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 <asm/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(&dir
->subdir
), struct proc_dir_entry
,
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_node
;
60 struct proc_dir_entry
*de
= container_of(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
*root
= &dir
->subdir
;
79 struct rb_node
**new = &root
->rb_node
, *parent
= NULL
;
81 /* Figure out where to put new node */
83 struct proc_dir_entry
*this =
84 container_of(*new, struct proc_dir_entry
, subdir_node
);
85 int result
= proc_match(de
->namelen
, de
->name
, this);
89 new = &(*new)->rb_left
;
91 new = &(*new)->rb_right
;
96 /* Add new node and rebalance tree. */
97 rb_link_node(&de
->subdir_node
, parent
, new);
98 rb_insert_color(&de
->subdir_node
, root
);
102 static int proc_notify_change(struct dentry
*dentry
, struct iattr
*iattr
)
104 struct inode
*inode
= d_inode(dentry
);
105 struct proc_dir_entry
*de
= PDE(inode
);
108 error
= inode_change_ok(inode
, iattr
);
112 setattr_copy(inode
, iattr
);
113 mark_inode_dirty(inode
);
115 proc_set_user(de
, inode
->i_uid
, inode
->i_gid
);
116 de
->mode
= inode
->i_mode
;
120 static int proc_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
,
123 struct inode
*inode
= d_inode(dentry
);
124 struct proc_dir_entry
*de
= PDE(inode
);
126 set_nlink(inode
, de
->nlink
);
128 generic_fillattr(inode
, stat
);
132 static const struct inode_operations proc_file_inode_operations
= {
133 .setattr
= proc_notify_change
,
137 * This function parses a name such as "tty/driver/serial", and
138 * returns the struct proc_dir_entry for "/proc/tty/driver", and
139 * returns "serial" in residual.
141 static int __xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
142 const char **residual
)
144 const char *cp
= name
, *next
;
145 struct proc_dir_entry
*de
;
153 next
= strchr(cp
, '/');
158 de
= pde_subdir_find(de
, cp
, len
);
160 WARN(1, "name '%s'\n", name
);
170 static int xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
171 const char **residual
)
175 read_lock(&proc_subdir_lock
);
176 rv
= __xlate_proc_name(name
, ret
, residual
);
177 read_unlock(&proc_subdir_lock
);
181 static DEFINE_IDA(proc_inum_ida
);
182 static DEFINE_SPINLOCK(proc_inum_lock
); /* protects the above */
184 #define PROC_DYNAMIC_FIRST 0xF0000000U
187 * Return an inode number between PROC_DYNAMIC_FIRST and
188 * 0xffffffff, or zero on failure.
190 int proc_alloc_inum(unsigned int *inum
)
196 if (!ida_pre_get(&proc_inum_ida
, GFP_KERNEL
))
199 spin_lock_irq(&proc_inum_lock
);
200 error
= ida_get_new(&proc_inum_ida
, &i
);
201 spin_unlock_irq(&proc_inum_lock
);
202 if (error
== -EAGAIN
)
207 if (i
> UINT_MAX
- PROC_DYNAMIC_FIRST
) {
208 spin_lock_irq(&proc_inum_lock
);
209 ida_remove(&proc_inum_ida
, i
);
210 spin_unlock_irq(&proc_inum_lock
);
213 *inum
= PROC_DYNAMIC_FIRST
+ i
;
217 void proc_free_inum(unsigned int inum
)
220 spin_lock_irqsave(&proc_inum_lock
, flags
);
221 ida_remove(&proc_inum_ida
, inum
- PROC_DYNAMIC_FIRST
);
222 spin_unlock_irqrestore(&proc_inum_lock
, flags
);
226 * Don't create negative dentries here, return -ENOENT by hand
229 struct dentry
*proc_lookup_de(struct proc_dir_entry
*de
, struct inode
*dir
,
230 struct dentry
*dentry
)
234 read_lock(&proc_subdir_lock
);
235 de
= pde_subdir_find(de
, dentry
->d_name
.name
, dentry
->d_name
.len
);
238 read_unlock(&proc_subdir_lock
);
239 inode
= proc_get_inode(dir
->i_sb
, de
);
241 return ERR_PTR(-ENOMEM
);
242 d_set_d_op(dentry
, &simple_dentry_operations
);
243 d_add(dentry
, inode
);
246 read_unlock(&proc_subdir_lock
);
247 return ERR_PTR(-ENOENT
);
250 struct dentry
*proc_lookup(struct inode
*dir
, struct dentry
*dentry
,
253 return proc_lookup_de(PDE(dir
), dir
, dentry
);
257 * This returns non-zero if at EOF, so that the /proc
258 * root directory can use this and check if it should
259 * continue with the <pid> entries..
261 * Note that the VFS-layer doesn't care about the return
262 * value of the readdir() call, as long as it's non-negative
265 int proc_readdir_de(struct proc_dir_entry
*de
, struct file
*file
,
266 struct dir_context
*ctx
)
270 if (!dir_emit_dots(file
, ctx
))
273 read_lock(&proc_subdir_lock
);
274 de
= pde_subdir_first(de
);
278 read_unlock(&proc_subdir_lock
);
283 de
= pde_subdir_next(de
);
288 struct proc_dir_entry
*next
;
290 read_unlock(&proc_subdir_lock
);
291 if (!dir_emit(ctx
, de
->name
, de
->namelen
,
292 de
->low_ino
, de
->mode
>> 12)) {
296 read_lock(&proc_subdir_lock
);
298 next
= pde_subdir_next(de
);
302 read_unlock(&proc_subdir_lock
);
306 int proc_readdir(struct file
*file
, struct dir_context
*ctx
)
308 struct inode
*inode
= file_inode(file
);
310 return proc_readdir_de(PDE(inode
), file
, ctx
);
314 * These are the generic /proc directory operations. They
315 * use the in-memory "struct proc_dir_entry" tree to parse
316 * the /proc directory.
318 static const struct file_operations proc_dir_operations
= {
319 .llseek
= generic_file_llseek
,
320 .read
= generic_read_dir
,
321 .iterate
= proc_readdir
,
325 * proc directories can do almost nothing..
327 static const struct inode_operations proc_dir_inode_operations
= {
328 .lookup
= proc_lookup
,
329 .getattr
= proc_getattr
,
330 .setattr
= proc_notify_change
,
333 static int proc_register(struct proc_dir_entry
* dir
, struct proc_dir_entry
* dp
)
337 ret
= proc_alloc_inum(&dp
->low_ino
);
341 write_lock(&proc_subdir_lock
);
343 if (pde_subdir_insert(dir
, dp
) == false) {
344 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
345 dir
->name
, dp
->name
);
346 write_unlock(&proc_subdir_lock
);
347 proc_free_inum(dp
->low_ino
);
350 write_unlock(&proc_subdir_lock
);
355 static struct proc_dir_entry
*__proc_create(struct proc_dir_entry
**parent
,
360 struct proc_dir_entry
*ent
= NULL
;
364 if (xlate_proc_name(name
, parent
, &fn
) != 0)
367 qstr
.len
= strlen(fn
);
368 if (qstr
.len
== 0 || qstr
.len
>= 256) {
369 WARN(1, "name len %u\n", qstr
.len
);
372 if (*parent
== &proc_root
&& name_to_int(&qstr
) != ~0U) {
373 WARN(1, "create '/proc/%s' by hand\n", qstr
.name
);
376 if (is_empty_pde(*parent
)) {
377 WARN(1, "attempt to add to permanently empty directory");
381 ent
= kzalloc(sizeof(struct proc_dir_entry
) + qstr
.len
+ 1, GFP_KERNEL
);
385 memcpy(ent
->name
, fn
, qstr
.len
+ 1);
386 ent
->namelen
= qstr
.len
;
389 ent
->subdir
= RB_ROOT
;
390 atomic_set(&ent
->count
, 1);
391 spin_lock_init(&ent
->pde_unload_lock
);
392 INIT_LIST_HEAD(&ent
->pde_openers
);
397 struct proc_dir_entry
*proc_symlink(const char *name
,
398 struct proc_dir_entry
*parent
, const char *dest
)
400 struct proc_dir_entry
*ent
;
402 ent
= __proc_create(&parent
, name
,
403 (S_IFLNK
| S_IRUGO
| S_IWUGO
| S_IXUGO
),1);
406 ent
->data
= kmalloc((ent
->size
=strlen(dest
))+1, GFP_KERNEL
);
408 strcpy((char*)ent
->data
,dest
);
409 ent
->proc_iops
= &proc_link_inode_operations
;
410 if (proc_register(parent
, ent
) < 0) {
422 EXPORT_SYMBOL(proc_symlink
);
424 struct proc_dir_entry
*proc_mkdir_data(const char *name
, umode_t mode
,
425 struct proc_dir_entry
*parent
, void *data
)
427 struct proc_dir_entry
*ent
;
430 mode
= S_IRUGO
| S_IXUGO
;
432 ent
= __proc_create(&parent
, name
, S_IFDIR
| mode
, 2);
435 ent
->proc_fops
= &proc_dir_operations
;
436 ent
->proc_iops
= &proc_dir_inode_operations
;
438 if (proc_register(parent
, ent
) < 0) {
446 EXPORT_SYMBOL_GPL(proc_mkdir_data
);
448 struct proc_dir_entry
*proc_mkdir_mode(const char *name
, umode_t mode
,
449 struct proc_dir_entry
*parent
)
451 return proc_mkdir_data(name
, mode
, parent
, NULL
);
453 EXPORT_SYMBOL(proc_mkdir_mode
);
455 struct proc_dir_entry
*proc_mkdir(const char *name
,
456 struct proc_dir_entry
*parent
)
458 return proc_mkdir_data(name
, 0, parent
, NULL
);
460 EXPORT_SYMBOL(proc_mkdir
);
462 struct proc_dir_entry
*proc_create_mount_point(const char *name
)
464 umode_t mode
= S_IFDIR
| S_IRUGO
| S_IXUGO
;
465 struct proc_dir_entry
*ent
, *parent
= NULL
;
467 ent
= __proc_create(&parent
, name
, mode
, 2);
470 ent
->proc_fops
= NULL
;
471 ent
->proc_iops
= NULL
;
472 if (proc_register(parent
, ent
) < 0) {
481 struct proc_dir_entry
*proc_create_data(const char *name
, umode_t mode
,
482 struct proc_dir_entry
*parent
,
483 const struct file_operations
*proc_fops
,
486 struct proc_dir_entry
*pde
;
487 if ((mode
& S_IFMT
) == 0)
490 if (!S_ISREG(mode
)) {
491 WARN_ON(1); /* use proc_mkdir() */
495 BUG_ON(proc_fops
== NULL
);
497 if ((mode
& S_IALLUGO
) == 0)
499 pde
= __proc_create(&parent
, name
, mode
, 1);
502 pde
->proc_fops
= proc_fops
;
504 pde
->proc_iops
= &proc_file_inode_operations
;
505 if (proc_register(parent
, pde
) < 0)
513 EXPORT_SYMBOL(proc_create_data
);
515 void proc_set_size(struct proc_dir_entry
*de
, loff_t size
)
519 EXPORT_SYMBOL(proc_set_size
);
521 void proc_set_user(struct proc_dir_entry
*de
, kuid_t uid
, kgid_t gid
)
526 EXPORT_SYMBOL(proc_set_user
);
528 static void free_proc_entry(struct proc_dir_entry
*de
)
530 proc_free_inum(de
->low_ino
);
532 if (S_ISLNK(de
->mode
))
537 void pde_put(struct proc_dir_entry
*pde
)
539 if (atomic_dec_and_test(&pde
->count
))
540 free_proc_entry(pde
);
544 * Remove a /proc entry and free it if it's not currently in use.
546 void remove_proc_entry(const char *name
, struct proc_dir_entry
*parent
)
548 struct proc_dir_entry
*de
= NULL
;
549 const char *fn
= name
;
552 write_lock(&proc_subdir_lock
);
553 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
554 write_unlock(&proc_subdir_lock
);
559 de
= pde_subdir_find(parent
, fn
, len
);
561 rb_erase(&de
->subdir_node
, &parent
->subdir
);
562 write_unlock(&proc_subdir_lock
);
564 WARN(1, "name '%s'\n", name
);
568 proc_entry_rundown(de
);
570 if (S_ISDIR(de
->mode
))
573 WARN(pde_subdir_first(de
),
574 "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
575 __func__
, de
->parent
->name
, de
->name
, pde_subdir_first(de
)->name
);
578 EXPORT_SYMBOL(remove_proc_entry
);
580 int remove_proc_subtree(const char *name
, struct proc_dir_entry
*parent
)
582 struct proc_dir_entry
*root
= NULL
, *de
, *next
;
583 const char *fn
= name
;
586 write_lock(&proc_subdir_lock
);
587 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
588 write_unlock(&proc_subdir_lock
);
593 root
= pde_subdir_find(parent
, fn
, len
);
595 write_unlock(&proc_subdir_lock
);
598 rb_erase(&root
->subdir_node
, &parent
->subdir
);
602 next
= pde_subdir_first(de
);
604 rb_erase(&next
->subdir_node
, &de
->subdir
);
608 write_unlock(&proc_subdir_lock
);
610 proc_entry_rundown(de
);
612 if (S_ISDIR(de
->mode
))
619 write_lock(&proc_subdir_lock
);
625 EXPORT_SYMBOL(remove_proc_subtree
);
627 void *proc_get_parent_data(const struct inode
*inode
)
629 struct proc_dir_entry
*de
= PDE(inode
);
630 return de
->parent
->data
;
632 EXPORT_SYMBOL_GPL(proc_get_parent_data
);
634 void proc_remove(struct proc_dir_entry
*de
)
637 remove_proc_subtree(de
->name
, de
->parent
);
639 EXPORT_SYMBOL(proc_remove
);
641 void *PDE_DATA(const struct inode
*inode
)
643 return __PDE_DATA(inode
);
645 EXPORT_SYMBOL(PDE_DATA
);