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(&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
= 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
*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 = rb_entry(*new,
84 struct proc_dir_entry
,
86 int result
= proc_match(de
->namelen
, de
->name
, this);
90 new = &(*new)->rb_left
;
92 new = &(*new)->rb_right
;
97 /* Add new node and rebalance tree. */
98 rb_link_node(&de
->subdir_node
, parent
, new);
99 rb_insert_color(&de
->subdir_node
, root
);
103 static int proc_notify_change(struct dentry
*dentry
, struct iattr
*iattr
)
105 struct inode
*inode
= d_inode(dentry
);
106 struct proc_dir_entry
*de
= PDE(inode
);
109 error
= setattr_prepare(dentry
, iattr
);
113 setattr_copy(inode
, iattr
);
114 mark_inode_dirty(inode
);
116 proc_set_user(de
, inode
->i_uid
, inode
->i_gid
);
117 de
->mode
= inode
->i_mode
;
121 static int proc_getattr(const struct path
*path
, struct kstat
*stat
,
122 u32 request_mask
, unsigned int query_flags
)
124 struct inode
*inode
= d_inode(path
->dentry
);
125 struct proc_dir_entry
*de
= PDE(inode
);
127 set_nlink(inode
, de
->nlink
);
129 generic_fillattr(inode
, stat
);
133 static const struct inode_operations proc_file_inode_operations
= {
134 .setattr
= proc_notify_change
,
138 * This function parses a name such as "tty/driver/serial", and
139 * returns the struct proc_dir_entry for "/proc/tty/driver", and
140 * returns "serial" in residual.
142 static int __xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
143 const char **residual
)
145 const char *cp
= name
, *next
;
146 struct proc_dir_entry
*de
;
154 next
= strchr(cp
, '/');
159 de
= pde_subdir_find(de
, cp
, len
);
161 WARN(1, "name '%s'\n", name
);
171 static int xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
172 const char **residual
)
176 read_lock(&proc_subdir_lock
);
177 rv
= __xlate_proc_name(name
, ret
, residual
);
178 read_unlock(&proc_subdir_lock
);
182 static DEFINE_IDA(proc_inum_ida
);
183 static DEFINE_SPINLOCK(proc_inum_lock
); /* protects the above */
185 #define PROC_DYNAMIC_FIRST 0xF0000000U
188 * Return an inode number between PROC_DYNAMIC_FIRST and
189 * 0xffffffff, or zero on failure.
191 int proc_alloc_inum(unsigned int *inum
)
197 if (!ida_pre_get(&proc_inum_ida
, GFP_KERNEL
))
200 spin_lock_irq(&proc_inum_lock
);
201 error
= ida_get_new(&proc_inum_ida
, &i
);
202 spin_unlock_irq(&proc_inum_lock
);
203 if (error
== -EAGAIN
)
208 if (i
> UINT_MAX
- PROC_DYNAMIC_FIRST
) {
209 spin_lock_irq(&proc_inum_lock
);
210 ida_remove(&proc_inum_ida
, i
);
211 spin_unlock_irq(&proc_inum_lock
);
214 *inum
= PROC_DYNAMIC_FIRST
+ i
;
218 void proc_free_inum(unsigned int inum
)
221 spin_lock_irqsave(&proc_inum_lock
, flags
);
222 ida_remove(&proc_inum_ida
, inum
- PROC_DYNAMIC_FIRST
);
223 spin_unlock_irqrestore(&proc_inum_lock
, flags
);
227 * Don't create negative dentries here, return -ENOENT by hand
230 struct dentry
*proc_lookup_de(struct proc_dir_entry
*de
, struct inode
*dir
,
231 struct dentry
*dentry
)
235 read_lock(&proc_subdir_lock
);
236 de
= pde_subdir_find(de
, dentry
->d_name
.name
, dentry
->d_name
.len
);
239 read_unlock(&proc_subdir_lock
);
240 inode
= proc_get_inode(dir
->i_sb
, de
);
242 return ERR_PTR(-ENOMEM
);
243 d_set_d_op(dentry
, &simple_dentry_operations
);
244 d_add(dentry
, inode
);
247 read_unlock(&proc_subdir_lock
);
248 return ERR_PTR(-ENOENT
);
251 struct dentry
*proc_lookup(struct inode
*dir
, struct dentry
*dentry
,
254 return proc_lookup_de(PDE(dir
), dir
, dentry
);
258 * This returns non-zero if at EOF, so that the /proc
259 * root directory can use this and check if it should
260 * continue with the <pid> entries..
262 * Note that the VFS-layer doesn't care about the return
263 * value of the readdir() call, as long as it's non-negative
266 int proc_readdir_de(struct proc_dir_entry
*de
, struct file
*file
,
267 struct dir_context
*ctx
)
271 if (!dir_emit_dots(file
, ctx
))
274 read_lock(&proc_subdir_lock
);
275 de
= pde_subdir_first(de
);
279 read_unlock(&proc_subdir_lock
);
284 de
= pde_subdir_next(de
);
289 struct proc_dir_entry
*next
;
291 read_unlock(&proc_subdir_lock
);
292 if (!dir_emit(ctx
, de
->name
, de
->namelen
,
293 de
->low_ino
, de
->mode
>> 12)) {
297 read_lock(&proc_subdir_lock
);
299 next
= pde_subdir_next(de
);
303 read_unlock(&proc_subdir_lock
);
307 int proc_readdir(struct file
*file
, struct dir_context
*ctx
)
309 struct inode
*inode
= file_inode(file
);
311 return proc_readdir_de(PDE(inode
), file
, ctx
);
315 * These are the generic /proc directory operations. They
316 * use the in-memory "struct proc_dir_entry" tree to parse
317 * the /proc directory.
319 static const struct file_operations proc_dir_operations
= {
320 .llseek
= generic_file_llseek
,
321 .read
= generic_read_dir
,
322 .iterate_shared
= proc_readdir
,
326 * proc directories can do almost nothing..
328 static const struct inode_operations proc_dir_inode_operations
= {
329 .lookup
= proc_lookup
,
330 .getattr
= proc_getattr
,
331 .setattr
= proc_notify_change
,
334 static int proc_register(struct proc_dir_entry
* dir
, struct proc_dir_entry
* dp
)
338 ret
= proc_alloc_inum(&dp
->low_ino
);
342 write_lock(&proc_subdir_lock
);
344 if (pde_subdir_insert(dir
, dp
) == false) {
345 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
346 dir
->name
, dp
->name
);
347 write_unlock(&proc_subdir_lock
);
348 proc_free_inum(dp
->low_ino
);
351 write_unlock(&proc_subdir_lock
);
356 static struct proc_dir_entry
*__proc_create(struct proc_dir_entry
**parent
,
361 struct proc_dir_entry
*ent
= NULL
;
365 if (xlate_proc_name(name
, parent
, &fn
) != 0)
368 qstr
.len
= strlen(fn
);
369 if (qstr
.len
== 0 || qstr
.len
>= 256) {
370 WARN(1, "name len %u\n", qstr
.len
);
373 if (*parent
== &proc_root
&& name_to_int(&qstr
) != ~0U) {
374 WARN(1, "create '/proc/%s' by hand\n", qstr
.name
);
377 if (is_empty_pde(*parent
)) {
378 WARN(1, "attempt to add to permanently empty directory");
382 ent
= kzalloc(sizeof(struct proc_dir_entry
) + qstr
.len
+ 1, GFP_KERNEL
);
386 memcpy(ent
->name
, fn
, qstr
.len
+ 1);
387 ent
->namelen
= qstr
.len
;
390 ent
->subdir
= RB_ROOT
;
391 atomic_set(&ent
->count
, 1);
392 spin_lock_init(&ent
->pde_unload_lock
);
393 INIT_LIST_HEAD(&ent
->pde_openers
);
394 proc_set_user(ent
, (*parent
)->uid
, (*parent
)->gid
);
400 struct proc_dir_entry
*proc_symlink(const char *name
,
401 struct proc_dir_entry
*parent
, const char *dest
)
403 struct proc_dir_entry
*ent
;
405 ent
= __proc_create(&parent
, name
,
406 (S_IFLNK
| S_IRUGO
| S_IWUGO
| S_IXUGO
),1);
409 ent
->data
= kmalloc((ent
->size
=strlen(dest
))+1, GFP_KERNEL
);
411 strcpy((char*)ent
->data
,dest
);
412 ent
->proc_iops
= &proc_link_inode_operations
;
413 if (proc_register(parent
, ent
) < 0) {
425 EXPORT_SYMBOL(proc_symlink
);
427 struct proc_dir_entry
*proc_mkdir_data(const char *name
, umode_t mode
,
428 struct proc_dir_entry
*parent
, void *data
)
430 struct proc_dir_entry
*ent
;
433 mode
= S_IRUGO
| S_IXUGO
;
435 ent
= __proc_create(&parent
, name
, S_IFDIR
| mode
, 2);
438 ent
->proc_fops
= &proc_dir_operations
;
439 ent
->proc_iops
= &proc_dir_inode_operations
;
441 if (proc_register(parent
, ent
) < 0) {
449 EXPORT_SYMBOL_GPL(proc_mkdir_data
);
451 struct proc_dir_entry
*proc_mkdir_mode(const char *name
, umode_t mode
,
452 struct proc_dir_entry
*parent
)
454 return proc_mkdir_data(name
, mode
, parent
, NULL
);
456 EXPORT_SYMBOL(proc_mkdir_mode
);
458 struct proc_dir_entry
*proc_mkdir(const char *name
,
459 struct proc_dir_entry
*parent
)
461 return proc_mkdir_data(name
, 0, parent
, NULL
);
463 EXPORT_SYMBOL(proc_mkdir
);
465 struct proc_dir_entry
*proc_create_mount_point(const char *name
)
467 umode_t mode
= S_IFDIR
| S_IRUGO
| S_IXUGO
;
468 struct proc_dir_entry
*ent
, *parent
= NULL
;
470 ent
= __proc_create(&parent
, name
, mode
, 2);
473 ent
->proc_fops
= NULL
;
474 ent
->proc_iops
= NULL
;
476 if (proc_register(parent
, ent
) < 0) {
484 EXPORT_SYMBOL(proc_create_mount_point
);
486 struct proc_dir_entry
*proc_create_data(const char *name
, umode_t mode
,
487 struct proc_dir_entry
*parent
,
488 const struct file_operations
*proc_fops
,
491 struct proc_dir_entry
*pde
;
492 if ((mode
& S_IFMT
) == 0)
495 if (!S_ISREG(mode
)) {
496 WARN_ON(1); /* use proc_mkdir() */
500 BUG_ON(proc_fops
== NULL
);
502 if ((mode
& S_IALLUGO
) == 0)
504 pde
= __proc_create(&parent
, name
, mode
, 1);
507 pde
->proc_fops
= proc_fops
;
509 pde
->proc_iops
= &proc_file_inode_operations
;
510 if (proc_register(parent
, pde
) < 0)
518 EXPORT_SYMBOL(proc_create_data
);
520 void proc_set_size(struct proc_dir_entry
*de
, loff_t size
)
524 EXPORT_SYMBOL(proc_set_size
);
526 void proc_set_user(struct proc_dir_entry
*de
, kuid_t uid
, kgid_t gid
)
531 EXPORT_SYMBOL(proc_set_user
);
533 static void free_proc_entry(struct proc_dir_entry
*de
)
535 proc_free_inum(de
->low_ino
);
537 if (S_ISLNK(de
->mode
))
542 void pde_put(struct proc_dir_entry
*pde
)
544 if (atomic_dec_and_test(&pde
->count
))
545 free_proc_entry(pde
);
549 * Remove a /proc entry and free it if it's not currently in use.
551 void remove_proc_entry(const char *name
, struct proc_dir_entry
*parent
)
553 struct proc_dir_entry
*de
= NULL
;
554 const char *fn
= name
;
557 write_lock(&proc_subdir_lock
);
558 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
559 write_unlock(&proc_subdir_lock
);
564 de
= pde_subdir_find(parent
, fn
, len
);
566 rb_erase(&de
->subdir_node
, &parent
->subdir
);
567 write_unlock(&proc_subdir_lock
);
569 WARN(1, "name '%s'\n", name
);
573 proc_entry_rundown(de
);
575 if (S_ISDIR(de
->mode
))
578 WARN(pde_subdir_first(de
),
579 "%s: removing non-empty directory '%s/%s', leaking at least '%s'\n",
580 __func__
, de
->parent
->name
, de
->name
, pde_subdir_first(de
)->name
);
583 EXPORT_SYMBOL(remove_proc_entry
);
585 int remove_proc_subtree(const char *name
, struct proc_dir_entry
*parent
)
587 struct proc_dir_entry
*root
= NULL
, *de
, *next
;
588 const char *fn
= name
;
591 write_lock(&proc_subdir_lock
);
592 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
593 write_unlock(&proc_subdir_lock
);
598 root
= pde_subdir_find(parent
, fn
, len
);
600 write_unlock(&proc_subdir_lock
);
603 rb_erase(&root
->subdir_node
, &parent
->subdir
);
607 next
= pde_subdir_first(de
);
609 rb_erase(&next
->subdir_node
, &de
->subdir
);
613 write_unlock(&proc_subdir_lock
);
615 proc_entry_rundown(de
);
617 if (S_ISDIR(de
->mode
))
624 write_lock(&proc_subdir_lock
);
630 EXPORT_SYMBOL(remove_proc_subtree
);
632 void *proc_get_parent_data(const struct inode
*inode
)
634 struct proc_dir_entry
*de
= PDE(inode
);
635 return de
->parent
->data
;
637 EXPORT_SYMBOL_GPL(proc_get_parent_data
);
639 void proc_remove(struct proc_dir_entry
*de
)
642 remove_proc_subtree(de
->name
, de
->parent
);
644 EXPORT_SYMBOL(proc_remove
);
646 void *PDE_DATA(const struct inode
*inode
)
648 return __PDE_DATA(inode
);
650 EXPORT_SYMBOL(PDE_DATA
);