xref: /linux/fs/tracefs/inode.c (revision 9466b6ae6befa87b27cc13ee7263e01b9c48efc0)
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 
get_dname(struct dentry * dentry)97 static char *get_dname(struct dentry *dentry)
98 {
99 	const char *dname;
100 	char *name;
101 	int len = dentry->d_name.len;
102 
103 	dname = dentry->d_name.name;
104 	name = kmalloc(len + 1, GFP_KERNEL);
105 	if (!name)
106 		return NULL;
107 	memcpy(name, dname, len);
108 	name[len] = 0;
109 	return name;
110 }
111 
tracefs_syscall_mkdir(struct mnt_idmap * idmap,struct inode * inode,struct dentry * dentry,umode_t mode)112 static int tracefs_syscall_mkdir(struct mnt_idmap *idmap,
113 				 struct inode *inode, struct dentry *dentry,
114 				 umode_t mode)
115 {
116 	struct tracefs_inode *ti;
117 	char *name;
118 	int ret;
119 
120 	name = get_dname(dentry);
121 	if (!name)
122 		return -ENOMEM;
123 
124 	/*
125 	 * This is a new directory that does not take the default of
126 	 * the rootfs. It becomes the default permissions for all the
127 	 * files and directories underneath it.
128 	 */
129 	ti = get_tracefs(inode);
130 	ti->flags |= TRACEFS_INSTANCE_INODE;
131 	ti->private = inode;
132 
133 	/*
134 	 * The mkdir call can call the generic functions that create
135 	 * the files within the tracefs system. It is up to the individual
136 	 * mkdir routine to handle races.
137 	 */
138 	inode_unlock(inode);
139 	ret = tracefs_ops.mkdir(name);
140 	inode_lock(inode);
141 
142 	kfree(name);
143 
144 	return ret;
145 }
146 
tracefs_syscall_rmdir(struct inode * inode,struct dentry * dentry)147 static int tracefs_syscall_rmdir(struct inode *inode, struct dentry *dentry)
148 {
149 	char *name;
150 	int ret;
151 
152 	name = get_dname(dentry);
153 	if (!name)
154 		return -ENOMEM;
155 
156 	/*
157 	 * The rmdir call can call the generic functions that create
158 	 * the files within the tracefs system. It is up to the individual
159 	 * rmdir routine to handle races.
160 	 * This time we need to unlock not only the parent (inode) but
161 	 * also the directory that is being deleted.
162 	 */
163 	inode_unlock(inode);
164 	inode_unlock(d_inode(dentry));
165 
166 	ret = tracefs_ops.rmdir(name);
167 
168 	inode_lock_nested(inode, I_MUTEX_PARENT);
169 	inode_lock(d_inode(dentry));
170 
171 	kfree(name);
172 
173 	return ret;
174 }
175 
set_tracefs_inode_owner(struct inode * inode)176 static void set_tracefs_inode_owner(struct inode *inode)
177 {
178 	struct tracefs_inode *ti = get_tracefs(inode);
179 	struct inode *root_inode = ti->private;
180 	kuid_t uid;
181 	kgid_t gid;
182 
183 	uid = root_inode->i_uid;
184 	gid = root_inode->i_gid;
185 
186 	/*
187 	 * If the root is not the mount point, then check the root's
188 	 * permissions. If it was never set, then default to the
189 	 * mount point.
190 	 */
191 	if (root_inode != d_inode(root_inode->i_sb->s_root)) {
192 		struct tracefs_inode *rti;
193 
194 		rti = get_tracefs(root_inode);
195 		root_inode = d_inode(root_inode->i_sb->s_root);
196 
197 		if (!(rti->flags & TRACEFS_UID_PERM_SET))
198 			uid = root_inode->i_uid;
199 
200 		if (!(rti->flags & TRACEFS_GID_PERM_SET))
201 			gid = root_inode->i_gid;
202 	}
203 
204 	/*
205 	 * If this inode has never been referenced, then update
206 	 * the permissions to the superblock.
207 	 */
208 	if (!(ti->flags & TRACEFS_UID_PERM_SET))
209 		inode->i_uid = uid;
210 
211 	if (!(ti->flags & TRACEFS_GID_PERM_SET))
212 		inode->i_gid = gid;
213 }
214 
tracefs_permission(struct mnt_idmap * idmap,struct inode * inode,int mask)215 static int tracefs_permission(struct mnt_idmap *idmap,
216 			      struct inode *inode, int mask)
217 {
218 	set_tracefs_inode_owner(inode);
219 	return generic_permission(idmap, inode, mask);
220 }
221 
tracefs_getattr(struct mnt_idmap * idmap,const struct path * path,struct kstat * stat,u32 request_mask,unsigned int flags)222 static int tracefs_getattr(struct mnt_idmap *idmap,
223 			   const struct path *path, struct kstat *stat,
224 			   u32 request_mask, unsigned int flags)
225 {
226 	struct inode *inode = d_backing_inode(path->dentry);
227 
228 	set_tracefs_inode_owner(inode);
229 	generic_fillattr(idmap, request_mask, inode, stat);
230 	return 0;
231 }
232 
tracefs_setattr(struct mnt_idmap * idmap,struct dentry * dentry,struct iattr * attr)233 static int tracefs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
234 			   struct iattr *attr)
235 {
236 	unsigned int ia_valid = attr->ia_valid;
237 	struct inode *inode = d_inode(dentry);
238 	struct tracefs_inode *ti = get_tracefs(inode);
239 
240 	if (ia_valid & ATTR_UID)
241 		ti->flags |= TRACEFS_UID_PERM_SET;
242 
243 	if (ia_valid & ATTR_GID)
244 		ti->flags |= TRACEFS_GID_PERM_SET;
245 
246 	return simple_setattr(idmap, dentry, attr);
247 }
248 
249 static const struct inode_operations tracefs_instance_dir_inode_operations = {
250 	.lookup		= simple_lookup,
251 	.mkdir		= tracefs_syscall_mkdir,
252 	.rmdir		= tracefs_syscall_rmdir,
253 	.permission	= tracefs_permission,
254 	.getattr	= tracefs_getattr,
255 	.setattr	= tracefs_setattr,
256 };
257 
258 static const struct inode_operations tracefs_dir_inode_operations = {
259 	.lookup		= simple_lookup,
260 	.permission	= tracefs_permission,
261 	.getattr	= tracefs_getattr,
262 	.setattr	= tracefs_setattr,
263 };
264 
265 static const struct inode_operations tracefs_file_inode_operations = {
266 	.permission	= tracefs_permission,
267 	.getattr	= tracefs_getattr,
268 	.setattr	= tracefs_setattr,
269 };
270 
tracefs_get_inode(struct super_block * sb)271 struct inode *tracefs_get_inode(struct super_block *sb)
272 {
273 	struct inode *inode = new_inode(sb);
274 	if (inode) {
275 		inode->i_ino = get_next_ino();
276 		simple_inode_init_ts(inode);
277 	}
278 	return inode;
279 }
280 
281 struct tracefs_fs_info {
282 	kuid_t uid;
283 	kgid_t gid;
284 	umode_t mode;
285 	/* Opt_* bitfield. */
286 	unsigned int opts;
287 };
288 
289 enum {
290 	Opt_uid,
291 	Opt_gid,
292 	Opt_mode,
293 };
294 
295 static const struct fs_parameter_spec tracefs_param_specs[] = {
296 	fsparam_gid	("gid",		Opt_gid),
297 	fsparam_u32oct	("mode",	Opt_mode),
298 	fsparam_uid	("uid",		Opt_uid),
299 	{}
300 };
301 
tracefs_parse_param(struct fs_context * fc,struct fs_parameter * param)302 static int tracefs_parse_param(struct fs_context *fc, struct fs_parameter *param)
303 {
304 	struct tracefs_fs_info *opts = fc->s_fs_info;
305 	struct fs_parse_result result;
306 	int opt;
307 
308 	opt = fs_parse(fc, tracefs_param_specs, param, &result);
309 	if (opt < 0)
310 		return opt;
311 
312 	switch (opt) {
313 	case Opt_uid:
314 		opts->uid = result.uid;
315 		break;
316 	case Opt_gid:
317 		opts->gid = result.gid;
318 		break;
319 	case Opt_mode:
320 		opts->mode = result.uint_32 & S_IALLUGO;
321 		break;
322 	/*
323 	 * We might like to report bad mount options here;
324 	 * but traditionally tracefs has ignored all mount options
325 	 */
326 	}
327 
328 	opts->opts |= BIT(opt);
329 
330 	return 0;
331 }
332 
tracefs_apply_options(struct super_block * sb,bool remount)333 static int tracefs_apply_options(struct super_block *sb, bool remount)
334 {
335 	struct tracefs_fs_info *fsi = sb->s_fs_info;
336 	struct inode *inode = d_inode(sb->s_root);
337 	struct tracefs_inode *ti;
338 	bool update_uid, update_gid;
339 	umode_t tmp_mode;
340 
341 	/*
342 	 * On remount, only reset mode/uid/gid if they were provided as mount
343 	 * options.
344 	 */
345 
346 	if (!remount || fsi->opts & BIT(Opt_mode)) {
347 		tmp_mode = READ_ONCE(inode->i_mode) & ~S_IALLUGO;
348 		tmp_mode |= fsi->mode;
349 		WRITE_ONCE(inode->i_mode, tmp_mode);
350 	}
351 
352 	if (!remount || fsi->opts & BIT(Opt_uid))
353 		inode->i_uid = fsi->uid;
354 
355 	if (!remount || fsi->opts & BIT(Opt_gid))
356 		inode->i_gid = fsi->gid;
357 
358 	if (remount && (fsi->opts & BIT(Opt_uid) || fsi->opts & BIT(Opt_gid))) {
359 
360 		update_uid = fsi->opts & BIT(Opt_uid);
361 		update_gid = fsi->opts & BIT(Opt_gid);
362 
363 		rcu_read_lock();
364 		list_for_each_entry_rcu(ti, &tracefs_inodes, list) {
365 			if (update_uid) {
366 				ti->flags &= ~TRACEFS_UID_PERM_SET;
367 				ti->vfs_inode.i_uid = fsi->uid;
368 			}
369 
370 			if (update_gid) {
371 				ti->flags &= ~TRACEFS_GID_PERM_SET;
372 				ti->vfs_inode.i_gid = fsi->gid;
373 			}
374 
375 			/*
376 			 * Note, the above ti->vfs_inode updates are
377 			 * used in eventfs_remount() so they must come
378 			 * before calling it.
379 			 */
380 			if (ti->flags & TRACEFS_EVENT_INODE)
381 				eventfs_remount(ti, update_uid, update_gid);
382 		}
383 		rcu_read_unlock();
384 	}
385 
386 	return 0;
387 }
388 
tracefs_reconfigure(struct fs_context * fc)389 static int tracefs_reconfigure(struct fs_context *fc)
390 {
391 	struct super_block *sb = fc->root->d_sb;
392 	struct tracefs_fs_info *sb_opts = sb->s_fs_info;
393 	struct tracefs_fs_info *new_opts = fc->s_fs_info;
394 
395 	sync_filesystem(sb);
396 	/* structure copy of new mount options to sb */
397 	*sb_opts = *new_opts;
398 
399 	return tracefs_apply_options(sb, true);
400 }
401 
tracefs_show_options(struct seq_file * m,struct dentry * root)402 static int tracefs_show_options(struct seq_file *m, struct dentry *root)
403 {
404 	struct tracefs_fs_info *fsi = root->d_sb->s_fs_info;
405 
406 	if (!uid_eq(fsi->uid, GLOBAL_ROOT_UID))
407 		seq_printf(m, ",uid=%u",
408 			   from_kuid_munged(&init_user_ns, fsi->uid));
409 	if (!gid_eq(fsi->gid, GLOBAL_ROOT_GID))
410 		seq_printf(m, ",gid=%u",
411 			   from_kgid_munged(&init_user_ns, fsi->gid));
412 	if (fsi->mode != TRACEFS_DEFAULT_MODE)
413 		seq_printf(m, ",mode=%o", fsi->mode);
414 
415 	return 0;
416 }
417 
tracefs_drop_inode(struct inode * inode)418 static int tracefs_drop_inode(struct inode *inode)
419 {
420 	struct tracefs_inode *ti = get_tracefs(inode);
421 
422 	/*
423 	 * This inode is being freed and cannot be used for
424 	 * eventfs. Clear the flag so that it doesn't call into
425 	 * eventfs during the remount flag updates. The eventfs_inode
426 	 * gets freed after an RCU cycle, so the content will still
427 	 * be safe if the iteration is going on now.
428 	 */
429 	ti->flags &= ~TRACEFS_EVENT_INODE;
430 
431 	return 1;
432 }
433 
434 static const struct super_operations tracefs_super_operations = {
435 	.alloc_inode    = tracefs_alloc_inode,
436 	.free_inode     = tracefs_free_inode,
437 	.destroy_inode  = tracefs_destroy_inode,
438 	.drop_inode     = tracefs_drop_inode,
439 	.statfs		= simple_statfs,
440 	.show_options	= tracefs_show_options,
441 };
442 
443 /*
444  * It would be cleaner if eventfs had its own dentry ops.
445  *
446  * Note that d_revalidate is called potentially under RCU,
447  * so it can't take the eventfs mutex etc. It's fine - if
448  * we open a file just as it's marked dead, things will
449  * still work just fine, and just see the old stale case.
450  */
tracefs_d_release(struct dentry * dentry)451 static void tracefs_d_release(struct dentry *dentry)
452 {
453 	if (dentry->d_fsdata)
454 		eventfs_d_release(dentry);
455 }
456 
tracefs_d_revalidate(struct dentry * dentry,unsigned int flags)457 static int tracefs_d_revalidate(struct dentry *dentry, unsigned int flags)
458 {
459 	struct eventfs_inode *ei = dentry->d_fsdata;
460 
461 	return !(ei && ei->is_freed);
462 }
463 
464 static const struct dentry_operations tracefs_dentry_operations = {
465 	.d_revalidate = tracefs_d_revalidate,
466 	.d_release = tracefs_d_release,
467 };
468 
tracefs_fill_super(struct super_block * sb,struct fs_context * fc)469 static int tracefs_fill_super(struct super_block *sb, struct fs_context *fc)
470 {
471 	static const struct tree_descr trace_files[] = {{""}};
472 	int err;
473 
474 	err = simple_fill_super(sb, TRACEFS_MAGIC, trace_files);
475 	if (err)
476 		return err;
477 
478 	sb->s_op = &tracefs_super_operations;
479 	sb->s_d_op = &tracefs_dentry_operations;
480 
481 	tracefs_apply_options(sb, false);
482 
483 	return 0;
484 }
485 
tracefs_get_tree(struct fs_context * fc)486 static int tracefs_get_tree(struct fs_context *fc)
487 {
488 	return get_tree_single(fc, tracefs_fill_super);
489 }
490 
tracefs_free_fc(struct fs_context * fc)491 static void tracefs_free_fc(struct fs_context *fc)
492 {
493 	kfree(fc->s_fs_info);
494 }
495 
496 static const struct fs_context_operations tracefs_context_ops = {
497 	.free		= tracefs_free_fc,
498 	.parse_param	= tracefs_parse_param,
499 	.get_tree	= tracefs_get_tree,
500 	.reconfigure	= tracefs_reconfigure,
501 };
502 
tracefs_init_fs_context(struct fs_context * fc)503 static int tracefs_init_fs_context(struct fs_context *fc)
504 {
505 	struct tracefs_fs_info *fsi;
506 
507 	fsi = kzalloc(sizeof(struct tracefs_fs_info), GFP_KERNEL);
508 	if (!fsi)
509 		return -ENOMEM;
510 
511 	fsi->mode = TRACEFS_DEFAULT_MODE;
512 
513 	fc->s_fs_info = fsi;
514 	fc->ops = &tracefs_context_ops;
515 	return 0;
516 }
517 
518 static struct file_system_type trace_fs_type = {
519 	.owner =	THIS_MODULE,
520 	.name =		"tracefs",
521 	.init_fs_context = tracefs_init_fs_context,
522 	.parameters	= tracefs_param_specs,
523 	.kill_sb =	kill_litter_super,
524 };
525 MODULE_ALIAS_FS("tracefs");
526 
tracefs_start_creating(const char * name,struct dentry * parent)527 struct dentry *tracefs_start_creating(const char *name, struct dentry *parent)
528 {
529 	struct dentry *dentry;
530 	int error;
531 
532 	pr_debug("tracefs: creating file '%s'\n",name);
533 
534 	error = simple_pin_fs(&trace_fs_type, &tracefs_mount,
535 			      &tracefs_mount_count);
536 	if (error)
537 		return ERR_PTR(error);
538 
539 	/* If the parent is not specified, we create it in the root.
540 	 * We need the root dentry to do this, which is in the super
541 	 * block. A pointer to that is in the struct vfsmount that we
542 	 * have around.
543 	 */
544 	if (!parent)
545 		parent = tracefs_mount->mnt_root;
546 
547 	inode_lock(d_inode(parent));
548 	if (unlikely(IS_DEADDIR(d_inode(parent))))
549 		dentry = ERR_PTR(-ENOENT);
550 	else
551 		dentry = lookup_one_len(name, parent, strlen(name));
552 	if (!IS_ERR(dentry) && d_inode(dentry)) {
553 		dput(dentry);
554 		dentry = ERR_PTR(-EEXIST);
555 	}
556 
557 	if (IS_ERR(dentry)) {
558 		inode_unlock(d_inode(parent));
559 		simple_release_fs(&tracefs_mount, &tracefs_mount_count);
560 	}
561 
562 	return dentry;
563 }
564 
tracefs_failed_creating(struct dentry * dentry)565 struct dentry *tracefs_failed_creating(struct dentry *dentry)
566 {
567 	inode_unlock(d_inode(dentry->d_parent));
568 	dput(dentry);
569 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
570 	return NULL;
571 }
572 
tracefs_end_creating(struct dentry * dentry)573 struct dentry *tracefs_end_creating(struct dentry *dentry)
574 {
575 	inode_unlock(d_inode(dentry->d_parent));
576 	return dentry;
577 }
578 
579 /* Find the inode that this will use for default */
instance_inode(struct dentry * parent,struct inode * inode)580 static struct inode *instance_inode(struct dentry *parent, struct inode *inode)
581 {
582 	struct tracefs_inode *ti;
583 
584 	/* If parent is NULL then use root inode */
585 	if (!parent)
586 		return d_inode(inode->i_sb->s_root);
587 
588 	/* Find the inode that is flagged as an instance or the root inode */
589 	while (!IS_ROOT(parent)) {
590 		ti = get_tracefs(d_inode(parent));
591 		if (ti->flags & TRACEFS_INSTANCE_INODE)
592 			break;
593 		parent = parent->d_parent;
594 	}
595 
596 	return d_inode(parent);
597 }
598 
599 /**
600  * tracefs_create_file - create a file in the tracefs filesystem
601  * @name: a pointer to a string containing the name of the file to create.
602  * @mode: the permission that the file should have.
603  * @parent: a pointer to the parent dentry for this file.  This should be a
604  *          directory dentry if set.  If this parameter is NULL, then the
605  *          file will be created in the root of the tracefs filesystem.
606  * @data: a pointer to something that the caller will want to get to later
607  *        on.  The inode.i_private pointer will point to this value on
608  *        the open() call.
609  * @fops: a pointer to a struct file_operations that should be used for
610  *        this file.
611  *
612  * This is the basic "create a file" function for tracefs.  It allows for a
613  * wide range of flexibility in creating a file, or a directory (if you want
614  * to create a directory, the tracefs_create_dir() function is
615  * recommended to be used instead.)
616  *
617  * This function will return a pointer to a dentry if it succeeds.  This
618  * pointer must be passed to the tracefs_remove() function when the file is
619  * to be removed (no automatic cleanup happens if your module is unloaded,
620  * you are responsible here.)  If an error occurs, %NULL will be returned.
621  *
622  * If tracefs is not enabled in the kernel, the value -%ENODEV will be
623  * returned.
624  */
tracefs_create_file(const char * name,umode_t mode,struct dentry * parent,void * data,const struct file_operations * fops)625 struct dentry *tracefs_create_file(const char *name, umode_t mode,
626 				   struct dentry *parent, void *data,
627 				   const struct file_operations *fops)
628 {
629 	struct tracefs_inode *ti;
630 	struct dentry *dentry;
631 	struct inode *inode;
632 
633 	if (security_locked_down(LOCKDOWN_TRACEFS))
634 		return NULL;
635 
636 	if (!(mode & S_IFMT))
637 		mode |= S_IFREG;
638 	BUG_ON(!S_ISREG(mode));
639 	dentry = tracefs_start_creating(name, parent);
640 
641 	if (IS_ERR(dentry))
642 		return NULL;
643 
644 	inode = tracefs_get_inode(dentry->d_sb);
645 	if (unlikely(!inode))
646 		return tracefs_failed_creating(dentry);
647 
648 	ti = get_tracefs(inode);
649 	ti->private = instance_inode(parent, inode);
650 
651 	inode->i_mode = mode;
652 	inode->i_op = &tracefs_file_inode_operations;
653 	inode->i_fop = fops ? fops : &tracefs_file_operations;
654 	inode->i_private = data;
655 	inode->i_uid = d_inode(dentry->d_parent)->i_uid;
656 	inode->i_gid = d_inode(dentry->d_parent)->i_gid;
657 	d_instantiate(dentry, inode);
658 	fsnotify_create(d_inode(dentry->d_parent), dentry);
659 	return tracefs_end_creating(dentry);
660 }
661 
__create_dir(const char * name,struct dentry * parent,const struct inode_operations * ops)662 static struct dentry *__create_dir(const char *name, struct dentry *parent,
663 				   const struct inode_operations *ops)
664 {
665 	struct tracefs_inode *ti;
666 	struct dentry *dentry = tracefs_start_creating(name, parent);
667 	struct inode *inode;
668 
669 	if (IS_ERR(dentry))
670 		return NULL;
671 
672 	inode = tracefs_get_inode(dentry->d_sb);
673 	if (unlikely(!inode))
674 		return tracefs_failed_creating(dentry);
675 
676 	/* Do not set bits for OTH */
677 	inode->i_mode = S_IFDIR | S_IRWXU | S_IRUSR| S_IRGRP | S_IXUSR | S_IXGRP;
678 	inode->i_op = ops;
679 	inode->i_fop = &simple_dir_operations;
680 	inode->i_uid = d_inode(dentry->d_parent)->i_uid;
681 	inode->i_gid = d_inode(dentry->d_parent)->i_gid;
682 
683 	ti = get_tracefs(inode);
684 	ti->private = instance_inode(parent, inode);
685 
686 	/* directory inodes start off with i_nlink == 2 (for "." entry) */
687 	inc_nlink(inode);
688 	d_instantiate(dentry, inode);
689 	inc_nlink(d_inode(dentry->d_parent));
690 	fsnotify_mkdir(d_inode(dentry->d_parent), dentry);
691 	return tracefs_end_creating(dentry);
692 }
693 
694 /**
695  * tracefs_create_dir - create a directory in the tracefs filesystem
696  * @name: a pointer to a string containing the name of the directory to
697  *        create.
698  * @parent: a pointer to the parent dentry for this file.  This should be a
699  *          directory dentry if set.  If this parameter is NULL, then the
700  *          directory will be created in the root of the tracefs filesystem.
701  *
702  * This function creates a directory in tracefs with the given name.
703  *
704  * This function will return a pointer to a dentry if it succeeds.  This
705  * pointer must be passed to the tracefs_remove() function when the file is
706  * to be removed. If an error occurs, %NULL will be returned.
707  *
708  * If tracing is not enabled in the kernel, the value -%ENODEV will be
709  * returned.
710  */
tracefs_create_dir(const char * name,struct dentry * parent)711 struct dentry *tracefs_create_dir(const char *name, struct dentry *parent)
712 {
713 	if (security_locked_down(LOCKDOWN_TRACEFS))
714 		return NULL;
715 
716 	return __create_dir(name, parent, &tracefs_dir_inode_operations);
717 }
718 
719 /**
720  * tracefs_create_instance_dir - create the tracing instances directory
721  * @name: The name of the instances directory to create
722  * @parent: The parent directory that the instances directory will exist
723  * @mkdir: The function to call when a mkdir is performed.
724  * @rmdir: The function to call when a rmdir is performed.
725  *
726  * Only one instances directory is allowed.
727  *
728  * The instances directory is special as it allows for mkdir and rmdir
729  * to be done by userspace. When a mkdir or rmdir is performed, the inode
730  * locks are released and the methods passed in (@mkdir and @rmdir) are
731  * called without locks and with the name of the directory being created
732  * within the instances directory.
733  *
734  * Returns the dentry of the instances directory.
735  */
tracefs_create_instance_dir(const char * name,struct dentry * parent,int (* mkdir)(const char * name),int (* rmdir)(const char * name))736 __init struct dentry *tracefs_create_instance_dir(const char *name,
737 					  struct dentry *parent,
738 					  int (*mkdir)(const char *name),
739 					  int (*rmdir)(const char *name))
740 {
741 	struct dentry *dentry;
742 
743 	/* Only allow one instance of the instances directory. */
744 	if (WARN_ON(tracefs_ops.mkdir || tracefs_ops.rmdir))
745 		return NULL;
746 
747 	dentry = __create_dir(name, parent, &tracefs_instance_dir_inode_operations);
748 	if (!dentry)
749 		return NULL;
750 
751 	tracefs_ops.mkdir = mkdir;
752 	tracefs_ops.rmdir = rmdir;
753 
754 	return dentry;
755 }
756 
remove_one(struct dentry * victim)757 static void remove_one(struct dentry *victim)
758 {
759 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
760 }
761 
762 /**
763  * tracefs_remove - recursively removes a directory
764  * @dentry: a pointer to a the dentry of the directory to be removed.
765  *
766  * This function recursively removes a directory tree in tracefs that
767  * was previously created with a call to another tracefs function
768  * (like tracefs_create_file() or variants thereof.)
769  */
tracefs_remove(struct dentry * dentry)770 void tracefs_remove(struct dentry *dentry)
771 {
772 	if (IS_ERR_OR_NULL(dentry))
773 		return;
774 
775 	simple_pin_fs(&trace_fs_type, &tracefs_mount, &tracefs_mount_count);
776 	simple_recursive_removal(dentry, remove_one);
777 	simple_release_fs(&tracefs_mount, &tracefs_mount_count);
778 }
779 
780 /**
781  * tracefs_initialized - Tells whether tracefs has been registered
782  */
tracefs_initialized(void)783 bool tracefs_initialized(void)
784 {
785 	return tracefs_registered;
786 }
787 
init_once(void * foo)788 static void init_once(void *foo)
789 {
790 	struct tracefs_inode *ti = (struct tracefs_inode *) foo;
791 
792 	/* inode_init_once() calls memset() on the vfs_inode portion */
793 	inode_init_once(&ti->vfs_inode);
794 
795 	/* Zero out the rest */
796 	memset_after(ti, 0, vfs_inode);
797 }
798 
tracefs_init(void)799 static int __init tracefs_init(void)
800 {
801 	int retval;
802 
803 	tracefs_inode_cachep = kmem_cache_create("tracefs_inode_cache",
804 						 sizeof(struct tracefs_inode),
805 						 0, (SLAB_RECLAIM_ACCOUNT|
806 						     SLAB_ACCOUNT),
807 						 init_once);
808 	if (!tracefs_inode_cachep)
809 		return -ENOMEM;
810 
811 	retval = sysfs_create_mount_point(kernel_kobj, "tracing");
812 	if (retval)
813 		return -EINVAL;
814 
815 	retval = register_filesystem(&trace_fs_type);
816 	if (!retval)
817 		tracefs_registered = true;
818 
819 	return retval;
820 }
821 core_initcall(tracefs_init);
822