Commit | Line | Data |
---|---|---|
77b14db5 EB |
1 | /* |
2 | * /proc/sys support | |
3 | */ | |
1e0edd3f | 4 | #include <linux/init.h> |
77b14db5 | 5 | #include <linux/sysctl.h> |
f1ecf068 | 6 | #include <linux/poll.h> |
77b14db5 EB |
7 | #include <linux/proc_fs.h> |
8 | #include <linux/security.h> | |
34286d66 | 9 | #include <linux/namei.h> |
1f87f0b5 | 10 | #include <linux/module.h> |
77b14db5 EB |
11 | #include "internal.h" |
12 | ||
d72f71eb | 13 | static const struct dentry_operations proc_sys_dentry_operations; |
77b14db5 | 14 | static const struct file_operations proc_sys_file_operations; |
03a44825 | 15 | static const struct inode_operations proc_sys_inode_operations; |
9043476f AV |
16 | static const struct file_operations proc_sys_dir_file_operations; |
17 | static const struct inode_operations proc_sys_dir_operations; | |
77b14db5 | 18 | |
f1ecf068 LDM |
19 | void proc_sys_poll_notify(struct ctl_table_poll *poll) |
20 | { | |
21 | if (!poll) | |
22 | return; | |
23 | ||
24 | atomic_inc(&poll->event); | |
25 | wake_up_interruptible(&poll->wait); | |
26 | } | |
27 | ||
1f87f0b5 EB |
28 | static struct ctl_table root_table[1]; |
29 | static struct ctl_table_root sysctl_table_root; | |
30 | static struct ctl_table_header root_table_header = { | |
31 | {{.count = 1, | |
32 | .ctl_table = root_table, | |
33 | .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}}, | |
34 | .root = &sysctl_table_root, | |
35 | .set = &sysctl_table_root.default_set, | |
36 | }; | |
37 | static struct ctl_table_root sysctl_table_root = { | |
38 | .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list), | |
39 | .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry), | |
40 | }; | |
41 | ||
42 | static DEFINE_SPINLOCK(sysctl_lock); | |
43 | ||
44 | /* called under sysctl_lock */ | |
45 | static int use_table(struct ctl_table_header *p) | |
46 | { | |
47 | if (unlikely(p->unregistering)) | |
48 | return 0; | |
49 | p->used++; | |
50 | return 1; | |
51 | } | |
52 | ||
53 | /* called under sysctl_lock */ | |
54 | static void unuse_table(struct ctl_table_header *p) | |
55 | { | |
56 | if (!--p->used) | |
57 | if (unlikely(p->unregistering)) | |
58 | complete(p->unregistering); | |
59 | } | |
60 | ||
61 | /* called under sysctl_lock, will reacquire if has to wait */ | |
62 | static void start_unregistering(struct ctl_table_header *p) | |
63 | { | |
64 | /* | |
65 | * if p->used is 0, nobody will ever touch that entry again; | |
66 | * we'll eliminate all paths to it before dropping sysctl_lock | |
67 | */ | |
68 | if (unlikely(p->used)) { | |
69 | struct completion wait; | |
70 | init_completion(&wait); | |
71 | p->unregistering = &wait; | |
72 | spin_unlock(&sysctl_lock); | |
73 | wait_for_completion(&wait); | |
74 | spin_lock(&sysctl_lock); | |
75 | } else { | |
76 | /* anything non-NULL; we'll never dereference it */ | |
77 | p->unregistering = ERR_PTR(-EINVAL); | |
78 | } | |
79 | /* | |
80 | * do not remove from the list until nobody holds it; walking the | |
81 | * list in do_sysctl() relies on that. | |
82 | */ | |
83 | list_del_init(&p->ctl_entry); | |
84 | } | |
85 | ||
86 | static void sysctl_head_get(struct ctl_table_header *head) | |
87 | { | |
88 | spin_lock(&sysctl_lock); | |
89 | head->count++; | |
90 | spin_unlock(&sysctl_lock); | |
91 | } | |
92 | ||
93 | void sysctl_head_put(struct ctl_table_header *head) | |
94 | { | |
95 | spin_lock(&sysctl_lock); | |
96 | if (!--head->count) | |
97 | kfree_rcu(head, rcu); | |
98 | spin_unlock(&sysctl_lock); | |
99 | } | |
100 | ||
101 | static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head) | |
102 | { | |
103 | if (!head) | |
104 | BUG(); | |
105 | spin_lock(&sysctl_lock); | |
106 | if (!use_table(head)) | |
107 | head = ERR_PTR(-ENOENT); | |
108 | spin_unlock(&sysctl_lock); | |
109 | return head; | |
110 | } | |
111 | ||
112 | static void sysctl_head_finish(struct ctl_table_header *head) | |
113 | { | |
114 | if (!head) | |
115 | return; | |
116 | spin_lock(&sysctl_lock); | |
117 | unuse_table(head); | |
118 | spin_unlock(&sysctl_lock); | |
119 | } | |
120 | ||
121 | static struct ctl_table_set * | |
122 | lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces) | |
123 | { | |
124 | struct ctl_table_set *set = &root->default_set; | |
125 | if (root->lookup) | |
126 | set = root->lookup(root, namespaces); | |
127 | return set; | |
128 | } | |
129 | ||
130 | static struct list_head * | |
131 | lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces) | |
132 | { | |
133 | struct ctl_table_set *set = lookup_header_set(root, namespaces); | |
134 | return &set->list; | |
135 | } | |
136 | ||
137 | static struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces, | |
138 | struct ctl_table_header *prev) | |
139 | { | |
140 | struct ctl_table_root *root; | |
141 | struct list_head *header_list; | |
142 | struct ctl_table_header *head; | |
143 | struct list_head *tmp; | |
144 | ||
145 | spin_lock(&sysctl_lock); | |
146 | if (prev) { | |
147 | head = prev; | |
148 | tmp = &prev->ctl_entry; | |
149 | unuse_table(prev); | |
150 | goto next; | |
151 | } | |
152 | tmp = &root_table_header.ctl_entry; | |
153 | for (;;) { | |
154 | head = list_entry(tmp, struct ctl_table_header, ctl_entry); | |
155 | ||
156 | if (!use_table(head)) | |
157 | goto next; | |
158 | spin_unlock(&sysctl_lock); | |
159 | return head; | |
160 | next: | |
161 | root = head->root; | |
162 | tmp = tmp->next; | |
163 | header_list = lookup_header_list(root, namespaces); | |
164 | if (tmp != header_list) | |
165 | continue; | |
166 | ||
167 | do { | |
168 | root = list_entry(root->root_list.next, | |
169 | struct ctl_table_root, root_list); | |
170 | if (root == &sysctl_table_root) | |
171 | goto out; | |
172 | header_list = lookup_header_list(root, namespaces); | |
173 | } while (list_empty(header_list)); | |
174 | tmp = header_list->next; | |
175 | } | |
176 | out: | |
177 | spin_unlock(&sysctl_lock); | |
178 | return NULL; | |
179 | } | |
180 | ||
181 | static struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev) | |
182 | { | |
183 | return __sysctl_head_next(current->nsproxy, prev); | |
184 | } | |
185 | ||
186 | void register_sysctl_root(struct ctl_table_root *root) | |
187 | { | |
188 | spin_lock(&sysctl_lock); | |
189 | list_add_tail(&root->root_list, &sysctl_table_root.root_list); | |
190 | spin_unlock(&sysctl_lock); | |
191 | } | |
192 | ||
193 | /* | |
194 | * sysctl_perm does NOT grant the superuser all rights automatically, because | |
195 | * some sysctl variables are readonly even to root. | |
196 | */ | |
197 | ||
198 | static int test_perm(int mode, int op) | |
199 | { | |
200 | if (!current_euid()) | |
201 | mode >>= 6; | |
202 | else if (in_egroup_p(0)) | |
203 | mode >>= 3; | |
204 | if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0) | |
205 | return 0; | |
206 | return -EACCES; | |
207 | } | |
208 | ||
209 | static int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op) | |
210 | { | |
211 | int mode; | |
212 | ||
213 | if (root->permissions) | |
214 | mode = root->permissions(root, current->nsproxy, table); | |
215 | else | |
216 | mode = table->mode; | |
217 | ||
218 | return test_perm(mode, op); | |
219 | } | |
220 | ||
9043476f AV |
221 | static struct inode *proc_sys_make_inode(struct super_block *sb, |
222 | struct ctl_table_header *head, struct ctl_table *table) | |
77b14db5 EB |
223 | { |
224 | struct inode *inode; | |
9043476f | 225 | struct proc_inode *ei; |
77b14db5 | 226 | |
9043476f | 227 | inode = new_inode(sb); |
77b14db5 EB |
228 | if (!inode) |
229 | goto out; | |
230 | ||
85fe4025 CH |
231 | inode->i_ino = get_next_ino(); |
232 | ||
9043476f | 233 | sysctl_head_get(head); |
77b14db5 | 234 | ei = PROC_I(inode); |
9043476f AV |
235 | ei->sysctl = head; |
236 | ei->sysctl_entry = table; | |
237 | ||
77b14db5 | 238 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; |
9043476f AV |
239 | inode->i_mode = table->mode; |
240 | if (!table->child) { | |
241 | inode->i_mode |= S_IFREG; | |
242 | inode->i_op = &proc_sys_inode_operations; | |
243 | inode->i_fop = &proc_sys_file_operations; | |
244 | } else { | |
245 | inode->i_mode |= S_IFDIR; | |
9043476f AV |
246 | inode->i_op = &proc_sys_dir_operations; |
247 | inode->i_fop = &proc_sys_dir_file_operations; | |
248 | } | |
77b14db5 EB |
249 | out: |
250 | return inode; | |
251 | } | |
252 | ||
9043476f | 253 | static struct ctl_table *find_in_table(struct ctl_table *p, struct qstr *name) |
77b14db5 | 254 | { |
2315ffa0 | 255 | for ( ; p->procname; p++) { |
36885d7b | 256 | if (strlen(p->procname) != name->len) |
77b14db5 EB |
257 | continue; |
258 | ||
36885d7b | 259 | if (memcmp(p->procname, name->name, name->len) != 0) |
77b14db5 EB |
260 | continue; |
261 | ||
262 | /* I have a match */ | |
9043476f | 263 | return p; |
77b14db5 EB |
264 | } |
265 | return NULL; | |
266 | } | |
267 | ||
81324364 | 268 | static struct ctl_table_header *grab_header(struct inode *inode) |
77b14db5 | 269 | { |
3cc3e046 EB |
270 | struct ctl_table_header *head = PROC_I(inode)->sysctl; |
271 | if (!head) | |
272 | head = &root_table_header; | |
273 | return sysctl_head_grab(head); | |
9043476f | 274 | } |
77b14db5 | 275 | |
9043476f AV |
276 | static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry, |
277 | struct nameidata *nd) | |
278 | { | |
279 | struct ctl_table_header *head = grab_header(dir); | |
280 | struct ctl_table *table = PROC_I(dir)->sysctl_entry; | |
281 | struct ctl_table_header *h = NULL; | |
282 | struct qstr *name = &dentry->d_name; | |
283 | struct ctl_table *p; | |
284 | struct inode *inode; | |
285 | struct dentry *err = ERR_PTR(-ENOENT); | |
77b14db5 | 286 | |
9043476f AV |
287 | if (IS_ERR(head)) |
288 | return ERR_CAST(head); | |
77b14db5 | 289 | |
9043476f AV |
290 | if (table && !table->child) { |
291 | WARN_ON(1); | |
292 | goto out; | |
77b14db5 | 293 | } |
77b14db5 | 294 | |
9043476f | 295 | table = table ? table->child : head->ctl_table; |
77b14db5 | 296 | |
9043476f AV |
297 | p = find_in_table(table, name); |
298 | if (!p) { | |
299 | for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) { | |
300 | if (h->attached_to != table) | |
301 | continue; | |
302 | p = find_in_table(h->attached_by, name); | |
303 | if (p) | |
304 | break; | |
305 | } | |
77b14db5 | 306 | } |
77b14db5 | 307 | |
9043476f | 308 | if (!p) |
77b14db5 EB |
309 | goto out; |
310 | ||
311 | err = ERR_PTR(-ENOMEM); | |
9043476f AV |
312 | inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p); |
313 | if (h) | |
314 | sysctl_head_finish(h); | |
315 | ||
77b14db5 EB |
316 | if (!inode) |
317 | goto out; | |
318 | ||
319 | err = NULL; | |
fb045adb | 320 | d_set_d_op(dentry, &proc_sys_dentry_operations); |
77b14db5 EB |
321 | d_add(dentry, inode); |
322 | ||
323 | out: | |
324 | sysctl_head_finish(head); | |
325 | return err; | |
326 | } | |
327 | ||
7708bfb1 PE |
328 | static ssize_t proc_sys_call_handler(struct file *filp, void __user *buf, |
329 | size_t count, loff_t *ppos, int write) | |
77b14db5 | 330 | { |
9043476f AV |
331 | struct inode *inode = filp->f_path.dentry->d_inode; |
332 | struct ctl_table_header *head = grab_header(inode); | |
333 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
2a2da53b DH |
334 | ssize_t error; |
335 | size_t res; | |
77b14db5 | 336 | |
9043476f AV |
337 | if (IS_ERR(head)) |
338 | return PTR_ERR(head); | |
77b14db5 EB |
339 | |
340 | /* | |
341 | * At this point we know that the sysctl was not unregistered | |
342 | * and won't be until we finish. | |
343 | */ | |
344 | error = -EPERM; | |
d7321cd6 | 345 | if (sysctl_perm(head->root, table, write ? MAY_WRITE : MAY_READ)) |
77b14db5 EB |
346 | goto out; |
347 | ||
9043476f AV |
348 | /* if that can happen at all, it should be -EINVAL, not -EISDIR */ |
349 | error = -EINVAL; | |
350 | if (!table->proc_handler) | |
351 | goto out; | |
352 | ||
77b14db5 EB |
353 | /* careful: calling conventions are nasty here */ |
354 | res = count; | |
8d65af78 | 355 | error = table->proc_handler(table, write, buf, &res, ppos); |
77b14db5 EB |
356 | if (!error) |
357 | error = res; | |
358 | out: | |
359 | sysctl_head_finish(head); | |
360 | ||
361 | return error; | |
362 | } | |
363 | ||
7708bfb1 | 364 | static ssize_t proc_sys_read(struct file *filp, char __user *buf, |
77b14db5 EB |
365 | size_t count, loff_t *ppos) |
366 | { | |
7708bfb1 PE |
367 | return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 0); |
368 | } | |
77b14db5 | 369 | |
7708bfb1 PE |
370 | static ssize_t proc_sys_write(struct file *filp, const char __user *buf, |
371 | size_t count, loff_t *ppos) | |
372 | { | |
373 | return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 1); | |
77b14db5 EB |
374 | } |
375 | ||
f1ecf068 LDM |
376 | static int proc_sys_open(struct inode *inode, struct file *filp) |
377 | { | |
378 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
379 | ||
380 | if (table->poll) | |
381 | filp->private_data = proc_sys_poll_event(table->poll); | |
382 | ||
383 | return 0; | |
384 | } | |
385 | ||
386 | static unsigned int proc_sys_poll(struct file *filp, poll_table *wait) | |
387 | { | |
388 | struct inode *inode = filp->f_path.dentry->d_inode; | |
389 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
390 | unsigned long event = (unsigned long)filp->private_data; | |
391 | unsigned int ret = DEFAULT_POLLMASK; | |
392 | ||
393 | if (!table->proc_handler) | |
394 | goto out; | |
395 | ||
396 | if (!table->poll) | |
397 | goto out; | |
398 | ||
399 | poll_wait(filp, &table->poll->wait, wait); | |
400 | ||
401 | if (event != atomic_read(&table->poll->event)) { | |
402 | filp->private_data = proc_sys_poll_event(table->poll); | |
403 | ret = POLLIN | POLLRDNORM | POLLERR | POLLPRI; | |
404 | } | |
405 | ||
406 | out: | |
407 | return ret; | |
408 | } | |
77b14db5 EB |
409 | |
410 | static int proc_sys_fill_cache(struct file *filp, void *dirent, | |
9043476f AV |
411 | filldir_t filldir, |
412 | struct ctl_table_header *head, | |
413 | struct ctl_table *table) | |
77b14db5 | 414 | { |
77b14db5 EB |
415 | struct dentry *child, *dir = filp->f_path.dentry; |
416 | struct inode *inode; | |
417 | struct qstr qname; | |
418 | ino_t ino = 0; | |
419 | unsigned type = DT_UNKNOWN; | |
77b14db5 EB |
420 | |
421 | qname.name = table->procname; | |
422 | qname.len = strlen(table->procname); | |
423 | qname.hash = full_name_hash(qname.name, qname.len); | |
424 | ||
77b14db5 EB |
425 | child = d_lookup(dir, &qname); |
426 | if (!child) { | |
9043476f AV |
427 | child = d_alloc(dir, &qname); |
428 | if (child) { | |
429 | inode = proc_sys_make_inode(dir->d_sb, head, table); | |
430 | if (!inode) { | |
431 | dput(child); | |
432 | return -ENOMEM; | |
433 | } else { | |
fb045adb | 434 | d_set_d_op(child, &proc_sys_dentry_operations); |
9043476f | 435 | d_add(child, inode); |
77b14db5 | 436 | } |
9043476f AV |
437 | } else { |
438 | return -ENOMEM; | |
77b14db5 EB |
439 | } |
440 | } | |
77b14db5 | 441 | inode = child->d_inode; |
9043476f AV |
442 | ino = inode->i_ino; |
443 | type = inode->i_mode >> 12; | |
77b14db5 | 444 | dput(child); |
9043476f AV |
445 | return !!filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type); |
446 | } | |
447 | ||
448 | static int scan(struct ctl_table_header *head, ctl_table *table, | |
449 | unsigned long *pos, struct file *file, | |
450 | void *dirent, filldir_t filldir) | |
451 | { | |
452 | ||
2315ffa0 | 453 | for (; table->procname; table++, (*pos)++) { |
9043476f AV |
454 | int res; |
455 | ||
9043476f AV |
456 | if (*pos < file->f_pos) |
457 | continue; | |
458 | ||
459 | res = proc_sys_fill_cache(file, dirent, filldir, head, table); | |
460 | if (res) | |
461 | return res; | |
462 | ||
463 | file->f_pos = *pos + 1; | |
464 | } | |
465 | return 0; | |
77b14db5 EB |
466 | } |
467 | ||
468 | static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir) | |
469 | { | |
9043476f | 470 | struct dentry *dentry = filp->f_path.dentry; |
77b14db5 | 471 | struct inode *inode = dentry->d_inode; |
9043476f AV |
472 | struct ctl_table_header *head = grab_header(inode); |
473 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
474 | struct ctl_table_header *h = NULL; | |
77b14db5 | 475 | unsigned long pos; |
9043476f AV |
476 | int ret = -EINVAL; |
477 | ||
478 | if (IS_ERR(head)) | |
479 | return PTR_ERR(head); | |
77b14db5 | 480 | |
9043476f AV |
481 | if (table && !table->child) { |
482 | WARN_ON(1); | |
77b14db5 | 483 | goto out; |
9043476f AV |
484 | } |
485 | ||
486 | table = table ? table->child : head->ctl_table; | |
77b14db5 EB |
487 | |
488 | ret = 0; | |
489 | /* Avoid a switch here: arm builds fail with missing __cmpdi2 */ | |
490 | if (filp->f_pos == 0) { | |
491 | if (filldir(dirent, ".", 1, filp->f_pos, | |
492 | inode->i_ino, DT_DIR) < 0) | |
493 | goto out; | |
494 | filp->f_pos++; | |
495 | } | |
496 | if (filp->f_pos == 1) { | |
497 | if (filldir(dirent, "..", 2, filp->f_pos, | |
498 | parent_ino(dentry), DT_DIR) < 0) | |
499 | goto out; | |
500 | filp->f_pos++; | |
501 | } | |
502 | pos = 2; | |
503 | ||
9043476f AV |
504 | ret = scan(head, table, &pos, filp, dirent, filldir); |
505 | if (ret) | |
506 | goto out; | |
77b14db5 | 507 | |
9043476f AV |
508 | for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) { |
509 | if (h->attached_to != table) | |
77b14db5 | 510 | continue; |
9043476f AV |
511 | ret = scan(h, h->attached_by, &pos, filp, dirent, filldir); |
512 | if (ret) { | |
513 | sysctl_head_finish(h); | |
514 | break; | |
77b14db5 EB |
515 | } |
516 | } | |
517 | ret = 1; | |
518 | out: | |
519 | sysctl_head_finish(head); | |
520 | return ret; | |
521 | } | |
522 | ||
10556cb2 | 523 | static int proc_sys_permission(struct inode *inode, int mask) |
77b14db5 EB |
524 | { |
525 | /* | |
526 | * sysctl entries that are not writeable, | |
527 | * are _NOT_ writeable, capabilities or not. | |
528 | */ | |
f696a365 MS |
529 | struct ctl_table_header *head; |
530 | struct ctl_table *table; | |
77b14db5 EB |
531 | int error; |
532 | ||
f696a365 MS |
533 | /* Executable files are not allowed under /proc/sys/ */ |
534 | if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode)) | |
535 | return -EACCES; | |
536 | ||
537 | head = grab_header(inode); | |
9043476f AV |
538 | if (IS_ERR(head)) |
539 | return PTR_ERR(head); | |
77b14db5 | 540 | |
f696a365 | 541 | table = PROC_I(inode)->sysctl_entry; |
9043476f AV |
542 | if (!table) /* global root - r-xr-xr-x */ |
543 | error = mask & MAY_WRITE ? -EACCES : 0; | |
544 | else /* Use the permissions on the sysctl table entry */ | |
1fc0f78c | 545 | error = sysctl_perm(head->root, table, mask & ~MAY_NOT_BLOCK); |
77b14db5 | 546 | |
77b14db5 EB |
547 | sysctl_head_finish(head); |
548 | return error; | |
549 | } | |
550 | ||
551 | static int proc_sys_setattr(struct dentry *dentry, struct iattr *attr) | |
552 | { | |
553 | struct inode *inode = dentry->d_inode; | |
554 | int error; | |
555 | ||
556 | if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID)) | |
557 | return -EPERM; | |
558 | ||
559 | error = inode_change_ok(inode, attr); | |
1025774c CH |
560 | if (error) |
561 | return error; | |
562 | ||
563 | if ((attr->ia_valid & ATTR_SIZE) && | |
564 | attr->ia_size != i_size_read(inode)) { | |
565 | error = vmtruncate(inode, attr->ia_size); | |
566 | if (error) | |
567 | return error; | |
568 | } | |
77b14db5 | 569 | |
1025774c CH |
570 | setattr_copy(inode, attr); |
571 | mark_inode_dirty(inode); | |
572 | return 0; | |
77b14db5 EB |
573 | } |
574 | ||
9043476f AV |
575 | static int proc_sys_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) |
576 | { | |
577 | struct inode *inode = dentry->d_inode; | |
578 | struct ctl_table_header *head = grab_header(inode); | |
579 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
580 | ||
581 | if (IS_ERR(head)) | |
582 | return PTR_ERR(head); | |
583 | ||
584 | generic_fillattr(inode, stat); | |
585 | if (table) | |
586 | stat->mode = (stat->mode & S_IFMT) | table->mode; | |
587 | ||
588 | sysctl_head_finish(head); | |
589 | return 0; | |
590 | } | |
591 | ||
77b14db5 | 592 | static const struct file_operations proc_sys_file_operations = { |
f1ecf068 LDM |
593 | .open = proc_sys_open, |
594 | .poll = proc_sys_poll, | |
77b14db5 EB |
595 | .read = proc_sys_read, |
596 | .write = proc_sys_write, | |
6038f373 | 597 | .llseek = default_llseek, |
9043476f AV |
598 | }; |
599 | ||
600 | static const struct file_operations proc_sys_dir_file_operations = { | |
887df078 | 601 | .read = generic_read_dir, |
77b14db5 | 602 | .readdir = proc_sys_readdir, |
3222a3e5 | 603 | .llseek = generic_file_llseek, |
77b14db5 EB |
604 | }; |
605 | ||
03a44825 | 606 | static const struct inode_operations proc_sys_inode_operations = { |
9043476f AV |
607 | .permission = proc_sys_permission, |
608 | .setattr = proc_sys_setattr, | |
609 | .getattr = proc_sys_getattr, | |
610 | }; | |
611 | ||
612 | static const struct inode_operations proc_sys_dir_operations = { | |
77b14db5 EB |
613 | .lookup = proc_sys_lookup, |
614 | .permission = proc_sys_permission, | |
615 | .setattr = proc_sys_setattr, | |
9043476f | 616 | .getattr = proc_sys_getattr, |
77b14db5 EB |
617 | }; |
618 | ||
619 | static int proc_sys_revalidate(struct dentry *dentry, struct nameidata *nd) | |
620 | { | |
34286d66 NP |
621 | if (nd->flags & LOOKUP_RCU) |
622 | return -ECHILD; | |
9043476f AV |
623 | return !PROC_I(dentry->d_inode)->sysctl->unregistering; |
624 | } | |
625 | ||
fe15ce44 | 626 | static int proc_sys_delete(const struct dentry *dentry) |
9043476f AV |
627 | { |
628 | return !!PROC_I(dentry->d_inode)->sysctl->unregistering; | |
629 | } | |
630 | ||
1f87f0b5 EB |
631 | static int sysctl_is_seen(struct ctl_table_header *p) |
632 | { | |
633 | struct ctl_table_set *set = p->set; | |
634 | int res; | |
635 | spin_lock(&sysctl_lock); | |
636 | if (p->unregistering) | |
637 | res = 0; | |
638 | else if (!set->is_seen) | |
639 | res = 1; | |
640 | else | |
641 | res = set->is_seen(set); | |
642 | spin_unlock(&sysctl_lock); | |
643 | return res; | |
644 | } | |
645 | ||
621e155a NP |
646 | static int proc_sys_compare(const struct dentry *parent, |
647 | const struct inode *pinode, | |
648 | const struct dentry *dentry, const struct inode *inode, | |
649 | unsigned int len, const char *str, const struct qstr *name) | |
9043476f | 650 | { |
dfef6dcd | 651 | struct ctl_table_header *head; |
31e6b01f NP |
652 | /* Although proc doesn't have negative dentries, rcu-walk means |
653 | * that inode here can be NULL */ | |
dfef6dcd | 654 | /* AV: can it, indeed? */ |
31e6b01f | 655 | if (!inode) |
dfef6dcd | 656 | return 1; |
621e155a | 657 | if (name->len != len) |
9043476f | 658 | return 1; |
621e155a | 659 | if (memcmp(name->name, str, len)) |
9043476f | 660 | return 1; |
dfef6dcd AV |
661 | head = rcu_dereference(PROC_I(inode)->sysctl); |
662 | return !head || !sysctl_is_seen(head); | |
77b14db5 EB |
663 | } |
664 | ||
d72f71eb | 665 | static const struct dentry_operations proc_sys_dentry_operations = { |
77b14db5 | 666 | .d_revalidate = proc_sys_revalidate, |
9043476f AV |
667 | .d_delete = proc_sys_delete, |
668 | .d_compare = proc_sys_compare, | |
77b14db5 EB |
669 | }; |
670 | ||
1f87f0b5 EB |
671 | static struct ctl_table *is_branch_in(struct ctl_table *branch, |
672 | struct ctl_table *table) | |
673 | { | |
674 | struct ctl_table *p; | |
675 | const char *s = branch->procname; | |
676 | ||
677 | /* branch should have named subdirectory as its first element */ | |
678 | if (!s || !branch->child) | |
679 | return NULL; | |
680 | ||
681 | /* ... and nothing else */ | |
682 | if (branch[1].procname) | |
683 | return NULL; | |
684 | ||
685 | /* table should contain subdirectory with the same name */ | |
686 | for (p = table; p->procname; p++) { | |
687 | if (!p->child) | |
688 | continue; | |
689 | if (p->procname && strcmp(p->procname, s) == 0) | |
690 | return p; | |
691 | } | |
692 | return NULL; | |
693 | } | |
694 | ||
695 | /* see if attaching q to p would be an improvement */ | |
696 | static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q) | |
697 | { | |
698 | struct ctl_table *to = p->ctl_table, *by = q->ctl_table; | |
699 | struct ctl_table *next; | |
700 | int is_better = 0; | |
701 | int not_in_parent = !p->attached_by; | |
702 | ||
703 | while ((next = is_branch_in(by, to)) != NULL) { | |
704 | if (by == q->attached_by) | |
705 | is_better = 1; | |
706 | if (to == p->attached_by) | |
707 | not_in_parent = 1; | |
708 | by = by->child; | |
709 | to = next->child; | |
710 | } | |
711 | ||
712 | if (is_better && not_in_parent) { | |
713 | q->attached_by = by; | |
714 | q->attached_to = to; | |
715 | q->parent = p; | |
716 | } | |
717 | } | |
718 | ||
7c60c48f EB |
719 | static int sysctl_check_table_dups(const char *path, struct ctl_table *old, |
720 | struct ctl_table *table) | |
1f87f0b5 | 721 | { |
7c60c48f EB |
722 | struct ctl_table *entry, *test; |
723 | int error = 0; | |
1f87f0b5 | 724 | |
7c60c48f EB |
725 | for (entry = old; entry->procname; entry++) { |
726 | for (test = table; test->procname; test++) { | |
727 | if (strcmp(entry->procname, test->procname) == 0) { | |
728 | printk(KERN_ERR "sysctl duplicate entry: %s/%s\n", | |
729 | path, test->procname); | |
730 | error = -EEXIST; | |
731 | } | |
732 | } | |
733 | } | |
734 | return error; | |
1f87f0b5 EB |
735 | } |
736 | ||
7c60c48f EB |
737 | static int sysctl_check_dups(struct nsproxy *namespaces, |
738 | struct ctl_table_header *header, | |
739 | const char *path, struct ctl_table *table) | |
1f87f0b5 | 740 | { |
7c60c48f EB |
741 | struct ctl_table_root *root; |
742 | struct ctl_table_set *set; | |
743 | struct ctl_table_header *dir_head, *head; | |
744 | struct ctl_table *dir_table; | |
745 | int error = 0; | |
1f87f0b5 | 746 | |
7c60c48f EB |
747 | /* No dups if we are the only member of our directory */ |
748 | if (header->attached_by != table) | |
749 | return 0; | |
1f87f0b5 | 750 | |
7c60c48f EB |
751 | dir_head = header->parent; |
752 | dir_table = header->attached_to; | |
1f87f0b5 | 753 | |
7c60c48f | 754 | error = sysctl_check_table_dups(path, dir_table, table); |
1f87f0b5 | 755 | |
7c60c48f EB |
756 | root = &sysctl_table_root; |
757 | do { | |
758 | set = lookup_header_set(root, namespaces); | |
1f87f0b5 | 759 | |
7c60c48f EB |
760 | list_for_each_entry(head, &set->list, ctl_entry) { |
761 | if (head->unregistering) | |
1f87f0b5 | 762 | continue; |
7c60c48f EB |
763 | if (head->attached_to != dir_table) |
764 | continue; | |
765 | error = sysctl_check_table_dups(path, head->attached_by, | |
766 | table); | |
1f87f0b5 | 767 | } |
7c60c48f EB |
768 | root = list_entry(root->root_list.next, |
769 | struct ctl_table_root, root_list); | |
770 | } while (root != &sysctl_table_root); | |
771 | return error; | |
1f87f0b5 EB |
772 | } |
773 | ||
7c60c48f | 774 | static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...) |
1f87f0b5 | 775 | { |
7c60c48f EB |
776 | struct va_format vaf; |
777 | va_list args; | |
1f87f0b5 | 778 | |
7c60c48f EB |
779 | va_start(args, fmt); |
780 | vaf.fmt = fmt; | |
781 | vaf.va = &args; | |
782 | ||
783 | printk(KERN_ERR "sysctl table check failed: %s/%s %pV\n", | |
784 | path, table->procname, &vaf); | |
1f87f0b5 | 785 | |
7c60c48f EB |
786 | va_end(args); |
787 | return -EINVAL; | |
1f87f0b5 EB |
788 | } |
789 | ||
7c60c48f | 790 | static int sysctl_check_table(const char *path, struct ctl_table *table) |
1f87f0b5 | 791 | { |
7c60c48f | 792 | int err = 0; |
1f87f0b5 | 793 | for (; table->procname; table++) { |
1f87f0b5 | 794 | if (table->child) |
7c60c48f EB |
795 | err = sysctl_err(path, table, "Not a file"); |
796 | ||
797 | if ((table->proc_handler == proc_dostring) || | |
798 | (table->proc_handler == proc_dointvec) || | |
799 | (table->proc_handler == proc_dointvec_minmax) || | |
800 | (table->proc_handler == proc_dointvec_jiffies) || | |
801 | (table->proc_handler == proc_dointvec_userhz_jiffies) || | |
802 | (table->proc_handler == proc_dointvec_ms_jiffies) || | |
803 | (table->proc_handler == proc_doulongvec_minmax) || | |
804 | (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) { | |
805 | if (!table->data) | |
806 | err = sysctl_err(path, table, "No data"); | |
807 | if (!table->maxlen) | |
808 | err = sysctl_err(path, table, "No maxlen"); | |
809 | } | |
810 | if (!table->proc_handler) | |
811 | err = sysctl_err(path, table, "No proc_handler"); | |
812 | ||
813 | if ((table->mode & (S_IRUGO|S_IWUGO)) != table->mode) | |
814 | err = sysctl_err(path, table, "bogus .mode 0%o", | |
815 | table->mode); | |
1f87f0b5 | 816 | } |
7c60c48f | 817 | return err; |
1f87f0b5 | 818 | } |
1f87f0b5 EB |
819 | |
820 | /** | |
f728019b | 821 | * __register_sysctl_table - register a leaf sysctl table |
1f87f0b5 EB |
822 | * @root: List of sysctl headers to register on |
823 | * @namespaces: Data to compute which lists of sysctl entries are visible | |
824 | * @path: The path to the directory the sysctl table is in. | |
825 | * @table: the top-level table structure | |
826 | * | |
827 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
828 | * array. A completely 0 filled entry terminates the table. | |
829 | * | |
830 | * The members of the &struct ctl_table structure are used as follows: | |
831 | * | |
832 | * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not | |
833 | * enter a sysctl file | |
834 | * | |
835 | * data - a pointer to data for use by proc_handler | |
836 | * | |
837 | * maxlen - the maximum size in bytes of the data | |
838 | * | |
f728019b | 839 | * mode - the file permissions for the /proc/sys file |
1f87f0b5 | 840 | * |
f728019b | 841 | * child - must be %NULL. |
1f87f0b5 EB |
842 | * |
843 | * proc_handler - the text handler routine (described below) | |
844 | * | |
1f87f0b5 EB |
845 | * extra1, extra2 - extra pointers usable by the proc handler routines |
846 | * | |
847 | * Leaf nodes in the sysctl tree will be represented by a single file | |
848 | * under /proc; non-leaf nodes will be represented by directories. | |
849 | * | |
f728019b EB |
850 | * There must be a proc_handler routine for any terminal nodes. |
851 | * Several default handlers are available to cover common cases - | |
1f87f0b5 EB |
852 | * |
853 | * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(), | |
854 | * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), | |
855 | * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax() | |
856 | * | |
857 | * It is the handler's job to read the input buffer from user memory | |
858 | * and process it. The handler should return 0 on success. | |
859 | * | |
860 | * This routine returns %NULL on a failure to register, and a pointer | |
861 | * to the table header on success. | |
862 | */ | |
6e9d5164 | 863 | struct ctl_table_header *__register_sysctl_table( |
1f87f0b5 EB |
864 | struct ctl_table_root *root, |
865 | struct nsproxy *namespaces, | |
6e9d5164 | 866 | const char *path, struct ctl_table *table) |
1f87f0b5 EB |
867 | { |
868 | struct ctl_table_header *header; | |
869 | struct ctl_table *new, **prevp; | |
6e9d5164 EB |
870 | const char *name, *nextname; |
871 | unsigned int npath = 0; | |
1f87f0b5 | 872 | struct ctl_table_set *set; |
f05e53a7 EB |
873 | size_t path_bytes = 0; |
874 | char *new_name; | |
1f87f0b5 EB |
875 | |
876 | /* Count the path components */ | |
6e9d5164 EB |
877 | for (name = path; name; name = nextname) { |
878 | int namelen; | |
879 | nextname = strchr(name, '/'); | |
880 | if (nextname) { | |
881 | namelen = nextname - name; | |
882 | nextname++; | |
883 | } else { | |
884 | namelen = strlen(name); | |
885 | } | |
886 | if (namelen == 0) | |
887 | continue; | |
888 | path_bytes += namelen + 1; | |
889 | npath++; | |
890 | } | |
1f87f0b5 EB |
891 | |
892 | /* | |
893 | * For each path component, allocate a 2-element ctl_table array. | |
894 | * The first array element will be filled with the sysctl entry | |
895 | * for this, the second will be the sentinel (procname == 0). | |
896 | * | |
897 | * We allocate everything in one go so that we don't have to | |
898 | * worry about freeing additional memory in unregister_sysctl_table. | |
899 | */ | |
f05e53a7 | 900 | header = kzalloc(sizeof(struct ctl_table_header) + path_bytes + |
1f87f0b5 EB |
901 | (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL); |
902 | if (!header) | |
903 | return NULL; | |
904 | ||
905 | new = (struct ctl_table *) (header + 1); | |
f05e53a7 | 906 | new_name = (char *)(new + (2 * npath)); |
1f87f0b5 EB |
907 | |
908 | /* Now connect the dots */ | |
909 | prevp = &header->ctl_table; | |
6e9d5164 EB |
910 | for (name = path; name; name = nextname) { |
911 | int namelen; | |
912 | nextname = strchr(name, '/'); | |
913 | if (nextname) { | |
914 | namelen = nextname - name; | |
915 | nextname++; | |
916 | } else { | |
917 | namelen = strlen(name); | |
918 | } | |
919 | if (namelen == 0) | |
920 | continue; | |
921 | memcpy(new_name, name, namelen); | |
922 | new_name[namelen] = '\0'; | |
923 | ||
f05e53a7 | 924 | new->procname = new_name; |
1f87f0b5 EB |
925 | new->mode = 0555; |
926 | ||
927 | *prevp = new; | |
928 | prevp = &new->child; | |
929 | ||
930 | new += 2; | |
6e9d5164 | 931 | new_name += namelen + 1; |
1f87f0b5 EB |
932 | } |
933 | *prevp = table; | |
934 | header->ctl_table_arg = table; | |
935 | ||
936 | INIT_LIST_HEAD(&header->ctl_entry); | |
937 | header->used = 0; | |
938 | header->unregistering = NULL; | |
939 | header->root = root; | |
1f87f0b5 | 940 | header->count = 1; |
7c60c48f EB |
941 | if (sysctl_check_table(path, table)) |
942 | goto fail; | |
8d6ecfcc | 943 | |
1f87f0b5 EB |
944 | spin_lock(&sysctl_lock); |
945 | header->set = lookup_header_set(root, namespaces); | |
946 | header->attached_by = header->ctl_table; | |
947 | header->attached_to = root_table; | |
948 | header->parent = &root_table_header; | |
bd295b56 EB |
949 | set = header->set; |
950 | root = header->root; | |
951 | for (;;) { | |
1f87f0b5 EB |
952 | struct ctl_table_header *p; |
953 | list_for_each_entry(p, &set->list, ctl_entry) { | |
954 | if (p->unregistering) | |
955 | continue; | |
956 | try_attach(p, header); | |
957 | } | |
bd295b56 EB |
958 | if (root == &sysctl_table_root) |
959 | break; | |
960 | root = list_entry(root->root_list.prev, | |
961 | struct ctl_table_root, root_list); | |
962 | set = lookup_header_set(root, namespaces); | |
1f87f0b5 | 963 | } |
7c60c48f EB |
964 | if (sysctl_check_dups(namespaces, header, path, table)) |
965 | goto fail_locked; | |
1f87f0b5 EB |
966 | header->parent->count++; |
967 | list_add_tail(&header->ctl_entry, &header->set->list); | |
968 | spin_unlock(&sysctl_lock); | |
969 | ||
970 | return header; | |
7c60c48f EB |
971 | fail_locked: |
972 | spin_unlock(&sysctl_lock); | |
973 | fail: | |
974 | kfree(header); | |
975 | dump_stack(); | |
976 | return NULL; | |
1f87f0b5 EB |
977 | } |
978 | ||
6e9d5164 EB |
979 | static char *append_path(const char *path, char *pos, const char *name) |
980 | { | |
981 | int namelen; | |
982 | namelen = strlen(name); | |
983 | if (((pos - path) + namelen + 2) >= PATH_MAX) | |
984 | return NULL; | |
985 | memcpy(pos, name, namelen); | |
986 | pos[namelen] = '/'; | |
987 | pos[namelen + 1] = '\0'; | |
988 | pos += namelen + 1; | |
989 | return pos; | |
990 | } | |
991 | ||
f728019b EB |
992 | static int count_subheaders(struct ctl_table *table) |
993 | { | |
994 | int has_files = 0; | |
995 | int nr_subheaders = 0; | |
996 | struct ctl_table *entry; | |
997 | ||
998 | /* special case: no directory and empty directory */ | |
999 | if (!table || !table->procname) | |
1000 | return 1; | |
1001 | ||
1002 | for (entry = table; entry->procname; entry++) { | |
1003 | if (entry->child) | |
1004 | nr_subheaders += count_subheaders(entry->child); | |
1005 | else | |
1006 | has_files = 1; | |
1007 | } | |
1008 | return nr_subheaders + has_files; | |
1009 | } | |
1010 | ||
1011 | static int register_leaf_sysctl_tables(const char *path, char *pos, | |
1012 | struct ctl_table_header ***subheader, | |
1013 | struct ctl_table_root *root, struct nsproxy *namespaces, | |
1014 | struct ctl_table *table) | |
1015 | { | |
1016 | struct ctl_table *ctl_table_arg = NULL; | |
1017 | struct ctl_table *entry, *files; | |
1018 | int nr_files = 0; | |
1019 | int nr_dirs = 0; | |
1020 | int err = -ENOMEM; | |
1021 | ||
1022 | for (entry = table; entry->procname; entry++) { | |
1023 | if (entry->child) | |
1024 | nr_dirs++; | |
1025 | else | |
1026 | nr_files++; | |
1027 | } | |
1028 | ||
1029 | files = table; | |
1030 | /* If there are mixed files and directories we need a new table */ | |
1031 | if (nr_dirs && nr_files) { | |
1032 | struct ctl_table *new; | |
1033 | files = kzalloc(sizeof(struct ctl_table) * (nr_files + 1), | |
1034 | GFP_KERNEL); | |
1035 | if (!files) | |
1036 | goto out; | |
1037 | ||
1038 | ctl_table_arg = files; | |
1039 | for (new = files, entry = table; entry->procname; entry++) { | |
1040 | if (entry->child) | |
1041 | continue; | |
1042 | *new = *entry; | |
1043 | new++; | |
1044 | } | |
1045 | } | |
1046 | ||
1047 | /* Register everything except a directory full of subdirectories */ | |
1048 | if (nr_files || !nr_dirs) { | |
1049 | struct ctl_table_header *header; | |
1050 | header = __register_sysctl_table(root, namespaces, path, files); | |
1051 | if (!header) { | |
1052 | kfree(ctl_table_arg); | |
1053 | goto out; | |
1054 | } | |
1055 | ||
1056 | /* Remember if we need to free the file table */ | |
1057 | header->ctl_table_arg = ctl_table_arg; | |
1058 | **subheader = header; | |
1059 | (*subheader)++; | |
1060 | } | |
1061 | ||
1062 | /* Recurse into the subdirectories. */ | |
1063 | for (entry = table; entry->procname; entry++) { | |
1064 | char *child_pos; | |
1065 | ||
1066 | if (!entry->child) | |
1067 | continue; | |
1068 | ||
1069 | err = -ENAMETOOLONG; | |
1070 | child_pos = append_path(path, pos, entry->procname); | |
1071 | if (!child_pos) | |
1072 | goto out; | |
1073 | ||
1074 | err = register_leaf_sysctl_tables(path, child_pos, subheader, | |
1075 | root, namespaces, entry->child); | |
1076 | pos[0] = '\0'; | |
1077 | if (err) | |
1078 | goto out; | |
1079 | } | |
1080 | err = 0; | |
1081 | out: | |
1082 | /* On failure our caller will unregister all registered subheaders */ | |
1083 | return err; | |
1084 | } | |
1085 | ||
6e9d5164 EB |
1086 | /** |
1087 | * __register_sysctl_paths - register a sysctl table hierarchy | |
1088 | * @root: List of sysctl headers to register on | |
1089 | * @namespaces: Data to compute which lists of sysctl entries are visible | |
1090 | * @path: The path to the directory the sysctl table is in. | |
1091 | * @table: the top-level table structure | |
1092 | * | |
1093 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1094 | * array. A completely 0 filled entry terminates the table. | |
1095 | * | |
1096 | * See __register_sysctl_table for more details. | |
1097 | */ | |
1098 | struct ctl_table_header *__register_sysctl_paths( | |
1099 | struct ctl_table_root *root, | |
1100 | struct nsproxy *namespaces, | |
1101 | const struct ctl_path *path, struct ctl_table *table) | |
1102 | { | |
ec6a5266 | 1103 | struct ctl_table *ctl_table_arg = table; |
f728019b EB |
1104 | int nr_subheaders = count_subheaders(table); |
1105 | struct ctl_table_header *header = NULL, **subheaders, **subheader; | |
6e9d5164 EB |
1106 | const struct ctl_path *component; |
1107 | char *new_path, *pos; | |
1108 | ||
1109 | pos = new_path = kmalloc(PATH_MAX, GFP_KERNEL); | |
1110 | if (!new_path) | |
1111 | return NULL; | |
1112 | ||
1113 | pos[0] = '\0'; | |
1114 | for (component = path; component->procname; component++) { | |
1115 | pos = append_path(new_path, pos, component->procname); | |
1116 | if (!pos) | |
1117 | goto out; | |
1118 | } | |
ec6a5266 EB |
1119 | while (table->procname && table->child && !table[1].procname) { |
1120 | pos = append_path(new_path, pos, table->procname); | |
1121 | if (!pos) | |
1122 | goto out; | |
1123 | table = table->child; | |
1124 | } | |
f728019b EB |
1125 | if (nr_subheaders == 1) { |
1126 | header = __register_sysctl_table(root, namespaces, new_path, table); | |
1127 | if (header) | |
1128 | header->ctl_table_arg = ctl_table_arg; | |
1129 | } else { | |
1130 | header = kzalloc(sizeof(*header) + | |
1131 | sizeof(*subheaders)*nr_subheaders, GFP_KERNEL); | |
1132 | if (!header) | |
1133 | goto out; | |
1134 | ||
1135 | subheaders = (struct ctl_table_header **) (header + 1); | |
1136 | subheader = subheaders; | |
ec6a5266 | 1137 | header->ctl_table_arg = ctl_table_arg; |
f728019b EB |
1138 | |
1139 | if (register_leaf_sysctl_tables(new_path, pos, &subheader, | |
1140 | root, namespaces, table)) | |
1141 | goto err_register_leaves; | |
1142 | } | |
1143 | ||
6e9d5164 EB |
1144 | out: |
1145 | kfree(new_path); | |
1146 | return header; | |
f728019b EB |
1147 | |
1148 | err_register_leaves: | |
1149 | while (subheader > subheaders) { | |
1150 | struct ctl_table_header *subh = *(--subheader); | |
1151 | struct ctl_table *table = subh->ctl_table_arg; | |
1152 | unregister_sysctl_table(subh); | |
1153 | kfree(table); | |
1154 | } | |
1155 | kfree(header); | |
1156 | header = NULL; | |
1157 | goto out; | |
6e9d5164 EB |
1158 | } |
1159 | ||
1f87f0b5 EB |
1160 | /** |
1161 | * register_sysctl_table_path - register a sysctl table hierarchy | |
1162 | * @path: The path to the directory the sysctl table is in. | |
1163 | * @table: the top-level table structure | |
1164 | * | |
1165 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1166 | * array. A completely 0 filled entry terminates the table. | |
1167 | * | |
1168 | * See __register_sysctl_paths for more details. | |
1169 | */ | |
1170 | struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path, | |
1171 | struct ctl_table *table) | |
1172 | { | |
1173 | return __register_sysctl_paths(&sysctl_table_root, current->nsproxy, | |
1174 | path, table); | |
1175 | } | |
1176 | EXPORT_SYMBOL(register_sysctl_paths); | |
1177 | ||
1178 | /** | |
1179 | * register_sysctl_table - register a sysctl table hierarchy | |
1180 | * @table: the top-level table structure | |
1181 | * | |
1182 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1183 | * array. A completely 0 filled entry terminates the table. | |
1184 | * | |
1185 | * See register_sysctl_paths for more details. | |
1186 | */ | |
1187 | struct ctl_table_header *register_sysctl_table(struct ctl_table *table) | |
1188 | { | |
1189 | static const struct ctl_path null_path[] = { {} }; | |
1190 | ||
1191 | return register_sysctl_paths(null_path, table); | |
1192 | } | |
1193 | EXPORT_SYMBOL(register_sysctl_table); | |
1194 | ||
1195 | /** | |
1196 | * unregister_sysctl_table - unregister a sysctl table hierarchy | |
1197 | * @header: the header returned from register_sysctl_table | |
1198 | * | |
1199 | * Unregisters the sysctl table and all children. proc entries may not | |
1200 | * actually be removed until they are no longer used by anyone. | |
1201 | */ | |
1202 | void unregister_sysctl_table(struct ctl_table_header * header) | |
1203 | { | |
f728019b | 1204 | int nr_subheaders; |
1f87f0b5 EB |
1205 | might_sleep(); |
1206 | ||
1207 | if (header == NULL) | |
1208 | return; | |
1209 | ||
f728019b EB |
1210 | nr_subheaders = count_subheaders(header->ctl_table_arg); |
1211 | if (unlikely(nr_subheaders > 1)) { | |
1212 | struct ctl_table_header **subheaders; | |
1213 | int i; | |
1214 | ||
1215 | subheaders = (struct ctl_table_header **)(header + 1); | |
1216 | for (i = nr_subheaders -1; i >= 0; i--) { | |
1217 | struct ctl_table_header *subh = subheaders[i]; | |
1218 | struct ctl_table *table = subh->ctl_table_arg; | |
1219 | unregister_sysctl_table(subh); | |
1220 | kfree(table); | |
1221 | } | |
1222 | kfree(header); | |
1223 | return; | |
1224 | } | |
1225 | ||
1f87f0b5 EB |
1226 | spin_lock(&sysctl_lock); |
1227 | start_unregistering(header); | |
1228 | if (!--header->parent->count) { | |
1229 | WARN_ON(1); | |
1230 | kfree_rcu(header->parent, rcu); | |
1231 | } | |
1232 | if (!--header->count) | |
1233 | kfree_rcu(header, rcu); | |
1234 | spin_unlock(&sysctl_lock); | |
1235 | } | |
1236 | EXPORT_SYMBOL(unregister_sysctl_table); | |
1237 | ||
1238 | void setup_sysctl_set(struct ctl_table_set *p, | |
1f87f0b5 EB |
1239 | int (*is_seen)(struct ctl_table_set *)) |
1240 | { | |
1241 | INIT_LIST_HEAD(&p->list); | |
1f87f0b5 EB |
1242 | p->is_seen = is_seen; |
1243 | } | |
1244 | ||
97324cd8 EB |
1245 | void retire_sysctl_set(struct ctl_table_set *set) |
1246 | { | |
1247 | WARN_ON(!list_empty(&set->list)); | |
1248 | } | |
1f87f0b5 | 1249 | |
1e0edd3f | 1250 | int __init proc_sys_init(void) |
77b14db5 | 1251 | { |
e1675231 AD |
1252 | struct proc_dir_entry *proc_sys_root; |
1253 | ||
77b14db5 | 1254 | proc_sys_root = proc_mkdir("sys", NULL); |
9043476f AV |
1255 | proc_sys_root->proc_iops = &proc_sys_dir_operations; |
1256 | proc_sys_root->proc_fops = &proc_sys_dir_file_operations; | |
77b14db5 | 1257 | proc_sys_root->nlink = 0; |
de4e83bd EB |
1258 | |
1259 | return sysctl_init(); | |
77b14db5 | 1260 | } |