blob: db7fa5cab988b1679ebc80a8135cb838a167eda5 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * proc/fs/generic.c --- generic routines for the proc-fs
3 *
4 * This file contains generic proc-fs routines for handling
5 * directories and files.
6 *
7 * Copyright (C) 1991, 1992 Linus Torvalds.
8 * Copyright (C) 1997 Theodore Ts'o
9 */
10
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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/init.h>
18#include <linux/idr.h>
19#include <linux/namei.h>
20#include <linux/bitops.h>
Steven Rostedt64a07bd2006-03-26 01:36:55 -080021#include <linux/spinlock.h>
Alexey Dobriyan786d7e12007-07-15 23:39:00 -070022#include <linux/completion.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#include <asm/uaccess.h>
24
Adrian Bunkfee781e2006-01-08 01:04:16 -080025#include "internal.h"
26
Steven Rostedt64a07bd2006-03-26 01:36:55 -080027DEFINE_SPINLOCK(proc_subdir_lock);
28
Eric W. Biederman77b14db2007-02-14 00:34:12 -080029static int proc_match(int len, const char *name, struct proc_dir_entry *de)
Linus Torvalds1da177e2005-04-16 15:20:36 -070030{
31 if (de->namelen != len)
32 return 0;
33 return !memcmp(name, de->name, len);
34}
35
Linus Torvalds1da177e2005-04-16 15:20:36 -070036/* buffer size is one page but our output routines use some slack for overruns */
37#define PROC_BLOCK_SIZE (PAGE_SIZE - 1024)
38
39static ssize_t
40proc_file_read(struct file *file, char __user *buf, size_t nbytes,
41 loff_t *ppos)
42{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -080043 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070044 char *page;
45 ssize_t retval=0;
46 int eof=0;
47 ssize_t n, count;
48 char *start;
49 struct proc_dir_entry * dp;
Linus Torvalds8b90db02005-12-30 08:39:10 -080050 unsigned long long pos;
51
52 /*
53 * Gaah, please just use "seq_file" instead. The legacy /proc
54 * interfaces cut loff_t down to off_t for reads, and ignore
55 * the offset entirely for writes..
56 */
57 pos = *ppos;
58 if (pos > MAX_NON_LFS)
59 return 0;
60 if (nbytes > MAX_NON_LFS - pos)
61 nbytes = MAX_NON_LFS - pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -070062
63 dp = PDE(inode);
Mel Gormane12ba742007-10-16 01:25:52 -070064 if (!(page = (char*) __get_free_page(GFP_TEMPORARY)))
Linus Torvalds1da177e2005-04-16 15:20:36 -070065 return -ENOMEM;
66
67 while ((nbytes > 0) && !eof) {
68 count = min_t(size_t, PROC_BLOCK_SIZE, nbytes);
69
70 start = NULL;
Alexey Dobriyan8731f142008-04-29 01:01:58 -070071 if (dp->read_proc) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 /*
73 * How to be a proc read function
74 * ------------------------------
75 * Prototype:
76 * int f(char *buffer, char **start, off_t offset,
77 * int count, int *peof, void *dat)
78 *
79 * Assume that the buffer is "count" bytes in size.
80 *
81 * If you know you have supplied all the data you
82 * have, set *peof.
83 *
84 * You have three ways to return data:
85 * 0) Leave *start = NULL. (This is the default.)
86 * Put the data of the requested offset at that
87 * offset within the buffer. Return the number (n)
88 * of bytes there are from the beginning of the
89 * buffer up to the last byte of data. If the
90 * number of supplied bytes (= n - offset) is
91 * greater than zero and you didn't signal eof
92 * and the reader is prepared to take more data
93 * you will be called again with the requested
94 * offset advanced by the number of bytes
95 * absorbed. This interface is useful for files
96 * no larger than the buffer.
97 * 1) Set *start = an unsigned long value less than
98 * the buffer address but greater than zero.
99 * Put the data of the requested offset at the
100 * beginning of the buffer. Return the number of
101 * bytes of data placed there. If this number is
102 * greater than zero and you didn't signal eof
103 * and the reader is prepared to take more data
104 * you will be called again with the requested
105 * offset advanced by *start. This interface is
106 * useful when you have a large file consisting
107 * of a series of blocks which you want to count
108 * and return as wholes.
109 * (Hack by Paul.Russell@rustcorp.com.au)
110 * 2) Set *start = an address within the buffer.
111 * Put the data of the requested offset at *start.
112 * Return the number of bytes of data placed there.
113 * If this number is greater than zero and you
114 * didn't signal eof and the reader is prepared to
115 * take more data you will be called again with the
116 * requested offset advanced by the number of bytes
117 * absorbed.
118 */
119 n = dp->read_proc(page, &start, *ppos,
120 count, &eof, dp->data);
121 } else
122 break;
123
124 if (n == 0) /* end of file */
125 break;
126 if (n < 0) { /* error */
127 if (retval == 0)
128 retval = n;
129 break;
130 }
131
132 if (start == NULL) {
133 if (n > PAGE_SIZE) {
134 printk(KERN_ERR
135 "proc_file_read: Apparent buffer overflow!\n");
136 n = PAGE_SIZE;
137 }
138 n -= *ppos;
139 if (n <= 0)
140 break;
141 if (n > count)
142 n = count;
143 start = page + *ppos;
144 } else if (start < page) {
145 if (n > PAGE_SIZE) {
146 printk(KERN_ERR
147 "proc_file_read: Apparent buffer overflow!\n");
148 n = PAGE_SIZE;
149 }
150 if (n > count) {
151 /*
152 * Don't reduce n because doing so might
153 * cut off part of a data block.
154 */
155 printk(KERN_WARNING
156 "proc_file_read: Read count exceeded\n");
157 }
158 } else /* start >= page */ {
159 unsigned long startoff = (unsigned long)(start - page);
160 if (n > (PAGE_SIZE - startoff)) {
161 printk(KERN_ERR
162 "proc_file_read: Apparent buffer overflow!\n");
163 n = PAGE_SIZE - startoff;
164 }
165 if (n > count)
166 n = count;
167 }
168
169 n -= copy_to_user(buf, start < page ? page : start, n);
170 if (n == 0) {
171 if (retval == 0)
172 retval = -EFAULT;
173 break;
174 }
175
176 *ppos += start < page ? (unsigned long)start : n;
177 nbytes -= n;
178 buf += n;
179 retval += n;
180 }
181 free_page((unsigned long) page);
182 return retval;
183}
184
185static ssize_t
186proc_file_write(struct file *file, const char __user *buffer,
187 size_t count, loff_t *ppos)
188{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800189 struct inode *inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 struct proc_dir_entry * dp;
191
192 dp = PDE(inode);
193
194 if (!dp->write_proc)
195 return -EIO;
196
197 /* FIXME: does this routine need ppos? probably... */
198 return dp->write_proc(file, buffer, count, dp->data);
199}
200
201
202static loff_t
203proc_file_lseek(struct file *file, loff_t offset, int orig)
204{
Linus Torvalds8b90db02005-12-30 08:39:10 -0800205 loff_t retval = -EINVAL;
206 switch (orig) {
207 case 1:
208 offset += file->f_pos;
209 /* fallthrough */
210 case 0:
211 if (offset < 0 || offset > MAX_NON_LFS)
212 break;
213 file->f_pos = retval = offset;
214 }
215 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216}
217
Alexey Dobriyan76df0c22008-02-08 04:18:27 -0800218static const struct file_operations proc_file_operations = {
219 .llseek = proc_file_lseek,
220 .read = proc_file_read,
221 .write = proc_file_write,
222};
223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224static int proc_notify_change(struct dentry *dentry, struct iattr *iattr)
225{
226 struct inode *inode = dentry->d_inode;
227 struct proc_dir_entry *de = PDE(inode);
228 int error;
229
230 error = inode_change_ok(inode, iattr);
231 if (error)
232 goto out;
233
234 error = inode_setattr(inode, iattr);
235 if (error)
236 goto out;
237
238 de->uid = inode->i_uid;
239 de->gid = inode->i_gid;
240 de->mode = inode->i_mode;
241out:
242 return error;
243}
244
Miklos Szeredi2b579be2005-09-06 15:17:18 -0700245static int proc_getattr(struct vfsmount *mnt, struct dentry *dentry,
246 struct kstat *stat)
247{
248 struct inode *inode = dentry->d_inode;
249 struct proc_dir_entry *de = PROC_I(inode)->pde;
250 if (de && de->nlink)
251 inode->i_nlink = de->nlink;
252
253 generic_fillattr(inode, stat);
254 return 0;
255}
256
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800257static const struct inode_operations proc_file_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258 .setattr = proc_notify_change,
259};
260
261/*
262 * This function parses a name such as "tty/driver/serial", and
263 * returns the struct proc_dir_entry for "/proc/tty/driver", and
264 * returns "serial" in residual.
265 */
266static int xlate_proc_name(const char *name,
267 struct proc_dir_entry **ret, const char **residual)
268{
269 const char *cp = name, *next;
270 struct proc_dir_entry *de;
271 int len;
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800272 int rtn = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273
Alexey Dobriyan7cee4e02008-04-29 01:01:40 -0700274 de = *ret;
275 if (!de)
276 de = &proc_root;
277
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800278 spin_lock(&proc_subdir_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 while (1) {
280 next = strchr(cp, '/');
281 if (!next)
282 break;
283
284 len = next - cp;
285 for (de = de->subdir; de ; de = de->next) {
286 if (proc_match(len, cp, de))
287 break;
288 }
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800289 if (!de) {
290 rtn = -ENOENT;
291 goto out;
292 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 cp += len + 1;
294 }
295 *residual = cp;
296 *ret = de;
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800297out:
298 spin_unlock(&proc_subdir_lock);
299 return rtn;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300}
301
Alexey Dobriyan9a185402008-07-26 11:21:37 +0400302static DEFINE_IDA(proc_inum_ida);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303static DEFINE_SPINLOCK(proc_inum_lock); /* protects the above */
304
Alexey Dobriyan67935df2008-07-26 11:18:28 +0400305#define PROC_DYNAMIC_FIRST 0xF0000000U
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306
307/*
308 * Return an inode number between PROC_DYNAMIC_FIRST and
309 * 0xffffffff, or zero on failure.
310 */
311static unsigned int get_inode_number(void)
312{
Alexey Dobriyan67935df2008-07-26 11:18:28 +0400313 unsigned int i;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 int error;
315
316retry:
Alexey Dobriyan9a185402008-07-26 11:21:37 +0400317 if (ida_pre_get(&proc_inum_ida, GFP_KERNEL) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 return 0;
319
320 spin_lock(&proc_inum_lock);
Alexey Dobriyan9a185402008-07-26 11:21:37 +0400321 error = ida_get_new(&proc_inum_ida, &i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 spin_unlock(&proc_inum_lock);
323 if (error == -EAGAIN)
324 goto retry;
325 else if (error)
326 return 0;
327
Alexey Dobriyan67935df2008-07-26 11:18:28 +0400328 if (i > UINT_MAX - PROC_DYNAMIC_FIRST) {
329 spin_lock(&proc_inum_lock);
Alexey Dobriyan9a185402008-07-26 11:21:37 +0400330 ida_remove(&proc_inum_ida, i);
Alexey Dobriyan67935df2008-07-26 11:18:28 +0400331 spin_unlock(&proc_inum_lock);
Alexey Dobriyancc996092008-08-02 07:30:48 +0400332 return 0;
Alexey Dobriyan67935df2008-07-26 11:18:28 +0400333 }
334 return PROC_DYNAMIC_FIRST + i;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335}
336
337static void release_inode_number(unsigned int inum)
338{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339 spin_lock(&proc_inum_lock);
Alexey Dobriyan9a185402008-07-26 11:21:37 +0400340 ida_remove(&proc_inum_ida, inum - PROC_DYNAMIC_FIRST);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 spin_unlock(&proc_inum_lock);
342}
343
Al Viro008b1502005-08-20 00:17:39 +0100344static void *proc_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345{
346 nd_set_link(nd, PDE(dentry->d_inode)->data);
Al Viro008b1502005-08-20 00:17:39 +0100347 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348}
349
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800350static const struct inode_operations proc_link_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 .readlink = generic_readlink,
352 .follow_link = proc_follow_link,
353};
354
355/*
356 * As some entries in /proc are volatile, we want to
357 * get rid of unused dentries. This could be made
358 * smarter: we could keep a "volatile" flag in the
359 * inode to indicate which ones to keep.
360 */
361static int proc_delete_dentry(struct dentry * dentry)
362{
363 return 1;
364}
365
366static struct dentry_operations proc_dentry_operations =
367{
368 .d_delete = proc_delete_dentry,
369};
370
371/*
372 * Don't create negative dentries here, return -ENOENT by hand
373 * instead.
374 */
Pavel Emelyanove9720ac2008-03-07 11:08:40 -0800375struct dentry *proc_lookup_de(struct proc_dir_entry *de, struct inode *dir,
376 struct dentry *dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377{
378 struct inode *inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 int error = -ENOENT;
380
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800381 spin_lock(&proc_subdir_lock);
Alexey Dobriyan5e971dc2008-04-29 01:01:41 -0700382 for (de = de->subdir; de ; de = de->next) {
383 if (de->namelen != dentry->d_name.len)
384 continue;
385 if (!memcmp(dentry->d_name.name, de->name, de->namelen)) {
386 unsigned int ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387
Alexey Dobriyan5e971dc2008-04-29 01:01:41 -0700388 ino = de->low_ino;
389 de_get(de);
390 spin_unlock(&proc_subdir_lock);
391 error = -EINVAL;
392 inode = proc_get_inode(dir->i_sb, ino, de);
393 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 }
395 }
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800396 spin_unlock(&proc_subdir_lock);
Alexey Dobriyan4237e0d2008-02-08 04:18:27 -0800397out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398
399 if (inode) {
400 dentry->d_op = &proc_dentry_operations;
401 d_add(dentry, inode);
402 return NULL;
403 }
Alexey Dobriyan5e971dc2008-04-29 01:01:41 -0700404 if (de)
405 de_put(de);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 return ERR_PTR(error);
407}
408
Pavel Emelyanove9720ac2008-03-07 11:08:40 -0800409struct dentry *proc_lookup(struct inode *dir, struct dentry *dentry,
410 struct nameidata *nd)
411{
412 return proc_lookup_de(PDE(dir), dir, dentry);
413}
414
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415/*
416 * This returns non-zero if at EOF, so that the /proc
417 * root directory can use this and check if it should
418 * continue with the <pid> entries..
419 *
420 * Note that the VFS-layer doesn't care about the return
421 * value of the readdir() call, as long as it's non-negative
422 * for success..
423 */
Pavel Emelyanove9720ac2008-03-07 11:08:40 -0800424int proc_readdir_de(struct proc_dir_entry *de, struct file *filp, void *dirent,
425 filldir_t filldir)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 unsigned int ino;
428 int i;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800429 struct inode *inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 int ret = 0;
431
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 ino = inode->i_ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 i = filp->f_pos;
434 switch (i) {
435 case 0:
436 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
437 goto out;
438 i++;
439 filp->f_pos++;
440 /* fall through */
441 case 1:
442 if (filldir(dirent, "..", 2, i,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800443 parent_ino(filp->f_path.dentry),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 DT_DIR) < 0)
445 goto out;
446 i++;
447 filp->f_pos++;
448 /* fall through */
449 default:
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800450 spin_lock(&proc_subdir_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 de = de->subdir;
452 i -= 2;
453 for (;;) {
454 if (!de) {
455 ret = 1;
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800456 spin_unlock(&proc_subdir_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 goto out;
458 }
459 if (!i)
460 break;
461 de = de->next;
462 i--;
463 }
464
465 do {
Darrick J. Wong59cd0cb2007-05-08 00:25:47 -0700466 struct proc_dir_entry *next;
467
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800468 /* filldir passes info to user space */
Darrick J. Wong59cd0cb2007-05-08 00:25:47 -0700469 de_get(de);
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800470 spin_unlock(&proc_subdir_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 if (filldir(dirent, de->name, de->namelen, filp->f_pos,
Darrick J. Wong59cd0cb2007-05-08 00:25:47 -0700472 de->low_ino, de->mode >> 12) < 0) {
473 de_put(de);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 goto out;
Darrick J. Wong59cd0cb2007-05-08 00:25:47 -0700475 }
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800476 spin_lock(&proc_subdir_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 filp->f_pos++;
Darrick J. Wong59cd0cb2007-05-08 00:25:47 -0700478 next = de->next;
479 de_put(de);
480 de = next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 } while (de);
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800482 spin_unlock(&proc_subdir_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 }
484 ret = 1;
Alexey Dobriyanb4df2b92008-10-27 22:48:36 +0300485out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 return ret;
487}
488
Pavel Emelyanove9720ac2008-03-07 11:08:40 -0800489int proc_readdir(struct file *filp, void *dirent, filldir_t filldir)
490{
491 struct inode *inode = filp->f_path.dentry->d_inode;
492
493 return proc_readdir_de(PDE(inode), filp, dirent, filldir);
494}
495
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496/*
497 * These are the generic /proc directory operations. They
498 * use the in-memory "struct proc_dir_entry" tree to parse
499 * the /proc directory.
500 */
Arjan van de Ven00977a52007-02-12 00:55:34 -0800501static const struct file_operations proc_dir_operations = {
Alexey Dobriyanb4df2b92008-10-27 22:48:36 +0300502 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 .read = generic_read_dir,
504 .readdir = proc_readdir,
505};
506
507/*
508 * proc directories can do almost nothing..
509 */
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800510static const struct inode_operations proc_dir_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 .lookup = proc_lookup,
Miklos Szeredi2b579be2005-09-06 15:17:18 -0700512 .getattr = proc_getattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 .setattr = proc_notify_change,
514};
515
516static int proc_register(struct proc_dir_entry * dir, struct proc_dir_entry * dp)
517{
518 unsigned int i;
Zhang Rui94413d82008-02-08 04:18:29 -0800519 struct proc_dir_entry *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520
521 i = get_inode_number();
522 if (i == 0)
523 return -EAGAIN;
524 dp->low_ino = i;
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800525
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 if (S_ISDIR(dp->mode)) {
527 if (dp->proc_iops == NULL) {
528 dp->proc_fops = &proc_dir_operations;
529 dp->proc_iops = &proc_dir_inode_operations;
530 }
531 dir->nlink++;
532 } else if (S_ISLNK(dp->mode)) {
533 if (dp->proc_iops == NULL)
534 dp->proc_iops = &proc_link_inode_operations;
535 } else if (S_ISREG(dp->mode)) {
536 if (dp->proc_fops == NULL)
537 dp->proc_fops = &proc_file_operations;
538 if (dp->proc_iops == NULL)
539 dp->proc_iops = &proc_file_inode_operations;
540 }
Changli Gao99fc06d2007-07-15 23:40:09 -0700541
542 spin_lock(&proc_subdir_lock);
Zhang Rui94413d82008-02-08 04:18:29 -0800543
544 for (tmp = dir->subdir; tmp; tmp = tmp->next)
545 if (strcmp(tmp->name, dp->name) == 0) {
Arjan van de Ven6c2f91e2008-09-13 19:51:30 -0700546 WARN(1, KERN_WARNING "proc_dir_entry '%s/%s' already registered\n",
Alexey Dobriyan665020c2008-09-13 02:33:06 -0700547 dir->name, dp->name);
Zhang Rui94413d82008-02-08 04:18:29 -0800548 break;
549 }
550
Changli Gao99fc06d2007-07-15 23:40:09 -0700551 dp->next = dir->subdir;
552 dp->parent = dir;
553 dir->subdir = dp;
554 spin_unlock(&proc_subdir_lock);
555
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 return 0;
557}
558
Alexey Dobriyan2d3a4e32008-02-08 04:18:37 -0800559static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 const char *name,
561 mode_t mode,
562 nlink_t nlink)
563{
564 struct proc_dir_entry *ent = NULL;
565 const char *fn = name;
566 int len;
567
568 /* make sure name is valid */
569 if (!name || !strlen(name)) goto out;
570
Alexey Dobriyan7cee4e02008-04-29 01:01:40 -0700571 if (xlate_proc_name(name, parent, &fn) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 goto out;
573
574 /* At this point there must not be any '/' characters beyond *fn */
575 if (strchr(fn, '/'))
576 goto out;
577
578 len = strlen(fn);
579
580 ent = kmalloc(sizeof(struct proc_dir_entry) + len + 1, GFP_KERNEL);
581 if (!ent) goto out;
582
583 memset(ent, 0, sizeof(struct proc_dir_entry));
584 memcpy(((char *) ent) + sizeof(struct proc_dir_entry), fn, len + 1);
585 ent->name = ((char *) ent) + sizeof(*ent);
586 ent->namelen = len;
587 ent->mode = mode;
588 ent->nlink = nlink;
Alexey Dobriyan5a622f22007-12-04 23:45:28 -0800589 atomic_set(&ent->count, 1);
Alexey Dobriyan786d7e12007-07-15 23:39:00 -0700590 ent->pde_users = 0;
591 spin_lock_init(&ent->pde_unload_lock);
592 ent->pde_unload_completion = NULL;
Alexey Dobriyan881adb82008-07-25 01:48:29 -0700593 INIT_LIST_HEAD(&ent->pde_openers);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 out:
595 return ent;
596}
597
598struct proc_dir_entry *proc_symlink(const char *name,
599 struct proc_dir_entry *parent, const char *dest)
600{
601 struct proc_dir_entry *ent;
602
Alexey Dobriyan2d3a4e32008-02-08 04:18:37 -0800603 ent = __proc_create(&parent, name,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 (S_IFLNK | S_IRUGO | S_IWUGO | S_IXUGO),1);
605
606 if (ent) {
607 ent->data = kmalloc((ent->size=strlen(dest))+1, GFP_KERNEL);
608 if (ent->data) {
609 strcpy((char*)ent->data,dest);
610 if (proc_register(parent, ent) < 0) {
611 kfree(ent->data);
612 kfree(ent);
613 ent = NULL;
614 }
615 } else {
616 kfree(ent);
617 ent = NULL;
618 }
619 }
620 return ent;
621}
622
623struct proc_dir_entry *proc_mkdir_mode(const char *name, mode_t mode,
624 struct proc_dir_entry *parent)
625{
626 struct proc_dir_entry *ent;
627
Alexey Dobriyan2d3a4e32008-02-08 04:18:37 -0800628 ent = __proc_create(&parent, name, S_IFDIR | mode, 2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 if (ent) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 if (proc_register(parent, ent) < 0) {
631 kfree(ent);
632 ent = NULL;
633 }
634 }
635 return ent;
636}
637
Denis V. Lunev78e92b92008-05-02 04:12:41 -0700638struct proc_dir_entry *proc_net_mkdir(struct net *net, const char *name,
639 struct proc_dir_entry *parent)
640{
641 struct proc_dir_entry *ent;
642
643 ent = __proc_create(&parent, name, S_IFDIR | S_IRUGO | S_IXUGO, 2);
644 if (ent) {
645 ent->data = net;
646 if (proc_register(parent, ent) < 0) {
647 kfree(ent);
648 ent = NULL;
649 }
650 }
651 return ent;
652}
653EXPORT_SYMBOL_GPL(proc_net_mkdir);
654
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655struct proc_dir_entry *proc_mkdir(const char *name,
656 struct proc_dir_entry *parent)
657{
658 return proc_mkdir_mode(name, S_IRUGO | S_IXUGO, parent);
659}
660
661struct proc_dir_entry *create_proc_entry(const char *name, mode_t mode,
662 struct proc_dir_entry *parent)
663{
664 struct proc_dir_entry *ent;
665 nlink_t nlink;
666
667 if (S_ISDIR(mode)) {
668 if ((mode & S_IALLUGO) == 0)
669 mode |= S_IRUGO | S_IXUGO;
670 nlink = 2;
671 } else {
672 if ((mode & S_IFMT) == 0)
673 mode |= S_IFREG;
674 if ((mode & S_IALLUGO) == 0)
675 mode |= S_IRUGO;
676 nlink = 1;
677 }
678
Alexey Dobriyan2d3a4e32008-02-08 04:18:37 -0800679 ent = __proc_create(&parent, name, mode, nlink);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 if (ent) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 if (proc_register(parent, ent) < 0) {
682 kfree(ent);
683 ent = NULL;
684 }
685 }
686 return ent;
687}
688
Denis V. Lunev59b74352008-04-29 01:02:00 -0700689struct proc_dir_entry *proc_create_data(const char *name, mode_t mode,
690 struct proc_dir_entry *parent,
691 const struct file_operations *proc_fops,
692 void *data)
Alexey Dobriyan2d3a4e32008-02-08 04:18:37 -0800693{
694 struct proc_dir_entry *pde;
695 nlink_t nlink;
696
697 if (S_ISDIR(mode)) {
698 if ((mode & S_IALLUGO) == 0)
699 mode |= S_IRUGO | S_IXUGO;
700 nlink = 2;
701 } else {
702 if ((mode & S_IFMT) == 0)
703 mode |= S_IFREG;
704 if ((mode & S_IALLUGO) == 0)
705 mode |= S_IRUGO;
706 nlink = 1;
707 }
708
709 pde = __proc_create(&parent, name, mode, nlink);
710 if (!pde)
711 goto out;
712 pde->proc_fops = proc_fops;
Denis V. Lunev59b74352008-04-29 01:02:00 -0700713 pde->data = data;
Alexey Dobriyan2d3a4e32008-02-08 04:18:37 -0800714 if (proc_register(parent, pde) < 0)
715 goto out_free;
716 return pde;
717out_free:
718 kfree(pde);
719out:
720 return NULL;
721}
722
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723void free_proc_entry(struct proc_dir_entry *de)
724{
725 unsigned int ino = de->low_ino;
726
727 if (ino < PROC_DYNAMIC_FIRST)
728 return;
729
730 release_inode_number(ino);
731
Alexey Dobriyanfd2cbe42008-02-08 04:18:28 -0800732 if (S_ISLNK(de->mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 kfree(de->data);
734 kfree(de);
735}
736
737/*
738 * Remove a /proc entry and free it if it's not currently in use.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 */
740void remove_proc_entry(const char *name, struct proc_dir_entry *parent)
741{
742 struct proc_dir_entry **p;
Alexey Dobriyanf649d6d2008-04-29 01:01:39 -0700743 struct proc_dir_entry *de = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 const char *fn = name;
745 int len;
746
Alexey Dobriyan7cee4e02008-04-29 01:01:40 -0700747 if (xlate_proc_name(name, &parent, &fn) != 0)
Alexey Dobriyanf649d6d2008-04-29 01:01:39 -0700748 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 len = strlen(fn);
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800750
751 spin_lock(&proc_subdir_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 for (p = &parent->subdir; *p; p=&(*p)->next ) {
Alexey Dobriyanf649d6d2008-04-29 01:01:39 -0700753 if (proc_match(len, fn, *p)) {
754 de = *p;
755 *p = de->next;
756 de->next = NULL;
757 break;
Alexey Dobriyan786d7e12007-07-15 23:39:00 -0700758 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 }
Steven Rostedt64a07bd2006-03-26 01:36:55 -0800760 spin_unlock(&proc_subdir_lock);
Alexey Dobriyanf649d6d2008-04-29 01:01:39 -0700761 if (!de)
762 return;
763
764 spin_lock(&de->pde_unload_lock);
765 /*
766 * Stop accepting new callers into module. If you're
767 * dynamically allocating ->proc_fops, save a pointer somewhere.
768 */
769 de->proc_fops = NULL;
770 /* Wait until all existing callers into module are done. */
771 if (de->pde_users > 0) {
772 DECLARE_COMPLETION_ONSTACK(c);
773
774 if (!de->pde_unload_completion)
775 de->pde_unload_completion = &c;
776
777 spin_unlock(&de->pde_unload_lock);
778
779 wait_for_completion(de->pde_unload_completion);
780
781 goto continue_removing;
782 }
783 spin_unlock(&de->pde_unload_lock);
784
785continue_removing:
Alexey Dobriyan881adb82008-07-25 01:48:29 -0700786 spin_lock(&de->pde_unload_lock);
787 while (!list_empty(&de->pde_openers)) {
788 struct pde_opener *pdeo;
789
790 pdeo = list_first_entry(&de->pde_openers, struct pde_opener, lh);
791 list_del(&pdeo->lh);
792 spin_unlock(&de->pde_unload_lock);
793 pdeo->release(pdeo->inode, pdeo->file);
794 kfree(pdeo);
795 spin_lock(&de->pde_unload_lock);
796 }
797 spin_unlock(&de->pde_unload_lock);
798
Alexey Dobriyanf649d6d2008-04-29 01:01:39 -0700799 if (S_ISDIR(de->mode))
800 parent->nlink--;
801 de->nlink = 0;
Arjan van de Ven267e2a92008-07-25 19:45:41 -0700802 WARN(de->subdir, KERN_WARNING "%s: removing non-empty directory "
Alexey Dobriyanf649d6d2008-04-29 01:01:39 -0700803 "'%s/%s', leaking at least '%s'\n", __func__,
804 de->parent->name, de->name, de->subdir->name);
Alexey Dobriyanf649d6d2008-04-29 01:01:39 -0700805 if (atomic_dec_and_test(&de->count))
806 free_proc_entry(de);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807}