xref: /linux/fs/tracefs/inode.c (revision 2e68039281932e6dc37718a1ea7cbb8e2cda42e6)
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 	int srcu_idx;
317 	umode_t tmp_mode;
318 
319 	/*
320 	 * On remount, only reset mode/uid/gid if they were provided as mount
321 	 * options.
322 	 */
323 
324 	if (!remount || fsi->opts & BIT(Opt_mode)) {
325 		tmp_mode = READ_ONCE(inode->i_mode) & ~S_IALLUGO;
326 		tmp_mode |= fsi->mode;
327 		WRITE_ONCE(inode->i_mode, tmp_mode);
328 	}
329 
330 	if (!remount || fsi->opts & BIT(Opt_uid))
331 		inode->i_uid = fsi->uid;
332 
333 	if (!remount || fsi->opts & BIT(Opt_gid))
334 		inode->i_gid = fsi->gid;
335 
336 	if (remount && (fsi->opts & BIT(Opt_uid) || fsi->opts & BIT(Opt_gid))) {
337 
338 		update_uid = fsi->opts & BIT(Opt_uid);
339 		update_gid = fsi->opts & BIT(Opt_gid);
340 
341 		srcu_idx = eventfs_remount_lock();
342 		rcu_read_lock();
343 		list_for_each_entry_rcu(ti, &tracefs_inodes, list) {
344 			if (update_uid) {
345 				ti->flags &= ~TRACEFS_UID_PERM_SET;
346 				ti->vfs_inode.i_uid = fsi->uid;
347 			}
348 
349 			if (update_gid) {
350 				ti->flags &= ~TRACEFS_GID_PERM_SET;
351 				ti->vfs_inode.i_gid = fsi->gid;
352 			}
353 
354 			/*
355 			 * Note, the above ti->vfs_inode updates are
356 			 * used in eventfs_remount() so they must come
357 			 * before calling it.
358 			 */
359 			if (ti->flags & TRACEFS_EVENT_INODE)
360 				eventfs_remount(ti, update_uid, update_gid);
361 		}
362 		rcu_read_unlock();
363 		eventfs_remount_unlock(srcu_idx);
364 	}
365 
366 	return 0;
367 }
368 
tracefs_reconfigure(struct fs_context * fc)369 static int tracefs_reconfigure(struct fs_context *fc)
370 {
371 	struct super_block *sb = fc->root->d_sb;
372 	struct tracefs_fs_info *sb_opts = sb->s_fs_info;
373 	struct tracefs_fs_info *new_opts = fc->s_fs_info;
374 
375 	if (!new_opts)
376 		return 0;
377 
378 	sync_filesystem(sb);
379 	/* structure copy of new mount options to sb */
380 	*sb_opts = *new_opts;
381 
382 	return tracefs_apply_options(sb, true);
383 }
384 
tracefs_show_options(struct seq_file * m,struct dentry * root)385 static int tracefs_show_options(struct seq_file *m, struct dentry *root)
386 {
387 	struct tracefs_fs_info *fsi = root->d_sb->s_fs_info;
388 
389 	if (!uid_eq(fsi->uid, GLOBAL_ROOT_UID))
390 		seq_printf(m, ",uid=%u",
391 			   from_kuid_munged(&init_user_ns, fsi->uid));
392 	if (!gid_eq(fsi->gid, GLOBAL_ROOT_GID))
393 		seq_printf(m, ",gid=%u",
394 			   from_kgid_munged(&init_user_ns, fsi->gid));
395 	if (fsi->mode != TRACEFS_DEFAULT_MODE)
396 		seq_printf(m, ",mode=%o", fsi->mode);
397 
398 	return 0;
399 }
400 
tracefs_drop_inode(struct inode * inode)401 static int tracefs_drop_inode(struct inode *inode)
402 {
403 	struct tracefs_inode *ti = get_tracefs(inode);
404 
405 	/*
406 	 * This inode is being freed and cannot be used for
407 	 * eventfs. Clear the flag so that it doesn't call into
408 	 * eventfs during the remount flag updates. The eventfs_inode
409 	 * gets freed after an SRCU cycle, so the content will still
410 	 * be safe if the iteration is going on now.
411 	 */
412 	ti->flags &= ~TRACEFS_EVENT_INODE;
413 
414 	return 1;
415 }
416 
417 static const struct super_operations tracefs_super_operations = {
418 	.alloc_inode    = tracefs_alloc_inode,
419 	.free_inode     = tracefs_free_inode,
420 	.destroy_inode  = tracefs_destroy_inode,
421 	.drop_inode     = tracefs_drop_inode,
422 	.statfs		= simple_statfs,
423 	.show_options	= tracefs_show_options,
424 };
425 
426 /*
427  * It would be cleaner if eventfs had its own dentry ops.
428  *
429  * Note that d_revalidate is called potentially under RCU,
430  * so it can't take the eventfs mutex etc. It's fine - if
431  * we open a file just as it's marked dead, things will
432  * still work just fine, and just see the old stale case.
433  */
tracefs_d_release(struct dentry * dentry)434 static void tracefs_d_release(struct dentry *dentry)
435 {
436 	if (dentry->d_fsdata)
437 		eventfs_d_release(dentry);
438 }
439 
tracefs_d_revalidate(struct inode * inode,const struct qstr * name,struct dentry * dentry,unsigned int flags)440 static int tracefs_d_revalidate(struct inode *inode, const struct qstr *name,
441 				struct dentry *dentry, unsigned int flags)
442 {
443 	struct eventfs_inode *ei = dentry->d_fsdata;
444 
445 	return !(ei && ei->is_freed);
446 }
447 
tracefs_d_delete(const struct dentry * dentry)448 static int tracefs_d_delete(const struct dentry *dentry)
449 {
450 	/*
451 	 * We want to keep eventfs dentries around but not tracefs
452 	 * ones. eventfs dentries have content in d_fsdata.
453 	 * Use d_fsdata to determine if it's a eventfs dentry or not.
454 	 */
455 	return dentry->d_fsdata == NULL;
456 }
457 
458 static const struct dentry_operations tracefs_dentry_operations = {
459 	.d_revalidate = tracefs_d_revalidate,
460 	.d_release = tracefs_d_release,
461 	.d_delete = tracefs_d_delete,
462 };
463 
tracefs_fill_super(struct super_block * sb,struct fs_context * fc)464 static int tracefs_fill_super(struct super_block *sb, struct fs_context *fc)
465 {
466 	static const struct tree_descr trace_files[] = {{""}};
467 	int err;
468 
469 	err = simple_fill_super(sb, TRACEFS_MAGIC, trace_files);
470 	if (err)
471 		return err;
472 
473 	sb->s_op = &tracefs_super_operations;
474 	tracefs_apply_options(sb, false);
475 	set_default_d_op(sb, &tracefs_dentry_operations);
476 
477 	return 0;
478 }
479 
tracefs_get_tree(struct fs_context * fc)480 static int tracefs_get_tree(struct fs_context *fc)
481 {
482 	int err = get_tree_single(fc, tracefs_fill_super);
483 
484 	if (err)
485 		return err;
486 
487 	return tracefs_reconfigure(fc);
488 }
489 
tracefs_free_fc(struct fs_context * fc)490 static void tracefs_free_fc(struct fs_context *fc)
491 {
492 	kfree(fc->s_fs_info);
493 }
494 
495 static const struct fs_context_operations tracefs_context_ops = {
496 	.free		= tracefs_free_fc,
497 	.parse_param	= tracefs_parse_param,
498 	.get_tree	= tracefs_get_tree,
499 	.reconfigure	= tracefs_reconfigure,
500 };
501 
tracefs_init_fs_context(struct fs_context * fc)502 static int tracefs_init_fs_context(struct fs_context *fc)
503 {
504 	struct tracefs_fs_info *fsi;
505 
506 	fsi = kzalloc_obj(struct tracefs_fs_info);
507 	if (!fsi)
508 		return -ENOMEM;
509 
510 	fsi->mode = TRACEFS_DEFAULT_MODE;
511 
512 	fc->s_fs_info = fsi;
513 	fc->ops = &tracefs_context_ops;
514 	return 0;
515 }
516 
517 static struct file_system_type trace_fs_type = {
518 	.owner =	THIS_MODULE,
519 	.name =		"tracefs",
520 	.init_fs_context = tracefs_init_fs_context,
521 	.parameters	= tracefs_param_specs,
522 	.kill_sb =	kill_anon_super,
523 };
524 MODULE_ALIAS_FS("tracefs");
525 
tracefs_start_creating(const char * name,struct dentry * parent)526 struct dentry *tracefs_start_creating(const char *name, struct dentry *parent)
527 {
528 	struct dentry *dentry;
529 	int error;
530 
531 	pr_debug("tracefs: creating file '%s'\n",name);
532 
533 	error = simple_pin_fs(&trace_fs_type, &tracefs_mount,
534 			      &tracefs_mount_count);
535 	if (error)
536 		return ERR_PTR(error);
537 
538 	/* If the parent is not specified, we create it in the root.
539 	 * We need the root dentry to do this, which is in the super
540 	 * block. A pointer to that is in the struct vfsmount that we
541 	 * have around.
542 	 */
543 	if (!parent)
544 		parent = tracefs_mount->mnt_root;
545 
546 	dentry = simple_start_creating(parent, name);
547 	if (IS_ERR(dentry))
548 		simple_release_fs(&tracefs_mount, &tracefs_mount_count);
549 
550 	return dentry;
551 }
552 
tracefs_failed_creating(struct dentry * dentry)553 struct dentry *tracefs_failed_creating(struct dentry *dentry)
554 {
555 	simple_done_creating(dentry);
556 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
557 	return NULL;
558 }
559 
tracefs_end_creating(struct dentry * dentry)560 struct dentry *tracefs_end_creating(struct dentry *dentry)
561 {
562 	simple_done_creating(dentry);
563 	return dentry;	// borrowed
564 }
565 
566 /* Find the inode that this will use for default */
instance_inode(struct dentry * parent,struct inode * inode)567 static struct inode *instance_inode(struct dentry *parent, struct inode *inode)
568 {
569 	struct tracefs_inode *ti;
570 
571 	/* If parent is NULL then use root inode */
572 	if (!parent)
573 		return d_inode(inode->i_sb->s_root);
574 
575 	/* Find the inode that is flagged as an instance or the root inode */
576 	while (!IS_ROOT(parent)) {
577 		ti = get_tracefs(d_inode(parent));
578 		if (ti->flags & TRACEFS_INSTANCE_INODE)
579 			break;
580 		parent = parent->d_parent;
581 	}
582 
583 	return d_inode(parent);
584 }
585 
586 /**
587  * tracefs_create_file - create a file in the tracefs filesystem
588  * @name: a pointer to a string containing the name of the file to create.
589  * @mode: the permission that the file should have.
590  * @parent: a pointer to the parent dentry for this file.  This should be a
591  *          directory dentry if set.  If this parameter is NULL, then the
592  *          file will be created in the root of the tracefs filesystem.
593  * @data: a pointer to something that the caller will want to get to later
594  *        on.  The inode.i_private pointer will point to this value on
595  *        the open() call.
596  * @fops: a pointer to a struct file_operations that should be used for
597  *        this file.
598  *
599  * This is the basic "create a file" function for tracefs.  It allows for a
600  * wide range of flexibility in creating a file, or a directory (if you want
601  * to create a directory, the tracefs_create_dir() function is
602  * recommended to be used instead.)
603  *
604  * This function will return a pointer to a dentry if it succeeds.  This
605  * pointer must be passed to the tracefs_remove() function when the file is
606  * to be removed (no automatic cleanup happens if your module is unloaded,
607  * you are responsible here.)  If an error occurs, %NULL will be returned.
608  *
609  * If tracefs is not enabled in the kernel, the value -%ENODEV will be
610  * returned.
611  */
tracefs_create_file(const char * name,umode_t mode,struct dentry * parent,void * data,const struct file_operations * fops)612 struct dentry *tracefs_create_file(const char *name, umode_t mode,
613 				   struct dentry *parent, void *data,
614 				   const struct file_operations *fops)
615 {
616 	struct tracefs_inode *ti;
617 	struct dentry *dentry;
618 	struct inode *inode;
619 
620 	if (security_locked_down(LOCKDOWN_TRACEFS))
621 		return NULL;
622 
623 	if (!(mode & S_IFMT))
624 		mode |= S_IFREG;
625 	BUG_ON(!S_ISREG(mode));
626 	dentry = tracefs_start_creating(name, parent);
627 
628 	if (IS_ERR(dentry))
629 		return NULL;
630 
631 	inode = tracefs_get_inode(dentry->d_sb);
632 	if (unlikely(!inode))
633 		return tracefs_failed_creating(dentry);
634 
635 	ti = get_tracefs(inode);
636 	ti->private = instance_inode(parent, inode);
637 
638 	inode->i_mode = mode;
639 	inode->i_op = &tracefs_file_inode_operations;
640 	inode->i_fop = fops ? fops : &tracefs_file_operations;
641 	inode->i_private = data;
642 	inode->i_uid = d_inode(dentry->d_parent)->i_uid;
643 	inode->i_gid = d_inode(dentry->d_parent)->i_gid;
644 	d_make_persistent(dentry, inode);
645 	fsnotify_create(d_inode(dentry->d_parent), dentry);
646 	return tracefs_end_creating(dentry);
647 }
648 EXPORT_SYMBOL_GPL(tracefs_create_file);
649 
__create_dir(const char * name,struct dentry * parent,const struct inode_operations * ops)650 static struct dentry *__create_dir(const char *name, struct dentry *parent,
651 				   const struct inode_operations *ops)
652 {
653 	struct tracefs_inode *ti;
654 	struct dentry *dentry = tracefs_start_creating(name, parent);
655 	struct inode *inode;
656 
657 	if (IS_ERR(dentry))
658 		return NULL;
659 
660 	inode = tracefs_get_inode(dentry->d_sb);
661 	if (unlikely(!inode))
662 		return tracefs_failed_creating(dentry);
663 
664 	/* Do not set bits for OTH */
665 	inode->i_mode = S_IFDIR | S_IRWXU | S_IRUSR| S_IRGRP | S_IXUSR | S_IXGRP;
666 	inode->i_op = ops;
667 	inode->i_fop = &simple_dir_operations;
668 	inode->i_uid = d_inode(dentry->d_parent)->i_uid;
669 	inode->i_gid = d_inode(dentry->d_parent)->i_gid;
670 
671 	ti = get_tracefs(inode);
672 	ti->private = instance_inode(parent, inode);
673 
674 	/* directory inodes start off with i_nlink == 2 (for "." entry) */
675 	inc_nlink(inode);
676 	d_make_persistent(dentry, inode);
677 	inc_nlink(d_inode(dentry->d_parent));
678 	fsnotify_mkdir(d_inode(dentry->d_parent), dentry);
679 	return tracefs_end_creating(dentry);
680 }
681 
682 /**
683  * tracefs_create_dir - create a directory in the tracefs filesystem
684  * @name: a pointer to a string containing the name of the directory to
685  *        create.
686  * @parent: a pointer to the parent dentry for this file.  This should be a
687  *          directory dentry if set.  If this parameter is NULL, then the
688  *          directory will be created in the root of the tracefs filesystem.
689  *
690  * This function creates a directory in tracefs with the given name.
691  *
692  * This function will return a pointer to a dentry if it succeeds.  This
693  * pointer must be passed to the tracefs_remove() function when the file is
694  * to be removed. If an error occurs, %NULL will be returned.
695  *
696  * If tracing is not enabled in the kernel, the value -%ENODEV will be
697  * returned.
698  */
tracefs_create_dir(const char * name,struct dentry * parent)699 struct dentry *tracefs_create_dir(const char *name, struct dentry *parent)
700 {
701 	if (security_locked_down(LOCKDOWN_TRACEFS))
702 		return NULL;
703 
704 	return __create_dir(name, parent, &tracefs_dir_inode_operations);
705 }
706 
707 /**
708  * tracefs_create_instance_dir - create the tracing instances directory
709  * @name: The name of the instances directory to create
710  * @parent: The parent directory that the instances directory will exist
711  * @mkdir: The function to call when a mkdir is performed.
712  * @rmdir: The function to call when a rmdir is performed.
713  *
714  * Only one instances directory is allowed.
715  *
716  * The instances directory is special as it allows for mkdir and rmdir
717  * to be done by userspace. When a mkdir or rmdir is performed, the inode
718  * locks are released and the methods passed in (@mkdir and @rmdir) are
719  * called without locks and with the name of the directory being created
720  * within the instances directory.
721  *
722  * Returns the dentry of the instances directory.
723  */
tracefs_create_instance_dir(const char * name,struct dentry * parent,int (* mkdir)(const char * name),int (* rmdir)(const char * name))724 __init struct dentry *tracefs_create_instance_dir(const char *name,
725 					  struct dentry *parent,
726 					  int (*mkdir)(const char *name),
727 					  int (*rmdir)(const char *name))
728 {
729 	struct dentry *dentry;
730 
731 	/* Only allow one instance of the instances directory. */
732 	if (WARN_ON(tracefs_ops.mkdir || tracefs_ops.rmdir))
733 		return NULL;
734 
735 	dentry = __create_dir(name, parent, &tracefs_instance_dir_inode_operations);
736 	if (!dentry)
737 		return NULL;
738 
739 	tracefs_ops.mkdir = mkdir;
740 	tracefs_ops.rmdir = rmdir;
741 
742 	return dentry;
743 }
744 
remove_one(struct dentry * victim)745 static void remove_one(struct dentry *victim)
746 {
747 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
748 }
749 
750 /**
751  * tracefs_remove - recursively removes a directory
752  * @dentry: a pointer to a the dentry of the directory to be removed.
753  *
754  * This function recursively removes a directory tree in tracefs that
755  * was previously created with a call to another tracefs function
756  * (like tracefs_create_file() or variants thereof.)
757  */
tracefs_remove(struct dentry * dentry)758 void tracefs_remove(struct dentry *dentry)
759 {
760 	if (IS_ERR_OR_NULL(dentry))
761 		return;
762 
763 	simple_pin_fs(&trace_fs_type, &tracefs_mount, &tracefs_mount_count);
764 	simple_recursive_removal(dentry, remove_one);
765 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
766 }
767 
768 /**
769  * tracefs_initialized - Tells whether tracefs has been registered
770  */
tracefs_initialized(void)771 bool tracefs_initialized(void)
772 {
773 	return tracefs_registered;
774 }
775 
init_once(void * foo)776 static void init_once(void *foo)
777 {
778 	struct tracefs_inode *ti = (struct tracefs_inode *) foo;
779 
780 	/* inode_init_once() calls memset() on the vfs_inode portion */
781 	inode_init_once(&ti->vfs_inode);
782 
783 	/* Zero out the rest */
784 	memset_after(ti, 0, vfs_inode);
785 }
786 
tracefs_init(void)787 static int __init tracefs_init(void)
788 {
789 	int retval;
790 
791 	tracefs_inode_cachep = kmem_cache_create("tracefs_inode_cache",
792 						 sizeof(struct tracefs_inode),
793 						 0, (SLAB_RECLAIM_ACCOUNT|
794 						     SLAB_ACCOUNT),
795 						 init_once);
796 	if (!tracefs_inode_cachep)
797 		return -ENOMEM;
798 
799 	retval = sysfs_create_mount_point(kernel_kobj, "tracing");
800 	if (retval)
801 		return -EINVAL;
802 
803 	retval = register_filesystem(&trace_fs_type);
804 	if (!retval)
805 		tracefs_registered = true;
806 
807 	return retval;
808 }
809 core_initcall(tracefs_init);
810