xref: /linux/fs/tracefs/inode.c (revision b7f84966b6f17626a8129723894dc315a076b391)
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 
tracefs_alloc_inode(struct super_block * sb)40 static struct inode *tracefs_alloc_inode(struct super_block *sb)
41 {
42 	struct tracefs_inode *ti;
43 	unsigned long flags;
44 
45 	ti = alloc_inode_sb(sb, 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 
tracefs_free_inode(struct inode * inode)56 static void tracefs_free_inode(struct inode *inode)
57 {
58 	struct tracefs_inode *ti = get_tracefs(inode);
59 
60 	kmem_cache_free(tracefs_inode_cachep, ti);
61 }
62 
tracefs_destroy_inode(struct inode * inode)63 static void tracefs_destroy_inode(struct inode *inode)
64 {
65 	struct tracefs_inode *ti = get_tracefs(inode);
66 	unsigned long flags;
67 
68 	spin_lock_irqsave(&tracefs_inode_lock, flags);
69 	list_del_rcu(&ti->list);
70 	spin_unlock_irqrestore(&tracefs_inode_lock, flags);
71 }
72 
default_read_file(struct file * file,char __user * buf,size_t count,loff_t * ppos)73 static ssize_t default_read_file(struct file *file, char __user *buf,
74 				 size_t count, loff_t *ppos)
75 {
76 	return 0;
77 }
78 
default_write_file(struct file * file,const char __user * buf,size_t count,loff_t * ppos)79 static ssize_t default_write_file(struct file *file, const char __user *buf,
80 				   size_t count, loff_t *ppos)
81 {
82 	return count;
83 }
84 
85 static const struct file_operations tracefs_file_operations = {
86 	.read =		default_read_file,
87 	.write =	default_write_file,
88 	.open =		simple_open,
89 	.llseek =	noop_llseek,
90 };
91 
92 static struct tracefs_dir_ops {
93 	int (*mkdir)(const char *name);
94 	int (*rmdir)(const char *name);
95 } tracefs_ops __ro_after_init;
96 
tracefs_syscall_mkdir(struct mnt_idmap * idmap,struct inode * inode,struct dentry * dentry,umode_t mode)97 static struct dentry *tracefs_syscall_mkdir(struct mnt_idmap *idmap,
98 					    struct inode *inode, struct dentry *dentry,
99 					    umode_t mode)
100 {
101 	struct tracefs_inode *ti;
102 	struct name_snapshot name;
103 	int ret;
104 
105 	/*
106 	 * This is a new directory that does not take the default of
107 	 * the rootfs. It becomes the default permissions for all the
108 	 * files and directories underneath it.
109 	 */
110 	ti = get_tracefs(inode);
111 	ti->flags |= TRACEFS_INSTANCE_INODE;
112 	ti->private = inode;
113 
114 	/*
115 	 * The mkdir call can call the generic functions that create
116 	 * the files within the tracefs system. It is up to the individual
117 	 * mkdir routine to handle races.
118 	 */
119 	take_dentry_name_snapshot(&name, dentry);
120 	inode_unlock(inode);
121 	ret = tracefs_ops.mkdir(name.name.name);
122 	inode_lock(inode);
123 	release_dentry_name_snapshot(&name);
124 
125 	return ERR_PTR(ret);
126 }
127 
tracefs_syscall_rmdir(struct inode * inode,struct dentry * dentry)128 static int tracefs_syscall_rmdir(struct inode *inode, struct dentry *dentry)
129 {
130 	struct name_snapshot name;
131 	int ret;
132 
133 	/*
134 	 * The rmdir call can call the generic functions that create
135 	 * the files within the tracefs system. It is up to the individual
136 	 * rmdir routine to handle races.
137 	 * This time we need to unlock not only the parent (inode) but
138 	 * also the directory that is being deleted.
139 	 */
140 	take_dentry_name_snapshot(&name, dentry);
141 	inode_unlock(inode);
142 	inode_unlock(d_inode(dentry));
143 
144 	ret = tracefs_ops.rmdir(name.name.name);
145 
146 	inode_lock_nested(inode, I_MUTEX_PARENT);
147 	inode_lock(d_inode(dentry));
148 	release_dentry_name_snapshot(&name);
149 
150 	return ret;
151 }
152 
set_tracefs_inode_owner(struct inode * inode)153 static void set_tracefs_inode_owner(struct inode *inode)
154 {
155 	struct tracefs_inode *ti = get_tracefs(inode);
156 	struct inode *root_inode = ti->private;
157 	kuid_t uid;
158 	kgid_t gid;
159 
160 	uid = root_inode->i_uid;
161 	gid = root_inode->i_gid;
162 
163 	/*
164 	 * If the root is not the mount point, then check the root's
165 	 * permissions. If it was never set, then default to the
166 	 * mount point.
167 	 */
168 	if (root_inode != d_inode(root_inode->i_sb->s_root)) {
169 		struct tracefs_inode *rti;
170 
171 		rti = get_tracefs(root_inode);
172 		root_inode = d_inode(root_inode->i_sb->s_root);
173 
174 		if (!(rti->flags & TRACEFS_UID_PERM_SET))
175 			uid = root_inode->i_uid;
176 
177 		if (!(rti->flags & TRACEFS_GID_PERM_SET))
178 			gid = root_inode->i_gid;
179 	}
180 
181 	/*
182 	 * If this inode has never been referenced, then update
183 	 * the permissions to the superblock.
184 	 */
185 	if (!(ti->flags & TRACEFS_UID_PERM_SET))
186 		inode->i_uid = uid;
187 
188 	if (!(ti->flags & TRACEFS_GID_PERM_SET))
189 		inode->i_gid = gid;
190 }
191 
tracefs_permission(struct mnt_idmap * idmap,struct inode * inode,int mask)192 static int tracefs_permission(struct mnt_idmap *idmap,
193 			      struct inode *inode, int mask)
194 {
195 	set_tracefs_inode_owner(inode);
196 	return generic_permission(idmap, inode, mask);
197 }
198 
tracefs_getattr(struct mnt_idmap * idmap,const struct path * path,struct kstat * stat,u32 request_mask,unsigned int flags)199 static int tracefs_getattr(struct mnt_idmap *idmap,
200 			   const struct path *path, struct kstat *stat,
201 			   u32 request_mask, unsigned int flags)
202 {
203 	struct inode *inode = d_backing_inode(path->dentry);
204 
205 	set_tracefs_inode_owner(inode);
206 	generic_fillattr(idmap, request_mask, inode, stat);
207 	return 0;
208 }
209 
tracefs_setattr(struct mnt_idmap * idmap,struct dentry * dentry,struct iattr * attr)210 static int tracefs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
211 			   struct iattr *attr)
212 {
213 	unsigned int ia_valid = attr->ia_valid;
214 	struct inode *inode = d_inode(dentry);
215 	struct tracefs_inode *ti = get_tracefs(inode);
216 
217 	if (ia_valid & ATTR_UID)
218 		ti->flags |= TRACEFS_UID_PERM_SET;
219 
220 	if (ia_valid & ATTR_GID)
221 		ti->flags |= TRACEFS_GID_PERM_SET;
222 
223 	return simple_setattr(idmap, dentry, attr);
224 }
225 
226 static const struct inode_operations tracefs_instance_dir_inode_operations = {
227 	.lookup		= simple_lookup,
228 	.mkdir		= tracefs_syscall_mkdir,
229 	.rmdir		= tracefs_syscall_rmdir,
230 	.permission	= tracefs_permission,
231 	.getattr	= tracefs_getattr,
232 	.setattr	= tracefs_setattr,
233 };
234 
235 static const struct inode_operations tracefs_dir_inode_operations = {
236 	.lookup		= simple_lookup,
237 	.permission	= tracefs_permission,
238 	.getattr	= tracefs_getattr,
239 	.setattr	= tracefs_setattr,
240 };
241 
242 static const struct inode_operations tracefs_file_inode_operations = {
243 	.permission	= tracefs_permission,
244 	.getattr	= tracefs_getattr,
245 	.setattr	= tracefs_setattr,
246 };
247 
tracefs_get_inode(struct super_block * sb)248 struct inode *tracefs_get_inode(struct super_block *sb)
249 {
250 	struct inode *inode = new_inode(sb);
251 	if (inode) {
252 		inode->i_ino = get_next_ino();
253 		simple_inode_init_ts(inode);
254 	}
255 	return inode;
256 }
257 
258 struct tracefs_fs_info {
259 	kuid_t uid;
260 	kgid_t gid;
261 	umode_t mode;
262 	/* Opt_* bitfield. */
263 	unsigned int opts;
264 };
265 
266 enum {
267 	Opt_uid,
268 	Opt_gid,
269 	Opt_mode,
270 };
271 
272 static const struct fs_parameter_spec tracefs_param_specs[] = {
273 	fsparam_gid	("gid",		Opt_gid),
274 	fsparam_u32oct	("mode",	Opt_mode),
275 	fsparam_uid	("uid",		Opt_uid),
276 	{}
277 };
278 
tracefs_parse_param(struct fs_context * fc,struct fs_parameter * param)279 static int tracefs_parse_param(struct fs_context *fc, struct fs_parameter *param)
280 {
281 	struct tracefs_fs_info *opts = fc->s_fs_info;
282 	struct fs_parse_result result;
283 	int opt;
284 
285 	opt = fs_parse(fc, tracefs_param_specs, param, &result);
286 	if (opt < 0)
287 		return opt;
288 
289 	switch (opt) {
290 	case Opt_uid:
291 		opts->uid = result.uid;
292 		break;
293 	case Opt_gid:
294 		opts->gid = result.gid;
295 		break;
296 	case Opt_mode:
297 		opts->mode = result.uint_32 & S_IALLUGO;
298 		break;
299 	/*
300 	 * We might like to report bad mount options here;
301 	 * but traditionally tracefs has ignored all mount options
302 	 */
303 	}
304 
305 	opts->opts |= BIT(opt);
306 
307 	return 0;
308 }
309 
tracefs_apply_options(struct super_block * sb,bool remount)310 static int tracefs_apply_options(struct super_block *sb, bool remount)
311 {
312 	struct tracefs_fs_info *fsi = sb->s_fs_info;
313 	struct inode *inode = d_inode(sb->s_root);
314 	struct tracefs_inode *ti;
315 	bool update_uid, update_gid;
316 	umode_t tmp_mode;
317 
318 	/*
319 	 * On remount, only reset mode/uid/gid if they were provided as mount
320 	 * options.
321 	 */
322 
323 	if (!remount || fsi->opts & BIT(Opt_mode)) {
324 		tmp_mode = READ_ONCE(inode->i_mode) & ~S_IALLUGO;
325 		tmp_mode |= fsi->mode;
326 		WRITE_ONCE(inode->i_mode, tmp_mode);
327 	}
328 
329 	if (!remount || fsi->opts & BIT(Opt_uid))
330 		inode->i_uid = fsi->uid;
331 
332 	if (!remount || fsi->opts & BIT(Opt_gid))
333 		inode->i_gid = fsi->gid;
334 
335 	if (remount && (fsi->opts & BIT(Opt_uid) || fsi->opts & BIT(Opt_gid))) {
336 
337 		update_uid = fsi->opts & BIT(Opt_uid);
338 		update_gid = fsi->opts & BIT(Opt_gid);
339 
340 		rcu_read_lock();
341 		list_for_each_entry_rcu(ti, &tracefs_inodes, list) {
342 			if (update_uid) {
343 				ti->flags &= ~TRACEFS_UID_PERM_SET;
344 				ti->vfs_inode.i_uid = fsi->uid;
345 			}
346 
347 			if (update_gid) {
348 				ti->flags &= ~TRACEFS_GID_PERM_SET;
349 				ti->vfs_inode.i_gid = fsi->gid;
350 			}
351 
352 			/*
353 			 * Note, the above ti->vfs_inode updates are
354 			 * used in eventfs_remount() so they must come
355 			 * before calling it.
356 			 */
357 			if (ti->flags & TRACEFS_EVENT_INODE)
358 				eventfs_remount(ti, update_uid, update_gid);
359 		}
360 		rcu_read_unlock();
361 	}
362 
363 	return 0;
364 }
365 
tracefs_reconfigure(struct fs_context * fc)366 static int tracefs_reconfigure(struct fs_context *fc)
367 {
368 	struct super_block *sb = fc->root->d_sb;
369 	struct tracefs_fs_info *sb_opts = sb->s_fs_info;
370 	struct tracefs_fs_info *new_opts = fc->s_fs_info;
371 
372 	if (!new_opts)
373 		return 0;
374 
375 	sync_filesystem(sb);
376 	/* structure copy of new mount options to sb */
377 	*sb_opts = *new_opts;
378 
379 	return tracefs_apply_options(sb, true);
380 }
381 
tracefs_show_options(struct seq_file * m,struct dentry * root)382 static int tracefs_show_options(struct seq_file *m, struct dentry *root)
383 {
384 	struct tracefs_fs_info *fsi = root->d_sb->s_fs_info;
385 
386 	if (!uid_eq(fsi->uid, GLOBAL_ROOT_UID))
387 		seq_printf(m, ",uid=%u",
388 			   from_kuid_munged(&init_user_ns, fsi->uid));
389 	if (!gid_eq(fsi->gid, GLOBAL_ROOT_GID))
390 		seq_printf(m, ",gid=%u",
391 			   from_kgid_munged(&init_user_ns, fsi->gid));
392 	if (fsi->mode != TRACEFS_DEFAULT_MODE)
393 		seq_printf(m, ",mode=%o", fsi->mode);
394 
395 	return 0;
396 }
397 
tracefs_drop_inode(struct inode * inode)398 static int tracefs_drop_inode(struct inode *inode)
399 {
400 	struct tracefs_inode *ti = get_tracefs(inode);
401 
402 	/*
403 	 * This inode is being freed and cannot be used for
404 	 * eventfs. Clear the flag so that it doesn't call into
405 	 * eventfs during the remount flag updates. The eventfs_inode
406 	 * gets freed after an RCU cycle, so the content will still
407 	 * be safe if the iteration is going on now.
408 	 */
409 	ti->flags &= ~TRACEFS_EVENT_INODE;
410 
411 	return 1;
412 }
413 
414 static const struct super_operations tracefs_super_operations = {
415 	.alloc_inode    = tracefs_alloc_inode,
416 	.free_inode     = tracefs_free_inode,
417 	.destroy_inode  = tracefs_destroy_inode,
418 	.drop_inode     = tracefs_drop_inode,
419 	.statfs		= simple_statfs,
420 	.show_options	= tracefs_show_options,
421 };
422 
423 /*
424  * It would be cleaner if eventfs had its own dentry ops.
425  *
426  * Note that d_revalidate is called potentially under RCU,
427  * so it can't take the eventfs mutex etc. It's fine - if
428  * we open a file just as it's marked dead, things will
429  * still work just fine, and just see the old stale case.
430  */
tracefs_d_release(struct dentry * dentry)431 static void tracefs_d_release(struct dentry *dentry)
432 {
433 	if (dentry->d_fsdata)
434 		eventfs_d_release(dentry);
435 }
436 
tracefs_d_revalidate(struct inode * inode,const struct qstr * name,struct dentry * dentry,unsigned int flags)437 static int tracefs_d_revalidate(struct inode *inode, const struct qstr *name,
438 				struct dentry *dentry, unsigned int flags)
439 {
440 	struct eventfs_inode *ei = dentry->d_fsdata;
441 
442 	return !(ei && ei->is_freed);
443 }
444 
tracefs_d_delete(const struct dentry * dentry)445 static int tracefs_d_delete(const struct dentry *dentry)
446 {
447 	/*
448 	 * We want to keep eventfs dentries around but not tracefs
449 	 * ones. eventfs dentries have content in d_fsdata.
450 	 * Use d_fsdata to determine if it's a eventfs dentry or not.
451 	 */
452 	return dentry->d_fsdata == NULL;
453 }
454 
455 static const struct dentry_operations tracefs_dentry_operations = {
456 	.d_revalidate = tracefs_d_revalidate,
457 	.d_release = tracefs_d_release,
458 	.d_delete = tracefs_d_delete,
459 };
460 
tracefs_fill_super(struct super_block * sb,struct fs_context * fc)461 static int tracefs_fill_super(struct super_block *sb, struct fs_context *fc)
462 {
463 	static const struct tree_descr trace_files[] = {{""}};
464 	int err;
465 
466 	err = simple_fill_super(sb, TRACEFS_MAGIC, trace_files);
467 	if (err)
468 		return err;
469 
470 	sb->s_op = &tracefs_super_operations;
471 	tracefs_apply_options(sb, false);
472 	set_default_d_op(sb, &tracefs_dentry_operations);
473 
474 	return 0;
475 }
476 
tracefs_get_tree(struct fs_context * fc)477 static int tracefs_get_tree(struct fs_context *fc)
478 {
479 	int err = get_tree_single(fc, tracefs_fill_super);
480 
481 	if (err)
482 		return err;
483 
484 	return tracefs_reconfigure(fc);
485 }
486 
tracefs_free_fc(struct fs_context * fc)487 static void tracefs_free_fc(struct fs_context *fc)
488 {
489 	kfree(fc->s_fs_info);
490 }
491 
492 static const struct fs_context_operations tracefs_context_ops = {
493 	.free		= tracefs_free_fc,
494 	.parse_param	= tracefs_parse_param,
495 	.get_tree	= tracefs_get_tree,
496 	.reconfigure	= tracefs_reconfigure,
497 };
498 
tracefs_init_fs_context(struct fs_context * fc)499 static int tracefs_init_fs_context(struct fs_context *fc)
500 {
501 	struct tracefs_fs_info *fsi;
502 
503 	fsi = kzalloc_obj(struct tracefs_fs_info);
504 	if (!fsi)
505 		return -ENOMEM;
506 
507 	fsi->mode = TRACEFS_DEFAULT_MODE;
508 
509 	fc->s_fs_info = fsi;
510 	fc->ops = &tracefs_context_ops;
511 	return 0;
512 }
513 
514 static struct file_system_type trace_fs_type = {
515 	.owner =	THIS_MODULE,
516 	.name =		"tracefs",
517 	.init_fs_context = tracefs_init_fs_context,
518 	.parameters	= tracefs_param_specs,
519 	.kill_sb =	kill_anon_super,
520 };
521 MODULE_ALIAS_FS("tracefs");
522 
tracefs_start_creating(const char * name,struct dentry * parent)523 struct dentry *tracefs_start_creating(const char *name, struct dentry *parent)
524 {
525 	struct dentry *dentry;
526 	int error;
527 
528 	pr_debug("tracefs: creating file '%s'\n",name);
529 
530 	error = simple_pin_fs(&trace_fs_type, &tracefs_mount,
531 			      &tracefs_mount_count);
532 	if (error)
533 		return ERR_PTR(error);
534 
535 	/* If the parent is not specified, we create it in the root.
536 	 * We need the root dentry to do this, which is in the super
537 	 * block. A pointer to that is in the struct vfsmount that we
538 	 * have around.
539 	 */
540 	if (!parent)
541 		parent = tracefs_mount->mnt_root;
542 
543 	dentry = simple_start_creating(parent, name);
544 	if (IS_ERR(dentry))
545 		simple_release_fs(&tracefs_mount, &tracefs_mount_count);
546 
547 	return dentry;
548 }
549 
tracefs_failed_creating(struct dentry * dentry)550 struct dentry *tracefs_failed_creating(struct dentry *dentry)
551 {
552 	simple_done_creating(dentry);
553 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
554 	return NULL;
555 }
556 
tracefs_end_creating(struct dentry * dentry)557 struct dentry *tracefs_end_creating(struct dentry *dentry)
558 {
559 	simple_done_creating(dentry);
560 	return dentry;	// borrowed
561 }
562 
563 /* Find the inode that this will use for default */
instance_inode(struct dentry * parent,struct inode * inode)564 static struct inode *instance_inode(struct dentry *parent, struct inode *inode)
565 {
566 	struct tracefs_inode *ti;
567 
568 	/* If parent is NULL then use root inode */
569 	if (!parent)
570 		return d_inode(inode->i_sb->s_root);
571 
572 	/* Find the inode that is flagged as an instance or the root inode */
573 	while (!IS_ROOT(parent)) {
574 		ti = get_tracefs(d_inode(parent));
575 		if (ti->flags & TRACEFS_INSTANCE_INODE)
576 			break;
577 		parent = parent->d_parent;
578 	}
579 
580 	return d_inode(parent);
581 }
582 
583 /**
584  * tracefs_create_file - create a file in the tracefs filesystem
585  * @name: a pointer to a string containing the name of the file to create.
586  * @mode: the permission that the file should have.
587  * @parent: a pointer to the parent dentry for this file.  This should be a
588  *          directory dentry if set.  If this parameter is NULL, then the
589  *          file will be created in the root of the tracefs filesystem.
590  * @data: a pointer to something that the caller will want to get to later
591  *        on.  The inode.i_private pointer will point to this value on
592  *        the open() call.
593  * @fops: a pointer to a struct file_operations that should be used for
594  *        this file.
595  *
596  * This is the basic "create a file" function for tracefs.  It allows for a
597  * wide range of flexibility in creating a file, or a directory (if you want
598  * to create a directory, the tracefs_create_dir() function is
599  * recommended to be used instead.)
600  *
601  * This function will return a pointer to a dentry if it succeeds.  This
602  * pointer must be passed to the tracefs_remove() function when the file is
603  * to be removed (no automatic cleanup happens if your module is unloaded,
604  * you are responsible here.)  If an error occurs, %NULL will be returned.
605  *
606  * If tracefs is not enabled in the kernel, the value -%ENODEV will be
607  * returned.
608  */
tracefs_create_file(const char * name,umode_t mode,struct dentry * parent,void * data,const struct file_operations * fops)609 struct dentry *tracefs_create_file(const char *name, umode_t mode,
610 				   struct dentry *parent, void *data,
611 				   const struct file_operations *fops)
612 {
613 	struct tracefs_inode *ti;
614 	struct dentry *dentry;
615 	struct inode *inode;
616 
617 	if (security_locked_down(LOCKDOWN_TRACEFS))
618 		return NULL;
619 
620 	if (!(mode & S_IFMT))
621 		mode |= S_IFREG;
622 	BUG_ON(!S_ISREG(mode));
623 	dentry = tracefs_start_creating(name, parent);
624 
625 	if (IS_ERR(dentry))
626 		return NULL;
627 
628 	inode = tracefs_get_inode(dentry->d_sb);
629 	if (unlikely(!inode))
630 		return tracefs_failed_creating(dentry);
631 
632 	ti = get_tracefs(inode);
633 	ti->private = instance_inode(parent, inode);
634 
635 	inode->i_mode = mode;
636 	inode->i_op = &tracefs_file_inode_operations;
637 	inode->i_fop = fops ? fops : &tracefs_file_operations;
638 	inode->i_private = data;
639 	inode->i_uid = d_inode(dentry->d_parent)->i_uid;
640 	inode->i_gid = d_inode(dentry->d_parent)->i_gid;
641 	d_make_persistent(dentry, inode);
642 	fsnotify_create(d_inode(dentry->d_parent), dentry);
643 	return tracefs_end_creating(dentry);
644 }
645 EXPORT_SYMBOL_GPL(tracefs_create_file);
646 
__create_dir(const char * name,struct dentry * parent,const struct inode_operations * ops)647 static struct dentry *__create_dir(const char *name, struct dentry *parent,
648 				   const struct inode_operations *ops)
649 {
650 	struct tracefs_inode *ti;
651 	struct dentry *dentry = tracefs_start_creating(name, parent);
652 	struct inode *inode;
653 
654 	if (IS_ERR(dentry))
655 		return NULL;
656 
657 	inode = tracefs_get_inode(dentry->d_sb);
658 	if (unlikely(!inode))
659 		return tracefs_failed_creating(dentry);
660 
661 	/* Do not set bits for OTH */
662 	inode->i_mode = S_IFDIR | S_IRWXU | S_IRUSR| S_IRGRP | S_IXUSR | S_IXGRP;
663 	inode->i_op = ops;
664 	inode->i_fop = &simple_dir_operations;
665 	inode->i_uid = d_inode(dentry->d_parent)->i_uid;
666 	inode->i_gid = d_inode(dentry->d_parent)->i_gid;
667 
668 	ti = get_tracefs(inode);
669 	ti->private = instance_inode(parent, inode);
670 
671 	/* directory inodes start off with i_nlink == 2 (for "." entry) */
672 	inc_nlink(inode);
673 	d_make_persistent(dentry, inode);
674 	inc_nlink(d_inode(dentry->d_parent));
675 	fsnotify_mkdir(d_inode(dentry->d_parent), dentry);
676 	return tracefs_end_creating(dentry);
677 }
678 
679 /**
680  * tracefs_create_dir - create a directory in the tracefs filesystem
681  * @name: a pointer to a string containing the name of the directory to
682  *        create.
683  * @parent: a pointer to the parent dentry for this file.  This should be a
684  *          directory dentry if set.  If this parameter is NULL, then the
685  *          directory will be created in the root of the tracefs filesystem.
686  *
687  * This function creates a directory in tracefs with the given name.
688  *
689  * This function will return a pointer to a dentry if it succeeds.  This
690  * pointer must be passed to the tracefs_remove() function when the file is
691  * to be removed. If an error occurs, %NULL will be returned.
692  *
693  * If tracing is not enabled in the kernel, the value -%ENODEV will be
694  * returned.
695  */
tracefs_create_dir(const char * name,struct dentry * parent)696 struct dentry *tracefs_create_dir(const char *name, struct dentry *parent)
697 {
698 	if (security_locked_down(LOCKDOWN_TRACEFS))
699 		return NULL;
700 
701 	return __create_dir(name, parent, &tracefs_dir_inode_operations);
702 }
703 
704 /**
705  * tracefs_create_instance_dir - create the tracing instances directory
706  * @name: The name of the instances directory to create
707  * @parent: The parent directory that the instances directory will exist
708  * @mkdir: The function to call when a mkdir is performed.
709  * @rmdir: The function to call when a rmdir is performed.
710  *
711  * Only one instances directory is allowed.
712  *
713  * The instances directory is special as it allows for mkdir and rmdir
714  * to be done by userspace. When a mkdir or rmdir is performed, the inode
715  * locks are released and the methods passed in (@mkdir and @rmdir) are
716  * called without locks and with the name of the directory being created
717  * within the instances directory.
718  *
719  * Returns the dentry of the instances directory.
720  */
tracefs_create_instance_dir(const char * name,struct dentry * parent,int (* mkdir)(const char * name),int (* rmdir)(const char * name))721 __init struct dentry *tracefs_create_instance_dir(const char *name,
722 					  struct dentry *parent,
723 					  int (*mkdir)(const char *name),
724 					  int (*rmdir)(const char *name))
725 {
726 	struct dentry *dentry;
727 
728 	/* Only allow one instance of the instances directory. */
729 	if (WARN_ON(tracefs_ops.mkdir || tracefs_ops.rmdir))
730 		return NULL;
731 
732 	dentry = __create_dir(name, parent, &tracefs_instance_dir_inode_operations);
733 	if (!dentry)
734 		return NULL;
735 
736 	tracefs_ops.mkdir = mkdir;
737 	tracefs_ops.rmdir = rmdir;
738 
739 	return dentry;
740 }
741 
remove_one(struct dentry * victim)742 static void remove_one(struct dentry *victim)
743 {
744 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
745 }
746 
747 /**
748  * tracefs_remove - recursively removes a directory
749  * @dentry: a pointer to a the dentry of the directory to be removed.
750  *
751  * This function recursively removes a directory tree in tracefs that
752  * was previously created with a call to another tracefs function
753  * (like tracefs_create_file() or variants thereof.)
754  */
tracefs_remove(struct dentry * dentry)755 void tracefs_remove(struct dentry *dentry)
756 {
757 	if (IS_ERR_OR_NULL(dentry))
758 		return;
759 
760 	simple_pin_fs(&trace_fs_type, &tracefs_mount, &tracefs_mount_count);
761 	simple_recursive_removal(dentry, remove_one);
762 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
763 }
764 
765 /**
766  * tracefs_initialized - Tells whether tracefs has been registered
767  */
tracefs_initialized(void)768 bool tracefs_initialized(void)
769 {
770 	return tracefs_registered;
771 }
772 
init_once(void * foo)773 static void init_once(void *foo)
774 {
775 	struct tracefs_inode *ti = (struct tracefs_inode *) foo;
776 
777 	/* inode_init_once() calls memset() on the vfs_inode portion */
778 	inode_init_once(&ti->vfs_inode);
779 
780 	/* Zero out the rest */
781 	memset_after(ti, 0, vfs_inode);
782 }
783 
tracefs_init(void)784 static int __init tracefs_init(void)
785 {
786 	int retval;
787 
788 	tracefs_inode_cachep = kmem_cache_create("tracefs_inode_cache",
789 						 sizeof(struct tracefs_inode),
790 						 0, (SLAB_RECLAIM_ACCOUNT|
791 						     SLAB_ACCOUNT),
792 						 init_once);
793 	if (!tracefs_inode_cachep)
794 		return -ENOMEM;
795 
796 	retval = sysfs_create_mount_point(kernel_kobj, "tracing");
797 	if (retval)
798 		return -EINVAL;
799 
800 	retval = register_filesystem(&trace_fs_type);
801 	if (!retval)
802 		tracefs_registered = true;
803 
804 	return retval;
805 }
806 core_initcall(tracefs_init);
807