xref: /openbmc/linux/fs/tracefs/inode.c (revision da097dcc)
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/mount.h>
15 #include <linux/kobject.h>
16 #include <linux/namei.h>
17 #include <linux/tracefs.h>
18 #include <linux/fsnotify.h>
19 #include <linux/security.h>
20 #include <linux/seq_file.h>
21 #include <linux/parser.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 		inode->i_atime = inode->i_mtime = inode_set_ctime_current(inode);
280 	}
281 	return inode;
282 }
283 
284 struct tracefs_mount_opts {
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 	Opt_err
297 };
298 
299 static const match_table_t tokens = {
300 	{Opt_uid, "uid=%u"},
301 	{Opt_gid, "gid=%u"},
302 	{Opt_mode, "mode=%o"},
303 	{Opt_err, NULL}
304 };
305 
306 struct tracefs_fs_info {
307 	struct tracefs_mount_opts mount_opts;
308 };
309 
310 static int tracefs_parse_options(char *data, struct tracefs_mount_opts *opts)
311 {
312 	substring_t args[MAX_OPT_ARGS];
313 	int option;
314 	int token;
315 	kuid_t uid;
316 	kgid_t gid;
317 	char *p;
318 
319 	opts->opts = 0;
320 	opts->mode = TRACEFS_DEFAULT_MODE;
321 
322 	while ((p = strsep(&data, ",")) != NULL) {
323 		if (!*p)
324 			continue;
325 
326 		token = match_token(p, tokens, args);
327 		switch (token) {
328 		case Opt_uid:
329 			if (match_int(&args[0], &option))
330 				return -EINVAL;
331 			uid = make_kuid(current_user_ns(), option);
332 			if (!uid_valid(uid))
333 				return -EINVAL;
334 			opts->uid = uid;
335 			break;
336 		case Opt_gid:
337 			if (match_int(&args[0], &option))
338 				return -EINVAL;
339 			gid = make_kgid(current_user_ns(), option);
340 			if (!gid_valid(gid))
341 				return -EINVAL;
342 			opts->gid = gid;
343 			break;
344 		case Opt_mode:
345 			if (match_octal(&args[0], &option))
346 				return -EINVAL;
347 			opts->mode = option & S_IALLUGO;
348 			break;
349 		/*
350 		 * We might like to report bad mount options here;
351 		 * but traditionally tracefs has ignored all mount options
352 		 */
353 		}
354 
355 		opts->opts |= BIT(token);
356 	}
357 
358 	return 0;
359 }
360 
361 static int tracefs_apply_options(struct super_block *sb, bool remount)
362 {
363 	struct tracefs_fs_info *fsi = sb->s_fs_info;
364 	struct inode *inode = d_inode(sb->s_root);
365 	struct tracefs_mount_opts *opts = &fsi->mount_opts;
366 	struct tracefs_inode *ti;
367 	bool update_uid, update_gid;
368 	umode_t tmp_mode;
369 
370 	/*
371 	 * On remount, only reset mode/uid/gid if they were provided as mount
372 	 * options.
373 	 */
374 
375 	if (!remount || opts->opts & BIT(Opt_mode)) {
376 		tmp_mode = READ_ONCE(inode->i_mode) & ~S_IALLUGO;
377 		tmp_mode |= opts->mode;
378 		WRITE_ONCE(inode->i_mode, tmp_mode);
379 	}
380 
381 	if (!remount || opts->opts & BIT(Opt_uid))
382 		inode->i_uid = opts->uid;
383 
384 	if (!remount || opts->opts & BIT(Opt_gid))
385 		inode->i_gid = opts->gid;
386 
387 	if (remount && (opts->opts & BIT(Opt_uid) || opts->opts & BIT(Opt_gid))) {
388 
389 		update_uid = opts->opts & BIT(Opt_uid);
390 		update_gid = opts->opts & BIT(Opt_gid);
391 
392 		rcu_read_lock();
393 		list_for_each_entry_rcu(ti, &tracefs_inodes, list) {
394 			if (update_uid)
395 				ti->flags &= ~TRACEFS_UID_PERM_SET;
396 
397 			if (update_gid)
398 				ti->flags &= ~TRACEFS_GID_PERM_SET;
399 
400 			if (ti->flags & TRACEFS_EVENT_INODE)
401 				eventfs_remount(ti, update_uid, update_gid);
402 		}
403 		rcu_read_unlock();
404 	}
405 
406 	return 0;
407 }
408 
409 static int tracefs_remount(struct super_block *sb, int *flags, char *data)
410 {
411 	int err;
412 	struct tracefs_fs_info *fsi = sb->s_fs_info;
413 
414 	sync_filesystem(sb);
415 	err = tracefs_parse_options(data, &fsi->mount_opts);
416 	if (err)
417 		goto fail;
418 
419 	tracefs_apply_options(sb, true);
420 
421 fail:
422 	return err;
423 }
424 
425 static int tracefs_show_options(struct seq_file *m, struct dentry *root)
426 {
427 	struct tracefs_fs_info *fsi = root->d_sb->s_fs_info;
428 	struct tracefs_mount_opts *opts = &fsi->mount_opts;
429 
430 	if (!uid_eq(opts->uid, GLOBAL_ROOT_UID))
431 		seq_printf(m, ",uid=%u",
432 			   from_kuid_munged(&init_user_ns, opts->uid));
433 	if (!gid_eq(opts->gid, GLOBAL_ROOT_GID))
434 		seq_printf(m, ",gid=%u",
435 			   from_kgid_munged(&init_user_ns, opts->gid));
436 	if (opts->mode != TRACEFS_DEFAULT_MODE)
437 		seq_printf(m, ",mode=%o", opts->mode);
438 
439 	return 0;
440 }
441 
442 static int tracefs_drop_inode(struct inode *inode)
443 {
444 	struct tracefs_inode *ti = get_tracefs(inode);
445 
446 	/*
447 	 * This inode is being freed and cannot be used for
448 	 * eventfs. Clear the flag so that it doesn't call into
449 	 * eventfs during the remount flag updates. The eventfs_inode
450 	 * gets freed after an RCU cycle, so the content will still
451 	 * be safe if the iteration is going on now.
452 	 */
453 	ti->flags &= ~TRACEFS_EVENT_INODE;
454 
455 	return 1;
456 }
457 
458 static const struct super_operations tracefs_super_operations = {
459 	.alloc_inode    = tracefs_alloc_inode,
460 	.free_inode     = tracefs_free_inode,
461 	.drop_inode     = tracefs_drop_inode,
462 	.statfs		= simple_statfs,
463 	.remount_fs	= tracefs_remount,
464 	.show_options	= tracefs_show_options,
465 };
466 
467 /*
468  * It would be cleaner if eventfs had its own dentry ops.
469  *
470  * Note that d_revalidate is called potentially under RCU,
471  * so it can't take the eventfs mutex etc. It's fine - if
472  * we open a file just as it's marked dead, things will
473  * still work just fine, and just see the old stale case.
474  */
475 static void tracefs_d_release(struct dentry *dentry)
476 {
477 	if (dentry->d_fsdata)
478 		eventfs_d_release(dentry);
479 }
480 
481 static int tracefs_d_revalidate(struct dentry *dentry, unsigned int flags)
482 {
483 	struct eventfs_inode *ei = dentry->d_fsdata;
484 
485 	return !(ei && ei->is_freed);
486 }
487 
488 static const struct dentry_operations tracefs_dentry_operations = {
489 	.d_revalidate = tracefs_d_revalidate,
490 	.d_release = tracefs_d_release,
491 };
492 
493 static int trace_fill_super(struct super_block *sb, void *data, int silent)
494 {
495 	static const struct tree_descr trace_files[] = {{""}};
496 	struct tracefs_fs_info *fsi;
497 	int err;
498 
499 	fsi = kzalloc(sizeof(struct tracefs_fs_info), GFP_KERNEL);
500 	sb->s_fs_info = fsi;
501 	if (!fsi) {
502 		err = -ENOMEM;
503 		goto fail;
504 	}
505 
506 	err = tracefs_parse_options(data, &fsi->mount_opts);
507 	if (err)
508 		goto fail;
509 
510 	err  =  simple_fill_super(sb, TRACEFS_MAGIC, trace_files);
511 	if (err)
512 		goto fail;
513 
514 	sb->s_op = &tracefs_super_operations;
515 	sb->s_d_op = &tracefs_dentry_operations;
516 
517 	tracefs_apply_options(sb, false);
518 
519 	return 0;
520 
521 fail:
522 	kfree(fsi);
523 	sb->s_fs_info = NULL;
524 	return err;
525 }
526 
527 static struct dentry *trace_mount(struct file_system_type *fs_type,
528 			int flags, const char *dev_name,
529 			void *data)
530 {
531 	return mount_single(fs_type, flags, data, trace_fill_super);
532 }
533 
534 static struct file_system_type trace_fs_type = {
535 	.owner =	THIS_MODULE,
536 	.name =		"tracefs",
537 	.mount =	trace_mount,
538 	.kill_sb =	kill_litter_super,
539 };
540 MODULE_ALIAS_FS("tracefs");
541 
542 struct dentry *tracefs_start_creating(const char *name, struct dentry *parent)
543 {
544 	struct dentry *dentry;
545 	int error;
546 
547 	pr_debug("tracefs: creating file '%s'\n",name);
548 
549 	error = simple_pin_fs(&trace_fs_type, &tracefs_mount,
550 			      &tracefs_mount_count);
551 	if (error)
552 		return ERR_PTR(error);
553 
554 	/* If the parent is not specified, we create it in the root.
555 	 * We need the root dentry to do this, which is in the super
556 	 * block. A pointer to that is in the struct vfsmount that we
557 	 * have around.
558 	 */
559 	if (!parent)
560 		parent = tracefs_mount->mnt_root;
561 
562 	inode_lock(d_inode(parent));
563 	if (unlikely(IS_DEADDIR(d_inode(parent))))
564 		dentry = ERR_PTR(-ENOENT);
565 	else
566 		dentry = lookup_one_len(name, parent, strlen(name));
567 	if (!IS_ERR(dentry) && d_inode(dentry)) {
568 		dput(dentry);
569 		dentry = ERR_PTR(-EEXIST);
570 	}
571 
572 	if (IS_ERR(dentry)) {
573 		inode_unlock(d_inode(parent));
574 		simple_release_fs(&tracefs_mount, &tracefs_mount_count);
575 	}
576 
577 	return dentry;
578 }
579 
580 struct dentry *tracefs_failed_creating(struct dentry *dentry)
581 {
582 	inode_unlock(d_inode(dentry->d_parent));
583 	dput(dentry);
584 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
585 	return NULL;
586 }
587 
588 struct dentry *tracefs_end_creating(struct dentry *dentry)
589 {
590 	inode_unlock(d_inode(dentry->d_parent));
591 	return dentry;
592 }
593 
594 /* Find the inode that this will use for default */
595 static struct inode *instance_inode(struct dentry *parent, struct inode *inode)
596 {
597 	struct tracefs_inode *ti;
598 
599 	/* If parent is NULL then use root inode */
600 	if (!parent)
601 		return d_inode(inode->i_sb->s_root);
602 
603 	/* Find the inode that is flagged as an instance or the root inode */
604 	while (!IS_ROOT(parent)) {
605 		ti = get_tracefs(d_inode(parent));
606 		if (ti->flags & TRACEFS_INSTANCE_INODE)
607 			break;
608 		parent = parent->d_parent;
609 	}
610 
611 	return d_inode(parent);
612 }
613 
614 /**
615  * tracefs_create_file - create a file in the tracefs filesystem
616  * @name: a pointer to a string containing the name of the file to create.
617  * @mode: the permission that the file should have.
618  * @parent: a pointer to the parent dentry for this file.  This should be a
619  *          directory dentry if set.  If this parameter is NULL, then the
620  *          file will be created in the root of the tracefs filesystem.
621  * @data: a pointer to something that the caller will want to get to later
622  *        on.  The inode.i_private pointer will point to this value on
623  *        the open() call.
624  * @fops: a pointer to a struct file_operations that should be used for
625  *        this file.
626  *
627  * This is the basic "create a file" function for tracefs.  It allows for a
628  * wide range of flexibility in creating a file, or a directory (if you want
629  * to create a directory, the tracefs_create_dir() function is
630  * recommended to be used instead.)
631  *
632  * This function will return a pointer to a dentry if it succeeds.  This
633  * pointer must be passed to the tracefs_remove() function when the file is
634  * to be removed (no automatic cleanup happens if your module is unloaded,
635  * you are responsible here.)  If an error occurs, %NULL will be returned.
636  *
637  * If tracefs is not enabled in the kernel, the value -%ENODEV will be
638  * returned.
639  */
640 struct dentry *tracefs_create_file(const char *name, umode_t mode,
641 				   struct dentry *parent, void *data,
642 				   const struct file_operations *fops)
643 {
644 	struct tracefs_inode *ti;
645 	struct dentry *dentry;
646 	struct inode *inode;
647 
648 	if (security_locked_down(LOCKDOWN_TRACEFS))
649 		return NULL;
650 
651 	if (!(mode & S_IFMT))
652 		mode |= S_IFREG;
653 	BUG_ON(!S_ISREG(mode));
654 	dentry = tracefs_start_creating(name, parent);
655 
656 	if (IS_ERR(dentry))
657 		return NULL;
658 
659 	inode = tracefs_get_inode(dentry->d_sb);
660 	if (unlikely(!inode))
661 		return tracefs_failed_creating(dentry);
662 
663 	ti = get_tracefs(inode);
664 	ti->private = instance_inode(parent, inode);
665 
666 	inode->i_mode = mode;
667 	inode->i_op = &tracefs_file_inode_operations;
668 	inode->i_fop = fops ? fops : &tracefs_file_operations;
669 	inode->i_private = data;
670 	inode->i_uid = d_inode(dentry->d_parent)->i_uid;
671 	inode->i_gid = d_inode(dentry->d_parent)->i_gid;
672 	d_instantiate(dentry, inode);
673 	fsnotify_create(d_inode(dentry->d_parent), dentry);
674 	return tracefs_end_creating(dentry);
675 }
676 
677 static struct dentry *__create_dir(const char *name, struct dentry *parent,
678 				   const struct inode_operations *ops)
679 {
680 	struct tracefs_inode *ti;
681 	struct dentry *dentry = tracefs_start_creating(name, parent);
682 	struct inode *inode;
683 
684 	if (IS_ERR(dentry))
685 		return NULL;
686 
687 	inode = tracefs_get_inode(dentry->d_sb);
688 	if (unlikely(!inode))
689 		return tracefs_failed_creating(dentry);
690 
691 	/* Do not set bits for OTH */
692 	inode->i_mode = S_IFDIR | S_IRWXU | S_IRUSR| S_IRGRP | S_IXUSR | S_IXGRP;
693 	inode->i_op = ops;
694 	inode->i_fop = &simple_dir_operations;
695 	inode->i_uid = d_inode(dentry->d_parent)->i_uid;
696 	inode->i_gid = d_inode(dentry->d_parent)->i_gid;
697 
698 	ti = get_tracefs(inode);
699 	ti->private = instance_inode(parent, inode);
700 
701 	/* directory inodes start off with i_nlink == 2 (for "." entry) */
702 	inc_nlink(inode);
703 	d_instantiate(dentry, inode);
704 	inc_nlink(d_inode(dentry->d_parent));
705 	fsnotify_mkdir(d_inode(dentry->d_parent), dentry);
706 	return tracefs_end_creating(dentry);
707 }
708 
709 /**
710  * tracefs_create_dir - create a directory in the tracefs filesystem
711  * @name: a pointer to a string containing the name of the directory to
712  *        create.
713  * @parent: a pointer to the parent dentry for this file.  This should be a
714  *          directory dentry if set.  If this parameter is NULL, then the
715  *          directory will be created in the root of the tracefs filesystem.
716  *
717  * This function creates a directory in tracefs with the given name.
718  *
719  * This function will return a pointer to a dentry if it succeeds.  This
720  * pointer must be passed to the tracefs_remove() function when the file is
721  * to be removed. If an error occurs, %NULL will be returned.
722  *
723  * If tracing is not enabled in the kernel, the value -%ENODEV will be
724  * returned.
725  */
726 struct dentry *tracefs_create_dir(const char *name, struct dentry *parent)
727 {
728 	if (security_locked_down(LOCKDOWN_TRACEFS))
729 		return NULL;
730 
731 	return __create_dir(name, parent, &tracefs_dir_inode_operations);
732 }
733 
734 /**
735  * tracefs_create_instance_dir - create the tracing instances directory
736  * @name: The name of the instances directory to create
737  * @parent: The parent directory that the instances directory will exist
738  * @mkdir: The function to call when a mkdir is performed.
739  * @rmdir: The function to call when a rmdir is performed.
740  *
741  * Only one instances directory is allowed.
742  *
743  * The instances directory is special as it allows for mkdir and rmdir
744  * to be done by userspace. When a mkdir or rmdir is performed, the inode
745  * locks are released and the methods passed in (@mkdir and @rmdir) are
746  * called without locks and with the name of the directory being created
747  * within the instances directory.
748  *
749  * Returns the dentry of the instances directory.
750  */
751 __init struct dentry *tracefs_create_instance_dir(const char *name,
752 					  struct dentry *parent,
753 					  int (*mkdir)(const char *name),
754 					  int (*rmdir)(const char *name))
755 {
756 	struct dentry *dentry;
757 
758 	/* Only allow one instance of the instances directory. */
759 	if (WARN_ON(tracefs_ops.mkdir || tracefs_ops.rmdir))
760 		return NULL;
761 
762 	dentry = __create_dir(name, parent, &tracefs_instance_dir_inode_operations);
763 	if (!dentry)
764 		return NULL;
765 
766 	tracefs_ops.mkdir = mkdir;
767 	tracefs_ops.rmdir = rmdir;
768 
769 	return dentry;
770 }
771 
772 static void remove_one(struct dentry *victim)
773 {
774 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
775 }
776 
777 /**
778  * tracefs_remove - recursively removes a directory
779  * @dentry: a pointer to a the dentry of the directory to be removed.
780  *
781  * This function recursively removes a directory tree in tracefs that
782  * was previously created with a call to another tracefs function
783  * (like tracefs_create_file() or variants thereof.)
784  */
785 void tracefs_remove(struct dentry *dentry)
786 {
787 	if (IS_ERR_OR_NULL(dentry))
788 		return;
789 
790 	simple_pin_fs(&trace_fs_type, &tracefs_mount, &tracefs_mount_count);
791 	simple_recursive_removal(dentry, remove_one);
792 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
793 }
794 
795 /**
796  * tracefs_initialized - Tells whether tracefs has been registered
797  */
798 bool tracefs_initialized(void)
799 {
800 	return tracefs_registered;
801 }
802 
803 static void init_once(void *foo)
804 {
805 	struct tracefs_inode *ti = (struct tracefs_inode *) foo;
806 
807 	/* inode_init_once() calls memset() on the vfs_inode portion */
808 	inode_init_once(&ti->vfs_inode);
809 
810 	/* Zero out the rest */
811 	memset_after(ti, 0, vfs_inode);
812 }
813 
814 static int __init tracefs_init(void)
815 {
816 	int retval;
817 
818 	tracefs_inode_cachep = kmem_cache_create("tracefs_inode_cache",
819 						 sizeof(struct tracefs_inode),
820 						 0, (SLAB_RECLAIM_ACCOUNT|
821 						     SLAB_MEM_SPREAD|
822 						     SLAB_ACCOUNT),
823 						 init_once);
824 	if (!tracefs_inode_cachep)
825 		return -ENOMEM;
826 
827 	retval = sysfs_create_mount_point(kernel_kobj, "tracing");
828 	if (retval)
829 		return -EINVAL;
830 
831 	retval = register_filesystem(&trace_fs_type);
832 	if (!retval)
833 		tracefs_registered = true;
834 
835 	return retval;
836 }
837 core_initcall(tracefs_init);
838