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(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
->f_path
.dentry
->d_inode
;
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
->f_path
.dentry
->d_inode
);
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
->f_path
.dentry
->d_inode
);
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 if ((iattr
->ia_valid
& ATTR_SIZE
) &&
265 iattr
->ia_size
!= i_size_read(inode
)) {
266 error
= vmtruncate(inode
, iattr
->ia_size
);
271 setattr_copy(inode
, iattr
);
272 mark_inode_dirty(inode
);
274 de
->uid
= inode
->i_uid
;
275 de
->gid
= inode
->i_gid
;
276 de
->mode
= inode
->i_mode
;
280 static int proc_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
,
283 struct inode
*inode
= dentry
->d_inode
;
284 struct proc_dir_entry
*de
= PROC_I(inode
)->pde
;
286 inode
->i_nlink
= de
->nlink
;
288 generic_fillattr(inode
, stat
);
292 static const struct inode_operations proc_file_inode_operations
= {
293 .setattr
= proc_notify_change
,
297 * This function parses a name such as "tty/driver/serial", and
298 * returns the struct proc_dir_entry for "/proc/tty/driver", and
299 * returns "serial" in residual.
301 static int __xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
302 const char **residual
)
304 const char *cp
= name
, *next
;
305 struct proc_dir_entry
*de
;
313 next
= strchr(cp
, '/');
318 for (de
= de
->subdir
; de
; de
= de
->next
) {
319 if (proc_match(len
, cp
, de
))
323 WARN(1, "name '%s'\n", name
);
333 static int xlate_proc_name(const char *name
, struct proc_dir_entry
**ret
,
334 const char **residual
)
338 spin_lock(&proc_subdir_lock
);
339 rv
= __xlate_proc_name(name
, ret
, residual
);
340 spin_unlock(&proc_subdir_lock
);
344 static DEFINE_IDA(proc_inum_ida
);
345 static DEFINE_SPINLOCK(proc_inum_lock
); /* protects the above */
347 #define PROC_DYNAMIC_FIRST 0xF0000000U
350 * Return an inode number between PROC_DYNAMIC_FIRST and
351 * 0xffffffff, or zero on failure.
353 static unsigned int get_inode_number(void)
359 if (ida_pre_get(&proc_inum_ida
, GFP_KERNEL
) == 0)
362 spin_lock(&proc_inum_lock
);
363 error
= ida_get_new(&proc_inum_ida
, &i
);
364 spin_unlock(&proc_inum_lock
);
365 if (error
== -EAGAIN
)
370 if (i
> UINT_MAX
- PROC_DYNAMIC_FIRST
) {
371 spin_lock(&proc_inum_lock
);
372 ida_remove(&proc_inum_ida
, i
);
373 spin_unlock(&proc_inum_lock
);
376 return PROC_DYNAMIC_FIRST
+ i
;
379 static void release_inode_number(unsigned int inum
)
381 spin_lock(&proc_inum_lock
);
382 ida_remove(&proc_inum_ida
, inum
- PROC_DYNAMIC_FIRST
);
383 spin_unlock(&proc_inum_lock
);
386 static void *proc_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
388 nd_set_link(nd
, PDE(dentry
->d_inode
)->data
);
392 static const struct inode_operations proc_link_inode_operations
= {
393 .readlink
= generic_readlink
,
394 .follow_link
= proc_follow_link
,
398 * As some entries in /proc are volatile, we want to
399 * get rid of unused dentries. This could be made
400 * smarter: we could keep a "volatile" flag in the
401 * inode to indicate which ones to keep.
403 static int proc_delete_dentry(struct dentry
* dentry
)
408 static const struct dentry_operations proc_dentry_operations
=
410 .d_delete
= proc_delete_dentry
,
414 * Don't create negative dentries here, return -ENOENT by hand
417 struct dentry
*proc_lookup_de(struct proc_dir_entry
*de
, struct inode
*dir
,
418 struct dentry
*dentry
)
420 struct inode
*inode
= NULL
;
423 spin_lock(&proc_subdir_lock
);
424 for (de
= de
->subdir
; de
; de
= de
->next
) {
425 if (de
->namelen
!= dentry
->d_name
.len
)
427 if (!memcmp(dentry
->d_name
.name
, de
->name
, de
->namelen
)) {
432 spin_unlock(&proc_subdir_lock
);
434 inode
= proc_get_inode(dir
->i_sb
, ino
, de
);
438 spin_unlock(&proc_subdir_lock
);
442 dentry
->d_op
= &proc_dentry_operations
;
443 d_add(dentry
, inode
);
448 return ERR_PTR(error
);
451 struct dentry
*proc_lookup(struct inode
*dir
, struct dentry
*dentry
,
452 struct nameidata
*nd
)
454 return proc_lookup_de(PDE(dir
), dir
, dentry
);
458 * This returns non-zero if at EOF, so that the /proc
459 * root directory can use this and check if it should
460 * continue with the <pid> entries..
462 * Note that the VFS-layer doesn't care about the return
463 * value of the readdir() call, as long as it's non-negative
466 int proc_readdir_de(struct proc_dir_entry
*de
, struct file
*filp
, void *dirent
,
471 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
478 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
484 if (filldir(dirent
, "..", 2, i
,
485 parent_ino(filp
->f_path
.dentry
),
492 spin_lock(&proc_subdir_lock
);
498 spin_unlock(&proc_subdir_lock
);
508 struct proc_dir_entry
*next
;
510 /* filldir passes info to user space */
512 spin_unlock(&proc_subdir_lock
);
513 if (filldir(dirent
, de
->name
, de
->namelen
, filp
->f_pos
,
514 de
->low_ino
, de
->mode
>> 12) < 0) {
518 spin_lock(&proc_subdir_lock
);
524 spin_unlock(&proc_subdir_lock
);
531 int proc_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
533 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
535 return proc_readdir_de(PDE(inode
), filp
, dirent
, filldir
);
539 * These are the generic /proc directory operations. They
540 * use the in-memory "struct proc_dir_entry" tree to parse
541 * the /proc directory.
543 static const struct file_operations proc_dir_operations
= {
544 .llseek
= generic_file_llseek
,
545 .read
= generic_read_dir
,
546 .readdir
= proc_readdir
,
550 * proc directories can do almost nothing..
552 static const struct inode_operations proc_dir_inode_operations
= {
553 .lookup
= proc_lookup
,
554 .getattr
= proc_getattr
,
555 .setattr
= proc_notify_change
,
558 static int proc_register(struct proc_dir_entry
* dir
, struct proc_dir_entry
* dp
)
561 struct proc_dir_entry
*tmp
;
563 i
= get_inode_number();
568 if (S_ISDIR(dp
->mode
)) {
569 if (dp
->proc_iops
== NULL
) {
570 dp
->proc_fops
= &proc_dir_operations
;
571 dp
->proc_iops
= &proc_dir_inode_operations
;
574 } else if (S_ISLNK(dp
->mode
)) {
575 if (dp
->proc_iops
== NULL
)
576 dp
->proc_iops
= &proc_link_inode_operations
;
577 } else if (S_ISREG(dp
->mode
)) {
578 if (dp
->proc_fops
== NULL
)
579 dp
->proc_fops
= &proc_file_operations
;
580 if (dp
->proc_iops
== NULL
)
581 dp
->proc_iops
= &proc_file_inode_operations
;
584 spin_lock(&proc_subdir_lock
);
586 for (tmp
= dir
->subdir
; tmp
; tmp
= tmp
->next
)
587 if (strcmp(tmp
->name
, dp
->name
) == 0) {
588 WARN(1, KERN_WARNING
"proc_dir_entry '%s/%s' already registered\n",
589 dir
->name
, dp
->name
);
593 dp
->next
= dir
->subdir
;
596 spin_unlock(&proc_subdir_lock
);
601 static struct proc_dir_entry
*__proc_create(struct proc_dir_entry
**parent
,
606 struct proc_dir_entry
*ent
= NULL
;
607 const char *fn
= name
;
610 /* make sure name is valid */
611 if (!name
|| !strlen(name
)) goto out
;
613 if (xlate_proc_name(name
, parent
, &fn
) != 0)
616 /* At this point there must not be any '/' characters beyond *fn */
622 ent
= kmalloc(sizeof(struct proc_dir_entry
) + len
+ 1, GFP_KERNEL
);
625 memset(ent
, 0, sizeof(struct proc_dir_entry
));
626 memcpy(((char *) ent
) + sizeof(struct proc_dir_entry
), fn
, len
+ 1);
627 ent
->name
= ((char *) ent
) + sizeof(*ent
);
631 atomic_set(&ent
->count
, 1);
633 spin_lock_init(&ent
->pde_unload_lock
);
634 ent
->pde_unload_completion
= NULL
;
635 INIT_LIST_HEAD(&ent
->pde_openers
);
640 struct proc_dir_entry
*proc_symlink(const char *name
,
641 struct proc_dir_entry
*parent
, const char *dest
)
643 struct proc_dir_entry
*ent
;
645 ent
= __proc_create(&parent
, name
,
646 (S_IFLNK
| S_IRUGO
| S_IWUGO
| S_IXUGO
),1);
649 ent
->data
= kmalloc((ent
->size
=strlen(dest
))+1, GFP_KERNEL
);
651 strcpy((char*)ent
->data
,dest
);
652 if (proc_register(parent
, ent
) < 0) {
664 EXPORT_SYMBOL(proc_symlink
);
666 struct proc_dir_entry
*proc_mkdir_mode(const char *name
, mode_t mode
,
667 struct proc_dir_entry
*parent
)
669 struct proc_dir_entry
*ent
;
671 ent
= __proc_create(&parent
, name
, S_IFDIR
| mode
, 2);
673 if (proc_register(parent
, ent
) < 0) {
681 struct proc_dir_entry
*proc_net_mkdir(struct net
*net
, const char *name
,
682 struct proc_dir_entry
*parent
)
684 struct proc_dir_entry
*ent
;
686 ent
= __proc_create(&parent
, name
, S_IFDIR
| S_IRUGO
| S_IXUGO
, 2);
689 if (proc_register(parent
, ent
) < 0) {
696 EXPORT_SYMBOL_GPL(proc_net_mkdir
);
698 struct proc_dir_entry
*proc_mkdir(const char *name
,
699 struct proc_dir_entry
*parent
)
701 return proc_mkdir_mode(name
, S_IRUGO
| S_IXUGO
, parent
);
703 EXPORT_SYMBOL(proc_mkdir
);
705 struct proc_dir_entry
*create_proc_entry(const char *name
, mode_t mode
,
706 struct proc_dir_entry
*parent
)
708 struct proc_dir_entry
*ent
;
712 if ((mode
& S_IALLUGO
) == 0)
713 mode
|= S_IRUGO
| S_IXUGO
;
716 if ((mode
& S_IFMT
) == 0)
718 if ((mode
& S_IALLUGO
) == 0)
723 ent
= __proc_create(&parent
, name
, mode
, nlink
);
725 if (proc_register(parent
, ent
) < 0) {
732 EXPORT_SYMBOL(create_proc_entry
);
734 struct proc_dir_entry
*proc_create_data(const char *name
, mode_t mode
,
735 struct proc_dir_entry
*parent
,
736 const struct file_operations
*proc_fops
,
739 struct proc_dir_entry
*pde
;
743 if ((mode
& S_IALLUGO
) == 0)
744 mode
|= S_IRUGO
| S_IXUGO
;
747 if ((mode
& S_IFMT
) == 0)
749 if ((mode
& S_IALLUGO
) == 0)
754 pde
= __proc_create(&parent
, name
, mode
, nlink
);
757 pde
->proc_fops
= proc_fops
;
759 if (proc_register(parent
, pde
) < 0)
767 EXPORT_SYMBOL(proc_create_data
);
769 static void free_proc_entry(struct proc_dir_entry
*de
)
771 unsigned int ino
= de
->low_ino
;
773 if (ino
< PROC_DYNAMIC_FIRST
)
776 release_inode_number(ino
);
778 if (S_ISLNK(de
->mode
))
783 void pde_put(struct proc_dir_entry
*pde
)
785 if (atomic_dec_and_test(&pde
->count
))
786 free_proc_entry(pde
);
790 * Remove a /proc entry and free it if it's not currently in use.
792 void remove_proc_entry(const char *name
, struct proc_dir_entry
*parent
)
794 struct proc_dir_entry
**p
;
795 struct proc_dir_entry
*de
= NULL
;
796 const char *fn
= name
;
799 spin_lock(&proc_subdir_lock
);
800 if (__xlate_proc_name(name
, &parent
, &fn
) != 0) {
801 spin_unlock(&proc_subdir_lock
);
806 for (p
= &parent
->subdir
; *p
; p
=&(*p
)->next
) {
807 if (proc_match(len
, fn
, *p
)) {
814 spin_unlock(&proc_subdir_lock
);
816 WARN(1, "name '%s'\n", name
);
820 spin_lock(&de
->pde_unload_lock
);
822 * Stop accepting new callers into module. If you're
823 * dynamically allocating ->proc_fops, save a pointer somewhere.
825 de
->proc_fops
= NULL
;
826 /* Wait until all existing callers into module are done. */
827 if (de
->pde_users
> 0) {
828 DECLARE_COMPLETION_ONSTACK(c
);
830 if (!de
->pde_unload_completion
)
831 de
->pde_unload_completion
= &c
;
833 spin_unlock(&de
->pde_unload_lock
);
835 wait_for_completion(de
->pde_unload_completion
);
837 goto continue_removing
;
839 spin_unlock(&de
->pde_unload_lock
);
842 spin_lock(&de
->pde_unload_lock
);
843 while (!list_empty(&de
->pde_openers
)) {
844 struct pde_opener
*pdeo
;
846 pdeo
= list_first_entry(&de
->pde_openers
, struct pde_opener
, lh
);
848 spin_unlock(&de
->pde_unload_lock
);
849 pdeo
->release(pdeo
->inode
, pdeo
->file
);
851 spin_lock(&de
->pde_unload_lock
);
853 spin_unlock(&de
->pde_unload_lock
);
855 if (S_ISDIR(de
->mode
))
858 WARN(de
->subdir
, KERN_WARNING
"%s: removing non-empty directory "
859 "'%s/%s', leaking at least '%s'\n", __func__
,
860 de
->parent
->name
, de
->name
, de
->subdir
->name
);
863 EXPORT_SYMBOL(remove_proc_entry
);