1 // SPDX-License-Identifier: GPL-2.0-only
3 * inode.c - part of tracefs, a pseudo file system for activating tracing
5 * Based on debugfs by: Greg Kroah-Hartman <greg@kroah.com>
7 * Copyright (C) 2014 Red Hat Inc, author: Steven Rostedt <srostedt@redhat.com>
9 * tracefs is the file system that is used by the tracing infrastructure.
12 #include <linux/module.h>
14 #include <linux/mount.h>
15 #include <linux/kobject.h>
16 #include <linux/namei.h>
17 #include <linux/tracefs.h>
18 #include <linux/fsnotify.h>
19 #include <linux/security.h>
20 #include <linux/seq_file.h>
21 #include <linux/parser.h>
22 #include <linux/magic.h>
23 #include <linux/slab.h>
25 #define TRACEFS_DEFAULT_MODE 0700
27 static struct vfsmount
*tracefs_mount
;
28 static int tracefs_mount_count
;
29 static bool tracefs_registered
;
31 static ssize_t
default_read_file(struct file
*file
, char __user
*buf
,
32 size_t count
, loff_t
*ppos
)
37 static ssize_t
default_write_file(struct file
*file
, const char __user
*buf
,
38 size_t count
, loff_t
*ppos
)
43 static const struct file_operations tracefs_file_operations
= {
44 .read
= default_read_file
,
45 .write
= default_write_file
,
47 .llseek
= noop_llseek
,
50 static struct tracefs_dir_ops
{
51 int (*mkdir
)(const char *name
);
52 int (*rmdir
)(const char *name
);
53 } tracefs_ops __ro_after_init
;
55 static char *get_dname(struct dentry
*dentry
)
59 int len
= dentry
->d_name
.len
;
61 dname
= dentry
->d_name
.name
;
62 name
= kmalloc(len
+ 1, GFP_KERNEL
);
65 memcpy(name
, dname
, len
);
70 static int tracefs_syscall_mkdir(struct user_namespace
*mnt_userns
,
71 struct inode
*inode
, struct dentry
*dentry
,
77 name
= get_dname(dentry
);
82 * The mkdir call can call the generic functions that create
83 * the files within the tracefs system. It is up to the individual
84 * mkdir routine to handle races.
87 ret
= tracefs_ops
.mkdir(name
);
95 static int tracefs_syscall_rmdir(struct inode
*inode
, struct dentry
*dentry
)
100 name
= get_dname(dentry
);
105 * The rmdir call can call the generic functions that create
106 * the files within the tracefs system. It is up to the individual
107 * rmdir routine to handle races.
108 * This time we need to unlock not only the parent (inode) but
109 * also the directory that is being deleted.
112 inode_unlock(d_inode(dentry
));
114 ret
= tracefs_ops
.rmdir(name
);
116 inode_lock_nested(inode
, I_MUTEX_PARENT
);
117 inode_lock(d_inode(dentry
));
124 static const struct inode_operations tracefs_dir_inode_operations
= {
125 .lookup
= simple_lookup
,
126 .mkdir
= tracefs_syscall_mkdir
,
127 .rmdir
= tracefs_syscall_rmdir
,
130 static struct inode
*tracefs_get_inode(struct super_block
*sb
)
132 struct inode
*inode
= new_inode(sb
);
134 inode
->i_ino
= get_next_ino();
135 inode
->i_atime
= inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
140 struct tracefs_mount_opts
{
153 static const match_table_t tokens
= {
156 {Opt_mode
, "mode=%o"},
160 struct tracefs_fs_info
{
161 struct tracefs_mount_opts mount_opts
;
164 static void change_gid(struct dentry
*dentry
, kgid_t gid
)
166 if (!dentry
->d_inode
)
168 dentry
->d_inode
->i_gid
= gid
;
172 * Taken from d_walk, but without he need for handling renames.
173 * Nothing can be renamed while walking the list, as tracefs
174 * does not support renames. This is only called when mounting
175 * or remounting the file system, to set all the files to
178 static void set_gid(struct dentry
*parent
, kgid_t gid
)
180 struct dentry
*this_parent
;
181 struct list_head
*next
;
183 this_parent
= parent
;
184 spin_lock(&this_parent
->d_lock
);
186 change_gid(this_parent
, gid
);
188 next
= this_parent
->d_subdirs
.next
;
190 while (next
!= &this_parent
->d_subdirs
) {
191 struct list_head
*tmp
= next
;
192 struct dentry
*dentry
= list_entry(tmp
, struct dentry
, d_child
);
195 spin_lock_nested(&dentry
->d_lock
, DENTRY_D_LOCK_NESTED
);
197 change_gid(dentry
, gid
);
199 if (!list_empty(&dentry
->d_subdirs
)) {
200 spin_unlock(&this_parent
->d_lock
);
201 spin_release(&dentry
->d_lock
.dep_map
, _RET_IP_
);
202 this_parent
= dentry
;
203 spin_acquire(&this_parent
->d_lock
.dep_map
, 0, 1, _RET_IP_
);
206 spin_unlock(&dentry
->d_lock
);
209 * All done at this level ... ascend and resume the search.
213 if (this_parent
!= parent
) {
214 struct dentry
*child
= this_parent
;
215 this_parent
= child
->d_parent
;
217 spin_unlock(&child
->d_lock
);
218 spin_lock(&this_parent
->d_lock
);
220 /* go into the first sibling still alive */
222 next
= child
->d_child
.next
;
223 if (next
== &this_parent
->d_subdirs
)
225 child
= list_entry(next
, struct dentry
, d_child
);
226 } while (unlikely(child
->d_flags
& DCACHE_DENTRY_KILLED
));
231 spin_unlock(&this_parent
->d_lock
);
235 static int tracefs_parse_options(char *data
, struct tracefs_mount_opts
*opts
)
237 substring_t args
[MAX_OPT_ARGS
];
244 opts
->mode
= TRACEFS_DEFAULT_MODE
;
246 while ((p
= strsep(&data
, ",")) != NULL
) {
250 token
= match_token(p
, tokens
, args
);
253 if (match_int(&args
[0], &option
))
255 uid
= make_kuid(current_user_ns(), option
);
261 if (match_int(&args
[0], &option
))
263 gid
= make_kgid(current_user_ns(), option
);
269 if (match_octal(&args
[0], &option
))
271 opts
->mode
= option
& S_IALLUGO
;
274 * We might like to report bad mount options here;
275 * but traditionally tracefs has ignored all mount options
283 static int tracefs_apply_options(struct super_block
*sb
)
285 struct tracefs_fs_info
*fsi
= sb
->s_fs_info
;
286 struct inode
*inode
= d_inode(sb
->s_root
);
287 struct tracefs_mount_opts
*opts
= &fsi
->mount_opts
;
289 inode
->i_mode
&= ~S_IALLUGO
;
290 inode
->i_mode
|= opts
->mode
;
292 inode
->i_uid
= opts
->uid
;
294 /* Set all the group ids to the mount option */
295 set_gid(sb
->s_root
, opts
->gid
);
300 static int tracefs_remount(struct super_block
*sb
, int *flags
, char *data
)
303 struct tracefs_fs_info
*fsi
= sb
->s_fs_info
;
306 err
= tracefs_parse_options(data
, &fsi
->mount_opts
);
310 tracefs_apply_options(sb
);
316 static int tracefs_show_options(struct seq_file
*m
, struct dentry
*root
)
318 struct tracefs_fs_info
*fsi
= root
->d_sb
->s_fs_info
;
319 struct tracefs_mount_opts
*opts
= &fsi
->mount_opts
;
321 if (!uid_eq(opts
->uid
, GLOBAL_ROOT_UID
))
322 seq_printf(m
, ",uid=%u",
323 from_kuid_munged(&init_user_ns
, opts
->uid
));
324 if (!gid_eq(opts
->gid
, GLOBAL_ROOT_GID
))
325 seq_printf(m
, ",gid=%u",
326 from_kgid_munged(&init_user_ns
, opts
->gid
));
327 if (opts
->mode
!= TRACEFS_DEFAULT_MODE
)
328 seq_printf(m
, ",mode=%o", opts
->mode
);
333 static const struct super_operations tracefs_super_operations
= {
334 .statfs
= simple_statfs
,
335 .remount_fs
= tracefs_remount
,
336 .show_options
= tracefs_show_options
,
339 static int trace_fill_super(struct super_block
*sb
, void *data
, int silent
)
341 static const struct tree_descr trace_files
[] = {{""}};
342 struct tracefs_fs_info
*fsi
;
345 fsi
= kzalloc(sizeof(struct tracefs_fs_info
), GFP_KERNEL
);
352 err
= tracefs_parse_options(data
, &fsi
->mount_opts
);
356 err
= simple_fill_super(sb
, TRACEFS_MAGIC
, trace_files
);
360 sb
->s_op
= &tracefs_super_operations
;
362 tracefs_apply_options(sb
);
368 sb
->s_fs_info
= NULL
;
372 static struct dentry
*trace_mount(struct file_system_type
*fs_type
,
373 int flags
, const char *dev_name
,
376 return mount_single(fs_type
, flags
, data
, trace_fill_super
);
379 static struct file_system_type trace_fs_type
= {
380 .owner
= THIS_MODULE
,
382 .mount
= trace_mount
,
383 .kill_sb
= kill_litter_super
,
385 MODULE_ALIAS_FS("tracefs");
387 static struct dentry
*start_creating(const char *name
, struct dentry
*parent
)
389 struct dentry
*dentry
;
392 pr_debug("tracefs: creating file '%s'\n",name
);
394 error
= simple_pin_fs(&trace_fs_type
, &tracefs_mount
,
395 &tracefs_mount_count
);
397 return ERR_PTR(error
);
399 /* If the parent is not specified, we create it in the root.
400 * We need the root dentry to do this, which is in the super
401 * block. A pointer to that is in the struct vfsmount that we
405 parent
= tracefs_mount
->mnt_root
;
407 inode_lock(d_inode(parent
));
408 if (unlikely(IS_DEADDIR(d_inode(parent
))))
409 dentry
= ERR_PTR(-ENOENT
);
411 dentry
= lookup_one_len(name
, parent
, strlen(name
));
412 if (!IS_ERR(dentry
) && d_inode(dentry
)) {
414 dentry
= ERR_PTR(-EEXIST
);
417 if (IS_ERR(dentry
)) {
418 inode_unlock(d_inode(parent
));
419 simple_release_fs(&tracefs_mount
, &tracefs_mount_count
);
425 static struct dentry
*failed_creating(struct dentry
*dentry
)
427 inode_unlock(d_inode(dentry
->d_parent
));
429 simple_release_fs(&tracefs_mount
, &tracefs_mount_count
);
433 static struct dentry
*end_creating(struct dentry
*dentry
)
435 inode_unlock(d_inode(dentry
->d_parent
));
440 * tracefs_create_file - create a file in the tracefs filesystem
441 * @name: a pointer to a string containing the name of the file to create.
442 * @mode: the permission that the file should have.
443 * @parent: a pointer to the parent dentry for this file. This should be a
444 * directory dentry if set. If this parameter is NULL, then the
445 * file will be created in the root of the tracefs filesystem.
446 * @data: a pointer to something that the caller will want to get to later
447 * on. The inode.i_private pointer will point to this value on
449 * @fops: a pointer to a struct file_operations that should be used for
452 * This is the basic "create a file" function for tracefs. It allows for a
453 * wide range of flexibility in creating a file, or a directory (if you want
454 * to create a directory, the tracefs_create_dir() function is
455 * recommended to be used instead.)
457 * This function will return a pointer to a dentry if it succeeds. This
458 * pointer must be passed to the tracefs_remove() function when the file is
459 * to be removed (no automatic cleanup happens if your module is unloaded,
460 * you are responsible here.) If an error occurs, %NULL will be returned.
462 * If tracefs is not enabled in the kernel, the value -%ENODEV will be
465 struct dentry
*tracefs_create_file(const char *name
, umode_t mode
,
466 struct dentry
*parent
, void *data
,
467 const struct file_operations
*fops
)
469 struct dentry
*dentry
;
472 if (security_locked_down(LOCKDOWN_TRACEFS
))
475 if (!(mode
& S_IFMT
))
477 BUG_ON(!S_ISREG(mode
));
478 dentry
= start_creating(name
, parent
);
483 inode
= tracefs_get_inode(dentry
->d_sb
);
484 if (unlikely(!inode
))
485 return failed_creating(dentry
);
487 inode
->i_mode
= mode
;
488 inode
->i_fop
= fops
? fops
: &tracefs_file_operations
;
489 inode
->i_private
= data
;
490 inode
->i_uid
= d_inode(dentry
->d_parent
)->i_uid
;
491 inode
->i_gid
= d_inode(dentry
->d_parent
)->i_gid
;
492 d_instantiate(dentry
, inode
);
493 fsnotify_create(d_inode(dentry
->d_parent
), dentry
);
494 return end_creating(dentry
);
497 static struct dentry
*__create_dir(const char *name
, struct dentry
*parent
,
498 const struct inode_operations
*ops
)
500 struct dentry
*dentry
= start_creating(name
, parent
);
506 inode
= tracefs_get_inode(dentry
->d_sb
);
507 if (unlikely(!inode
))
508 return failed_creating(dentry
);
510 /* Do not set bits for OTH */
511 inode
->i_mode
= S_IFDIR
| S_IRWXU
| S_IRUSR
| S_IRGRP
| S_IXUSR
| S_IXGRP
;
513 inode
->i_fop
= &simple_dir_operations
;
514 inode
->i_uid
= d_inode(dentry
->d_parent
)->i_uid
;
515 inode
->i_gid
= d_inode(dentry
->d_parent
)->i_gid
;
517 /* directory inodes start off with i_nlink == 2 (for "." entry) */
519 d_instantiate(dentry
, inode
);
520 inc_nlink(d_inode(dentry
->d_parent
));
521 fsnotify_mkdir(d_inode(dentry
->d_parent
), dentry
);
522 return end_creating(dentry
);
526 * tracefs_create_dir - create a directory in the tracefs filesystem
527 * @name: a pointer to a string containing the name of the directory to
529 * @parent: a pointer to the parent dentry for this file. This should be a
530 * directory dentry if set. If this parameter is NULL, then the
531 * directory will be created in the root of the tracefs filesystem.
533 * This function creates a directory in tracefs with the given name.
535 * This function will return a pointer to a dentry if it succeeds. This
536 * pointer must be passed to the tracefs_remove() function when the file is
537 * to be removed. If an error occurs, %NULL will be returned.
539 * If tracing is not enabled in the kernel, the value -%ENODEV will be
542 struct dentry
*tracefs_create_dir(const char *name
, struct dentry
*parent
)
544 return __create_dir(name
, parent
, &simple_dir_inode_operations
);
548 * tracefs_create_instance_dir - create the tracing instances directory
549 * @name: The name of the instances directory to create
550 * @parent: The parent directory that the instances directory will exist
551 * @mkdir: The function to call when a mkdir is performed.
552 * @rmdir: The function to call when a rmdir is performed.
554 * Only one instances directory is allowed.
556 * The instances directory is special as it allows for mkdir and rmdir
557 * to be done by userspace. When a mkdir or rmdir is performed, the inode
558 * locks are released and the methods passed in (@mkdir and @rmdir) are
559 * called without locks and with the name of the directory being created
560 * within the instances directory.
562 * Returns the dentry of the instances directory.
564 __init
struct dentry
*tracefs_create_instance_dir(const char *name
,
565 struct dentry
*parent
,
566 int (*mkdir
)(const char *name
),
567 int (*rmdir
)(const char *name
))
569 struct dentry
*dentry
;
571 /* Only allow one instance of the instances directory. */
572 if (WARN_ON(tracefs_ops
.mkdir
|| tracefs_ops
.rmdir
))
575 dentry
= __create_dir(name
, parent
, &tracefs_dir_inode_operations
);
579 tracefs_ops
.mkdir
= mkdir
;
580 tracefs_ops
.rmdir
= rmdir
;
585 static void remove_one(struct dentry
*victim
)
587 simple_release_fs(&tracefs_mount
, &tracefs_mount_count
);
591 * tracefs_remove - recursively removes a directory
592 * @dentry: a pointer to a the dentry of the directory to be removed.
594 * This function recursively removes a directory tree in tracefs that
595 * was previously created with a call to another tracefs function
596 * (like tracefs_create_file() or variants thereof.)
598 void tracefs_remove(struct dentry
*dentry
)
600 if (IS_ERR_OR_NULL(dentry
))
603 simple_pin_fs(&trace_fs_type
, &tracefs_mount
, &tracefs_mount_count
);
604 simple_recursive_removal(dentry
, remove_one
);
605 simple_release_fs(&tracefs_mount
, &tracefs_mount_count
);
609 * tracefs_initialized - Tells whether tracefs has been registered
611 bool tracefs_initialized(void)
613 return tracefs_registered
;
616 static int __init
tracefs_init(void)
620 retval
= sysfs_create_mount_point(kernel_kobj
, "tracing");
624 retval
= register_filesystem(&trace_fs_type
);
626 tracefs_registered
= true;
630 core_initcall(tracefs_init
);