]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NETLINK Kernel-user communication protocol. | |
3 | * | |
113aa838 | 4 | * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk> |
1da177e4 | 5 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> |
cd1df525 | 6 | * Patrick McHardy <kaber@trash.net> |
1da177e4 LT |
7 | * |
8 | * This program is free software; you can redistribute it and/or | |
9 | * modify it under the terms of the GNU General Public License | |
10 | * as published by the Free Software Foundation; either version | |
11 | * 2 of the License, or (at your option) any later version. | |
746fac4d | 12 | * |
1da177e4 LT |
13 | * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith |
14 | * added netlink_proto_exit | |
15 | * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br> | |
16 | * use nlk_sk, as sk->protinfo is on a diet 8) | |
4fdb3bb7 HW |
17 | * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org> |
18 | * - inc module use count of module that owns | |
19 | * the kernel socket in case userspace opens | |
20 | * socket of same protocol | |
21 | * - remove all module support, since netlink is | |
22 | * mandatory if CONFIG_NET=y these days | |
1da177e4 LT |
23 | */ |
24 | ||
1da177e4 LT |
25 | #include <linux/module.h> |
26 | ||
4fc268d2 | 27 | #include <linux/capability.h> |
1da177e4 LT |
28 | #include <linux/kernel.h> |
29 | #include <linux/init.h> | |
1da177e4 LT |
30 | #include <linux/signal.h> |
31 | #include <linux/sched.h> | |
32 | #include <linux/errno.h> | |
33 | #include <linux/string.h> | |
34 | #include <linux/stat.h> | |
35 | #include <linux/socket.h> | |
36 | #include <linux/un.h> | |
37 | #include <linux/fcntl.h> | |
38 | #include <linux/termios.h> | |
39 | #include <linux/sockios.h> | |
40 | #include <linux/net.h> | |
41 | #include <linux/fs.h> | |
42 | #include <linux/slab.h> | |
7c0f6ba6 | 43 | #include <linux/uaccess.h> |
1da177e4 LT |
44 | #include <linux/skbuff.h> |
45 | #include <linux/netdevice.h> | |
46 | #include <linux/rtnetlink.h> | |
47 | #include <linux/proc_fs.h> | |
48 | #include <linux/seq_file.h> | |
1da177e4 LT |
49 | #include <linux/notifier.h> |
50 | #include <linux/security.h> | |
51 | #include <linux/jhash.h> | |
52 | #include <linux/jiffies.h> | |
53 | #include <linux/random.h> | |
54 | #include <linux/bitops.h> | |
55 | #include <linux/mm.h> | |
56 | #include <linux/types.h> | |
54e0f520 | 57 | #include <linux/audit.h> |
af65bdfc | 58 | #include <linux/mutex.h> |
ccdfcc39 | 59 | #include <linux/vmalloc.h> |
bcbde0d4 | 60 | #include <linux/if_arp.h> |
e341694e | 61 | #include <linux/rhashtable.h> |
9652e931 | 62 | #include <asm/cacheflush.h> |
e341694e | 63 | #include <linux/hash.h> |
ee1c2442 | 64 | #include <linux/genetlink.h> |
54e0f520 | 65 | |
457c4cbc | 66 | #include <net/net_namespace.h> |
1da177e4 LT |
67 | #include <net/sock.h> |
68 | #include <net/scm.h> | |
82ace47a | 69 | #include <net/netlink.h> |
1da177e4 | 70 | |
0f29c768 | 71 | #include "af_netlink.h" |
1da177e4 | 72 | |
5c398dc8 ED |
73 | struct listeners { |
74 | struct rcu_head rcu; | |
75 | unsigned long masks[0]; | |
6c04bb18 JB |
76 | }; |
77 | ||
cd967e05 | 78 | /* state bits */ |
cc3a572f | 79 | #define NETLINK_S_CONGESTED 0x0 |
cd967e05 | 80 | |
035c4c16 | 81 | static inline int netlink_is_kernel(struct sock *sk) |
aed81560 | 82 | { |
cc3a572f | 83 | return nlk_sk(sk)->flags & NETLINK_F_KERNEL_SOCKET; |
aed81560 DL |
84 | } |
85 | ||
91dd93f9 | 86 | struct netlink_table *nl_table __read_mostly; |
0f29c768 | 87 | EXPORT_SYMBOL_GPL(nl_table); |
1da177e4 LT |
88 | |
89 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
90 | ||
8a0f5ccf HX |
91 | static struct lock_class_key nlk_cb_mutex_keys[MAX_LINKS]; |
92 | ||
93 | static const char *const nlk_cb_mutex_key_strings[MAX_LINKS + 1] = { | |
94 | "nlk_cb_mutex-ROUTE", | |
95 | "nlk_cb_mutex-1", | |
96 | "nlk_cb_mutex-USERSOCK", | |
97 | "nlk_cb_mutex-FIREWALL", | |
98 | "nlk_cb_mutex-SOCK_DIAG", | |
99 | "nlk_cb_mutex-NFLOG", | |
100 | "nlk_cb_mutex-XFRM", | |
101 | "nlk_cb_mutex-SELINUX", | |
102 | "nlk_cb_mutex-ISCSI", | |
103 | "nlk_cb_mutex-AUDIT", | |
104 | "nlk_cb_mutex-FIB_LOOKUP", | |
105 | "nlk_cb_mutex-CONNECTOR", | |
106 | "nlk_cb_mutex-NETFILTER", | |
107 | "nlk_cb_mutex-IP6_FW", | |
108 | "nlk_cb_mutex-DNRTMSG", | |
109 | "nlk_cb_mutex-KOBJECT_UEVENT", | |
110 | "nlk_cb_mutex-GENERIC", | |
111 | "nlk_cb_mutex-17", | |
112 | "nlk_cb_mutex-SCSITRANSPORT", | |
113 | "nlk_cb_mutex-ECRYPTFS", | |
114 | "nlk_cb_mutex-RDMA", | |
115 | "nlk_cb_mutex-CRYPTO", | |
116 | "nlk_cb_mutex-SMC", | |
117 | "nlk_cb_mutex-23", | |
118 | "nlk_cb_mutex-24", | |
119 | "nlk_cb_mutex-25", | |
120 | "nlk_cb_mutex-26", | |
121 | "nlk_cb_mutex-27", | |
122 | "nlk_cb_mutex-28", | |
123 | "nlk_cb_mutex-29", | |
124 | "nlk_cb_mutex-30", | |
125 | "nlk_cb_mutex-31", | |
126 | "nlk_cb_mutex-MAX_LINKS" | |
127 | }; | |
128 | ||
1da177e4 | 129 | static int netlink_dump(struct sock *sk); |
9652e931 | 130 | static void netlink_skb_destructor(struct sk_buff *skb); |
1da177e4 | 131 | |
78fd1d0a | 132 | /* nl_table locking explained: |
21e4902a | 133 | * Lookup and traversal are protected with an RCU read-side lock. Insertion |
c5adde94 | 134 | * and removal are protected with per bucket lock while using RCU list |
21e4902a TG |
135 | * modification primitives and may run in parallel to RCU protected lookups. |
136 | * Destruction of the Netlink socket may only occur *after* nl_table_lock has | |
137 | * been acquired * either during or after the socket has been removed from | |
138 | * the list and after an RCU grace period. | |
78fd1d0a | 139 | */ |
0f29c768 AV |
140 | DEFINE_RWLOCK(nl_table_lock); |
141 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
142 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
143 | ||
6d772ac5 ED |
144 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
145 | ||
efa172f4 | 146 | static BLOCKING_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 147 | |
bcbde0d4 DB |
148 | static DEFINE_SPINLOCK(netlink_tap_lock); |
149 | static struct list_head netlink_tap_all __read_mostly; | |
150 | ||
c428ecd1 HX |
151 | static const struct rhashtable_params netlink_rhashtable_params; |
152 | ||
b57ef81f | 153 | static inline u32 netlink_group_mask(u32 group) |
d629b836 PM |
154 | { |
155 | return group ? 1 << (group - 1) : 0; | |
156 | } | |
157 | ||
1853c949 DB |
158 | static struct sk_buff *netlink_to_full_skb(const struct sk_buff *skb, |
159 | gfp_t gfp_mask) | |
160 | { | |
161 | unsigned int len = skb_end_offset(skb); | |
162 | struct sk_buff *new; | |
163 | ||
164 | new = alloc_skb(len, gfp_mask); | |
165 | if (new == NULL) | |
166 | return NULL; | |
167 | ||
168 | NETLINK_CB(new).portid = NETLINK_CB(skb).portid; | |
169 | NETLINK_CB(new).dst_group = NETLINK_CB(skb).dst_group; | |
170 | NETLINK_CB(new).creds = NETLINK_CB(skb).creds; | |
171 | ||
172 | memcpy(skb_put(new, len), skb->data, len); | |
173 | return new; | |
174 | } | |
175 | ||
bcbde0d4 DB |
176 | int netlink_add_tap(struct netlink_tap *nt) |
177 | { | |
178 | if (unlikely(nt->dev->type != ARPHRD_NETLINK)) | |
179 | return -EINVAL; | |
180 | ||
181 | spin_lock(&netlink_tap_lock); | |
182 | list_add_rcu(&nt->list, &netlink_tap_all); | |
183 | spin_unlock(&netlink_tap_lock); | |
184 | ||
fcd4d35e | 185 | __module_get(nt->module); |
bcbde0d4 DB |
186 | |
187 | return 0; | |
188 | } | |
189 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
190 | ||
2173f8d9 | 191 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 DB |
192 | { |
193 | bool found = false; | |
194 | struct netlink_tap *tmp; | |
195 | ||
196 | spin_lock(&netlink_tap_lock); | |
197 | ||
198 | list_for_each_entry(tmp, &netlink_tap_all, list) { | |
199 | if (nt == tmp) { | |
200 | list_del_rcu(&nt->list); | |
201 | found = true; | |
202 | goto out; | |
203 | } | |
204 | } | |
205 | ||
206 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
207 | out: | |
208 | spin_unlock(&netlink_tap_lock); | |
209 | ||
92b80eb3 | 210 | if (found) |
bcbde0d4 DB |
211 | module_put(nt->module); |
212 | ||
213 | return found ? 0 : -ENODEV; | |
214 | } | |
bcbde0d4 DB |
215 | |
216 | int netlink_remove_tap(struct netlink_tap *nt) | |
217 | { | |
218 | int ret; | |
219 | ||
220 | ret = __netlink_remove_tap(nt); | |
221 | synchronize_net(); | |
222 | ||
223 | return ret; | |
224 | } | |
225 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
226 | ||
5ffd5cdd DB |
227 | static bool netlink_filter_tap(const struct sk_buff *skb) |
228 | { | |
229 | struct sock *sk = skb->sk; | |
5ffd5cdd DB |
230 | |
231 | /* We take the more conservative approach and | |
232 | * whitelist socket protocols that may pass. | |
233 | */ | |
234 | switch (sk->sk_protocol) { | |
235 | case NETLINK_ROUTE: | |
236 | case NETLINK_USERSOCK: | |
237 | case NETLINK_SOCK_DIAG: | |
238 | case NETLINK_NFLOG: | |
239 | case NETLINK_XFRM: | |
240 | case NETLINK_FIB_LOOKUP: | |
241 | case NETLINK_NETFILTER: | |
242 | case NETLINK_GENERIC: | |
498044bb | 243 | return true; |
5ffd5cdd DB |
244 | } |
245 | ||
498044bb | 246 | return false; |
5ffd5cdd DB |
247 | } |
248 | ||
bcbde0d4 DB |
249 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
250 | struct net_device *dev) | |
251 | { | |
252 | struct sk_buff *nskb; | |
5ffd5cdd | 253 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
254 | int ret = -ENOMEM; |
255 | ||
256 | dev_hold(dev); | |
1853c949 | 257 | |
d1b4c689 | 258 | if (is_vmalloc_addr(skb->head)) |
1853c949 DB |
259 | nskb = netlink_to_full_skb(skb, GFP_ATOMIC); |
260 | else | |
261 | nskb = skb_clone(skb, GFP_ATOMIC); | |
bcbde0d4 DB |
262 | if (nskb) { |
263 | nskb->dev = dev; | |
5ffd5cdd | 264 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
265 | nskb->pkt_type = netlink_is_kernel(sk) ? |
266 | PACKET_KERNEL : PACKET_USER; | |
4e48ed88 | 267 | skb_reset_network_header(nskb); |
bcbde0d4 DB |
268 | ret = dev_queue_xmit(nskb); |
269 | if (unlikely(ret > 0)) | |
270 | ret = net_xmit_errno(ret); | |
271 | } | |
272 | ||
273 | dev_put(dev); | |
274 | return ret; | |
275 | } | |
276 | ||
277 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
278 | { | |
279 | int ret; | |
280 | struct netlink_tap *tmp; | |
281 | ||
5ffd5cdd DB |
282 | if (!netlink_filter_tap(skb)) |
283 | return; | |
284 | ||
bcbde0d4 DB |
285 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
286 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
287 | if (unlikely(ret)) | |
288 | break; | |
289 | } | |
290 | } | |
291 | ||
292 | static void netlink_deliver_tap(struct sk_buff *skb) | |
293 | { | |
294 | rcu_read_lock(); | |
295 | ||
296 | if (unlikely(!list_empty(&netlink_tap_all))) | |
297 | __netlink_deliver_tap(skb); | |
298 | ||
299 | rcu_read_unlock(); | |
300 | } | |
301 | ||
73bfd370 DB |
302 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
303 | struct sk_buff *skb) | |
304 | { | |
305 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
306 | netlink_deliver_tap(skb); | |
307 | } | |
308 | ||
cd1df525 PM |
309 | static void netlink_overrun(struct sock *sk) |
310 | { | |
311 | struct netlink_sock *nlk = nlk_sk(sk); | |
312 | ||
cc3a572f ND |
313 | if (!(nlk->flags & NETLINK_F_RECV_NO_ENOBUFS)) { |
314 | if (!test_and_set_bit(NETLINK_S_CONGESTED, | |
315 | &nlk_sk(sk)->state)) { | |
cd1df525 PM |
316 | sk->sk_err = ENOBUFS; |
317 | sk->sk_error_report(sk); | |
318 | } | |
319 | } | |
320 | atomic_inc(&sk->sk_drops); | |
321 | } | |
322 | ||
323 | static void netlink_rcv_wake(struct sock *sk) | |
324 | { | |
325 | struct netlink_sock *nlk = nlk_sk(sk); | |
326 | ||
327 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
cc3a572f ND |
328 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); |
329 | if (!test_bit(NETLINK_S_CONGESTED, &nlk->state)) | |
cd1df525 PM |
330 | wake_up_interruptible(&nlk->wait); |
331 | } | |
332 | ||
cf0a018a PM |
333 | static void netlink_skb_destructor(struct sk_buff *skb) |
334 | { | |
c05cdb1b | 335 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
336 | if (!skb->cloned || |
337 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
338 | vfree(skb->head); | |
339 | ||
c05cdb1b PNA |
340 | skb->head = NULL; |
341 | } | |
9652e931 PM |
342 | if (skb->sk != NULL) |
343 | sock_rfree(skb); | |
cf0a018a PM |
344 | } |
345 | ||
346 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
347 | { | |
348 | WARN_ON(skb->sk != NULL); | |
349 | skb->sk = sk; | |
350 | skb->destructor = netlink_skb_destructor; | |
351 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
352 | sk_mem_charge(sk, skb->truesize); | |
353 | } | |
354 | ||
ed5d7788 | 355 | static void netlink_sock_destruct(struct sock *sk) |
1da177e4 | 356 | { |
3f660d66 HX |
357 | struct netlink_sock *nlk = nlk_sk(sk); |
358 | ||
16b304f3 | 359 | if (nlk->cb_running) { |
ed5d7788 HX |
360 | if (nlk->cb.done) |
361 | nlk->cb.done(&nlk->cb); | |
16b304f3 PS |
362 | module_put(nlk->cb.module); |
363 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
364 | } |
365 | ||
1da177e4 LT |
366 | skb_queue_purge(&sk->sk_receive_queue); |
367 | ||
368 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 369 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
370 | return; |
371 | } | |
547b792c IJ |
372 | |
373 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
374 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
375 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
376 | } |
377 | ||
707693c8 HX |
378 | static void netlink_sock_destruct_work(struct work_struct *work) |
379 | { | |
380 | struct netlink_sock *nlk = container_of(work, struct netlink_sock, | |
381 | work); | |
382 | ||
ed5d7788 | 383 | sk_free(&nlk->sk); |
707693c8 HX |
384 | } |
385 | ||
6ac552fd PM |
386 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
387 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
388 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
389 | * this, _but_ remember, it adds useless work on UP machines. | |
390 | */ | |
391 | ||
d136f1bd | 392 | void netlink_table_grab(void) |
9a429c49 | 393 | __acquires(nl_table_lock) |
1da177e4 | 394 | { |
d136f1bd JB |
395 | might_sleep(); |
396 | ||
6abd219c | 397 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
398 | |
399 | if (atomic_read(&nl_table_users)) { | |
400 | DECLARE_WAITQUEUE(wait, current); | |
401 | ||
402 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 403 | for (;;) { |
1da177e4 LT |
404 | set_current_state(TASK_UNINTERRUPTIBLE); |
405 | if (atomic_read(&nl_table_users) == 0) | |
406 | break; | |
6abd219c | 407 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 408 | schedule(); |
6abd219c | 409 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
410 | } |
411 | ||
412 | __set_current_state(TASK_RUNNING); | |
413 | remove_wait_queue(&nl_table_wait, &wait); | |
414 | } | |
415 | } | |
416 | ||
d136f1bd | 417 | void netlink_table_ungrab(void) |
9a429c49 | 418 | __releases(nl_table_lock) |
1da177e4 | 419 | { |
6abd219c | 420 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
421 | wake_up(&nl_table_wait); |
422 | } | |
423 | ||
6ac552fd | 424 | static inline void |
1da177e4 LT |
425 | netlink_lock_table(void) |
426 | { | |
427 | /* read_lock() synchronizes us to netlink_table_grab */ | |
428 | ||
429 | read_lock(&nl_table_lock); | |
430 | atomic_inc(&nl_table_users); | |
431 | read_unlock(&nl_table_lock); | |
432 | } | |
433 | ||
6ac552fd | 434 | static inline void |
1da177e4 LT |
435 | netlink_unlock_table(void) |
436 | { | |
437 | if (atomic_dec_and_test(&nl_table_users)) | |
438 | wake_up(&nl_table_wait); | |
439 | } | |
440 | ||
e341694e | 441 | struct netlink_compare_arg |
1da177e4 | 442 | { |
c428ecd1 | 443 | possible_net_t pnet; |
e341694e TG |
444 | u32 portid; |
445 | }; | |
1da177e4 | 446 | |
8f2ddaac HX |
447 | /* Doing sizeof directly may yield 4 extra bytes on 64-bit. */ |
448 | #define netlink_compare_arg_len \ | |
449 | (offsetof(struct netlink_compare_arg, portid) + sizeof(u32)) | |
c428ecd1 HX |
450 | |
451 | static inline int netlink_compare(struct rhashtable_compare_arg *arg, | |
452 | const void *ptr) | |
1da177e4 | 453 | { |
c428ecd1 HX |
454 | const struct netlink_compare_arg *x = arg->key; |
455 | const struct netlink_sock *nlk = ptr; | |
1da177e4 | 456 | |
da314c99 | 457 | return nlk->portid != x->portid || |
c428ecd1 HX |
458 | !net_eq(sock_net(&nlk->sk), read_pnet(&x->pnet)); |
459 | } | |
460 | ||
461 | static void netlink_compare_arg_init(struct netlink_compare_arg *arg, | |
462 | struct net *net, u32 portid) | |
463 | { | |
464 | memset(arg, 0, sizeof(*arg)); | |
465 | write_pnet(&arg->pnet, net); | |
466 | arg->portid = portid; | |
1da177e4 LT |
467 | } |
468 | ||
e341694e TG |
469 | static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid, |
470 | struct net *net) | |
1da177e4 | 471 | { |
c428ecd1 | 472 | struct netlink_compare_arg arg; |
1da177e4 | 473 | |
c428ecd1 HX |
474 | netlink_compare_arg_init(&arg, net, portid); |
475 | return rhashtable_lookup_fast(&table->hash, &arg, | |
476 | netlink_rhashtable_params); | |
1da177e4 LT |
477 | } |
478 | ||
c428ecd1 | 479 | static int __netlink_insert(struct netlink_table *table, struct sock *sk) |
c5adde94 | 480 | { |
c428ecd1 | 481 | struct netlink_compare_arg arg; |
c5adde94 | 482 | |
da314c99 | 483 | netlink_compare_arg_init(&arg, sock_net(sk), nlk_sk(sk)->portid); |
c428ecd1 HX |
484 | return rhashtable_lookup_insert_key(&table->hash, &arg, |
485 | &nlk_sk(sk)->node, | |
486 | netlink_rhashtable_params); | |
c5adde94 YX |
487 | } |
488 | ||
e341694e | 489 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 490 | { |
e341694e TG |
491 | struct netlink_table *table = &nl_table[protocol]; |
492 | struct sock *sk; | |
1da177e4 | 493 | |
e341694e TG |
494 | rcu_read_lock(); |
495 | sk = __netlink_lookup(table, portid, net); | |
496 | if (sk) | |
497 | sock_hold(sk); | |
498 | rcu_read_unlock(); | |
1da177e4 | 499 | |
e341694e | 500 | return sk; |
1da177e4 LT |
501 | } |
502 | ||
90ddc4f0 | 503 | static const struct proto_ops netlink_ops; |
1da177e4 | 504 | |
4277a083 PM |
505 | static void |
506 | netlink_update_listeners(struct sock *sk) | |
507 | { | |
508 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
509 | unsigned long mask; |
510 | unsigned int i; | |
6d772ac5 ED |
511 | struct listeners *listeners; |
512 | ||
513 | listeners = nl_deref_protected(tbl->listeners); | |
514 | if (!listeners) | |
515 | return; | |
4277a083 | 516 | |
b4ff4f04 | 517 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 518 | mask = 0; |
b67bfe0d | 519 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
520 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
521 | mask |= nlk_sk(sk)->groups[i]; | |
522 | } | |
6d772ac5 | 523 | listeners->masks[i] = mask; |
4277a083 PM |
524 | } |
525 | /* this function is only called with the netlink table "grabbed", which | |
526 | * makes sure updates are visible before bind or setsockopt return. */ | |
527 | } | |
528 | ||
8ea65f4a | 529 | static int netlink_insert(struct sock *sk, u32 portid) |
1da177e4 | 530 | { |
da12c90e | 531 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
919d9db9 | 532 | int err; |
1da177e4 | 533 | |
c5adde94 | 534 | lock_sock(sk); |
1da177e4 | 535 | |
da314c99 HX |
536 | err = nlk_sk(sk)->portid == portid ? 0 : -EBUSY; |
537 | if (nlk_sk(sk)->bound) | |
1da177e4 LT |
538 | goto err; |
539 | ||
540 | err = -ENOMEM; | |
97defe1e TG |
541 | if (BITS_PER_LONG > 32 && |
542 | unlikely(atomic_read(&table->hash.nelems) >= UINT_MAX)) | |
1da177e4 LT |
543 | goto err; |
544 | ||
da314c99 | 545 | nlk_sk(sk)->portid = portid; |
e341694e | 546 | sock_hold(sk); |
919d9db9 | 547 | |
c428ecd1 HX |
548 | err = __netlink_insert(table, sk); |
549 | if (err) { | |
4e7c1330 DB |
550 | /* In case the hashtable backend returns with -EBUSY |
551 | * from here, it must not escape to the caller. | |
552 | */ | |
553 | if (unlikely(err == -EBUSY)) | |
554 | err = -EOVERFLOW; | |
c428ecd1 HX |
555 | if (err == -EEXIST) |
556 | err = -EADDRINUSE; | |
c5adde94 | 557 | sock_put(sk); |
1f770c0a | 558 | goto err; |
919d9db9 HX |
559 | } |
560 | ||
da314c99 HX |
561 | /* We need to ensure that the socket is hashed and visible. */ |
562 | smp_wmb(); | |
563 | nlk_sk(sk)->bound = portid; | |
1f770c0a | 564 | |
1da177e4 | 565 | err: |
c5adde94 | 566 | release_sock(sk); |
1da177e4 LT |
567 | return err; |
568 | } | |
569 | ||
570 | static void netlink_remove(struct sock *sk) | |
571 | { | |
e341694e TG |
572 | struct netlink_table *table; |
573 | ||
e341694e | 574 | table = &nl_table[sk->sk_protocol]; |
c428ecd1 HX |
575 | if (!rhashtable_remove_fast(&table->hash, &nlk_sk(sk)->node, |
576 | netlink_rhashtable_params)) { | |
e341694e TG |
577 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
578 | __sock_put(sk); | |
579 | } | |
e341694e | 580 | |
1da177e4 | 581 | netlink_table_grab(); |
b10dcb3b | 582 | if (nlk_sk(sk)->subscriptions) { |
1da177e4 | 583 | __sk_del_bind_node(sk); |
b10dcb3b JB |
584 | netlink_update_listeners(sk); |
585 | } | |
ee1c2442 JB |
586 | if (sk->sk_protocol == NETLINK_GENERIC) |
587 | atomic_inc(&genl_sk_destructing_cnt); | |
1da177e4 LT |
588 | netlink_table_ungrab(); |
589 | } | |
590 | ||
591 | static struct proto netlink_proto = { | |
592 | .name = "NETLINK", | |
593 | .owner = THIS_MODULE, | |
594 | .obj_size = sizeof(struct netlink_sock), | |
595 | }; | |
596 | ||
1b8d7ae4 | 597 | static int __netlink_create(struct net *net, struct socket *sock, |
11aa9c28 EB |
598 | struct mutex *cb_mutex, int protocol, |
599 | int kern) | |
1da177e4 LT |
600 | { |
601 | struct sock *sk; | |
602 | struct netlink_sock *nlk; | |
ab33a171 PM |
603 | |
604 | sock->ops = &netlink_ops; | |
605 | ||
11aa9c28 | 606 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto, kern); |
ab33a171 PM |
607 | if (!sk) |
608 | return -ENOMEM; | |
609 | ||
610 | sock_init_data(sock, sk); | |
611 | ||
612 | nlk = nlk_sk(sk); | |
658cb354 | 613 | if (cb_mutex) { |
ffa4d721 | 614 | nlk->cb_mutex = cb_mutex; |
658cb354 | 615 | } else { |
ffa4d721 PM |
616 | nlk->cb_mutex = &nlk->cb_def_mutex; |
617 | mutex_init(nlk->cb_mutex); | |
8a0f5ccf HX |
618 | lockdep_set_class_and_name(nlk->cb_mutex, |
619 | nlk_cb_mutex_keys + protocol, | |
620 | nlk_cb_mutex_key_strings[protocol]); | |
ffa4d721 | 621 | } |
ab33a171 PM |
622 | init_waitqueue_head(&nlk->wait); |
623 | ||
624 | sk->sk_destruct = netlink_sock_destruct; | |
625 | sk->sk_protocol = protocol; | |
626 | return 0; | |
627 | } | |
628 | ||
3f378b68 EP |
629 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
630 | int kern) | |
ab33a171 PM |
631 | { |
632 | struct module *module = NULL; | |
af65bdfc | 633 | struct mutex *cb_mutex; |
f7fa9b10 | 634 | struct netlink_sock *nlk; |
023e2cfa JB |
635 | int (*bind)(struct net *net, int group); |
636 | void (*unbind)(struct net *net, int group); | |
ab33a171 | 637 | int err = 0; |
1da177e4 LT |
638 | |
639 | sock->state = SS_UNCONNECTED; | |
640 | ||
641 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
642 | return -ESOCKTNOSUPPORT; | |
643 | ||
6ac552fd | 644 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
645 | return -EPROTONOSUPPORT; |
646 | ||
77247bbb | 647 | netlink_lock_table(); |
95a5afca | 648 | #ifdef CONFIG_MODULES |
ab33a171 | 649 | if (!nl_table[protocol].registered) { |
77247bbb | 650 | netlink_unlock_table(); |
4fdb3bb7 | 651 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 652 | netlink_lock_table(); |
4fdb3bb7 | 653 | } |
ab33a171 PM |
654 | #endif |
655 | if (nl_table[protocol].registered && | |
656 | try_module_get(nl_table[protocol].module)) | |
657 | module = nl_table[protocol].module; | |
974c37e9 AD |
658 | else |
659 | err = -EPROTONOSUPPORT; | |
af65bdfc | 660 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 661 | bind = nl_table[protocol].bind; |
4f520900 | 662 | unbind = nl_table[protocol].unbind; |
77247bbb | 663 | netlink_unlock_table(); |
4fdb3bb7 | 664 | |
974c37e9 AD |
665 | if (err < 0) |
666 | goto out; | |
667 | ||
11aa9c28 | 668 | err = __netlink_create(net, sock, cb_mutex, protocol, kern); |
6ac552fd | 669 | if (err < 0) |
f7fa9b10 PM |
670 | goto out_module; |
671 | ||
6f756a8c | 672 | local_bh_disable(); |
c1fd3b94 | 673 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
674 | local_bh_enable(); |
675 | ||
f7fa9b10 | 676 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 677 | nlk->module = module; |
03292745 | 678 | nlk->netlink_bind = bind; |
4f520900 | 679 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
680 | out: |
681 | return err; | |
1da177e4 | 682 | |
ab33a171 PM |
683 | out_module: |
684 | module_put(module); | |
685 | goto out; | |
1da177e4 LT |
686 | } |
687 | ||
21e4902a TG |
688 | static void deferred_put_nlk_sk(struct rcu_head *head) |
689 | { | |
690 | struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu); | |
ed5d7788 HX |
691 | struct sock *sk = &nlk->sk; |
692 | ||
693 | if (!atomic_dec_and_test(&sk->sk_refcnt)) | |
694 | return; | |
695 | ||
696 | if (nlk->cb_running && nlk->cb.done) { | |
697 | INIT_WORK(&nlk->work, netlink_sock_destruct_work); | |
698 | schedule_work(&nlk->work); | |
699 | return; | |
700 | } | |
21e4902a | 701 | |
ed5d7788 | 702 | sk_free(sk); |
21e4902a TG |
703 | } |
704 | ||
1da177e4 LT |
705 | static int netlink_release(struct socket *sock) |
706 | { | |
707 | struct sock *sk = sock->sk; | |
708 | struct netlink_sock *nlk; | |
709 | ||
710 | if (!sk) | |
711 | return 0; | |
712 | ||
713 | netlink_remove(sk); | |
ac57b3a9 | 714 | sock_orphan(sk); |
1da177e4 LT |
715 | nlk = nlk_sk(sk); |
716 | ||
3f660d66 HX |
717 | /* |
718 | * OK. Socket is unlinked, any packets that arrive now | |
719 | * will be purged. | |
720 | */ | |
1da177e4 | 721 | |
ee1c2442 JB |
722 | /* must not acquire netlink_table_lock in any way again before unbind |
723 | * and notifying genetlink is done as otherwise it might deadlock | |
724 | */ | |
725 | if (nlk->netlink_unbind) { | |
726 | int i; | |
727 | ||
728 | for (i = 0; i < nlk->ngroups; i++) | |
729 | if (test_bit(i, nlk->groups)) | |
730 | nlk->netlink_unbind(sock_net(sk), i + 1); | |
731 | } | |
732 | if (sk->sk_protocol == NETLINK_GENERIC && | |
733 | atomic_dec_return(&genl_sk_destructing_cnt) == 0) | |
734 | wake_up(&genl_sk_destructing_waitq); | |
735 | ||
1da177e4 LT |
736 | sock->sk = NULL; |
737 | wake_up_interruptible_all(&nlk->wait); | |
738 | ||
739 | skb_queue_purge(&sk->sk_write_queue); | |
740 | ||
e2726020 | 741 | if (nlk->portid && nlk->bound) { |
1da177e4 | 742 | struct netlink_notify n = { |
3b1e0a65 | 743 | .net = sock_net(sk), |
1da177e4 | 744 | .protocol = sk->sk_protocol, |
15e47304 | 745 | .portid = nlk->portid, |
1da177e4 | 746 | }; |
efa172f4 | 747 | blocking_notifier_call_chain(&netlink_chain, |
e041c683 | 748 | NETLINK_URELEASE, &n); |
746fac4d | 749 | } |
4fdb3bb7 | 750 | |
5e7c001c | 751 | module_put(nlk->module); |
4fdb3bb7 | 752 | |
aed81560 | 753 | if (netlink_is_kernel(sk)) { |
b10dcb3b | 754 | netlink_table_grab(); |
869e58f8 DL |
755 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
756 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
757 | struct listeners *old; |
758 | ||
759 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
760 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
761 | kfree_rcu(old, rcu); | |
869e58f8 | 762 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 763 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 764 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 765 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
766 | nl_table[sk->sk_protocol].registered = 0; |
767 | } | |
b10dcb3b | 768 | netlink_table_ungrab(); |
658cb354 | 769 | } |
77247bbb | 770 | |
f7fa9b10 PM |
771 | kfree(nlk->groups); |
772 | nlk->groups = NULL; | |
773 | ||
3755810c | 774 | local_bh_disable(); |
c1fd3b94 | 775 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 776 | local_bh_enable(); |
21e4902a | 777 | call_rcu(&nlk->rcu, deferred_put_nlk_sk); |
1da177e4 LT |
778 | return 0; |
779 | } | |
780 | ||
781 | static int netlink_autobind(struct socket *sock) | |
782 | { | |
783 | struct sock *sk = sock->sk; | |
3b1e0a65 | 784 | struct net *net = sock_net(sk); |
da12c90e | 785 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
15e47304 | 786 | s32 portid = task_tgid_vnr(current); |
1da177e4 | 787 | int err; |
b9fbe709 HX |
788 | s32 rover = -4096; |
789 | bool ok; | |
1da177e4 LT |
790 | |
791 | retry: | |
792 | cond_resched(); | |
e341694e | 793 | rcu_read_lock(); |
b9fbe709 HX |
794 | ok = !__netlink_lookup(table, portid, net); |
795 | rcu_read_unlock(); | |
796 | if (!ok) { | |
e341694e | 797 | /* Bind collision, search negative portid values. */ |
b9fbe709 HX |
798 | if (rover == -4096) |
799 | /* rover will be in range [S32_MIN, -4097] */ | |
800 | rover = S32_MIN + prandom_u32_max(-4096 - S32_MIN); | |
801 | else if (rover >= -4096) | |
e341694e | 802 | rover = -4097; |
b9fbe709 | 803 | portid = rover--; |
e341694e | 804 | goto retry; |
1da177e4 | 805 | } |
1da177e4 | 806 | |
8ea65f4a | 807 | err = netlink_insert(sk, portid); |
1da177e4 LT |
808 | if (err == -EADDRINUSE) |
809 | goto retry; | |
d470e3b4 DM |
810 | |
811 | /* If 2 threads race to autobind, that is fine. */ | |
812 | if (err == -EBUSY) | |
813 | err = 0; | |
814 | ||
815 | return err; | |
1da177e4 LT |
816 | } |
817 | ||
aa4cf945 EB |
818 | /** |
819 | * __netlink_ns_capable - General netlink message capability test | |
820 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
821 | * @user_ns: The user namespace of the capability to use | |
822 | * @cap: The capability to use | |
823 | * | |
824 | * Test to see if the opener of the socket we received the message | |
825 | * from had when the netlink socket was created and the sender of the | |
826 | * message has has the capability @cap in the user namespace @user_ns. | |
827 | */ | |
828 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
829 | struct user_namespace *user_ns, int cap) | |
830 | { | |
2d7a85f4 EB |
831 | return ((nsp->flags & NETLINK_SKB_DST) || |
832 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
833 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
834 | } |
835 | EXPORT_SYMBOL(__netlink_ns_capable); | |
836 | ||
837 | /** | |
838 | * netlink_ns_capable - General netlink message capability test | |
839 | * @skb: socket buffer holding a netlink command from userspace | |
840 | * @user_ns: The user namespace of the capability to use | |
841 | * @cap: The capability to use | |
842 | * | |
843 | * Test to see if the opener of the socket we received the message | |
844 | * from had when the netlink socket was created and the sender of the | |
845 | * message has has the capability @cap in the user namespace @user_ns. | |
846 | */ | |
847 | bool netlink_ns_capable(const struct sk_buff *skb, | |
848 | struct user_namespace *user_ns, int cap) | |
849 | { | |
850 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
851 | } | |
852 | EXPORT_SYMBOL(netlink_ns_capable); | |
853 | ||
854 | /** | |
855 | * netlink_capable - Netlink global message capability test | |
856 | * @skb: socket buffer holding a netlink command from userspace | |
857 | * @cap: The capability to use | |
858 | * | |
859 | * Test to see if the opener of the socket we received the message | |
860 | * from had when the netlink socket was created and the sender of the | |
861 | * message has has the capability @cap in all user namespaces. | |
862 | */ | |
863 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
864 | { | |
865 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
866 | } | |
867 | EXPORT_SYMBOL(netlink_capable); | |
868 | ||
869 | /** | |
870 | * netlink_net_capable - Netlink network namespace message capability test | |
871 | * @skb: socket buffer holding a netlink command from userspace | |
872 | * @cap: The capability to use | |
873 | * | |
874 | * Test to see if the opener of the socket we received the message | |
875 | * from had when the netlink socket was created and the sender of the | |
876 | * message has has the capability @cap over the network namespace of | |
877 | * the socket we received the message from. | |
878 | */ | |
879 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
880 | { | |
881 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
882 | } | |
883 | EXPORT_SYMBOL(netlink_net_capable); | |
884 | ||
5187cd05 | 885 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 886 | { |
9785e10a | 887 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 888 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 889 | } |
1da177e4 | 890 | |
f7fa9b10 PM |
891 | static void |
892 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
893 | { | |
894 | struct netlink_sock *nlk = nlk_sk(sk); | |
895 | ||
896 | if (nlk->subscriptions && !subscriptions) | |
897 | __sk_del_bind_node(sk); | |
898 | else if (!nlk->subscriptions && subscriptions) | |
899 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
900 | nlk->subscriptions = subscriptions; | |
901 | } | |
902 | ||
b4ff4f04 | 903 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
904 | { |
905 | struct netlink_sock *nlk = nlk_sk(sk); | |
906 | unsigned int groups; | |
b4ff4f04 | 907 | unsigned long *new_groups; |
513c2500 PM |
908 | int err = 0; |
909 | ||
b4ff4f04 JB |
910 | netlink_table_grab(); |
911 | ||
513c2500 | 912 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 913 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 914 | err = -ENOENT; |
b4ff4f04 JB |
915 | goto out_unlock; |
916 | } | |
513c2500 | 917 | |
b4ff4f04 JB |
918 | if (nlk->ngroups >= groups) |
919 | goto out_unlock; | |
513c2500 | 920 | |
b4ff4f04 JB |
921 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
922 | if (new_groups == NULL) { | |
923 | err = -ENOMEM; | |
924 | goto out_unlock; | |
925 | } | |
6ac552fd | 926 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
927 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
928 | ||
929 | nlk->groups = new_groups; | |
513c2500 | 930 | nlk->ngroups = groups; |
b4ff4f04 JB |
931 | out_unlock: |
932 | netlink_table_ungrab(); | |
933 | return err; | |
513c2500 PM |
934 | } |
935 | ||
02c81ab9 | 936 | static void netlink_undo_bind(int group, long unsigned int groups, |
023e2cfa | 937 | struct sock *sk) |
4f520900 | 938 | { |
023e2cfa | 939 | struct netlink_sock *nlk = nlk_sk(sk); |
4f520900 RGB |
940 | int undo; |
941 | ||
942 | if (!nlk->netlink_unbind) | |
943 | return; | |
944 | ||
945 | for (undo = 0; undo < group; undo++) | |
6251edd9 | 946 | if (test_bit(undo, &groups)) |
8b7c36d8 | 947 | nlk->netlink_unbind(sock_net(sk), undo + 1); |
4f520900 RGB |
948 | } |
949 | ||
6ac552fd PM |
950 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
951 | int addr_len) | |
1da177e4 LT |
952 | { |
953 | struct sock *sk = sock->sk; | |
3b1e0a65 | 954 | struct net *net = sock_net(sk); |
1da177e4 LT |
955 | struct netlink_sock *nlk = nlk_sk(sk); |
956 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
957 | int err; | |
4f520900 | 958 | long unsigned int groups = nladdr->nl_groups; |
da314c99 | 959 | bool bound; |
746fac4d | 960 | |
4e4b5376 HFS |
961 | if (addr_len < sizeof(struct sockaddr_nl)) |
962 | return -EINVAL; | |
963 | ||
1da177e4 LT |
964 | if (nladdr->nl_family != AF_NETLINK) |
965 | return -EINVAL; | |
966 | ||
967 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 968 | if (groups) { |
5187cd05 | 969 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 970 | return -EPERM; |
b4ff4f04 JB |
971 | err = netlink_realloc_groups(sk); |
972 | if (err) | |
973 | return err; | |
513c2500 | 974 | } |
1da177e4 | 975 | |
da314c99 HX |
976 | bound = nlk->bound; |
977 | if (bound) { | |
978 | /* Ensure nlk->portid is up-to-date. */ | |
979 | smp_rmb(); | |
980 | ||
15e47304 | 981 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 982 | return -EINVAL; |
da314c99 | 983 | } |
4f520900 RGB |
984 | |
985 | if (nlk->netlink_bind && groups) { | |
986 | int group; | |
987 | ||
988 | for (group = 0; group < nlk->ngroups; group++) { | |
989 | if (!test_bit(group, &groups)) | |
990 | continue; | |
8b7c36d8 | 991 | err = nlk->netlink_bind(net, group + 1); |
4f520900 RGB |
992 | if (!err) |
993 | continue; | |
023e2cfa | 994 | netlink_undo_bind(group, groups, sk); |
4f520900 RGB |
995 | return err; |
996 | } | |
997 | } | |
998 | ||
da314c99 HX |
999 | /* No need for barriers here as we return to user-space without |
1000 | * using any of the bound attributes. | |
1001 | */ | |
1002 | if (!bound) { | |
1da177e4 | 1003 | err = nladdr->nl_pid ? |
8ea65f4a | 1004 | netlink_insert(sk, nladdr->nl_pid) : |
1da177e4 | 1005 | netlink_autobind(sock); |
4f520900 | 1006 | if (err) { |
023e2cfa | 1007 | netlink_undo_bind(nlk->ngroups, groups, sk); |
1da177e4 | 1008 | return err; |
4f520900 | 1009 | } |
1da177e4 LT |
1010 | } |
1011 | ||
4f520900 | 1012 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1013 | return 0; |
1014 | ||
1015 | netlink_table_grab(); | |
f7fa9b10 | 1016 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1017 | hweight32(groups) - |
746fac4d | 1018 | hweight32(nlk->groups[0])); |
4f520900 | 1019 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1020 | netlink_update_listeners(sk); |
1da177e4 LT |
1021 | netlink_table_ungrab(); |
1022 | ||
1023 | return 0; | |
1024 | } | |
1025 | ||
1026 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1027 | int alen, int flags) | |
1028 | { | |
1029 | int err = 0; | |
1030 | struct sock *sk = sock->sk; | |
1031 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1032 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1033 | |
6503d961 CG |
1034 | if (alen < sizeof(addr->sa_family)) |
1035 | return -EINVAL; | |
1036 | ||
1da177e4 LT |
1037 | if (addr->sa_family == AF_UNSPEC) { |
1038 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1039 | nlk->dst_portid = 0; |
d629b836 | 1040 | nlk->dst_group = 0; |
1da177e4 LT |
1041 | return 0; |
1042 | } | |
1043 | if (addr->sa_family != AF_NETLINK) | |
1044 | return -EINVAL; | |
1045 | ||
46833a86 | 1046 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1047 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1048 | return -EPERM; |
1049 | ||
da314c99 HX |
1050 | /* No need for barriers here as we return to user-space without |
1051 | * using any of the bound attributes. | |
1052 | */ | |
1053 | if (!nlk->bound) | |
1da177e4 LT |
1054 | err = netlink_autobind(sock); |
1055 | ||
1056 | if (err == 0) { | |
1057 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1058 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1059 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1060 | } |
1061 | ||
1062 | return err; | |
1063 | } | |
1064 | ||
6ac552fd PM |
1065 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1066 | int *addr_len, int peer) | |
1da177e4 LT |
1067 | { |
1068 | struct sock *sk = sock->sk; | |
1069 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1070 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1071 | |
1da177e4 LT |
1072 | nladdr->nl_family = AF_NETLINK; |
1073 | nladdr->nl_pad = 0; | |
1074 | *addr_len = sizeof(*nladdr); | |
1075 | ||
1076 | if (peer) { | |
15e47304 | 1077 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1078 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1079 | } else { |
15e47304 | 1080 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1081 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1082 | } |
1083 | return 0; | |
1084 | } | |
1085 | ||
025c6818 DD |
1086 | static int netlink_ioctl(struct socket *sock, unsigned int cmd, |
1087 | unsigned long arg) | |
1088 | { | |
1089 | /* try to hand this ioctl down to the NIC drivers. | |
1090 | */ | |
1091 | return -ENOIOCTLCMD; | |
1092 | } | |
1093 | ||
15e47304 | 1094 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1095 | { |
1da177e4 LT |
1096 | struct sock *sock; |
1097 | struct netlink_sock *nlk; | |
1098 | ||
15e47304 | 1099 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1100 | if (!sock) |
1101 | return ERR_PTR(-ECONNREFUSED); | |
1102 | ||
1103 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1104 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1105 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1106 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1107 | sock_put(sock); |
1108 | return ERR_PTR(-ECONNREFUSED); | |
1109 | } | |
1110 | return sock; | |
1111 | } | |
1112 | ||
1113 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1114 | { | |
496ad9aa | 1115 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1116 | struct sock *sock; |
1117 | ||
1118 | if (!S_ISSOCK(inode->i_mode)) | |
1119 | return ERR_PTR(-ENOTSOCK); | |
1120 | ||
1121 | sock = SOCKET_I(inode)->sk; | |
1122 | if (sock->sk_family != AF_NETLINK) | |
1123 | return ERR_PTR(-EINVAL); | |
1124 | ||
1125 | sock_hold(sock); | |
1126 | return sock; | |
1127 | } | |
1128 | ||
3a36515f PN |
1129 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1130 | int broadcast) | |
c05cdb1b PNA |
1131 | { |
1132 | struct sk_buff *skb; | |
1133 | void *data; | |
1134 | ||
3a36515f | 1135 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1136 | return alloc_skb(size, GFP_KERNEL); |
1137 | ||
3a36515f PN |
1138 | size = SKB_DATA_ALIGN(size) + |
1139 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1140 | |
1141 | data = vmalloc(size); | |
1142 | if (data == NULL) | |
3a36515f | 1143 | return NULL; |
c05cdb1b | 1144 | |
2ea2f62c | 1145 | skb = __build_skb(data, size); |
3a36515f PN |
1146 | if (skb == NULL) |
1147 | vfree(data); | |
2ea2f62c | 1148 | else |
3a36515f | 1149 | skb->destructor = netlink_skb_destructor; |
c05cdb1b PNA |
1150 | |
1151 | return skb; | |
c05cdb1b PNA |
1152 | } |
1153 | ||
1da177e4 LT |
1154 | /* |
1155 | * Attach a skb to a netlink socket. | |
1156 | * The caller must hold a reference to the destination socket. On error, the | |
1157 | * reference is dropped. The skb is not send to the destination, just all | |
1158 | * all error checks are performed and memory in the queue is reserved. | |
1159 | * Return values: | |
1160 | * < 0: error. skb freed, reference to sock dropped. | |
1161 | * 0: continue | |
1162 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1163 | */ | |
9457afee | 1164 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1165 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1166 | { |
1167 | struct netlink_sock *nlk; | |
1168 | ||
1169 | nlk = nlk_sk(sk); | |
1170 | ||
5fd96123 | 1171 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
d1b4c689 | 1172 | test_bit(NETLINK_S_CONGESTED, &nlk->state))) { |
1da177e4 | 1173 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1174 | if (!*timeo) { |
aed81560 | 1175 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1176 | netlink_overrun(sk); |
1177 | sock_put(sk); | |
1178 | kfree_skb(skb); | |
1179 | return -EAGAIN; | |
1180 | } | |
1181 | ||
1182 | __set_current_state(TASK_INTERRUPTIBLE); | |
1183 | add_wait_queue(&nlk->wait, &wait); | |
1184 | ||
1185 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cc3a572f | 1186 | test_bit(NETLINK_S_CONGESTED, &nlk->state)) && |
1da177e4 | 1187 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1188 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1189 | |
1190 | __set_current_state(TASK_RUNNING); | |
1191 | remove_wait_queue(&nlk->wait, &wait); | |
1192 | sock_put(sk); | |
1193 | ||
1194 | if (signal_pending(current)) { | |
1195 | kfree_skb(skb); | |
c3d8d1e3 | 1196 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1197 | } |
1198 | return 1; | |
1199 | } | |
cf0a018a | 1200 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1201 | return 0; |
1202 | } | |
1203 | ||
4a7e7c2a | 1204 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1205 | { |
1da177e4 LT |
1206 | int len = skb->len; |
1207 | ||
bcbde0d4 DB |
1208 | netlink_deliver_tap(skb); |
1209 | ||
d1b4c689 | 1210 | skb_queue_tail(&sk->sk_receive_queue, skb); |
676d2369 | 1211 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1212 | return len; |
1213 | } | |
1214 | ||
1215 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1216 | { | |
1217 | int len = __netlink_sendskb(sk, skb); | |
1218 | ||
1da177e4 LT |
1219 | sock_put(sk); |
1220 | return len; | |
1221 | } | |
1222 | ||
1223 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1224 | { | |
1225 | kfree_skb(skb); | |
1226 | sock_put(sk); | |
1227 | } | |
1228 | ||
b57ef81f | 1229 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1230 | { |
1231 | int delta; | |
1232 | ||
1298ca46 | 1233 | WARN_ON(skb->sk != NULL); |
4305b541 | 1234 | delta = skb->end - skb->tail; |
c05cdb1b | 1235 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1236 | return skb; |
1237 | ||
1238 | if (skb_shared(skb)) { | |
1239 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1240 | if (!nskb) | |
1241 | return skb; | |
8460c00f | 1242 | consume_skb(skb); |
1da177e4 LT |
1243 | skb = nskb; |
1244 | } | |
1245 | ||
158f323b ED |
1246 | pskb_expand_head(skb, 0, -delta, |
1247 | (allocation & ~__GFP_DIRECT_RECLAIM) | | |
1248 | __GFP_NOWARN | __GFP_NORETRY); | |
1da177e4 LT |
1249 | return skb; |
1250 | } | |
1251 | ||
3fbc2905 EB |
1252 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1253 | struct sock *ssk) | |
cd40b7d3 DL |
1254 | { |
1255 | int ret; | |
1256 | struct netlink_sock *nlk = nlk_sk(sk); | |
1257 | ||
1258 | ret = -ECONNREFUSED; | |
1259 | if (nlk->netlink_rcv != NULL) { | |
1260 | ret = skb->len; | |
cf0a018a | 1261 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1262 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1263 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1264 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1265 | consume_skb(skb); |
1266 | } else { | |
1267 | kfree_skb(skb); | |
cd40b7d3 | 1268 | } |
cd40b7d3 DL |
1269 | sock_put(sk); |
1270 | return ret; | |
1271 | } | |
1272 | ||
1273 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1274 | u32 portid, int nonblock) |
1da177e4 LT |
1275 | { |
1276 | struct sock *sk; | |
1277 | int err; | |
1278 | long timeo; | |
1279 | ||
1280 | skb = netlink_trim(skb, gfp_any()); | |
1281 | ||
1282 | timeo = sock_sndtimeo(ssk, nonblock); | |
1283 | retry: | |
15e47304 | 1284 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1285 | if (IS_ERR(sk)) { |
1286 | kfree_skb(skb); | |
1287 | return PTR_ERR(sk); | |
1288 | } | |
cd40b7d3 | 1289 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1290 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1291 | |
b1153f29 | 1292 | if (sk_filter(sk, skb)) { |
84874607 | 1293 | err = skb->len; |
b1153f29 SH |
1294 | kfree_skb(skb); |
1295 | sock_put(sk); | |
1296 | return err; | |
1297 | } | |
1298 | ||
9457afee | 1299 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1300 | if (err == 1) |
1301 | goto retry; | |
1302 | if (err) | |
1303 | return err; | |
1304 | ||
7ee015e0 | 1305 | return netlink_sendskb(sk, skb); |
1da177e4 | 1306 | } |
6ac552fd | 1307 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1308 | |
4277a083 PM |
1309 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1310 | { | |
1311 | int res = 0; | |
5c398dc8 | 1312 | struct listeners *listeners; |
4277a083 | 1313 | |
aed81560 | 1314 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1315 | |
1316 | rcu_read_lock(); | |
1317 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1318 | ||
6d772ac5 | 1319 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1320 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1321 | |
1322 | rcu_read_unlock(); | |
1323 | ||
4277a083 PM |
1324 | return res; |
1325 | } | |
1326 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1327 | ||
b57ef81f | 1328 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1329 | { |
1330 | struct netlink_sock *nlk = nlk_sk(sk); | |
1331 | ||
1332 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cc3a572f | 1333 | !test_bit(NETLINK_S_CONGESTED, &nlk->state)) { |
cf0a018a | 1334 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1335 | __netlink_sendskb(sk, skb); |
2c645800 | 1336 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1337 | } |
1338 | return -1; | |
1339 | } | |
1340 | ||
1341 | struct netlink_broadcast_data { | |
1342 | struct sock *exclude_sk; | |
b4b51029 | 1343 | struct net *net; |
15e47304 | 1344 | u32 portid; |
1da177e4 LT |
1345 | u32 group; |
1346 | int failure; | |
ff491a73 | 1347 | int delivery_failure; |
1da177e4 LT |
1348 | int congested; |
1349 | int delivered; | |
7d877f3b | 1350 | gfp_t allocation; |
1da177e4 | 1351 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1352 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1353 | void *tx_data; | |
1da177e4 LT |
1354 | }; |
1355 | ||
46c9521f RR |
1356 | static void do_one_broadcast(struct sock *sk, |
1357 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1358 | { |
1359 | struct netlink_sock *nlk = nlk_sk(sk); | |
1360 | int val; | |
1361 | ||
1362 | if (p->exclude_sk == sk) | |
46c9521f | 1363 | return; |
1da177e4 | 1364 | |
15e47304 | 1365 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1366 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1367 | return; |
1da177e4 | 1368 | |
59324cf3 ND |
1369 | if (!net_eq(sock_net(sk), p->net)) { |
1370 | if (!(nlk->flags & NETLINK_F_LISTEN_ALL_NSID)) | |
1371 | return; | |
1372 | ||
1373 | if (!peernet_has_id(sock_net(sk), p->net)) | |
1374 | return; | |
1375 | ||
1376 | if (!file_ns_capable(sk->sk_socket->file, p->net->user_ns, | |
1377 | CAP_NET_BROADCAST)) | |
1378 | return; | |
1379 | } | |
b4b51029 | 1380 | |
1da177e4 LT |
1381 | if (p->failure) { |
1382 | netlink_overrun(sk); | |
46c9521f | 1383 | return; |
1da177e4 LT |
1384 | } |
1385 | ||
1386 | sock_hold(sk); | |
1387 | if (p->skb2 == NULL) { | |
68acc024 | 1388 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1389 | p->skb2 = skb_clone(p->skb, p->allocation); |
1390 | } else { | |
68acc024 TC |
1391 | p->skb2 = skb_get(p->skb); |
1392 | /* | |
1393 | * skb ownership may have been set when | |
1394 | * delivered to a previous socket. | |
1395 | */ | |
1396 | skb_orphan(p->skb2); | |
1da177e4 LT |
1397 | } |
1398 | } | |
1399 | if (p->skb2 == NULL) { | |
1400 | netlink_overrun(sk); | |
1401 | /* Clone failed. Notify ALL listeners. */ | |
1402 | p->failure = 1; | |
cc3a572f | 1403 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 1404 | p->delivery_failure = 1; |
59324cf3 ND |
1405 | goto out; |
1406 | } | |
1407 | if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { | |
910a7e90 EB |
1408 | kfree_skb(p->skb2); |
1409 | p->skb2 = NULL; | |
59324cf3 ND |
1410 | goto out; |
1411 | } | |
1412 | if (sk_filter(sk, p->skb2)) { | |
b1153f29 SH |
1413 | kfree_skb(p->skb2); |
1414 | p->skb2 = NULL; | |
59324cf3 ND |
1415 | goto out; |
1416 | } | |
1417 | NETLINK_CB(p->skb2).nsid = peernet2id(sock_net(sk), p->net); | |
1418 | NETLINK_CB(p->skb2).nsid_is_set = true; | |
1419 | val = netlink_broadcast_deliver(sk, p->skb2); | |
1420 | if (val < 0) { | |
1da177e4 | 1421 | netlink_overrun(sk); |
cc3a572f | 1422 | if (nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR) |
be0c22a4 | 1423 | p->delivery_failure = 1; |
1da177e4 LT |
1424 | } else { |
1425 | p->congested |= val; | |
1426 | p->delivered = 1; | |
1427 | p->skb2 = NULL; | |
1428 | } | |
59324cf3 | 1429 | out: |
1da177e4 | 1430 | sock_put(sk); |
1da177e4 LT |
1431 | } |
1432 | ||
15e47304 | 1433 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1434 | u32 group, gfp_t allocation, |
1435 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1436 | void *filter_data) | |
1da177e4 | 1437 | { |
3b1e0a65 | 1438 | struct net *net = sock_net(ssk); |
1da177e4 | 1439 | struct netlink_broadcast_data info; |
1da177e4 LT |
1440 | struct sock *sk; |
1441 | ||
1442 | skb = netlink_trim(skb, allocation); | |
1443 | ||
1444 | info.exclude_sk = ssk; | |
b4b51029 | 1445 | info.net = net; |
15e47304 | 1446 | info.portid = portid; |
1da177e4 LT |
1447 | info.group = group; |
1448 | info.failure = 0; | |
ff491a73 | 1449 | info.delivery_failure = 0; |
1da177e4 LT |
1450 | info.congested = 0; |
1451 | info.delivered = 0; | |
1452 | info.allocation = allocation; | |
1453 | info.skb = skb; | |
1454 | info.skb2 = NULL; | |
910a7e90 EB |
1455 | info.tx_filter = filter; |
1456 | info.tx_data = filter_data; | |
1da177e4 LT |
1457 | |
1458 | /* While we sleep in clone, do not allow to change socket list */ | |
1459 | ||
1460 | netlink_lock_table(); | |
1461 | ||
b67bfe0d | 1462 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1463 | do_one_broadcast(sk, &info); |
1464 | ||
70d4bf6d | 1465 | consume_skb(skb); |
aa1c6a6f | 1466 | |
1da177e4 LT |
1467 | netlink_unlock_table(); |
1468 | ||
70d4bf6d NH |
1469 | if (info.delivery_failure) { |
1470 | kfree_skb(info.skb2); | |
ff491a73 | 1471 | return -ENOBUFS; |
658cb354 ED |
1472 | } |
1473 | consume_skb(info.skb2); | |
ff491a73 | 1474 | |
1da177e4 | 1475 | if (info.delivered) { |
d0164adc | 1476 | if (info.congested && gfpflags_allow_blocking(allocation)) |
1da177e4 LT |
1477 | yield(); |
1478 | return 0; | |
1479 | } | |
1da177e4 LT |
1480 | return -ESRCH; |
1481 | } | |
910a7e90 EB |
1482 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
1483 | ||
15e47304 | 1484 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1485 | u32 group, gfp_t allocation) |
1486 | { | |
15e47304 | 1487 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
1488 | NULL, NULL); |
1489 | } | |
6ac552fd | 1490 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
1491 | |
1492 | struct netlink_set_err_data { | |
1493 | struct sock *exclude_sk; | |
15e47304 | 1494 | u32 portid; |
1da177e4 LT |
1495 | u32 group; |
1496 | int code; | |
1497 | }; | |
1498 | ||
b57ef81f | 1499 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
1500 | { |
1501 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 1502 | int ret = 0; |
1da177e4 LT |
1503 | |
1504 | if (sk == p->exclude_sk) | |
1505 | goto out; | |
1506 | ||
09ad9bc7 | 1507 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
1508 | goto out; |
1509 | ||
15e47304 | 1510 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1511 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
1512 | goto out; |
1513 | ||
cc3a572f | 1514 | if (p->code == ENOBUFS && nlk->flags & NETLINK_F_RECV_NO_ENOBUFS) { |
1a50307b PNA |
1515 | ret = 1; |
1516 | goto out; | |
1517 | } | |
1518 | ||
1da177e4 LT |
1519 | sk->sk_err = p->code; |
1520 | sk->sk_error_report(sk); | |
1521 | out: | |
1a50307b | 1522 | return ret; |
1da177e4 LT |
1523 | } |
1524 | ||
4843b93c PNA |
1525 | /** |
1526 | * netlink_set_err - report error to broadcast listeners | |
1527 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 1528 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 1529 | * @group: the broadcast group that will notice the error |
4843b93c | 1530 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
1531 | * |
1532 | * This function returns the number of broadcast listeners that have set the | |
cc3a572f | 1533 | * NETLINK_NO_ENOBUFS socket option. |
4843b93c | 1534 | */ |
15e47304 | 1535 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
1536 | { |
1537 | struct netlink_set_err_data info; | |
1da177e4 | 1538 | struct sock *sk; |
1a50307b | 1539 | int ret = 0; |
1da177e4 LT |
1540 | |
1541 | info.exclude_sk = ssk; | |
15e47304 | 1542 | info.portid = portid; |
1da177e4 | 1543 | info.group = group; |
4843b93c PNA |
1544 | /* sk->sk_err wants a positive error value */ |
1545 | info.code = -code; | |
1da177e4 LT |
1546 | |
1547 | read_lock(&nl_table_lock); | |
1548 | ||
b67bfe0d | 1549 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 1550 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
1551 | |
1552 | read_unlock(&nl_table_lock); | |
1a50307b | 1553 | return ret; |
1da177e4 | 1554 | } |
dd5b6ce6 | 1555 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 1556 | |
84659eb5 JB |
1557 | /* must be called with netlink table grabbed */ |
1558 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
1559 | unsigned int group, | |
1560 | int is_new) | |
1561 | { | |
1562 | int old, new = !!is_new, subscriptions; | |
1563 | ||
1564 | old = test_bit(group - 1, nlk->groups); | |
1565 | subscriptions = nlk->subscriptions - old + new; | |
1566 | if (new) | |
1567 | __set_bit(group - 1, nlk->groups); | |
1568 | else | |
1569 | __clear_bit(group - 1, nlk->groups); | |
1570 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
1571 | netlink_update_listeners(&nlk->sk); | |
1572 | } | |
1573 | ||
9a4595bc | 1574 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 1575 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
1576 | { |
1577 | struct sock *sk = sock->sk; | |
1578 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
1579 | unsigned int val = 0; |
1580 | int err; | |
9a4595bc PM |
1581 | |
1582 | if (level != SOL_NETLINK) | |
1583 | return -ENOPROTOOPT; | |
1584 | ||
d1b4c689 | 1585 | if (optlen >= sizeof(int) && |
eb496534 | 1586 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
1587 | return -EFAULT; |
1588 | ||
1589 | switch (optname) { | |
1590 | case NETLINK_PKTINFO: | |
1591 | if (val) | |
cc3a572f | 1592 | nlk->flags |= NETLINK_F_RECV_PKTINFO; |
9a4595bc | 1593 | else |
cc3a572f | 1594 | nlk->flags &= ~NETLINK_F_RECV_PKTINFO; |
9a4595bc PM |
1595 | err = 0; |
1596 | break; | |
1597 | case NETLINK_ADD_MEMBERSHIP: | |
1598 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 1599 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 1600 | return -EPERM; |
b4ff4f04 JB |
1601 | err = netlink_realloc_groups(sk); |
1602 | if (err) | |
1603 | return err; | |
9a4595bc PM |
1604 | if (!val || val - 1 >= nlk->ngroups) |
1605 | return -EINVAL; | |
7774d5e0 | 1606 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
023e2cfa | 1607 | err = nlk->netlink_bind(sock_net(sk), val); |
4f520900 RGB |
1608 | if (err) |
1609 | return err; | |
1610 | } | |
9a4595bc | 1611 | netlink_table_grab(); |
84659eb5 JB |
1612 | netlink_update_socket_mc(nlk, val, |
1613 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 1614 | netlink_table_ungrab(); |
7774d5e0 | 1615 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
023e2cfa | 1616 | nlk->netlink_unbind(sock_net(sk), val); |
03292745 | 1617 | |
9a4595bc PM |
1618 | err = 0; |
1619 | break; | |
1620 | } | |
be0c22a4 PNA |
1621 | case NETLINK_BROADCAST_ERROR: |
1622 | if (val) | |
cc3a572f | 1623 | nlk->flags |= NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 | 1624 | else |
cc3a572f | 1625 | nlk->flags &= ~NETLINK_F_BROADCAST_SEND_ERROR; |
be0c22a4 PNA |
1626 | err = 0; |
1627 | break; | |
38938bfe PNA |
1628 | case NETLINK_NO_ENOBUFS: |
1629 | if (val) { | |
cc3a572f ND |
1630 | nlk->flags |= NETLINK_F_RECV_NO_ENOBUFS; |
1631 | clear_bit(NETLINK_S_CONGESTED, &nlk->state); | |
38938bfe | 1632 | wake_up_interruptible(&nlk->wait); |
658cb354 | 1633 | } else { |
cc3a572f | 1634 | nlk->flags &= ~NETLINK_F_RECV_NO_ENOBUFS; |
658cb354 | 1635 | } |
38938bfe PNA |
1636 | err = 0; |
1637 | break; | |
59324cf3 ND |
1638 | case NETLINK_LISTEN_ALL_NSID: |
1639 | if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_BROADCAST)) | |
1640 | return -EPERM; | |
1641 | ||
1642 | if (val) | |
1643 | nlk->flags |= NETLINK_F_LISTEN_ALL_NSID; | |
1644 | else | |
1645 | nlk->flags &= ~NETLINK_F_LISTEN_ALL_NSID; | |
1646 | err = 0; | |
1647 | break; | |
0a6a3a23 CR |
1648 | case NETLINK_CAP_ACK: |
1649 | if (val) | |
1650 | nlk->flags |= NETLINK_F_CAP_ACK; | |
1651 | else | |
1652 | nlk->flags &= ~NETLINK_F_CAP_ACK; | |
1653 | err = 0; | |
1654 | break; | |
2d4bc933 JB |
1655 | case NETLINK_EXT_ACK: |
1656 | if (val) | |
1657 | nlk->flags |= NETLINK_F_EXT_ACK; | |
1658 | else | |
1659 | nlk->flags &= ~NETLINK_F_EXT_ACK; | |
1660 | err = 0; | |
1661 | break; | |
9a4595bc PM |
1662 | default: |
1663 | err = -ENOPROTOOPT; | |
1664 | } | |
1665 | return err; | |
1666 | } | |
1667 | ||
1668 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 1669 | char __user *optval, int __user *optlen) |
9a4595bc PM |
1670 | { |
1671 | struct sock *sk = sock->sk; | |
1672 | struct netlink_sock *nlk = nlk_sk(sk); | |
1673 | int len, val, err; | |
1674 | ||
1675 | if (level != SOL_NETLINK) | |
1676 | return -ENOPROTOOPT; | |
1677 | ||
1678 | if (get_user(len, optlen)) | |
1679 | return -EFAULT; | |
1680 | if (len < 0) | |
1681 | return -EINVAL; | |
1682 | ||
1683 | switch (optname) { | |
1684 | case NETLINK_PKTINFO: | |
1685 | if (len < sizeof(int)) | |
1686 | return -EINVAL; | |
1687 | len = sizeof(int); | |
cc3a572f | 1688 | val = nlk->flags & NETLINK_F_RECV_PKTINFO ? 1 : 0; |
a27b58fe HC |
1689 | if (put_user(len, optlen) || |
1690 | put_user(val, optval)) | |
1691 | return -EFAULT; | |
9a4595bc PM |
1692 | err = 0; |
1693 | break; | |
be0c22a4 PNA |
1694 | case NETLINK_BROADCAST_ERROR: |
1695 | if (len < sizeof(int)) | |
1696 | return -EINVAL; | |
1697 | len = sizeof(int); | |
cc3a572f | 1698 | val = nlk->flags & NETLINK_F_BROADCAST_SEND_ERROR ? 1 : 0; |
be0c22a4 PNA |
1699 | if (put_user(len, optlen) || |
1700 | put_user(val, optval)) | |
1701 | return -EFAULT; | |
1702 | err = 0; | |
1703 | break; | |
38938bfe PNA |
1704 | case NETLINK_NO_ENOBUFS: |
1705 | if (len < sizeof(int)) | |
1706 | return -EINVAL; | |
1707 | len = sizeof(int); | |
cc3a572f | 1708 | val = nlk->flags & NETLINK_F_RECV_NO_ENOBUFS ? 1 : 0; |
38938bfe PNA |
1709 | if (put_user(len, optlen) || |
1710 | put_user(val, optval)) | |
1711 | return -EFAULT; | |
1712 | err = 0; | |
1713 | break; | |
b42be38b DH |
1714 | case NETLINK_LIST_MEMBERSHIPS: { |
1715 | int pos, idx, shift; | |
1716 | ||
1717 | err = 0; | |
47191d65 | 1718 | netlink_lock_table(); |
b42be38b DH |
1719 | for (pos = 0; pos * 8 < nlk->ngroups; pos += sizeof(u32)) { |
1720 | if (len - pos < sizeof(u32)) | |
1721 | break; | |
1722 | ||
1723 | idx = pos / sizeof(unsigned long); | |
1724 | shift = (pos % sizeof(unsigned long)) * 8; | |
1725 | if (put_user((u32)(nlk->groups[idx] >> shift), | |
1726 | (u32 __user *)(optval + pos))) { | |
1727 | err = -EFAULT; | |
1728 | break; | |
1729 | } | |
1730 | } | |
1731 | if (put_user(ALIGN(nlk->ngroups / 8, sizeof(u32)), optlen)) | |
1732 | err = -EFAULT; | |
47191d65 | 1733 | netlink_unlock_table(); |
b42be38b DH |
1734 | break; |
1735 | } | |
0a6a3a23 CR |
1736 | case NETLINK_CAP_ACK: |
1737 | if (len < sizeof(int)) | |
1738 | return -EINVAL; | |
1739 | len = sizeof(int); | |
1740 | val = nlk->flags & NETLINK_F_CAP_ACK ? 1 : 0; | |
1741 | if (put_user(len, optlen) || | |
1742 | put_user(val, optval)) | |
1743 | return -EFAULT; | |
1744 | err = 0; | |
1745 | break; | |
2d4bc933 JB |
1746 | case NETLINK_EXT_ACK: |
1747 | if (len < sizeof(int)) | |
1748 | return -EINVAL; | |
1749 | len = sizeof(int); | |
1750 | val = nlk->flags & NETLINK_F_EXT_ACK ? 1 : 0; | |
1751 | if (put_user(len, optlen) || put_user(val, optval)) | |
1752 | return -EFAULT; | |
1753 | err = 0; | |
1754 | break; | |
9a4595bc PM |
1755 | default: |
1756 | err = -ENOPROTOOPT; | |
1757 | } | |
1758 | return err; | |
1759 | } | |
1760 | ||
1761 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
1762 | { | |
1763 | struct nl_pktinfo info; | |
1764 | ||
1765 | info.group = NETLINK_CB(skb).dst_group; | |
1766 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
1767 | } | |
1768 | ||
59324cf3 ND |
1769 | static void netlink_cmsg_listen_all_nsid(struct sock *sk, struct msghdr *msg, |
1770 | struct sk_buff *skb) | |
1771 | { | |
1772 | if (!NETLINK_CB(skb).nsid_is_set) | |
1773 | return; | |
1774 | ||
1775 | put_cmsg(msg, SOL_NETLINK, NETLINK_LISTEN_ALL_NSID, sizeof(int), | |
1776 | &NETLINK_CB(skb).nsid); | |
1777 | } | |
1778 | ||
1b784140 | 1779 | static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) |
1da177e4 | 1780 | { |
1da177e4 LT |
1781 | struct sock *sk = sock->sk; |
1782 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 1783 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 1784 | u32 dst_portid; |
d629b836 | 1785 | u32 dst_group; |
1da177e4 LT |
1786 | struct sk_buff *skb; |
1787 | int err; | |
1788 | struct scm_cookie scm; | |
2d7a85f4 | 1789 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
1790 | |
1791 | if (msg->msg_flags&MSG_OOB) | |
1792 | return -EOPNOTSUPP; | |
1793 | ||
7cc05662 | 1794 | err = scm_send(sock, msg, &scm, true); |
1da177e4 LT |
1795 | if (err < 0) |
1796 | return err; | |
1797 | ||
1798 | if (msg->msg_namelen) { | |
b47030c7 | 1799 | err = -EINVAL; |
1da177e4 | 1800 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 1801 | goto out; |
15e47304 | 1802 | dst_portid = addr->nl_pid; |
d629b836 | 1803 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 1804 | err = -EPERM; |
15e47304 | 1805 | if ((dst_group || dst_portid) && |
5187cd05 | 1806 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 1807 | goto out; |
2d7a85f4 | 1808 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 1809 | } else { |
15e47304 | 1810 | dst_portid = nlk->dst_portid; |
d629b836 | 1811 | dst_group = nlk->dst_group; |
1da177e4 LT |
1812 | } |
1813 | ||
da314c99 | 1814 | if (!nlk->bound) { |
1da177e4 LT |
1815 | err = netlink_autobind(sock); |
1816 | if (err) | |
1817 | goto out; | |
da314c99 HX |
1818 | } else { |
1819 | /* Ensure nlk is hashed and visible. */ | |
1820 | smp_rmb(); | |
1da177e4 LT |
1821 | } |
1822 | ||
1823 | err = -EMSGSIZE; | |
1824 | if (len > sk->sk_sndbuf - 32) | |
1825 | goto out; | |
1826 | err = -ENOBUFS; | |
3a36515f | 1827 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 1828 | if (skb == NULL) |
1da177e4 LT |
1829 | goto out; |
1830 | ||
15e47304 | 1831 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 1832 | NETLINK_CB(skb).dst_group = dst_group; |
7cc05662 | 1833 | NETLINK_CB(skb).creds = scm.creds; |
2d7a85f4 | 1834 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 1835 | |
1da177e4 | 1836 | err = -EFAULT; |
6ce8e9ce | 1837 | if (memcpy_from_msg(skb_put(skb, len), msg, len)) { |
1da177e4 LT |
1838 | kfree_skb(skb); |
1839 | goto out; | |
1840 | } | |
1841 | ||
1842 | err = security_netlink_send(sk, skb); | |
1843 | if (err) { | |
1844 | kfree_skb(skb); | |
1845 | goto out; | |
1846 | } | |
1847 | ||
d629b836 | 1848 | if (dst_group) { |
1da177e4 | 1849 | atomic_inc(&skb->users); |
15e47304 | 1850 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 1851 | } |
15e47304 | 1852 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
1853 | |
1854 | out: | |
7cc05662 | 1855 | scm_destroy(&scm); |
1da177e4 LT |
1856 | return err; |
1857 | } | |
1858 | ||
1b784140 | 1859 | static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
1da177e4 LT |
1860 | int flags) |
1861 | { | |
1da177e4 LT |
1862 | struct scm_cookie scm; |
1863 | struct sock *sk = sock->sk; | |
1864 | struct netlink_sock *nlk = nlk_sk(sk); | |
1865 | int noblock = flags&MSG_DONTWAIT; | |
1866 | size_t copied; | |
68d6ac6d | 1867 | struct sk_buff *skb, *data_skb; |
b44d211e | 1868 | int err, ret; |
1da177e4 LT |
1869 | |
1870 | if (flags&MSG_OOB) | |
1871 | return -EOPNOTSUPP; | |
1872 | ||
1873 | copied = 0; | |
1874 | ||
6ac552fd PM |
1875 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
1876 | if (skb == NULL) | |
1da177e4 LT |
1877 | goto out; |
1878 | ||
68d6ac6d JB |
1879 | data_skb = skb; |
1880 | ||
1dacc76d JB |
1881 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
1882 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 1883 | /* |
68d6ac6d JB |
1884 | * If this skb has a frag_list, then here that means that we |
1885 | * will have to use the frag_list skb's data for compat tasks | |
1886 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 1887 | * |
68d6ac6d JB |
1888 | * If we need to send the compat skb, assign it to the |
1889 | * 'data_skb' variable so that it will be used below for data | |
1890 | * copying. We keep 'skb' for everything else, including | |
1891 | * freeing both later. | |
1dacc76d | 1892 | */ |
68d6ac6d JB |
1893 | if (flags & MSG_CMSG_COMPAT) |
1894 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
1895 | } |
1896 | #endif | |
1897 | ||
9063e21f ED |
1898 | /* Record the max length of recvmsg() calls for future allocations */ |
1899 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
1900 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
d35c99ff | 1901 | SKB_WITH_OVERHEAD(32768)); |
9063e21f | 1902 | |
68d6ac6d | 1903 | copied = data_skb->len; |
1da177e4 LT |
1904 | if (len < copied) { |
1905 | msg->msg_flags |= MSG_TRUNC; | |
1906 | copied = len; | |
1907 | } | |
1908 | ||
68d6ac6d | 1909 | skb_reset_transport_header(data_skb); |
51f3d02b | 1910 | err = skb_copy_datagram_msg(data_skb, 0, msg, copied); |
1da177e4 LT |
1911 | |
1912 | if (msg->msg_name) { | |
342dfc30 | 1913 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
1914 | addr->nl_family = AF_NETLINK; |
1915 | addr->nl_pad = 0; | |
15e47304 | 1916 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 1917 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
1918 | msg->msg_namelen = sizeof(*addr); |
1919 | } | |
1920 | ||
cc3a572f | 1921 | if (nlk->flags & NETLINK_F_RECV_PKTINFO) |
cc9a06cd | 1922 | netlink_cmsg_recv_pktinfo(msg, skb); |
59324cf3 ND |
1923 | if (nlk->flags & NETLINK_F_LISTEN_ALL_NSID) |
1924 | netlink_cmsg_listen_all_nsid(sk, msg, skb); | |
cc9a06cd | 1925 | |
7cc05662 CH |
1926 | memset(&scm, 0, sizeof(scm)); |
1927 | scm.creds = *NETLINK_CREDS(skb); | |
188ccb55 | 1928 | if (flags & MSG_TRUNC) |
68d6ac6d | 1929 | copied = data_skb->len; |
daa3766e | 1930 | |
1da177e4 LT |
1931 | skb_free_datagram(sk, skb); |
1932 | ||
16b304f3 PS |
1933 | if (nlk->cb_running && |
1934 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
1935 | ret = netlink_dump(sk); |
1936 | if (ret) { | |
ac30ef83 | 1937 | sk->sk_err = -ret; |
b44d211e AV |
1938 | sk->sk_error_report(sk); |
1939 | } | |
1940 | } | |
1da177e4 | 1941 | |
7cc05662 | 1942 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
1943 | out: |
1944 | netlink_rcv_wake(sk); | |
1945 | return err ? : copied; | |
1946 | } | |
1947 | ||
676d2369 | 1948 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 1949 | { |
cd40b7d3 | 1950 | BUG(); |
1da177e4 LT |
1951 | } |
1952 | ||
1953 | /* | |
746fac4d | 1954 | * We export these functions to other modules. They provide a |
1da177e4 LT |
1955 | * complete set of kernel non-blocking support for message |
1956 | * queueing. | |
1957 | */ | |
1958 | ||
1959 | struct sock * | |
9f00d977 PNA |
1960 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
1961 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
1962 | { |
1963 | struct socket *sock; | |
1964 | struct sock *sk; | |
77247bbb | 1965 | struct netlink_sock *nlk; |
5c398dc8 | 1966 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
1967 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
1968 | unsigned int groups; | |
1da177e4 | 1969 | |
fab2caf6 | 1970 | BUG_ON(!nl_table); |
1da177e4 | 1971 | |
6ac552fd | 1972 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
1973 | return NULL; |
1974 | ||
1975 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
1976 | return NULL; | |
13d3078e EB |
1977 | |
1978 | if (__netlink_create(net, sock, cb_mutex, unit, 1) < 0) | |
23fe1866 PE |
1979 | goto out_sock_release_nosk; |
1980 | ||
1981 | sk = sock->sk; | |
4fdb3bb7 | 1982 | |
a31f2d17 | 1983 | if (!cfg || cfg->groups < 32) |
4277a083 | 1984 | groups = 32; |
a31f2d17 PNA |
1985 | else |
1986 | groups = cfg->groups; | |
4277a083 | 1987 | |
5c398dc8 | 1988 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
1989 | if (!listeners) |
1990 | goto out_sock_release; | |
1991 | ||
1da177e4 | 1992 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
1993 | if (cfg && cfg->input) |
1994 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 1995 | |
8ea65f4a | 1996 | if (netlink_insert(sk, 0)) |
77247bbb | 1997 | goto out_sock_release; |
4fdb3bb7 | 1998 | |
77247bbb | 1999 | nlk = nlk_sk(sk); |
cc3a572f | 2000 | nlk->flags |= NETLINK_F_KERNEL_SOCKET; |
4fdb3bb7 | 2001 | |
4fdb3bb7 | 2002 | netlink_table_grab(); |
b4b51029 EB |
2003 | if (!nl_table[unit].registered) { |
2004 | nl_table[unit].groups = groups; | |
5c398dc8 | 2005 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2006 | nl_table[unit].cb_mutex = cb_mutex; |
2007 | nl_table[unit].module = module; | |
9785e10a PNA |
2008 | if (cfg) { |
2009 | nl_table[unit].bind = cfg->bind; | |
6251edd9 | 2010 | nl_table[unit].unbind = cfg->unbind; |
9785e10a | 2011 | nl_table[unit].flags = cfg->flags; |
da12c90e G |
2012 | if (cfg->compare) |
2013 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2014 | } |
b4b51029 | 2015 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2016 | } else { |
2017 | kfree(listeners); | |
869e58f8 | 2018 | nl_table[unit].registered++; |
b4b51029 | 2019 | } |
4fdb3bb7 | 2020 | netlink_table_ungrab(); |
77247bbb PM |
2021 | return sk; |
2022 | ||
4fdb3bb7 | 2023 | out_sock_release: |
4277a083 | 2024 | kfree(listeners); |
9dfbec1f | 2025 | netlink_kernel_release(sk); |
23fe1866 PE |
2026 | return NULL; |
2027 | ||
2028 | out_sock_release_nosk: | |
4fdb3bb7 | 2029 | sock_release(sock); |
77247bbb | 2030 | return NULL; |
1da177e4 | 2031 | } |
9f00d977 | 2032 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2033 | |
2034 | void | |
2035 | netlink_kernel_release(struct sock *sk) | |
2036 | { | |
13d3078e EB |
2037 | if (sk == NULL || sk->sk_socket == NULL) |
2038 | return; | |
2039 | ||
2040 | sock_release(sk->sk_socket); | |
b7c6ba6e DL |
2041 | } |
2042 | EXPORT_SYMBOL(netlink_kernel_release); | |
2043 | ||
d136f1bd | 2044 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2045 | { |
5c398dc8 | 2046 | struct listeners *new, *old; |
b4ff4f04 | 2047 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2048 | |
2049 | if (groups < 32) | |
2050 | groups = 32; | |
2051 | ||
b4ff4f04 | 2052 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2053 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2054 | if (!new) | |
d136f1bd | 2055 | return -ENOMEM; |
6d772ac5 | 2056 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2057 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2058 | rcu_assign_pointer(tbl->listeners, new); | |
2059 | ||
37b6b935 | 2060 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2061 | } |
2062 | tbl->groups = groups; | |
2063 | ||
d136f1bd JB |
2064 | return 0; |
2065 | } | |
2066 | ||
2067 | /** | |
2068 | * netlink_change_ngroups - change number of multicast groups | |
2069 | * | |
2070 | * This changes the number of multicast groups that are available | |
2071 | * on a certain netlink family. Note that it is not possible to | |
2072 | * change the number of groups to below 32. Also note that it does | |
2073 | * not implicitly call netlink_clear_multicast_users() when the | |
2074 | * number of groups is reduced. | |
2075 | * | |
2076 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2077 | * @groups: The new number of groups. | |
2078 | */ | |
2079 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2080 | { | |
2081 | int err; | |
2082 | ||
2083 | netlink_table_grab(); | |
2084 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2085 | netlink_table_ungrab(); |
d136f1bd | 2086 | |
b4ff4f04 JB |
2087 | return err; |
2088 | } | |
b4ff4f04 | 2089 | |
b8273570 JB |
2090 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2091 | { | |
2092 | struct sock *sk; | |
b8273570 JB |
2093 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2094 | ||
b67bfe0d | 2095 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2096 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2097 | } | |
2098 | ||
a46621a3 | 2099 | struct nlmsghdr * |
15e47304 | 2100 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2101 | { |
2102 | struct nlmsghdr *nlh; | |
573ce260 | 2103 | int size = nlmsg_msg_size(len); |
a46621a3 | 2104 | |
23b45672 | 2105 | nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2106 | nlh->nlmsg_type = type; |
2107 | nlh->nlmsg_len = size; | |
2108 | nlh->nlmsg_flags = flags; | |
15e47304 | 2109 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2110 | nlh->nlmsg_seq = seq; |
2111 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2112 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2113 | return nlh; |
2114 | } | |
2115 | EXPORT_SYMBOL(__nlmsg_put); | |
2116 | ||
1da177e4 LT |
2117 | /* |
2118 | * It looks a bit ugly. | |
2119 | * It would be better to create kernel thread. | |
2120 | */ | |
2121 | ||
2122 | static int netlink_dump(struct sock *sk) | |
2123 | { | |
2124 | struct netlink_sock *nlk = nlk_sk(sk); | |
2125 | struct netlink_callback *cb; | |
c7ac8679 | 2126 | struct sk_buff *skb = NULL; |
1da177e4 | 2127 | struct nlmsghdr *nlh; |
92964c79 | 2128 | struct module *module; |
bf8b79e4 | 2129 | int len, err = -ENOBUFS; |
db65a3aa | 2130 | int alloc_min_size; |
c7ac8679 | 2131 | int alloc_size; |
1da177e4 | 2132 | |
af65bdfc | 2133 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2134 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2135 | err = -EINVAL; |
2136 | goto errout_skb; | |
1da177e4 LT |
2137 | } |
2138 | ||
d1b4c689 | 2139 | if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) |
f9c22888 | 2140 | goto errout_skb; |
9063e21f ED |
2141 | |
2142 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2143 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2144 | * to reduce number of system calls on dump operations, if user | |
2145 | * ever provided a big enough buffer. | |
2146 | */ | |
db65a3aa AR |
2147 | cb = &nlk->cb; |
2148 | alloc_min_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); | |
2149 | ||
2150 | if (alloc_min_size < nlk->max_recvmsg_len) { | |
2151 | alloc_size = nlk->max_recvmsg_len; | |
d35c99ff ED |
2152 | skb = alloc_skb(alloc_size, |
2153 | (GFP_KERNEL & ~__GFP_DIRECT_RECLAIM) | | |
2154 | __GFP_NOWARN | __GFP_NORETRY); | |
9063e21f | 2155 | } |
db65a3aa AR |
2156 | if (!skb) { |
2157 | alloc_size = alloc_min_size; | |
c5b0db32 | 2158 | skb = alloc_skb(alloc_size, GFP_KERNEL); |
db65a3aa | 2159 | } |
c7ac8679 | 2160 | if (!skb) |
c63d6ea3 | 2161 | goto errout_skb; |
db65a3aa AR |
2162 | |
2163 | /* Trim skb to allocated size. User is expected to provide buffer as | |
2164 | * large as max(min_dump_alloc, 16KiB (mac_recvmsg_len capped at | |
2165 | * netlink_recvmsg())). dump will pack as many smaller messages as | |
2166 | * could fit within the allocated skb. skb is typically allocated | |
2167 | * with larger space than required (could be as much as near 2x the | |
2168 | * requested size with align to next power of 2 approach). Allowing | |
2169 | * dump to use the excess space makes it difficult for a user to have a | |
2170 | * reasonable static buffer based on the expected largest dump of a | |
2171 | * single netdev. The outcome is MSG_TRUNC error. | |
2172 | */ | |
d1b4c689 | 2173 | skb_reserve(skb, skb_tailroom(skb) - alloc_size); |
f9c22888 | 2174 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2175 | |
1da177e4 LT |
2176 | len = cb->dump(skb, cb); |
2177 | ||
2178 | if (len > 0) { | |
af65bdfc | 2179 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2180 | |
2181 | if (sk_filter(sk, skb)) | |
2182 | kfree_skb(skb); | |
4a7e7c2a ED |
2183 | else |
2184 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2185 | return 0; |
2186 | } | |
2187 | ||
bf8b79e4 TG |
2188 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2189 | if (!nlh) | |
2190 | goto errout_skb; | |
2191 | ||
670dc283 JB |
2192 | nl_dump_check_consistent(cb, nlh); |
2193 | ||
bf8b79e4 TG |
2194 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2195 | ||
b1153f29 SH |
2196 | if (sk_filter(sk, skb)) |
2197 | kfree_skb(skb); | |
4a7e7c2a ED |
2198 | else |
2199 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2200 | |
a8f74b22 TG |
2201 | if (cb->done) |
2202 | cb->done(cb); | |
1da177e4 | 2203 | |
16b304f3 | 2204 | nlk->cb_running = false; |
92964c79 HX |
2205 | module = cb->module; |
2206 | skb = cb->skb; | |
16b304f3 | 2207 | mutex_unlock(nlk->cb_mutex); |
92964c79 HX |
2208 | module_put(module); |
2209 | consume_skb(skb); | |
1da177e4 | 2210 | return 0; |
1797754e | 2211 | |
bf8b79e4 | 2212 | errout_skb: |
af65bdfc | 2213 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2214 | kfree_skb(skb); |
bf8b79e4 | 2215 | return err; |
1da177e4 LT |
2216 | } |
2217 | ||
6dc878a8 G |
2218 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2219 | const struct nlmsghdr *nlh, | |
2220 | struct netlink_dump_control *control) | |
1da177e4 LT |
2221 | { |
2222 | struct netlink_callback *cb; | |
2223 | struct sock *sk; | |
2224 | struct netlink_sock *nlk; | |
b44d211e | 2225 | int ret; |
1da177e4 | 2226 | |
d1b4c689 | 2227 | atomic_inc(&skb->users); |
f9c22888 | 2228 | |
15e47304 | 2229 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2230 | if (sk == NULL) { |
16b304f3 PS |
2231 | ret = -ECONNREFUSED; |
2232 | goto error_free; | |
1da177e4 | 2233 | } |
6dc878a8 | 2234 | |
16b304f3 | 2235 | nlk = nlk_sk(sk); |
af65bdfc | 2236 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2237 | /* A dump is in progress... */ |
16b304f3 | 2238 | if (nlk->cb_running) { |
6dc878a8 | 2239 | ret = -EBUSY; |
16b304f3 | 2240 | goto error_unlock; |
1da177e4 | 2241 | } |
6dc878a8 | 2242 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2243 | if (!try_module_get(control->module)) { |
6dc878a8 | 2244 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2245 | goto error_unlock; |
6dc878a8 G |
2246 | } |
2247 | ||
16b304f3 PS |
2248 | cb = &nlk->cb; |
2249 | memset(cb, 0, sizeof(*cb)); | |
fc9e50f5 | 2250 | cb->start = control->start; |
16b304f3 PS |
2251 | cb->dump = control->dump; |
2252 | cb->done = control->done; | |
2253 | cb->nlh = nlh; | |
2254 | cb->data = control->data; | |
2255 | cb->module = control->module; | |
2256 | cb->min_dump_alloc = control->min_dump_alloc; | |
2257 | cb->skb = skb; | |
2258 | ||
2259 | nlk->cb_running = true; | |
2260 | ||
af65bdfc | 2261 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2262 | |
fc9e50f5 TH |
2263 | if (cb->start) |
2264 | cb->start(cb); | |
2265 | ||
b44d211e | 2266 | ret = netlink_dump(sk); |
1da177e4 | 2267 | sock_put(sk); |
5c58298c | 2268 | |
b44d211e AV |
2269 | if (ret) |
2270 | return ret; | |
2271 | ||
5c58298c DL |
2272 | /* We successfully started a dump, by returning -EINTR we |
2273 | * signal not to send ACK even if it was requested. | |
2274 | */ | |
2275 | return -EINTR; | |
16b304f3 PS |
2276 | |
2277 | error_unlock: | |
2278 | sock_put(sk); | |
2279 | mutex_unlock(nlk->cb_mutex); | |
2280 | error_free: | |
2281 | kfree_skb(skb); | |
2282 | return ret; | |
1da177e4 | 2283 | } |
6dc878a8 | 2284 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 | 2285 | |
2d4bc933 JB |
2286 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err, |
2287 | const struct netlink_ext_ack *extack) | |
1da177e4 LT |
2288 | { |
2289 | struct sk_buff *skb; | |
2290 | struct nlmsghdr *rep; | |
2291 | struct nlmsgerr *errmsg; | |
339bf98f | 2292 | size_t payload = sizeof(*errmsg); |
2d4bc933 | 2293 | size_t tlvlen = 0; |
0a6a3a23 | 2294 | struct netlink_sock *nlk = nlk_sk(NETLINK_CB(in_skb).sk); |
2d4bc933 | 2295 | unsigned int flags = 0; |
1da177e4 | 2296 | |
0a6a3a23 | 2297 | /* Error messages get the original request appened, unless the user |
2d4bc933 JB |
2298 | * requests to cap the error message, and get extra error data if |
2299 | * requested. | |
0a6a3a23 | 2300 | */ |
2d4bc933 JB |
2301 | if (err) { |
2302 | if (!(nlk->flags & NETLINK_F_CAP_ACK)) | |
2303 | payload += nlmsg_len(nlh); | |
2304 | else | |
2305 | flags |= NLM_F_CAPPED; | |
2306 | if (nlk->flags & NETLINK_F_EXT_ACK && extack) { | |
2307 | if (extack->_msg) | |
2308 | tlvlen += nla_total_size(strlen(extack->_msg) + 1); | |
2309 | if (extack->bad_attr) | |
2310 | tlvlen += nla_total_size(sizeof(u32)); | |
2311 | } | |
2312 | } else { | |
2313 | flags |= NLM_F_CAPPED; | |
2314 | } | |
1da177e4 | 2315 | |
2d4bc933 JB |
2316 | if (tlvlen) |
2317 | flags |= NLM_F_ACK_TLVS; | |
2318 | ||
2319 | skb = nlmsg_new(payload + tlvlen, GFP_KERNEL); | |
1da177e4 LT |
2320 | if (!skb) { |
2321 | struct sock *sk; | |
2322 | ||
3b1e0a65 | 2323 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2324 | in_skb->sk->sk_protocol, |
15e47304 | 2325 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2326 | if (sk) { |
2327 | sk->sk_err = ENOBUFS; | |
2328 | sk->sk_error_report(sk); | |
2329 | sock_put(sk); | |
2330 | } | |
2331 | return; | |
2332 | } | |
2333 | ||
15e47304 | 2334 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
2d4bc933 | 2335 | NLMSG_ERROR, payload, flags); |
bf8b79e4 | 2336 | errmsg = nlmsg_data(rep); |
1da177e4 | 2337 | errmsg->error = err; |
0a6a3a23 | 2338 | memcpy(&errmsg->msg, nlh, payload > sizeof(*errmsg) ? nlh->nlmsg_len : sizeof(*nlh)); |
2d4bc933 JB |
2339 | |
2340 | if (err && nlk->flags & NETLINK_F_EXT_ACK && extack) { | |
2341 | if (extack->_msg) | |
2342 | WARN_ON(nla_put_string(skb, NLMSGERR_ATTR_MSG, | |
2343 | extack->_msg)); | |
2344 | if (extack->bad_attr && | |
2345 | !WARN_ON((u8 *)extack->bad_attr < in_skb->data || | |
2346 | (u8 *)extack->bad_attr >= in_skb->data + | |
2347 | in_skb->len)) | |
2348 | WARN_ON(nla_put_u32(skb, NLMSGERR_ATTR_OFFS, | |
2349 | (u8 *)extack->bad_attr - | |
2350 | in_skb->data)); | |
2351 | } | |
2352 | ||
2353 | nlmsg_end(skb, rep); | |
2354 | ||
15e47304 | 2355 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2356 | } |
6ac552fd | 2357 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2358 | |
cd40b7d3 | 2359 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
2d4bc933 JB |
2360 | struct nlmsghdr *, |
2361 | struct netlink_ext_ack *)) | |
82ace47a | 2362 | { |
2d4bc933 | 2363 | struct netlink_ext_ack extack = {}; |
82ace47a TG |
2364 | struct nlmsghdr *nlh; |
2365 | int err; | |
2366 | ||
2367 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2368 | int msglen; |
2369 | ||
b529ccf2 | 2370 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2371 | err = 0; |
82ace47a | 2372 | |
ad8e4b75 | 2373 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2374 | return 0; |
2375 | ||
d35b6856 TG |
2376 | /* Only requests are handled by the kernel */ |
2377 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2378 | goto ack; |
45e7ae7f TG |
2379 | |
2380 | /* Skip control messages */ | |
2381 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2382 | goto ack; |
d35b6856 | 2383 | |
2d4bc933 | 2384 | err = cb(skb, nlh, &extack); |
5c58298c DL |
2385 | if (err == -EINTR) |
2386 | goto skip; | |
2387 | ||
2388 | ack: | |
d35b6856 | 2389 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
2d4bc933 | 2390 | netlink_ack(skb, nlh, err, &extack); |
82ace47a | 2391 | |
5c58298c | 2392 | skip: |
6ac552fd | 2393 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2394 | if (msglen > skb->len) |
2395 | msglen = skb->len; | |
2396 | skb_pull(skb, msglen); | |
82ace47a TG |
2397 | } |
2398 | ||
2399 | return 0; | |
2400 | } | |
6ac552fd | 2401 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2402 | |
d387f6ad TG |
2403 | /** |
2404 | * nlmsg_notify - send a notification netlink message | |
2405 | * @sk: netlink socket to use | |
2406 | * @skb: notification message | |
15e47304 | 2407 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2408 | * @group: destination multicast group or 0 |
2409 | * @report: 1 to report back, 0 to disable | |
2410 | * @flags: allocation flags | |
2411 | */ | |
15e47304 | 2412 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2413 | unsigned int group, int report, gfp_t flags) |
2414 | { | |
2415 | int err = 0; | |
2416 | ||
2417 | if (group) { | |
15e47304 | 2418 | int exclude_portid = 0; |
d387f6ad TG |
2419 | |
2420 | if (report) { | |
2421 | atomic_inc(&skb->users); | |
15e47304 | 2422 | exclude_portid = portid; |
d387f6ad TG |
2423 | } |
2424 | ||
1ce85fe4 PNA |
2425 | /* errors reported via destination sk->sk_err, but propagate |
2426 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2427 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2428 | } |
2429 | ||
1ce85fe4 PNA |
2430 | if (report) { |
2431 | int err2; | |
2432 | ||
15e47304 | 2433 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2434 | if (!err || err == -ESRCH) |
2435 | err = err2; | |
2436 | } | |
d387f6ad TG |
2437 | |
2438 | return err; | |
2439 | } | |
6ac552fd | 2440 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2441 | |
1da177e4 LT |
2442 | #ifdef CONFIG_PROC_FS |
2443 | struct nl_seq_iter { | |
e372c414 | 2444 | struct seq_net_private p; |
56d28b1e | 2445 | struct rhashtable_iter hti; |
1da177e4 | 2446 | int link; |
1da177e4 LT |
2447 | }; |
2448 | ||
56d28b1e | 2449 | static int netlink_walk_start(struct nl_seq_iter *iter) |
1da177e4 | 2450 | { |
56d28b1e | 2451 | int err; |
1da177e4 | 2452 | |
8f6fd83c BC |
2453 | err = rhashtable_walk_init(&nl_table[iter->link].hash, &iter->hti, |
2454 | GFP_KERNEL); | |
56d28b1e HX |
2455 | if (err) { |
2456 | iter->link = MAX_LINKS; | |
2457 | return err; | |
1da177e4 | 2458 | } |
56d28b1e HX |
2459 | |
2460 | err = rhashtable_walk_start(&iter->hti); | |
2461 | return err == -EAGAIN ? 0 : err; | |
1da177e4 LT |
2462 | } |
2463 | ||
56d28b1e | 2464 | static void netlink_walk_stop(struct nl_seq_iter *iter) |
1da177e4 | 2465 | { |
56d28b1e HX |
2466 | rhashtable_walk_stop(&iter->hti); |
2467 | rhashtable_walk_exit(&iter->hti); | |
1da177e4 LT |
2468 | } |
2469 | ||
56d28b1e | 2470 | static void *__netlink_seq_next(struct seq_file *seq) |
1da177e4 | 2471 | { |
56d28b1e | 2472 | struct nl_seq_iter *iter = seq->private; |
e341694e | 2473 | struct netlink_sock *nlk; |
1da177e4 | 2474 | |
56d28b1e HX |
2475 | do { |
2476 | for (;;) { | |
2477 | int err; | |
1da177e4 | 2478 | |
56d28b1e | 2479 | nlk = rhashtable_walk_next(&iter->hti); |
746fac4d | 2480 | |
56d28b1e HX |
2481 | if (IS_ERR(nlk)) { |
2482 | if (PTR_ERR(nlk) == -EAGAIN) | |
2483 | continue; | |
e341694e | 2484 | |
56d28b1e HX |
2485 | return nlk; |
2486 | } | |
1da177e4 | 2487 | |
56d28b1e HX |
2488 | if (nlk) |
2489 | break; | |
1da177e4 | 2490 | |
56d28b1e HX |
2491 | netlink_walk_stop(iter); |
2492 | if (++iter->link >= MAX_LINKS) | |
2493 | return NULL; | |
da12c90e | 2494 | |
56d28b1e HX |
2495 | err = netlink_walk_start(iter); |
2496 | if (err) | |
2497 | return ERR_PTR(err); | |
1da177e4 | 2498 | } |
56d28b1e | 2499 | } while (sock_net(&nlk->sk) != seq_file_net(seq)); |
1da177e4 | 2500 | |
56d28b1e HX |
2501 | return nlk; |
2502 | } | |
1da177e4 | 2503 | |
56d28b1e HX |
2504 | static void *netlink_seq_start(struct seq_file *seq, loff_t *posp) |
2505 | { | |
2506 | struct nl_seq_iter *iter = seq->private; | |
2507 | void *obj = SEQ_START_TOKEN; | |
2508 | loff_t pos; | |
2509 | int err; | |
2510 | ||
2511 | iter->link = 0; | |
2512 | ||
2513 | err = netlink_walk_start(iter); | |
2514 | if (err) | |
2515 | return ERR_PTR(err); | |
2516 | ||
2517 | for (pos = *posp; pos && obj && !IS_ERR(obj); pos--) | |
2518 | obj = __netlink_seq_next(seq); | |
2519 | ||
2520 | return obj; | |
2521 | } | |
2522 | ||
2523 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2524 | { | |
2525 | ++*pos; | |
2526 | return __netlink_seq_next(seq); | |
1da177e4 LT |
2527 | } |
2528 | ||
2529 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
2530 | { | |
56d28b1e HX |
2531 | struct nl_seq_iter *iter = seq->private; |
2532 | ||
2533 | if (iter->link >= MAX_LINKS) | |
2534 | return; | |
2535 | ||
2536 | netlink_walk_stop(iter); | |
1da177e4 LT |
2537 | } |
2538 | ||
2539 | ||
2540 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2541 | { | |
658cb354 | 2542 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
2543 | seq_puts(seq, |
2544 | "sk Eth Pid Groups " | |
cf0aa4e0 | 2545 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 2546 | } else { |
1da177e4 LT |
2547 | struct sock *s = v; |
2548 | struct netlink_sock *nlk = nlk_sk(s); | |
2549 | ||
16b304f3 | 2550 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
2551 | s, |
2552 | s->sk_protocol, | |
15e47304 | 2553 | nlk->portid, |
513c2500 | 2554 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2555 | sk_rmem_alloc_get(s), |
2556 | sk_wmem_alloc_get(s), | |
16b304f3 | 2557 | nlk->cb_running, |
38938bfe | 2558 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
2559 | atomic_read(&s->sk_drops), |
2560 | sock_i_ino(s) | |
1da177e4 LT |
2561 | ); |
2562 | ||
2563 | } | |
2564 | return 0; | |
2565 | } | |
2566 | ||
56b3d975 | 2567 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
2568 | .start = netlink_seq_start, |
2569 | .next = netlink_seq_next, | |
2570 | .stop = netlink_seq_stop, | |
2571 | .show = netlink_seq_show, | |
2572 | }; | |
2573 | ||
2574 | ||
2575 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
2576 | { | |
e372c414 DL |
2577 | return seq_open_net(inode, file, &netlink_seq_ops, |
2578 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
2579 | } |
2580 | ||
da7071d7 | 2581 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
2582 | .owner = THIS_MODULE, |
2583 | .open = netlink_seq_open, | |
2584 | .read = seq_read, | |
2585 | .llseek = seq_lseek, | |
e372c414 | 2586 | .release = seq_release_net, |
1da177e4 LT |
2587 | }; |
2588 | ||
2589 | #endif | |
2590 | ||
2591 | int netlink_register_notifier(struct notifier_block *nb) | |
2592 | { | |
efa172f4 | 2593 | return blocking_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 2594 | } |
6ac552fd | 2595 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
2596 | |
2597 | int netlink_unregister_notifier(struct notifier_block *nb) | |
2598 | { | |
efa172f4 | 2599 | return blocking_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 2600 | } |
6ac552fd | 2601 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 2602 | |
90ddc4f0 | 2603 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
2604 | .family = PF_NETLINK, |
2605 | .owner = THIS_MODULE, | |
2606 | .release = netlink_release, | |
2607 | .bind = netlink_bind, | |
2608 | .connect = netlink_connect, | |
2609 | .socketpair = sock_no_socketpair, | |
2610 | .accept = sock_no_accept, | |
2611 | .getname = netlink_getname, | |
d1b4c689 | 2612 | .poll = datagram_poll, |
025c6818 | 2613 | .ioctl = netlink_ioctl, |
1da177e4 LT |
2614 | .listen = sock_no_listen, |
2615 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
2616 | .setsockopt = netlink_setsockopt, |
2617 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
2618 | .sendmsg = netlink_sendmsg, |
2619 | .recvmsg = netlink_recvmsg, | |
d1b4c689 | 2620 | .mmap = sock_no_mmap, |
1da177e4 LT |
2621 | .sendpage = sock_no_sendpage, |
2622 | }; | |
2623 | ||
ec1b4cf7 | 2624 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
2625 | .family = PF_NETLINK, |
2626 | .create = netlink_create, | |
2627 | .owner = THIS_MODULE, /* for consistency 8) */ | |
2628 | }; | |
2629 | ||
4665079c | 2630 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
2631 | { |
2632 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 2633 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
2634 | return -ENOMEM; |
2635 | #endif | |
2636 | return 0; | |
2637 | } | |
2638 | ||
4665079c | 2639 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
2640 | { |
2641 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 2642 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
2643 | #endif |
2644 | } | |
2645 | ||
b963ea89 DM |
2646 | static void __init netlink_add_usersock_entry(void) |
2647 | { | |
5c398dc8 | 2648 | struct listeners *listeners; |
b963ea89 DM |
2649 | int groups = 32; |
2650 | ||
5c398dc8 | 2651 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 2652 | if (!listeners) |
5c398dc8 | 2653 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
2654 | |
2655 | netlink_table_grab(); | |
2656 | ||
2657 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 2658 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
2659 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
2660 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 2661 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
2662 | |
2663 | netlink_table_ungrab(); | |
2664 | } | |
2665 | ||
022cbae6 | 2666 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
2667 | .init = netlink_net_init, |
2668 | .exit = netlink_net_exit, | |
2669 | }; | |
2670 | ||
49f7b33e | 2671 | static inline u32 netlink_hash(const void *data, u32 len, u32 seed) |
c428ecd1 HX |
2672 | { |
2673 | const struct netlink_sock *nlk = data; | |
2674 | struct netlink_compare_arg arg; | |
2675 | ||
da314c99 | 2676 | netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->portid); |
11b58ba1 | 2677 | return jhash2((u32 *)&arg, netlink_compare_arg_len / sizeof(u32), seed); |
c428ecd1 HX |
2678 | } |
2679 | ||
2680 | static const struct rhashtable_params netlink_rhashtable_params = { | |
2681 | .head_offset = offsetof(struct netlink_sock, node), | |
2682 | .key_len = netlink_compare_arg_len, | |
c428ecd1 HX |
2683 | .obj_hashfn = netlink_hash, |
2684 | .obj_cmpfn = netlink_compare, | |
b5e2c150 | 2685 | .automatic_shrinking = true, |
c428ecd1 HX |
2686 | }; |
2687 | ||
1da177e4 LT |
2688 | static int __init netlink_proto_init(void) |
2689 | { | |
1da177e4 | 2690 | int i; |
1da177e4 LT |
2691 | int err = proto_register(&netlink_proto, 0); |
2692 | ||
2693 | if (err != 0) | |
2694 | goto out; | |
2695 | ||
fab25745 | 2696 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 2697 | |
0da974f4 | 2698 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
2699 | if (!nl_table) |
2700 | goto panic; | |
1da177e4 | 2701 | |
1da177e4 | 2702 | for (i = 0; i < MAX_LINKS; i++) { |
c428ecd1 HX |
2703 | if (rhashtable_init(&nl_table[i].hash, |
2704 | &netlink_rhashtable_params) < 0) { | |
e341694e TG |
2705 | while (--i > 0) |
2706 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 2707 | kfree(nl_table); |
fab2caf6 | 2708 | goto panic; |
1da177e4 | 2709 | } |
1da177e4 LT |
2710 | } |
2711 | ||
bcbde0d4 DB |
2712 | INIT_LIST_HEAD(&netlink_tap_all); |
2713 | ||
b963ea89 DM |
2714 | netlink_add_usersock_entry(); |
2715 | ||
1da177e4 | 2716 | sock_register(&netlink_family_ops); |
b4b51029 | 2717 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 2718 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
2719 | rtnetlink_init(); |
2720 | out: | |
2721 | return err; | |
fab2caf6 AM |
2722 | panic: |
2723 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
2724 | } |
2725 | ||
1da177e4 | 2726 | core_initcall(netlink_proto_init); |