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/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>
26 #define TRACEFS_DEFAULT_MODE 0700
27 static struct kmem_cache
*tracefs_inode_cachep __ro_after_init
;
29 static struct vfsmount
*tracefs_mount
;
30 static int tracefs_mount_count
;
31 static bool tracefs_registered
;
34 * Keep track of all tracefs_inodes in order to update their
35 * flags if necessary on a remount.
37 static DEFINE_SPINLOCK(tracefs_inode_lock
);
38 static LIST_HEAD(tracefs_inodes
);
40 static struct inode
*tracefs_alloc_inode(struct super_block
*sb
)
42 struct tracefs_inode
*ti
;
45 ti
= alloc_inode_sb(sb
, tracefs_inode_cachep
, GFP_KERNEL
);
49 spin_lock_irqsave(&tracefs_inode_lock
, flags
);
50 list_add_rcu(&ti
->list
, &tracefs_inodes
);
51 spin_unlock_irqrestore(&tracefs_inode_lock
, flags
);
53 return &ti
->vfs_inode
;
56 static void tracefs_free_inode(struct inode
*inode
)
58 struct tracefs_inode
*ti
= get_tracefs(inode
);
60 kmem_cache_free(tracefs_inode_cachep
, ti
);
63 static void tracefs_destroy_inode(struct inode
*inode
)
65 struct tracefs_inode
*ti
= get_tracefs(inode
);
68 spin_lock_irqsave(&tracefs_inode_lock
, flags
);
69 list_del_rcu(&ti
->list
);
70 spin_unlock_irqrestore(&tracefs_inode_lock
, flags
);
73 static ssize_t
default_read_file(struct file
*file
, char __user
*buf
,
74 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
)
85 static const struct file_operations tracefs_file_operations
= {
86 .read
= default_read_file
,
87 .write
= default_write_file
,
89 .llseek
= noop_llseek
,
92 static struct tracefs_dir_ops
{
93 int (*mkdir
)(const char *name
);
94 int (*rmdir
)(const char *name
);
95 } tracefs_ops __ro_after_init
;
97 static char *get_dname(struct dentry
*dentry
)
101 int len
= dentry
->d_name
.len
;
103 dname
= dentry
->d_name
.name
;
104 name
= kmalloc(len
+ 1, GFP_KERNEL
);
107 memcpy(name
, dname
, len
);
112 static int tracefs_syscall_mkdir(struct mnt_idmap
*idmap
,
113 struct inode
*inode
, struct dentry
*dentry
,
116 struct tracefs_inode
*ti
;
120 name
= get_dname(dentry
);
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.
129 ti
= get_tracefs(inode
);
130 ti
->flags
|= TRACEFS_INSTANCE_INODE
;
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.
139 ret
= tracefs_ops
.mkdir(name
);
147 static int tracefs_syscall_rmdir(struct inode
*inode
, struct dentry
*dentry
)
152 name
= get_dname(dentry
);
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.
164 inode_unlock(d_inode(dentry
));
166 ret
= tracefs_ops
.rmdir(name
);
168 inode_lock_nested(inode
, I_MUTEX_PARENT
);
169 inode_lock(d_inode(dentry
));
176 static void set_tracefs_inode_owner(struct inode
*inode
)
178 struct tracefs_inode
*ti
= get_tracefs(inode
);
179 struct inode
*root_inode
= ti
->private;
183 uid
= root_inode
->i_uid
;
184 gid
= root_inode
->i_gid
;
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
191 if (root_inode
!= d_inode(root_inode
->i_sb
->s_root
)) {
192 struct tracefs_inode
*rti
;
194 rti
= get_tracefs(root_inode
);
195 root_inode
= d_inode(root_inode
->i_sb
->s_root
);
197 if (!(rti
->flags
& TRACEFS_UID_PERM_SET
))
198 uid
= root_inode
->i_uid
;
200 if (!(rti
->flags
& TRACEFS_GID_PERM_SET
))
201 gid
= root_inode
->i_gid
;
205 * If this inode has never been referenced, then update
206 * the permissions to the superblock.
208 if (!(ti
->flags
& TRACEFS_UID_PERM_SET
))
211 if (!(ti
->flags
& TRACEFS_GID_PERM_SET
))
215 static int tracefs_permission(struct mnt_idmap
*idmap
,
216 struct inode
*inode
, int mask
)
218 set_tracefs_inode_owner(inode
);
219 return generic_permission(idmap
, inode
, mask
);
222 static int tracefs_getattr(struct mnt_idmap
*idmap
,
223 const struct path
*path
, struct kstat
*stat
,
224 u32 request_mask
, unsigned int flags
)
226 struct inode
*inode
= d_backing_inode(path
->dentry
);
228 set_tracefs_inode_owner(inode
);
229 generic_fillattr(idmap
, request_mask
, inode
, stat
);
233 static int tracefs_setattr(struct mnt_idmap
*idmap
, struct dentry
*dentry
,
236 unsigned int ia_valid
= attr
->ia_valid
;
237 struct inode
*inode
= d_inode(dentry
);
238 struct tracefs_inode
*ti
= get_tracefs(inode
);
240 if (ia_valid
& ATTR_UID
)
241 ti
->flags
|= TRACEFS_UID_PERM_SET
;
243 if (ia_valid
& ATTR_GID
)
244 ti
->flags
|= TRACEFS_GID_PERM_SET
;
246 return simple_setattr(idmap
, dentry
, attr
);
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
,
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
,
265 static const struct inode_operations tracefs_file_inode_operations
= {
266 .permission
= tracefs_permission
,
267 .getattr
= tracefs_getattr
,
268 .setattr
= tracefs_setattr
,
271 struct inode
*tracefs_get_inode(struct super_block
*sb
)
273 struct inode
*inode
= new_inode(sb
);
275 inode
->i_ino
= get_next_ino();
276 simple_inode_init_ts(inode
);
281 struct tracefs_fs_info
{
285 /* Opt_* bitfield. */
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
),
302 static int tracefs_parse_param(struct fs_context
*fc
, struct fs_parameter
*param
)
304 struct tracefs_fs_info
*opts
= fc
->s_fs_info
;
305 struct fs_parse_result result
;
308 opt
= fs_parse(fc
, tracefs_param_specs
, param
, &result
);
314 opts
->uid
= result
.uid
;
317 opts
->gid
= result
.gid
;
320 opts
->mode
= result
.uint_32
& S_IALLUGO
;
323 * We might like to report bad mount options here;
324 * but traditionally tracefs has ignored all mount options
328 opts
->opts
|= BIT(opt
);
333 static int tracefs_apply_options(struct super_block
*sb
, bool remount
)
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
;
342 * On remount, only reset mode/uid/gid if they were provided as mount
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
);
352 if (!remount
|| fsi
->opts
& BIT(Opt_uid
))
353 inode
->i_uid
= fsi
->uid
;
355 if (!remount
|| fsi
->opts
& BIT(Opt_gid
))
356 inode
->i_gid
= fsi
->gid
;
358 if (remount
&& (fsi
->opts
& BIT(Opt_uid
) || fsi
->opts
& BIT(Opt_gid
))) {
360 update_uid
= fsi
->opts
& BIT(Opt_uid
);
361 update_gid
= fsi
->opts
& BIT(Opt_gid
);
364 list_for_each_entry_rcu(ti
, &tracefs_inodes
, list
) {
366 ti
->flags
&= ~TRACEFS_UID_PERM_SET
;
367 ti
->vfs_inode
.i_uid
= fsi
->uid
;
371 ti
->flags
&= ~TRACEFS_GID_PERM_SET
;
372 ti
->vfs_inode
.i_gid
= fsi
->gid
;
376 * Note, the above ti->vfs_inode updates are
377 * used in eventfs_remount() so they must come
380 if (ti
->flags
& TRACEFS_EVENT_INODE
)
381 eventfs_remount(ti
, update_uid
, update_gid
);
389 static int tracefs_reconfigure(struct fs_context
*fc
)
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
;
399 /* structure copy of new mount options to sb */
400 *sb_opts
= *new_opts
;
402 return tracefs_apply_options(sb
, true);
405 static int tracefs_show_options(struct seq_file
*m
, struct dentry
*root
)
407 struct tracefs_fs_info
*fsi
= root
->d_sb
->s_fs_info
;
409 if (!uid_eq(fsi
->uid
, GLOBAL_ROOT_UID
))
410 seq_printf(m
, ",uid=%u",
411 from_kuid_munged(&init_user_ns
, fsi
->uid
));
412 if (!gid_eq(fsi
->gid
, GLOBAL_ROOT_GID
))
413 seq_printf(m
, ",gid=%u",
414 from_kgid_munged(&init_user_ns
, fsi
->gid
));
415 if (fsi
->mode
!= TRACEFS_DEFAULT_MODE
)
416 seq_printf(m
, ",mode=%o", fsi
->mode
);
421 static int tracefs_drop_inode(struct inode
*inode
)
423 struct tracefs_inode
*ti
= get_tracefs(inode
);
426 * This inode is being freed and cannot be used for
427 * eventfs. Clear the flag so that it doesn't call into
428 * eventfs during the remount flag updates. The eventfs_inode
429 * gets freed after an RCU cycle, so the content will still
430 * be safe if the iteration is going on now.
432 ti
->flags
&= ~TRACEFS_EVENT_INODE
;
437 static const struct super_operations tracefs_super_operations
= {
438 .alloc_inode
= tracefs_alloc_inode
,
439 .free_inode
= tracefs_free_inode
,
440 .destroy_inode
= tracefs_destroy_inode
,
441 .drop_inode
= tracefs_drop_inode
,
442 .statfs
= simple_statfs
,
443 .show_options
= tracefs_show_options
,
447 * It would be cleaner if eventfs had its own dentry ops.
449 * Note that d_revalidate is called potentially under RCU,
450 * so it can't take the eventfs mutex etc. It's fine - if
451 * we open a file just as it's marked dead, things will
452 * still work just fine, and just see the old stale case.
454 static void tracefs_d_release(struct dentry
*dentry
)
456 if (dentry
->d_fsdata
)
457 eventfs_d_release(dentry
);
460 static int tracefs_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
462 struct eventfs_inode
*ei
= dentry
->d_fsdata
;
464 return !(ei
&& ei
->is_freed
);
467 static const struct dentry_operations tracefs_dentry_operations
= {
468 .d_revalidate
= tracefs_d_revalidate
,
469 .d_release
= tracefs_d_release
,
472 static int tracefs_fill_super(struct super_block
*sb
, struct fs_context
*fc
)
474 static const struct tree_descr trace_files
[] = {{""}};
477 err
= simple_fill_super(sb
, TRACEFS_MAGIC
, trace_files
);
481 sb
->s_op
= &tracefs_super_operations
;
482 sb
->s_d_op
= &tracefs_dentry_operations
;
487 static int tracefs_get_tree(struct fs_context
*fc
)
489 int err
= get_tree_single(fc
, tracefs_fill_super
);
494 return tracefs_reconfigure(fc
);
497 static void tracefs_free_fc(struct fs_context
*fc
)
499 kfree(fc
->s_fs_info
);
502 static const struct fs_context_operations tracefs_context_ops
= {
503 .free
= tracefs_free_fc
,
504 .parse_param
= tracefs_parse_param
,
505 .get_tree
= tracefs_get_tree
,
506 .reconfigure
= tracefs_reconfigure
,
509 static int tracefs_init_fs_context(struct fs_context
*fc
)
511 struct tracefs_fs_info
*fsi
;
513 fsi
= kzalloc(sizeof(struct tracefs_fs_info
), GFP_KERNEL
);
517 fsi
->mode
= TRACEFS_DEFAULT_MODE
;
520 fc
->ops
= &tracefs_context_ops
;
524 static struct file_system_type trace_fs_type
= {
525 .owner
= THIS_MODULE
,
527 .init_fs_context
= tracefs_init_fs_context
,
528 .parameters
= tracefs_param_specs
,
529 .kill_sb
= kill_litter_super
,
531 MODULE_ALIAS_FS("tracefs");
533 struct dentry
*tracefs_start_creating(const char *name
, struct dentry
*parent
)
535 struct dentry
*dentry
;
538 pr_debug("tracefs: creating file '%s'\n",name
);
540 error
= simple_pin_fs(&trace_fs_type
, &tracefs_mount
,
541 &tracefs_mount_count
);
543 return ERR_PTR(error
);
545 /* If the parent is not specified, we create it in the root.
546 * We need the root dentry to do this, which is in the super
547 * block. A pointer to that is in the struct vfsmount that we
551 parent
= tracefs_mount
->mnt_root
;
553 inode_lock(d_inode(parent
));
554 if (unlikely(IS_DEADDIR(d_inode(parent
))))
555 dentry
= ERR_PTR(-ENOENT
);
557 dentry
= lookup_one_len(name
, parent
, strlen(name
));
558 if (!IS_ERR(dentry
) && d_inode(dentry
)) {
560 dentry
= ERR_PTR(-EEXIST
);
563 if (IS_ERR(dentry
)) {
564 inode_unlock(d_inode(parent
));
565 simple_release_fs(&tracefs_mount
, &tracefs_mount_count
);
571 struct dentry
*tracefs_failed_creating(struct dentry
*dentry
)
573 inode_unlock(d_inode(dentry
->d_parent
));
575 simple_release_fs(&tracefs_mount
, &tracefs_mount_count
);
579 struct dentry
*tracefs_end_creating(struct dentry
*dentry
)
581 inode_unlock(d_inode(dentry
->d_parent
));
585 /* Find the inode that this will use for default */
586 static struct inode
*instance_inode(struct dentry
*parent
, struct inode
*inode
)
588 struct tracefs_inode
*ti
;
590 /* If parent is NULL then use root inode */
592 return d_inode(inode
->i_sb
->s_root
);
594 /* Find the inode that is flagged as an instance or the root inode */
595 while (!IS_ROOT(parent
)) {
596 ti
= get_tracefs(d_inode(parent
));
597 if (ti
->flags
& TRACEFS_INSTANCE_INODE
)
599 parent
= parent
->d_parent
;
602 return d_inode(parent
);
606 * tracefs_create_file - create a file in the tracefs filesystem
607 * @name: a pointer to a string containing the name of the file to create.
608 * @mode: the permission that the file should have.
609 * @parent: a pointer to the parent dentry for this file. This should be a
610 * directory dentry if set. If this parameter is NULL, then the
611 * file will be created in the root of the tracefs filesystem.
612 * @data: a pointer to something that the caller will want to get to later
613 * on. The inode.i_private pointer will point to this value on
615 * @fops: a pointer to a struct file_operations that should be used for
618 * This is the basic "create a file" function for tracefs. It allows for a
619 * wide range of flexibility in creating a file, or a directory (if you want
620 * to create a directory, the tracefs_create_dir() function is
621 * recommended to be used instead.)
623 * This function will return a pointer to a dentry if it succeeds. This
624 * pointer must be passed to the tracefs_remove() function when the file is
625 * to be removed (no automatic cleanup happens if your module is unloaded,
626 * you are responsible here.) If an error occurs, %NULL will be returned.
628 * If tracefs is not enabled in the kernel, the value -%ENODEV will be
631 struct dentry
*tracefs_create_file(const char *name
, umode_t mode
,
632 struct dentry
*parent
, void *data
,
633 const struct file_operations
*fops
)
635 struct tracefs_inode
*ti
;
636 struct dentry
*dentry
;
639 if (security_locked_down(LOCKDOWN_TRACEFS
))
642 if (!(mode
& S_IFMT
))
644 BUG_ON(!S_ISREG(mode
));
645 dentry
= tracefs_start_creating(name
, parent
);
650 inode
= tracefs_get_inode(dentry
->d_sb
);
651 if (unlikely(!inode
))
652 return tracefs_failed_creating(dentry
);
654 ti
= get_tracefs(inode
);
655 ti
->private = instance_inode(parent
, inode
);
657 inode
->i_mode
= mode
;
658 inode
->i_op
= &tracefs_file_inode_operations
;
659 inode
->i_fop
= fops
? fops
: &tracefs_file_operations
;
660 inode
->i_private
= data
;
661 inode
->i_uid
= d_inode(dentry
->d_parent
)->i_uid
;
662 inode
->i_gid
= d_inode(dentry
->d_parent
)->i_gid
;
663 d_instantiate(dentry
, inode
);
664 fsnotify_create(d_inode(dentry
->d_parent
), dentry
);
665 return tracefs_end_creating(dentry
);
668 static struct dentry
*__create_dir(const char *name
, struct dentry
*parent
,
669 const struct inode_operations
*ops
)
671 struct tracefs_inode
*ti
;
672 struct dentry
*dentry
= tracefs_start_creating(name
, parent
);
678 inode
= tracefs_get_inode(dentry
->d_sb
);
679 if (unlikely(!inode
))
680 return tracefs_failed_creating(dentry
);
682 /* Do not set bits for OTH */
683 inode
->i_mode
= S_IFDIR
| S_IRWXU
| S_IRUSR
| S_IRGRP
| S_IXUSR
| S_IXGRP
;
685 inode
->i_fop
= &simple_dir_operations
;
686 inode
->i_uid
= d_inode(dentry
->d_parent
)->i_uid
;
687 inode
->i_gid
= d_inode(dentry
->d_parent
)->i_gid
;
689 ti
= get_tracefs(inode
);
690 ti
->private = instance_inode(parent
, inode
);
692 /* directory inodes start off with i_nlink == 2 (for "." entry) */
694 d_instantiate(dentry
, inode
);
695 inc_nlink(d_inode(dentry
->d_parent
));
696 fsnotify_mkdir(d_inode(dentry
->d_parent
), dentry
);
697 return tracefs_end_creating(dentry
);
701 * tracefs_create_dir - create a directory in the tracefs filesystem
702 * @name: a pointer to a string containing the name of the directory to
704 * @parent: a pointer to the parent dentry for this file. This should be a
705 * directory dentry if set. If this parameter is NULL, then the
706 * directory will be created in the root of the tracefs filesystem.
708 * This function creates a directory in tracefs with the given name.
710 * This function will return a pointer to a dentry if it succeeds. This
711 * pointer must be passed to the tracefs_remove() function when the file is
712 * to be removed. If an error occurs, %NULL will be returned.
714 * If tracing is not enabled in the kernel, the value -%ENODEV will be
717 struct dentry
*tracefs_create_dir(const char *name
, struct dentry
*parent
)
719 if (security_locked_down(LOCKDOWN_TRACEFS
))
722 return __create_dir(name
, parent
, &tracefs_dir_inode_operations
);
726 * tracefs_create_instance_dir - create the tracing instances directory
727 * @name: The name of the instances directory to create
728 * @parent: The parent directory that the instances directory will exist
729 * @mkdir: The function to call when a mkdir is performed.
730 * @rmdir: The function to call when a rmdir is performed.
732 * Only one instances directory is allowed.
734 * The instances directory is special as it allows for mkdir and rmdir
735 * to be done by userspace. When a mkdir or rmdir is performed, the inode
736 * locks are released and the methods passed in (@mkdir and @rmdir) are
737 * called without locks and with the name of the directory being created
738 * within the instances directory.
740 * Returns the dentry of the instances directory.
742 __init
struct dentry
*tracefs_create_instance_dir(const char *name
,
743 struct dentry
*parent
,
744 int (*mkdir
)(const char *name
),
745 int (*rmdir
)(const char *name
))
747 struct dentry
*dentry
;
749 /* Only allow one instance of the instances directory. */
750 if (WARN_ON(tracefs_ops
.mkdir
|| tracefs_ops
.rmdir
))
753 dentry
= __create_dir(name
, parent
, &tracefs_instance_dir_inode_operations
);
757 tracefs_ops
.mkdir
= mkdir
;
758 tracefs_ops
.rmdir
= rmdir
;
763 static void remove_one(struct dentry
*victim
)
765 simple_release_fs(&tracefs_mount
, &tracefs_mount_count
);
769 * tracefs_remove - recursively removes a directory
770 * @dentry: a pointer to a the dentry of the directory to be removed.
772 * This function recursively removes a directory tree in tracefs that
773 * was previously created with a call to another tracefs function
774 * (like tracefs_create_file() or variants thereof.)
776 void tracefs_remove(struct dentry
*dentry
)
778 if (IS_ERR_OR_NULL(dentry
))
781 simple_pin_fs(&trace_fs_type
, &tracefs_mount
, &tracefs_mount_count
);
782 simple_recursive_removal(dentry
, remove_one
);
783 simple_release_fs(&tracefs_mount
, &tracefs_mount_count
);
787 * tracefs_initialized - Tells whether tracefs has been registered
789 bool tracefs_initialized(void)
791 return tracefs_registered
;
794 static void init_once(void *foo
)
796 struct tracefs_inode
*ti
= (struct tracefs_inode
*) foo
;
798 /* inode_init_once() calls memset() on the vfs_inode portion */
799 inode_init_once(&ti
->vfs_inode
);
801 /* Zero out the rest */
802 memset_after(ti
, 0, vfs_inode
);
805 static int __init
tracefs_init(void)
809 tracefs_inode_cachep
= kmem_cache_create("tracefs_inode_cache",
810 sizeof(struct tracefs_inode
),
811 0, (SLAB_RECLAIM_ACCOUNT
|
814 if (!tracefs_inode_cachep
)
817 retval
= sysfs_create_mount_point(kernel_kobj
, "tracing");
821 retval
= register_filesystem(&trace_fs_type
);
823 tracefs_registered
= true;
827 core_initcall(tracefs_init
);