOR)
115 			continue;
116 		if (simple_positive(d) && !--count) {
117 			spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
118 			if (simple_positive(d))
119 				found = dget_dlock(d);
120 			spin_unlock(&d->d_lock);
121 			if (likely(found))
122 				break;
123 			count = 1;
124 		}
125 		if (need_resched()) {
126 			list_move(&cursor->d_child, p);
127 			p = &cursor->d_child;
128 			spin_unlock(&dentry->d_lock);
129 			cond_resched();
130 			spin_lock(&dentry->d_lock);
131 		}
132 	}
133 	spin_unlock(&dentry->d_lock);
134 	dput(last);
135 	return found;
136 }
137 
dcache_dir_lseek(struct file * file,loff_t offset,int whence)138 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
139 {
140 	struct dentry *dentry = file->f_path.dentry;
141 	switch (whence) {
142 		case 1:
143 			offset += file->f_pos;
144 			fallthrough;
145 		case 0:
146 			if (offset >= 0)
147 				break;
148 			fallthrough;
149 		default:
150 			return -EINVAL;
151 	}
152 	if (offset != file->f_pos) {
153 		struct dentry *cursor = file->private_data;
154 		struct dentry *to = NULL;
155 
156 		inode_lock_shared(dentry->d_inode);
157 
158 		if (offset > 2)
159 			to = scan_positives(cursor, &dentry->d_subdirs,
160 					    offset - 2, NULL);
161 		spin_lock(&dentry->d_lock);
162 		if (to)
163 			list_move(&cursor->d_child, &to->d_child);
164 		else
165 			list_del_init(&cursor->d_child);
166 		spin_unlock(&dentry->d_lock);
167 		dput(to);
168 
169 		file->f_pos = offset;
170 
171 		inode_unlock_shared(dentry->d_inode);
172 	}
173 	return offset;
174 }
175 EXPORT_SYMBOL(dcache_dir_lseek);
176 
177 /*
178  * Directory is locked and all positive dentries in it are safe, since
179  * for ramfs-type trees they can't go away without unlink() or rmdir(),
180  * both impossible due to the lock on directory.
181  */
182 
dcache_readdir(struct file * file,struct dir_context * ctx)183 int dcache_readdir(struct file *file, struct dir_context *ctx)
184 {
185 	struct dentry *dentry = file->f_path.dentry;
186 	struct dentry *cursor = file->private_data;
187 	struct list_head *anchor = &dentry->d_subdirs;
188 	struct dentry *next = NULL;
189 	struct list_head *p;
190 
191 	if (!dir_emit_dots(file, ctx))
192 		return 0;
193 
194 	if (ctx->pos == 2)
195 		p = anchor;
196 	else if (!list_empty(&cursor->d_child))
197 		p = &cursor->d_child;
198 	else
199 		return 0;
200 
201 	while ((next = scan_positives(cursor, p, 1, next)) != NULL) {
202 		if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
203 			      d_inode(next)->i_ino,
204 			      fs_umode_to_dtype(d_inode(next)->i_mode)))
205 			break;
206 		ctx->pos++;
207 		p = &next->d_child;
208 	}
209 	spin_lock(&dentry->d_lock);
210 	if (next)
211 		list_move_tail(&cursor->d_child, &next->d_child);
212 	else
213 		list_del_init(&cursor->d_child);
214 	spin_unlock(&dentry->d_lock);
215 	dput(next);
216 
217 	return 0;
218 }
219 EXPORT_SYMBOL(dcache_readdir);
220 
generic_read_dir(struct file * filp,char __user * buf,size_t siz,loff_t * ppos)221 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
222 {
223 	return -EISDIR;
224 }
225 EXPORT_SYMBOL(generic_read_dir);
226 
227 const struct file_operations simple_dir_operations = {
228 	.open		= dcache_dir_open,
229 	.release	= dcache_dir_close,
230 	.llseek		= dcache_dir_lseek,
231 	.read		= generic_read_dir,
232 	.iterate_shared	= dcache_readdir,
233 	.fsync		= noop_fsync,
234 };
235 EXPORT_SYMBOL(simple_dir_operations);
236 
237 const struct inode_operations simple_dir_inode_operations = {
238 	.lookup		= simple_lookup,
239 };
240 EXPORT_SYMBOL(simple_dir_inode_operations);
241 
offset_set(struct dentry * dentry,u32 offset)242 static void offset_set(struct dentry *dentry, u32 offset)
243 {
244 	dentry->d_fsdata = (void *)((uintptr_t)(offset));
245 }
246 
dentry2offset(struct dentry * dentry)247 static u32 dentry2offset(struct dentry *dentry)
248 {
249 	return (u32)((uintptr_t)(dentry->d_fsdata));
250 }
251 
252 static struct lock_class_key simple_offset_xa_lock;
253 
254 /**
255  * simple_offset_init - initialize an offset_ctx
256  * @octx: directory offset map to be initialized
257  *
258  */
simple_offset_init(struct offset_ctx * octx)259 void simple_offset_init(struct offset_ctx *octx)
260 {
261 	xa_init_flags(&octx->xa, XA_FLAGS_ALLOC1);
262 	lockdep_set_class(&octx->xa.xa_lock, &simple_offset_xa_lock);
263 
264 	/* 0 is '.', 1 is '..', so always start with offset 2 */
265 	octx->next_offset = 2;
266 }
267 
268 /**
269  * simple_offset_add - Add an entry to a directory's offset map
270  * @octx: directory offset ctx to be updated
271  * @dentry: new dentry being added
272  *
273  * Returns zero on success. @so_ctx and the dentry offset are updated.
274  * Otherwise, a negative errno value is returned.
275  */
simple_offset_add(struct offset_ctx * octx,struct dentry * dentry)276 int simple_offset_add(struct offset_ctx *octx, struct dentry *dentry)
277 {
278 	static const struct xa_limit limit = XA_LIMIT(2, U32_MAX);
279 	u32 offset;
280 	int ret;
281 
282 	if (dentry2offset(dentry) != 0)
283 		return -EBUSY;
284 
285 	ret = xa_alloc_cyclic(&octx->xa, &offset, dentry, limit,
286 			      &octx->next_offset, GFP_KERNEL);
287 	if (ret < 0)
288 		return ret;
289 
290 	offset_set(dentry, offset);
291 	return 0;
292 }
293 
294 /**
295  * simple_offset_remove - Remove an entry to a directory's offset map
296  * @octx: directory offset ctx to be updated
297  * @dentry: dentry being removed
298  *
299  */
simple_offset_remove(struct offset_ctx * octx,struct dentry * dentry)300 void simple_offset_remove(struct offset_ctx *octx, struct dentry *dentry)
301 {
302 	u32 offset;
303 
304 	offset = dentry2offset(dentry);
305 	if (offset == 0)
306 		return;
307 
308 	xa_erase(&octx->xa, offset);
309 	offset_set(dentry, 0);
310 }
311 
312 /**
313  * simple_offset_rename_exchange - exchange rename with directory offsets
314  * @old_dir: parent of dentry being moved
315  * @old_dentry: dentry being moved
316  * @new_dir: destination parent
317  * @new_dentry: destination dentry
318  *
319  * Returns zero on success. Otherwise a negative errno is returned and the
320  * rename is rolled back.
321  */
simple_offset_rename_exchange(struct inode * old_dir,struct dentry * old_dentry,struct inode * new_dir,struct dentry * new_dentry)322 int simple_offset_rename_exchange(struct inode *old_dir,
323 				  struct dentry *old_dentry,
324 				  struct inode *new_dir,
325 				  struct dentry *new_dentry)
326 {
327 	struct offset_ctx *old_ctx = old_dir->i_op->get_offset_ctx(old_dir);
328 	struct offset_ctx *new_ctx = new_dir->i_op->get_offset_ctx(new_dir);
329 	u32 old_index = dentry2offset(old_dentry);
330 	u32 new_index = dentry2offset(new_dentry);
331 	int ret;
332 
333 	simple_offset_remove(old_ctx, old_dentry);
334 	simple_offset_remove(new_ctx, new_dentry);
335 
336 	ret = simple_offset_add(new_ctx, old_dentry);
337 	if (ret)
338 		goto out_restore;
339 
340 	ret = simple_offset_add(old_ctx, new_dentry);
341 	if (ret) {
342 		simple_offset_remove(new_ctx, old_dentry);
343 		goto out_restore;
344 	}
345 
346 	ret = simple_rename_exchange(old_dir, old_dentry, new_dir, new_dentry);
347 	if (ret) {
348 		simple_offset_remove(new_ctx, old_dentry);
349 		simple_offset_remove(old_ctx, new_dentry);
350 		goto out_restore;
351 	}
352 	return 0;
353 
354 out_restore:
355 	offset_set(old_dentry, old_index);
356 	xa_store(&old_ctx->xa, old_index, old_dentry, GFP_KERNEL);
357 	offset_set(new_dentry, new_index);
358 	xa_store(&new_ctx->xa, new_index, new_dentry, GFP_KERNEL);
359 	return ret;
360 }
361 
362 /**
363  * simple_offset_destroy - Release offset map
364  * @octx: directory offset ctx that is about to be destroyed
365  *
366  * During fs teardown (eg. umount), a directory's offset map might still
367  * contain entries. xa_destroy() cleans out anything that remains.
368  */
simple_offset_destroy(struct offset_ctx * octx)369 void simple_offset_destroy(struct offset_ctx *octx)
370 {
371 	xa_destroy(&octx->xa);
372 }
373 
374 /**
375  * offset_dir_llseek - Advance the read position of a directory descriptor
376  * @file: an open directory whose position is to be updated
377  * @offset: a byte offset
378  * @whence: enumerator describing the starting position for this update
379  *
380  * SEEK_END, SEEK_DATA, and SEEK_HOLE are not supported for directories.
381  *
382  * Returns the updated read position if successful; otherwise a
383  * negative errno is returned and the read position remains unchanged.
384  */
offset_dir_llseek(struct file * file,loff_t offset,int whence)385 static loff_t offset_dir_llseek(struct file *file, loff_t offset, int whence)
386 {
387 	switch (whence) {
388 	case SEEK_CUR:
389 		offset += file->f_pos;
390 		fallthrough;
391 	case SEEK_SET:
392 		if (offset >= 0)
393 			break;
394 		fallthrough;
395 	default:
396 		return -EINVAL;
397 	}
398 
399 	/* In this case, ->private_data is protected by f_pos_lock */
400 	file->private_data = NULL;
401 	return vfs_setpos(file, offset, U32_MAX);
402 }
403 
offset_find_next(struct xa_state * xas)404 static struct dentry *offset_find_next(struct xa_state *xas)
405 {
406 	struct dentry *child, *found = NULL;
407 
408 	rcu_read_lock();
409 	child = xas_next_entry(xas, U32_MAX);
410 	if (!child)
411 		goto out;
412 	spin_lock(&child->d_lock);
413 	if (simple_positive(child))
414 		found = dget_dlock(child);
415 	spin_unlock(&child->d_lock);
416 out:
417 	rcu_read_unlock();
418 	return found;
419 }
420 
offset_dir_emit(struct dir_context * ctx,struct dentry * dentry)421 static bool offset_dir_emit(struct dir_context *ctx, struct dentry *dentry)
422 {
423 	u32 offset = dentry2offset(dentry);
424 	struct inode *inode = d_inode(dentry);
425 
426 	return ctx->actor(ctx, dentry->d_name.name, dentry->d_name.len, offset,
427 			  inode->i_ino, fs_umode_to_dtype(inode->i_mode));
428 }
429 
offset_iterate_dir(struct inode * inode,struct dir_context * ctx)430 static void *offset_iterate_dir(struct inode *inode, struct dir_context *ctx)
431 {
432 	struct offset_ctx *so_ctx = inode->i_op->get_offset_ctx(inode);
433 	XA_STATE(xas, &so_ctx->xa, ctx->pos);
434 	struct dentry *dentry;
435 
436 	while (true) {
437 		dentry = offset_find_next(&xas);
438 		if (!dentry)
439 			return ERR_PTR(-ENOENT);
440 
441 		if (!offset_dir_emit(ctx, dentry)) {
442 			dput(dentry);
443 			break;
444 		}
445 
446 		dput(dentry);
447 		ctx->pos = xas.xa_index + 1;
448 	}
449 	return NULL;
450 }
451 
452 /**
453  * offset_readdir - Emit entries starting at offset @ctx->pos
454  * @file: an open directory to iterate over
455  * @ctx: directory iteration context
456  *
457  * Caller must hold @file's i_rwsem to prevent insertion or removal of
458  * entries during this call.
459  *
460  * On entry, @ctx->pos contains an offset that represents the first entry
461  * to be read from the directory.
462  *
463  * The operation continues until there are no more entries to read, or
464  * until the ctx->actor indicates there is no more space in the caller's
465  * output buffer.
466  *
467  * On return, @ctx->pos contains an offset that will read the next entry
468  * in this directory when offset_readdir() is called again with @ctx.
469  *
470  * Return values:
471  *   %0 - Complete
472  */
offset_readdir(struct file * file,struct dir_context * ctx)473 static int offset_readdir(struct file *file, struct dir_context *ctx)
474 {
475 	struct dentry *dir = file->f_path.dentry;
476 
477 	lockdep_assert_held(&d_inode(dir)->i_rwsem);
478 
479 	if (!dir_emit_dots(file, ctx))
480 		return 0;
481 
482 	/* In this case, ->private_data is protected by f_pos_lock */
483 	if (ctx->pos == 2)
484 		file->private_data = NULL;
485 	else if (file->private_data == ERR_PTR(-ENOENT))
486 		return 0;
487 	file->private_data = offset_iterate_dir(d_inode(dir), ctx);
488 	return 0;
489 }
490 
491 const struct file_operations simple_offset_dir_operations = {
492 	.llseek		= offset_dir_llseek,
493 	.iterate_shared	= offset_readdir,
494 	.read		= generic_read_dir,
495 	.fsync		= noop_fsync,
496 };
497 
find_next_child(struct dentry * parent,struct dentry * prev)498 static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev)
499 {
500 	struct dentry *child = NULL;
501 	struct list_head *p = prev ? &prev->d_child : &parent->d_subdirs;
502 
503 	spin_lock(&parent->d_lock);
504 	while ((p = p->next) != &parent->d_subdirs) {
505 		struct dentry *d = container_of(p, struct dentry, d_child);
506 		if (simple_positive(d)) {
507 			spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
508 			if (simple_positive(d))
509 				child = dget_dlock(d);
510 			spin_unlock(&d->d_lock);
511 			if (likely(child))
512 				break;
513 		}
514 	}
515 	spin_unlock(&parent->d_lock);
516 	dput(prev);
517 	return child;
518 }
519 
simple_recursive_removal(struct dentry * dentry,void (* callback)(struct dentry *))520 void simple_recursive_removal(struct dentry *dentry,
521                               void (*callback)(struct dentry *))
522 {
523 	struct dentry *this = dget(dentry);
524 	while (true) {
525 		struct dentry *victim = NULL, *child;
526 		struct inode *inode = this->d_inode;
527 
528 		inode_lock(inode);
529 		if (d_is_dir(this))
530 			inode->i_flags |= S_DEAD;
531 		while ((child = find_next_child(this, victim)) == NULL) {
532 			// kill and ascend
533 			// update metadata while it's still locked
534 			inode_set_ctime_current(inode);
535 			clear_nlink(inode);
536 			inode_unlock(inode);
537 			victim = this;
538 			this = this->d_parent;
539 			inode = this->d_inode;
540 			inode_lock(inode);
541 			if (simple_positive(victim)) {
542 				d_invalidate(victim);	// avoid lost mounts
543 				if (d_is_dir(victim))
544 					fsnotify_rmdir(inode, victim);
545 				else
546 					fsnotify_unlink(inode, victim);
547 				if (callback)
548 					callback(victim);
549 				dput(victim);		// unpin it
550 			}
551 			if (victim == dentry) {
552 				inode_set_mtime_to_ts(inode,
553 						      inode_set_ctime_current(inode));
554 				if (d_is_dir(dentry))
555 					drop_nlink(inode);
556 				inode_unlock(inode);
557 				dput(dentry);
558 				return;
559 			}
560 		}
561 		inode_unlock(inode);
562 		this = child;
563 	}
564 }
565 EXPORT_SYMBOL(simple_recursive_removal);
566 
567 static const struct super_operations simple_super_operations = {
568 	.statfs		= simple_statfs,
569 };
570 
pseudo_fs_fill_super(struct super_block * s,struct fs_context * fc)571 static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc)
572 {
573 	struct pseudo_fs_context *ctx = fc->fs_private;
574 	struct inode *root;
575 
576 	s->s_maxbytes = MAX_LFS_FILESIZE;
577 	s->s_blocksize = PAGE_SIZE;
578 	s->s_blocksize_bits = PAGE_SHIFT;
579 	s->s_magic = ctx->magic;
580 	s->s_op = ctx->ops ?: &simple_super_operations;
581 	s->s_xattr = ctx->xattr;
582 	s->s_time_gran = 1;
583 	root = new_inode(s);
584 	if (!root)
585 		return -ENOMEM;
586 
587 	/*
588 	 * since this is the first inode, make it number 1. New inodes created
589 	 * after this must take care not to collide with it (by passing
590 	 * max_reserved of 1 to iunique).
591 	 */
592 	root->i_ino = 1;
593 	root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
594 	simple_inode_init_ts(root);
595 	s->s_root = d_make_root(root);
596 	if (!s->s_root)
597 		return -ENOMEM;
598 	s->s_d_op = ctx->dops;
599 	return 0;
600 }
601 
pseudo_fs_get_tree(struct fs_context * fc)602 static int pseudo_fs_get_tree(struct fs_context *fc)
603 {
604 	return get_tree_nodev(fc, pseudo_fs_fill_super);
605 }
606 
pseudo_fs_free(struct fs_context * fc)607 static void pseudo_fs_free(struct fs_context *fc)
608 {
609 	kfree(fc->fs_private);
610 }
611 
612 static const struct fs_context_operations pseudo_fs_context_ops = {
613 	.free		= pseudo_fs_free,
614 	.get_tree	= pseudo_fs_get_tree,
615 };
616 
617 /*
618  * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
619  * will never be mountable)
620  */
init_pseudo(struct fs_context * fc,unsigned long magic)621 struct pseudo_fs_context *init_pseudo(struct fs_context *fc,
622 					unsigned long magic)
623 {
624 	struct pseudo_fs_context *ctx;
625 
626 	ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL);
627 	if (likely(ctx)) {
628 		ctx->magic = magic;
629 		fc->fs_private = ctx;
630 		fc->ops = &pseudo_fs_context_ops;
631 		fc->sb_flags |= SB_NOUSER;
632 		fc->global = true;
633 	}
634 	return ctx;
635 }
636 EXPORT_SYMBOL(init_pseudo);
637 
simple_open(struct inode * inode,struct file * file)638 int simple_open(struct inode *inode, struct file *file)
639 {
640 	if (inode->i_private)
641 		file->private_data = inode->i_private;
642 	return 0;
643 }
644 EXPORT_SYMBOL(simple_open);
645 
simple_link(struct dentry * old_dentry,struct inode * dir,struct dentry * dentry)646 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
647 {
648 	struct inode *inode = d_inode(old_dentry);
649 
650 	inode_set_mtime_to_ts(dir,
651 			      inode_set_ctime_to_ts(dir, inode_set_ctime_current(inode)));
652 	inc_nlink(inode);
653 	ihold(inode);
654 	dget(dentry);
655 	d_instantiate(dentry, inode);
656 	return 0;
657 }
658 EXPORT_SYMBOL(simple_link);
659 
simple_empty(struct dentry * dentry)660 int simple_empty(struct dentry *dentry)
661 {
662 	struct dentry *child;
663 	int ret = 0;
664 
665 	spin_lock(&dentry->d_lock);
666 	list_for_each_entry(child, &dentry->d_subdirs, d_child) {
667 		spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
668 		if (simple_positive(child)) {
669 			spin_unlock(&child->d_lock);
670 			goto out;
671 		}
672 		spin_unlock(&child->d_lock);
673 	}
674 	ret = 1;
675 out:
676 	spin_unlock(&dentry->d_lock);
677 	return ret;
678 }
679 EXPORT_SYMBOL(simple_empty);
680 
simple_unlink(struct inode * dir,struct dentry * dentry)681 int simple_unlink(struct inode *dir, struct dentry *dentry)
682 {
683 	struct inode *inode = d_inode(dentry);
684 
685 	inode_set_mtime_to_ts(dir,
686 			      inode_set_ctime_to_ts(dir, inode_set_ctime_current(inode)));
687 	drop_nlink(inode);
688 	dput(dentry);
689 	return 0;
690 }
691 EXPORT_SYMBOL(simple_unlink);
692 
simple_rmdir(struct inode * dir,struct dentry * dentry)693 int simple_rmdir(struct inode *dir, struct dentry *dentry)
694 {
695 	if (!simple_empty(dentry))
696 		return -ENOTEMPTY;
697 
698 	drop_nlink(d_inode(dentry));
699 	simple_unlink(dir, dentry);
700 	drop_nlink(dir);
701 	return 0;
702 }
703 EXPORT_SYMBOL(simple_rmdir);
704 
705 /**
706  * simple_rename_timestamp - update the various inode timestamps for rename
707  * @old_dir: old parent directory
708  * @old_dentry: dentry that is being renamed
709  * @new_dir: new parent directory
710  * @new_dentry: target for rename
711  *
712  * POSIX mandates that the old and new parent directories have their ctime and
713  * mtime updated, and that inodes of @old_dentry and @new_dentry (if any), have
714  * their ctime updated.
715  */
simple_rename_timestamp(struct inode * old_dir,struct dentry * old_dentry,struct inode * new_dir,struct dentry * new_dentry)716 void simple_rename_timestamp(struct inode *old_dir, struct dentry *old_dentry,
717 			     struct inode *new_dir, struct dentry *new_dentry)
718 {
719 	struct inode *newino = d_inode(new_dentry);
720 
721 	inode_set_mtime_to_ts(old_dir, inode_set_ctime_current(old_dir));
722 	if (new_dir != old_dir)
723 		inode_set_mtime_to_ts(new_dir,
724 				      inode_set_ctime_current(new_dir));
725 	inode_set_ctime_current(d_inode(old_dentry));
726 	if (newino)
727 		inode_set_ctime_current(newino);
728 }
729 EXPORT_SYMBOL_GPL(simple_rename_timestamp);
730 
simple_rename_exchange(struct inode * old_dir,struct dentry * old_dentry,struct inode * new_dir,struct dentry * new_dentry)731 int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
732 			   struct inode *new_dir, struct dentry *new_dentry)
733 {
734 	bool old_is_dir = d_is_dir(old_dentry);
735 	bool new_is_dir = d_is_dir(new_dentry);
736 
737 	if (old_dir != new_dir && old_is_dir != new_is_dir) {
738 		if (old_is_dir) {
739 			drop_nlink(old_dir);
740 			inc_nlink(new_dir);
741 		} else {
742 			drop_nlink(new_dir);
743 			inc_nlink(old_dir);
744 		}
745 	}
746 	simple_rename_timestamp(old_dir, old_dentry, new_dir, new_dentry);
747 	return 0;
748 }
749 EXPORT_SYMBOL_GPL(simple_rename_exchange);
750 
simple_rename(struct mnt_idmap * idmap,struct inode * old_dir,struct dentry * old_dentry,struct inode * new_dir,struct dentry * new_dentry,unsigned int flags)751 int simple_rename(struct mnt_idmap *idmap, struct inode *old_dir,
752 		  struct dentry *old_dentry, struct inode *new_dir,
753 		  struct dentry *new_dentry, unsigned int flags)
754 {
755 	int they_are_dirs = d_is_dir(old_dentry);
756 
757 	if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE))
758 		return -EINVAL;
759 
760 	if (flags & RENAME_EXCHANGE)
761 		return simple_rename_exchange(old_dir, old_dentry, new_dir, new_dentry);
762 
763 	if (!simple_empty(new_dentry))
764 		return -ENOTEMPTY;
765 
766 	if (d_really_is_positive(new_dentry)) {
767 		simple_unlink(new_dir, new_dentry);
768 		if (they_are_dirs) {
769 			drop_nlink(d_inode(new_dentry));
770 			drop_nlink(old_dir);
771 		}
772 	} else if (they_are_dirs) {
773 		drop_nlink(old_dir);
774 		inc_nlink(new_dir);
775 	}
776 
777 	simple_rename_timestamp(old_dir, old_dentry, new_dir, new_dentry);
778 	return 0;
779 }
780 EXPORT_SYMBOL(simple_rename);
781 
782 /**
783  * simple_setattr - setattr for simple filesystem
784  * @idmap: idmap of the target mount
785  * @dentry: dentry
786  * @iattr: iattr structure
787  *
788  * Returns 0 on success, -error on failure.
789  *
790  * simple_setattr is a simple ->setattr implementation without a proper
791  * implementation of size changes.
792  *
793  * It can either be used for in-memory filesystems or special files
794  * on simple regular filesystems.  Anything that needs to change on-disk
795  * or wire state on size changes needs its own setattr method.
796  */
simple_setattr(struct mnt_idmap * idmap,struct dentry * dentry,struct iattr * iattr)797 int simple_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
798 		   struct iattr *iattr)
799 {
800 	struct inode *inode = d_inode(dentry);
801 	int error;
802 
803 	error = setattr_prepare(idmap, dentry, iattr);
804 	if (error)
805 		return error;
806 
807 	if (iattr->ia_valid & ATTR_SIZE)
808 		truncate_setsize(inode, iattr->ia_size);
809 	setattr_copy(idmap, inode, iattr);
810 	mark_inode_dirty(inode);
811 	return 0;
812 }
813 EXPORT_SYMBOL(simple_setattr);
814 
simple_read_folio(struct file * file,struct folio * folio)815 static int simple_read_folio(struct file *file, struct folio *folio)
816 {
817 	folio_zero_range(folio, 0, folio_size(folio));
818 	flush_dcache_folio(folio);
819 	folio_mark_uptodate(folio);
820 	folio_unlock(folio);
821 	return 0;
822 }
823 
simple_write_begin(struct file * file,struct address_space * mapping,loff_t pos,unsigned len,struct page ** pagep,void ** fsdata)824 int simple_write_begin(struct file *file, struct address_space *mapping,
825 			loff_t pos, unsigned len,
826 			struct page **pagep, void **fsdata)
827 {
828 	struct folio *folio;
829 
830 	folio = __filemap_get_folio(mapping, pos / PAGE_SIZE, FGP_WRITEBEGIN,
831 			mapping_gfp_mask(mapping));
832 	if (IS_ERR(folio))
833 		return PTR_ERR(folio);
834 
835 	*pagep = &folio->page;
836 
837 	if (!folio_test_uptodate(folio) && (len != folio_size(folio))) {
838 		size_t from = offset_in_folio(folio, pos);
839 
840 		folio_zero_segments(folio, 0, from,
841 				from + len, folio_size(folio));
842 	}
843 	return 0;
844 }
845 EXPORT_SYMBOL(simple_write_begin);
846 
847 /**
848  * simple_write_end - .write_end helper for non-block-device FSes
849  * @file: See .write_end of address_space_operations
850  * @mapping: 		"
851  * @pos: 		"
852  * @len: 		"
853  * @copied: 		"
854  * @page: 		"
855  * @fsdata: 		"
856  *
857  * simple_write_end does the minimum needed for updating a page after writing is
858  * done. It has the same API signature as the .write_end of
859  * address_space_operations vector. So it can just be set onto .write_end for
860  * FSes that don't need any other processing. i_mutex is assumed to be held.
861  * Block based filesystems should use generic_write_end().
862  * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
863  * is not called, so a filesystem that actually does store data in .write_inode
864  * should extend on what's done here with a call to mark_inode_dirty() in the
865  * case that i_size has changed.
866  *
867  * Use *ONLY* with simple_read_folio()
868  */
simple_write_end(struct file * file,struct address_space * mapping,loff_t pos,unsigned len,unsigned copied,struct page * page,void * fsdata)869 static int simple_write_end(struct file *file, struct address_space *mapping,
870 			loff_t pos, unsigned len, unsigned copied,
871 			struct page *page, void *fsdata)
872 {
873 	struct folio *folio = page_folio(page);
874 	struct inode *inode = folio->mapping->host;
875 	loff_t last_pos = pos + copied;
876 
877 	/* zero the stale part of the folio if we did a short copy */
878 	if (!folio_test_uptodate(folio)) {
879 		if (copied < len) {
880 			size_t from = offset_in_folio(folio, pos);
881 
882 			folio_zero_range(folio, from + copied, len - copied);
883 		}
884 		folio_mark_uptodate(folio);
885 	}
886 	/*
887 	 * No need to use i_size_read() here, the i_size
888 	 * cannot change under us because we hold the i_mutex.
889 	 */
890 	if (last_pos > inode->i_size)
891 		i_size_write(inode, last_pos);
892 
893 	folio_mark_dirty(folio);
894 	folio_unlock(folio);
895 	folio_put(folio);
896 
897 	return copied;
898 }
899 
900 /*
901  * Provides ramfs-style behavior: data in the pagecache, but no writeback.
902  */
903 const struct address_space_operations ram_aops = {
904 	.read_folio	= simple_read_folio,
905 	.write_begin	= simple_write_begin,
906 	.write_end	= simple_write_end,
907 	.dirty_folio	= noop_dirty_folio,
908 };
909 EXPORT_SYMBOL(ram_aops);
910 
911 /*
912  * the inodes created here are not hashed. If you use iunique to generate
913  * unique inode values later for this filesystem, then you must take care
914  * to pass it an appropriate max_reserved value to avoid collisions.
915  */
simple_fill_super(struct super_block * s,unsigned long magic,const struct tree_descr * files)916 int simple_fill_super(struct super_block *s, unsigned long magic,
917 		      const struct tree_descr *files)
918 {
919 	struct inode *inode;
920 	struct dentry *root;
921 	struct dentry *dentry;
922 	int i;
923 
924 	s->s_blocksize = PAGE_SIZE;
925 	s->s_blocksize_bits = PAGE_SHIFT;
926 	s->s_magic = magic;
927 	s->s_op = &simple_super_operations;
928 	s->s_time_gran = 1;
929 
930 	inode = new_inode(s);
931 	if (!inode)
932 		return -ENOMEM;
933 	/*
934 	 * because the root inode is 1, the files array must not contain an
935 	 * entry at index 1
936 	 */
937 	inode->i_ino = 1;
938 	inode->i_mode = S_IFDIR | 0755;
939 	simple_inode_init_ts(inode);
940 	inode->i_op = &simple_dir_inode_operations;
941 	inode->i_fop = &simple_dir_operations;
942 	set_nlink(inode, 2);
943 	root = d_make_root(inode);
944 	if (!root)
945 		return -ENOMEM;
946 	for (i = 0; !files->name || files->name[0]; i++, files++) {
947 		if (!files->name)
948 			continue;
949 
950 		/* warn if it tries to conflict with the root inode */
951 		if (unlikely(i == 1))
952 			printk(KERN_WARNING "%s: %s passed in a files array"
953 				"with an index of 1!\n", __func__,
954 				s->s_type->name);
955 
956 		dentry = d_alloc_name(root, files->name);
957 		if (!dentry)
958 			goto out;
959 		inode = new_inode(s);
960 		if (!inode) {
961 			dput(dentry);
962 			goto out;
963 		}
964 		inode->i_mode = S_IFREG | files->mode;
965 		simple_inode_init_ts(inode);
966 		inode->i_fop = files->ops;
967 		inode->i_ino = i;
968 		d_add(dentry, inode);
969 	}
970 	s->s_root = root;
971 	return 0;
972 out:
973 	d_genocide(root);
974 	shrink_dcache_parent(root);
975 	dput(root);
976 	return -ENOMEM;
977 }
978 EXPORT_SYMBOL(simple_fill_super);
979 
980 static DEFINE_SPINLOCK(pin_fs_lock);
981 
simple_pin_fs(struct file_system_type * type,struct vfsmount ** mount,int * count)982 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
983 {
984 	struct vfsmount *mnt = NULL;
985 	spin_lock(&pin_fs_lock);
986 	if (unlikely(!*mount)) {
987 		spin_unlock(&pin_fs_lock);
988 		mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
989 		if (IS_ERR(mnt))
990 			return PTR_ERR(mnt);
991 		spin_lock(&pin_fs_lock);
992 		if (!*mount)
993 			*mount = mnt;
994 	}
995 	mntget(*mount);
996 	++*count;
997 	spin_unlock(&pin_fs_lock);
998 	mntput(mnt);
999 	return 0;
1000 }
1001 EXPORT_SYMBOL(simple_pin_fs);
1002 
simple_release_fs(struct vfsmount ** mount,int * count)1003 void simple_release_fs(struct vfsmount **mount, int *count)
1004 {
1005 	struct vfsmount *mnt;
1006 	spin_lock(&pin_fs_lock);
1007 	mnt = *mount;
1008 	if (!--*count)
1009 		*mount = NULL;
1010 	spin_unlock(&pin_fs_lock);
1011 	mntput(mnt);
1012 }
1013 EXPORT_SYMBOL(simple_release_fs);
1014 
1015 /**
1016  * simple_read_from_buffer - copy data from the buffer to user space
1017  * @to: the user space buffer to read to
1018  * @count: the maximum number of bytes to read
1019  * @ppos: the current position in the buffer
1020  * @from: the buffer to read from
1021  * @available: the size of the buffer
1022  *
1023  * The simple_read_from_buffer() function reads up to @count bytes from the
1024  * buffer @from at offset @ppos into the user space address starting at @to.
1025  *
1026  * On success, the number of bytes read is returned and the offset @ppos is
1027  * advanced by this number, or negative value is returned on error.
1028  **/
simple_read_from_buffer(void __user * to,size_t count,loff_t * ppos,const void * from,size_t available)1029 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
1030 				const void *from, size_t available)
1031 {
1032 	loff_t pos = *ppos;
1033 	size_t ret;
1034 
1035 	if (pos < 0)
1036 		return -EINVAL;
1037 	if (pos >= available || !count)
1038 		return 0;
1039 	if (count > available - pos)
1040 		count = available - pos;
1041 	ret = copy_to_user(to, from + pos, count);
1042 	if (ret == count)
1043 		return -EFAULT;
1044 	count -= ret;
1045 	*ppos = pos + count;
1046 	return count;
1047 }
1048 EXPORT_SYMBOL(simple_read_from_buffer);
1049 
1050 /**
1051  * simple_write_to_buffer - copy data from user space to the buffer
1052  * @to: the buffer to write to
1053  * @available: the size of the buffer
1054  * @ppos: the current position in the buffer
1055  * @from: the user space buffer to read from
1056  * @count: the maximum number of bytes to read
1057  *
1058  * The simple_write_to_buffer() function reads up to @count bytes from the user
1059  * space address starting at @from into the buffer @to at offset @ppos.
1060  *
1061  * On success, the number of bytes written is returned and the offset @ppos is
1062  * advanced by this number, or negative value is returned on error.
1063  **/
simple_write_to_buffer(void * to,size_t available,loff_t * ppos,const void __user * from,size_t count)1064 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
1065 		const void __user *from, size_t count)
1066 {
1067 	loff_t pos = *ppos;
1068 	size_t res;
1069 
1070 	if (pos < 0)
1071 		return -EINVAL;
1072 	if (pos >= available || !count)
1073 		return 0;
1074 	if (count > available - pos)
1075 		count = available - pos;
1076 	res = copy_from_user(to + pos, from, count);
1077 	if (res == count)
1078 		return -EFAULT;
1079 	count -= res;
1080 	*ppos = pos + count;
1081 	return count;
1082 }
1083 EXPORT_SYMBOL(simple_write_to_buffer);
1084 
1085 /**
1086  * memory_read_from_buffer - copy data from the buffer
1087  * @to: the kernel space buffer to read to
1088  * @count: the maximum number of bytes to read
1089  * @ppos: the current position in the buffer
1090  * @from: the buffer to read from
1091  * @available: the size of the buffer
1092  *
1093  * The memory_read_from_buffer() function reads up to @count bytes from the
1094  * buffer @from at offset @ppos into the kernel space address starting at @to.
1095  *
1096  * On success, the number of bytes read is returned and the offset @ppos is
1097  * advanced by this number, or negative value is returned on error.
1098  **/
memory_read_from_buffer(void * to,size_t count,loff_t * ppos,const void * from,size_t available)1099 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
1100 				const void *from, size_t available)
1101 {
1102 	loff_t pos = *ppos;
1103 
1104 	if (pos < 0)
1105 		return -EINVAL;
1106 	if (pos >= available)
1107 		return 0;
1108 	if (count > available - pos)
1109 		count = available - pos;
1110 	memcpy(to, from + pos, count);
1111 	*ppos = pos + count;
1112 
1113 	return count;
1114 }
1115 EXPORT_SYMBOL(memory_read_from_buffer);
1116 
1117 /*
1118  * Transaction based IO.
1119  * The file expects a single write which triggers the transaction, and then
1120  * possibly a read which collects the result - which is stored in a
1121  * file-local buffer.
1122  */
1123 
simple_transaction_set(struct file * file,size_t n)1124 void simple_transaction_set(struct file *file, size_t n)
1125 {
1126 	struct simple_transaction_argresp *ar = file->private_data;
1127 
1128 	BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
1129 
1130 	/*
1131 	 * The barrier ensures that ar->size will really remain zero until
1132 	 * ar->data is ready for reading.
1133 	 */
1134 	smp_mb();
1135 	ar->size = n;
1136 }
1137 EXPORT_SYMBOL(simple_transaction_set);
1138 
simple_transaction_get(struct file * file,const char __user * buf,size_t size)1139 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
1140 {
1141 	struct simple_transaction_argresp *ar;
1142 	static DEFINE_SPINLOCK(simple_transaction_lock);
1143 
1144 	if (size > SIMPLE_TRANSACTION_LIMIT - 1)
1145 		return ERR_PTR(-EFBIG);
1146 
1147 	ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
1148 	if (!ar)
1149 		return ERR_PTR(-ENOMEM);
1150 
1151 	spin_lock(&simple_transaction_lock);
1152 
1153 	/* only one write allowed per open */
1154 	if (file->private_data) {
1155 		spin_unlock(&simple_transaction_lock);
1156 		free_page((unsigned long)ar);
1157 		return ERR_PTR(-EBUSY);
1158 	}
1159 
1160 	file->private_data = ar;
1161 
1162 	spin_unlock(&simple_transaction_lock);
1163 
1164 	if (copy_from_user(ar->data, buf, size))
1165 		return ERR_PTR(-EFAULT);
1166 
1167 	return ar->data;
1168 }
1169 EXPORT_SYMBOL(simple_transaction_get);
1170 
simple_transaction_read(struct file * file,char __user * buf,size_t size,loff_t * pos)1171 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
1172 {
1173 	struct simple_transaction_argresp *ar = file->private_data;
1174 
1175 	if (!ar)
1176 		return 0;
1177 	return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
1178 }
1179 EXPORT_SYMBOL(simple_transaction_read);
1180 
simple_transaction_release(struct inode * inode,struct file * file)1181 int simple_transaction_release(struct inode *inode, struct file *file)
1182 {
1183 	free_page((unsigned long)file->private_data);
1184 	return 0;
1185 }
1186 EXPORT_SYMBOL(simple_transaction_release);
1187 
1188 /* Simple attribute files */
1189 
1190 struct simple_attr {
1191 	int (*get)(void *, u64 *);
1192 	int (*set)(void *, u64);
1193 	char get_buf[24];	/* enough to store a u64 and "\n\0" */
1194 	char set_buf[24];
1195 	void *data;
1196 	const char *fmt;	/* format for read operation */
1197 	struct mutex mutex;	/* protects access to these buffers */
1198 };
1199 
1200 /* simple_attr_open is called by an actual attribute open file operation
1201  * to set the attribute specific access operations. */
simple_attr_open(struct inode * inode,struct file * file,int (* get)(void *,u64 *),int (* set)(void *,u64),const char * fmt)1202 int simple_attr_open(struct inode *inode, struct file *file,
1203 		     int (*get)(void *, u64 *), int (*set)(void *, u64),
1204 		     const char *fmt)
1205 {
1206 	struct simple_attr *attr;
1207 
1208 	attr = kzalloc(sizeof(*attr), GFP_KERNEL);
1209 	if (!attr)
1210 		return -ENOMEM;
1211 
1212 	attr->get = get;
1213 	attr->set = set;
1214 	attr->data = inode->i_private;
1215 	attr->fmt = fmt;
1216 	mutex_init(&attr->mutex);
1217 
1218 	file->private_data = attr;
1219 
1220 	return nonseekable_open(inode, file);
1221 }
1222 EXPORT_SYMBOL_GPL(simple_attr_open);
1223 
simple_attr_release(struct inode * inode,struct file * file)1224 int simple_attr_release(struct inode *inode, struct file *file)
1225 {
1226 	kfree(file->private_data);
1227 	return 0;
1228 }
1229 EXPORT_SYMBOL_GPL(simple_attr_release);	/* GPL-only?  This?  Really? */
1230 
1231 /* read from the buffer that is filled with the get function */
simple_attr_read(struct file * file,char __user * buf,size_t len,loff_t * ppos)1232 ssize_t simple_attr_read(struct file *file, char __user *buf,
1233 			 size_t len, loff_t *ppos)
1234 {
1235 	struct simple_attr *attr;
1236 	size_t size;
1237 	ssize_t ret;
1238 
1239 	attr = file->private_data;
1240 
1241 	if (!attr->get)
1242 		return -EACCES;
1243 
1244 	ret = mutex_lock_interruptible(&attr->mutex);
1245 	if (ret)
1246 		return ret;
1247 
1248 	if (*ppos && attr->get_buf[0]) {
1249 		/* continued read */
1250 		size = strlen(attr->get_buf);
1251 	} else {
1252 		/* first read */
1253 		u64 val;
1254 		ret = attr->get(attr->data, &val);
1255 		if (ret)
1256 			goto out;
1257 
1258 		size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
1259 				 attr->fmt, (unsigned long long)val);
1260 	}
1261 
1262 	ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
1263 out:
1264 	mutex_unlock(&attr->mutex);
1265 	return ret;
1266 }
1267 EXPORT_SYMBOL_GPL(simple_attr_read);
1268 
1269 /* interpret the buffer as a number to call the set function with */
simple_attr_write_xsigned(struct file * file,const char __user * buf,size_t len,loff_t * ppos,bool is_signed)1270 static ssize_t simple_attr_write_xsigned(struct file *file, const char __user *buf,
1271 			  size_t len, loff_t *ppos, bool is_signed)
1272 {
1273 	struct simple_attr *attr;
1274 	unsigned long long val;
1275 	size_t size;
1276 	ssize_t ret;
1277 
1278 	attr = file->private_data;
1279 	if (!attr->set)
1280 		return -EACCES;
1281 
1282 	ret = mutex_lock_interruptible(&attr->mutex);
1283 	if (ret)
1284 		return ret;
1285 
1286 	ret = -EFAULT;
1287 	size = min(sizeof(attr->set_buf) - 1, len);
1288 	if (copy_from_user(attr->set_buf, buf, size))
1289 		goto out;
1290 
1291 	attr->set_buf[size] = '\0';
1292 	if (is_signed)
1293 		ret = kstrtoll(attr->set_buf, 0, &val);
1294 	else
1295 		ret = kstrtoull(attr->set_buf, 0, &val);
1296 	if (ret)
1297 		goto out;
1298 	ret = attr->set(attr->data, val);
1299 	if (ret == 0)
1300 		ret = len; /* on success, claim we got the whole input */
1301 out:
1302 	mutex_unlock(&attr->mutex);
1303 	return ret;
1304 }
1305 
simple_attr_write(struct file * file,const char __user * buf,size_t len,loff_t * ppos)1306 ssize_t simple_attr_write(struct file *file, const char __user *buf,
1307 			  size_t len, loff_t *ppos)
1308 {
1309 	return simple_attr_write_xsigned(file, buf, len, ppos, false);
1310 }
1311 EXPORT_SYMBOL_GPL(simple_attr_write);
1312 
simple_attr_write_signed(struct file * file,const char __user * buf,size_t len,loff_t * ppos)1313 ssize_t simple_attr_write_signed(struct file *file, const char __user *buf,
1314 			  size_t len, loff_t *ppos)
1315 {
1316 	return simple_attr_write_xsigned(file, buf, len, ppos, true);
1317 }
1318 EXPORT_SYMBOL_GPL(simple_attr_write_signed);
1319 
1320 /**
1321  * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
1322  * @sb:		filesystem to do the file handle conversion on
1323  * @fid:	file handle to convert
1324  * @fh_len:	length of the file handle in bytes
1325  * @fh_type:	type of file handle
1326  * @get_inode:	filesystem callback to retrieve inode
1327  *
1328  * This function decodes @fid as long as it has one of the well-known
1329  * Linux filehandle types and calls @get_inode on it to retrieve the
1330  * inode for the object specified in the file handle.
1331  */
generic_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type,struct inode * (* get_inode)(struct super_block * sb,u64 ino,u32 gen))1332 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
1333 		int fh_len, int fh_type, struct inode *(*get_inode)
1334 			(struct super_block *sb, u64 ino, u32 gen))
1335 {
1336 	struct inode *inode = NULL;
1337 
1338 	if (fh_len < 2)
1339 		return NULL;
1340 
1341 	switch (fh_type) {
1342 	case FILEID_INO32_GEN:
1343 	case FILEID_INO32_GEN_PARENT:
1344 		inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
1345 		break;
1346 	}
1347 
1348 	return d_obtain_alias(inode);
1349 }
1350 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
1351 
1352 /**
1353  * generic_fh_to_parent - generic helper for the fh_to_parent export operation
1354  * @sb:		filesystem to do the file handle conversion on
1355  * @fid:	file handle to convert
1356  * @fh_len:	length of the file handle in bytes
1357  * @fh_type:	type of file handle
1358  * @get_inode:	filesystem callback to retrieve inode
1359  *
1360  * This function decodes @fid as long as it has one of the well-known
1361  * Linux filehandle types and calls @get_inode on it to retrieve the
1362  * inode for the _parent_ object specified in the file handle if it
1363  * is specified in the file handle, or NULL otherwise.
1364  */
generic_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type,struct inode * (* get_inode)(struct super_block * sb,u64 ino,u32 gen))1365 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
1366 		int fh_len, int fh_type, struct inode *(*get_inode)
1367 			(struct super_block *sb, u64 ino, u32 gen))
1368 {
1369 	struct inode *inode = NULL;
1370 
1371 	if (fh_len <= 2)
1372 		return NULL;
1373 
1374 	switch (fh_type) {
1375 	case FILEID_INO32_GEN_PARENT:
1376 		inode = get_inode(sb, fid->i32.parent_ino,
1377 				  (fh_len > 3 ? fid->i32.parent_gen : 0));
1378 		break;
1379 	}
1380 
1381 	return d_obtain_alias(inode);
1382 }
1383 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
1384 
1385 /**
1386  * __generic_file_fsync - generic fsync implementation for simple filesystems
1387  *
1388  * @file:	file to synchronize
1389  * @start:	start offset in bytes
1390  * @end:	end offset in bytes (inclusive)
1391  * @datasync:	only synchronize essential metadata if true
1392  *
1393  * This is a generic implementation of the fsync method for simple
1394  * filesystems which track all non-inode metadata in the buffers list
1395  * hanging off the address_space structure.
1396  */
__generic_file_fsync(struct file * file,loff_t start,loff_t end,int datasync)1397 int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
1398 				 int datasync)
1399 {
1400 	struct inode *inode = file->f_mapping->host;
1401 	int err;
1402 	int ret;
1403 
1404 	err = file_write_and_wait_range(file, start, end);
1405 	if (err)
1406 		return err;
1407 
1408 	inode_lock(inode);
1409 	ret = sync_mapping_buffers(inode->i_mapping);
1410 	if (!(inode->i_state & I_DIRTY_ALL))
1411 		goto out;
1412 	if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
1413 		goto out;
1414 
1415 	err = sync_inode_metadata(inode, 1);
1416 	if (ret == 0)
1417 		ret = err;
1418 
1419 out:
1420 	inode_unlock(inode);
1421 	/* check and advance again to catch errors after syncing out buffers */
1422 	err = file_check_and_advance_wb_err(file);
1423 	if (ret == 0)
1424 		ret = err;
1425 	return ret;
1426 }
1427 EXPORT_SYMBOL(__generic_file_fsync);
1428 
1429 /**
1430  * generic_file_fsync - generic fsync implementation for simple filesystems
1431  *			with flush
1432  * @file:	file to synchronize
1433  * @start:	start offset in bytes
1434  * @end:	end offset in bytes (inclusive)
1435  * @datasync:	only synchronize essential metadata if true
1436  *
1437  */
1438 
generic_file_fsync(struct file * file,loff_t start,loff_t end,int datasync)1439 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1440 		       int datasync)
1441 {
1442 	struct inode *inode = file->f_mapping->host;
1443 	int err;
1444 
1445 	err = __generic_file_fsync(file, start, end, datasync);
1446 	if (err)
1447 		return err;
1448 	return blkdev_issue_flush(inode->i_sb->s_bdev);
1449 }
1450 EXPORT_SYMBOL(generic_file_fsync);
1451 
1452 /**
1453  * generic_check_addressable - Check addressability of file system
1454  * @blocksize_bits:	log of file system block size
1455  * @num_blocks:		number of blocks in file system
1456  *
1457  * Determine whether a file system with @num_blocks blocks (and a
1458  * block size of 2**@blocksize_bits) is addressable by the sector_t
1459  * and page cache of the system.  Return 0 if so and -EFBIG otherwise.
1460  */
generic_check_addressable(unsigned blocksize_bits,u64 num_blocks)1461 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1462 {
1463 	u64 last_fs_block = num_blocks - 1;
1464 	u64 last_fs_page =
1465 		last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1466 
1467 	if (unlikely(num_blocks == 0))
1468 		return 0;
1469 
1470 	if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1471 		return -EINVAL;
1472 
1473 	if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1474 	    (last_fs_page > (pgoff_t)(~0ULL))) {
1475 		return -EFBIG;
1476 	}
1477 	return 0;
1478 }
1479 EXPORT_SYMBOL(generic_check_addressable);
1480 
1481 /*
1482  * No-op implementation of ->fsync for in-memory filesystems.
1483  */
noop_fsync(struct file * file,loff_t start,loff_t end,int datasync)1484 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1485 {
1486 	return 0;
1487 }
1488 EXPORT_SYMBOL(noop_fsync);
1489 
noop_direct_IO(struct kiocb * iocb,struct iov_iter * iter)1490 ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1491 {
1492 	/*
1493 	 * iomap based filesystems support direct I/O without need for
1494 	 * this callback. However, it still needs to be set in
1495 	 * inode->a_ops so that open/fcntl know that direct I/O is
1496 	 * generally supported.
1497 	 */
1498 	return -EINVAL;
1499 }
1500 EXPORT_SYMBOL_GPL(noop_direct_IO);
1501 
1502 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
kfree_link(void * p)1503 void kfree_link(void *p)
1504 {
1505 	kfree(p);
1506 }
1507 EXPORT_SYMBOL(kfree_link);
1508 
alloc_anon_inode(struct super_block * s)1509 struct inode *alloc_anon_inode(struct super_block *s)
1510 {
1511 	static const struct address_space_operations anon_aops = {
1512 		.dirty_folio	= noop_dirty_folio,
1513 	};
1514 	struct inode *inode = new_inode_pseudo(s);
1515 
1516 	if (!inode)
1517 		return ERR_PTR(-ENOMEM);
1518 
1519 	inode->i_ino = get_next_ino();
1520 	inode->i_mapping->a_ops = &anon_aops;
1521 
1522 	/*
1523 	 * Mark the inode dirty from the very beginning,
1524 	 * that way it will never be moved to the dirty
1525 	 * list because mark_inode_dirty() will think
1526 	 * that it already _is_ on the dirty list.
1527 	 */
1528 	inode->i_state = I_DIRTY;
1529 	inode->i_mode = S_IRUSR | S_IWUSR;
1530 	inode->i_uid = current_fsuid();
1531 	inode->i_gid = current_fsgid();
1532 	inode->i_flags |= S_PRIVATE;
1533 	simple_inode_init_ts(inode);
1534 	return inode;
1535 }
1536 EXPORT_SYMBOL(alloc_anon_inode);
1537 
1538 /**
1539  * simple_nosetlease - generic helper for prohibiting leases
1540  * @filp: file pointer
1541  * @arg: type of lease to obtain
1542  * @flp: new lease supplied for insertion
1543  * @priv: private data for lm_setup operation
1544  *
1545  * Generic helper for filesystems that do not wish to allow leases to be set.
1546  * All arguments are ignored and it just returns -EINVAL.
1547  */
1548 int
simple_nosetlease(struct file * filp,int arg,struct file_lock ** flp,void ** priv)1549 simple_nosetlease(struct file *filp, int arg, struct file_lock **flp,
1550 		  void **priv)
1551 {
1552 	return -EINVAL;
1553 }
1554 EXPORT_SYMBOL(simple_nosetlease);
1555 
1556 /**
1557  * simple_get_link - generic helper to get the target of "fast" symlinks
1558  * @dentry: not used here
1559  * @inode: the symlink inode
1560  * @done: not used here
1561  *
1562  * Generic helper for filesystems to use for symlink inodes where a pointer to
1563  * the symlink target is stored in ->i_link.  NOTE: this isn't normally called,
1564  * since as an optimization the path lookup code uses any non-NULL ->i_link
1565  * directly, without calling ->get_link().  But ->get_link() still must be set,
1566  * to mark the inode_operations as being for a symlink.
1567  *
1568  * Return: the symlink target
1569  */
simple_get_link(struct dentry * dentry,struct inode * inode,struct delayed_call * done)1570 const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1571 			    struct delayed_call *done)
1572 {
1573 	return inode->i_link;
1574 }
1575 EXPORT_SYMBOL(simple_get_link);
1576 
1577 const struct inode_operations simple_symlink_inode_operations = {
1578 	.get_link = simple_get_link,
1579 };
1580 EXPORT_SYMBOL(simple_symlink_inode_operations);
1581 
1582 /*
1583  * Operations for a permanently empty directory.
1584  */
empty_dir_lookup(struct inode * dir,struct dentry * dentry,unsigned int flags)1585 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1586 {
1587 	return ERR_PTR(-ENOENT);
1588 }
1589 
empty_dir_getattr(struct mnt_idmap * idmap,const struct path * path,struct kstat * stat,u32 request_mask,unsigned int query_flags)1590 static int empty_dir_getattr(struct mnt_idmap *idmap,
1591 			     const struct path *path, struct kstat *stat,
1592 			     u32 request_mask, unsigned int query_flags)
1593 {
1594 	struct inode *inode = d_inode(path->dentry);
1595 	generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
1596 	return 0;
1597 }
1598 
empty_dir_setattr(struct mnt_idmap * idmap,struct dentry * dentry,struct iattr * attr)1599 static int empty_dir_setattr(struct mnt_idmap *idmap,
1600 			     struct dentry *dentry, struct iattr *attr)
1601 {
1602 	return -EPERM;
1603 }
1604 
empty_dir_listxattr(struct dentry * dentry,char * list,size_t size)1605 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1606 {
1607 	return -EOPNOTSUPP;
1608 }
1609 
1610 static const struct inode_operations empty_dir_inode_operations = {
1611 	.lookup		= empty_dir_lookup,
1612 	.permission	= generic_permission,
1613 	.setattr	= empty_dir_setattr,
1614 	.getattr	= empty_dir_getattr,
1615 	.listxattr	= empty_dir_listxattr,
1616 };
1617 
empty_dir_llseek(struct file * file,loff_t offset,int whence)1618 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1619 {
1620 	/* An empty directory has two entries . and .. at offsets 0 and 1 */
1621 	return generic_file_llseek_size(file, offset, whence, 2, 2);
1622 }
1623 
empty_dir_readdir(struct file * file,struct dir_context * ctx)1624 static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1625 {
1626 	dir_emit_dots(file, ctx);
1627 	return 0;
1628 }
1629 
1630 static const struct file_operations empty_dir_operations = {
1631 	.llseek		= empty_dir_llseek,
1632 	.read		= generic_read_dir,
1633 	.iterate_shared	= empty_dir_readdir,
1634 	.fsync		= noop_fsync,
1635 };
1636 
1637 
make_empty_dir_inode(struct inode * inode)1638 void make_empty_dir_inode(struct inode *inode)
1639 {
1640 	set_nlink(inode, 2);
1641 	inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1642 	inode->i_uid = GLOBAL_ROOT_UID;
1643 	inode->i_gid = GLOBAL_ROOT_GID;
1644 	inode->i_rdev = 0;
1645 	inode->i_size = 0;
1646 	inode->i_blkbits = PAGE_SHIFT;
1647 	inode->i_blocks = 0;
1648 
1649 	inode->i_op = &empty_dir_inode_operations;
1650 	inode->i_opflags &= ~IOP_XATTR;
1651 	inode->i_fop = &empty_dir_operations;
1652 }
1653 
is_empty_dir_inode(struct inode * inode)1654 bool is_empty_dir_inode(struct inode *inode)
1655 {
1656 	return (inode->i_fop == &empty_dir_operations) &&
1657 		(inode->i_op == &empty_dir_inode_operations);
1658 }
1659 
1660 #if IS_ENABLED(CONFIG_UNICODE)
1661 /**
1662  * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
1663  * @dentry:	dentry whose name we are checking against
1664  * @len:	len of name of dentry
1665  * @str:	str pointer to name of dentry
1666  * @name:	Name to compare against
1667  *
1668  * Return: 0 if names match, 1 if mismatch, or -ERRNO
1669  */
generic_ci_d_compare(const struct dentry * dentry,unsigned int len,const char * str,const struct qstr * name)1670 static int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
1671 				const char *str, const struct qstr *name)
1672 {
1673 	const struct dentry *parent = READ_ONCE(dentry->d_parent);
1674 	const struct inode *dir = READ_ONCE(parent->d_inode);
1675 	const struct super_block *sb = dentry->d_sb;
1676 	const struct unicode_map *um = sb->s_encoding;
1677 	struct qstr qstr = QSTR_INIT(str, len);
1678 	char strbuf[DNAME_INLINE_LEN];
1679 	int ret;
1680 
1681 	if (!dir || !IS_CASEFOLDED(dir))
1682 		goto fallback;
1683 	/*
1684 	 * If the dentry name is stored in-line, then it may be concurrently
1685 	 * modified by a rename.  If this happens, the VFS will eventually retry
1686 	 * the lookup, so it doesn't matter what ->d_compare() returns.
1687 	 * However, it's unsafe to call utf8_strncasecmp() with an unstable
1688 	 * string.  Therefore, we have to copy the name into a temporary buffer.
1689 	 */
1690 	if (len <= DNAME_INLINE_LEN - 1) {
1691 		memcpy(strbuf, str, len);
1692 		strbuf[len] = 0;
1693 		qstr.name = strbuf;
1694 		/* prevent compiler from optimizing out the temporary buffer */
1695 		barrier();
1696 	}
1697 	ret = utf8_strncasecmp(um, name, &qstr);
1698 	if (ret >= 0)
1699 		return ret;
1700 
1701 	if (sb_has_strict_encoding(sb))
1702 		return -EINVAL;
1703 fallback:
1704 	if (len != name->len)
1705 		return 1;
1706 	return !!memcmp(str, name->name, len);
1707 }
1708 
1709 /**
1710  * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
1711  * @dentry:	dentry of the parent directory
1712  * @str:	qstr of name whose hash we should fill in
1713  *
1714  * Return: 0 if hash was successful or unchanged, and -EINVAL on error
1715  */
generic_ci_d_hash(const struct dentry * dentry,struct qstr * str)1716 static int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str)
1717 {
1718 	const struct inode *dir = READ_ONCE(dentry->d_inode);
1719 	struct super_block *sb = dentry->d_sb;
1720 	const struct unicode_map *um = sb->s_encoding;
1721 	int ret = 0;
1722 
1723 	if (!dir || !IS_CASEFOLDED(dir))
1724 		return 0;
1725 
1726 	ret = utf8_casefold_hash(um, dentry, str);
1727 	if (ret < 0 && sb_has_strict_encoding(sb))
1728 		return -EINVAL;
1729 	return 0;
1730 }
1731 
1732 static const struct dentry_operations generic_ci_dentry_ops = {
1733 	.d_hash = generic_ci_d_hash,
1734 	.d_compare = generic_ci_d_compare,
1735 };
1736 #endif
1737 
1738 #ifdef CONFIG_FS_ENCRYPTION
1739 static const struct dentry_operations generic_encrypted_dentry_ops = {
1740 	.d_revalidate = fscrypt_d_revalidate,
1741 };
1742 #endif
1743 
1744 #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
1745 static const struct dentry_operations generic_encrypted_ci_dentry_ops = {
1746 	.d_hash = generic_ci_d_hash,
1747 	.d_compare = generic_ci_d_compare,
1748 	.d_revalidate = fscrypt_d_revalidate,
1749 };
1750 #endif
1751 
1752 /**
1753  * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry
1754  * @dentry:	dentry to set ops on
1755  *
1756  * Casefolded directories need d_hash and d_compare set, so that the dentries
1757  * contained in them are handled case-insensitively.  Note that these operations
1758  * are needed on the parent directory rather than on the dentries in it, and
1759  * while the casefolding flag can be toggled on and off on an empty directory,
1760  * dentry_operations can't be changed later.  As a result, if the filesystem has
1761  * casefolding support enabled at all, we have to give all dentries the
1762  * casefolding operations even if their inode doesn't have the casefolding flag
1763  * currently (and thus the casefolding ops would be no-ops for now).
1764  *
1765  * Encryption works differently in that the only dentry operation it needs is
1766  * d_revalidate, which it only needs on dentries that have the no-key name flag.
1767  * The no-key flag can't be set "later", so we don't have to worry about that.
1768  *
1769  * Finally, to maximize compatibility with overlayfs (which isn't compatible
1770  * with certain dentry operations) and to avoid taking an unnecessary
1771  * performance hit, we use custom dentry_operations for each possible
1772  * combination rather than always installing all operations.
1773  */
generic_set_encrypted_ci_d_ops(struct dentry * dentry)1774 void generic_set_encrypted_ci_d_ops(struct dentry *dentry)
1775 {
1776 #ifdef CONFIG_FS_ENCRYPTION
1777 	bool needs_encrypt_ops = dentry->d_flags & DCACHE_NOKEY_NAME;
1778 #endif
1779 #if IS_ENABLED(CONFIG_UNICODE)
1780 	bool needs_ci_ops = dentry->d_sb->s_encoding;
1781 #endif
1782 #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
1783 	if (needs_encrypt_ops && needs_ci_ops) {
1784 		d_set_d_op(dentry, &generic_encrypted_ci_dentry_ops);
1785 		return;
1786 	}
1787 #endif
1788 #ifdef CONFIG_FS_ENCRYPTION
1789 	if (needs_encrypt_ops) {
1790 		d_set_d_op(dentry, &generic_encrypted_dentry_ops);
1791 		return;
1792 	}
1793 #endif
1794 #if IS_ENABLED(CONFIG_UNICODE)
1795 	if (needs_ci_ops) {
1796 		d_set_d_op(dentry, &generic_ci_dentry_ops);
1797 		return;
1798 	}
1799 #endif
1800 }
1801 EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops);
1802 
1803 /**
1804  * inode_maybe_inc_iversion - increments i_version
1805  * @inode: inode with the i_version that should be updated
1806  * @force: increment the counter even if it's not necessary?
1807  *
1808  * Every time the inode is modified, the i_version field must be seen to have
1809  * changed by any observer.
1810  *
1811  * If "force" is set or the QUERIED flag is set, then ensure that we increment
1812  * the value, and clear the queried flag.
1813  *
1814  * In the common case where neither is set, then we can return "false" without
1815  * updating i_version.
1816  *
1817  * If this function returns false, and no other metadata has changed, then we
1818  * can avoid logging the metadata.
1819  */
inode_maybe_inc_iversion(struct inode * inode,bool force)1820 bool inode_maybe_inc_iversion(struct inode *inode, bool force)
1821 {
1822 	u64 cur, new;
1823 
1824 	/*
1825 	 * The i_version field is not strictly ordered with any other inode
1826 	 * information, but the legacy inode_inc_iversion code used a spinlock
1827 	 * to serialize increments.
1828 	 *
1829 	 * Here, we add full memory barriers to ensure that any de-facto
1830 	 * ordering with other info is preserved.
1831 	 *
1832 	 * This barrier pairs with the barrier in inode_query_iversion()
1833 	 */
1834 	smp_mb();
1835 	cur = inode_peek_iversion_raw(inode);
1836 	do {
1837 		/* If flag is clear then we needn't do anything */
1838 		if (!force && !(cur & I_VERSION_QUERIED))
1839 			return false;
1840 
1841 		/* Since lowest bit is flag, add 2 to avoid it */
1842 		new = (cur & ~I_VERSION_QUERIED) + I_VERSION_INCREMENT;
1843 	} while (!atomic64_try_cmpxchg(&inode->i_version, &cur, new));
1844 	return true;
1845 }
1846 EXPORT_SYMBOL(inode_maybe_inc_iversion);
1847 
1848 /**
1849  * inode_query_iversion - read i_version for later use
1850  * @inode: inode from which i_version should be read
1851  *
1852  * Read the inode i_version counter. This should be used by callers that wish
1853  * to store the returned i_version for later comparison. This will guarantee
1854  * that a later query of the i_version will result in a different value if
1855  * anything has changed.
1856  *
1857  * In this implementation, we fetch the current value, set the QUERIED flag and
1858  * then try to swap it into place with a cmpxchg, if it wasn't already set. If
1859  * that fails, we try again with the newly fetched value from the cmpxchg.
1860  */
inode_query_iversion(struct inode * inode)1861 u64 inode_query_iversion(struct inode *inode)
1862 {
1863 	u64 cur, new;
1864 
1865 	cur = inode_peek_iversion_raw(inode);
1866 	do {
1867 		/* If flag is already set, then no need to swap */
1868 		if (cur & I_VERSION_QUERIED) {
1869 			/*
1870 			 * This barrier (and the implicit barrier in the
1871 			 * cmpxchg below) pairs with the barrier in
1872 			 * inode_maybe_inc_iversion().
1873 			 */
1874 			smp_mb();
1875 			break;
1876 		}
1877 
1878 		new = cur | I_VERSION_QUERIED;
1879 	} while (!atomic64_try_cmpxchg(&inode->i_version, &cur, new));
1880 	return cur >> I_VERSION_QUERIED_SHIFT;
1881 }
1882 EXPORT_SYMBOL(inode_query_iversion);
1883 
direct_write_fallback(struct kiocb * iocb,struct iov_iter * iter,ssize_t direct_written,ssize_t buffered_written)1884 ssize_t direct_write_fallback(struct kiocb *iocb, struct iov_iter *iter,
1885 		ssize_t direct_written, ssize_t buffered_written)
1886 {
1887 	struct address_space *mapping = iocb->ki_filp->f_mapping;
1888 	loff_t pos = iocb->ki_pos - buffered_written;
1889 	loff_t end = iocb->ki_pos - 1;
1890 	int err;
1891 
1892 	/*
1893 	 * If the buffered write fallback returned an error, we want to return
1894 	 * the number of bytes which were written by direct I/O, or the error
1895 	 * code if that was zero.
1896 	 *
1897 	 * Note that this differs from normal direct-io semantics, which will
1898 	 * return -EFOO even if some bytes were written.
1899 	 */
1900 	if (unlikely(buffered_written < 0)) {
1901 		if (direct_written)
1902 			return direct_written;
1903 		return buffered_written;
1904 	}
1905 
1906 	/*
1907 	 * We need to ensure that the page cache pages are written to disk and
1908 	 * invalidated to preserve the expected O_DIRECT semantics.
1909 	 */
1910 	err = filemap_write_and_wait_range(mapping, pos, end);
1911 	if (err < 0) {
1912 		/*
1913 		 * We don't know how much we wrote, so just return the number of
1914 		 * bytes which were direct-written
1915 		 */
1916 		iocb->ki_pos -= buffered_written;
1917 		if (direct_written)
1918 			return direct_written;
1919 		return err;
1920 	}
1921 	invalidate_mapping_pages(mapping, pos >> PAGE_SHIFT, end >> PAGE_SHIFT);
1922 	return direct_written + buffered_written;
1923 }
1924 EXPORT_SYMBOL_GPL(direct_write_fallback);
1925 
1926 /**
1927  * simple_inode_init_ts - initialize the timestamps for a new inode
1928  * @inode: inode to be initialized
1929  *
1930  * When a new inode is created, most filesystems set the timestamps to the
1931  * current time. Add a helper to do this.
1932  */
simple_inode_init_ts(struct inode * inode)1933 struct timespec64 simple_inode_init_ts(struct inode *inode)
1934 {
1935 	struct timespec64 ts = inode_set_ctime_current(inode);
1936 
1937 	inode_set_atime_to_ts(inode, ts);
1938 	inode_set_mtime_to_ts(inode, ts);
1939 	return ts;
1940 }
1941 EXPORT_SYMBOL(simple_inode_init_ts);
1942