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/namei.h>
23 #include <linux/bitops.h>
24 #include <linux/spinlock.h>
25 #include <linux/completion.h>
26 #include <asm/uaccess.h>
30 DEFINE_SPINLOCK(proc_subdir_lock
);
32 static int proc_match(unsigned int len
, const char *name
, struct proc_dir_entry
*de
)
34 if (de
->namelen
!= len
)
36 return !memcmp(name
, de
->name
, len
);
39 /* buffer size is one page but our output routines use some slack for overruns */
40 #define PROC_BLOCK_SIZE (PAGE_SIZE - 1024)
43 __proc_file_read(struct file
*file
, char __user
*buf
, size_t nbytes
,
46 struct inode
* inode
= file_inode(file
);
52 struct proc_dir_entry
* dp
;
53 unsigned long long pos
;
56 * Gaah, please just use "seq_file" instead. The legacy /proc
57 * interfaces cut loff_t down to off_t for reads, and ignore
58 * the offset entirely for writes..
61 if (pos
> MAX_NON_LFS
)
63 if (nbytes
> MAX_NON_LFS
- pos
)
64 nbytes
= MAX_NON_LFS
- pos
;
67 if (!(page
= (char*) __get_free_page(GFP_TEMPORARY
)))
70 while ((nbytes
> 0) && !eof
) {
71 count
= min_t(size_t, PROC_BLOCK_SIZE
, nbytes
);
76 * How to be a proc read function
77 * ------------------------------
79 * int f(char *buffer, char **start, off_t offset,
80 * int count, int *peof, void *dat)
82 * Assume that the buffer is "count" bytes in size.
84 * If you know you have supplied all the data you
87 * You have three ways to return data:
88 * 0) Leave *start = NULL. (This is the default.)
89 * Put the data of the requested offset at that
90 * offset within the buffer. Return the number (n)
91 * of bytes there are from the beginning of the
92 * buffer up to the last byte of data. If the
93 * number of supplied bytes (= n - offset) is
94 * greater than zero and you didn't signal eof
95 * and the reader is prepared to take more data
96 * you will be called again with the requested
97 * offset advanced by the number of bytes
98 * absorbed. This interface is useful for files
99 * no larger than the buffer.
100 * 1) Set *start = an unsigned long value less than
101 * the buffer address but greater than zero.
102 * Put the data of the requested offset at the
103 * beginning of the buffer. Return the number of
104 * bytes of data placed there. If this number is
105 * greater than zero and you didn't signal eof
106 * and the reader is prepared to take more data
107 * you will be called again with the requested
108 * offset advanced by *start. This interface is
109 * useful when you have a large file consisting
110 * of a series of blocks which you want to count
111 * and return as wholes.
112 * (Hack by Paul.Russell@rustcorp.com.au)
113 * 2) Set *start = an address within the buffer.
114 * Put the data of the requested offset at *start.
115 * Return the number of bytes of data placed there.
116 * If this number is greater than zero and you
117 * didn't signal eof and the reader is prepared to
118 * take more data you will be called again with the
119 * requested offset advanced by the number of bytes
122 n
= dp
->read_proc(page
, &start
, *ppos
,
123 count
, &eof
, dp
->data
);
127 if (n
== 0) /* end of file */
129 if (n
< 0) { /* error */
136 if (n
> PAGE_SIZE
) /* Apparent buffer overflow */
143 start
= page
+ *ppos
;
144 } else if (start
< page
) {
145 if (n
> PAGE_SIZE
) /* Apparent buffer overflow */
149 * Don't reduce n because doing so might
150 * cut off part of a data block.
152 pr_warn("proc_file_read: count exceeded\n");
154 } else /* start >= page */ {
155 unsigned long startoff
= (unsigned long)(start
- page
);
156 if (n
> (PAGE_SIZE
- startoff
)) /* buffer overflow? */
157 n
= PAGE_SIZE
- startoff
;
162 n
-= copy_to_user(buf
, start
< page
? page
: start
, n
);
169 *ppos
+= start
< page
? (unsigned long)start
: n
;
174 free_page((unsigned long) page
);
179 proc_file_read(struct file
*file
, char __user
*buf
, size_t nbytes
,
182 struct proc_dir_entry
*pde
= PDE(file_inode(file
));
185 spin_lock(&pde
->pde_unload_lock
);
186 if (!pde
->proc_fops
) {
187 spin_unlock(&pde
->pde_unload_lock
);
191 spin_unlock(&pde
->pde_unload_lock
);
193 rv
= __proc_file_read(file
, buf
, nbytes
, ppos
);
200 proc_file_lseek(struct file
*file
, loff_t offset
, int orig
)
202 loff_t retval
= -EINVAL
;
205 offset
+= file
->f_pos
;
208 if (offset
< 0 || offset
> MAX_NON_LFS
)
210 file
->f_pos
= retval
= offset
;
215 static const struct file_operations proc_file_operations
= {
216 .llseek
= proc_file_lseek
,
217 .read
= proc_file_read
,
220 static int proc_notify_change(struct dentry
*dentry
, struct iattr
*iattr
)
222 struct inode
*inode
= dentry
->d_inode
;
223 struct proc_dir_entry
*de
= PDE(inode
);
226 error
= inode_change_ok(inode
, iattr
);
230 setattr_copy(inode
, iattr
);
231 mark_inode_dirty(inode
);
233 de
->uid
= inode
->i_uid
;
234 de
->gid
= inode
->i_gid
;
235 de
->mode
= inode
->i_mode
;
239 static int proc_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
,
242 struct inode
*inode
= dentry
->d_inode
;
243 struct proc_dir_entry
*de
= PROC_I(inode
)->pde
;
245 set_nlink(inode
, de
->nlink
);
247 generic_fillattr(inode
, stat
);
251 static const struct inode_operations proc_file_inode_operations
= {
252 .setattr
= proc_notify_change
,
256 * This function parses a name such as "tty/driver/serial", and
257 * returns the struct proc_dir_entry for "/proc/tty/driver", and
258 * returns "serial" in residual.
260 static int __xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
261 const char **residual
)
263 const char *cp
= name
, *next
;
264 struct proc_dir_entry
*de
;
272 next
= strchr(cp
, '/');
277 for (de
= de
->subdir
; de
; de
= de
->next
) {
278 if (proc_match(len
, cp
, de
))
282 WARN(1, "name '%s'\n", name
);
292 static int xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
293 const char **residual
)
297 spin_lock(&proc_subdir_lock
);
298 rv
= __xlate_proc_name(name
, ret
, residual
);
299 spin_unlock(&proc_subdir_lock
);
303 static DEFINE_IDA(proc_inum_ida
);
304 static DEFINE_SPINLOCK(proc_inum_lock
); /* protects the above */
306 #define PROC_DYNAMIC_FIRST 0xF0000000U
309 * Return an inode number between PROC_DYNAMIC_FIRST and
310 * 0xffffffff, or zero on failure.
312 int proc_alloc_inum(unsigned int *inum
)
318 if (!ida_pre_get(&proc_inum_ida
, GFP_KERNEL
))
321 spin_lock_irq(&proc_inum_lock
);
322 error
= ida_get_new(&proc_inum_ida
, &i
);
323 spin_unlock_irq(&proc_inum_lock
);
324 if (error
== -EAGAIN
)
329 if (i
> UINT_MAX
- PROC_DYNAMIC_FIRST
) {
330 spin_lock_irq(&proc_inum_lock
);
331 ida_remove(&proc_inum_ida
, i
);
332 spin_unlock_irq(&proc_inum_lock
);
335 *inum
= PROC_DYNAMIC_FIRST
+ i
;
339 void proc_free_inum(unsigned int inum
)
342 spin_lock_irqsave(&proc_inum_lock
, flags
);
343 ida_remove(&proc_inum_ida
, inum
- PROC_DYNAMIC_FIRST
);
344 spin_unlock_irqrestore(&proc_inum_lock
, flags
);
347 static void *proc_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
349 nd_set_link(nd
, PDE_DATA(dentry
->d_inode
));
353 static const struct inode_operations proc_link_inode_operations
= {
354 .readlink
= generic_readlink
,
355 .follow_link
= proc_follow_link
,
359 * As some entries in /proc are volatile, we want to
360 * get rid of unused dentries. This could be made
361 * smarter: we could keep a "volatile" flag in the
362 * inode to indicate which ones to keep.
364 static int proc_delete_dentry(const struct dentry
* dentry
)
369 static const struct dentry_operations proc_dentry_operations
=
371 .d_delete
= proc_delete_dentry
,
375 * Don't create negative dentries here, return -ENOENT by hand
378 struct dentry
*proc_lookup_de(struct proc_dir_entry
*de
, struct inode
*dir
,
379 struct dentry
*dentry
)
383 spin_lock(&proc_subdir_lock
);
384 for (de
= de
->subdir
; de
; de
= de
->next
) {
385 if (de
->namelen
!= dentry
->d_name
.len
)
387 if (!memcmp(dentry
->d_name
.name
, de
->name
, de
->namelen
)) {
389 spin_unlock(&proc_subdir_lock
);
390 inode
= proc_get_inode(dir
->i_sb
, de
);
392 return ERR_PTR(-ENOMEM
);
393 d_set_d_op(dentry
, &proc_dentry_operations
);
394 d_add(dentry
, inode
);
398 spin_unlock(&proc_subdir_lock
);
399 return ERR_PTR(-ENOENT
);
402 struct dentry
*proc_lookup(struct inode
*dir
, struct dentry
*dentry
,
405 return proc_lookup_de(PDE(dir
), dir
, dentry
);
409 * This returns non-zero if at EOF, so that the /proc
410 * root directory can use this and check if it should
411 * continue with the <pid> entries..
413 * Note that the VFS-layer doesn't care about the return
414 * value of the readdir() call, as long as it's non-negative
417 int proc_readdir_de(struct proc_dir_entry
*de
, struct file
*filp
, void *dirent
,
422 struct inode
*inode
= file_inode(filp
);
429 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
435 if (filldir(dirent
, "..", 2, i
,
436 parent_ino(filp
->f_path
.dentry
),
443 spin_lock(&proc_subdir_lock
);
449 spin_unlock(&proc_subdir_lock
);
459 struct proc_dir_entry
*next
;
461 /* filldir passes info to user space */
463 spin_unlock(&proc_subdir_lock
);
464 if (filldir(dirent
, de
->name
, de
->namelen
, filp
->f_pos
,
465 de
->low_ino
, de
->mode
>> 12) < 0) {
469 spin_lock(&proc_subdir_lock
);
475 spin_unlock(&proc_subdir_lock
);
482 int proc_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
484 struct inode
*inode
= file_inode(filp
);
486 return proc_readdir_de(PDE(inode
), filp
, dirent
, filldir
);
490 * These are the generic /proc directory operations. They
491 * use the in-memory "struct proc_dir_entry" tree to parse
492 * the /proc directory.
494 static const struct file_operations proc_dir_operations
= {
495 .llseek
= generic_file_llseek
,
496 .read
= generic_read_dir
,
497 .readdir
= proc_readdir
,
501 * proc directories can do almost nothing..
503 static const struct inode_operations proc_dir_inode_operations
= {
504 .lookup
= proc_lookup
,
505 .getattr
= proc_getattr
,
506 .setattr
= proc_notify_change
,
509 static int proc_register(struct proc_dir_entry
* dir
, struct proc_dir_entry
* dp
)
511 struct proc_dir_entry
*tmp
;
514 ret
= proc_alloc_inum(&dp
->low_ino
);
518 if (S_ISDIR(dp
->mode
)) {
519 dp
->proc_fops
= &proc_dir_operations
;
520 dp
->proc_iops
= &proc_dir_inode_operations
;
522 } else if (S_ISLNK(dp
->mode
)) {
523 dp
->proc_iops
= &proc_link_inode_operations
;
524 } else if (S_ISREG(dp
->mode
)) {
525 if (dp
->proc_fops
== NULL
)
526 dp
->proc_fops
= &proc_file_operations
;
527 dp
->proc_iops
= &proc_file_inode_operations
;
533 spin_lock(&proc_subdir_lock
);
535 for (tmp
= dir
->subdir
; tmp
; tmp
= tmp
->next
)
536 if (strcmp(tmp
->name
, dp
->name
) == 0) {
537 WARN(1, "proc_dir_entry '%s/%s' already registered\n",
538 dir
->name
, dp
->name
);
542 dp
->next
= dir
->subdir
;
545 spin_unlock(&proc_subdir_lock
);
550 static struct proc_dir_entry
*__proc_create(struct proc_dir_entry
**parent
,
555 struct proc_dir_entry
*ent
= NULL
;
556 const char *fn
= name
;
559 /* make sure name is valid */
560 if (!name
|| !strlen(name
))
563 if (xlate_proc_name(name
, parent
, &fn
) != 0)
566 /* At this point there must not be any '/' characters beyond *fn */
572 ent
= kzalloc(sizeof(struct proc_dir_entry
) + len
+ 1, GFP_KERNEL
);
576 memcpy(ent
->name
, fn
, len
+ 1);
580 atomic_set(&ent
->count
, 1);
581 spin_lock_init(&ent
->pde_unload_lock
);
582 INIT_LIST_HEAD(&ent
->pde_openers
);
587 struct proc_dir_entry
*proc_symlink(const char *name
,
588 struct proc_dir_entry
*parent
, const char *dest
)
590 struct proc_dir_entry
*ent
;
592 ent
= __proc_create(&parent
, name
,
593 (S_IFLNK
| S_IRUGO
| S_IWUGO
| S_IXUGO
),1);
596 ent
->data
= kmalloc((ent
->size
=strlen(dest
))+1, GFP_KERNEL
);
598 strcpy((char*)ent
->data
,dest
);
599 if (proc_register(parent
, ent
) < 0) {
611 EXPORT_SYMBOL(proc_symlink
);
613 struct proc_dir_entry
*proc_mkdir_mode(const char *name
, umode_t mode
,
614 struct proc_dir_entry
*parent
)
616 struct proc_dir_entry
*ent
;
618 ent
= __proc_create(&parent
, name
, S_IFDIR
| mode
, 2);
620 if (proc_register(parent
, ent
) < 0) {
627 EXPORT_SYMBOL(proc_mkdir_mode
);
629 struct proc_dir_entry
*proc_net_mkdir(struct net
*net
, const char *name
,
630 struct proc_dir_entry
*parent
)
632 struct proc_dir_entry
*ent
;
634 ent
= __proc_create(&parent
, name
, S_IFDIR
| S_IRUGO
| S_IXUGO
, 2);
637 if (proc_register(parent
, ent
) < 0) {
644 EXPORT_SYMBOL_GPL(proc_net_mkdir
);
646 struct proc_dir_entry
*proc_mkdir(const char *name
,
647 struct proc_dir_entry
*parent
)
649 return proc_mkdir_mode(name
, S_IRUGO
| S_IXUGO
, parent
);
651 EXPORT_SYMBOL(proc_mkdir
);
653 struct proc_dir_entry
*create_proc_entry(const char *name
, umode_t mode
,
654 struct proc_dir_entry
*parent
)
656 struct proc_dir_entry
*ent
;
658 if ((mode
& S_IFMT
) == 0)
661 if (!S_ISREG(mode
)) {
662 WARN_ON(1); /* use proc_mkdir(), damnit */
666 if ((mode
& S_IALLUGO
) == 0)
669 ent
= __proc_create(&parent
, name
, mode
, 1);
671 if (proc_register(parent
, ent
) < 0) {
678 EXPORT_SYMBOL(create_proc_entry
);
680 struct proc_dir_entry
*proc_create_data(const char *name
, umode_t mode
,
681 struct proc_dir_entry
*parent
,
682 const struct file_operations
*proc_fops
,
685 struct proc_dir_entry
*pde
;
686 if ((mode
& S_IFMT
) == 0)
689 if (!S_ISREG(mode
)) {
690 WARN_ON(1); /* use proc_mkdir() */
694 if ((mode
& S_IALLUGO
) == 0)
696 pde
= __proc_create(&parent
, name
, mode
, 1);
699 pde
->proc_fops
= proc_fops
;
701 if (proc_register(parent
, pde
) < 0)
709 EXPORT_SYMBOL(proc_create_data
);
711 static void free_proc_entry(struct proc_dir_entry
*de
)
713 proc_free_inum(de
->low_ino
);
715 if (S_ISLNK(de
->mode
))
720 void pde_put(struct proc_dir_entry
*pde
)
722 if (atomic_dec_and_test(&pde
->count
))
723 free_proc_entry(pde
);
726 static void entry_rundown(struct proc_dir_entry
*de
)
728 spin_lock(&de
->pde_unload_lock
);
730 * Stop accepting new callers into module. If you're
731 * dynamically allocating ->proc_fops, save a pointer somewhere.
733 de
->proc_fops
= NULL
;
734 /* Wait until all existing callers into module are done. */
735 if (de
->pde_users
> 0) {
736 DECLARE_COMPLETION_ONSTACK(c
);
738 if (!de
->pde_unload_completion
)
739 de
->pde_unload_completion
= &c
;
741 spin_unlock(&de
->pde_unload_lock
);
743 wait_for_completion(de
->pde_unload_completion
);
745 spin_lock(&de
->pde_unload_lock
);
748 while (!list_empty(&de
->pde_openers
)) {
749 struct pde_opener
*pdeo
;
751 pdeo
= list_first_entry(&de
->pde_openers
, struct pde_opener
, lh
);
753 spin_unlock(&de
->pde_unload_lock
);
754 pdeo
->release(pdeo
->inode
, pdeo
->file
);
756 spin_lock(&de
->pde_unload_lock
);
758 spin_unlock(&de
->pde_unload_lock
);
762 * Remove a /proc entry and free it if it's not currently in use.
764 void remove_proc_entry(const char *name
, struct proc_dir_entry
*parent
)
766 struct proc_dir_entry
**p
;
767 struct proc_dir_entry
*de
= NULL
;
768 const char *fn
= name
;
771 spin_lock(&proc_subdir_lock
);
772 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
773 spin_unlock(&proc_subdir_lock
);
778 for (p
= &parent
->subdir
; *p
; p
=&(*p
)->next
) {
779 if (proc_match(len
, fn
, *p
)) {
786 spin_unlock(&proc_subdir_lock
);
788 WARN(1, "name '%s'\n", name
);
794 if (S_ISDIR(de
->mode
))
797 WARN(de
->subdir
, "%s: removing non-empty directory "
798 "'%s/%s', leaking at least '%s'\n", __func__
,
799 de
->parent
->name
, de
->name
, de
->subdir
->name
);
802 EXPORT_SYMBOL(remove_proc_entry
);
804 int remove_proc_subtree(const char *name
, struct proc_dir_entry
*parent
)
806 struct proc_dir_entry
**p
;
807 struct proc_dir_entry
*root
= NULL
, *de
, *next
;
808 const char *fn
= name
;
811 spin_lock(&proc_subdir_lock
);
812 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
813 spin_unlock(&proc_subdir_lock
);
818 for (p
= &parent
->subdir
; *p
; p
=&(*p
)->next
) {
819 if (proc_match(len
, fn
, *p
)) {
827 spin_unlock(&proc_subdir_lock
);
834 de
->subdir
= next
->next
;
839 spin_unlock(&proc_subdir_lock
);
843 if (S_ISDIR(de
->mode
))
850 spin_lock(&proc_subdir_lock
);
856 EXPORT_SYMBOL(remove_proc_subtree
);