xref: /linux/fs/tracefs/inode.c (revision 001821b0e79716c4e17c71d8e053a23599a7a508)
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_u32	("gid",		Opt_gid),
300 	fsparam_u32oct	("mode",	Opt_mode),
301 	fsparam_u32	("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 	kuid_t uid;
310 	kgid_t gid;
311 	int opt;
312 
313 	opt = fs_parse(fc, tracefs_param_specs, param, &result);
314 	if (opt < 0)
315 		return opt;
316 
317 	switch (opt) {
318 	case Opt_uid:
319 		uid = make_kuid(current_user_ns(), result.uint_32);
320 		if (!uid_valid(uid))
321 			return invalf(fc, "Unknown uid");
322 		opts->uid = uid;
323 		break;
324 	case Opt_gid:
325 		gid = make_kgid(current_user_ns(), result.uint_32);
326 		if (!gid_valid(gid))
327 			return invalf(fc, "Unknown gid");
328 		opts->gid = gid;
329 		break;
330 	case Opt_mode:
331 		opts->mode = result.uint_32 & S_IALLUGO;
332 		break;
333 	/*
334 	 * We might like to report bad mount options here;
335 	 * but traditionally tracefs has ignored all mount options
336 	 */
337 	}
338 
339 	opts->opts |= BIT(opt);
340 
341 	return 0;
342 }
343 
344 static int tracefs_apply_options(struct super_block *sb, bool remount)
345 {
346 	struct tracefs_fs_info *fsi = sb->s_fs_info;
347 	struct inode *inode = d_inode(sb->s_root);
348 	struct tracefs_inode *ti;
349 	bool update_uid, update_gid;
350 	umode_t tmp_mode;
351 
352 	/*
353 	 * On remount, only reset mode/uid/gid if they were provided as mount
354 	 * options.
355 	 */
356 
357 	if (!remount || fsi->opts & BIT(Opt_mode)) {
358 		tmp_mode = READ_ONCE(inode->i_mode) & ~S_IALLUGO;
359 		tmp_mode |= fsi->mode;
360 		WRITE_ONCE(inode->i_mode, tmp_mode);
361 	}
362 
363 	if (!remount || fsi->opts & BIT(Opt_uid))
364 		inode->i_uid = fsi->uid;
365 
366 	if (!remount || fsi->opts & BIT(Opt_gid))
367 		inode->i_gid = fsi->gid;
368 
369 	if (remount && (fsi->opts & BIT(Opt_uid) || fsi->opts & BIT(Opt_gid))) {
370 
371 		update_uid = fsi->opts & BIT(Opt_uid);
372 		update_gid = fsi->opts & BIT(Opt_gid);
373 
374 		rcu_read_lock();
375 		list_for_each_entry_rcu(ti, &tracefs_inodes, list) {
376 			if (update_uid) {
377 				ti->flags &= ~TRACEFS_UID_PERM_SET;
378 				ti->vfs_inode.i_uid = fsi->uid;
379 			}
380 
381 			if (update_gid) {
382 				ti->flags &= ~TRACEFS_GID_PERM_SET;
383 				ti->vfs_inode.i_gid = fsi->gid;
384 			}
385 
386 			/*
387 			 * Note, the above ti->vfs_inode updates are
388 			 * used in eventfs_remount() so they must come
389 			 * before calling it.
390 			 */
391 			if (ti->flags & TRACEFS_EVENT_INODE)
392 				eventfs_remount(ti, update_uid, update_gid);
393 		}
394 		rcu_read_unlock();
395 	}
396 
397 	return 0;
398 }
399 
400 static int tracefs_reconfigure(struct fs_context *fc)
401 {
402 	struct super_block *sb = fc->root->d_sb;
403 	struct tracefs_fs_info *sb_opts = sb->s_fs_info;
404 	struct tracefs_fs_info *new_opts = fc->s_fs_info;
405 
406 	sync_filesystem(sb);
407 	/* structure copy of new mount options to sb */
408 	*sb_opts = *new_opts;
409 
410 	return tracefs_apply_options(sb, true);
411 }
412 
413 static int tracefs_show_options(struct seq_file *m, struct dentry *root)
414 {
415 	struct tracefs_fs_info *fsi = root->d_sb->s_fs_info;
416 
417 	if (!uid_eq(fsi->uid, GLOBAL_ROOT_UID))
418 		seq_printf(m, ",uid=%u",
419 			   from_kuid_munged(&init_user_ns, fsi->uid));
420 	if (!gid_eq(fsi->gid, GLOBAL_ROOT_GID))
421 		seq_printf(m, ",gid=%u",
422 			   from_kgid_munged(&init_user_ns, fsi->gid));
423 	if (fsi->mode != TRACEFS_DEFAULT_MODE)
424 		seq_printf(m, ",mode=%o", fsi->mode);
425 
426 	return 0;
427 }
428 
429 static int tracefs_drop_inode(struct inode *inode)
430 {
431 	struct tracefs_inode *ti = get_tracefs(inode);
432 
433 	/*
434 	 * This inode is being freed and cannot be used for
435 	 * eventfs. Clear the flag so that it doesn't call into
436 	 * eventfs during the remount flag updates. The eventfs_inode
437 	 * gets freed after an RCU cycle, so the content will still
438 	 * be safe if the iteration is going on now.
439 	 */
440 	ti->flags &= ~TRACEFS_EVENT_INODE;
441 
442 	return 1;
443 }
444 
445 static const struct super_operations tracefs_super_operations = {
446 	.alloc_inode    = tracefs_alloc_inode,
447 	.free_inode     = tracefs_free_inode,
448 	.drop_inode     = tracefs_drop_inode,
449 	.statfs		= simple_statfs,
450 	.show_options	= tracefs_show_options,
451 };
452 
453 /*
454  * It would be cleaner if eventfs had its own dentry ops.
455  *
456  * Note that d_revalidate is called potentially under RCU,
457  * so it can't take the eventfs mutex etc. It's fine - if
458  * we open a file just as it's marked dead, things will
459  * still work just fine, and just see the old stale case.
460  */
461 static void tracefs_d_release(struct dentry *dentry)
462 {
463 	if (dentry->d_fsdata)
464 		eventfs_d_release(dentry);
465 }
466 
467 static int tracefs_d_revalidate(struct dentry *dentry, unsigned int flags)
468 {
469 	struct eventfs_inode *ei = dentry->d_fsdata;
470 
471 	return !(ei && ei->is_freed);
472 }
473 
474 static const struct dentry_operations tracefs_dentry_operations = {
475 	.d_revalidate = tracefs_d_revalidate,
476 	.d_release = tracefs_d_release,
477 };
478 
479 static int tracefs_fill_super(struct super_block *sb, struct fs_context *fc)
480 {
481 	static const struct tree_descr trace_files[] = {{""}};
482 	int err;
483 
484 	err = simple_fill_super(sb, TRACEFS_MAGIC, trace_files);
485 	if (err)
486 		return err;
487 
488 	sb->s_op = &tracefs_super_operations;
489 	sb->s_d_op = &tracefs_dentry_operations;
490 
491 	tracefs_apply_options(sb, false);
492 
493 	return 0;
494 }
495 
496 static int tracefs_get_tree(struct fs_context *fc)
497 {
498 	return get_tree_single(fc, tracefs_fill_super);
499 }
500 
501 static void tracefs_free_fc(struct fs_context *fc)
502 {
503 	kfree(fc->s_fs_info);
504 }
505 
506 static const struct fs_context_operations tracefs_context_ops = {
507 	.free		= tracefs_free_fc,
508 	.parse_param	= tracefs_parse_param,
509 	.get_tree	= tracefs_get_tree,
510 	.reconfigure	= tracefs_reconfigure,
511 };
512 
513 static int tracefs_init_fs_context(struct fs_context *fc)
514 {
515 	struct tracefs_fs_info *fsi;
516 
517 	fsi = kzalloc(sizeof(struct tracefs_fs_info), GFP_KERNEL);
518 	if (!fsi)
519 		return -ENOMEM;
520 
521 	fsi->mode = TRACEFS_DEFAULT_MODE;
522 
523 	fc->s_fs_info = fsi;
524 	fc->ops = &tracefs_context_ops;
525 	return 0;
526 }
527 
528 static struct file_system_type trace_fs_type = {
529 	.owner =	THIS_MODULE,
530 	.name =		"tracefs",
531 	.init_fs_context = tracefs_init_fs_context,
532 	.parameters	= tracefs_param_specs,
533 	.kill_sb =	kill_litter_super,
534 };
535 MODULE_ALIAS_FS("tracefs");
536 
537 struct dentry *tracefs_start_creating(const char *name, struct dentry *parent)
538 {
539 	struct dentry *dentry;
540 	int error;
541 
542 	pr_debug("tracefs: creating file '%s'\n",name);
543 
544 	error = simple_pin_fs(&trace_fs_type, &tracefs_mount,
545 			      &tracefs_mount_count);
546 	if (error)
547 		return ERR_PTR(error);
548 
549 	/* If the parent is not specified, we create it in the root.
550 	 * We need the root dentry to do this, which is in the super
551 	 * block. A pointer to that is in the struct vfsmount that we
552 	 * have around.
553 	 */
554 	if (!parent)
555 		parent = tracefs_mount->mnt_root;
556 
557 	inode_lock(d_inode(parent));
558 	if (unlikely(IS_DEADDIR(d_inode(parent))))
559 		dentry = ERR_PTR(-ENOENT);
560 	else
561 		dentry = lookup_one_len(name, parent, strlen(name));
562 	if (!IS_ERR(dentry) && d_inode(dentry)) {
563 		dput(dentry);
564 		dentry = ERR_PTR(-EEXIST);
565 	}
566 
567 	if (IS_ERR(dentry)) {
568 		inode_unlock(d_inode(parent));
569 		simple_release_fs(&tracefs_mount, &tracefs_mount_count);
570 	}
571 
572 	return dentry;
573 }
574 
575 struct dentry *tracefs_failed_creating(struct dentry *dentry)
576 {
577 	inode_unlock(d_inode(dentry->d_parent));
578 	dput(dentry);
579 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
580 	return NULL;
581 }
582 
583 struct dentry *tracefs_end_creating(struct dentry *dentry)
584 {
585 	inode_unlock(d_inode(dentry->d_parent));
586 	return dentry;
587 }
588 
589 /* Find the inode that this will use for default */
590 static struct inode *instance_inode(struct dentry *parent, struct inode *inode)
591 {
592 	struct tracefs_inode *ti;
593 
594 	/* If parent is NULL then use root inode */
595 	if (!parent)
596 		return d_inode(inode->i_sb->s_root);
597 
598 	/* Find the inode that is flagged as an instance or the root inode */
599 	while (!IS_ROOT(parent)) {
600 		ti = get_tracefs(d_inode(parent));
601 		if (ti->flags & TRACEFS_INSTANCE_INODE)
602 			break;
603 		parent = parent->d_parent;
604 	}
605 
606 	return d_inode(parent);
607 }
608 
609 /**
610  * tracefs_create_file - create a file in the tracefs filesystem
611  * @name: a pointer to a string containing the name of the file to create.
612  * @mode: the permission that the file should have.
613  * @parent: a pointer to the parent dentry for this file.  This should be a
614  *          directory dentry if set.  If this parameter is NULL, then the
615  *          file will be created in the root of the tracefs filesystem.
616  * @data: a pointer to something that the caller will want to get to later
617  *        on.  The inode.i_private pointer will point to this value on
618  *        the open() call.
619  * @fops: a pointer to a struct file_operations that should be used for
620  *        this file.
621  *
622  * This is the basic "create a file" function for tracefs.  It allows for a
623  * wide range of flexibility in creating a file, or a directory (if you want
624  * to create a directory, the tracefs_create_dir() function is
625  * recommended to be used instead.)
626  *
627  * This function will return a pointer to a dentry if it succeeds.  This
628  * pointer must be passed to the tracefs_remove() function when the file is
629  * to be removed (no automatic cleanup happens if your module is unloaded,
630  * you are responsible here.)  If an error occurs, %NULL will be returned.
631  *
632  * If tracefs is not enabled in the kernel, the value -%ENODEV will be
633  * returned.
634  */
635 struct dentry *tracefs_create_file(const char *name, umode_t mode,
636 				   struct dentry *parent, void *data,
637 				   const struct file_operations *fops)
638 {
639 	struct tracefs_inode *ti;
640 	struct dentry *dentry;
641 	struct inode *inode;
642 
643 	if (security_locked_down(LOCKDOWN_TRACEFS))
644 		return NULL;
645 
646 	if (!(mode & S_IFMT))
647 		mode |= S_IFREG;
648 	BUG_ON(!S_ISREG(mode));
649 	dentry = tracefs_start_creating(name, parent);
650 
651 	if (IS_ERR(dentry))
652 		return NULL;
653 
654 	inode = tracefs_get_inode(dentry->d_sb);
655 	if (unlikely(!inode))
656 		return tracefs_failed_creating(dentry);
657 
658 	ti = get_tracefs(inode);
659 	ti->private = instance_inode(parent, inode);
660 
661 	inode->i_mode = mode;
662 	inode->i_op = &tracefs_file_inode_operations;
663 	inode->i_fop = fops ? fops : &tracefs_file_operations;
664 	inode->i_private = data;
665 	inode->i_uid = d_inode(dentry->d_parent)->i_uid;
666 	inode->i_gid = d_inode(dentry->d_parent)->i_gid;
667 	d_instantiate(dentry, inode);
668 	fsnotify_create(d_inode(dentry->d_parent), dentry);
669 	return tracefs_end_creating(dentry);
670 }
671 
672 static struct dentry *__create_dir(const char *name, struct dentry *parent,
673 				   const struct inode_operations *ops)
674 {
675 	struct tracefs_inode *ti;
676 	struct dentry *dentry = tracefs_start_creating(name, parent);
677 	struct inode *inode;
678 
679 	if (IS_ERR(dentry))
680 		return NULL;
681 
682 	inode = tracefs_get_inode(dentry->d_sb);
683 	if (unlikely(!inode))
684 		return tracefs_failed_creating(dentry);
685 
686 	/* Do not set bits for OTH */
687 	inode->i_mode = S_IFDIR | S_IRWXU | S_IRUSR| S_IRGRP | S_IXUSR | S_IXGRP;
688 	inode->i_op = ops;
689 	inode->i_fop = &simple_dir_operations;
690 	inode->i_uid = d_inode(dentry->d_parent)->i_uid;
691 	inode->i_gid = d_inode(dentry->d_parent)->i_gid;
692 
693 	ti = get_tracefs(inode);
694 	ti->private = instance_inode(parent, inode);
695 
696 	/* directory inodes start off with i_nlink == 2 (for "." entry) */
697 	inc_nlink(inode);
698 	d_instantiate(dentry, inode);
699 	inc_nlink(d_inode(dentry->d_parent));
700 	fsnotify_mkdir(d_inode(dentry->d_parent), dentry);
701 	return tracefs_end_creating(dentry);
702 }
703 
704 /**
705  * tracefs_create_dir - create a directory in the tracefs filesystem
706  * @name: a pointer to a string containing the name of the directory to
707  *        create.
708  * @parent: a pointer to the parent dentry for this file.  This should be a
709  *          directory dentry if set.  If this parameter is NULL, then the
710  *          directory will be created in the root of the tracefs filesystem.
711  *
712  * This function creates a directory in tracefs with the given name.
713  *
714  * This function will return a pointer to a dentry if it succeeds.  This
715  * pointer must be passed to the tracefs_remove() function when the file is
716  * to be removed. If an error occurs, %NULL will be returned.
717  *
718  * If tracing is not enabled in the kernel, the value -%ENODEV will be
719  * returned.
720  */
721 struct dentry *tracefs_create_dir(const char *name, struct dentry *parent)
722 {
723 	if (security_locked_down(LOCKDOWN_TRACEFS))
724 		return NULL;
725 
726 	return __create_dir(name, parent, &tracefs_dir_inode_operations);
727 }
728 
729 /**
730  * tracefs_create_instance_dir - create the tracing instances directory
731  * @name: The name of the instances directory to create
732  * @parent: The parent directory that the instances directory will exist
733  * @mkdir: The function to call when a mkdir is performed.
734  * @rmdir: The function to call when a rmdir is performed.
735  *
736  * Only one instances directory is allowed.
737  *
738  * The instances directory is special as it allows for mkdir and rmdir
739  * to be done by userspace. When a mkdir or rmdir is performed, the inode
740  * locks are released and the methods passed in (@mkdir and @rmdir) are
741  * called without locks and with the name of the directory being created
742  * within the instances directory.
743  *
744  * Returns the dentry of the instances directory.
745  */
746 __init struct dentry *tracefs_create_instance_dir(const char *name,
747 					  struct dentry *parent,
748 					  int (*mkdir)(const char *name),
749 					  int (*rmdir)(const char *name))
750 {
751 	struct dentry *dentry;
752 
753 	/* Only allow one instance of the instances directory. */
754 	if (WARN_ON(tracefs_ops.mkdir || tracefs_ops.rmdir))
755 		return NULL;
756 
757 	dentry = __create_dir(name, parent, &tracefs_instance_dir_inode_operations);
758 	if (!dentry)
759 		return NULL;
760 
761 	tracefs_ops.mkdir = mkdir;
762 	tracefs_ops.rmdir = rmdir;
763 
764 	return dentry;
765 }
766 
767 static void remove_one(struct dentry *victim)
768 {
769 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
770 }
771 
772 /**
773  * tracefs_remove - recursively removes a directory
774  * @dentry: a pointer to a the dentry of the directory to be removed.
775  *
776  * This function recursively removes a directory tree in tracefs that
777  * was previously created with a call to another tracefs function
778  * (like tracefs_create_file() or variants thereof.)
779  */
780 void tracefs_remove(struct dentry *dentry)
781 {
782 	if (IS_ERR_OR_NULL(dentry))
783 		return;
784 
785 	simple_pin_fs(&trace_fs_type, &tracefs_mount, &tracefs_mount_count);
786 	simple_recursive_removal(dentry, remove_one);
787 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
788 }
789 
790 /**
791  * tracefs_initialized - Tells whether tracefs has been registered
792  */
793 bool tracefs_initialized(void)
794 {
795 	return tracefs_registered;
796 }
797 
798 static void init_once(void *foo)
799 {
800 	struct tracefs_inode *ti = (struct tracefs_inode *) foo;
801 
802 	/* inode_init_once() calls memset() on the vfs_inode portion */
803 	inode_init_once(&ti->vfs_inode);
804 
805 	/* Zero out the rest */
806 	memset_after(ti, 0, vfs_inode);
807 }
808 
809 static int __init tracefs_init(void)
810 {
811 	int retval;
812 
813 	tracefs_inode_cachep = kmem_cache_create("tracefs_inode_cache",
814 						 sizeof(struct tracefs_inode),
815 						 0, (SLAB_RECLAIM_ACCOUNT|
816 						     SLAB_ACCOUNT),
817 						 init_once);
818 	if (!tracefs_inode_cachep)
819 		return -ENOMEM;
820 
821 	retval = sysfs_create_mount_point(kernel_kobj, "tracing");
822 	if (retval)
823 		return -EINVAL;
824 
825 	retval = register_filesystem(&trace_fs_type);
826 	if (!retval)
827 		tracefs_registered = true;
828 
829 	return retval;
830 }
831 core_initcall(tracefs_init);
832