2 * proc/fs/generic.c --- generic routines for the proc-fs
4 * This file contains generic proc-fs routines for handling
5 * directories and files.
7 * Copyright (C) 1991, 1992 Linus Torvalds.
8 * Copyright (C) 1997 Theodore Ts'o
11 #include <linux/errno.h>
12 #include <linux/time.h>
13 #include <linux/proc_fs.h>
14 #include <linux/stat.h>
16 #include <linux/module.h>
17 #include <linux/slab.h>
18 #include <linux/printk.h>
19 #include <linux/mount.h>
20 #include <linux/init.h>
21 #include <linux/idr.h>
22 #include <linux/bitops.h>
23 #include <linux/spinlock.h>
24 #include <linux/completion.h>
25 #include <linux/uaccess.h>
29 static DEFINE_RWLOCK(proc_subdir_lock
);
31 static int proc_match(unsigned int len
, const char *name
, struct proc_dir_entry
*de
)
33 if (len
< de
->namelen
)
35 if (len
> de
->namelen
)
38 return memcmp(name
, de
->name
, len
);
41 static struct proc_dir_entry
*pde_subdir_first(struct proc_dir_entry
*dir
)
43 return rb_entry_safe(rb_first_cached(&dir
->subdir
),
44 struct proc_dir_entry
, subdir_node
);
47 static struct proc_dir_entry
*pde_subdir_next(struct proc_dir_entry
*dir
)
49 return rb_entry_safe(rb_next(&dir
->subdir_node
), struct proc_dir_entry
,
53 static struct proc_dir_entry
*pde_subdir_find(struct proc_dir_entry
*dir
,
57 struct rb_node
*node
= dir
->subdir
.rb_root
.rb_node
;
60 struct proc_dir_entry
*de
= rb_entry(node
,
61 struct proc_dir_entry
,
63 int result
= proc_match(len
, name
, de
);
68 node
= node
->rb_right
;
75 static bool pde_subdir_insert(struct proc_dir_entry
*dir
,
76 struct proc_dir_entry
*de
)
78 struct rb_root_cached
*root
= &dir
->subdir
;
79 struct rb_node
**new = &root
->rb_root
.rb_node
, *parent
= NULL
;
82 /* Figure out where to put new node */
84 struct proc_dir_entry
*this = rb_entry(*new,
85 struct proc_dir_entry
,
87 int result
= proc_match(de
->namelen
, de
->name
, this);
91 new = &(*new)->rb_left
;
92 else if (result
> 0) {
93 new = &(*new)->rb_right
;
99 /* Add new node and rebalance tree. */
100 rb_link_node(&de
->subdir_node
, parent
, new);
101 rb_insert_color_cached(&de
->subdir_node
, root
, leftmost
);
105 static int proc_notify_change(struct dentry
*dentry
, struct iattr
*iattr
)
107 struct inode
*inode
= d_inode(dentry
);
108 struct proc_dir_entry
*de
= PDE(inode
);
109 struct user_namespace
*s_user_ns
;
112 /* Don't let anyone mess with weird proc files */
113 s_user_ns
= inode
->i_sb
->s_user_ns
;
114 if (!kuid_has_mapping(s_user_ns
, inode
->i_uid
) ||
115 !kgid_has_mapping(s_user_ns
, inode
->i_gid
))
118 error
= setattr_prepare(dentry
, iattr
);
122 setattr_copy(inode
, iattr
);
123 mark_inode_dirty(inode
);
125 proc_set_user(de
, inode
->i_uid
, inode
->i_gid
);
126 de
->mode
= inode
->i_mode
;
130 static int proc_getattr(const struct path
*path
, struct kstat
*stat
,
131 u32 request_mask
, unsigned int query_flags
)
133 struct inode
*inode
= d_inode(path
->dentry
);
134 struct proc_dir_entry
*de
= PDE(inode
);
136 set_nlink(inode
, de
->nlink
);
138 generic_fillattr(inode
, stat
);
142 static const struct inode_operations proc_file_inode_operations
= {
143 .setattr
= proc_notify_change
,
147 * This function parses a name such as "tty/driver/serial", and
148 * returns the struct proc_dir_entry for "/proc/tty/driver", and
149 * returns "serial" in residual.
151 static int __xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
152 const char **residual
)
154 const char *cp
= name
, *next
;
155 struct proc_dir_entry
*de
;
163 next
= strchr(cp
, '/');
168 de
= pde_subdir_find(de
, cp
, len
);
170 WARN(1, "name '%s'\n", name
);
180 static int xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
181 const char **residual
)
185 read_lock(&proc_subdir_lock
);
186 rv
= __xlate_proc_name(name
, ret
, residual
);
187 read_unlock(&proc_subdir_lock
);
191 static DEFINE_IDA(proc_inum_ida
);
193 #define PROC_DYNAMIC_FIRST 0xF0000000U
196 * Return an inode number between PROC_DYNAMIC_FIRST and
197 * 0xffffffff, or zero on failure.
199 int proc_alloc_inum(unsigned int *inum
)
203 i
= ida_simple_get(&proc_inum_ida
, 0, UINT_MAX
- PROC_DYNAMIC_FIRST
+ 1,
208 *inum
= PROC_DYNAMIC_FIRST
+ (unsigned int)i
;
212 void proc_free_inum(unsigned int inum
)
214 ida_simple_remove(&proc_inum_ida
, inum
- PROC_DYNAMIC_FIRST
);
218 * Don't create negative dentries here, return -ENOENT by hand
221 struct dentry
*proc_lookup_de(struct proc_dir_entry
*de
, struct inode
*dir
,
222 struct dentry
*dentry
)
226 read_lock(&proc_subdir_lock
);
227 de
= pde_subdir_find(de
, dentry
->d_name
.name
, dentry
->d_name
.len
);
230 read_unlock(&proc_subdir_lock
);
231 inode
= proc_get_inode(dir
->i_sb
, de
);
233 return ERR_PTR(-ENOMEM
);
234 d_set_d_op(dentry
, &simple_dentry_operations
);
235 d_add(dentry
, inode
);
238 read_unlock(&proc_subdir_lock
);
239 return ERR_PTR(-ENOENT
);
242 struct dentry
*proc_lookup(struct inode
*dir
, struct dentry
*dentry
,
245 return proc_lookup_de(PDE(dir
), dir
, dentry
);
249 * This returns non-zero if at EOF, so that the /proc
250 * root directory can use this and check if it should
251 * continue with the <pid> entries..
253 * Note that the VFS-layer doesn't care about the return
254 * value of the readdir() call, as long as it's non-negative
257 int proc_readdir_de(struct proc_dir_entry
*de
, struct file
*file
,
258 struct dir_context
*ctx
)
262 if (!dir_emit_dots(file
, ctx
))
265 read_lock(&proc_subdir_lock
);
266 de
= pde_subdir_first(de
);
270 read_unlock(&proc_subdir_lock
);
275 de
= pde_subdir_next(de
);
280 struct proc_dir_entry
*next
;
282 read_unlock(&proc_subdir_lock
);
283 if (!dir_emit(ctx
, de
->name
, de
->namelen
,
284 de
->low_ino
, de
->mode
>> 12)) {
288 read_lock(&proc_subdir_lock
);
290 next
= pde_subdir_next(de
);
294 read_unlock(&proc_subdir_lock
);
298 int proc_readdir(struct file
*file
, struct dir_context
*ctx
)
300 struct inode
*inode
= file_inode(file
);
302 return proc_readdir_de(PDE(inode
), file
, ctx
);
306 * These are the generic /proc directory operations. They
307 * use the in-memory "struct proc_dir_entry" tree to parse
308 * the /proc directory.
310 static const struct file_operations proc_dir_operations
= {
311 .llseek
= generic_file_llseek
,
312 .read
= generic_read_dir
,
313 .iterate_shared
= proc_readdir
,
317 * proc directories can do almost nothing..
319 static const struct inode_operations proc_dir_inode_operations
= {
320 .lookup
= proc_lookup
,
321 .getattr
= proc_getattr
,
322 .setattr
= proc_notify_change
,
325 static int proc_register(struct proc_dir_entry
* dir
, struct proc_dir_entry
* dp
)
329 ret
= proc_alloc_inum(&dp
->low_ino
);
333 write_lock(&proc_subdir_lock
);
335 if (pde_subdir_insert(dir
, dp
) == false) {
336 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
337 dir
->name
, dp
->name
);
338 write_unlock(&proc_subdir_lock
);
339 proc_free_inum(dp
->low_ino
);
342 write_unlock(&proc_subdir_lock
);
347 static struct proc_dir_entry
*__proc_create(struct proc_dir_entry
**parent
,
352 struct proc_dir_entry
*ent
= NULL
;
356 if (xlate_proc_name(name
, parent
, &fn
) != 0)
359 qstr
.len
= strlen(fn
);
360 if (qstr
.len
== 0 || qstr
.len
>= 256) {
361 WARN(1, "name len %u\n", qstr
.len
);
364 if (*parent
== &proc_root
&& name_to_int(&qstr
) != ~0U) {
365 WARN(1, "create '/proc/%s' by hand\n", qstr
.name
);
368 if (is_empty_pde(*parent
)) {
369 WARN(1, "attempt to add to permanently empty directory");
373 ent
= kzalloc(sizeof(struct proc_dir_entry
) + qstr
.len
+ 1, GFP_KERNEL
);
377 memcpy(ent
->name
, fn
, qstr
.len
+ 1);
378 ent
->namelen
= qstr
.len
;
381 ent
->subdir
= RB_ROOT_CACHED
;
382 atomic_set(&ent
->count
, 1);
383 spin_lock_init(&ent
->pde_unload_lock
);
384 INIT_LIST_HEAD(&ent
->pde_openers
);
385 proc_set_user(ent
, (*parent
)->uid
, (*parent
)->gid
);
391 struct proc_dir_entry
*proc_symlink(const char *name
,
392 struct proc_dir_entry
*parent
, const char *dest
)
394 struct proc_dir_entry
*ent
;
396 ent
= __proc_create(&parent
, name
,
397 (S_IFLNK
| S_IRUGO
| S_IWUGO
| S_IXUGO
),1);
400 ent
->data
= kmalloc((ent
->size
=strlen(dest
))+1, GFP_KERNEL
);
402 strcpy((char*)ent
->data
,dest
);
403 ent
->proc_iops
= &proc_link_inode_operations
;
404 if (proc_register(parent
, ent
) < 0) {
416 EXPORT_SYMBOL(proc_symlink
);
418 struct proc_dir_entry
*proc_mkdir_data(const char *name
, umode_t mode
,
419 struct proc_dir_entry
*parent
, void *data
)
421 struct proc_dir_entry
*ent
;
424 mode
= S_IRUGO
| S_IXUGO
;
426 ent
= __proc_create(&parent
, name
, S_IFDIR
| mode
, 2);
429 ent
->proc_fops
= &proc_dir_operations
;
430 ent
->proc_iops
= &proc_dir_inode_operations
;
432 if (proc_register(parent
, ent
) < 0) {
440 EXPORT_SYMBOL_GPL(proc_mkdir_data
);
442 struct proc_dir_entry
*proc_mkdir_mode(const char *name
, umode_t mode
,
443 struct proc_dir_entry
*parent
)
445 return proc_mkdir_data(name
, mode
, parent
, NULL
);
447 EXPORT_SYMBOL(proc_mkdir_mode
);
449 struct proc_dir_entry
*proc_mkdir(const char *name
,
450 struct proc_dir_entry
*parent
)
452 return proc_mkdir_data(name
, 0, parent
, NULL
);
454 EXPORT_SYMBOL(proc_mkdir
);
456 struct proc_dir_entry
*proc_create_mount_point(const char *name
)
458 umode_t mode
= S_IFDIR
| S_IRUGO
| S_IXUGO
;
459 struct proc_dir_entry
*ent
, *parent
= NULL
;
461 ent
= __proc_create(&parent
, name
, mode
, 2);
464 ent
->proc_fops
= NULL
;
465 ent
->proc_iops
= NULL
;
467 if (proc_register(parent
, ent
) < 0) {
475 EXPORT_SYMBOL(proc_create_mount_point
);
477 struct proc_dir_entry
*proc_create_data(const char *name
, umode_t mode
,
478 struct proc_dir_entry
*parent
,
479 const struct file_operations
*proc_fops
,
482 struct proc_dir_entry
*pde
;
483 if ((mode
& S_IFMT
) == 0)
486 if (!S_ISREG(mode
)) {
487 WARN_ON(1); /* use proc_mkdir() */
491 BUG_ON(proc_fops
== NULL
);
493 if ((mode
& S_IALLUGO
) == 0)
495 pde
= __proc_create(&parent
, name
, mode
, 1);
498 pde
->proc_fops
= proc_fops
;
500 pde
->proc_iops
= &proc_file_inode_operations
;
501 if (proc_register(parent
, pde
) < 0)
509 EXPORT_SYMBOL(proc_create_data
);
511 struct proc_dir_entry
*proc_create(const char *name
, umode_t mode
,
512 struct proc_dir_entry
*parent
,
513 const struct file_operations
*proc_fops
)
515 return proc_create_data(name
, mode
, parent
, proc_fops
, NULL
);
517 EXPORT_SYMBOL(proc_create
);
519 void proc_set_size(struct proc_dir_entry
*de
, loff_t size
)
523 EXPORT_SYMBOL(proc_set_size
);
525 void proc_set_user(struct proc_dir_entry
*de
, kuid_t uid
, kgid_t gid
)
530 EXPORT_SYMBOL(proc_set_user
);
532 static void free_proc_entry(struct proc_dir_entry
*de
)
534 proc_free_inum(de
->low_ino
);
536 if (S_ISLNK(de
->mode
))
541 void pde_put(struct proc_dir_entry
*pde
)
543 if (atomic_dec_and_test(&pde
->count
))
544 free_proc_entry(pde
);
548 * Remove a /proc entry and free it if it's not currently in use.
550 void remove_proc_entry(const char *name
, struct proc_dir_entry
*parent
)
552 struct proc_dir_entry
*de
= NULL
;
553 const char *fn
= name
;
556 write_lock(&proc_subdir_lock
);
557 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
558 write_unlock(&proc_subdir_lock
);
563 de
= pde_subdir_find(parent
, fn
, len
);
565 rb_erase_cached(&de
->subdir_node
, &parent
->subdir
);
566 write_unlock(&proc_subdir_lock
);
568 WARN(1, "name '%s'\n", name
);
572 proc_entry_rundown(de
);
574 if (S_ISDIR(de
->mode
))
577 WARN(pde_subdir_first(de
),
578 "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
579 __func__
, de
->parent
->name
, de
->name
, pde_subdir_first(de
)->name
);
582 EXPORT_SYMBOL(remove_proc_entry
);
584 int remove_proc_subtree(const char *name
, struct proc_dir_entry
*parent
)
586 struct proc_dir_entry
*root
= NULL
, *de
, *next
;
587 const char *fn
= name
;
590 write_lock(&proc_subdir_lock
);
591 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
592 write_unlock(&proc_subdir_lock
);
597 root
= pde_subdir_find(parent
, fn
, len
);
599 write_unlock(&proc_subdir_lock
);
602 rb_erase_cached(&root
->subdir_node
, &parent
->subdir
);
606 next
= pde_subdir_first(de
);
608 rb_erase_cached(&next
->subdir_node
, &de
->subdir
);
612 write_unlock(&proc_subdir_lock
);
614 proc_entry_rundown(de
);
616 if (S_ISDIR(de
->mode
))
623 write_lock(&proc_subdir_lock
);
629 EXPORT_SYMBOL(remove_proc_subtree
);
631 void *proc_get_parent_data(const struct inode
*inode
)
633 struct proc_dir_entry
*de
= PDE(inode
);
634 return de
->parent
->data
;
636 EXPORT_SYMBOL_GPL(proc_get_parent_data
);
638 void proc_remove(struct proc_dir_entry
*de
)
641 remove_proc_subtree(de
->name
, de
->parent
);
643 EXPORT_SYMBOL(proc_remove
);
645 void *PDE_DATA(const struct inode
*inode
)
647 return __PDE_DATA(inode
);
649 EXPORT_SYMBOL(PDE_DATA
);