2 * inode.c - part of tracefs, a pseudo file system for activating tracing
4 * Based on debugfs by: Greg Kroah-Hartman <greg@kroah.com>
6 * Copyright (C) 2014 Red Hat Inc, author: Steven Rostedt <srostedt@redhat.com>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License version
10 * 2 as published by the Free Software Foundation.
12 * tracefs is the file system that is used by the tracing infrastructure.
16 #include <linux/module.h>
18 #include <linux/mount.h>
19 #include <linux/kobject.h>
20 #include <linux/namei.h>
21 #include <linux/tracefs.h>
22 #include <linux/fsnotify.h>
23 #include <linux/seq_file.h>
24 #include <linux/parser.h>
25 #include <linux/magic.h>
26 #include <linux/slab.h>
28 #define TRACEFS_DEFAULT_MODE 0700
30 static struct vfsmount *tracefs_mount;
31 static int tracefs_mount_count;
32 static bool tracefs_registered;
34 static ssize_t default_read_file(struct file *file, char __user *buf,
35 size_t count, loff_t *ppos)
40 static ssize_t default_write_file(struct file *file, const char __user *buf,
41 size_t count, loff_t *ppos)
46 static const struct file_operations tracefs_file_operations = {
47 .read = default_read_file,
48 .write = default_write_file,
50 .llseek = noop_llseek,
53 static struct tracefs_dir_ops {
54 int (*mkdir)(const char *name);
55 int (*rmdir)(const char *name);
58 static char *get_dname(struct dentry *dentry)
62 int len = dentry->d_name.len;
64 dname = dentry->d_name.name;
65 name = kmalloc(len + 1, GFP_KERNEL);
68 memcpy(name, dname, len);
73 static int tracefs_syscall_mkdir(struct inode *inode, struct dentry *dentry, umode_t mode)
78 name = get_dname(dentry);
83 * The mkdir call can call the generic functions that create
84 * the files within the tracefs system. It is up to the individual
85 * mkdir routine to handle races.
87 mutex_unlock(&inode->i_mutex);
88 ret = tracefs_ops.mkdir(name);
89 mutex_lock(&inode->i_mutex);
96 static int tracefs_syscall_rmdir(struct inode *inode, struct dentry *dentry)
101 name = get_dname(dentry);
106 * The rmdir call can call the generic functions that create
107 * the files within the tracefs system. It is up to the individual
108 * rmdir routine to handle races.
109 * This time we need to unlock not only the parent (inode) but
110 * also the directory that is being deleted.
112 mutex_unlock(&inode->i_mutex);
113 mutex_unlock(&dentry->d_inode->i_mutex);
115 ret = tracefs_ops.rmdir(name);
117 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
118 mutex_lock(&dentry->d_inode->i_mutex);
125 static const struct inode_operations tracefs_dir_inode_operations = {
126 .lookup = simple_lookup,
127 .mkdir = tracefs_syscall_mkdir,
128 .rmdir = tracefs_syscall_rmdir,
131 static struct inode *tracefs_get_inode(struct super_block *sb)
133 struct inode *inode = new_inode(sb);
135 inode->i_ino = get_next_ino();
136 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
141 struct tracefs_mount_opts {
154 static const match_table_t tokens = {
157 {Opt_mode, "mode=%o"},
161 struct tracefs_fs_info {
162 struct tracefs_mount_opts mount_opts;
165 static int tracefs_parse_options(char *data, struct tracefs_mount_opts *opts)
167 substring_t args[MAX_OPT_ARGS];
174 opts->mode = TRACEFS_DEFAULT_MODE;
176 while ((p = strsep(&data, ",")) != NULL) {
180 token = match_token(p, tokens, args);
183 if (match_int(&args[0], &option))
185 uid = make_kuid(current_user_ns(), option);
191 if (match_int(&args[0], &option))
193 gid = make_kgid(current_user_ns(), option);
199 if (match_octal(&args[0], &option))
201 opts->mode = option & S_IALLUGO;
204 * We might like to report bad mount options here;
205 * but traditionally tracefs has ignored all mount options
213 static int tracefs_apply_options(struct super_block *sb)
215 struct tracefs_fs_info *fsi = sb->s_fs_info;
216 struct inode *inode = sb->s_root->d_inode;
217 struct tracefs_mount_opts *opts = &fsi->mount_opts;
219 inode->i_mode &= ~S_IALLUGO;
220 inode->i_mode |= opts->mode;
222 inode->i_uid = opts->uid;
223 inode->i_gid = opts->gid;
228 static int tracefs_remount(struct super_block *sb, int *flags, char *data)
231 struct tracefs_fs_info *fsi = sb->s_fs_info;
234 err = tracefs_parse_options(data, &fsi->mount_opts);
238 tracefs_apply_options(sb);
244 static int tracefs_show_options(struct seq_file *m, struct dentry *root)
246 struct tracefs_fs_info *fsi = root->d_sb->s_fs_info;
247 struct tracefs_mount_opts *opts = &fsi->mount_opts;
249 if (!uid_eq(opts->uid, GLOBAL_ROOT_UID))
250 seq_printf(m, ",uid=%u",
251 from_kuid_munged(&init_user_ns, opts->uid));
252 if (!gid_eq(opts->gid, GLOBAL_ROOT_GID))
253 seq_printf(m, ",gid=%u",
254 from_kgid_munged(&init_user_ns, opts->gid));
255 if (opts->mode != TRACEFS_DEFAULT_MODE)
256 seq_printf(m, ",mode=%o", opts->mode);
261 static const struct super_operations tracefs_super_operations = {
262 .statfs = simple_statfs,
263 .remount_fs = tracefs_remount,
264 .show_options = tracefs_show_options,
267 static int trace_fill_super(struct super_block *sb, void *data, int silent)
269 static struct tree_descr trace_files[] = {{""}};
270 struct tracefs_fs_info *fsi;
273 save_mount_options(sb, data);
275 fsi = kzalloc(sizeof(struct tracefs_fs_info), GFP_KERNEL);
282 err = tracefs_parse_options(data, &fsi->mount_opts);
286 err = simple_fill_super(sb, TRACEFS_MAGIC, trace_files);
290 sb->s_op = &tracefs_super_operations;
292 tracefs_apply_options(sb);
298 sb->s_fs_info = NULL;
302 static struct dentry *trace_mount(struct file_system_type *fs_type,
303 int flags, const char *dev_name,
306 return mount_single(fs_type, flags, data, trace_fill_super);
309 static struct file_system_type trace_fs_type = {
310 .owner = THIS_MODULE,
312 .mount = trace_mount,
313 .kill_sb = kill_litter_super,
315 MODULE_ALIAS_FS("tracefs");
317 static struct dentry *start_creating(const char *name, struct dentry *parent)
319 struct dentry *dentry;
322 pr_debug("tracefs: creating file '%s'\n",name);
324 error = simple_pin_fs(&trace_fs_type, &tracefs_mount,
325 &tracefs_mount_count);
327 return ERR_PTR(error);
329 /* If the parent is not specified, we create it in the root.
330 * We need the root dentry to do this, which is in the super
331 * block. A pointer to that is in the struct vfsmount that we
335 parent = tracefs_mount->mnt_root;
337 mutex_lock(&parent->d_inode->i_mutex);
338 dentry = lookup_one_len(name, parent, strlen(name));
339 if (!IS_ERR(dentry) && dentry->d_inode) {
341 dentry = ERR_PTR(-EEXIST);
344 if (IS_ERR(dentry)) {
345 mutex_unlock(&parent->d_inode->i_mutex);
346 simple_release_fs(&tracefs_mount, &tracefs_mount_count);
352 static struct dentry *failed_creating(struct dentry *dentry)
354 mutex_unlock(&dentry->d_parent->d_inode->i_mutex);
356 simple_release_fs(&tracefs_mount, &tracefs_mount_count);
360 static struct dentry *end_creating(struct dentry *dentry)
362 mutex_unlock(&dentry->d_parent->d_inode->i_mutex);
367 * tracefs_create_file - create a file in the tracefs filesystem
368 * @name: a pointer to a string containing the name of the file to create.
369 * @mode: the permission that the file should have.
370 * @parent: a pointer to the parent dentry for this file. This should be a
371 * directory dentry if set. If this parameter is NULL, then the
372 * file will be created in the root of the tracefs filesystem.
373 * @data: a pointer to something that the caller will want to get to later
374 * on. The inode.i_private pointer will point to this value on
376 * @fops: a pointer to a struct file_operations that should be used for
379 * This is the basic "create a file" function for tracefs. It allows for a
380 * wide range of flexibility in creating a file, or a directory (if you want
381 * to create a directory, the tracefs_create_dir() function is
382 * recommended to be used instead.)
384 * This function will return a pointer to a dentry if it succeeds. This
385 * pointer must be passed to the tracefs_remove() function when the file is
386 * to be removed (no automatic cleanup happens if your module is unloaded,
387 * you are responsible here.) If an error occurs, %NULL will be returned.
389 * If tracefs is not enabled in the kernel, the value -%ENODEV will be
392 struct dentry *tracefs_create_file(const char *name, umode_t mode,
393 struct dentry *parent, void *data,
394 const struct file_operations *fops)
396 struct dentry *dentry;
399 if (!(mode & S_IFMT))
401 BUG_ON(!S_ISREG(mode));
402 dentry = start_creating(name, parent);
407 inode = tracefs_get_inode(dentry->d_sb);
408 if (unlikely(!inode))
409 return failed_creating(dentry);
411 inode->i_mode = mode;
412 inode->i_fop = fops ? fops : &tracefs_file_operations;
413 inode->i_private = data;
414 d_instantiate(dentry, inode);
415 fsnotify_create(dentry->d_parent->d_inode, dentry);
416 return end_creating(dentry);
419 static struct dentry *__create_dir(const char *name, struct dentry *parent,
420 const struct inode_operations *ops)
422 struct dentry *dentry = start_creating(name, parent);
428 inode = tracefs_get_inode(dentry->d_sb);
429 if (unlikely(!inode))
430 return failed_creating(dentry);
432 inode->i_mode = S_IFDIR | S_IRWXU | S_IRUGO | S_IXUGO;
434 inode->i_fop = &simple_dir_operations;
436 /* directory inodes start off with i_nlink == 2 (for "." entry) */
438 d_instantiate(dentry, inode);
439 inc_nlink(dentry->d_parent->d_inode);
440 fsnotify_mkdir(dentry->d_parent->d_inode, dentry);
441 return end_creating(dentry);
445 * tracefs_create_dir - create a directory in the tracefs filesystem
446 * @name: a pointer to a string containing the name of the directory to
448 * @parent: a pointer to the parent dentry for this file. This should be a
449 * directory dentry if set. If this parameter is NULL, then the
450 * directory will be created in the root of the tracefs filesystem.
452 * This function creates a directory in tracefs with the given name.
454 * This function will return a pointer to a dentry if it succeeds. This
455 * pointer must be passed to the tracefs_remove() function when the file is
456 * to be removed. If an error occurs, %NULL will be returned.
458 * If tracing is not enabled in the kernel, the value -%ENODEV will be
461 struct dentry *tracefs_create_dir(const char *name, struct dentry *parent)
463 return __create_dir(name, parent, &simple_dir_inode_operations);
467 * tracefs_create_instance_dir - create the tracing instances directory
468 * @name: The name of the instances directory to create
469 * @parent: The parent directory that the instances directory will exist
470 * @mkdir: The function to call when a mkdir is performed.
471 * @rmdir: The function to call when a rmdir is performed.
473 * Only one instances directory is allowed.
475 * The instances directory is special as it allows for mkdir and rmdir to
476 * to be done by userspace. When a mkdir or rmdir is performed, the inode
477 * locks are released and the methhods passed in (@mkdir and @rmdir) are
478 * called without locks and with the name of the directory being created
479 * within the instances directory.
481 * Returns the dentry of the instances directory.
483 struct dentry *tracefs_create_instance_dir(const char *name, struct dentry *parent,
484 int (*mkdir)(const char *name),
485 int (*rmdir)(const char *name))
487 struct dentry *dentry;
489 /* Only allow one instance of the instances directory. */
490 if (WARN_ON(tracefs_ops.mkdir || tracefs_ops.rmdir))
493 dentry = __create_dir(name, parent, &tracefs_dir_inode_operations);
497 tracefs_ops.mkdir = mkdir;
498 tracefs_ops.rmdir = rmdir;
503 static int __tracefs_remove(struct dentry *dentry, struct dentry *parent)
507 if (simple_positive(dentry)) {
508 if (dentry->d_inode) {
510 switch (dentry->d_inode->i_mode & S_IFMT) {
512 ret = simple_rmdir(parent->d_inode, dentry);
515 simple_unlink(parent->d_inode, dentry);
527 * tracefs_remove - removes a file or directory from the tracefs filesystem
528 * @dentry: a pointer to a the dentry of the file or directory to be
531 * This function removes a file or directory in tracefs that was previously
532 * created with a call to another tracefs function (like
533 * tracefs_create_file() or variants thereof.)
535 void tracefs_remove(struct dentry *dentry)
537 struct dentry *parent;
540 if (IS_ERR_OR_NULL(dentry))
543 parent = dentry->d_parent;
544 if (!parent || !parent->d_inode)
547 mutex_lock(&parent->d_inode->i_mutex);
548 ret = __tracefs_remove(dentry, parent);
549 mutex_unlock(&parent->d_inode->i_mutex);
551 simple_release_fs(&tracefs_mount, &tracefs_mount_count);
555 * tracefs_remove_recursive - recursively removes a directory
556 * @dentry: a pointer to a the dentry of the directory to be removed.
558 * This function recursively removes a directory tree in tracefs that
559 * was previously created with a call to another tracefs function
560 * (like tracefs_create_file() or variants thereof.)
562 void tracefs_remove_recursive(struct dentry *dentry)
564 struct dentry *child, *parent;
566 if (IS_ERR_OR_NULL(dentry))
569 parent = dentry->d_parent;
570 if (!parent || !parent->d_inode)
575 mutex_lock(&parent->d_inode->i_mutex);
578 * The parent->d_subdirs is protected by the d_lock. Outside that
579 * lock, the child can be unlinked and set to be freed which can
580 * use the d_u.d_child as the rcu head and corrupt this list.
582 spin_lock(&parent->d_lock);
583 list_for_each_entry(child, &parent->d_subdirs, d_child) {
584 if (!simple_positive(child))
587 /* perhaps simple_empty(child) makes more sense */
588 if (!list_empty(&child->d_subdirs)) {
589 spin_unlock(&parent->d_lock);
590 mutex_unlock(&parent->d_inode->i_mutex);
595 spin_unlock(&parent->d_lock);
597 if (!__tracefs_remove(child, parent))
598 simple_release_fs(&tracefs_mount, &tracefs_mount_count);
601 * The parent->d_lock protects agaist child from unlinking
602 * from d_subdirs. When releasing the parent->d_lock we can
603 * no longer trust that the next pointer is valid.
604 * Restart the loop. We'll skip this one with the
605 * simple_positive() check.
609 spin_unlock(&parent->d_lock);
611 mutex_unlock(&parent->d_inode->i_mutex);
613 parent = parent->d_parent;
614 mutex_lock(&parent->d_inode->i_mutex);
620 if (!__tracefs_remove(child, parent))
621 simple_release_fs(&tracefs_mount, &tracefs_mount_count);
622 mutex_unlock(&parent->d_inode->i_mutex);
626 * tracefs_initialized - Tells whether tracefs has been registered
628 bool tracefs_initialized(void)
630 return tracefs_registered;
633 static int __init tracefs_init(void)
637 retval = sysfs_create_mount_point(kernel_kobj, "tracing");
641 retval = register_filesystem(&trace_fs_type);
643 tracefs_registered = true;
647 core_initcall(tracefs_init);