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/mount.h>
19 #include <linux/init.h>
20 #include <linux/idr.h>
21 #include <linux/namei.h>
22 #include <linux/bitops.h>
23 #include <linux/spinlock.h>
24 #include <linux/completion.h>
25 #include <asm/uaccess.h>
29 DEFINE_SPINLOCK(proc_subdir_lock
);
31 static int proc_match(unsigned int len
, const char *name
, struct proc_dir_entry
*de
)
33 if (de
->namelen
!= len
)
35 return !memcmp(name
, de
->name
, len
);
38 /* buffer size is one page but our output routines use some slack for overruns */
39 #define PROC_BLOCK_SIZE (PAGE_SIZE - 1024)
42 __proc_file_read(struct file
*file
, char __user
*buf
, size_t nbytes
,
45 struct inode
* inode
= file_inode(file
);
51 struct proc_dir_entry
* dp
;
52 unsigned long long pos
;
55 * Gaah, please just use "seq_file" instead. The legacy /proc
56 * interfaces cut loff_t down to off_t for reads, and ignore
57 * the offset entirely for writes..
60 if (pos
> MAX_NON_LFS
)
62 if (nbytes
> MAX_NON_LFS
- pos
)
63 nbytes
= MAX_NON_LFS
- pos
;
66 if (!(page
= (char*) __get_free_page(GFP_TEMPORARY
)))
69 while ((nbytes
> 0) && !eof
) {
70 count
= min_t(size_t, PROC_BLOCK_SIZE
, nbytes
);
75 * How to be a proc read function
76 * ------------------------------
78 * int f(char *buffer, char **start, off_t offset,
79 * int count, int *peof, void *dat)
81 * Assume that the buffer is "count" bytes in size.
83 * If you know you have supplied all the data you
86 * You have three ways to return data:
87 * 0) Leave *start = NULL. (This is the default.)
88 * Put the data of the requested offset at that
89 * offset within the buffer. Return the number (n)
90 * of bytes there are from the beginning of the
91 * buffer up to the last byte of data. If the
92 * number of supplied bytes (= n - offset) is
93 * greater than zero and you didn't signal eof
94 * and the reader is prepared to take more data
95 * you will be called again with the requested
96 * offset advanced by the number of bytes
97 * absorbed. This interface is useful for files
98 * no larger than the buffer.
99 * 1) Set *start = an unsigned long value less than
100 * the buffer address but greater than zero.
101 * Put the data of the requested offset at the
102 * beginning of the buffer. Return the number of
103 * bytes of data placed there. If this number is
104 * greater than zero and you didn't signal eof
105 * and the reader is prepared to take more data
106 * you will be called again with the requested
107 * offset advanced by *start. This interface is
108 * useful when you have a large file consisting
109 * of a series of blocks which you want to count
110 * and return as wholes.
111 * (Hack by Paul.Russell@rustcorp.com.au)
112 * 2) Set *start = an address within the buffer.
113 * Put the data of the requested offset at *start.
114 * Return the number of bytes of data placed there.
115 * If this number is greater than zero and you
116 * didn't signal eof and the reader is prepared to
117 * take more data you will be called again with the
118 * requested offset advanced by the number of bytes
121 n
= dp
->read_proc(page
, &start
, *ppos
,
122 count
, &eof
, dp
->data
);
126 if (n
== 0) /* end of file */
128 if (n
< 0) { /* error */
137 "proc_file_read: Apparent buffer overflow!\n");
145 start
= page
+ *ppos
;
146 } else if (start
< page
) {
149 "proc_file_read: Apparent buffer overflow!\n");
154 * Don't reduce n because doing so might
155 * cut off part of a data block.
158 "proc_file_read: Read count exceeded\n");
160 } else /* start >= page */ {
161 unsigned long startoff
= (unsigned long)(start
- page
);
162 if (n
> (PAGE_SIZE
- startoff
)) {
164 "proc_file_read: Apparent buffer overflow!\n");
165 n
= PAGE_SIZE
- startoff
;
171 n
-= copy_to_user(buf
, start
< page
? page
: start
, n
);
178 *ppos
+= start
< page
? (unsigned long)start
: n
;
183 free_page((unsigned long) page
);
188 proc_file_read(struct file
*file
, char __user
*buf
, size_t nbytes
,
191 struct proc_dir_entry
*pde
= PDE(file_inode(file
));
194 spin_lock(&pde
->pde_unload_lock
);
195 if (!pde
->proc_fops
) {
196 spin_unlock(&pde
->pde_unload_lock
);
200 spin_unlock(&pde
->pde_unload_lock
);
202 rv
= __proc_file_read(file
, buf
, nbytes
, ppos
);
209 proc_file_write(struct file
*file
, const char __user
*buffer
,
210 size_t count
, loff_t
*ppos
)
212 struct proc_dir_entry
*pde
= PDE(file_inode(file
));
215 if (pde
->write_proc
) {
216 spin_lock(&pde
->pde_unload_lock
);
217 if (!pde
->proc_fops
) {
218 spin_unlock(&pde
->pde_unload_lock
);
222 spin_unlock(&pde
->pde_unload_lock
);
224 /* FIXME: does this routine need ppos? probably... */
225 rv
= pde
->write_proc(file
, buffer
, count
, pde
->data
);
233 proc_file_lseek(struct file
*file
, loff_t offset
, int orig
)
235 loff_t retval
= -EINVAL
;
238 offset
+= file
->f_pos
;
241 if (offset
< 0 || offset
> MAX_NON_LFS
)
243 file
->f_pos
= retval
= offset
;
248 static const struct file_operations proc_file_operations
= {
249 .llseek
= proc_file_lseek
,
250 .read
= proc_file_read
,
251 .write
= proc_file_write
,
254 static int proc_notify_change(struct dentry
*dentry
, struct iattr
*iattr
)
256 struct inode
*inode
= dentry
->d_inode
;
257 struct proc_dir_entry
*de
= PDE(inode
);
260 error
= inode_change_ok(inode
, iattr
);
264 setattr_copy(inode
, iattr
);
265 mark_inode_dirty(inode
);
267 de
->uid
= inode
->i_uid
;
268 de
->gid
= inode
->i_gid
;
269 de
->mode
= inode
->i_mode
;
273 static int proc_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
,
276 struct inode
*inode
= dentry
->d_inode
;
277 struct proc_dir_entry
*de
= PROC_I(inode
)->pde
;
279 set_nlink(inode
, de
->nlink
);
281 generic_fillattr(inode
, stat
);
285 static const struct inode_operations proc_file_inode_operations
= {
286 .setattr
= proc_notify_change
,
290 * This function parses a name such as "tty/driver/serial", and
291 * returns the struct proc_dir_entry for "/proc/tty/driver", and
292 * returns "serial" in residual.
294 static int __xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
295 const char **residual
)
297 const char *cp
= name
, *next
;
298 struct proc_dir_entry
*de
;
306 next
= strchr(cp
, '/');
311 for (de
= de
->subdir
; de
; de
= de
->next
) {
312 if (proc_match(len
, cp
, de
))
316 WARN(1, "name '%s'\n", name
);
326 static int xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
327 const char **residual
)
331 spin_lock(&proc_subdir_lock
);
332 rv
= __xlate_proc_name(name
, ret
, residual
);
333 spin_unlock(&proc_subdir_lock
);
337 static DEFINE_IDA(proc_inum_ida
);
338 static DEFINE_SPINLOCK(proc_inum_lock
); /* protects the above */
340 #define PROC_DYNAMIC_FIRST 0xF0000000U
343 * Return an inode number between PROC_DYNAMIC_FIRST and
344 * 0xffffffff, or zero on failure.
346 int proc_alloc_inum(unsigned int *inum
)
352 if (!ida_pre_get(&proc_inum_ida
, GFP_KERNEL
))
355 spin_lock_irq(&proc_inum_lock
);
356 error
= ida_get_new(&proc_inum_ida
, &i
);
357 spin_unlock_irq(&proc_inum_lock
);
358 if (error
== -EAGAIN
)
363 if (i
> UINT_MAX
- PROC_DYNAMIC_FIRST
) {
364 spin_lock_irq(&proc_inum_lock
);
365 ida_remove(&proc_inum_ida
, i
);
366 spin_unlock_irq(&proc_inum_lock
);
369 *inum
= PROC_DYNAMIC_FIRST
+ i
;
373 void proc_free_inum(unsigned int inum
)
376 spin_lock_irqsave(&proc_inum_lock
, flags
);
377 ida_remove(&proc_inum_ida
, inum
- PROC_DYNAMIC_FIRST
);
378 spin_unlock_irqrestore(&proc_inum_lock
, flags
);
381 static void *proc_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
383 nd_set_link(nd
, PDE(dentry
->d_inode
)->data
);
387 static const struct inode_operations proc_link_inode_operations
= {
388 .readlink
= generic_readlink
,
389 .follow_link
= proc_follow_link
,
393 * As some entries in /proc are volatile, we want to
394 * get rid of unused dentries. This could be made
395 * smarter: we could keep a "volatile" flag in the
396 * inode to indicate which ones to keep.
398 static int proc_delete_dentry(const struct dentry
* dentry
)
403 static const struct dentry_operations proc_dentry_operations
=
405 .d_delete
= proc_delete_dentry
,
409 * Don't create negative dentries here, return -ENOENT by hand
412 struct dentry
*proc_lookup_de(struct proc_dir_entry
*de
, struct inode
*dir
,
413 struct dentry
*dentry
)
417 spin_lock(&proc_subdir_lock
);
418 for (de
= de
->subdir
; de
; de
= de
->next
) {
419 if (de
->namelen
!= dentry
->d_name
.len
)
421 if (!memcmp(dentry
->d_name
.name
, de
->name
, de
->namelen
)) {
423 spin_unlock(&proc_subdir_lock
);
424 inode
= proc_get_inode(dir
->i_sb
, de
);
426 return ERR_PTR(-ENOMEM
);
427 d_set_d_op(dentry
, &proc_dentry_operations
);
428 d_add(dentry
, inode
);
432 spin_unlock(&proc_subdir_lock
);
433 return ERR_PTR(-ENOENT
);
436 struct dentry
*proc_lookup(struct inode
*dir
, struct dentry
*dentry
,
439 return proc_lookup_de(PDE(dir
), dir
, dentry
);
443 * This returns non-zero if at EOF, so that the /proc
444 * root directory can use this and check if it should
445 * continue with the <pid> entries..
447 * Note that the VFS-layer doesn't care about the return
448 * value of the readdir() call, as long as it's non-negative
451 int proc_readdir_de(struct proc_dir_entry
*de
, struct file
*filp
, void *dirent
,
456 struct inode
*inode
= file_inode(filp
);
463 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
469 if (filldir(dirent
, "..", 2, i
,
470 parent_ino(filp
->f_path
.dentry
),
477 spin_lock(&proc_subdir_lock
);
483 spin_unlock(&proc_subdir_lock
);
493 struct proc_dir_entry
*next
;
495 /* filldir passes info to user space */
497 spin_unlock(&proc_subdir_lock
);
498 if (filldir(dirent
, de
->name
, de
->namelen
, filp
->f_pos
,
499 de
->low_ino
, de
->mode
>> 12) < 0) {
503 spin_lock(&proc_subdir_lock
);
509 spin_unlock(&proc_subdir_lock
);
516 int proc_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
518 struct inode
*inode
= file_inode(filp
);
520 return proc_readdir_de(PDE(inode
), filp
, dirent
, filldir
);
524 * These are the generic /proc directory operations. They
525 * use the in-memory "struct proc_dir_entry" tree to parse
526 * the /proc directory.
528 static const struct file_operations proc_dir_operations
= {
529 .llseek
= generic_file_llseek
,
530 .read
= generic_read_dir
,
531 .readdir
= proc_readdir
,
535 * proc directories can do almost nothing..
537 static const struct inode_operations proc_dir_inode_operations
= {
538 .lookup
= proc_lookup
,
539 .getattr
= proc_getattr
,
540 .setattr
= proc_notify_change
,
543 static int proc_register(struct proc_dir_entry
* dir
, struct proc_dir_entry
* dp
)
545 struct proc_dir_entry
*tmp
;
548 ret
= proc_alloc_inum(&dp
->low_ino
);
552 if (S_ISDIR(dp
->mode
)) {
553 if (dp
->proc_iops
== NULL
) {
554 dp
->proc_fops
= &proc_dir_operations
;
555 dp
->proc_iops
= &proc_dir_inode_operations
;
558 } else if (S_ISLNK(dp
->mode
)) {
559 if (dp
->proc_iops
== NULL
)
560 dp
->proc_iops
= &proc_link_inode_operations
;
561 } else if (S_ISREG(dp
->mode
)) {
562 if (dp
->proc_fops
== NULL
)
563 dp
->proc_fops
= &proc_file_operations
;
564 if (dp
->proc_iops
== NULL
)
565 dp
->proc_iops
= &proc_file_inode_operations
;
568 spin_lock(&proc_subdir_lock
);
570 for (tmp
= dir
->subdir
; tmp
; tmp
= tmp
->next
)
571 if (strcmp(tmp
->name
, dp
->name
) == 0) {
572 WARN(1, KERN_WARNING
"proc_dir_entry '%s/%s' already registered\n",
573 dir
->name
, dp
->name
);
577 dp
->next
= dir
->subdir
;
580 spin_unlock(&proc_subdir_lock
);
585 static struct proc_dir_entry
*__proc_create(struct proc_dir_entry
**parent
,
590 struct proc_dir_entry
*ent
= NULL
;
591 const char *fn
= name
;
594 /* make sure name is valid */
595 if (!name
|| !strlen(name
))
598 if (xlate_proc_name(name
, parent
, &fn
) != 0)
601 /* At this point there must not be any '/' characters beyond *fn */
607 ent
= kzalloc(sizeof(struct proc_dir_entry
) + len
+ 1, GFP_KERNEL
);
611 memcpy(ent
->name
, fn
, len
+ 1);
615 atomic_set(&ent
->count
, 1);
616 spin_lock_init(&ent
->pde_unload_lock
);
617 INIT_LIST_HEAD(&ent
->pde_openers
);
622 struct proc_dir_entry
*proc_symlink(const char *name
,
623 struct proc_dir_entry
*parent
, const char *dest
)
625 struct proc_dir_entry
*ent
;
627 ent
= __proc_create(&parent
, name
,
628 (S_IFLNK
| S_IRUGO
| S_IWUGO
| S_IXUGO
),1);
631 ent
->data
= kmalloc((ent
->size
=strlen(dest
))+1, GFP_KERNEL
);
633 strcpy((char*)ent
->data
,dest
);
634 if (proc_register(parent
, ent
) < 0) {
646 EXPORT_SYMBOL(proc_symlink
);
648 struct proc_dir_entry
*proc_mkdir_mode(const char *name
, umode_t mode
,
649 struct proc_dir_entry
*parent
)
651 struct proc_dir_entry
*ent
;
653 ent
= __proc_create(&parent
, name
, S_IFDIR
| mode
, 2);
655 if (proc_register(parent
, ent
) < 0) {
662 EXPORT_SYMBOL(proc_mkdir_mode
);
664 struct proc_dir_entry
*proc_net_mkdir(struct net
*net
, const char *name
,
665 struct proc_dir_entry
*parent
)
667 struct proc_dir_entry
*ent
;
669 ent
= __proc_create(&parent
, name
, S_IFDIR
| S_IRUGO
| S_IXUGO
, 2);
672 if (proc_register(parent
, ent
) < 0) {
679 EXPORT_SYMBOL_GPL(proc_net_mkdir
);
681 struct proc_dir_entry
*proc_mkdir(const char *name
,
682 struct proc_dir_entry
*parent
)
684 return proc_mkdir_mode(name
, S_IRUGO
| S_IXUGO
, parent
);
686 EXPORT_SYMBOL(proc_mkdir
);
688 struct proc_dir_entry
*create_proc_entry(const char *name
, umode_t mode
,
689 struct proc_dir_entry
*parent
)
691 struct proc_dir_entry
*ent
;
695 if ((mode
& S_IALLUGO
) == 0)
696 mode
|= S_IRUGO
| S_IXUGO
;
699 if ((mode
& S_IFMT
) == 0)
701 if ((mode
& S_IALLUGO
) == 0)
706 ent
= __proc_create(&parent
, name
, mode
, nlink
);
708 if (proc_register(parent
, ent
) < 0) {
715 EXPORT_SYMBOL(create_proc_entry
);
717 struct proc_dir_entry
*proc_create_data(const char *name
, umode_t mode
,
718 struct proc_dir_entry
*parent
,
719 const struct file_operations
*proc_fops
,
722 struct proc_dir_entry
*pde
;
726 if ((mode
& S_IALLUGO
) == 0)
727 mode
|= S_IRUGO
| S_IXUGO
;
730 if ((mode
& S_IFMT
) == 0)
732 if ((mode
& S_IALLUGO
) == 0)
737 pde
= __proc_create(&parent
, name
, mode
, nlink
);
740 pde
->proc_fops
= proc_fops
;
742 if (proc_register(parent
, pde
) < 0)
750 EXPORT_SYMBOL(proc_create_data
);
752 static void free_proc_entry(struct proc_dir_entry
*de
)
754 proc_free_inum(de
->low_ino
);
756 if (S_ISLNK(de
->mode
))
761 void pde_put(struct proc_dir_entry
*pde
)
763 if (atomic_dec_and_test(&pde
->count
))
764 free_proc_entry(pde
);
768 * Remove a /proc entry and free it if it's not currently in use.
770 void remove_proc_entry(const char *name
, struct proc_dir_entry
*parent
)
772 struct proc_dir_entry
**p
;
773 struct proc_dir_entry
*de
= NULL
;
774 const char *fn
= name
;
777 spin_lock(&proc_subdir_lock
);
778 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
779 spin_unlock(&proc_subdir_lock
);
784 for (p
= &parent
->subdir
; *p
; p
=&(*p
)->next
) {
785 if (proc_match(len
, fn
, *p
)) {
792 spin_unlock(&proc_subdir_lock
);
794 WARN(1, "name '%s'\n", name
);
798 spin_lock(&de
->pde_unload_lock
);
800 * Stop accepting new callers into module. If you're
801 * dynamically allocating ->proc_fops, save a pointer somewhere.
803 de
->proc_fops
= NULL
;
804 /* Wait until all existing callers into module are done. */
805 if (de
->pde_users
> 0) {
806 DECLARE_COMPLETION_ONSTACK(c
);
808 if (!de
->pde_unload_completion
)
809 de
->pde_unload_completion
= &c
;
811 spin_unlock(&de
->pde_unload_lock
);
813 wait_for_completion(de
->pde_unload_completion
);
815 spin_lock(&de
->pde_unload_lock
);
818 while (!list_empty(&de
->pde_openers
)) {
819 struct pde_opener
*pdeo
;
821 pdeo
= list_first_entry(&de
->pde_openers
, struct pde_opener
, lh
);
823 spin_unlock(&de
->pde_unload_lock
);
824 pdeo
->release(pdeo
->inode
, pdeo
->file
);
826 spin_lock(&de
->pde_unload_lock
);
828 spin_unlock(&de
->pde_unload_lock
);
830 if (S_ISDIR(de
->mode
))
833 WARN(de
->subdir
, KERN_WARNING
"%s: removing non-empty directory "
834 "'%s/%s', leaking at least '%s'\n", __func__
,
835 de
->parent
->name
, de
->name
, de
->subdir
->name
);
838 EXPORT_SYMBOL(remove_proc_entry
);