1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * common UDP/RAW code
4 * Linux INET6 implementation
5 *
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 */
9
10 #include <linux/capability.h>
11 #include <linux/errno.h>
12 #include <linux/types.h>
13 #include <linux/kernel.h>
14 #include <linux/interrupt.h>
15 #include <linux/socket.h>
16 #include <linux/sockios.h>
17 #include <linux/in6.h>
18 #include <linux/ipv6.h>
19 #include <linux/route.h>
20 #include <linux/slab.h>
21 #include <linux/export.h>
22 #include <linux/icmp.h>
23
24 #include <net/ipv6.h>
25 #include <net/ndisc.h>
26 #include <net/addrconf.h>
27 #include <net/transp_v6.h>
28 #include <net/ip6_route.h>
29 #include <net/tcp_states.h>
30 #include <net/dsfield.h>
31 #include <net/sock_reuseport.h>
32
33 #include <linux/errqueue.h>
34 #include <linux/uaccess.h>
35
ipv6_mapped_addr_any(const struct in6_addr * a)36 static bool ipv6_mapped_addr_any(const struct in6_addr *a)
37 {
38 return ipv6_addr_v4mapped(a) && (a->s6_addr32[3] == 0);
39 }
40
ip6_datagram_flow_key_init(struct flowi6 * fl6,struct sock * sk)41 static void ip6_datagram_flow_key_init(struct flowi6 *fl6, struct sock *sk)
42 {
43 struct inet_sock *inet = inet_sk(sk);
44 struct ipv6_pinfo *np = inet6_sk(sk);
45 int oif = sk->sk_bound_dev_if;
46
47 memset(fl6, 0, sizeof(*fl6));
48 fl6->flowi6_proto = sk->sk_protocol;
49 fl6->daddr = sk->sk_v6_daddr;
50 fl6->saddr = np->saddr;
51 fl6->flowi6_mark = sk->sk_mark;
52 fl6->fl6_dport = inet->inet_dport;
53 fl6->fl6_sport = inet->inet_sport;
54 fl6->flowlabel = np->flow_label;
55 fl6->flowi6_uid = sk->sk_uid;
56
57 if (!oif)
58 oif = np->sticky_pktinfo.ipi6_ifindex;
59
60 if (!oif) {
61 if (ipv6_addr_is_multicast(&fl6->daddr))
62 oif = np->mcast_oif;
63 else
64 oif = np->ucast_oif;
65 }
66
67 fl6->flowi6_oif = oif;
68 security_sk_classify_flow(sk, flowi6_to_flowi_common(fl6));
69 }
70
ip6_datagram_dst_update(struct sock * sk,bool fix_sk_saddr)71 int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr)
72 {
73 struct ip6_flowlabel *flowlabel = NULL;
74 struct in6_addr *final_p, final;
75 struct ipv6_txoptions *opt;
76 struct dst_entry *dst;
77 struct inet_sock *inet = inet_sk(sk);
78 struct ipv6_pinfo *np = inet6_sk(sk);
79 struct flowi6 fl6;
80 int err = 0;
81
82 if (np->sndflow && (np->flow_label & IPV6_FLOWLABEL_MASK)) {
83 flowlabel = fl6_sock_lookup(sk, np->flow_label);
84 if (IS_ERR(flowlabel))
85 return -EINVAL;
86 }
87 ip6_datagram_flow_key_init(&fl6, sk);
88
89 rcu_read_lock();
90 opt = flowlabel ? flowlabel->opt : rcu_dereference(np->opt);
91 final_p = fl6_update_dst(&fl6, opt, &final);
92 rcu_read_unlock();
93
94 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
95 if (IS_ERR(dst)) {
96 err = PTR_ERR(dst);
97 goto out;
98 }
99
100 if (fix_sk_saddr) {
101 if (ipv6_addr_any(&np->saddr))
102 np->saddr = fl6.saddr;
103
104 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
105 sk->sk_v6_rcv_saddr = fl6.saddr;
106 inet->inet_rcv_saddr = LOOPBACK4_IPV6;
107 if (sk->sk_prot->rehash)
108 sk->sk_prot->rehash(sk);
109 }
110 }
111
112 ip6_sk_dst_store_flow(sk, dst, &fl6);
113
114 out:
115 fl6_sock_release(flowlabel);
116 return err;
117 }
118
ip6_datagram_release_cb(struct sock * sk)119 void ip6_datagram_release_cb(struct sock *sk)
120 {
121 struct dst_entry *dst;
122
123 if (ipv6_addr_v4mapped(&sk->sk_v6_daddr))
124 return;
125
126 rcu_read_lock();
127 dst = __sk_dst_get(sk);
128 if (!dst || !dst->obsolete ||
129 dst->ops->check(dst, inet6_sk(sk)->dst_cookie)) {
130 rcu_read_unlock();
131 return;
132 }
133 rcu_read_unlock();
134
135 ip6_datagram_dst_update(sk, false);
136 }
137 EXPORT_SYMBOL_GPL(ip6_datagram_release_cb);
138
__ip6_datagram_connect(struct sock * sk,struct sockaddr * uaddr,int addr_len)139 int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr,
140 int addr_len)
141 {
142 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
143 struct inet_sock *inet = inet_sk(sk);
144 struct ipv6_pinfo *np = inet6_sk(sk);
145 struct in6_addr *daddr, old_daddr;
146 __be32 fl6_flowlabel = 0;
147 __be32 old_fl6_flowlabel;
148 __be16 old_dport;
149 int addr_type;
150 int err;
151
152 if (usin->sin6_family == AF_INET) {
153 if (ipv6_only_sock(sk))
154 return -EAFNOSUPPORT;
155 err = __ip4_datagram_connect(sk, uaddr, addr_len);
156 goto ipv4_connected;
157 }
158
159 if (addr_len < SIN6_LEN_RFC2133)
160 return -EINVAL;
161
162 if (usin->sin6_family != AF_INET6)
163 return -EAFNOSUPPORT;
164
165 if (np->sndflow)
166 fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
167
168 if (ipv6_addr_any(&usin->sin6_addr)) {
169 /*
170 * connect to self
171 */
172 if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr))
173 ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK),
174 &usin->sin6_addr);
175 else
176 usin->sin6_addr = in6addr_loopback;
177 }
178
179 addr_type = ipv6_addr_type(&usin->sin6_addr);
180
181 daddr = &usin->sin6_addr;
182
183 if (addr_type & IPV6_ADDR_MAPPED) {
184 struct sockaddr_in sin;
185
186 if (ipv6_only_sock(sk)) {
187 err = -ENETUNREACH;
188 goto out;
189 }
190 sin.sin_family = AF_INET;
191 sin.sin_addr.s_addr = daddr->s6_addr32[3];
192 sin.sin_port = usin->sin6_port;
193
194 err = __ip4_datagram_connect(sk,
195 (struct sockaddr *) &sin,
196 sizeof(sin));
197
198 ipv4_connected:
199 if (err)
200 goto out;
201
202 ipv6_addr_set_v4mapped(inet->inet_daddr, &sk->sk_v6_daddr);
203
204 if (ipv6_addr_any(&np->saddr) ||
205 ipv6_mapped_addr_any(&np->saddr))
206 ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
207
208 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) ||
209 ipv6_mapped_addr_any(&sk->sk_v6_rcv_saddr)) {
210 ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
211 &sk->sk_v6_rcv_saddr);
212 if (sk->sk_prot->rehash)
213 sk->sk_prot->rehash(sk);
214 }
215
216 goto out;
217 }
218
219 if (__ipv6_addr_needs_scope_id(addr_type)) {
220 if (addr_len >= sizeof(struct sockaddr_in6) &&
221 usin->sin6_scope_id) {
222 if (!sk_dev_equal_l3scope(sk, usin->sin6_scope_id)) {
223 err = -EINVAL;
224 goto out;
225 }
226 WRITE_ONCE(sk->sk_bound_dev_if, usin->sin6_scope_id);
227 }
228
229 if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
230 WRITE_ONCE(sk->sk_bound_dev_if, np->mcast_oif);
231
232 /* Connect to link-local address requires an interface */
233 if (!sk->sk_bound_dev_if) {
234 err = -EINVAL;
235 goto out;
236 }
237 }
238
239 /* save the current peer information before updating it */
240 old_daddr = sk->sk_v6_daddr;
241 old_fl6_flowlabel = np->flow_label;
242 old_dport = inet->inet_dport;
243
244 sk->sk_v6_daddr = *daddr;
245 np->flow_label = fl6_flowlabel;
246 inet->inet_dport = usin->sin6_port;
247
248 /*
249 * Check for a route to destination an obtain the
250 * destination cache for it.
251 */
252
253 err = ip6_datagram_dst_update(sk, true);
254 if (err) {
255 /* Restore the socket peer info, to keep it consistent with
256 * the old socket state
257 */
258 sk->sk_v6_daddr = old_daddr;
259 np->flow_label = old_fl6_flowlabel;
260 inet->inet_dport = old_dport;
261 goto out;
262 }
263
264 reuseport_has_conns_set(sk);
265 sk->sk_state = TCP_ESTABLISHED;
266 sk_set_txhash(sk);
267 out:
268 return err;
269 }
270 EXPORT_SYMBOL_GPL(__ip6_datagram_connect);
271
ip6_datagram_connect(struct sock * sk,struct sockaddr * uaddr,int addr_len)272 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
273 {
274 int res;
275
276 lock_sock(sk);
277 res = __ip6_datagram_connect(sk, uaddr, addr_len);
278 release_sock(sk);
279 return res;
280 }
281 EXPORT_SYMBOL_GPL(ip6_datagram_connect);
282
ip6_datagram_connect_v6_only(struct sock * sk,struct sockaddr * uaddr,int addr_len)283 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr,
284 int addr_len)
285 {
286 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr);
287 if (sin6->sin6_family != AF_INET6)
288 return -EAFNOSUPPORT;
289 return ip6_datagram_connect(sk, uaddr, addr_len);
290 }
291 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only);
292
ipv6_icmp_error_rfc4884(const struct sk_buff * skb,struct sock_ee_data_rfc4884 * out)293 static void ipv6_icmp_error_rfc4884(const struct sk_buff *skb,
294 struct sock_ee_data_rfc4884 *out)
295 {
296 switch (icmp6_hdr(skb)->icmp6_type) {
297 case ICMPV6_TIME_EXCEED:
298 case ICMPV6_DEST_UNREACH:
299 ip_icmp_error_rfc4884(skb, out, sizeof(struct icmp6hdr),
300 icmp6_hdr(skb)->icmp6_datagram_len * 8);
301 }
302 }
303
ipv6_icmp_error(struct sock * sk,struct sk_buff * skb,int err,__be16 port,u32 info,u8 * payload)304 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
305 __be16 port, u32 info, u8 *payload)
306 {
307 struct ipv6_pinfo *np = inet6_sk(sk);
308 struct icmp6hdr *icmph = icmp6_hdr(skb);
309 struct sock_exterr_skb *serr;
310
311 if (!np->recverr)
312 return;
313
314 skb = skb_clone(skb, GFP_ATOMIC);
315 if (!skb)
316 return;
317
318 skb->protocol = htons(ETH_P_IPV6);
319
320 serr = SKB_EXT_ERR(skb);
321 serr->ee.ee_errno = err;
322 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
323 serr->ee.ee_type = icmph->icmp6_type;
324 serr->ee.ee_code = icmph->icmp6_code;
325 serr->ee.ee_pad = 0;
326 serr->ee.ee_info = info;
327 serr->ee.ee_data = 0;
328 serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
329 skb_network_header(skb);
330 serr->port = port;
331
332 __skb_pull(skb, payload - skb->data);
333
334 if (inet6_sk(sk)->recverr_rfc4884)
335 ipv6_icmp_error_rfc4884(skb, &serr->ee.ee_rfc4884);
336
337 skb_reset_transport_header(skb);
338
339 if (sock_queue_err_skb(sk, skb))
340 kfree_skb(skb);
341 }
342
ipv6_local_error(struct sock * sk,int err,struct flowi6 * fl6,u32 info)343 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
344 {
345 const struct ipv6_pinfo *np = inet6_sk(sk);
346 struct sock_exterr_skb *serr;
347 struct ipv6hdr *iph;
348 struct sk_buff *skb;
349
350 if (!np->recverr)
351 return;
352
353 skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
354 if (!skb)
355 return;
356
357 skb->protocol = htons(ETH_P_IPV6);
358
359 skb_put(skb, sizeof(struct ipv6hdr));
360 skb_reset_network_header(skb);
361 iph = ipv6_hdr(skb);
362 iph->daddr = fl6->daddr;
363 ip6_flow_hdr(iph, 0, 0);
364
365 serr = SKB_EXT_ERR(skb);
366 serr->ee.ee_errno = err;
367 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
368 serr->ee.ee_type = 0;
369 serr->ee.ee_code = 0;
370 serr->ee.ee_pad = 0;
371 serr->ee.ee_info = info;
372 serr->ee.ee_data = 0;
373 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
374 serr->port = fl6->fl6_dport;
375
376 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
377 skb_reset_transport_header(skb);
378
379 if (sock_queue_err_skb(sk, skb))
380 kfree_skb(skb);
381 }
382
ipv6_local_rxpmtu(struct sock * sk,struct flowi6 * fl6,u32 mtu)383 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
384 {
385 struct ipv6_pinfo *np = inet6_sk(sk);
386 struct ipv6hdr *iph;
387 struct sk_buff *skb;
388 struct ip6_mtuinfo *mtu_info;
389
390 if (!np->rxopt.bits.rxpmtu)
391 return;
392
393 skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
394 if (!skb)
395 return;
396
397 skb_put(skb, sizeof(struct ipv6hdr));
398 skb_reset_network_header(skb);
399 iph = ipv6_hdr(skb);
400 iph->daddr = fl6->daddr;
401
402 mtu_info = IP6CBMTU(skb);
403
404 mtu_info->ip6m_mtu = mtu;
405 mtu_info->ip6m_addr.sin6_family = AF_INET6;
406 mtu_info->ip6m_addr.sin6_port = 0;
407 mtu_info->ip6m_addr.sin6_flowinfo = 0;
408 mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
409 mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
410
411 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
412 skb_reset_transport_header(skb);
413
414 skb = xchg(&np->rxpmtu, skb);
415 kfree_skb(skb);
416 }
417
418 /* For some errors we have valid addr_offset even with zero payload and
419 * zero port. Also, addr_offset should be supported if port is set.
420 */
ipv6_datagram_support_addr(struct sock_exterr_skb * serr)421 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr)
422 {
423 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 ||
424 serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
425 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
426 }
427
428 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL.
429 *
430 * At one point, excluding local errors was a quick test to identify icmp/icmp6
431 * errors. This is no longer true, but the test remained, so the v6 stack,
432 * unlike v4, also honors cmsg requests on all wifi and timestamp errors.
433 */
ip6_datagram_support_cmsg(struct sk_buff * skb,struct sock_exterr_skb * serr)434 static bool ip6_datagram_support_cmsg(struct sk_buff *skb,
435 struct sock_exterr_skb *serr)
436 {
437 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
438 serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6)
439 return true;
440
441 if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL)
442 return false;
443
444 if (!IP6CB(skb)->iif)
445 return false;
446
447 return true;
448 }
449
450 /*
451 * Handle MSG_ERRQUEUE
452 */
ipv6_recv_error(struct sock * sk,struct msghdr * msg,int len,int * addr_len)453 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
454 {
455 struct ipv6_pinfo *np = inet6_sk(sk);
456 struct sock_exterr_skb *serr;
457 struct sk_buff *skb;
458 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
459 struct {
460 struct sock_extended_err ee;
461 struct sockaddr_in6 offender;
462 } errhdr;
463 int err;
464 int copied;
465
466 err = -EAGAIN;
467 skb = sock_dequeue_err_skb(sk);
468 if (!skb)
469 goto out;
470
471 copied = skb->len;
472 if (copied > len) {
473 msg->msg_flags |= MSG_TRUNC;
474 copied = len;
475 }
476 err = skb_copy_datagram_msg(skb, 0, msg, copied);
477 if (unlikely(err)) {
478 kfree_skb(skb);
479 return err;
480 }
481 sock_recv_timestamp(msg, sk, skb);
482
483 serr = SKB_EXT_ERR(skb);
484
485 if (sin && ipv6_datagram_support_addr(serr)) {
486 const unsigned char *nh = skb_network_header(skb);
487 sin->sin6_family = AF_INET6;
488 sin->sin6_flowinfo = 0;
489 sin->sin6_port = serr->port;
490 if (skb->protocol == htons(ETH_P_IPV6)) {
491 const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset),
492 struct ipv6hdr, daddr);
493 sin->sin6_addr = ip6h->daddr;
494 if (np->sndflow)
495 sin->sin6_flowinfo = ip6_flowinfo(ip6h);
496 sin->sin6_scope_id =
497 ipv6_iface_scope_id(&sin->sin6_addr,
498 IP6CB(skb)->iif);
499 } else {
500 ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
501 &sin->sin6_addr);
502 sin->sin6_scope_id = 0;
503 }
504 *addr_len = sizeof(*sin);
505 }
506
507 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
508 sin = &errhdr.offender;
509 memset(sin, 0, sizeof(*sin));
510
511 if (ip6_datagram_support_cmsg(skb, serr)) {
512 sin->sin6_family = AF_INET6;
513 if (np->rxopt.all)
514 ip6_datagram_recv_common_ctl(sk, msg, skb);
515 if (skb->protocol == htons(ETH_P_IPV6)) {
516 sin->sin6_addr = ipv6_hdr(skb)->saddr;
517 if (np->rxopt.all)
518 ip6_datagram_recv_specific_ctl(sk, msg, skb);
519 sin->sin6_scope_id =
520 ipv6_iface_scope_id(&sin->sin6_addr,
521 IP6CB(skb)->iif);
522 } else {
523 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
524 &sin->sin6_addr);
525 if (inet_sk(sk)->cmsg_flags)
526 ip_cmsg_recv(msg, skb);
527 }
528 }
529
530 put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
531
532 /* Now we could try to dump offended packet options */
533
534 msg->msg_flags |= MSG_ERRQUEUE;
535 err = copied;
536
537 consume_skb(skb);
538 out:
539 return err;
540 }
541 EXPORT_SYMBOL_GPL(ipv6_recv_error);
542
543 /*
544 * Handle IPV6_RECVPATHMTU
545 */
ipv6_recv_rxpmtu(struct sock * sk,struct msghdr * msg,int len,int * addr_len)546 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
547 int *addr_len)
548 {
549 struct ipv6_pinfo *np = inet6_sk(sk);
550 struct sk_buff *skb;
551 struct ip6_mtuinfo mtu_info;
552 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
553 int err;
554 int copied;
555
556 err = -EAGAIN;
557 skb = xchg(&np->rxpmtu, NULL);
558 if (!skb)
559 goto out;
560
561 copied = skb->len;
562 if (copied > len) {
563 msg->msg_flags |= MSG_TRUNC;
564 copied = len;
565 }
566 err = skb_copy_datagram_msg(skb, 0, msg, copied);
567 if (err)
568 goto out_free_skb;
569
570 sock_recv_timestamp(msg, sk, skb);
571
572 memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info));
573
574 if (sin) {
575 sin->sin6_family = AF_INET6;
576 sin->sin6_flowinfo = 0;
577 sin->sin6_port = 0;
578 sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id;
579 sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr;
580 *addr_len = sizeof(*sin);
581 }
582
583 put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info);
584
585 err = copied;
586
587 out_free_skb:
588 kfree_skb(skb);
589 out:
590 return err;
591 }
592
593
ip6_datagram_recv_common_ctl(struct sock * sk,struct msghdr * msg,struct sk_buff * skb)594 void ip6_datagram_recv_common_ctl(struct sock *sk, struct msghdr *msg,
595 struct sk_buff *skb)
596 {
597 struct ipv6_pinfo *np = inet6_sk(sk);
598 bool is_ipv6 = skb->protocol == htons(ETH_P_IPV6);
599
600 if (np->rxopt.bits.rxinfo) {
601 struct in6_pktinfo src_info;
602
603 if (is_ipv6) {
604 src_info.ipi6_ifindex = IP6CB(skb)->iif;
605 src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
606 } else {
607 src_info.ipi6_ifindex =
608 PKTINFO_SKB_CB(skb)->ipi_ifindex;
609 ipv6_addr_set_v4mapped(ip_hdr(skb)->daddr,
610 &src_info.ipi6_addr);
611 }
612
613 if (src_info.ipi6_ifindex >= 0)
614 put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO,
615 sizeof(src_info), &src_info);
616 }
617 }
618
ip6_datagram_recv_specific_ctl(struct sock * sk,struct msghdr * msg,struct sk_buff * skb)619 void ip6_datagram_recv_specific_ctl(struct sock *sk, struct msghdr *msg,
620 struct sk_buff *skb)
621 {
622 struct ipv6_pinfo *np = inet6_sk(sk);
623 struct inet6_skb_parm *opt = IP6CB(skb);
624 unsigned char *nh = skb_network_header(skb);
625
626 if (np->rxopt.bits.rxhlim) {
627 int hlim = ipv6_hdr(skb)->hop_limit;
628 put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
629 }
630
631 if (np->rxopt.bits.rxtclass) {
632 int tclass = ipv6_get_dsfield(ipv6_hdr(skb));
633 put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
634 }
635
636 if (np->rxopt.bits.rxflow) {
637 __be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh);
638 if (flowinfo)
639 put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
640 }
641
642 /* HbH is allowed only once */
643 if (np->rxopt.bits.hopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
644 u8 *ptr = nh + sizeof(struct ipv6hdr);
645 put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
646 }
647
648 if (opt->lastopt &&
649 (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
650 /*
651 * Silly enough, but we need to reparse in order to
652 * report extension headers (except for HbH)
653 * in order.
654 *
655 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
656 * (and WILL NOT be) defined because
657 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
658 */
659 unsigned int off = sizeof(struct ipv6hdr);
660 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
661
662 while (off <= opt->lastopt) {
663 unsigned int len;
664 u8 *ptr = nh + off;
665
666 switch (nexthdr) {
667 case IPPROTO_DSTOPTS:
668 nexthdr = ptr[0];
669 len = (ptr[1] + 1) << 3;
670 if (np->rxopt.bits.dstopts)
671 put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
672 break;
673 case IPPROTO_ROUTING:
674 nexthdr = ptr[0];
675 len = (ptr[1] + 1) << 3;
676 if (np->rxopt.bits.srcrt)
677 put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
678 break;
679 case IPPROTO_AH:
680 nexthdr = ptr[0];
681 len = (ptr[1] + 2) << 2;
682 break;
683 default:
684 nexthdr = ptr[0];
685 len = (ptr[1] + 1) << 3;
686 break;
687 }
688
689 off += len;
690 }
691 }
692
693 /* socket options in old style */
694 if (np->rxopt.bits.rxoinfo) {
695 struct in6_pktinfo src_info;
696
697 src_info.ipi6_ifindex = opt->iif;
698 src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
699 put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
700 }
701 if (np->rxopt.bits.rxohlim) {
702 int hlim = ipv6_hdr(skb)->hop_limit;
703 put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
704 }
705 if (np->rxopt.bits.ohopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
706 u8 *ptr = nh + sizeof(struct ipv6hdr);
707 put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
708 }
709 if (np->rxopt.bits.odstopts && opt->dst0) {
710 u8 *ptr = nh + opt->dst0;
711 put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
712 }
713 if (np->rxopt.bits.osrcrt && opt->srcrt) {
714 struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt);
715 put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
716 }
717 if (np->rxopt.bits.odstopts && opt->dst1) {
718 u8 *ptr = nh + opt->dst1;
719 put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
720 }
721 if (np->rxopt.bits.rxorigdstaddr) {
722 struct sockaddr_in6 sin6;
723 __be16 _ports[2], *ports;
724
725 ports = skb_header_pointer(skb, skb_transport_offset(skb),
726 sizeof(_ports), &_ports);
727 if (ports) {
728 /* All current transport protocols have the port numbers in the
729 * first four bytes of the transport header and this function is
730 * written with this assumption in mind.
731 */
732 sin6.sin6_family = AF_INET6;
733 sin6.sin6_addr = ipv6_hdr(skb)->daddr;
734 sin6.sin6_port = ports[1];
735 sin6.sin6_flowinfo = 0;
736 sin6.sin6_scope_id =
737 ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr,
738 opt->iif);
739
740 put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
741 }
742 }
743 if (np->rxopt.bits.recvfragsize && opt->frag_max_size) {
744 int val = opt->frag_max_size;
745
746 put_cmsg(msg, SOL_IPV6, IPV6_RECVFRAGSIZE, sizeof(val), &val);
747 }
748 }
749
ip6_datagram_recv_ctl(struct sock * sk,struct msghdr * msg,struct sk_buff * skb)750 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
751 struct sk_buff *skb)
752 {
753 ip6_datagram_recv_common_ctl(sk, msg, skb);
754 ip6_datagram_recv_specific_ctl(sk, msg, skb);
755 }
756 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
757
ip6_datagram_send_ctl(struct net * net,struct sock * sk,struct msghdr * msg,struct flowi6 * fl6,struct ipcm6_cookie * ipc6)758 int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
759 struct msghdr *msg, struct flowi6 *fl6,
760 struct ipcm6_cookie *ipc6)
761 {
762 struct in6_pktinfo *src_info;
763 struct cmsghdr *cmsg;
764 struct ipv6_rt_hdr *rthdr;
765 struct ipv6_opt_hdr *hdr;
766 struct ipv6_txoptions *opt = ipc6->opt;
767 int len;
768 int err = 0;
769
770 for_each_cmsghdr(cmsg, msg) {
771 int addr_type;
772
773 if (!CMSG_OK(msg, cmsg)) {
774 err = -EINVAL;
775 goto exit_f;
776 }
777
778 if (cmsg->cmsg_level == SOL_SOCKET) {
779 err = __sock_cmsg_send(sk, msg, cmsg, &ipc6->sockc);
780 if (err)
781 return err;
782 continue;
783 }
784
785 if (cmsg->cmsg_level != SOL_IPV6)
786 continue;
787
788 switch (cmsg->cmsg_type) {
789 case IPV6_PKTINFO:
790 case IPV6_2292PKTINFO:
791 {
792 struct net_device *dev = NULL;
793 int src_idx;
794
795 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
796 err = -EINVAL;
797 goto exit_f;
798 }
799
800 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
801 src_idx = src_info->ipi6_ifindex;
802
803 if (src_idx) {
804 if (fl6->flowi6_oif &&
805 src_idx != fl6->flowi6_oif &&
806 (READ_ONCE(sk->sk_bound_dev_if) != fl6->flowi6_oif ||
807 !sk_dev_equal_l3scope(sk, src_idx)))
808 return -EINVAL;
809 fl6->flowi6_oif = src_idx;
810 }
811
812 addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
813
814 rcu_read_lock();
815 if (fl6->flowi6_oif) {
816 dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
817 if (!dev) {
818 rcu_read_unlock();
819 return -ENODEV;
820 }
821 } else if (addr_type & IPV6_ADDR_LINKLOCAL) {
822 rcu_read_unlock();
823 return -EINVAL;
824 }
825
826 if (addr_type != IPV6_ADDR_ANY) {
827 int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
828 if (!ipv6_can_nonlocal_bind(net, inet_sk(sk)) &&
829 !ipv6_chk_addr_and_flags(net, &src_info->ipi6_addr,
830 dev, !strict, 0,
831 IFA_F_TENTATIVE) &&
832 !ipv6_chk_acast_addr_src(net, dev,
833 &src_info->ipi6_addr))
834 err = -EINVAL;
835 else
836 fl6->saddr = src_info->ipi6_addr;
837 }
838
839 rcu_read_unlock();
840
841 if (err)
842 goto exit_f;
843
844 break;
845 }
846
847 case IPV6_FLOWINFO:
848 if (cmsg->cmsg_len < CMSG_LEN(4)) {
849 err = -EINVAL;
850 goto exit_f;
851 }
852
853 if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
854 if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
855 err = -EINVAL;
856 goto exit_f;
857 }
858 }
859 fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
860 break;
861
862 case IPV6_2292HOPOPTS:
863 case IPV6_HOPOPTS:
864 if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
865 err = -EINVAL;
866 goto exit_f;
867 }
868
869 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
870 len = ((hdr->hdrlen + 1) << 3);
871 if (cmsg->cmsg_len < CMSG_LEN(len)) {
872 err = -EINVAL;
873 goto exit_f;
874 }
875 if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
876 err = -EPERM;
877 goto exit_f;
878 }
879 opt->opt_nflen += len;
880 opt->hopopt = hdr;
881 break;
882
883 case IPV6_2292DSTOPTS:
884 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
885 err = -EINVAL;
886 goto exit_f;
887 }
888
889 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
890 len = ((hdr->hdrlen + 1) << 3);
891 if (cmsg->cmsg_len < CMSG_LEN(len)) {
892 err = -EINVAL;
893 goto exit_f;
894 }
895 if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
896 err = -EPERM;
897 goto exit_f;
898 }
899 if (opt->dst1opt) {
900 err = -EINVAL;
901 goto exit_f;
902 }
903 opt->opt_flen += len;
904 opt->dst1opt = hdr;
905 break;
906
907 case IPV6_DSTOPTS:
908 case IPV6_RTHDRDSTOPTS:
909 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
910 err = -EINVAL;
911 goto exit_f;
912 }
913
914 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
915 len = ((hdr->hdrlen + 1) << 3);
916 if (cmsg->cmsg_len < CMSG_LEN(len)) {
917 err = -EINVAL;
918 goto exit_f;
919 }
920 if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
921 err = -EPERM;
922 goto exit_f;
923 }
924 if (cmsg->cmsg_type == IPV6_DSTOPTS) {
925 opt->opt_flen += len;
926 opt->dst1opt = hdr;
927 } else {
928 opt->opt_nflen += len;
929 opt->dst0opt = hdr;
930 }
931 break;
932
933 case IPV6_2292RTHDR:
934 case IPV6_RTHDR:
935 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
936 err = -EINVAL;
937 goto exit_f;
938 }
939
940 rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
941
942 switch (rthdr->type) {
943 #if IS_ENABLED(CONFIG_IPV6_MIP6)
944 case IPV6_SRCRT_TYPE_2:
945 if (rthdr->hdrlen != 2 ||
946 rthdr->segments_left != 1) {
947 err = -EINVAL;
948 goto exit_f;
949 }
950 break;
951 #endif
952 default:
953 err = -EINVAL;
954 goto exit_f;
955 }
956
957 len = ((rthdr->hdrlen + 1) << 3);
958
959 if (cmsg->cmsg_len < CMSG_LEN(len)) {
960 err = -EINVAL;
961 goto exit_f;
962 }
963
964 /* segments left must also match */
965 if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
966 err = -EINVAL;
967 goto exit_f;
968 }
969
970 opt->opt_nflen += len;
971 opt->srcrt = rthdr;
972
973 if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
974 int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
975
976 opt->opt_nflen += dsthdrlen;
977 opt->dst0opt = opt->dst1opt;
978 opt->dst1opt = NULL;
979 opt->opt_flen -= dsthdrlen;
980 }
981
982 break;
983
984 case IPV6_2292HOPLIMIT:
985 case IPV6_HOPLIMIT:
986 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
987 err = -EINVAL;
988 goto exit_f;
989 }
990
991 ipc6->hlimit = *(int *)CMSG_DATA(cmsg);
992 if (ipc6->hlimit < -1 || ipc6->hlimit > 0xff) {
993 err = -EINVAL;
994 goto exit_f;
995 }
996
997 break;
998
999 case IPV6_TCLASS:
1000 {
1001 int tc;
1002
1003 err = -EINVAL;
1004 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
1005 goto exit_f;
1006
1007 tc = *(int *)CMSG_DATA(cmsg);
1008 if (tc < -1 || tc > 0xff)
1009 goto exit_f;
1010
1011 err = 0;
1012 ipc6->tclass = tc;
1013
1014 break;
1015 }
1016
1017 case IPV6_DONTFRAG:
1018 {
1019 int df;
1020
1021 err = -EINVAL;
1022 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
1023 goto exit_f;
1024
1025 df = *(int *)CMSG_DATA(cmsg);
1026 if (df < 0 || df > 1)
1027 goto exit_f;
1028
1029 err = 0;
1030 ipc6->dontfrag = df;
1031
1032 break;
1033 }
1034 default:
1035 net_dbg_ratelimited("invalid cmsg type: %d\n",
1036 cmsg->cmsg_type);
1037 err = -EINVAL;
1038 goto exit_f;
1039 }
1040 }
1041
1042 exit_f:
1043 return err;
1044 }
1045 EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl);
1046
__ip6_dgram_sock_seq_show(struct seq_file * seq,struct sock * sp,__u16 srcp,__u16 destp,int rqueue,int bucket)1047 void __ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp,
1048 __u16 srcp, __u16 destp, int rqueue, int bucket)
1049 {
1050 const struct in6_addr *dest, *src;
1051
1052 dest = &sp->sk_v6_daddr;
1053 src = &sp->sk_v6_rcv_saddr;
1054 seq_printf(seq,
1055 "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1056 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u\n",
1057 bucket,
1058 src->s6_addr32[0], src->s6_addr32[1],
1059 src->s6_addr32[2], src->s6_addr32[3], srcp,
1060 dest->s6_addr32[0], dest->s6_addr32[1],
1061 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1062 sp->sk_state,
1063 sk_wmem_alloc_get(sp),
1064 rqueue,
1065 0, 0L, 0,
1066 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1067 0,
1068 sock_i_ino(sp),
1069 refcount_read(&sp->sk_refcnt), sp,
1070 atomic_read(&sp->sk_drops));
1071 }
1072