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.
88 ret
= tracefs_ops
.mkdir(name
);
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.
113 inode_unlock(dentry
->d_inode
);
115 ret
= tracefs_ops
.rmdir(name
);
117 inode_lock_nested(inode
, I_MUTEX_PARENT
);
118 inode_lock(dentry
->d_inode
);
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(inode
);
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 const struct tree_descr trace_files
[] = {{""}};
270 struct tracefs_fs_info
*fsi
;
273 fsi
= kzalloc(sizeof(struct tracefs_fs_info
), GFP_KERNEL
);
280 err
= tracefs_parse_options(data
, &fsi
->mount_opts
);
284 err
= simple_fill_super(sb
, TRACEFS_MAGIC
, trace_files
);
288 sb
->s_op
= &tracefs_super_operations
;
290 tracefs_apply_options(sb
);
296 sb
->s_fs_info
= NULL
;
300 static struct dentry
*trace_mount(struct file_system_type
*fs_type
,
301 int flags
, const char *dev_name
,
304 return mount_single(fs_type
, flags
, data
, trace_fill_super
);
307 static struct file_system_type trace_fs_type
= {
308 .owner
= THIS_MODULE
,
310 .mount
= trace_mount
,
311 .kill_sb
= kill_litter_super
,
313 MODULE_ALIAS_FS("tracefs");
315 static struct dentry
*start_creating(const char *name
, struct dentry
*parent
)
317 struct dentry
*dentry
;
320 pr_debug("tracefs: creating file '%s'\n",name
);
322 error
= simple_pin_fs(&trace_fs_type
, &tracefs_mount
,
323 &tracefs_mount_count
);
325 return ERR_PTR(error
);
327 /* If the parent is not specified, we create it in the root.
328 * We need the root dentry to do this, which is in the super
329 * block. A pointer to that is in the struct vfsmount that we
333 parent
= tracefs_mount
->mnt_root
;
335 inode_lock(parent
->d_inode
);
336 dentry
= lookup_one_len(name
, parent
, strlen(name
));
337 if (!IS_ERR(dentry
) && dentry
->d_inode
) {
339 dentry
= ERR_PTR(-EEXIST
);
342 if (IS_ERR(dentry
)) {
343 inode_unlock(parent
->d_inode
);
344 simple_release_fs(&tracefs_mount
, &tracefs_mount_count
);
350 static struct dentry
*failed_creating(struct dentry
*dentry
)
352 inode_unlock(dentry
->d_parent
->d_inode
);
354 simple_release_fs(&tracefs_mount
, &tracefs_mount_count
);
358 static struct dentry
*end_creating(struct dentry
*dentry
)
360 inode_unlock(dentry
->d_parent
->d_inode
);
365 * tracefs_create_file - create a file in the tracefs filesystem
366 * @name: a pointer to a string containing the name of the file to create.
367 * @mode: the permission that the file should have.
368 * @parent: a pointer to the parent dentry for this file. This should be a
369 * directory dentry if set. If this parameter is NULL, then the
370 * file will be created in the root of the tracefs filesystem.
371 * @data: a pointer to something that the caller will want to get to later
372 * on. The inode.i_private pointer will point to this value on
374 * @fops: a pointer to a struct file_operations that should be used for
377 * This is the basic "create a file" function for tracefs. It allows for a
378 * wide range of flexibility in creating a file, or a directory (if you want
379 * to create a directory, the tracefs_create_dir() function is
380 * recommended to be used instead.)
382 * This function will return a pointer to a dentry if it succeeds. This
383 * pointer must be passed to the tracefs_remove() function when the file is
384 * to be removed (no automatic cleanup happens if your module is unloaded,
385 * you are responsible here.) If an error occurs, %NULL will be returned.
387 * If tracefs is not enabled in the kernel, the value -%ENODEV will be
390 struct dentry
*tracefs_create_file(const char *name
, umode_t mode
,
391 struct dentry
*parent
, void *data
,
392 const struct file_operations
*fops
)
394 struct dentry
*dentry
;
397 if (!(mode
& S_IFMT
))
399 BUG_ON(!S_ISREG(mode
));
400 dentry
= start_creating(name
, parent
);
405 inode
= tracefs_get_inode(dentry
->d_sb
);
406 if (unlikely(!inode
))
407 return failed_creating(dentry
);
409 inode
->i_mode
= mode
;
410 inode
->i_fop
= fops
? fops
: &tracefs_file_operations
;
411 inode
->i_private
= data
;
412 d_instantiate(dentry
, inode
);
413 fsnotify_create(dentry
->d_parent
->d_inode
, dentry
);
414 return end_creating(dentry
);
417 static struct dentry
*__create_dir(const char *name
, struct dentry
*parent
,
418 const struct inode_operations
*ops
)
420 struct dentry
*dentry
= start_creating(name
, parent
);
426 inode
= tracefs_get_inode(dentry
->d_sb
);
427 if (unlikely(!inode
))
428 return failed_creating(dentry
);
430 inode
->i_mode
= S_IFDIR
| S_IRWXU
| S_IRUGO
| S_IXUGO
;
432 inode
->i_fop
= &simple_dir_operations
;
434 /* directory inodes start off with i_nlink == 2 (for "." entry) */
436 d_instantiate(dentry
, inode
);
437 inc_nlink(dentry
->d_parent
->d_inode
);
438 fsnotify_mkdir(dentry
->d_parent
->d_inode
, dentry
);
439 return end_creating(dentry
);
443 * tracefs_create_dir - create a directory in the tracefs filesystem
444 * @name: a pointer to a string containing the name of the directory to
446 * @parent: a pointer to the parent dentry for this file. This should be a
447 * directory dentry if set. If this parameter is NULL, then the
448 * directory will be created in the root of the tracefs filesystem.
450 * This function creates a directory in tracefs with the given name.
452 * This function will return a pointer to a dentry if it succeeds. This
453 * pointer must be passed to the tracefs_remove() function when the file is
454 * to be removed. If an error occurs, %NULL will be returned.
456 * If tracing is not enabled in the kernel, the value -%ENODEV will be
459 struct dentry
*tracefs_create_dir(const char *name
, struct dentry
*parent
)
461 return __create_dir(name
, parent
, &simple_dir_inode_operations
);
465 * tracefs_create_instance_dir - create the tracing instances directory
466 * @name: The name of the instances directory to create
467 * @parent: The parent directory that the instances directory will exist
468 * @mkdir: The function to call when a mkdir is performed.
469 * @rmdir: The function to call when a rmdir is performed.
471 * Only one instances directory is allowed.
473 * The instances directory is special as it allows for mkdir and rmdir to
474 * to be done by userspace. When a mkdir or rmdir is performed, the inode
475 * locks are released and the methhods passed in (@mkdir and @rmdir) are
476 * called without locks and with the name of the directory being created
477 * within the instances directory.
479 * Returns the dentry of the instances directory.
481 struct dentry
*tracefs_create_instance_dir(const char *name
, struct dentry
*parent
,
482 int (*mkdir
)(const char *name
),
483 int (*rmdir
)(const char *name
))
485 struct dentry
*dentry
;
487 /* Only allow one instance of the instances directory. */
488 if (WARN_ON(tracefs_ops
.mkdir
|| tracefs_ops
.rmdir
))
491 dentry
= __create_dir(name
, parent
, &tracefs_dir_inode_operations
);
495 tracefs_ops
.mkdir
= mkdir
;
496 tracefs_ops
.rmdir
= rmdir
;
501 static int __tracefs_remove(struct dentry
*dentry
, struct dentry
*parent
)
505 if (simple_positive(dentry
)) {
506 if (dentry
->d_inode
) {
508 switch (dentry
->d_inode
->i_mode
& S_IFMT
) {
510 ret
= simple_rmdir(parent
->d_inode
, dentry
);
513 simple_unlink(parent
->d_inode
, dentry
);
525 * tracefs_remove - removes a file or directory from the tracefs filesystem
526 * @dentry: a pointer to a the dentry of the file or directory to be
529 * This function removes a file or directory in tracefs that was previously
530 * created with a call to another tracefs function (like
531 * tracefs_create_file() or variants thereof.)
533 void tracefs_remove(struct dentry
*dentry
)
535 struct dentry
*parent
;
538 if (IS_ERR_OR_NULL(dentry
))
541 parent
= dentry
->d_parent
;
542 inode_lock(parent
->d_inode
);
543 ret
= __tracefs_remove(dentry
, parent
);
544 inode_unlock(parent
->d_inode
);
546 simple_release_fs(&tracefs_mount
, &tracefs_mount_count
);
550 * tracefs_remove_recursive - recursively removes a directory
551 * @dentry: a pointer to a the dentry of the directory to be removed.
553 * This function recursively removes a directory tree in tracefs that
554 * was previously created with a call to another tracefs function
555 * (like tracefs_create_file() or variants thereof.)
557 void tracefs_remove_recursive(struct dentry
*dentry
)
559 struct dentry
*child
, *parent
;
561 if (IS_ERR_OR_NULL(dentry
))
566 inode_lock(parent
->d_inode
);
569 * The parent->d_subdirs is protected by the d_lock. Outside that
570 * lock, the child can be unlinked and set to be freed which can
571 * use the d_u.d_child as the rcu head and corrupt this list.
573 spin_lock(&parent
->d_lock
);
574 list_for_each_entry(child
, &parent
->d_subdirs
, d_child
) {
575 if (!simple_positive(child
))
578 /* perhaps simple_empty(child) makes more sense */
579 if (!list_empty(&child
->d_subdirs
)) {
580 spin_unlock(&parent
->d_lock
);
581 inode_unlock(parent
->d_inode
);
586 spin_unlock(&parent
->d_lock
);
588 if (!__tracefs_remove(child
, parent
))
589 simple_release_fs(&tracefs_mount
, &tracefs_mount_count
);
592 * The parent->d_lock protects agaist child from unlinking
593 * from d_subdirs. When releasing the parent->d_lock we can
594 * no longer trust that the next pointer is valid.
595 * Restart the loop. We'll skip this one with the
596 * simple_positive() check.
600 spin_unlock(&parent
->d_lock
);
602 inode_unlock(parent
->d_inode
);
604 parent
= parent
->d_parent
;
605 inode_lock(parent
->d_inode
);
611 if (!__tracefs_remove(child
, parent
))
612 simple_release_fs(&tracefs_mount
, &tracefs_mount_count
);
613 inode_unlock(parent
->d_inode
);
617 * tracefs_initialized - Tells whether tracefs has been registered
619 bool tracefs_initialized(void)
621 return tracefs_registered
;
624 static int __init
tracefs_init(void)
628 retval
= sysfs_create_mount_point(kernel_kobj
, "tracing");
632 retval
= register_filesystem(&trace_fs_type
);
634 tracefs_registered
= true;
638 core_initcall(tracefs_init
);