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> | |
40401530 | 9 | #include <linux/sched.h> |
34286d66 | 10 | #include <linux/namei.h> |
40401530 | 11 | #include <linux/mm.h> |
1f87f0b5 | 12 | #include <linux/module.h> |
77b14db5 EB |
13 | #include "internal.h" |
14 | ||
d72f71eb | 15 | static const struct dentry_operations proc_sys_dentry_operations; |
77b14db5 | 16 | static const struct file_operations proc_sys_file_operations; |
03a44825 | 17 | static const struct inode_operations proc_sys_inode_operations; |
9043476f AV |
18 | static const struct file_operations proc_sys_dir_file_operations; |
19 | static const struct inode_operations proc_sys_dir_operations; | |
77b14db5 | 20 | |
f1ecf068 LDM |
21 | void proc_sys_poll_notify(struct ctl_table_poll *poll) |
22 | { | |
23 | if (!poll) | |
24 | return; | |
25 | ||
26 | atomic_inc(&poll->event); | |
27 | wake_up_interruptible(&poll->wait); | |
28 | } | |
29 | ||
a194558e EB |
30 | static struct ctl_table root_table[] = { |
31 | { | |
32 | .procname = "", | |
7ec66d06 | 33 | .mode = S_IFDIR|S_IRUGO|S_IXUGO, |
a194558e EB |
34 | }, |
35 | { } | |
36 | }; | |
0e47c99d | 37 | static struct ctl_table_root sysctl_table_root = { |
0e47c99d | 38 | .default_set.dir.header = { |
7ec66d06 EB |
39 | {{.count = 1, |
40 | .nreg = 1, | |
ac13ac6f | 41 | .ctl_table = root_table }}, |
0e47c99d | 42 | .ctl_table_arg = root_table, |
7ec66d06 EB |
43 | .root = &sysctl_table_root, |
44 | .set = &sysctl_table_root.default_set, | |
45 | }, | |
1f87f0b5 | 46 | }; |
1f87f0b5 EB |
47 | |
48 | static DEFINE_SPINLOCK(sysctl_lock); | |
49 | ||
7ec66d06 | 50 | static void drop_sysctl_table(struct ctl_table_header *header); |
0e47c99d EB |
51 | static int sysctl_follow_link(struct ctl_table_header **phead, |
52 | struct ctl_table **pentry, struct nsproxy *namespaces); | |
53 | static int insert_links(struct ctl_table_header *head); | |
54 | static void put_links(struct ctl_table_header *header); | |
7ec66d06 | 55 | |
6980128f EB |
56 | static void sysctl_print_dir(struct ctl_dir *dir) |
57 | { | |
58 | if (dir->header.parent) | |
59 | sysctl_print_dir(dir->header.parent); | |
60 | printk(KERN_CONT "%s/", dir->header.ctl_table[0].procname); | |
61 | } | |
62 | ||
076c3eed EB |
63 | static int namecmp(const char *name1, int len1, const char *name2, int len2) |
64 | { | |
65 | int minlen; | |
66 | int cmp; | |
67 | ||
68 | minlen = len1; | |
69 | if (minlen > len2) | |
70 | minlen = len2; | |
71 | ||
72 | cmp = memcmp(name1, name2, minlen); | |
73 | if (cmp == 0) | |
74 | cmp = len1 - len2; | |
75 | return cmp; | |
76 | } | |
77 | ||
60f126d9 | 78 | /* Called under sysctl_lock */ |
076c3eed | 79 | static struct ctl_table *find_entry(struct ctl_table_header **phead, |
0e47c99d | 80 | struct ctl_dir *dir, const char *name, int namelen) |
076c3eed EB |
81 | { |
82 | struct ctl_table_header *head; | |
83 | struct ctl_table *entry; | |
ac13ac6f | 84 | struct rb_node *node = dir->root.rb_node; |
076c3eed | 85 | |
ac13ac6f EB |
86 | while (node) |
87 | { | |
88 | struct ctl_node *ctl_node; | |
89 | const char *procname; | |
90 | int cmp; | |
91 | ||
92 | ctl_node = rb_entry(node, struct ctl_node, node); | |
93 | head = ctl_node->header; | |
94 | entry = &head->ctl_table[ctl_node - head->node]; | |
95 | procname = entry->procname; | |
96 | ||
97 | cmp = namecmp(name, namelen, procname, strlen(procname)); | |
98 | if (cmp < 0) | |
99 | node = node->rb_left; | |
100 | else if (cmp > 0) | |
101 | node = node->rb_right; | |
102 | else { | |
103 | *phead = head; | |
104 | return entry; | |
076c3eed EB |
105 | } |
106 | } | |
107 | return NULL; | |
108 | } | |
109 | ||
ac13ac6f EB |
110 | static int insert_entry(struct ctl_table_header *head, struct ctl_table *entry) |
111 | { | |
112 | struct rb_node *node = &head->node[entry - head->ctl_table].node; | |
113 | struct rb_node **p = &head->parent->root.rb_node; | |
114 | struct rb_node *parent = NULL; | |
115 | const char *name = entry->procname; | |
116 | int namelen = strlen(name); | |
117 | ||
118 | while (*p) { | |
119 | struct ctl_table_header *parent_head; | |
120 | struct ctl_table *parent_entry; | |
121 | struct ctl_node *parent_node; | |
122 | const char *parent_name; | |
123 | int cmp; | |
124 | ||
125 | parent = *p; | |
126 | parent_node = rb_entry(parent, struct ctl_node, node); | |
127 | parent_head = parent_node->header; | |
128 | parent_entry = &parent_head->ctl_table[parent_node - parent_head->node]; | |
129 | parent_name = parent_entry->procname; | |
130 | ||
131 | cmp = namecmp(name, namelen, parent_name, strlen(parent_name)); | |
132 | if (cmp < 0) | |
133 | p = &(*p)->rb_left; | |
134 | else if (cmp > 0) | |
135 | p = &(*p)->rb_right; | |
136 | else { | |
137 | printk(KERN_ERR "sysctl duplicate entry: "); | |
138 | sysctl_print_dir(head->parent); | |
139 | printk(KERN_CONT "/%s\n", entry->procname); | |
140 | return -EEXIST; | |
141 | } | |
142 | } | |
143 | ||
144 | rb_link_node(node, parent, p); | |
145 | return 0; | |
146 | } | |
147 | ||
148 | static void erase_entry(struct ctl_table_header *head, struct ctl_table *entry) | |
149 | { | |
150 | struct rb_node *node = &head->node[entry - head->ctl_table].node; | |
151 | ||
152 | rb_erase(node, &head->parent->root); | |
153 | } | |
154 | ||
e0d04529 EB |
155 | static void init_header(struct ctl_table_header *head, |
156 | struct ctl_table_root *root, struct ctl_table_set *set, | |
ac13ac6f | 157 | struct ctl_node *node, struct ctl_table *table) |
e0d04529 | 158 | { |
7ec66d06 | 159 | head->ctl_table = table; |
e0d04529 | 160 | head->ctl_table_arg = table; |
e0d04529 EB |
161 | head->used = 0; |
162 | head->count = 1; | |
163 | head->nreg = 1; | |
164 | head->unregistering = NULL; | |
165 | head->root = root; | |
166 | head->set = set; | |
167 | head->parent = NULL; | |
ac13ac6f EB |
168 | head->node = node; |
169 | if (node) { | |
170 | struct ctl_table *entry; | |
171 | for (entry = table; entry->procname; entry++, node++) { | |
172 | rb_init_node(&node->node); | |
173 | node->header = head; | |
174 | } | |
175 | } | |
e0d04529 EB |
176 | } |
177 | ||
8425d6aa EB |
178 | static void erase_header(struct ctl_table_header *head) |
179 | { | |
ac13ac6f EB |
180 | struct ctl_table *entry; |
181 | for (entry = head->ctl_table; entry->procname; entry++) | |
182 | erase_entry(head, entry); | |
8425d6aa EB |
183 | } |
184 | ||
0e47c99d | 185 | static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header) |
8425d6aa | 186 | { |
ac13ac6f | 187 | struct ctl_table *entry; |
0e47c99d EB |
188 | int err; |
189 | ||
190 | dir->header.nreg++; | |
7ec66d06 | 191 | header->parent = dir; |
0e47c99d EB |
192 | err = insert_links(header); |
193 | if (err) | |
194 | goto fail_links; | |
ac13ac6f EB |
195 | for (entry = header->ctl_table; entry->procname; entry++) { |
196 | err = insert_entry(header, entry); | |
197 | if (err) | |
198 | goto fail; | |
199 | } | |
0e47c99d | 200 | return 0; |
ac13ac6f EB |
201 | fail: |
202 | erase_header(header); | |
203 | put_links(header); | |
0e47c99d EB |
204 | fail_links: |
205 | header->parent = NULL; | |
206 | drop_sysctl_table(&dir->header); | |
207 | return err; | |
8425d6aa EB |
208 | } |
209 | ||
1f87f0b5 EB |
210 | /* called under sysctl_lock */ |
211 | static int use_table(struct ctl_table_header *p) | |
212 | { | |
213 | if (unlikely(p->unregistering)) | |
214 | return 0; | |
215 | p->used++; | |
216 | return 1; | |
217 | } | |
218 | ||
219 | /* called under sysctl_lock */ | |
220 | static void unuse_table(struct ctl_table_header *p) | |
221 | { | |
222 | if (!--p->used) | |
223 | if (unlikely(p->unregistering)) | |
224 | complete(p->unregistering); | |
225 | } | |
226 | ||
227 | /* called under sysctl_lock, will reacquire if has to wait */ | |
228 | static void start_unregistering(struct ctl_table_header *p) | |
229 | { | |
230 | /* | |
231 | * if p->used is 0, nobody will ever touch that entry again; | |
232 | * we'll eliminate all paths to it before dropping sysctl_lock | |
233 | */ | |
234 | if (unlikely(p->used)) { | |
235 | struct completion wait; | |
236 | init_completion(&wait); | |
237 | p->unregistering = &wait; | |
238 | spin_unlock(&sysctl_lock); | |
239 | wait_for_completion(&wait); | |
240 | spin_lock(&sysctl_lock); | |
241 | } else { | |
242 | /* anything non-NULL; we'll never dereference it */ | |
243 | p->unregistering = ERR_PTR(-EINVAL); | |
244 | } | |
245 | /* | |
246 | * do not remove from the list until nobody holds it; walking the | |
247 | * list in do_sysctl() relies on that. | |
248 | */ | |
8425d6aa | 249 | erase_header(p); |
1f87f0b5 EB |
250 | } |
251 | ||
252 | static void sysctl_head_get(struct ctl_table_header *head) | |
253 | { | |
254 | spin_lock(&sysctl_lock); | |
255 | head->count++; | |
256 | spin_unlock(&sysctl_lock); | |
257 | } | |
258 | ||
259 | void sysctl_head_put(struct ctl_table_header *head) | |
260 | { | |
261 | spin_lock(&sysctl_lock); | |
262 | if (!--head->count) | |
263 | kfree_rcu(head, rcu); | |
264 | spin_unlock(&sysctl_lock); | |
265 | } | |
266 | ||
267 | static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head) | |
268 | { | |
269 | if (!head) | |
270 | BUG(); | |
271 | spin_lock(&sysctl_lock); | |
272 | if (!use_table(head)) | |
273 | head = ERR_PTR(-ENOENT); | |
274 | spin_unlock(&sysctl_lock); | |
275 | return head; | |
276 | } | |
277 | ||
278 | static void sysctl_head_finish(struct ctl_table_header *head) | |
279 | { | |
280 | if (!head) | |
281 | return; | |
282 | spin_lock(&sysctl_lock); | |
283 | unuse_table(head); | |
284 | spin_unlock(&sysctl_lock); | |
285 | } | |
286 | ||
287 | static struct ctl_table_set * | |
288 | lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces) | |
289 | { | |
290 | struct ctl_table_set *set = &root->default_set; | |
291 | if (root->lookup) | |
292 | set = root->lookup(root, namespaces); | |
293 | return set; | |
294 | } | |
295 | ||
076c3eed | 296 | static struct ctl_table *lookup_entry(struct ctl_table_header **phead, |
7ec66d06 | 297 | struct ctl_dir *dir, |
076c3eed EB |
298 | const char *name, int namelen) |
299 | { | |
300 | struct ctl_table_header *head; | |
301 | struct ctl_table *entry; | |
076c3eed EB |
302 | |
303 | spin_lock(&sysctl_lock); | |
0e47c99d EB |
304 | entry = find_entry(&head, dir, name, namelen); |
305 | if (entry && use_table(head)) | |
306 | *phead = head; | |
307 | else | |
308 | entry = NULL; | |
076c3eed EB |
309 | spin_unlock(&sysctl_lock); |
310 | return entry; | |
311 | } | |
312 | ||
ac13ac6f | 313 | static struct ctl_node *first_usable_entry(struct rb_node *node) |
1f87f0b5 | 314 | { |
ac13ac6f | 315 | struct ctl_node *ctl_node; |
1f87f0b5 | 316 | |
ac13ac6f EB |
317 | for (;node; node = rb_next(node)) { |
318 | ctl_node = rb_entry(node, struct ctl_node, node); | |
319 | if (use_table(ctl_node->header)) | |
320 | return ctl_node; | |
1f87f0b5 | 321 | } |
1f87f0b5 EB |
322 | return NULL; |
323 | } | |
324 | ||
7ec66d06 | 325 | static void first_entry(struct ctl_dir *dir, |
6a75ce16 EB |
326 | struct ctl_table_header **phead, struct ctl_table **pentry) |
327 | { | |
ac13ac6f | 328 | struct ctl_table_header *head = NULL; |
7ec66d06 | 329 | struct ctl_table *entry = NULL; |
ac13ac6f | 330 | struct ctl_node *ctl_node; |
6a75ce16 EB |
331 | |
332 | spin_lock(&sysctl_lock); | |
ac13ac6f | 333 | ctl_node = first_usable_entry(rb_first(&dir->root)); |
6a75ce16 | 334 | spin_unlock(&sysctl_lock); |
ac13ac6f EB |
335 | if (ctl_node) { |
336 | head = ctl_node->header; | |
337 | entry = &head->ctl_table[ctl_node - head->node]; | |
338 | } | |
6a75ce16 EB |
339 | *phead = head; |
340 | *pentry = entry; | |
341 | } | |
342 | ||
7ec66d06 | 343 | static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry) |
1f87f0b5 | 344 | { |
6a75ce16 EB |
345 | struct ctl_table_header *head = *phead; |
346 | struct ctl_table *entry = *pentry; | |
ac13ac6f | 347 | struct ctl_node *ctl_node = &head->node[entry - head->ctl_table]; |
6a75ce16 | 348 | |
ac13ac6f EB |
349 | spin_lock(&sysctl_lock); |
350 | unuse_table(head); | |
351 | ||
352 | ctl_node = first_usable_entry(rb_next(&ctl_node->node)); | |
353 | spin_unlock(&sysctl_lock); | |
354 | head = NULL; | |
355 | if (ctl_node) { | |
356 | head = ctl_node->header; | |
357 | entry = &head->ctl_table[ctl_node - head->node]; | |
6a75ce16 EB |
358 | } |
359 | *phead = head; | |
360 | *pentry = entry; | |
1f87f0b5 EB |
361 | } |
362 | ||
363 | void register_sysctl_root(struct ctl_table_root *root) | |
364 | { | |
1f87f0b5 EB |
365 | } |
366 | ||
367 | /* | |
368 | * sysctl_perm does NOT grant the superuser all rights automatically, because | |
369 | * some sysctl variables are readonly even to root. | |
370 | */ | |
371 | ||
372 | static int test_perm(int mode, int op) | |
373 | { | |
091bd3ea | 374 | if (uid_eq(current_euid(), GLOBAL_ROOT_UID)) |
1f87f0b5 | 375 | mode >>= 6; |
091bd3ea | 376 | else if (in_egroup_p(GLOBAL_ROOT_GID)) |
1f87f0b5 EB |
377 | mode >>= 3; |
378 | if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0) | |
379 | return 0; | |
380 | return -EACCES; | |
381 | } | |
382 | ||
383 | static int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op) | |
384 | { | |
385 | int mode; | |
386 | ||
387 | if (root->permissions) | |
388 | mode = root->permissions(root, current->nsproxy, table); | |
389 | else | |
390 | mode = table->mode; | |
391 | ||
392 | return test_perm(mode, op); | |
393 | } | |
394 | ||
9043476f AV |
395 | static struct inode *proc_sys_make_inode(struct super_block *sb, |
396 | struct ctl_table_header *head, struct ctl_table *table) | |
77b14db5 EB |
397 | { |
398 | struct inode *inode; | |
9043476f | 399 | struct proc_inode *ei; |
77b14db5 | 400 | |
9043476f | 401 | inode = new_inode(sb); |
77b14db5 EB |
402 | if (!inode) |
403 | goto out; | |
404 | ||
85fe4025 CH |
405 | inode->i_ino = get_next_ino(); |
406 | ||
9043476f | 407 | sysctl_head_get(head); |
77b14db5 | 408 | ei = PROC_I(inode); |
9043476f AV |
409 | ei->sysctl = head; |
410 | ei->sysctl_entry = table; | |
411 | ||
77b14db5 | 412 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; |
9043476f | 413 | inode->i_mode = table->mode; |
7ec66d06 | 414 | if (!S_ISDIR(table->mode)) { |
9043476f AV |
415 | inode->i_mode |= S_IFREG; |
416 | inode->i_op = &proc_sys_inode_operations; | |
417 | inode->i_fop = &proc_sys_file_operations; | |
418 | } else { | |
419 | inode->i_mode |= S_IFDIR; | |
9043476f AV |
420 | inode->i_op = &proc_sys_dir_operations; |
421 | inode->i_fop = &proc_sys_dir_file_operations; | |
422 | } | |
77b14db5 EB |
423 | out: |
424 | return inode; | |
425 | } | |
426 | ||
81324364 | 427 | static struct ctl_table_header *grab_header(struct inode *inode) |
77b14db5 | 428 | { |
3cc3e046 EB |
429 | struct ctl_table_header *head = PROC_I(inode)->sysctl; |
430 | if (!head) | |
0e47c99d | 431 | head = &sysctl_table_root.default_set.dir.header; |
3cc3e046 | 432 | return sysctl_head_grab(head); |
9043476f | 433 | } |
77b14db5 | 434 | |
9043476f AV |
435 | static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry, |
436 | struct nameidata *nd) | |
437 | { | |
438 | struct ctl_table_header *head = grab_header(dir); | |
9043476f AV |
439 | struct ctl_table_header *h = NULL; |
440 | struct qstr *name = &dentry->d_name; | |
441 | struct ctl_table *p; | |
442 | struct inode *inode; | |
443 | struct dentry *err = ERR_PTR(-ENOENT); | |
7ec66d06 | 444 | struct ctl_dir *ctl_dir; |
0e47c99d | 445 | int ret; |
77b14db5 | 446 | |
9043476f AV |
447 | if (IS_ERR(head)) |
448 | return ERR_CAST(head); | |
77b14db5 | 449 | |
7ec66d06 | 450 | ctl_dir = container_of(head, struct ctl_dir, header); |
77b14db5 | 451 | |
7ec66d06 | 452 | p = lookup_entry(&h, ctl_dir, name->name, name->len); |
9043476f | 453 | if (!p) |
77b14db5 EB |
454 | goto out; |
455 | ||
4e757320 EB |
456 | if (S_ISLNK(p->mode)) { |
457 | ret = sysctl_follow_link(&h, &p, current->nsproxy); | |
458 | err = ERR_PTR(ret); | |
459 | if (ret) | |
460 | goto out; | |
461 | } | |
0e47c99d | 462 | |
77b14db5 | 463 | err = ERR_PTR(-ENOMEM); |
9043476f AV |
464 | inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p); |
465 | if (h) | |
466 | sysctl_head_finish(h); | |
467 | ||
77b14db5 EB |
468 | if (!inode) |
469 | goto out; | |
470 | ||
471 | err = NULL; | |
fb045adb | 472 | d_set_d_op(dentry, &proc_sys_dentry_operations); |
77b14db5 EB |
473 | d_add(dentry, inode); |
474 | ||
475 | out: | |
476 | sysctl_head_finish(head); | |
477 | return err; | |
478 | } | |
479 | ||
7708bfb1 PE |
480 | static ssize_t proc_sys_call_handler(struct file *filp, void __user *buf, |
481 | size_t count, loff_t *ppos, int write) | |
77b14db5 | 482 | { |
9043476f AV |
483 | struct inode *inode = filp->f_path.dentry->d_inode; |
484 | struct ctl_table_header *head = grab_header(inode); | |
485 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
2a2da53b DH |
486 | ssize_t error; |
487 | size_t res; | |
77b14db5 | 488 | |
9043476f AV |
489 | if (IS_ERR(head)) |
490 | return PTR_ERR(head); | |
77b14db5 EB |
491 | |
492 | /* | |
493 | * At this point we know that the sysctl was not unregistered | |
494 | * and won't be until we finish. | |
495 | */ | |
496 | error = -EPERM; | |
d7321cd6 | 497 | if (sysctl_perm(head->root, table, write ? MAY_WRITE : MAY_READ)) |
77b14db5 EB |
498 | goto out; |
499 | ||
9043476f AV |
500 | /* if that can happen at all, it should be -EINVAL, not -EISDIR */ |
501 | error = -EINVAL; | |
502 | if (!table->proc_handler) | |
503 | goto out; | |
504 | ||
77b14db5 EB |
505 | /* careful: calling conventions are nasty here */ |
506 | res = count; | |
8d65af78 | 507 | error = table->proc_handler(table, write, buf, &res, ppos); |
77b14db5 EB |
508 | if (!error) |
509 | error = res; | |
510 | out: | |
511 | sysctl_head_finish(head); | |
512 | ||
513 | return error; | |
514 | } | |
515 | ||
7708bfb1 | 516 | static ssize_t proc_sys_read(struct file *filp, char __user *buf, |
77b14db5 EB |
517 | size_t count, loff_t *ppos) |
518 | { | |
7708bfb1 PE |
519 | return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 0); |
520 | } | |
77b14db5 | 521 | |
7708bfb1 PE |
522 | static ssize_t proc_sys_write(struct file *filp, const char __user *buf, |
523 | size_t count, loff_t *ppos) | |
524 | { | |
525 | return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 1); | |
77b14db5 EB |
526 | } |
527 | ||
f1ecf068 LDM |
528 | static int proc_sys_open(struct inode *inode, struct file *filp) |
529 | { | |
4e474a00 | 530 | struct ctl_table_header *head = grab_header(inode); |
f1ecf068 LDM |
531 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; |
532 | ||
4e474a00 LDM |
533 | /* sysctl was unregistered */ |
534 | if (IS_ERR(head)) | |
535 | return PTR_ERR(head); | |
536 | ||
f1ecf068 LDM |
537 | if (table->poll) |
538 | filp->private_data = proc_sys_poll_event(table->poll); | |
539 | ||
4e474a00 LDM |
540 | sysctl_head_finish(head); |
541 | ||
f1ecf068 LDM |
542 | return 0; |
543 | } | |
544 | ||
545 | static unsigned int proc_sys_poll(struct file *filp, poll_table *wait) | |
546 | { | |
547 | struct inode *inode = filp->f_path.dentry->d_inode; | |
4e474a00 | 548 | struct ctl_table_header *head = grab_header(inode); |
f1ecf068 | 549 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; |
f1ecf068 | 550 | unsigned int ret = DEFAULT_POLLMASK; |
4e474a00 LDM |
551 | unsigned long event; |
552 | ||
553 | /* sysctl was unregistered */ | |
554 | if (IS_ERR(head)) | |
555 | return POLLERR | POLLHUP; | |
f1ecf068 LDM |
556 | |
557 | if (!table->proc_handler) | |
558 | goto out; | |
559 | ||
560 | if (!table->poll) | |
561 | goto out; | |
562 | ||
4e474a00 | 563 | event = (unsigned long)filp->private_data; |
f1ecf068 LDM |
564 | poll_wait(filp, &table->poll->wait, wait); |
565 | ||
566 | if (event != atomic_read(&table->poll->event)) { | |
567 | filp->private_data = proc_sys_poll_event(table->poll); | |
568 | ret = POLLIN | POLLRDNORM | POLLERR | POLLPRI; | |
569 | } | |
570 | ||
571 | out: | |
4e474a00 LDM |
572 | sysctl_head_finish(head); |
573 | ||
f1ecf068 LDM |
574 | return ret; |
575 | } | |
77b14db5 EB |
576 | |
577 | static int proc_sys_fill_cache(struct file *filp, void *dirent, | |
9043476f AV |
578 | filldir_t filldir, |
579 | struct ctl_table_header *head, | |
580 | struct ctl_table *table) | |
77b14db5 | 581 | { |
77b14db5 EB |
582 | struct dentry *child, *dir = filp->f_path.dentry; |
583 | struct inode *inode; | |
584 | struct qstr qname; | |
585 | ino_t ino = 0; | |
586 | unsigned type = DT_UNKNOWN; | |
77b14db5 EB |
587 | |
588 | qname.name = table->procname; | |
589 | qname.len = strlen(table->procname); | |
590 | qname.hash = full_name_hash(qname.name, qname.len); | |
591 | ||
77b14db5 EB |
592 | child = d_lookup(dir, &qname); |
593 | if (!child) { | |
9043476f AV |
594 | child = d_alloc(dir, &qname); |
595 | if (child) { | |
596 | inode = proc_sys_make_inode(dir->d_sb, head, table); | |
597 | if (!inode) { | |
598 | dput(child); | |
599 | return -ENOMEM; | |
600 | } else { | |
fb045adb | 601 | d_set_d_op(child, &proc_sys_dentry_operations); |
9043476f | 602 | d_add(child, inode); |
77b14db5 | 603 | } |
9043476f AV |
604 | } else { |
605 | return -ENOMEM; | |
77b14db5 EB |
606 | } |
607 | } | |
77b14db5 | 608 | inode = child->d_inode; |
9043476f AV |
609 | ino = inode->i_ino; |
610 | type = inode->i_mode >> 12; | |
77b14db5 | 611 | dput(child); |
9043476f AV |
612 | return !!filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type); |
613 | } | |
614 | ||
0e47c99d EB |
615 | static int proc_sys_link_fill_cache(struct file *filp, void *dirent, |
616 | filldir_t filldir, | |
617 | struct ctl_table_header *head, | |
618 | struct ctl_table *table) | |
619 | { | |
620 | int err, ret = 0; | |
621 | head = sysctl_head_grab(head); | |
622 | ||
4e757320 EB |
623 | if (S_ISLNK(table->mode)) { |
624 | /* It is not an error if we can not follow the link ignore it */ | |
625 | err = sysctl_follow_link(&head, &table, current->nsproxy); | |
626 | if (err) | |
627 | goto out; | |
628 | } | |
0e47c99d EB |
629 | |
630 | ret = proc_sys_fill_cache(filp, dirent, filldir, head, table); | |
631 | out: | |
632 | sysctl_head_finish(head); | |
633 | return ret; | |
634 | } | |
635 | ||
9043476f AV |
636 | static int scan(struct ctl_table_header *head, ctl_table *table, |
637 | unsigned long *pos, struct file *file, | |
638 | void *dirent, filldir_t filldir) | |
639 | { | |
6a75ce16 | 640 | int res; |
9043476f | 641 | |
6a75ce16 EB |
642 | if ((*pos)++ < file->f_pos) |
643 | return 0; | |
9043476f | 644 | |
0e47c99d EB |
645 | if (unlikely(S_ISLNK(table->mode))) |
646 | res = proc_sys_link_fill_cache(file, dirent, filldir, head, table); | |
647 | else | |
9043476f | 648 | res = proc_sys_fill_cache(file, dirent, filldir, head, table); |
9043476f | 649 | |
6a75ce16 EB |
650 | if (res == 0) |
651 | file->f_pos = *pos; | |
9043476f | 652 | |
6a75ce16 | 653 | return res; |
77b14db5 EB |
654 | } |
655 | ||
656 | static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir) | |
657 | { | |
9043476f | 658 | struct dentry *dentry = filp->f_path.dentry; |
77b14db5 | 659 | struct inode *inode = dentry->d_inode; |
9043476f | 660 | struct ctl_table_header *head = grab_header(inode); |
9043476f | 661 | struct ctl_table_header *h = NULL; |
6a75ce16 | 662 | struct ctl_table *entry; |
7ec66d06 | 663 | struct ctl_dir *ctl_dir; |
77b14db5 | 664 | unsigned long pos; |
9043476f AV |
665 | int ret = -EINVAL; |
666 | ||
667 | if (IS_ERR(head)) | |
668 | return PTR_ERR(head); | |
77b14db5 | 669 | |
7ec66d06 | 670 | ctl_dir = container_of(head, struct ctl_dir, header); |
77b14db5 EB |
671 | |
672 | ret = 0; | |
673 | /* Avoid a switch here: arm builds fail with missing __cmpdi2 */ | |
674 | if (filp->f_pos == 0) { | |
675 | if (filldir(dirent, ".", 1, filp->f_pos, | |
676 | inode->i_ino, DT_DIR) < 0) | |
677 | goto out; | |
678 | filp->f_pos++; | |
679 | } | |
680 | if (filp->f_pos == 1) { | |
681 | if (filldir(dirent, "..", 2, filp->f_pos, | |
682 | parent_ino(dentry), DT_DIR) < 0) | |
683 | goto out; | |
684 | filp->f_pos++; | |
685 | } | |
686 | pos = 2; | |
687 | ||
7ec66d06 | 688 | for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) { |
6a75ce16 | 689 | ret = scan(h, entry, &pos, filp, dirent, filldir); |
9043476f AV |
690 | if (ret) { |
691 | sysctl_head_finish(h); | |
692 | break; | |
77b14db5 EB |
693 | } |
694 | } | |
695 | ret = 1; | |
696 | out: | |
697 | sysctl_head_finish(head); | |
698 | return ret; | |
699 | } | |
700 | ||
10556cb2 | 701 | static int proc_sys_permission(struct inode *inode, int mask) |
77b14db5 EB |
702 | { |
703 | /* | |
704 | * sysctl entries that are not writeable, | |
705 | * are _NOT_ writeable, capabilities or not. | |
706 | */ | |
f696a365 MS |
707 | struct ctl_table_header *head; |
708 | struct ctl_table *table; | |
77b14db5 EB |
709 | int error; |
710 | ||
f696a365 MS |
711 | /* Executable files are not allowed under /proc/sys/ */ |
712 | if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode)) | |
713 | return -EACCES; | |
714 | ||
715 | head = grab_header(inode); | |
9043476f AV |
716 | if (IS_ERR(head)) |
717 | return PTR_ERR(head); | |
77b14db5 | 718 | |
f696a365 | 719 | table = PROC_I(inode)->sysctl_entry; |
9043476f AV |
720 | if (!table) /* global root - r-xr-xr-x */ |
721 | error = mask & MAY_WRITE ? -EACCES : 0; | |
722 | else /* Use the permissions on the sysctl table entry */ | |
1fc0f78c | 723 | error = sysctl_perm(head->root, table, mask & ~MAY_NOT_BLOCK); |
77b14db5 | 724 | |
77b14db5 EB |
725 | sysctl_head_finish(head); |
726 | return error; | |
727 | } | |
728 | ||
729 | static int proc_sys_setattr(struct dentry *dentry, struct iattr *attr) | |
730 | { | |
731 | struct inode *inode = dentry->d_inode; | |
732 | int error; | |
733 | ||
734 | if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID)) | |
735 | return -EPERM; | |
736 | ||
737 | error = inode_change_ok(inode, attr); | |
1025774c CH |
738 | if (error) |
739 | return error; | |
740 | ||
741 | if ((attr->ia_valid & ATTR_SIZE) && | |
742 | attr->ia_size != i_size_read(inode)) { | |
743 | error = vmtruncate(inode, attr->ia_size); | |
744 | if (error) | |
745 | return error; | |
746 | } | |
77b14db5 | 747 | |
1025774c CH |
748 | setattr_copy(inode, attr); |
749 | mark_inode_dirty(inode); | |
750 | return 0; | |
77b14db5 EB |
751 | } |
752 | ||
9043476f AV |
753 | static int proc_sys_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) |
754 | { | |
755 | struct inode *inode = dentry->d_inode; | |
756 | struct ctl_table_header *head = grab_header(inode); | |
757 | struct ctl_table *table = PROC_I(inode)->sysctl_entry; | |
758 | ||
759 | if (IS_ERR(head)) | |
760 | return PTR_ERR(head); | |
761 | ||
762 | generic_fillattr(inode, stat); | |
763 | if (table) | |
764 | stat->mode = (stat->mode & S_IFMT) | table->mode; | |
765 | ||
766 | sysctl_head_finish(head); | |
767 | return 0; | |
768 | } | |
769 | ||
77b14db5 | 770 | static const struct file_operations proc_sys_file_operations = { |
f1ecf068 LDM |
771 | .open = proc_sys_open, |
772 | .poll = proc_sys_poll, | |
77b14db5 EB |
773 | .read = proc_sys_read, |
774 | .write = proc_sys_write, | |
6038f373 | 775 | .llseek = default_llseek, |
9043476f AV |
776 | }; |
777 | ||
778 | static const struct file_operations proc_sys_dir_file_operations = { | |
887df078 | 779 | .read = generic_read_dir, |
77b14db5 | 780 | .readdir = proc_sys_readdir, |
3222a3e5 | 781 | .llseek = generic_file_llseek, |
77b14db5 EB |
782 | }; |
783 | ||
03a44825 | 784 | static const struct inode_operations proc_sys_inode_operations = { |
9043476f AV |
785 | .permission = proc_sys_permission, |
786 | .setattr = proc_sys_setattr, | |
787 | .getattr = proc_sys_getattr, | |
788 | }; | |
789 | ||
790 | static const struct inode_operations proc_sys_dir_operations = { | |
77b14db5 EB |
791 | .lookup = proc_sys_lookup, |
792 | .permission = proc_sys_permission, | |
793 | .setattr = proc_sys_setattr, | |
9043476f | 794 | .getattr = proc_sys_getattr, |
77b14db5 EB |
795 | }; |
796 | ||
797 | static int proc_sys_revalidate(struct dentry *dentry, struct nameidata *nd) | |
798 | { | |
34286d66 NP |
799 | if (nd->flags & LOOKUP_RCU) |
800 | return -ECHILD; | |
9043476f AV |
801 | return !PROC_I(dentry->d_inode)->sysctl->unregistering; |
802 | } | |
803 | ||
fe15ce44 | 804 | static int proc_sys_delete(const struct dentry *dentry) |
9043476f AV |
805 | { |
806 | return !!PROC_I(dentry->d_inode)->sysctl->unregistering; | |
807 | } | |
808 | ||
1f87f0b5 EB |
809 | static int sysctl_is_seen(struct ctl_table_header *p) |
810 | { | |
811 | struct ctl_table_set *set = p->set; | |
812 | int res; | |
813 | spin_lock(&sysctl_lock); | |
814 | if (p->unregistering) | |
815 | res = 0; | |
816 | else if (!set->is_seen) | |
817 | res = 1; | |
818 | else | |
819 | res = set->is_seen(set); | |
820 | spin_unlock(&sysctl_lock); | |
821 | return res; | |
822 | } | |
823 | ||
621e155a NP |
824 | static int proc_sys_compare(const struct dentry *parent, |
825 | const struct inode *pinode, | |
826 | const struct dentry *dentry, const struct inode *inode, | |
827 | unsigned int len, const char *str, const struct qstr *name) | |
9043476f | 828 | { |
dfef6dcd | 829 | struct ctl_table_header *head; |
31e6b01f NP |
830 | /* Although proc doesn't have negative dentries, rcu-walk means |
831 | * that inode here can be NULL */ | |
dfef6dcd | 832 | /* AV: can it, indeed? */ |
31e6b01f | 833 | if (!inode) |
dfef6dcd | 834 | return 1; |
621e155a | 835 | if (name->len != len) |
9043476f | 836 | return 1; |
621e155a | 837 | if (memcmp(name->name, str, len)) |
9043476f | 838 | return 1; |
dfef6dcd AV |
839 | head = rcu_dereference(PROC_I(inode)->sysctl); |
840 | return !head || !sysctl_is_seen(head); | |
77b14db5 EB |
841 | } |
842 | ||
d72f71eb | 843 | static const struct dentry_operations proc_sys_dentry_operations = { |
77b14db5 | 844 | .d_revalidate = proc_sys_revalidate, |
9043476f AV |
845 | .d_delete = proc_sys_delete, |
846 | .d_compare = proc_sys_compare, | |
77b14db5 EB |
847 | }; |
848 | ||
0e47c99d EB |
849 | static struct ctl_dir *find_subdir(struct ctl_dir *dir, |
850 | const char *name, int namelen) | |
1f87f0b5 | 851 | { |
7ec66d06 EB |
852 | struct ctl_table_header *head; |
853 | struct ctl_table *entry; | |
1f87f0b5 | 854 | |
0e47c99d | 855 | entry = find_entry(&head, dir, name, namelen); |
7ec66d06 EB |
856 | if (!entry) |
857 | return ERR_PTR(-ENOENT); | |
51f72f4a EB |
858 | if (!S_ISDIR(entry->mode)) |
859 | return ERR_PTR(-ENOTDIR); | |
860 | return container_of(head, struct ctl_dir, header); | |
7ec66d06 EB |
861 | } |
862 | ||
863 | static struct ctl_dir *new_dir(struct ctl_table_set *set, | |
0e47c99d | 864 | const char *name, int namelen) |
7ec66d06 EB |
865 | { |
866 | struct ctl_table *table; | |
867 | struct ctl_dir *new; | |
ac13ac6f | 868 | struct ctl_node *node; |
7ec66d06 | 869 | char *new_name; |
1f87f0b5 | 870 | |
ac13ac6f EB |
871 | new = kzalloc(sizeof(*new) + sizeof(struct ctl_node) + |
872 | sizeof(struct ctl_table)*2 + namelen + 1, | |
873 | GFP_KERNEL); | |
7ec66d06 | 874 | if (!new) |
1f87f0b5 EB |
875 | return NULL; |
876 | ||
ac13ac6f EB |
877 | node = (struct ctl_node *)(new + 1); |
878 | table = (struct ctl_table *)(node + 1); | |
7ec66d06 EB |
879 | new_name = (char *)(table + 2); |
880 | memcpy(new_name, name, namelen); | |
881 | new_name[namelen] = '\0'; | |
882 | table[0].procname = new_name; | |
883 | table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
ac13ac6f | 884 | init_header(&new->header, set->dir.header.root, set, node, table); |
7ec66d06 EB |
885 | |
886 | return new; | |
1f87f0b5 EB |
887 | } |
888 | ||
60f126d9 EB |
889 | /** |
890 | * get_subdir - find or create a subdir with the specified name. | |
891 | * @dir: Directory to create the subdirectory in | |
892 | * @name: The name of the subdirectory to find or create | |
893 | * @namelen: The length of name | |
894 | * | |
895 | * Takes a directory with an elevated reference count so we know that | |
896 | * if we drop the lock the directory will not go away. Upon success | |
897 | * the reference is moved from @dir to the returned subdirectory. | |
898 | * Upon error an error code is returned and the reference on @dir is | |
899 | * simply dropped. | |
900 | */ | |
0e47c99d EB |
901 | static struct ctl_dir *get_subdir(struct ctl_dir *dir, |
902 | const char *name, int namelen) | |
1f87f0b5 | 903 | { |
0e47c99d | 904 | struct ctl_table_set *set = dir->header.set; |
7ec66d06 | 905 | struct ctl_dir *subdir, *new = NULL; |
0eb97f38 | 906 | int err; |
1f87f0b5 | 907 | |
7ec66d06 | 908 | spin_lock(&sysctl_lock); |
0e47c99d | 909 | subdir = find_subdir(dir, name, namelen); |
7ec66d06 EB |
910 | if (!IS_ERR(subdir)) |
911 | goto found; | |
912 | if (PTR_ERR(subdir) != -ENOENT) | |
913 | goto failed; | |
914 | ||
915 | spin_unlock(&sysctl_lock); | |
916 | new = new_dir(set, name, namelen); | |
917 | spin_lock(&sysctl_lock); | |
918 | subdir = ERR_PTR(-ENOMEM); | |
919 | if (!new) | |
920 | goto failed; | |
921 | ||
60f126d9 | 922 | /* Was the subdir added while we dropped the lock? */ |
0e47c99d | 923 | subdir = find_subdir(dir, name, namelen); |
7ec66d06 EB |
924 | if (!IS_ERR(subdir)) |
925 | goto found; | |
926 | if (PTR_ERR(subdir) != -ENOENT) | |
927 | goto failed; | |
928 | ||
60f126d9 | 929 | /* Nope. Use the our freshly made directory entry. */ |
0eb97f38 EB |
930 | err = insert_header(dir, &new->header); |
931 | subdir = ERR_PTR(err); | |
932 | if (err) | |
0e47c99d | 933 | goto failed; |
7ec66d06 EB |
934 | subdir = new; |
935 | found: | |
936 | subdir->header.nreg++; | |
937 | failed: | |
938 | if (unlikely(IS_ERR(subdir))) { | |
6980128f EB |
939 | printk(KERN_ERR "sysctl could not get directory: "); |
940 | sysctl_print_dir(dir); | |
941 | printk(KERN_CONT "/%*.*s %ld\n", | |
7ec66d06 | 942 | namelen, namelen, name, PTR_ERR(subdir)); |
1f87f0b5 | 943 | } |
7ec66d06 EB |
944 | drop_sysctl_table(&dir->header); |
945 | if (new) | |
946 | drop_sysctl_table(&new->header); | |
947 | spin_unlock(&sysctl_lock); | |
948 | return subdir; | |
1f87f0b5 EB |
949 | } |
950 | ||
0e47c99d EB |
951 | static struct ctl_dir *xlate_dir(struct ctl_table_set *set, struct ctl_dir *dir) |
952 | { | |
953 | struct ctl_dir *parent; | |
954 | const char *procname; | |
955 | if (!dir->header.parent) | |
956 | return &set->dir; | |
957 | parent = xlate_dir(set, dir->header.parent); | |
958 | if (IS_ERR(parent)) | |
959 | return parent; | |
960 | procname = dir->header.ctl_table[0].procname; | |
961 | return find_subdir(parent, procname, strlen(procname)); | |
962 | } | |
963 | ||
964 | static int sysctl_follow_link(struct ctl_table_header **phead, | |
965 | struct ctl_table **pentry, struct nsproxy *namespaces) | |
966 | { | |
967 | struct ctl_table_header *head; | |
968 | struct ctl_table_root *root; | |
969 | struct ctl_table_set *set; | |
970 | struct ctl_table *entry; | |
971 | struct ctl_dir *dir; | |
972 | int ret; | |
973 | ||
0e47c99d EB |
974 | ret = 0; |
975 | spin_lock(&sysctl_lock); | |
976 | root = (*pentry)->data; | |
977 | set = lookup_header_set(root, namespaces); | |
978 | dir = xlate_dir(set, (*phead)->parent); | |
979 | if (IS_ERR(dir)) | |
980 | ret = PTR_ERR(dir); | |
981 | else { | |
982 | const char *procname = (*pentry)->procname; | |
983 | head = NULL; | |
984 | entry = find_entry(&head, dir, procname, strlen(procname)); | |
985 | ret = -ENOENT; | |
986 | if (entry && use_table(head)) { | |
987 | unuse_table(*phead); | |
988 | *phead = head; | |
989 | *pentry = entry; | |
990 | ret = 0; | |
991 | } | |
992 | } | |
993 | ||
994 | spin_unlock(&sysctl_lock); | |
995 | return ret; | |
996 | } | |
997 | ||
7c60c48f | 998 | static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...) |
1f87f0b5 | 999 | { |
7c60c48f EB |
1000 | struct va_format vaf; |
1001 | va_list args; | |
1f87f0b5 | 1002 | |
7c60c48f EB |
1003 | va_start(args, fmt); |
1004 | vaf.fmt = fmt; | |
1005 | vaf.va = &args; | |
1006 | ||
1007 | printk(KERN_ERR "sysctl table check failed: %s/%s %pV\n", | |
1008 | path, table->procname, &vaf); | |
1f87f0b5 | 1009 | |
7c60c48f EB |
1010 | va_end(args); |
1011 | return -EINVAL; | |
1f87f0b5 EB |
1012 | } |
1013 | ||
7c60c48f | 1014 | static int sysctl_check_table(const char *path, struct ctl_table *table) |
1f87f0b5 | 1015 | { |
7c60c48f | 1016 | int err = 0; |
1f87f0b5 | 1017 | for (; table->procname; table++) { |
1f87f0b5 | 1018 | if (table->child) |
7c60c48f EB |
1019 | err = sysctl_err(path, table, "Not a file"); |
1020 | ||
1021 | if ((table->proc_handler == proc_dostring) || | |
1022 | (table->proc_handler == proc_dointvec) || | |
1023 | (table->proc_handler == proc_dointvec_minmax) || | |
1024 | (table->proc_handler == proc_dointvec_jiffies) || | |
1025 | (table->proc_handler == proc_dointvec_userhz_jiffies) || | |
1026 | (table->proc_handler == proc_dointvec_ms_jiffies) || | |
1027 | (table->proc_handler == proc_doulongvec_minmax) || | |
1028 | (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) { | |
1029 | if (!table->data) | |
1030 | err = sysctl_err(path, table, "No data"); | |
1031 | if (!table->maxlen) | |
1032 | err = sysctl_err(path, table, "No maxlen"); | |
1033 | } | |
1034 | if (!table->proc_handler) | |
1035 | err = sysctl_err(path, table, "No proc_handler"); | |
1036 | ||
1037 | if ((table->mode & (S_IRUGO|S_IWUGO)) != table->mode) | |
1038 | err = sysctl_err(path, table, "bogus .mode 0%o", | |
1039 | table->mode); | |
1f87f0b5 | 1040 | } |
7c60c48f | 1041 | return err; |
1f87f0b5 | 1042 | } |
1f87f0b5 | 1043 | |
0e47c99d EB |
1044 | static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table *table, |
1045 | struct ctl_table_root *link_root) | |
1046 | { | |
1047 | struct ctl_table *link_table, *entry, *link; | |
1048 | struct ctl_table_header *links; | |
ac13ac6f | 1049 | struct ctl_node *node; |
0e47c99d EB |
1050 | char *link_name; |
1051 | int nr_entries, name_bytes; | |
1052 | ||
1053 | name_bytes = 0; | |
1054 | nr_entries = 0; | |
1055 | for (entry = table; entry->procname; entry++) { | |
1056 | nr_entries++; | |
1057 | name_bytes += strlen(entry->procname) + 1; | |
1058 | } | |
1059 | ||
1060 | links = kzalloc(sizeof(struct ctl_table_header) + | |
ac13ac6f | 1061 | sizeof(struct ctl_node)*nr_entries + |
0e47c99d EB |
1062 | sizeof(struct ctl_table)*(nr_entries + 1) + |
1063 | name_bytes, | |
1064 | GFP_KERNEL); | |
1065 | ||
1066 | if (!links) | |
1067 | return NULL; | |
1068 | ||
ac13ac6f EB |
1069 | node = (struct ctl_node *)(links + 1); |
1070 | link_table = (struct ctl_table *)(node + nr_entries); | |
0e47c99d EB |
1071 | link_name = (char *)&link_table[nr_entries + 1]; |
1072 | ||
1073 | for (link = link_table, entry = table; entry->procname; link++, entry++) { | |
1074 | int len = strlen(entry->procname) + 1; | |
1075 | memcpy(link_name, entry->procname, len); | |
1076 | link->procname = link_name; | |
1077 | link->mode = S_IFLNK|S_IRWXUGO; | |
1078 | link->data = link_root; | |
1079 | link_name += len; | |
1080 | } | |
ac13ac6f | 1081 | init_header(links, dir->header.root, dir->header.set, node, link_table); |
0e47c99d EB |
1082 | links->nreg = nr_entries; |
1083 | ||
1084 | return links; | |
1085 | } | |
1086 | ||
1087 | static bool get_links(struct ctl_dir *dir, | |
1088 | struct ctl_table *table, struct ctl_table_root *link_root) | |
1089 | { | |
1090 | struct ctl_table_header *head; | |
1091 | struct ctl_table *entry, *link; | |
1092 | ||
1093 | /* Are there links available for every entry in table? */ | |
1094 | for (entry = table; entry->procname; entry++) { | |
1095 | const char *procname = entry->procname; | |
1096 | link = find_entry(&head, dir, procname, strlen(procname)); | |
1097 | if (!link) | |
1098 | return false; | |
1099 | if (S_ISDIR(link->mode) && S_ISDIR(entry->mode)) | |
1100 | continue; | |
1101 | if (S_ISLNK(link->mode) && (link->data == link_root)) | |
1102 | continue; | |
1103 | return false; | |
1104 | } | |
1105 | ||
1106 | /* The checks passed. Increase the registration count on the links */ | |
1107 | for (entry = table; entry->procname; entry++) { | |
1108 | const char *procname = entry->procname; | |
1109 | link = find_entry(&head, dir, procname, strlen(procname)); | |
1110 | head->nreg++; | |
1111 | } | |
1112 | return true; | |
1113 | } | |
1114 | ||
1115 | static int insert_links(struct ctl_table_header *head) | |
1116 | { | |
1117 | struct ctl_table_set *root_set = &sysctl_table_root.default_set; | |
1118 | struct ctl_dir *core_parent = NULL; | |
1119 | struct ctl_table_header *links; | |
1120 | int err; | |
1121 | ||
1122 | if (head->set == root_set) | |
1123 | return 0; | |
1124 | ||
1125 | core_parent = xlate_dir(root_set, head->parent); | |
1126 | if (IS_ERR(core_parent)) | |
1127 | return 0; | |
1128 | ||
1129 | if (get_links(core_parent, head->ctl_table, head->root)) | |
1130 | return 0; | |
1131 | ||
1132 | core_parent->header.nreg++; | |
1133 | spin_unlock(&sysctl_lock); | |
1134 | ||
1135 | links = new_links(core_parent, head->ctl_table, head->root); | |
1136 | ||
1137 | spin_lock(&sysctl_lock); | |
1138 | err = -ENOMEM; | |
1139 | if (!links) | |
1140 | goto out; | |
1141 | ||
1142 | err = 0; | |
1143 | if (get_links(core_parent, head->ctl_table, head->root)) { | |
1144 | kfree(links); | |
1145 | goto out; | |
1146 | } | |
1147 | ||
1148 | err = insert_header(core_parent, links); | |
1149 | if (err) | |
1150 | kfree(links); | |
1151 | out: | |
1152 | drop_sysctl_table(&core_parent->header); | |
1153 | return err; | |
1154 | } | |
1155 | ||
1f87f0b5 | 1156 | /** |
f728019b | 1157 | * __register_sysctl_table - register a leaf sysctl table |
60a47a2e | 1158 | * @set: Sysctl tree to register on |
1f87f0b5 EB |
1159 | * @path: The path to the directory the sysctl table is in. |
1160 | * @table: the top-level table structure | |
1161 | * | |
1162 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1163 | * array. A completely 0 filled entry terminates the table. | |
1164 | * | |
1165 | * The members of the &struct ctl_table structure are used as follows: | |
1166 | * | |
1167 | * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not | |
1168 | * enter a sysctl file | |
1169 | * | |
1170 | * data - a pointer to data for use by proc_handler | |
1171 | * | |
1172 | * maxlen - the maximum size in bytes of the data | |
1173 | * | |
f728019b | 1174 | * mode - the file permissions for the /proc/sys file |
1f87f0b5 | 1175 | * |
f728019b | 1176 | * child - must be %NULL. |
1f87f0b5 EB |
1177 | * |
1178 | * proc_handler - the text handler routine (described below) | |
1179 | * | |
1f87f0b5 EB |
1180 | * extra1, extra2 - extra pointers usable by the proc handler routines |
1181 | * | |
1182 | * Leaf nodes in the sysctl tree will be represented by a single file | |
1183 | * under /proc; non-leaf nodes will be represented by directories. | |
1184 | * | |
f728019b EB |
1185 | * There must be a proc_handler routine for any terminal nodes. |
1186 | * Several default handlers are available to cover common cases - | |
1f87f0b5 EB |
1187 | * |
1188 | * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(), | |
1189 | * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), | |
1190 | * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax() | |
1191 | * | |
1192 | * It is the handler's job to read the input buffer from user memory | |
1193 | * and process it. The handler should return 0 on success. | |
1194 | * | |
1195 | * This routine returns %NULL on a failure to register, and a pointer | |
1196 | * to the table header on success. | |
1197 | */ | |
6e9d5164 | 1198 | struct ctl_table_header *__register_sysctl_table( |
60a47a2e | 1199 | struct ctl_table_set *set, |
6e9d5164 | 1200 | const char *path, struct ctl_table *table) |
1f87f0b5 | 1201 | { |
60a47a2e | 1202 | struct ctl_table_root *root = set->dir.header.root; |
1f87f0b5 | 1203 | struct ctl_table_header *header; |
6e9d5164 | 1204 | const char *name, *nextname; |
7ec66d06 | 1205 | struct ctl_dir *dir; |
ac13ac6f EB |
1206 | struct ctl_table *entry; |
1207 | struct ctl_node *node; | |
1208 | int nr_entries = 0; | |
1209 | ||
1210 | for (entry = table; entry->procname; entry++) | |
1211 | nr_entries++; | |
1f87f0b5 | 1212 | |
ac13ac6f EB |
1213 | header = kzalloc(sizeof(struct ctl_table_header) + |
1214 | sizeof(struct ctl_node)*nr_entries, GFP_KERNEL); | |
1f87f0b5 EB |
1215 | if (!header) |
1216 | return NULL; | |
1217 | ||
ac13ac6f EB |
1218 | node = (struct ctl_node *)(header + 1); |
1219 | init_header(header, root, set, node, table); | |
7ec66d06 EB |
1220 | if (sysctl_check_table(path, table)) |
1221 | goto fail; | |
1222 | ||
1223 | spin_lock(&sysctl_lock); | |
0e47c99d | 1224 | dir = &set->dir; |
60f126d9 | 1225 | /* Reference moved down the diretory tree get_subdir */ |
7ec66d06 EB |
1226 | dir->header.nreg++; |
1227 | spin_unlock(&sysctl_lock); | |
1f87f0b5 | 1228 | |
7ec66d06 | 1229 | /* Find the directory for the ctl_table */ |
6e9d5164 EB |
1230 | for (name = path; name; name = nextname) { |
1231 | int namelen; | |
1232 | nextname = strchr(name, '/'); | |
1233 | if (nextname) { | |
1234 | namelen = nextname - name; | |
1235 | nextname++; | |
1236 | } else { | |
1237 | namelen = strlen(name); | |
1238 | } | |
1239 | if (namelen == 0) | |
1240 | continue; | |
1f87f0b5 | 1241 | |
0e47c99d | 1242 | dir = get_subdir(dir, name, namelen); |
7ec66d06 EB |
1243 | if (IS_ERR(dir)) |
1244 | goto fail; | |
1f87f0b5 | 1245 | } |
0e47c99d | 1246 | |
1f87f0b5 | 1247 | spin_lock(&sysctl_lock); |
0e47c99d | 1248 | if (insert_header(dir, header)) |
7ec66d06 | 1249 | goto fail_put_dir_locked; |
0e47c99d | 1250 | |
7ec66d06 | 1251 | drop_sysctl_table(&dir->header); |
1f87f0b5 EB |
1252 | spin_unlock(&sysctl_lock); |
1253 | ||
1254 | return header; | |
0e47c99d | 1255 | |
7ec66d06 EB |
1256 | fail_put_dir_locked: |
1257 | drop_sysctl_table(&dir->header); | |
7c60c48f EB |
1258 | spin_unlock(&sysctl_lock); |
1259 | fail: | |
1260 | kfree(header); | |
1261 | dump_stack(); | |
1262 | return NULL; | |
1f87f0b5 EB |
1263 | } |
1264 | ||
fea478d4 EB |
1265 | /** |
1266 | * register_sysctl - register a sysctl table | |
1267 | * @path: The path to the directory the sysctl table is in. | |
1268 | * @table: the table structure | |
1269 | * | |
1270 | * Register a sysctl table. @table should be a filled in ctl_table | |
1271 | * array. A completely 0 filled entry terminates the table. | |
1272 | * | |
1273 | * See __register_sysctl_table for more details. | |
1274 | */ | |
1275 | struct ctl_table_header *register_sysctl(const char *path, struct ctl_table *table) | |
1276 | { | |
1277 | return __register_sysctl_table(&sysctl_table_root.default_set, | |
1278 | path, table); | |
1279 | } | |
1280 | EXPORT_SYMBOL(register_sysctl); | |
1281 | ||
6e9d5164 EB |
1282 | static char *append_path(const char *path, char *pos, const char *name) |
1283 | { | |
1284 | int namelen; | |
1285 | namelen = strlen(name); | |
1286 | if (((pos - path) + namelen + 2) >= PATH_MAX) | |
1287 | return NULL; | |
1288 | memcpy(pos, name, namelen); | |
1289 | pos[namelen] = '/'; | |
1290 | pos[namelen + 1] = '\0'; | |
1291 | pos += namelen + 1; | |
1292 | return pos; | |
1293 | } | |
1294 | ||
f728019b EB |
1295 | static int count_subheaders(struct ctl_table *table) |
1296 | { | |
1297 | int has_files = 0; | |
1298 | int nr_subheaders = 0; | |
1299 | struct ctl_table *entry; | |
1300 | ||
1301 | /* special case: no directory and empty directory */ | |
1302 | if (!table || !table->procname) | |
1303 | return 1; | |
1304 | ||
1305 | for (entry = table; entry->procname; entry++) { | |
1306 | if (entry->child) | |
1307 | nr_subheaders += count_subheaders(entry->child); | |
1308 | else | |
1309 | has_files = 1; | |
1310 | } | |
1311 | return nr_subheaders + has_files; | |
1312 | } | |
1313 | ||
1314 | static int register_leaf_sysctl_tables(const char *path, char *pos, | |
60a47a2e | 1315 | struct ctl_table_header ***subheader, struct ctl_table_set *set, |
f728019b EB |
1316 | struct ctl_table *table) |
1317 | { | |
1318 | struct ctl_table *ctl_table_arg = NULL; | |
1319 | struct ctl_table *entry, *files; | |
1320 | int nr_files = 0; | |
1321 | int nr_dirs = 0; | |
1322 | int err = -ENOMEM; | |
1323 | ||
1324 | for (entry = table; entry->procname; entry++) { | |
1325 | if (entry->child) | |
1326 | nr_dirs++; | |
1327 | else | |
1328 | nr_files++; | |
1329 | } | |
1330 | ||
1331 | files = table; | |
1332 | /* If there are mixed files and directories we need a new table */ | |
1333 | if (nr_dirs && nr_files) { | |
1334 | struct ctl_table *new; | |
1335 | files = kzalloc(sizeof(struct ctl_table) * (nr_files + 1), | |
1336 | GFP_KERNEL); | |
1337 | if (!files) | |
1338 | goto out; | |
1339 | ||
1340 | ctl_table_arg = files; | |
1341 | for (new = files, entry = table; entry->procname; entry++) { | |
1342 | if (entry->child) | |
1343 | continue; | |
1344 | *new = *entry; | |
1345 | new++; | |
1346 | } | |
1347 | } | |
1348 | ||
1349 | /* Register everything except a directory full of subdirectories */ | |
1350 | if (nr_files || !nr_dirs) { | |
1351 | struct ctl_table_header *header; | |
60a47a2e | 1352 | header = __register_sysctl_table(set, path, files); |
f728019b EB |
1353 | if (!header) { |
1354 | kfree(ctl_table_arg); | |
1355 | goto out; | |
1356 | } | |
1357 | ||
1358 | /* Remember if we need to free the file table */ | |
1359 | header->ctl_table_arg = ctl_table_arg; | |
1360 | **subheader = header; | |
1361 | (*subheader)++; | |
1362 | } | |
1363 | ||
1364 | /* Recurse into the subdirectories. */ | |
1365 | for (entry = table; entry->procname; entry++) { | |
1366 | char *child_pos; | |
1367 | ||
1368 | if (!entry->child) | |
1369 | continue; | |
1370 | ||
1371 | err = -ENAMETOOLONG; | |
1372 | child_pos = append_path(path, pos, entry->procname); | |
1373 | if (!child_pos) | |
1374 | goto out; | |
1375 | ||
1376 | err = register_leaf_sysctl_tables(path, child_pos, subheader, | |
60a47a2e | 1377 | set, entry->child); |
f728019b EB |
1378 | pos[0] = '\0'; |
1379 | if (err) | |
1380 | goto out; | |
1381 | } | |
1382 | err = 0; | |
1383 | out: | |
1384 | /* On failure our caller will unregister all registered subheaders */ | |
1385 | return err; | |
1386 | } | |
1387 | ||
6e9d5164 EB |
1388 | /** |
1389 | * __register_sysctl_paths - register a sysctl table hierarchy | |
60a47a2e | 1390 | * @set: Sysctl tree to register on |
6e9d5164 EB |
1391 | * @path: The path to the directory the sysctl table is in. |
1392 | * @table: the top-level table structure | |
1393 | * | |
1394 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1395 | * array. A completely 0 filled entry terminates the table. | |
1396 | * | |
1397 | * See __register_sysctl_table for more details. | |
1398 | */ | |
1399 | struct ctl_table_header *__register_sysctl_paths( | |
60a47a2e | 1400 | struct ctl_table_set *set, |
6e9d5164 EB |
1401 | const struct ctl_path *path, struct ctl_table *table) |
1402 | { | |
ec6a5266 | 1403 | struct ctl_table *ctl_table_arg = table; |
f728019b EB |
1404 | int nr_subheaders = count_subheaders(table); |
1405 | struct ctl_table_header *header = NULL, **subheaders, **subheader; | |
6e9d5164 EB |
1406 | const struct ctl_path *component; |
1407 | char *new_path, *pos; | |
1408 | ||
1409 | pos = new_path = kmalloc(PATH_MAX, GFP_KERNEL); | |
1410 | if (!new_path) | |
1411 | return NULL; | |
1412 | ||
1413 | pos[0] = '\0'; | |
1414 | for (component = path; component->procname; component++) { | |
1415 | pos = append_path(new_path, pos, component->procname); | |
1416 | if (!pos) | |
1417 | goto out; | |
1418 | } | |
ec6a5266 EB |
1419 | while (table->procname && table->child && !table[1].procname) { |
1420 | pos = append_path(new_path, pos, table->procname); | |
1421 | if (!pos) | |
1422 | goto out; | |
1423 | table = table->child; | |
1424 | } | |
f728019b | 1425 | if (nr_subheaders == 1) { |
60a47a2e | 1426 | header = __register_sysctl_table(set, new_path, table); |
f728019b EB |
1427 | if (header) |
1428 | header->ctl_table_arg = ctl_table_arg; | |
1429 | } else { | |
1430 | header = kzalloc(sizeof(*header) + | |
1431 | sizeof(*subheaders)*nr_subheaders, GFP_KERNEL); | |
1432 | if (!header) | |
1433 | goto out; | |
1434 | ||
1435 | subheaders = (struct ctl_table_header **) (header + 1); | |
1436 | subheader = subheaders; | |
ec6a5266 | 1437 | header->ctl_table_arg = ctl_table_arg; |
f728019b EB |
1438 | |
1439 | if (register_leaf_sysctl_tables(new_path, pos, &subheader, | |
60a47a2e | 1440 | set, table)) |
f728019b EB |
1441 | goto err_register_leaves; |
1442 | } | |
1443 | ||
6e9d5164 EB |
1444 | out: |
1445 | kfree(new_path); | |
1446 | return header; | |
f728019b EB |
1447 | |
1448 | err_register_leaves: | |
1449 | while (subheader > subheaders) { | |
1450 | struct ctl_table_header *subh = *(--subheader); | |
1451 | struct ctl_table *table = subh->ctl_table_arg; | |
1452 | unregister_sysctl_table(subh); | |
1453 | kfree(table); | |
1454 | } | |
1455 | kfree(header); | |
1456 | header = NULL; | |
1457 | goto out; | |
6e9d5164 EB |
1458 | } |
1459 | ||
1f87f0b5 EB |
1460 | /** |
1461 | * register_sysctl_table_path - register a sysctl table hierarchy | |
1462 | * @path: The path to the directory the sysctl table is in. | |
1463 | * @table: the top-level table structure | |
1464 | * | |
1465 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1466 | * array. A completely 0 filled entry terminates the table. | |
1467 | * | |
1468 | * See __register_sysctl_paths for more details. | |
1469 | */ | |
1470 | struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path, | |
1471 | struct ctl_table *table) | |
1472 | { | |
60a47a2e | 1473 | return __register_sysctl_paths(&sysctl_table_root.default_set, |
1f87f0b5 EB |
1474 | path, table); |
1475 | } | |
1476 | EXPORT_SYMBOL(register_sysctl_paths); | |
1477 | ||
1478 | /** | |
1479 | * register_sysctl_table - register a sysctl table hierarchy | |
1480 | * @table: the top-level table structure | |
1481 | * | |
1482 | * Register a sysctl table hierarchy. @table should be a filled in ctl_table | |
1483 | * array. A completely 0 filled entry terminates the table. | |
1484 | * | |
1485 | * See register_sysctl_paths for more details. | |
1486 | */ | |
1487 | struct ctl_table_header *register_sysctl_table(struct ctl_table *table) | |
1488 | { | |
1489 | static const struct ctl_path null_path[] = { {} }; | |
1490 | ||
1491 | return register_sysctl_paths(null_path, table); | |
1492 | } | |
1493 | EXPORT_SYMBOL(register_sysctl_table); | |
1494 | ||
0e47c99d EB |
1495 | static void put_links(struct ctl_table_header *header) |
1496 | { | |
1497 | struct ctl_table_set *root_set = &sysctl_table_root.default_set; | |
1498 | struct ctl_table_root *root = header->root; | |
1499 | struct ctl_dir *parent = header->parent; | |
1500 | struct ctl_dir *core_parent; | |
1501 | struct ctl_table *entry; | |
1502 | ||
1503 | if (header->set == root_set) | |
1504 | return; | |
1505 | ||
1506 | core_parent = xlate_dir(root_set, parent); | |
1507 | if (IS_ERR(core_parent)) | |
1508 | return; | |
1509 | ||
1510 | for (entry = header->ctl_table; entry->procname; entry++) { | |
1511 | struct ctl_table_header *link_head; | |
1512 | struct ctl_table *link; | |
1513 | const char *name = entry->procname; | |
1514 | ||
1515 | link = find_entry(&link_head, core_parent, name, strlen(name)); | |
1516 | if (link && | |
1517 | ((S_ISDIR(link->mode) && S_ISDIR(entry->mode)) || | |
1518 | (S_ISLNK(link->mode) && (link->data == root)))) { | |
1519 | drop_sysctl_table(link_head); | |
1520 | } | |
1521 | else { | |
1522 | printk(KERN_ERR "sysctl link missing during unregister: "); | |
1523 | sysctl_print_dir(parent); | |
1524 | printk(KERN_CONT "/%s\n", name); | |
1525 | } | |
1526 | } | |
1527 | } | |
1528 | ||
938aaa4f EB |
1529 | static void drop_sysctl_table(struct ctl_table_header *header) |
1530 | { | |
7ec66d06 EB |
1531 | struct ctl_dir *parent = header->parent; |
1532 | ||
938aaa4f EB |
1533 | if (--header->nreg) |
1534 | return; | |
1535 | ||
0e47c99d | 1536 | put_links(header); |
938aaa4f | 1537 | start_unregistering(header); |
938aaa4f EB |
1538 | if (!--header->count) |
1539 | kfree_rcu(header, rcu); | |
7ec66d06 EB |
1540 | |
1541 | if (parent) | |
1542 | drop_sysctl_table(&parent->header); | |
938aaa4f EB |
1543 | } |
1544 | ||
1f87f0b5 EB |
1545 | /** |
1546 | * unregister_sysctl_table - unregister a sysctl table hierarchy | |
1547 | * @header: the header returned from register_sysctl_table | |
1548 | * | |
1549 | * Unregisters the sysctl table and all children. proc entries may not | |
1550 | * actually be removed until they are no longer used by anyone. | |
1551 | */ | |
1552 | void unregister_sysctl_table(struct ctl_table_header * header) | |
1553 | { | |
f728019b | 1554 | int nr_subheaders; |
1f87f0b5 EB |
1555 | might_sleep(); |
1556 | ||
1557 | if (header == NULL) | |
1558 | return; | |
1559 | ||
f728019b EB |
1560 | nr_subheaders = count_subheaders(header->ctl_table_arg); |
1561 | if (unlikely(nr_subheaders > 1)) { | |
1562 | struct ctl_table_header **subheaders; | |
1563 | int i; | |
1564 | ||
1565 | subheaders = (struct ctl_table_header **)(header + 1); | |
1566 | for (i = nr_subheaders -1; i >= 0; i--) { | |
1567 | struct ctl_table_header *subh = subheaders[i]; | |
1568 | struct ctl_table *table = subh->ctl_table_arg; | |
1569 | unregister_sysctl_table(subh); | |
1570 | kfree(table); | |
1571 | } | |
1572 | kfree(header); | |
1573 | return; | |
1574 | } | |
1575 | ||
1f87f0b5 | 1576 | spin_lock(&sysctl_lock); |
938aaa4f | 1577 | drop_sysctl_table(header); |
1f87f0b5 EB |
1578 | spin_unlock(&sysctl_lock); |
1579 | } | |
1580 | EXPORT_SYMBOL(unregister_sysctl_table); | |
1581 | ||
0e47c99d | 1582 | void setup_sysctl_set(struct ctl_table_set *set, |
9eb47c26 | 1583 | struct ctl_table_root *root, |
1f87f0b5 EB |
1584 | int (*is_seen)(struct ctl_table_set *)) |
1585 | { | |
1347440d | 1586 | memset(set, 0, sizeof(*set)); |
0e47c99d | 1587 | set->is_seen = is_seen; |
ac13ac6f | 1588 | init_header(&set->dir.header, root, set, NULL, root_table); |
1f87f0b5 EB |
1589 | } |
1590 | ||
97324cd8 EB |
1591 | void retire_sysctl_set(struct ctl_table_set *set) |
1592 | { | |
ac13ac6f | 1593 | WARN_ON(!RB_EMPTY_ROOT(&set->dir.root)); |
97324cd8 | 1594 | } |
1f87f0b5 | 1595 | |
1e0edd3f | 1596 | int __init proc_sys_init(void) |
77b14db5 | 1597 | { |
e1675231 AD |
1598 | struct proc_dir_entry *proc_sys_root; |
1599 | ||
77b14db5 | 1600 | proc_sys_root = proc_mkdir("sys", NULL); |
9043476f AV |
1601 | proc_sys_root->proc_iops = &proc_sys_dir_operations; |
1602 | proc_sys_root->proc_fops = &proc_sys_dir_file_operations; | |
77b14db5 | 1603 | proc_sys_root->nlink = 0; |
de4e83bd EB |
1604 | |
1605 | return sysctl_init(); | |
77b14db5 | 1606 | } |