xref: /linux/fs/tracefs/inode.c (revision ba199dc909a20fe62270ae4e93f263987bb9d119)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  inode.c - part of tracefs, a pseudo file system for activating tracing
4  *
5  * Based on debugfs by: Greg Kroah-Hartman <greg@kroah.com>
6  *
7  *  Copyright (C) 2014 Red Hat Inc, author: Steven Rostedt <srostedt@redhat.com>
8  *
9  * tracefs is the file system that is used by the tracing infrastructure.
10  */
11 
12 #include <linux/module.h>
13 #include <linux/fs.h>
14 #include <linux/fs_context.h>
15 #include <linux/fs_parser.h>
16 #include <linux/kobject.h>
17 #include <linux/namei.h>
18 #include <linux/tracefs.h>
19 #include <linux/fsnotify.h>
20 #include <linux/security.h>
21 #include <linux/seq_file.h>
22 #include <linux/magic.h>
23 #include <linux/slab.h>
24 #include "internal.h"
25 
26 #define TRACEFS_DEFAULT_MODE	0700
27 static struct kmem_cache *tracefs_inode_cachep __ro_after_init;
28 
29 static struct vfsmount *tracefs_mount;
30 static int tracefs_mount_count;
31 static bool tracefs_registered;
32 
33 /*
34  * Keep track of all tracefs_inodes in order to update their
35  * flags if necessary on a remount.
36  */
37 static DEFINE_SPINLOCK(tracefs_inode_lock);
38 static LIST_HEAD(tracefs_inodes);
39 
40 static struct inode *tracefs_alloc_inode(struct super_block *sb)
41 {
42 	struct tracefs_inode *ti;
43 	unsigned long flags;
44 
45 	ti = kmem_cache_alloc(tracefs_inode_cachep, GFP_KERNEL);
46 	if (!ti)
47 		return NULL;
48 
49 	spin_lock_irqsave(&tracefs_inode_lock, flags);
50 	list_add_rcu(&ti->list, &tracefs_inodes);
51 	spin_unlock_irqrestore(&tracefs_inode_lock, flags);
52 
53 	return &ti->vfs_inode;
54 }
55 
56 static void tracefs_free_inode_rcu(struct rcu_head *rcu)
57 {
58 	struct tracefs_inode *ti;
59 
60 	ti = container_of(rcu, struct tracefs_inode, rcu);
61 	kmem_cache_free(tracefs_inode_cachep, ti);
62 }
63 
64 static void tracefs_free_inode(struct inode *inode)
65 {
66 	struct tracefs_inode *ti = get_tracefs(inode);
67 	unsigned long flags;
68 
69 	spin_lock_irqsave(&tracefs_inode_lock, flags);
70 	list_del_rcu(&ti->list);
71 	spin_unlock_irqrestore(&tracefs_inode_lock, flags);
72 
73 	call_rcu(&ti->rcu, tracefs_free_inode_rcu);
74 }
75 
76 static ssize_t default_read_file(struct file *file, char __user *buf,
77 				 size_t count, loff_t *ppos)
78 {
79 	return 0;
80 }
81 
82 static ssize_t default_write_file(struct file *file, const char __user *buf,
83 				   size_t count, loff_t *ppos)
84 {
85 	return count;
86 }
87 
88 static const struct file_operations tracefs_file_operations = {
89 	.read =		default_read_file,
90 	.write =	default_write_file,
91 	.open =		simple_open,
92 	.llseek =	noop_llseek,
93 };
94 
95 static struct tracefs_dir_ops {
96 	int (*mkdir)(const char *name);
97 	int (*rmdir)(const char *name);
98 } tracefs_ops __ro_after_init;
99 
100 static char *get_dname(struct dentry *dentry)
101 {
102 	const char *dname;
103 	char *name;
104 	int len = dentry->d_name.len;
105 
106 	dname = dentry->d_name.name;
107 	name = kmalloc(len + 1, GFP_KERNEL);
108 	if (!name)
109 		return NULL;
110 	memcpy(name, dname, len);
111 	name[len] = 0;
112 	return name;
113 }
114 
115 static int tracefs_syscall_mkdir(struct mnt_idmap *idmap,
116 				 struct inode *inode, struct dentry *dentry,
117 				 umode_t mode)
118 {
119 	struct tracefs_inode *ti;
120 	char *name;
121 	int ret;
122 
123 	name = get_dname(dentry);
124 	if (!name)
125 		return -ENOMEM;
126 
127 	/*
128 	 * This is a new directory that does not take the default of
129 	 * the rootfs. It becomes the default permissions for all the
130 	 * files and directories underneath it.
131 	 */
132 	ti = get_tracefs(inode);
133 	ti->flags |= TRACEFS_INSTANCE_INODE;
134 	ti->private = inode;
135 
136 	/*
137 	 * The mkdir call can call the generic functions that create
138 	 * the files within the tracefs system. It is up to the individual
139 	 * mkdir routine to handle races.
140 	 */
141 	inode_unlock(inode);
142 	ret = tracefs_ops.mkdir(name);
143 	inode_lock(inode);
144 
145 	kfree(name);
146 
147 	return ret;
148 }
149 
150 static int tracefs_syscall_rmdir(struct inode *inode, struct dentry *dentry)
151 {
152 	char *name;
153 	int ret;
154 
155 	name = get_dname(dentry);
156 	if (!name)
157 		return -ENOMEM;
158 
159 	/*
160 	 * The rmdir call can call the generic functions that create
161 	 * the files within the tracefs system. It is up to the individual
162 	 * rmdir routine to handle races.
163 	 * This time we need to unlock not only the parent (inode) but
164 	 * also the directory that is being deleted.
165 	 */
166 	inode_unlock(inode);
167 	inode_unlock(d_inode(dentry));
168 
169 	ret = tracefs_ops.rmdir(name);
170 
171 	inode_lock_nested(inode, I_MUTEX_PARENT);
172 	inode_lock(d_inode(dentry));
173 
174 	kfree(name);
175 
176 	return ret;
177 }
178 
179 static void set_tracefs_inode_owner(struct inode *inode)
180 {
181 	struct tracefs_inode *ti = get_tracefs(inode);
182 	struct inode *root_inode = ti->private;
183 	kuid_t uid;
184 	kgid_t gid;
185 
186 	uid = root_inode->i_uid;
187 	gid = root_inode->i_gid;
188 
189 	/*
190 	 * If the root is not the mount point, then check the root's
191 	 * permissions. If it was never set, then default to the
192 	 * mount point.
193 	 */
194 	if (root_inode != d_inode(root_inode->i_sb->s_root)) {
195 		struct tracefs_inode *rti;
196 
197 		rti = get_tracefs(root_inode);
198 		root_inode = d_inode(root_inode->i_sb->s_root);
199 
200 		if (!(rti->flags & TRACEFS_UID_PERM_SET))
201 			uid = root_inode->i_uid;
202 
203 		if (!(rti->flags & TRACEFS_GID_PERM_SET))
204 			gid = root_inode->i_gid;
205 	}
206 
207 	/*
208 	 * If this inode has never been referenced, then update
209 	 * the permissions to the superblock.
210 	 */
211 	if (!(ti->flags & TRACEFS_UID_PERM_SET))
212 		inode->i_uid = uid;
213 
214 	if (!(ti->flags & TRACEFS_GID_PERM_SET))
215 		inode->i_gid = gid;
216 }
217 
218 static int tracefs_permission(struct mnt_idmap *idmap,
219 			      struct inode *inode, int mask)
220 {
221 	set_tracefs_inode_owner(inode);
222 	return generic_permission(idmap, inode, mask);
223 }
224 
225 static int tracefs_getattr(struct mnt_idmap *idmap,
226 			   const struct path *path, struct kstat *stat,
227 			   u32 request_mask, unsigned int flags)
228 {
229 	struct inode *inode = d_backing_inode(path->dentry);
230 
231 	set_tracefs_inode_owner(inode);
232 	generic_fillattr(idmap, request_mask, inode, stat);
233 	return 0;
234 }
235 
236 static int tracefs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
237 			   struct iattr *attr)
238 {
239 	unsigned int ia_valid = attr->ia_valid;
240 	struct inode *inode = d_inode(dentry);
241 	struct tracefs_inode *ti = get_tracefs(inode);
242 
243 	if (ia_valid & ATTR_UID)
244 		ti->flags |= TRACEFS_UID_PERM_SET;
245 
246 	if (ia_valid & ATTR_GID)
247 		ti->flags |= TRACEFS_GID_PERM_SET;
248 
249 	return simple_setattr(idmap, dentry, attr);
250 }
251 
252 static const struct inode_operations tracefs_instance_dir_inode_operations = {
253 	.lookup		= simple_lookup,
254 	.mkdir		= tracefs_syscall_mkdir,
255 	.rmdir		= tracefs_syscall_rmdir,
256 	.permission	= tracefs_permission,
257 	.getattr	= tracefs_getattr,
258 	.setattr	= tracefs_setattr,
259 };
260 
261 static const struct inode_operations tracefs_dir_inode_operations = {
262 	.lookup		= simple_lookup,
263 	.permission	= tracefs_permission,
264 	.getattr	= tracefs_getattr,
265 	.setattr	= tracefs_setattr,
266 };
267 
268 static const struct inode_operations tracefs_file_inode_operations = {
269 	.permission	= tracefs_permission,
270 	.getattr	= tracefs_getattr,
271 	.setattr	= tracefs_setattr,
272 };
273 
274 struct inode *tracefs_get_inode(struct super_block *sb)
275 {
276 	struct inode *inode = new_inode(sb);
277 	if (inode) {
278 		inode->i_ino = get_next_ino();
279 		simple_inode_init_ts(inode);
280 	}
281 	return inode;
282 }
283 
284 struct tracefs_fs_info {
285 	kuid_t uid;
286 	kgid_t gid;
287 	umode_t mode;
288 	/* Opt_* bitfield. */
289 	unsigned int opts;
290 };
291 
292 enum {
293 	Opt_uid,
294 	Opt_gid,
295 	Opt_mode,
296 };
297 
298 static const struct fs_parameter_spec tracefs_param_specs[] = {
299 	fsparam_gid	("gid",		Opt_gid),
300 	fsparam_u32oct	("mode",	Opt_mode),
301 	fsparam_uid	("uid",		Opt_uid),
302 	{}
303 };
304 
305 static int tracefs_parse_param(struct fs_context *fc, struct fs_parameter *param)
306 {
307 	struct tracefs_fs_info *opts = fc->s_fs_info;
308 	struct fs_parse_result result;
309 	int opt;
310 
311 	opt = fs_parse(fc, tracefs_param_specs, param, &result);
312 	if (opt < 0)
313 		return opt;
314 
315 	switch (opt) {
316 	case Opt_uid:
317 		opts->uid = result.uid;
318 		break;
319 	case Opt_gid:
320 		opts->gid = result.gid;
321 		break;
322 	case Opt_mode:
323 		opts->mode = result.uint_32 & S_IALLUGO;
324 		break;
325 	/*
326 	 * We might like to report bad mount options here;
327 	 * but traditionally tracefs has ignored all mount options
328 	 */
329 	}
330 
331 	opts->opts |= BIT(opt);
332 
333 	return 0;
334 }
335 
336 static int tracefs_apply_options(struct super_block *sb, bool remount)
337 {
338 	struct tracefs_fs_info *fsi = sb->s_fs_info;
339 	struct inode *inode = d_inode(sb->s_root);
340 	struct tracefs_inode *ti;
341 	bool update_uid, update_gid;
342 	umode_t tmp_mode;
343 
344 	/*
345 	 * On remount, only reset mode/uid/gid if they were provided as mount
346 	 * options.
347 	 */
348 
349 	if (!remount || fsi->opts & BIT(Opt_mode)) {
350 		tmp_mode = READ_ONCE(inode->i_mode) & ~S_IALLUGO;
351 		tmp_mode |= fsi->mode;
352 		WRITE_ONCE(inode->i_mode, tmp_mode);
353 	}
354 
355 	if (!remount || fsi->opts & BIT(Opt_uid))
356 		inode->i_uid = fsi->uid;
357 
358 	if (!remount || fsi->opts & BIT(Opt_gid))
359 		inode->i_gid = fsi->gid;
360 
361 	if (remount && (fsi->opts & BIT(Opt_uid) || fsi->opts & BIT(Opt_gid))) {
362 
363 		update_uid = fsi->opts & BIT(Opt_uid);
364 		update_gid = fsi->opts & BIT(Opt_gid);
365 
366 		rcu_read_lock();
367 		list_for_each_entry_rcu(ti, &tracefs_inodes, list) {
368 			if (update_uid) {
369 				ti->flags &= ~TRACEFS_UID_PERM_SET;
370 				ti->vfs_inode.i_uid = fsi->uid;
371 			}
372 
373 			if (update_gid) {
374 				ti->flags &= ~TRACEFS_GID_PERM_SET;
375 				ti->vfs_inode.i_gid = fsi->gid;
376 			}
377 
378 			/*
379 			 * Note, the above ti->vfs_inode updates are
380 			 * used in eventfs_remount() so they must come
381 			 * before calling it.
382 			 */
383 			if (ti->flags & TRACEFS_EVENT_INODE)
384 				eventfs_remount(ti, update_uid, update_gid);
385 		}
386 		rcu_read_unlock();
387 	}
388 
389 	return 0;
390 }
391 
392 static int tracefs_reconfigure(struct fs_context *fc)
393 {
394 	struct super_block *sb = fc->root->d_sb;
395 	struct tracefs_fs_info *sb_opts = sb->s_fs_info;
396 	struct tracefs_fs_info *new_opts = fc->s_fs_info;
397 
398 	sync_filesystem(sb);
399 	/* structure copy of new mount options to sb */
400 	*sb_opts = *new_opts;
401 
402 	return tracefs_apply_options(sb, true);
403 }
404 
405 static int tracefs_show_options(struct seq_file *m, struct dentry *root)
406 {
407 	struct tracefs_fs_info *fsi = root->d_sb->s_fs_info;
408 
409 	if (!uid_eq(fsi->uid, GLOBAL_ROOT_UID))
410 		seq_printf(m, ",uid=%u",
411 			   from_kuid_munged(&init_user_ns, fsi->uid));
412 	if (!gid_eq(fsi->gid, GLOBAL_ROOT_GID))
413 		seq_printf(m, ",gid=%u",
414 			   from_kgid_munged(&init_user_ns, fsi->gid));
415 	if (fsi->mode != TRACEFS_DEFAULT_MODE)
416 		seq_printf(m, ",mode=%o", fsi->mode);
417 
418 	return 0;
419 }
420 
421 static int tracefs_drop_inode(struct inode *inode)
422 {
423 	struct tracefs_inode *ti = get_tracefs(inode);
424 
425 	/*
426 	 * This inode is being freed and cannot be used for
427 	 * eventfs. Clear the flag so that it doesn't call into
428 	 * eventfs during the remount flag updates. The eventfs_inode
429 	 * gets freed after an RCU cycle, so the content will still
430 	 * be safe if the iteration is going on now.
431 	 */
432 	ti->flags &= ~TRACEFS_EVENT_INODE;
433 
434 	return 1;
435 }
436 
437 static const struct super_operations tracefs_super_operations = {
438 	.alloc_inode    = tracefs_alloc_inode,
439 	.free_inode     = tracefs_free_inode,
440 	.drop_inode     = tracefs_drop_inode,
441 	.statfs		= simple_statfs,
442 	.show_options	= tracefs_show_options,
443 };
444 
445 /*
446  * It would be cleaner if eventfs had its own dentry ops.
447  *
448  * Note that d_revalidate is called potentially under RCU,
449  * so it can't take the eventfs mutex etc. It's fine - if
450  * we open a file just as it's marked dead, things will
451  * still work just fine, and just see the old stale case.
452  */
453 static void tracefs_d_release(struct dentry *dentry)
454 {
455 	if (dentry->d_fsdata)
456 		eventfs_d_release(dentry);
457 }
458 
459 static int tracefs_d_revalidate(struct dentry *dentry, unsigned int flags)
460 {
461 	struct eventfs_inode *ei = dentry->d_fsdata;
462 
463 	return !(ei && ei->is_freed);
464 }
465 
466 static const struct dentry_operations tracefs_dentry_operations = {
467 	.d_revalidate = tracefs_d_revalidate,
468 	.d_release = tracefs_d_release,
469 };
470 
471 static int tracefs_fill_super(struct super_block *sb, struct fs_context *fc)
472 {
473 	static const struct tree_descr trace_files[] = {{""}};
474 	int err;
475 
476 	err = simple_fill_super(sb, TRACEFS_MAGIC, trace_files);
477 	if (err)
478 		return err;
479 
480 	sb->s_op = &tracefs_super_operations;
481 	sb->s_d_op = &tracefs_dentry_operations;
482 
483 	tracefs_apply_options(sb, false);
484 
485 	return 0;
486 }
487 
488 static int tracefs_get_tree(struct fs_context *fc)
489 {
490 	return get_tree_single(fc, tracefs_fill_super);
491 }
492 
493 static void tracefs_free_fc(struct fs_context *fc)
494 {
495 	kfree(fc->s_fs_info);
496 }
497 
498 static const struct fs_context_operations tracefs_context_ops = {
499 	.free		= tracefs_free_fc,
500 	.parse_param	= tracefs_parse_param,
501 	.get_tree	= tracefs_get_tree,
502 	.reconfigure	= tracefs_reconfigure,
503 };
504 
505 static int tracefs_init_fs_context(struct fs_context *fc)
506 {
507 	struct tracefs_fs_info *fsi;
508 
509 	fsi = kzalloc(sizeof(struct tracefs_fs_info), GFP_KERNEL);
510 	if (!fsi)
511 		return -ENOMEM;
512 
513 	fsi->mode = TRACEFS_DEFAULT_MODE;
514 
515 	fc->s_fs_info = fsi;
516 	fc->ops = &tracefs_context_ops;
517 	return 0;
518 }
519 
520 static struct file_system_type trace_fs_type = {
521 	.owner =	THIS_MODULE,
522 	.name =		"tracefs",
523 	.init_fs_context = tracefs_init_fs_context,
524 	.parameters	= tracefs_param_specs,
525 	.kill_sb =	kill_litter_super,
526 };
527 MODULE_ALIAS_FS("tracefs");
528 
529 struct dentry *tracefs_start_creating(const char *name, struct dentry *parent)
530 {
531 	struct dentry *dentry;
532 	int error;
533 
534 	pr_debug("tracefs: creating file '%s'\n",name);
535 
536 	error = simple_pin_fs(&trace_fs_type, &tracefs_mount,
537 			      &tracefs_mount_count);
538 	if (error)
539 		return ERR_PTR(error);
540 
541 	/* If the parent is not specified, we create it in the root.
542 	 * We need the root dentry to do this, which is in the super
543 	 * block. A pointer to that is in the struct vfsmount that we
544 	 * have around.
545 	 */
546 	if (!parent)
547 		parent = tracefs_mount->mnt_root;
548 
549 	inode_lock(d_inode(parent));
550 	if (unlikely(IS_DEADDIR(d_inode(parent))))
551 		dentry = ERR_PTR(-ENOENT);
552 	else
553 		dentry = lookup_one_len(name, parent, strlen(name));
554 	if (!IS_ERR(dentry) && d_inode(dentry)) {
555 		dput(dentry);
556 		dentry = ERR_PTR(-EEXIST);
557 	}
558 
559 	if (IS_ERR(dentry)) {
560 		inode_unlock(d_inode(parent));
561 		simple_release_fs(&tracefs_mount, &tracefs_mount_count);
562 	}
563 
564 	return dentry;
565 }
566 
567 struct dentry *tracefs_failed_creating(struct dentry *dentry)
568 {
569 	inode_unlock(d_inode(dentry->d_parent));
570 	dput(dentry);
571 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
572 	return NULL;
573 }
574 
575 struct dentry *tracefs_end_creating(struct dentry *dentry)
576 {
577 	inode_unlock(d_inode(dentry->d_parent));
578 	return dentry;
579 }
580 
581 /* Find the inode that this will use for default */
582 static struct inode *instance_inode(struct dentry *parent, struct inode *inode)
583 {
584 	struct tracefs_inode *ti;
585 
586 	/* If parent is NULL then use root inode */
587 	if (!parent)
588 		return d_inode(inode->i_sb->s_root);
589 
590 	/* Find the inode that is flagged as an instance or the root inode */
591 	while (!IS_ROOT(parent)) {
592 		ti = get_tracefs(d_inode(parent));
593 		if (ti->flags & TRACEFS_INSTANCE_INODE)
594 			break;
595 		parent = parent->d_parent;
596 	}
597 
598 	return d_inode(parent);
599 }
600 
601 /**
602  * tracefs_create_file - create a file in the tracefs filesystem
603  * @name: a pointer to a string containing the name of the file to create.
604  * @mode: the permission that the file should have.
605  * @parent: a pointer to the parent dentry for this file.  This should be a
606  *          directory dentry if set.  If this parameter is NULL, then the
607  *          file will be created in the root of the tracefs filesystem.
608  * @data: a pointer to something that the caller will want to get to later
609  *        on.  The inode.i_private pointer will point to this value on
610  *        the open() call.
611  * @fops: a pointer to a struct file_operations that should be used for
612  *        this file.
613  *
614  * This is the basic "create a file" function for tracefs.  It allows for a
615  * wide range of flexibility in creating a file, or a directory (if you want
616  * to create a directory, the tracefs_create_dir() function is
617  * recommended to be used instead.)
618  *
619  * This function will return a pointer to a dentry if it succeeds.  This
620  * pointer must be passed to the tracefs_remove() function when the file is
621  * to be removed (no automatic cleanup happens if your module is unloaded,
622  * you are responsible here.)  If an error occurs, %NULL will be returned.
623  *
624  * If tracefs is not enabled in the kernel, the value -%ENODEV will be
625  * returned.
626  */
627 struct dentry *tracefs_create_file(const char *name, umode_t mode,
628 				   struct dentry *parent, void *data,
629 				   const struct file_operations *fops)
630 {
631 	struct tracefs_inode *ti;
632 	struct dentry *dentry;
633 	struct inode *inode;
634 
635 	if (security_locked_down(LOCKDOWN_TRACEFS))
636 		return NULL;
637 
638 	if (!(mode & S_IFMT))
639 		mode |= S_IFREG;
640 	BUG_ON(!S_ISREG(mode));
641 	dentry = tracefs_start_creating(name, parent);
642 
643 	if (IS_ERR(dentry))
644 		return NULL;
645 
646 	inode = tracefs_get_inode(dentry->d_sb);
647 	if (unlikely(!inode))
648 		return tracefs_failed_creating(dentry);
649 
650 	ti = get_tracefs(inode);
651 	ti->private = instance_inode(parent, inode);
652 
653 	inode->i_mode = mode;
654 	inode->i_op = &tracefs_file_inode_operations;
655 	inode->i_fop = fops ? fops : &tracefs_file_operations;
656 	inode->i_private = data;
657 	inode->i_uid = d_inode(dentry->d_parent)->i_uid;
658 	inode->i_gid = d_inode(dentry->d_parent)->i_gid;
659 	d_instantiate(dentry, inode);
660 	fsnotify_create(d_inode(dentry->d_parent), dentry);
661 	return tracefs_end_creating(dentry);
662 }
663 
664 static struct dentry *__create_dir(const char *name, struct dentry *parent,
665 				   const struct inode_operations *ops)
666 {
667 	struct tracefs_inode *ti;
668 	struct dentry *dentry = tracefs_start_creating(name, parent);
669 	struct inode *inode;
670 
671 	if (IS_ERR(dentry))
672 		return NULL;
673 
674 	inode = tracefs_get_inode(dentry->d_sb);
675 	if (unlikely(!inode))
676 		return tracefs_failed_creating(dentry);
677 
678 	/* Do not set bits for OTH */
679 	inode->i_mode = S_IFDIR | S_IRWXU | S_IRUSR| S_IRGRP | S_IXUSR | S_IXGRP;
680 	inode->i_op = ops;
681 	inode->i_fop = &simple_dir_operations;
682 	inode->i_uid = d_inode(dentry->d_parent)->i_uid;
683 	inode->i_gid = d_inode(dentry->d_parent)->i_gid;
684 
685 	ti = get_tracefs(inode);
686 	ti->private = instance_inode(parent, inode);
687 
688 	/* directory inodes start off with i_nlink == 2 (for "." entry) */
689 	inc_nlink(inode);
690 	d_instantiate(dentry, inode);
691 	inc_nlink(d_inode(dentry->d_parent));
692 	fsnotify_mkdir(d_inode(dentry->d_parent), dentry);
693 	return tracefs_end_creating(dentry);
694 }
695 
696 /**
697  * tracefs_create_dir - create a directory in the tracefs filesystem
698  * @name: a pointer to a string containing the name of the directory to
699  *        create.
700  * @parent: a pointer to the parent dentry for this file.  This should be a
701  *          directory dentry if set.  If this parameter is NULL, then the
702  *          directory will be created in the root of the tracefs filesystem.
703  *
704  * This function creates a directory in tracefs with the given name.
705  *
706  * This function will return a pointer to a dentry if it succeeds.  This
707  * pointer must be passed to the tracefs_remove() function when the file is
708  * to be removed. If an error occurs, %NULL will be returned.
709  *
710  * If tracing is not enabled in the kernel, the value -%ENODEV will be
711  * returned.
712  */
713 struct dentry *tracefs_create_dir(const char *name, struct dentry *parent)
714 {
715 	if (security_locked_down(LOCKDOWN_TRACEFS))
716 		return NULL;
717 
718 	return __create_dir(name, parent, &tracefs_dir_inode_operations);
719 }
720 
721 /**
722  * tracefs_create_instance_dir - create the tracing instances directory
723  * @name: The name of the instances directory to create
724  * @parent: The parent directory that the instances directory will exist
725  * @mkdir: The function to call when a mkdir is performed.
726  * @rmdir: The function to call when a rmdir is performed.
727  *
728  * Only one instances directory is allowed.
729  *
730  * The instances directory is special as it allows for mkdir and rmdir
731  * to be done by userspace. When a mkdir or rmdir is performed, the inode
732  * locks are released and the methods passed in (@mkdir and @rmdir) are
733  * called without locks and with the name of the directory being created
734  * within the instances directory.
735  *
736  * Returns the dentry of the instances directory.
737  */
738 __init struct dentry *tracefs_create_instance_dir(const char *name,
739 					  struct dentry *parent,
740 					  int (*mkdir)(const char *name),
741 					  int (*rmdir)(const char *name))
742 {
743 	struct dentry *dentry;
744 
745 	/* Only allow one instance of the instances directory. */
746 	if (WARN_ON(tracefs_ops.mkdir || tracefs_ops.rmdir))
747 		return NULL;
748 
749 	dentry = __create_dir(name, parent, &tracefs_instance_dir_inode_operations);
750 	if (!dentry)
751 		return NULL;
752 
753 	tracefs_ops.mkdir = mkdir;
754 	tracefs_ops.rmdir = rmdir;
755 
756 	return dentry;
757 }
758 
759 static void remove_one(struct dentry *victim)
760 {
761 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
762 }
763 
764 /**
765  * tracefs_remove - recursively removes a directory
766  * @dentry: a pointer to a the dentry of the directory to be removed.
767  *
768  * This function recursively removes a directory tree in tracefs that
769  * was previously created with a call to another tracefs function
770  * (like tracefs_create_file() or variants thereof.)
771  */
772 void tracefs_remove(struct dentry *dentry)
773 {
774 	if (IS_ERR_OR_NULL(dentry))
775 		return;
776 
777 	simple_pin_fs(&trace_fs_type, &tracefs_mount, &tracefs_mount_count);
778 	simple_recursive_removal(dentry, remove_one);
779 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
780 }
781 
782 /**
783  * tracefs_initialized - Tells whether tracefs has been registered
784  */
785 bool tracefs_initialized(void)
786 {
787 	return tracefs_registered;
788 }
789 
790 static void init_once(void *foo)
791 {
792 	struct tracefs_inode *ti = (struct tracefs_inode *) foo;
793 
794 	/* inode_init_once() calls memset() on the vfs_inode portion */
795 	inode_init_once(&ti->vfs_inode);
796 
797 	/* Zero out the rest */
798 	memset_after(ti, 0, vfs_inode);
799 }
800 
801 static int __init tracefs_init(void)
802 {
803 	int retval;
804 
805 	tracefs_inode_cachep = kmem_cache_create("tracefs_inode_cache",
806 						 sizeof(struct tracefs_inode),
807 						 0, (SLAB_RECLAIM_ACCOUNT|
808 						     SLAB_ACCOUNT),
809 						 init_once);
810 	if (!tracefs_inode_cachep)
811 		return -ENOMEM;
812 
813 	retval = sysfs_create_mount_point(kernel_kobj, "tracing");
814 	if (retval)
815 		return -EINVAL;
816 
817 	retval = register_filesystem(&trace_fs_type);
818 	if (!retval)
819 		tracefs_registered = true;
820 
821 	return retval;
822 }
823 core_initcall(tracefs_init);
824