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>
15 #include <linux/module.h>
16 #include <linux/mount.h>
17 #include <linux/smp_lock.h>
18 #include <linux/init.h>
19 #include <linux/idr.h>
20 #include <linux/namei.h>
21 #include <linux/bitops.h>
22 #include <asm/uaccess.h>
26 static ssize_t
proc_file_read(struct file
*file
, char __user
*buf
,
27 size_t nbytes
, loff_t
*ppos
);
28 static ssize_t
proc_file_write(struct file
*file
, const char __user
*buffer
,
29 size_t count
, loff_t
*ppos
);
30 static loff_t
proc_file_lseek(struct file
*, loff_t
, int);
32 int proc_match(int len
, const char *name
, struct proc_dir_entry
*de
)
34 if (de
->namelen
!= len
)
36 return !memcmp(name
, de
->name
, len
);
39 static struct file_operations proc_file_operations
= {
40 .llseek
= proc_file_lseek
,
41 .read
= proc_file_read
,
42 .write
= proc_file_write
,
45 /* buffer size is one page but our output routines use some slack for overruns */
46 #define PROC_BLOCK_SIZE (PAGE_SIZE - 1024)
49 proc_file_read(struct file
*file
, char __user
*buf
, size_t nbytes
,
52 struct inode
* inode
= file
->f_dentry
->d_inode
;
58 struct proc_dir_entry
* dp
;
59 unsigned long long pos
;
62 * Gaah, please just use "seq_file" instead. The legacy /proc
63 * interfaces cut loff_t down to off_t for reads, and ignore
64 * the offset entirely for writes..
67 if (pos
> MAX_NON_LFS
)
69 if (nbytes
> MAX_NON_LFS
- pos
)
70 nbytes
= MAX_NON_LFS
- pos
;
73 if (!(page
= (char*) __get_free_page(GFP_KERNEL
)))
76 while ((nbytes
> 0) && !eof
) {
77 count
= min_t(size_t, PROC_BLOCK_SIZE
, nbytes
);
81 /* Handle old net routines */
82 n
= dp
->get_info(page
, &start
, *ppos
, count
);
85 } else if (dp
->read_proc
) {
87 * How to be a proc read function
88 * ------------------------------
90 * int f(char *buffer, char **start, off_t offset,
91 * int count, int *peof, void *dat)
93 * Assume that the buffer is "count" bytes in size.
95 * If you know you have supplied all the data you
98 * You have three ways to return data:
99 * 0) Leave *start = NULL. (This is the default.)
100 * Put the data of the requested offset at that
101 * offset within the buffer. Return the number (n)
102 * of bytes there are from the beginning of the
103 * buffer up to the last byte of data. If the
104 * number of supplied bytes (= n - offset) 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 the number of bytes
109 * absorbed. This interface is useful for files
110 * no larger than the buffer.
111 * 1) Set *start = an unsigned long value less than
112 * the buffer address but greater than zero.
113 * Put the data of the requested offset at the
114 * beginning of the buffer. Return the number of
115 * bytes of data placed there. If this number is
116 * greater than zero and you didn't signal eof
117 * and the reader is prepared to take more data
118 * you will be called again with the requested
119 * offset advanced by *start. This interface is
120 * useful when you have a large file consisting
121 * of a series of blocks which you want to count
122 * and return as wholes.
123 * (Hack by Paul.Russell@rustcorp.com.au)
124 * 2) Set *start = an address within the buffer.
125 * Put the data of the requested offset at *start.
126 * Return the number of bytes of data placed there.
127 * If this number is greater than zero and you
128 * didn't signal eof and the reader is prepared to
129 * take more data you will be called again with the
130 * requested offset advanced by the number of bytes
133 n
= dp
->read_proc(page
, &start
, *ppos
,
134 count
, &eof
, dp
->data
);
138 if (n
== 0) /* end of file */
140 if (n
< 0) { /* error */
149 "proc_file_read: Apparent buffer overflow!\n");
157 start
= page
+ *ppos
;
158 } else if (start
< page
) {
161 "proc_file_read: Apparent buffer overflow!\n");
166 * Don't reduce n because doing so might
167 * cut off part of a data block.
170 "proc_file_read: Read count exceeded\n");
172 } else /* start >= page */ {
173 unsigned long startoff
= (unsigned long)(start
- page
);
174 if (n
> (PAGE_SIZE
- startoff
)) {
176 "proc_file_read: Apparent buffer overflow!\n");
177 n
= PAGE_SIZE
- startoff
;
183 n
-= copy_to_user(buf
, start
< page
? page
: start
, n
);
190 *ppos
+= start
< page
? (unsigned long)start
: n
;
195 free_page((unsigned long) page
);
200 proc_file_write(struct file
*file
, const char __user
*buffer
,
201 size_t count
, loff_t
*ppos
)
203 struct inode
*inode
= file
->f_dentry
->d_inode
;
204 struct proc_dir_entry
* dp
;
211 /* FIXME: does this routine need ppos? probably... */
212 return dp
->write_proc(file
, buffer
, count
, dp
->data
);
217 proc_file_lseek(struct file
*file
, loff_t offset
, int orig
)
219 loff_t retval
= -EINVAL
;
222 offset
+= file
->f_pos
;
225 if (offset
< 0 || offset
> MAX_NON_LFS
)
227 file
->f_pos
= retval
= offset
;
232 static int proc_notify_change(struct dentry
*dentry
, struct iattr
*iattr
)
234 struct inode
*inode
= dentry
->d_inode
;
235 struct proc_dir_entry
*de
= PDE(inode
);
238 error
= inode_change_ok(inode
, iattr
);
242 error
= inode_setattr(inode
, iattr
);
246 de
->uid
= inode
->i_uid
;
247 de
->gid
= inode
->i_gid
;
248 de
->mode
= inode
->i_mode
;
253 static int proc_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
,
256 struct inode
*inode
= dentry
->d_inode
;
257 struct proc_dir_entry
*de
= PROC_I(inode
)->pde
;
259 inode
->i_nlink
= de
->nlink
;
261 generic_fillattr(inode
, stat
);
265 static struct inode_operations proc_file_inode_operations
= {
266 .setattr
= proc_notify_change
,
270 * This function parses a name such as "tty/driver/serial", and
271 * returns the struct proc_dir_entry for "/proc/tty/driver", and
272 * returns "serial" in residual.
274 static int xlate_proc_name(const char *name
,
275 struct proc_dir_entry
**ret
, const char **residual
)
277 const char *cp
= name
, *next
;
278 struct proc_dir_entry
*de
;
283 next
= strchr(cp
, '/');
288 for (de
= de
->subdir
; de
; de
= de
->next
) {
289 if (proc_match(len
, cp
, de
))
301 static DEFINE_IDR(proc_inum_idr
);
302 static DEFINE_SPINLOCK(proc_inum_lock
); /* protects the above */
304 #define PROC_DYNAMIC_FIRST 0xF0000000UL
307 * Return an inode number between PROC_DYNAMIC_FIRST and
308 * 0xffffffff, or zero on failure.
310 static unsigned int get_inode_number(void)
316 if (idr_pre_get(&proc_inum_idr
, GFP_KERNEL
) == 0)
319 spin_lock(&proc_inum_lock
);
320 error
= idr_get_new(&proc_inum_idr
, NULL
, &i
);
321 spin_unlock(&proc_inum_lock
);
322 if (error
== -EAGAIN
)
327 inum
= (i
& MAX_ID_MASK
) + PROC_DYNAMIC_FIRST
;
329 /* inum will never be more than 0xf0ffffff, so no check
336 static void release_inode_number(unsigned int inum
)
338 int id
= (inum
- PROC_DYNAMIC_FIRST
) | ~MAX_ID_MASK
;
340 spin_lock(&proc_inum_lock
);
341 idr_remove(&proc_inum_idr
, id
);
342 spin_unlock(&proc_inum_lock
);
345 static void *proc_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
347 nd_set_link(nd
, PDE(dentry
->d_inode
)->data
);
351 static struct inode_operations proc_link_inode_operations
= {
352 .readlink
= generic_readlink
,
353 .follow_link
= proc_follow_link
,
357 * As some entries in /proc are volatile, we want to
358 * get rid of unused dentries. This could be made
359 * smarter: we could keep a "volatile" flag in the
360 * inode to indicate which ones to keep.
362 static int proc_delete_dentry(struct dentry
* dentry
)
367 static struct dentry_operations proc_dentry_operations
=
369 .d_delete
= proc_delete_dentry
,
373 * Don't create negative dentries here, return -ENOENT by hand
376 struct dentry
*proc_lookup(struct inode
* dir
, struct dentry
*dentry
, struct nameidata
*nd
)
378 struct inode
*inode
= NULL
;
379 struct proc_dir_entry
* de
;
385 for (de
= de
->subdir
; de
; de
= de
->next
) {
386 if (de
->namelen
!= dentry
->d_name
.len
)
388 if (!memcmp(dentry
->d_name
.name
, de
->name
, de
->namelen
)) {
389 unsigned int ino
= de
->low_ino
;
392 inode
= proc_get_inode(dir
->i_sb
, ino
, de
);
400 dentry
->d_op
= &proc_dentry_operations
;
401 d_add(dentry
, inode
);
404 return ERR_PTR(error
);
408 * This returns non-zero if at EOF, so that the /proc
409 * root directory can use this and check if it should
410 * continue with the <pid> entries..
412 * Note that the VFS-layer doesn't care about the return
413 * value of the readdir() call, as long as it's non-negative
416 int proc_readdir(struct file
* filp
,
417 void * dirent
, filldir_t filldir
)
419 struct proc_dir_entry
* de
;
422 struct inode
*inode
= filp
->f_dentry
->d_inode
;
436 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
442 if (filldir(dirent
, "..", 2, i
,
443 parent_ino(filp
->f_dentry
),
464 if (filldir(dirent
, de
->name
, de
->namelen
, filp
->f_pos
,
465 de
->low_ino
, de
->mode
>> 12) < 0)
472 out
: unlock_kernel();
477 * These are the generic /proc directory operations. They
478 * use the in-memory "struct proc_dir_entry" tree to parse
479 * the /proc directory.
481 static struct file_operations proc_dir_operations
= {
482 .read
= generic_read_dir
,
483 .readdir
= proc_readdir
,
487 * proc directories can do almost nothing..
489 static struct inode_operations proc_dir_inode_operations
= {
490 .lookup
= proc_lookup
,
491 .getattr
= proc_getattr
,
492 .setattr
= proc_notify_change
,
495 static int proc_register(struct proc_dir_entry
* dir
, struct proc_dir_entry
* dp
)
499 i
= get_inode_number();
503 dp
->next
= dir
->subdir
;
506 if (S_ISDIR(dp
->mode
)) {
507 if (dp
->proc_iops
== NULL
) {
508 dp
->proc_fops
= &proc_dir_operations
;
509 dp
->proc_iops
= &proc_dir_inode_operations
;
512 } else if (S_ISLNK(dp
->mode
)) {
513 if (dp
->proc_iops
== NULL
)
514 dp
->proc_iops
= &proc_link_inode_operations
;
515 } else if (S_ISREG(dp
->mode
)) {
516 if (dp
->proc_fops
== NULL
)
517 dp
->proc_fops
= &proc_file_operations
;
518 if (dp
->proc_iops
== NULL
)
519 dp
->proc_iops
= &proc_file_inode_operations
;
525 * Kill an inode that got unregistered..
527 static void proc_kill_inodes(struct proc_dir_entry
*de
)
530 struct super_block
*sb
= proc_mnt
->mnt_sb
;
533 * Actually it's a partial revoke().
536 list_for_each(p
, &sb
->s_files
) {
537 struct file
* filp
= list_entry(p
, struct file
, f_u
.fu_list
);
538 struct dentry
* dentry
= filp
->f_dentry
;
539 struct inode
* inode
;
540 struct file_operations
*fops
;
542 if (dentry
->d_op
!= &proc_dentry_operations
)
544 inode
= dentry
->d_inode
;
545 if (PDE(inode
) != de
)
554 static struct proc_dir_entry
*proc_create(struct proc_dir_entry
**parent
,
559 struct proc_dir_entry
*ent
= NULL
;
560 const char *fn
= name
;
563 /* make sure name is valid */
564 if (!name
|| !strlen(name
)) goto out
;
566 if (!(*parent
) && xlate_proc_name(name
, parent
, &fn
) != 0)
569 /* At this point there must not be any '/' characters beyond *fn */
575 ent
= kmalloc(sizeof(struct proc_dir_entry
) + len
+ 1, GFP_KERNEL
);
578 memset(ent
, 0, sizeof(struct proc_dir_entry
));
579 memcpy(((char *) ent
) + sizeof(struct proc_dir_entry
), fn
, len
+ 1);
580 ent
->name
= ((char *) ent
) + sizeof(*ent
);
588 struct proc_dir_entry
*proc_symlink(const char *name
,
589 struct proc_dir_entry
*parent
, const char *dest
)
591 struct proc_dir_entry
*ent
;
593 ent
= proc_create(&parent
,name
,
594 (S_IFLNK
| S_IRUGO
| S_IWUGO
| S_IXUGO
),1);
597 ent
->data
= kmalloc((ent
->size
=strlen(dest
))+1, GFP_KERNEL
);
599 strcpy((char*)ent
->data
,dest
);
600 if (proc_register(parent
, ent
) < 0) {
613 struct proc_dir_entry
*proc_mkdir_mode(const char *name
, mode_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 ent
->proc_fops
= &proc_dir_operations
;
621 ent
->proc_iops
= &proc_dir_inode_operations
;
623 if (proc_register(parent
, ent
) < 0) {
631 struct proc_dir_entry
*proc_mkdir(const char *name
,
632 struct proc_dir_entry
*parent
)
634 return proc_mkdir_mode(name
, S_IRUGO
| S_IXUGO
, parent
);
637 struct proc_dir_entry
*create_proc_entry(const char *name
, mode_t mode
,
638 struct proc_dir_entry
*parent
)
640 struct proc_dir_entry
*ent
;
644 if ((mode
& S_IALLUGO
) == 0)
645 mode
|= S_IRUGO
| S_IXUGO
;
648 if ((mode
& S_IFMT
) == 0)
650 if ((mode
& S_IALLUGO
) == 0)
655 ent
= proc_create(&parent
,name
,mode
,nlink
);
658 ent
->proc_fops
= &proc_dir_operations
;
659 ent
->proc_iops
= &proc_dir_inode_operations
;
661 if (proc_register(parent
, ent
) < 0) {
669 void free_proc_entry(struct proc_dir_entry
*de
)
671 unsigned int ino
= de
->low_ino
;
673 if (ino
< PROC_DYNAMIC_FIRST
)
676 release_inode_number(ino
);
678 if (S_ISLNK(de
->mode
) && de
->data
)
684 * Remove a /proc entry and free it if it's not currently in use.
685 * If it is in use, we set the 'deleted' flag.
687 void remove_proc_entry(const char *name
, struct proc_dir_entry
*parent
)
689 struct proc_dir_entry
**p
;
690 struct proc_dir_entry
*de
;
691 const char *fn
= name
;
694 if (!parent
&& xlate_proc_name(name
, &parent
, &fn
) != 0)
697 for (p
= &parent
->subdir
; *p
; p
=&(*p
)->next
) {
698 if (!proc_match(len
, fn
, *p
))
703 if (S_ISDIR(de
->mode
))
705 proc_kill_inodes(de
);
708 if (!atomic_read(&de
->count
))
712 printk("remove_proc_entry: %s/%s busy, count=%d\n",
713 parent
->name
, de
->name
, atomic_read(&de
->count
));