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
);
111 error
= setattr_prepare(dentry
, iattr
);
115 setattr_copy(inode
, iattr
);
116 mark_inode_dirty(inode
);
118 proc_set_user(de
, inode
->i_uid
, inode
->i_gid
);
119 de
->mode
= inode
->i_mode
;
123 static int proc_getattr(const struct path
*path
, struct kstat
*stat
,
124 u32 request_mask
, unsigned int query_flags
)
126 struct inode
*inode
= d_inode(path
->dentry
);
127 struct proc_dir_entry
*de
= PDE(inode
);
129 set_nlink(inode
, de
->nlink
);
131 generic_fillattr(inode
, stat
);
135 static const struct inode_operations proc_file_inode_operations
= {
136 .setattr
= proc_notify_change
,
140 * This function parses a name such as "tty/driver/serial", and
141 * returns the struct proc_dir_entry for "/proc/tty/driver", and
142 * returns "serial" in residual.
144 static int __xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
145 const char **residual
)
147 const char *cp
= name
, *next
;
148 struct proc_dir_entry
*de
;
156 next
= strchr(cp
, '/');
161 de
= pde_subdir_find(de
, cp
, len
);
163 WARN(1, "name '%s'\n", name
);
173 static int xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
174 const char **residual
)
178 read_lock(&proc_subdir_lock
);
179 rv
= __xlate_proc_name(name
, ret
, residual
);
180 read_unlock(&proc_subdir_lock
);
184 static DEFINE_IDA(proc_inum_ida
);
186 #define PROC_DYNAMIC_FIRST 0xF0000000U
189 * Return an inode number between PROC_DYNAMIC_FIRST and
190 * 0xffffffff, or zero on failure.
192 int proc_alloc_inum(unsigned int *inum
)
196 i
= ida_simple_get(&proc_inum_ida
, 0, UINT_MAX
- PROC_DYNAMIC_FIRST
+ 1,
201 *inum
= PROC_DYNAMIC_FIRST
+ (unsigned int)i
;
205 void proc_free_inum(unsigned int inum
)
207 ida_simple_remove(&proc_inum_ida
, inum
- PROC_DYNAMIC_FIRST
);
211 * Don't create negative dentries here, return -ENOENT by hand
214 struct dentry
*proc_lookup_de(struct proc_dir_entry
*de
, struct inode
*dir
,
215 struct dentry
*dentry
)
219 read_lock(&proc_subdir_lock
);
220 de
= pde_subdir_find(de
, dentry
->d_name
.name
, dentry
->d_name
.len
);
223 read_unlock(&proc_subdir_lock
);
224 inode
= proc_get_inode(dir
->i_sb
, de
);
226 return ERR_PTR(-ENOMEM
);
227 d_set_d_op(dentry
, &simple_dentry_operations
);
228 d_add(dentry
, inode
);
231 read_unlock(&proc_subdir_lock
);
232 return ERR_PTR(-ENOENT
);
235 struct dentry
*proc_lookup(struct inode
*dir
, struct dentry
*dentry
,
238 return proc_lookup_de(PDE(dir
), dir
, dentry
);
242 * This returns non-zero if at EOF, so that the /proc
243 * root directory can use this and check if it should
244 * continue with the <pid> entries..
246 * Note that the VFS-layer doesn't care about the return
247 * value of the readdir() call, as long as it's non-negative
250 int proc_readdir_de(struct proc_dir_entry
*de
, struct file
*file
,
251 struct dir_context
*ctx
)
255 if (!dir_emit_dots(file
, ctx
))
258 read_lock(&proc_subdir_lock
);
259 de
= pde_subdir_first(de
);
263 read_unlock(&proc_subdir_lock
);
268 de
= pde_subdir_next(de
);
273 struct proc_dir_entry
*next
;
275 read_unlock(&proc_subdir_lock
);
276 if (!dir_emit(ctx
, de
->name
, de
->namelen
,
277 de
->low_ino
, de
->mode
>> 12)) {
281 read_lock(&proc_subdir_lock
);
283 next
= pde_subdir_next(de
);
287 read_unlock(&proc_subdir_lock
);
291 int proc_readdir(struct file
*file
, struct dir_context
*ctx
)
293 struct inode
*inode
= file_inode(file
);
295 return proc_readdir_de(PDE(inode
), file
, ctx
);
299 * These are the generic /proc directory operations. They
300 * use the in-memory "struct proc_dir_entry" tree to parse
301 * the /proc directory.
303 static const struct file_operations proc_dir_operations
= {
304 .llseek
= generic_file_llseek
,
305 .read
= generic_read_dir
,
306 .iterate_shared
= proc_readdir
,
310 * proc directories can do almost nothing..
312 static const struct inode_operations proc_dir_inode_operations
= {
313 .lookup
= proc_lookup
,
314 .getattr
= proc_getattr
,
315 .setattr
= proc_notify_change
,
318 static int proc_register(struct proc_dir_entry
* dir
, struct proc_dir_entry
* dp
)
322 ret
= proc_alloc_inum(&dp
->low_ino
);
326 write_lock(&proc_subdir_lock
);
328 if (pde_subdir_insert(dir
, dp
) == false) {
329 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
330 dir
->name
, dp
->name
);
331 write_unlock(&proc_subdir_lock
);
332 proc_free_inum(dp
->low_ino
);
335 write_unlock(&proc_subdir_lock
);
340 static struct proc_dir_entry
*__proc_create(struct proc_dir_entry
**parent
,
345 struct proc_dir_entry
*ent
= NULL
;
349 if (xlate_proc_name(name
, parent
, &fn
) != 0)
352 qstr
.len
= strlen(fn
);
353 if (qstr
.len
== 0 || qstr
.len
>= 256) {
354 WARN(1, "name len %u\n", qstr
.len
);
357 if (*parent
== &proc_root
&& name_to_int(&qstr
) != ~0U) {
358 WARN(1, "create '/proc/%s' by hand\n", qstr
.name
);
361 if (is_empty_pde(*parent
)) {
362 WARN(1, "attempt to add to permanently empty directory");
366 ent
= kzalloc(sizeof(struct proc_dir_entry
) + qstr
.len
+ 1, GFP_KERNEL
);
370 memcpy(ent
->name
, fn
, qstr
.len
+ 1);
371 ent
->namelen
= qstr
.len
;
374 ent
->subdir
= RB_ROOT_CACHED
;
375 atomic_set(&ent
->count
, 1);
376 spin_lock_init(&ent
->pde_unload_lock
);
377 INIT_LIST_HEAD(&ent
->pde_openers
);
378 proc_set_user(ent
, (*parent
)->uid
, (*parent
)->gid
);
384 struct proc_dir_entry
*proc_symlink(const char *name
,
385 struct proc_dir_entry
*parent
, const char *dest
)
387 struct proc_dir_entry
*ent
;
389 ent
= __proc_create(&parent
, name
,
390 (S_IFLNK
| S_IRUGO
| S_IWUGO
| S_IXUGO
),1);
393 ent
->data
= kmalloc((ent
->size
=strlen(dest
))+1, GFP_KERNEL
);
395 strcpy((char*)ent
->data
,dest
);
396 ent
->proc_iops
= &proc_link_inode_operations
;
397 if (proc_register(parent
, ent
) < 0) {
409 EXPORT_SYMBOL(proc_symlink
);
411 struct proc_dir_entry
*proc_mkdir_data(const char *name
, umode_t mode
,
412 struct proc_dir_entry
*parent
, void *data
)
414 struct proc_dir_entry
*ent
;
417 mode
= S_IRUGO
| S_IXUGO
;
419 ent
= __proc_create(&parent
, name
, S_IFDIR
| mode
, 2);
422 ent
->proc_fops
= &proc_dir_operations
;
423 ent
->proc_iops
= &proc_dir_inode_operations
;
425 if (proc_register(parent
, ent
) < 0) {
433 EXPORT_SYMBOL_GPL(proc_mkdir_data
);
435 struct proc_dir_entry
*proc_mkdir_mode(const char *name
, umode_t mode
,
436 struct proc_dir_entry
*parent
)
438 return proc_mkdir_data(name
, mode
, parent
, NULL
);
440 EXPORT_SYMBOL(proc_mkdir_mode
);
442 struct proc_dir_entry
*proc_mkdir(const char *name
,
443 struct proc_dir_entry
*parent
)
445 return proc_mkdir_data(name
, 0, parent
, NULL
);
447 EXPORT_SYMBOL(proc_mkdir
);
449 struct proc_dir_entry
*proc_create_mount_point(const char *name
)
451 umode_t mode
= S_IFDIR
| S_IRUGO
| S_IXUGO
;
452 struct proc_dir_entry
*ent
, *parent
= NULL
;
454 ent
= __proc_create(&parent
, name
, mode
, 2);
457 ent
->proc_fops
= NULL
;
458 ent
->proc_iops
= NULL
;
460 if (proc_register(parent
, ent
) < 0) {
468 EXPORT_SYMBOL(proc_create_mount_point
);
470 struct proc_dir_entry
*proc_create_data(const char *name
, umode_t mode
,
471 struct proc_dir_entry
*parent
,
472 const struct file_operations
*proc_fops
,
475 struct proc_dir_entry
*pde
;
476 if ((mode
& S_IFMT
) == 0)
479 if (!S_ISREG(mode
)) {
480 WARN_ON(1); /* use proc_mkdir() */
484 BUG_ON(proc_fops
== NULL
);
486 if ((mode
& S_IALLUGO
) == 0)
488 pde
= __proc_create(&parent
, name
, mode
, 1);
491 pde
->proc_fops
= proc_fops
;
493 pde
->proc_iops
= &proc_file_inode_operations
;
494 if (proc_register(parent
, pde
) < 0)
502 EXPORT_SYMBOL(proc_create_data
);
504 struct proc_dir_entry
*proc_create(const char *name
, umode_t mode
,
505 struct proc_dir_entry
*parent
,
506 const struct file_operations
*proc_fops
)
508 return proc_create_data(name
, mode
, parent
, proc_fops
, NULL
);
510 EXPORT_SYMBOL(proc_create
);
512 void proc_set_size(struct proc_dir_entry
*de
, loff_t size
)
516 EXPORT_SYMBOL(proc_set_size
);
518 void proc_set_user(struct proc_dir_entry
*de
, kuid_t uid
, kgid_t gid
)
523 EXPORT_SYMBOL(proc_set_user
);
525 static void free_proc_entry(struct proc_dir_entry
*de
)
527 proc_free_inum(de
->low_ino
);
529 if (S_ISLNK(de
->mode
))
534 void pde_put(struct proc_dir_entry
*pde
)
536 if (atomic_dec_and_test(&pde
->count
))
537 free_proc_entry(pde
);
541 * Remove a /proc entry and free it if it's not currently in use.
543 void remove_proc_entry(const char *name
, struct proc_dir_entry
*parent
)
545 struct proc_dir_entry
*de
= NULL
;
546 const char *fn
= name
;
549 write_lock(&proc_subdir_lock
);
550 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
551 write_unlock(&proc_subdir_lock
);
556 de
= pde_subdir_find(parent
, fn
, len
);
558 rb_erase_cached(&de
->subdir_node
, &parent
->subdir
);
559 write_unlock(&proc_subdir_lock
);
561 WARN(1, "name '%s'\n", name
);
565 proc_entry_rundown(de
);
567 if (S_ISDIR(de
->mode
))
570 WARN(pde_subdir_first(de
),
571 "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
572 __func__
, de
->parent
->name
, de
->name
, pde_subdir_first(de
)->name
);
575 EXPORT_SYMBOL(remove_proc_entry
);
577 int remove_proc_subtree(const char *name
, struct proc_dir_entry
*parent
)
579 struct proc_dir_entry
*root
= NULL
, *de
, *next
;
580 const char *fn
= name
;
583 write_lock(&proc_subdir_lock
);
584 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
585 write_unlock(&proc_subdir_lock
);
590 root
= pde_subdir_find(parent
, fn
, len
);
592 write_unlock(&proc_subdir_lock
);
595 rb_erase_cached(&root
->subdir_node
, &parent
->subdir
);
599 next
= pde_subdir_first(de
);
601 rb_erase_cached(&next
->subdir_node
, &de
->subdir
);
605 write_unlock(&proc_subdir_lock
);
607 proc_entry_rundown(de
);
609 if (S_ISDIR(de
->mode
))
616 write_lock(&proc_subdir_lock
);
622 EXPORT_SYMBOL(remove_proc_subtree
);
624 void *proc_get_parent_data(const struct inode
*inode
)
626 struct proc_dir_entry
*de
= PDE(inode
);
627 return de
->parent
->data
;
629 EXPORT_SYMBOL_GPL(proc_get_parent_data
);
631 void proc_remove(struct proc_dir_entry
*de
)
634 remove_proc_subtree(de
->name
, de
->parent
);
636 EXPORT_SYMBOL(proc_remove
);
638 void *PDE_DATA(const struct inode
*inode
)
640 return __PDE_DATA(inode
);
642 EXPORT_SYMBOL(PDE_DATA
);