1 // SPDX-License-Identifier: GPL-2.0-only
3 * proc/fs/generic.c --- generic routines for the proc-fs
5 * This file contains generic proc-fs routines for handling
6 * directories and files.
8 * Copyright (C) 1991, 1992 Linus Torvalds.
9 * Copyright (C) 1997 Theodore Ts'o
12 #include <linux/cache.h>
13 #include <linux/errno.h>
14 #include <linux/time.h>
15 #include <linux/proc_fs.h>
16 #include <linux/stat.h>
18 #include <linux/module.h>
19 #include <linux/namei.h>
20 #include <linux/slab.h>
21 #include <linux/printk.h>
22 #include <linux/mount.h>
23 #include <linux/init.h>
24 #include <linux/idr.h>
25 #include <linux/bitops.h>
26 #include <linux/spinlock.h>
27 #include <linux/completion.h>
28 #include <linux/uaccess.h>
29 #include <linux/seq_file.h>
33 static DEFINE_RWLOCK(proc_subdir_lock
);
35 struct kmem_cache
*proc_dir_entry_cache __ro_after_init
;
37 void pde_free(struct proc_dir_entry
*pde
)
39 if (S_ISLNK(pde
->mode
))
41 if (pde
->name
!= pde
->inline_name
)
43 kmem_cache_free(proc_dir_entry_cache
, pde
);
46 static int proc_match(const char *name
, struct proc_dir_entry
*de
, unsigned int len
)
48 if (len
< de
->namelen
)
50 if (len
> de
->namelen
)
53 return memcmp(name
, de
->name
, len
);
56 static struct proc_dir_entry
*pde_subdir_first(struct proc_dir_entry
*dir
)
58 return rb_entry_safe(rb_first(&dir
->subdir
), struct proc_dir_entry
,
62 static struct proc_dir_entry
*pde_subdir_next(struct proc_dir_entry
*dir
)
64 return rb_entry_safe(rb_next(&dir
->subdir_node
), struct proc_dir_entry
,
68 static struct proc_dir_entry
*pde_subdir_find(struct proc_dir_entry
*dir
,
72 struct rb_node
*node
= dir
->subdir
.rb_node
;
75 struct proc_dir_entry
*de
= rb_entry(node
,
76 struct proc_dir_entry
,
78 int result
= proc_match(name
, de
, len
);
83 node
= node
->rb_right
;
90 static bool pde_subdir_insert(struct proc_dir_entry
*dir
,
91 struct proc_dir_entry
*de
)
93 struct rb_root
*root
= &dir
->subdir
;
94 struct rb_node
**new = &root
->rb_node
, *parent
= NULL
;
96 /* Figure out where to put new node */
98 struct proc_dir_entry
*this = rb_entry(*new,
99 struct proc_dir_entry
,
101 int result
= proc_match(de
->name
, this, de
->namelen
);
105 new = &(*new)->rb_left
;
107 new = &(*new)->rb_right
;
112 /* Add new node and rebalance tree. */
113 rb_link_node(&de
->subdir_node
, parent
, new);
114 rb_insert_color(&de
->subdir_node
, root
);
118 static int proc_notify_change(struct dentry
*dentry
, struct iattr
*iattr
)
120 struct inode
*inode
= d_inode(dentry
);
121 struct proc_dir_entry
*de
= PDE(inode
);
124 error
= setattr_prepare(dentry
, iattr
);
128 setattr_copy(inode
, iattr
);
129 mark_inode_dirty(inode
);
131 proc_set_user(de
, inode
->i_uid
, inode
->i_gid
);
132 de
->mode
= inode
->i_mode
;
136 static int proc_getattr(const struct path
*path
, struct kstat
*stat
,
137 u32 request_mask
, unsigned int query_flags
)
139 struct inode
*inode
= d_inode(path
->dentry
);
140 struct proc_dir_entry
*de
= PDE(inode
);
142 nlink_t nlink
= READ_ONCE(de
->nlink
);
144 set_nlink(inode
, nlink
);
148 generic_fillattr(inode
, stat
);
152 static const struct inode_operations proc_file_inode_operations
= {
153 .setattr
= proc_notify_change
,
157 * This function parses a name such as "tty/driver/serial", and
158 * returns the struct proc_dir_entry for "/proc/tty/driver", and
159 * returns "serial" in residual.
161 static int __xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
162 const char **residual
)
164 const char *cp
= name
, *next
;
165 struct proc_dir_entry
*de
;
172 next
= strchr(cp
, '/');
176 de
= pde_subdir_find(de
, cp
, next
- cp
);
178 WARN(1, "name '%s'\n", name
);
188 static int xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
189 const char **residual
)
193 read_lock(&proc_subdir_lock
);
194 rv
= __xlate_proc_name(name
, ret
, residual
);
195 read_unlock(&proc_subdir_lock
);
199 static DEFINE_IDA(proc_inum_ida
);
201 #define PROC_DYNAMIC_FIRST 0xF0000000U
204 * Return an inode number between PROC_DYNAMIC_FIRST and
205 * 0xffffffff, or zero on failure.
207 int proc_alloc_inum(unsigned int *inum
)
211 i
= ida_simple_get(&proc_inum_ida
, 0, UINT_MAX
- PROC_DYNAMIC_FIRST
+ 1,
216 *inum
= PROC_DYNAMIC_FIRST
+ (unsigned int)i
;
220 void proc_free_inum(unsigned int inum
)
222 ida_simple_remove(&proc_inum_ida
, inum
- PROC_DYNAMIC_FIRST
);
225 static int proc_misc_d_revalidate(struct dentry
*dentry
, unsigned int flags
)
227 if (flags
& LOOKUP_RCU
)
230 if (atomic_read(&PDE(d_inode(dentry
))->in_use
) < 0)
231 return 0; /* revalidate */
235 static int proc_misc_d_delete(const struct dentry
*dentry
)
237 return atomic_read(&PDE(d_inode(dentry
))->in_use
) < 0;
240 static const struct dentry_operations proc_misc_dentry_ops
= {
241 .d_revalidate
= proc_misc_d_revalidate
,
242 .d_delete
= proc_misc_d_delete
,
246 * Don't create negative dentries here, return -ENOENT by hand
249 struct dentry
*proc_lookup_de(struct inode
*dir
, struct dentry
*dentry
,
250 struct proc_dir_entry
*de
)
254 read_lock(&proc_subdir_lock
);
255 de
= pde_subdir_find(de
, dentry
->d_name
.name
, dentry
->d_name
.len
);
258 read_unlock(&proc_subdir_lock
);
259 inode
= proc_get_inode(dir
->i_sb
, de
);
261 return ERR_PTR(-ENOMEM
);
262 d_set_d_op(dentry
, de
->proc_dops
);
263 return d_splice_alias(inode
, dentry
);
265 read_unlock(&proc_subdir_lock
);
266 return ERR_PTR(-ENOENT
);
269 struct dentry
*proc_lookup(struct inode
*dir
, struct dentry
*dentry
,
272 return proc_lookup_de(dir
, dentry
, PDE(dir
));
276 * This returns non-zero if at EOF, so that the /proc
277 * root directory can use this and check if it should
278 * continue with the <pid> entries..
280 * Note that the VFS-layer doesn't care about the return
281 * value of the readdir() call, as long as it's non-negative
284 int proc_readdir_de(struct file
*file
, struct dir_context
*ctx
,
285 struct proc_dir_entry
*de
)
289 if (!dir_emit_dots(file
, ctx
))
293 read_lock(&proc_subdir_lock
);
294 de
= pde_subdir_first(de
);
297 read_unlock(&proc_subdir_lock
);
302 de
= pde_subdir_next(de
);
307 struct proc_dir_entry
*next
;
309 read_unlock(&proc_subdir_lock
);
310 if (!dir_emit(ctx
, de
->name
, de
->namelen
,
311 de
->low_ino
, de
->mode
>> 12)) {
316 read_lock(&proc_subdir_lock
);
317 next
= pde_subdir_next(de
);
321 read_unlock(&proc_subdir_lock
);
325 int proc_readdir(struct file
*file
, struct dir_context
*ctx
)
327 struct inode
*inode
= file_inode(file
);
329 return proc_readdir_de(file
, ctx
, PDE(inode
));
333 * These are the generic /proc directory operations. They
334 * use the in-memory "struct proc_dir_entry" tree to parse
335 * the /proc directory.
337 static const struct file_operations proc_dir_operations
= {
338 .llseek
= generic_file_llseek
,
339 .read
= generic_read_dir
,
340 .iterate_shared
= proc_readdir
,
344 * proc directories can do almost nothing..
346 static const struct inode_operations proc_dir_inode_operations
= {
347 .lookup
= proc_lookup
,
348 .getattr
= proc_getattr
,
349 .setattr
= proc_notify_change
,
352 /* returns the registered entry, or frees dp and returns NULL on failure */
353 struct proc_dir_entry
*proc_register(struct proc_dir_entry
*dir
,
354 struct proc_dir_entry
*dp
)
356 if (proc_alloc_inum(&dp
->low_ino
))
359 write_lock(&proc_subdir_lock
);
361 if (pde_subdir_insert(dir
, dp
) == false) {
362 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
363 dir
->name
, dp
->name
);
364 write_unlock(&proc_subdir_lock
);
368 write_unlock(&proc_subdir_lock
);
372 proc_free_inum(dp
->low_ino
);
378 static struct proc_dir_entry
*__proc_create(struct proc_dir_entry
**parent
,
383 struct proc_dir_entry
*ent
= NULL
;
387 if (xlate_proc_name(name
, parent
, &fn
) != 0)
390 qstr
.len
= strlen(fn
);
391 if (qstr
.len
== 0 || qstr
.len
>= 256) {
392 WARN(1, "name len %u\n", qstr
.len
);
395 if (qstr
.len
== 1 && fn
[0] == '.') {
396 WARN(1, "name '.'\n");
399 if (qstr
.len
== 2 && fn
[0] == '.' && fn
[1] == '.') {
400 WARN(1, "name '..'\n");
403 if (*parent
== &proc_root
&& name_to_int(&qstr
) != ~0U) {
404 WARN(1, "create '/proc/%s' by hand\n", qstr
.name
);
407 if (is_empty_pde(*parent
)) {
408 WARN(1, "attempt to add to permanently empty directory");
412 ent
= kmem_cache_zalloc(proc_dir_entry_cache
, GFP_KERNEL
);
416 if (qstr
.len
+ 1 <= SIZEOF_PDE_INLINE_NAME
) {
417 ent
->name
= ent
->inline_name
;
419 ent
->name
= kmalloc(qstr
.len
+ 1, GFP_KERNEL
);
426 memcpy(ent
->name
, fn
, qstr
.len
+ 1);
427 ent
->namelen
= qstr
.len
;
430 ent
->subdir
= RB_ROOT
;
431 refcount_set(&ent
->refcnt
, 1);
432 spin_lock_init(&ent
->pde_unload_lock
);
433 INIT_LIST_HEAD(&ent
->pde_openers
);
434 proc_set_user(ent
, (*parent
)->uid
, (*parent
)->gid
);
436 ent
->proc_dops
= &proc_misc_dentry_ops
;
442 struct proc_dir_entry
*proc_symlink(const char *name
,
443 struct proc_dir_entry
*parent
, const char *dest
)
445 struct proc_dir_entry
*ent
;
447 ent
= __proc_create(&parent
, name
,
448 (S_IFLNK
| S_IRUGO
| S_IWUGO
| S_IXUGO
),1);
451 ent
->data
= kmalloc((ent
->size
=strlen(dest
))+1, GFP_KERNEL
);
453 strcpy((char*)ent
->data
,dest
);
454 ent
->proc_iops
= &proc_link_inode_operations
;
455 ent
= proc_register(parent
, ent
);
463 EXPORT_SYMBOL(proc_symlink
);
465 struct proc_dir_entry
*proc_mkdir_data(const char *name
, umode_t mode
,
466 struct proc_dir_entry
*parent
, void *data
)
468 struct proc_dir_entry
*ent
;
471 mode
= S_IRUGO
| S_IXUGO
;
473 ent
= __proc_create(&parent
, name
, S_IFDIR
| mode
, 2);
476 ent
->proc_dir_ops
= &proc_dir_operations
;
477 ent
->proc_iops
= &proc_dir_inode_operations
;
478 ent
= proc_register(parent
, ent
);
482 EXPORT_SYMBOL_GPL(proc_mkdir_data
);
484 struct proc_dir_entry
*proc_mkdir_mode(const char *name
, umode_t mode
,
485 struct proc_dir_entry
*parent
)
487 return proc_mkdir_data(name
, mode
, parent
, NULL
);
489 EXPORT_SYMBOL(proc_mkdir_mode
);
491 struct proc_dir_entry
*proc_mkdir(const char *name
,
492 struct proc_dir_entry
*parent
)
494 return proc_mkdir_data(name
, 0, parent
, NULL
);
496 EXPORT_SYMBOL(proc_mkdir
);
498 struct proc_dir_entry
*proc_create_mount_point(const char *name
)
500 umode_t mode
= S_IFDIR
| S_IRUGO
| S_IXUGO
;
501 struct proc_dir_entry
*ent
, *parent
= NULL
;
503 ent
= __proc_create(&parent
, name
, mode
, 2);
506 ent
->proc_dir_ops
= NULL
;
507 ent
->proc_iops
= NULL
;
508 ent
= proc_register(parent
, ent
);
512 EXPORT_SYMBOL(proc_create_mount_point
);
514 struct proc_dir_entry
*proc_create_reg(const char *name
, umode_t mode
,
515 struct proc_dir_entry
**parent
, void *data
)
517 struct proc_dir_entry
*p
;
519 if ((mode
& S_IFMT
) == 0)
521 if ((mode
& S_IALLUGO
) == 0)
523 if (WARN_ON_ONCE(!S_ISREG(mode
)))
526 p
= __proc_create(parent
, name
, mode
, 1);
528 p
->proc_iops
= &proc_file_inode_operations
;
534 static inline void pde_set_flags(struct proc_dir_entry
*pde
)
536 if (pde
->proc_ops
->proc_flags
& PROC_ENTRY_PERMANENT
)
537 pde
->flags
|= PROC_ENTRY_PERMANENT
;
540 struct proc_dir_entry
*proc_create_data(const char *name
, umode_t mode
,
541 struct proc_dir_entry
*parent
,
542 const struct proc_ops
*proc_ops
, void *data
)
544 struct proc_dir_entry
*p
;
546 p
= proc_create_reg(name
, mode
, &parent
, data
);
549 p
->proc_ops
= proc_ops
;
551 return proc_register(parent
, p
);
553 EXPORT_SYMBOL(proc_create_data
);
555 struct proc_dir_entry
*proc_create(const char *name
, umode_t mode
,
556 struct proc_dir_entry
*parent
,
557 const struct proc_ops
*proc_ops
)
559 return proc_create_data(name
, mode
, parent
, proc_ops
, NULL
);
561 EXPORT_SYMBOL(proc_create
);
563 static int proc_seq_open(struct inode
*inode
, struct file
*file
)
565 struct proc_dir_entry
*de
= PDE(inode
);
568 return seq_open_private(file
, de
->seq_ops
, de
->state_size
);
569 return seq_open(file
, de
->seq_ops
);
572 static int proc_seq_release(struct inode
*inode
, struct file
*file
)
574 struct proc_dir_entry
*de
= PDE(inode
);
577 return seq_release_private(inode
, file
);
578 return seq_release(inode
, file
);
581 static const struct proc_ops proc_seq_ops
= {
582 /* not permanent -- can call into arbitrary seq_operations */
583 .proc_open
= proc_seq_open
,
584 .proc_read
= seq_read
,
585 .proc_lseek
= seq_lseek
,
586 .proc_release
= proc_seq_release
,
589 struct proc_dir_entry
*proc_create_seq_private(const char *name
, umode_t mode
,
590 struct proc_dir_entry
*parent
, const struct seq_operations
*ops
,
591 unsigned int state_size
, void *data
)
593 struct proc_dir_entry
*p
;
595 p
= proc_create_reg(name
, mode
, &parent
, data
);
598 p
->proc_ops
= &proc_seq_ops
;
600 p
->state_size
= state_size
;
601 return proc_register(parent
, p
);
603 EXPORT_SYMBOL(proc_create_seq_private
);
605 static int proc_single_open(struct inode
*inode
, struct file
*file
)
607 struct proc_dir_entry
*de
= PDE(inode
);
609 return single_open(file
, de
->single_show
, de
->data
);
612 static const struct proc_ops proc_single_ops
= {
613 /* not permanent -- can call into arbitrary ->single_show */
614 .proc_open
= proc_single_open
,
615 .proc_read
= seq_read
,
616 .proc_lseek
= seq_lseek
,
617 .proc_release
= single_release
,
620 struct proc_dir_entry
*proc_create_single_data(const char *name
, umode_t mode
,
621 struct proc_dir_entry
*parent
,
622 int (*show
)(struct seq_file
*, void *), void *data
)
624 struct proc_dir_entry
*p
;
626 p
= proc_create_reg(name
, mode
, &parent
, data
);
629 p
->proc_ops
= &proc_single_ops
;
630 p
->single_show
= show
;
631 return proc_register(parent
, p
);
633 EXPORT_SYMBOL(proc_create_single_data
);
635 void proc_set_size(struct proc_dir_entry
*de
, loff_t size
)
639 EXPORT_SYMBOL(proc_set_size
);
641 void proc_set_user(struct proc_dir_entry
*de
, kuid_t uid
, kgid_t gid
)
646 EXPORT_SYMBOL(proc_set_user
);
648 void pde_put(struct proc_dir_entry
*pde
)
650 if (refcount_dec_and_test(&pde
->refcnt
)) {
651 proc_free_inum(pde
->low_ino
);
657 * Remove a /proc entry and free it if it's not currently in use.
659 void remove_proc_entry(const char *name
, struct proc_dir_entry
*parent
)
661 struct proc_dir_entry
*de
= NULL
;
662 const char *fn
= name
;
665 write_lock(&proc_subdir_lock
);
666 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
667 write_unlock(&proc_subdir_lock
);
672 de
= pde_subdir_find(parent
, fn
, len
);
674 if (unlikely(pde_is_permanent(de
))) {
675 WARN(1, "removing permanent /proc entry '%s'", de
->name
);
678 rb_erase(&de
->subdir_node
, &parent
->subdir
);
679 if (S_ISDIR(de
->mode
))
683 write_unlock(&proc_subdir_lock
);
685 WARN(1, "name '%s'\n", name
);
689 proc_entry_rundown(de
);
691 WARN(pde_subdir_first(de
),
692 "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
693 __func__
, de
->parent
->name
, de
->name
, pde_subdir_first(de
)->name
);
696 EXPORT_SYMBOL(remove_proc_entry
);
698 int remove_proc_subtree(const char *name
, struct proc_dir_entry
*parent
)
700 struct proc_dir_entry
*root
= NULL
, *de
, *next
;
701 const char *fn
= name
;
704 write_lock(&proc_subdir_lock
);
705 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
706 write_unlock(&proc_subdir_lock
);
711 root
= pde_subdir_find(parent
, fn
, len
);
713 write_unlock(&proc_subdir_lock
);
716 if (unlikely(pde_is_permanent(root
))) {
717 write_unlock(&proc_subdir_lock
);
718 WARN(1, "removing permanent /proc entry '%s/%s'",
719 root
->parent
->name
, root
->name
);
722 rb_erase(&root
->subdir_node
, &parent
->subdir
);
726 next
= pde_subdir_first(de
);
728 if (unlikely(pde_is_permanent(root
))) {
729 write_unlock(&proc_subdir_lock
);
730 WARN(1, "removing permanent /proc entry '%s/%s'",
731 next
->parent
->name
, next
->name
);
734 rb_erase(&next
->subdir_node
, &de
->subdir
);
739 if (S_ISDIR(de
->mode
))
741 write_unlock(&proc_subdir_lock
);
743 proc_entry_rundown(de
);
748 write_lock(&proc_subdir_lock
);
754 EXPORT_SYMBOL(remove_proc_subtree
);
756 void *proc_get_parent_data(const struct inode
*inode
)
758 struct proc_dir_entry
*de
= PDE(inode
);
759 return de
->parent
->data
;
761 EXPORT_SYMBOL_GPL(proc_get_parent_data
);
763 void proc_remove(struct proc_dir_entry
*de
)
766 remove_proc_subtree(de
->name
, de
->parent
);
768 EXPORT_SYMBOL(proc_remove
);
770 void *PDE_DATA(const struct inode
*inode
)
772 return __PDE_DATA(inode
);
774 EXPORT_SYMBOL(PDE_DATA
);
777 * Pull a user buffer into memory and pass it to the file's write handler if
778 * one is supplied. The ->write() method is permitted to modify the
779 * kernel-side buffer.
781 ssize_t
proc_simple_write(struct file
*f
, const char __user
*ubuf
, size_t size
,
784 struct proc_dir_entry
*pde
= PDE(file_inode(f
));
790 if (size
== 0 || size
> PAGE_SIZE
- 1)
792 buf
= memdup_user_nul(ubuf
, size
);
795 ret
= pde
->write(f
, buf
, size
);
797 return ret
== 0 ? size
: ret
;