Merge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
[deliverable/linux.git] / fs / proc / proc_sysctl.c
1 /*
2 * /proc/sys support
3 */
4 #include <linux/init.h>
5 #include <linux/sysctl.h>
6 #include <linux/proc_fs.h>
7 #include <linux/security.h>
8 #include <linux/namei.h>
9 #include "internal.h"
10
11 static const struct dentry_operations proc_sys_dentry_operations;
12 static const struct file_operations proc_sys_file_operations;
13 static const struct inode_operations proc_sys_inode_operations;
14 static const struct file_operations proc_sys_dir_file_operations;
15 static const struct inode_operations proc_sys_dir_operations;
16
17 static struct inode *proc_sys_make_inode(struct super_block *sb,
18 struct ctl_table_header *head, struct ctl_table *table)
19 {
20 struct inode *inode;
21 struct proc_inode *ei;
22
23 inode = new_inode(sb);
24 if (!inode)
25 goto out;
26
27 inode->i_ino = get_next_ino();
28
29 sysctl_head_get(head);
30 ei = PROC_I(inode);
31 ei->sysctl = head;
32 ei->sysctl_entry = table;
33
34 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
35 inode->i_mode = table->mode;
36 if (!table->child) {
37 inode->i_mode |= S_IFREG;
38 inode->i_op = &proc_sys_inode_operations;
39 inode->i_fop = &proc_sys_file_operations;
40 } else {
41 inode->i_mode |= S_IFDIR;
42 inode->i_nlink = 0;
43 inode->i_op = &proc_sys_dir_operations;
44 inode->i_fop = &proc_sys_dir_file_operations;
45 }
46 out:
47 return inode;
48 }
49
50 static struct ctl_table *find_in_table(struct ctl_table *p, struct qstr *name)
51 {
52 int len;
53 for ( ; p->procname; p++) {
54
55 if (!p->procname)
56 continue;
57
58 len = strlen(p->procname);
59 if (len != name->len)
60 continue;
61
62 if (memcmp(p->procname, name->name, len) != 0)
63 continue;
64
65 /* I have a match */
66 return p;
67 }
68 return NULL;
69 }
70
71 static struct ctl_table_header *grab_header(struct inode *inode)
72 {
73 if (PROC_I(inode)->sysctl)
74 return sysctl_head_grab(PROC_I(inode)->sysctl);
75 else
76 return sysctl_head_next(NULL);
77 }
78
79 static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
80 struct nameidata *nd)
81 {
82 struct ctl_table_header *head = grab_header(dir);
83 struct ctl_table *table = PROC_I(dir)->sysctl_entry;
84 struct ctl_table_header *h = NULL;
85 struct qstr *name = &dentry->d_name;
86 struct ctl_table *p;
87 struct inode *inode;
88 struct dentry *err = ERR_PTR(-ENOENT);
89
90 if (IS_ERR(head))
91 return ERR_CAST(head);
92
93 if (table && !table->child) {
94 WARN_ON(1);
95 goto out;
96 }
97
98 table = table ? table->child : head->ctl_table;
99
100 p = find_in_table(table, name);
101 if (!p) {
102 for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) {
103 if (h->attached_to != table)
104 continue;
105 p = find_in_table(h->attached_by, name);
106 if (p)
107 break;
108 }
109 }
110
111 if (!p)
112 goto out;
113
114 err = ERR_PTR(-ENOMEM);
115 inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p);
116 if (h)
117 sysctl_head_finish(h);
118
119 if (!inode)
120 goto out;
121
122 err = NULL;
123 d_set_d_op(dentry, &proc_sys_dentry_operations);
124 d_add(dentry, inode);
125
126 out:
127 sysctl_head_finish(head);
128 return err;
129 }
130
131 static ssize_t proc_sys_call_handler(struct file *filp, void __user *buf,
132 size_t count, loff_t *ppos, int write)
133 {
134 struct inode *inode = filp->f_path.dentry->d_inode;
135 struct ctl_table_header *head = grab_header(inode);
136 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
137 ssize_t error;
138 size_t res;
139
140 if (IS_ERR(head))
141 return PTR_ERR(head);
142
143 /*
144 * At this point we know that the sysctl was not unregistered
145 * and won't be until we finish.
146 */
147 error = -EPERM;
148 if (sysctl_perm(head->root, table, write ? MAY_WRITE : MAY_READ))
149 goto out;
150
151 /* if that can happen at all, it should be -EINVAL, not -EISDIR */
152 error = -EINVAL;
153 if (!table->proc_handler)
154 goto out;
155
156 /* careful: calling conventions are nasty here */
157 res = count;
158 error = table->proc_handler(table, write, buf, &res, ppos);
159 if (!error)
160 error = res;
161 out:
162 sysctl_head_finish(head);
163
164 return error;
165 }
166
167 static ssize_t proc_sys_read(struct file *filp, char __user *buf,
168 size_t count, loff_t *ppos)
169 {
170 return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 0);
171 }
172
173 static ssize_t proc_sys_write(struct file *filp, const char __user *buf,
174 size_t count, loff_t *ppos)
175 {
176 return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 1);
177 }
178
179
180 static int proc_sys_fill_cache(struct file *filp, void *dirent,
181 filldir_t filldir,
182 struct ctl_table_header *head,
183 struct ctl_table *table)
184 {
185 struct dentry *child, *dir = filp->f_path.dentry;
186 struct inode *inode;
187 struct qstr qname;
188 ino_t ino = 0;
189 unsigned type = DT_UNKNOWN;
190
191 qname.name = table->procname;
192 qname.len = strlen(table->procname);
193 qname.hash = full_name_hash(qname.name, qname.len);
194
195 child = d_lookup(dir, &qname);
196 if (!child) {
197 child = d_alloc(dir, &qname);
198 if (child) {
199 inode = proc_sys_make_inode(dir->d_sb, head, table);
200 if (!inode) {
201 dput(child);
202 return -ENOMEM;
203 } else {
204 d_set_d_op(child, &proc_sys_dentry_operations);
205 d_add(child, inode);
206 }
207 } else {
208 return -ENOMEM;
209 }
210 }
211 inode = child->d_inode;
212 ino = inode->i_ino;
213 type = inode->i_mode >> 12;
214 dput(child);
215 return !!filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type);
216 }
217
218 static int scan(struct ctl_table_header *head, ctl_table *table,
219 unsigned long *pos, struct file *file,
220 void *dirent, filldir_t filldir)
221 {
222
223 for (; table->procname; table++, (*pos)++) {
224 int res;
225
226 /* Can't do anything without a proc name */
227 if (!table->procname)
228 continue;
229
230 if (*pos < file->f_pos)
231 continue;
232
233 res = proc_sys_fill_cache(file, dirent, filldir, head, table);
234 if (res)
235 return res;
236
237 file->f_pos = *pos + 1;
238 }
239 return 0;
240 }
241
242 static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
243 {
244 struct dentry *dentry = filp->f_path.dentry;
245 struct inode *inode = dentry->d_inode;
246 struct ctl_table_header *head = grab_header(inode);
247 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
248 struct ctl_table_header *h = NULL;
249 unsigned long pos;
250 int ret = -EINVAL;
251
252 if (IS_ERR(head))
253 return PTR_ERR(head);
254
255 if (table && !table->child) {
256 WARN_ON(1);
257 goto out;
258 }
259
260 table = table ? table->child : head->ctl_table;
261
262 ret = 0;
263 /* Avoid a switch here: arm builds fail with missing __cmpdi2 */
264 if (filp->f_pos == 0) {
265 if (filldir(dirent, ".", 1, filp->f_pos,
266 inode->i_ino, DT_DIR) < 0)
267 goto out;
268 filp->f_pos++;
269 }
270 if (filp->f_pos == 1) {
271 if (filldir(dirent, "..", 2, filp->f_pos,
272 parent_ino(dentry), DT_DIR) < 0)
273 goto out;
274 filp->f_pos++;
275 }
276 pos = 2;
277
278 ret = scan(head, table, &pos, filp, dirent, filldir);
279 if (ret)
280 goto out;
281
282 for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) {
283 if (h->attached_to != table)
284 continue;
285 ret = scan(h, h->attached_by, &pos, filp, dirent, filldir);
286 if (ret) {
287 sysctl_head_finish(h);
288 break;
289 }
290 }
291 ret = 1;
292 out:
293 sysctl_head_finish(head);
294 return ret;
295 }
296
297 static int proc_sys_permission(struct inode *inode, int mask,unsigned int flags)
298 {
299 /*
300 * sysctl entries that are not writeable,
301 * are _NOT_ writeable, capabilities or not.
302 */
303 struct ctl_table_header *head;
304 struct ctl_table *table;
305 int error;
306
307 if (flags & IPERM_FLAG_RCU)
308 return -ECHILD;
309
310 /* Executable files are not allowed under /proc/sys/ */
311 if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode))
312 return -EACCES;
313
314 head = grab_header(inode);
315 if (IS_ERR(head))
316 return PTR_ERR(head);
317
318 table = PROC_I(inode)->sysctl_entry;
319 if (!table) /* global root - r-xr-xr-x */
320 error = mask & MAY_WRITE ? -EACCES : 0;
321 else /* Use the permissions on the sysctl table entry */
322 error = sysctl_perm(head->root, table, mask);
323
324 sysctl_head_finish(head);
325 return error;
326 }
327
328 static int proc_sys_setattr(struct dentry *dentry, struct iattr *attr)
329 {
330 struct inode *inode = dentry->d_inode;
331 int error;
332
333 if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID))
334 return -EPERM;
335
336 error = inode_change_ok(inode, attr);
337 if (error)
338 return error;
339
340 if ((attr->ia_valid & ATTR_SIZE) &&
341 attr->ia_size != i_size_read(inode)) {
342 error = vmtruncate(inode, attr->ia_size);
343 if (error)
344 return error;
345 }
346
347 setattr_copy(inode, attr);
348 mark_inode_dirty(inode);
349 return 0;
350 }
351
352 static int proc_sys_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
353 {
354 struct inode *inode = dentry->d_inode;
355 struct ctl_table_header *head = grab_header(inode);
356 struct ctl_table *table = PROC_I(inode)->sysctl_entry;
357
358 if (IS_ERR(head))
359 return PTR_ERR(head);
360
361 generic_fillattr(inode, stat);
362 if (table)
363 stat->mode = (stat->mode & S_IFMT) | table->mode;
364
365 sysctl_head_finish(head);
366 return 0;
367 }
368
369 static const struct file_operations proc_sys_file_operations = {
370 .read = proc_sys_read,
371 .write = proc_sys_write,
372 .llseek = default_llseek,
373 };
374
375 static const struct file_operations proc_sys_dir_file_operations = {
376 .readdir = proc_sys_readdir,
377 .llseek = generic_file_llseek,
378 };
379
380 static const struct inode_operations proc_sys_inode_operations = {
381 .permission = proc_sys_permission,
382 .setattr = proc_sys_setattr,
383 .getattr = proc_sys_getattr,
384 };
385
386 static const struct inode_operations proc_sys_dir_operations = {
387 .lookup = proc_sys_lookup,
388 .permission = proc_sys_permission,
389 .setattr = proc_sys_setattr,
390 .getattr = proc_sys_getattr,
391 };
392
393 static int proc_sys_revalidate(struct dentry *dentry, struct nameidata *nd)
394 {
395 if (nd->flags & LOOKUP_RCU)
396 return -ECHILD;
397 return !PROC_I(dentry->d_inode)->sysctl->unregistering;
398 }
399
400 static int proc_sys_delete(const struct dentry *dentry)
401 {
402 return !!PROC_I(dentry->d_inode)->sysctl->unregistering;
403 }
404
405 static int proc_sys_compare(const struct dentry *parent,
406 const struct inode *pinode,
407 const struct dentry *dentry, const struct inode *inode,
408 unsigned int len, const char *str, const struct qstr *name)
409 {
410 struct ctl_table_header *head;
411 /* Although proc doesn't have negative dentries, rcu-walk means
412 * that inode here can be NULL */
413 /* AV: can it, indeed? */
414 if (!inode)
415 return 1;
416 if (name->len != len)
417 return 1;
418 if (memcmp(name->name, str, len))
419 return 1;
420 head = rcu_dereference(PROC_I(inode)->sysctl);
421 return !head || !sysctl_is_seen(head);
422 }
423
424 static const struct dentry_operations proc_sys_dentry_operations = {
425 .d_revalidate = proc_sys_revalidate,
426 .d_delete = proc_sys_delete,
427 .d_compare = proc_sys_compare,
428 };
429
430 int __init proc_sys_init(void)
431 {
432 struct proc_dir_entry *proc_sys_root;
433
434 proc_sys_root = proc_mkdir("sys", NULL);
435 proc_sys_root->proc_iops = &proc_sys_dir_operations;
436 proc_sys_root->proc_fops = &proc_sys_dir_file_operations;
437 proc_sys_root->nlink = 0;
438 return 0;
439 }
This page took 0.039386 seconds and 5 git commands to generate.