1 /*
2 * UDP over IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Based on linux/ipv4/udp.c
9 *
10 * Fixes:
11 * Hideaki YOSHIFUJI : sin6_scope_id support
12 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
13 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind
14 * a single port at the same time.
15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
16 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/udp6 to seq_file.
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 */
23
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/ipv6.h>
33 #include <linux/icmpv6.h>
34 #include <linux/init.h>
35 #include <linux/module.h>
36 #include <linux/skbuff.h>
37 #include <asm/uaccess.h>
38
39 #include <net/ndisc.h>
40 #include <net/protocol.h>
41 #include <net/transp_v6.h>
42 #include <net/ip6_route.h>
43 #include <net/raw.h>
44 #include <net/tcp_states.h>
45 #include <net/ip6_checksum.h>
46 #include <net/xfrm.h>
47
48 #include <linux/proc_fs.h>
49 #include <linux/seq_file.h>
50 #include "udp_impl.h"
51
udp_v6_get_port(struct sock * sk,unsigned short snum)52 int udp_v6_get_port(struct sock *sk, unsigned short snum)
53 {
54 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal);
55 }
56
compute_score(struct sock * sk,struct net * net,unsigned short hnum,struct in6_addr * saddr,__be16 sport,struct in6_addr * daddr,__be16 dport,int dif)57 static inline int compute_score(struct sock *sk, struct net *net,
58 unsigned short hnum,
59 struct in6_addr *saddr, __be16 sport,
60 struct in6_addr *daddr, __be16 dport,
61 int dif)
62 {
63 int score = -1;
64
65 if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum &&
66 sk->sk_family == PF_INET6) {
67 struct ipv6_pinfo *np = inet6_sk(sk);
68 struct inet_sock *inet = inet_sk(sk);
69
70 score = 0;
71 if (inet->dport) {
72 if (inet->dport != sport)
73 return -1;
74 score++;
75 }
76 if (!ipv6_addr_any(&np->rcv_saddr)) {
77 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
78 return -1;
79 score++;
80 }
81 if (!ipv6_addr_any(&np->daddr)) {
82 if (!ipv6_addr_equal(&np->daddr, saddr))
83 return -1;
84 score++;
85 }
86 if (sk->sk_bound_dev_if) {
87 if (sk->sk_bound_dev_if != dif)
88 return -1;
89 score++;
90 }
91 }
92 return score;
93 }
94
__udp6_lib_lookup(struct net * net,struct in6_addr * saddr,__be16 sport,struct in6_addr * daddr,__be16 dport,int dif,struct udp_table * udptable)95 static struct sock *__udp6_lib_lookup(struct net *net,
96 struct in6_addr *saddr, __be16 sport,
97 struct in6_addr *daddr, __be16 dport,
98 int dif, struct udp_table *udptable)
99 {
100 struct sock *sk, *result;
101 struct hlist_nulls_node *node;
102 unsigned short hnum = ntohs(dport);
103 unsigned int hash = udp_hashfn(net, hnum);
104 struct udp_hslot *hslot = &udptable->hash[hash];
105 int score, badness;
106
107 rcu_read_lock();
108 begin:
109 result = NULL;
110 badness = -1;
111 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
112 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
113 if (score > badness) {
114 result = sk;
115 badness = score;
116 }
117 }
118 /*
119 * if the nulls value we got at the end of this lookup is
120 * not the expected one, we must restart lookup.
121 * We probably met an item that was moved to another chain.
122 */
123 if (get_nulls_value(node) != hash)
124 goto begin;
125
126 if (result) {
127 if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt)))
128 result = NULL;
129 else if (unlikely(compute_score(result, net, hnum, saddr, sport,
130 daddr, dport, dif) < badness)) {
131 sock_put(result);
132 goto begin;
133 }
134 }
135 rcu_read_unlock();
136 return result;
137 }
138
__udp6_lib_lookup_skb(struct sk_buff * skb,__be16 sport,__be16 dport,struct udp_table * udptable)139 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
140 __be16 sport, __be16 dport,
141 struct udp_table *udptable)
142 {
143 struct sock *sk;
144 struct ipv6hdr *iph = ipv6_hdr(skb);
145
146 if (unlikely(sk = skb_steal_sock(skb)))
147 return sk;
148 else
149 return __udp6_lib_lookup(dev_net(skb->dst->dev), &iph->saddr, sport,
150 &iph->daddr, dport, inet6_iif(skb),
151 udptable);
152 }
153
154 /*
155 * This should be easy, if there is something there we
156 * return it, otherwise we block.
157 */
158
udpv6_recvmsg(struct kiocb * iocb,struct sock * sk,struct msghdr * msg,size_t len,int noblock,int flags,int * addr_len)159 int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
160 struct msghdr *msg, size_t len,
161 int noblock, int flags, int *addr_len)
162 {
163 struct ipv6_pinfo *np = inet6_sk(sk);
164 struct inet_sock *inet = inet_sk(sk);
165 struct sk_buff *skb;
166 unsigned int ulen, copied;
167 int peeked;
168 int err;
169 int is_udplite = IS_UDPLITE(sk);
170 int is_udp4;
171
172 if (addr_len)
173 *addr_len=sizeof(struct sockaddr_in6);
174
175 if (flags & MSG_ERRQUEUE)
176 return ipv6_recv_error(sk, msg, len);
177
178 try_again:
179 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
180 &peeked, &err);
181 if (!skb)
182 goto out;
183
184 ulen = skb->len - sizeof(struct udphdr);
185 copied = len;
186 if (copied > ulen)
187 copied = ulen;
188 else if (copied < ulen)
189 msg->msg_flags |= MSG_TRUNC;
190
191 is_udp4 = (skb->protocol == htons(ETH_P_IP));
192
193 /*
194 * If checksum is needed at all, try to do it while copying the
195 * data. If the data is truncated, or if we only want a partial
196 * coverage checksum (UDP-Lite), do it before the copy.
197 */
198
199 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
200 if (udp_lib_checksum_complete(skb))
201 goto csum_copy_err;
202 }
203
204 if (skb_csum_unnecessary(skb))
205 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
206 msg->msg_iov, copied );
207 else {
208 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
209 if (err == -EINVAL)
210 goto csum_copy_err;
211 }
212 if (err)
213 goto out_free;
214
215 if (!peeked) {
216 if (is_udp4)
217 UDP_INC_STATS_USER(sock_net(sk),
218 UDP_MIB_INDATAGRAMS, is_udplite);
219 else
220 UDP6_INC_STATS_USER(sock_net(sk),
221 UDP_MIB_INDATAGRAMS, is_udplite);
222 }
223
224 sock_recv_timestamp(msg, sk, skb);
225
226 /* Copy the address. */
227 if (msg->msg_name) {
228 struct sockaddr_in6 *sin6;
229
230 sin6 = (struct sockaddr_in6 *) msg->msg_name;
231 sin6->sin6_family = AF_INET6;
232 sin6->sin6_port = udp_hdr(skb)->source;
233 sin6->sin6_flowinfo = 0;
234 sin6->sin6_scope_id = 0;
235
236 if (is_udp4)
237 ipv6_addr_set(&sin6->sin6_addr, 0, 0,
238 htonl(0xffff), ip_hdr(skb)->saddr);
239 else {
240 ipv6_addr_copy(&sin6->sin6_addr,
241 &ipv6_hdr(skb)->saddr);
242 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
243 sin6->sin6_scope_id = IP6CB(skb)->iif;
244 }
245
246 }
247 if (is_udp4) {
248 if (inet->cmsg_flags)
249 ip_cmsg_recv(msg, skb);
250 } else {
251 if (np->rxopt.all)
252 datagram_recv_ctl(sk, msg, skb);
253 }
254
255 err = copied;
256 if (flags & MSG_TRUNC)
257 err = ulen;
258
259 out_free:
260 lock_sock(sk);
261 skb_free_datagram(sk, skb);
262 release_sock(sk);
263 out:
264 return err;
265
266 csum_copy_err:
267 lock_sock(sk);
268 if (!skb_kill_datagram(sk, skb, flags)) {
269 if (is_udp4)
270 UDP_INC_STATS_USER(sock_net(sk),
271 UDP_MIB_INERRORS, is_udplite);
272 else
273 UDP6_INC_STATS_USER(sock_net(sk),
274 UDP_MIB_INERRORS, is_udplite);
275 }
276 release_sock(sk);
277
278 if (flags & MSG_DONTWAIT)
279 return -EAGAIN;
280 goto try_again;
281 }
282
__udp6_lib_err(struct sk_buff * skb,struct inet6_skb_parm * opt,int type,int code,int offset,__be32 info,struct udp_table * udptable)283 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
284 int type, int code, int offset, __be32 info,
285 struct udp_table *udptable)
286 {
287 struct ipv6_pinfo *np;
288 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
289 struct in6_addr *saddr = &hdr->saddr;
290 struct in6_addr *daddr = &hdr->daddr;
291 struct udphdr *uh = (struct udphdr*)(skb->data+offset);
292 struct sock *sk;
293 int err;
294
295 sk = __udp6_lib_lookup(dev_net(skb->dev), daddr, uh->dest,
296 saddr, uh->source, inet6_iif(skb), udptable);
297 if (sk == NULL)
298 return;
299
300 np = inet6_sk(sk);
301
302 if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
303 goto out;
304
305 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
306 goto out;
307
308 if (np->recverr)
309 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
310
311 sk->sk_err = err;
312 sk->sk_error_report(sk);
313 out:
314 sock_put(sk);
315 }
316
udpv6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,int type,int code,int offset,__be32 info)317 static __inline__ void udpv6_err(struct sk_buff *skb,
318 struct inet6_skb_parm *opt, int type,
319 int code, int offset, __be32 info )
320 {
321 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
322 }
323
udpv6_queue_rcv_skb(struct sock * sk,struct sk_buff * skb)324 int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
325 {
326 struct udp_sock *up = udp_sk(sk);
327 int rc;
328 int is_udplite = IS_UDPLITE(sk);
329
330 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
331 goto drop;
332
333 /*
334 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
335 */
336 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
337
338 if (up->pcrlen == 0) { /* full coverage was set */
339 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage"
340 " %d while full coverage %d requested\n",
341 UDP_SKB_CB(skb)->cscov, skb->len);
342 goto drop;
343 }
344 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
345 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d "
346 "too small, need min %d\n",
347 UDP_SKB_CB(skb)->cscov, up->pcrlen);
348 goto drop;
349 }
350 }
351
352 if (sk->sk_filter) {
353 if (udp_lib_checksum_complete(skb))
354 goto drop;
355 }
356
357 if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) {
358 /* Note that an ENOMEM error is charged twice */
359 if (rc == -ENOMEM) {
360 UDP6_INC_STATS_BH(sock_net(sk),
361 UDP_MIB_RCVBUFERRORS, is_udplite);
362 atomic_inc(&sk->sk_drops);
363 }
364 goto drop;
365 }
366
367 return 0;
368 drop:
369 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
370 kfree_skb(skb);
371 return -1;
372 }
373
udp_v6_mcast_next(struct net * net,struct sock * sk,__be16 loc_port,struct in6_addr * loc_addr,__be16 rmt_port,struct in6_addr * rmt_addr,int dif)374 static struct sock *udp_v6_mcast_next(struct net *net, struct sock *sk,
375 __be16 loc_port, struct in6_addr *loc_addr,
376 __be16 rmt_port, struct in6_addr *rmt_addr,
377 int dif)
378 {
379 struct hlist_nulls_node *node;
380 struct sock *s = sk;
381 unsigned short num = ntohs(loc_port);
382
383 sk_nulls_for_each_from(s, node) {
384 struct inet_sock *inet = inet_sk(s);
385
386 if (!net_eq(sock_net(s), net))
387 continue;
388
389 if (s->sk_hash == num && s->sk_family == PF_INET6) {
390 struct ipv6_pinfo *np = inet6_sk(s);
391 if (inet->dport) {
392 if (inet->dport != rmt_port)
393 continue;
394 }
395 if (!ipv6_addr_any(&np->daddr) &&
396 !ipv6_addr_equal(&np->daddr, rmt_addr))
397 continue;
398
399 if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
400 continue;
401
402 if (!ipv6_addr_any(&np->rcv_saddr)) {
403 if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
404 continue;
405 }
406 if (!inet6_mc_check(s, loc_addr, rmt_addr))
407 continue;
408 return s;
409 }
410 }
411 return NULL;
412 }
413
414 /*
415 * Note: called only from the BH handler context,
416 * so we don't need to lock the hashes.
417 */
__udp6_lib_mcast_deliver(struct net * net,struct sk_buff * skb,struct in6_addr * saddr,struct in6_addr * daddr,struct udp_table * udptable)418 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
419 struct in6_addr *saddr, struct in6_addr *daddr,
420 struct udp_table *udptable)
421 {
422 struct sock *sk, *sk2;
423 const struct udphdr *uh = udp_hdr(skb);
424 struct udp_hslot *hslot = &udptable->hash[udp_hashfn(net, ntohs(uh->dest))];
425 int dif;
426
427 spin_lock(&hslot->lock);
428 sk = sk_nulls_head(&hslot->head);
429 dif = inet6_iif(skb);
430 sk = udp_v6_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
431 if (!sk) {
432 kfree_skb(skb);
433 goto out;
434 }
435
436 sk2 = sk;
437 while ((sk2 = udp_v6_mcast_next(net, sk_nulls_next(sk2), uh->dest, daddr,
438 uh->source, saddr, dif))) {
439 struct sk_buff *buff = skb_clone(skb, GFP_ATOMIC);
440 if (buff) {
441 bh_lock_sock(sk2);
442 if (!sock_owned_by_user(sk2))
443 udpv6_queue_rcv_skb(sk2, buff);
444 else
445 sk_add_backlog(sk2, buff);
446 bh_unlock_sock(sk2);
447 }
448 }
449 bh_lock_sock(sk);
450 if (!sock_owned_by_user(sk))
451 udpv6_queue_rcv_skb(sk, skb);
452 else
453 sk_add_backlog(sk, skb);
454 bh_unlock_sock(sk);
455 out:
456 spin_unlock(&hslot->lock);
457 return 0;
458 }
459
udp6_csum_init(struct sk_buff * skb,struct udphdr * uh,int proto)460 static inline int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh,
461 int proto)
462 {
463 int err;
464
465 UDP_SKB_CB(skb)->partial_cov = 0;
466 UDP_SKB_CB(skb)->cscov = skb->len;
467
468 if (proto == IPPROTO_UDPLITE) {
469 err = udplite_checksum_init(skb, uh);
470 if (err)
471 return err;
472 }
473
474 if (uh->check == 0) {
475 /* RFC 2460 section 8.1 says that we SHOULD log
476 this error. Well, it is reasonable.
477 */
478 LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n");
479 return 1;
480 }
481 if (skb->ip_summed == CHECKSUM_COMPLETE &&
482 !csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
483 skb->len, proto, skb->csum))
484 skb->ip_summed = CHECKSUM_UNNECESSARY;
485
486 if (!skb_csum_unnecessary(skb))
487 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
488 &ipv6_hdr(skb)->daddr,
489 skb->len, proto, 0));
490
491 return 0;
492 }
493
__udp6_lib_rcv(struct sk_buff * skb,struct udp_table * udptable,int proto)494 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
495 int proto)
496 {
497 struct sock *sk;
498 struct udphdr *uh;
499 struct net_device *dev = skb->dev;
500 struct in6_addr *saddr, *daddr;
501 u32 ulen = 0;
502 struct net *net = dev_net(skb->dev);
503
504 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
505 goto short_packet;
506
507 saddr = &ipv6_hdr(skb)->saddr;
508 daddr = &ipv6_hdr(skb)->daddr;
509 uh = udp_hdr(skb);
510
511 ulen = ntohs(uh->len);
512 if (ulen > skb->len)
513 goto short_packet;
514
515 if (proto == IPPROTO_UDP) {
516 /* UDP validates ulen. */
517
518 /* Check for jumbo payload */
519 if (ulen == 0)
520 ulen = skb->len;
521
522 if (ulen < sizeof(*uh))
523 goto short_packet;
524
525 if (ulen < skb->len) {
526 if (pskb_trim_rcsum(skb, ulen))
527 goto short_packet;
528 saddr = &ipv6_hdr(skb)->saddr;
529 daddr = &ipv6_hdr(skb)->daddr;
530 uh = udp_hdr(skb);
531 }
532 }
533
534 if (udp6_csum_init(skb, uh, proto))
535 goto discard;
536
537 /*
538 * Multicast receive code
539 */
540 if (ipv6_addr_is_multicast(daddr))
541 return __udp6_lib_mcast_deliver(net, skb,
542 saddr, daddr, udptable);
543
544 /* Unicast */
545
546 /*
547 * check socket cache ... must talk to Alan about his plans
548 * for sock caches... i'll skip this for now.
549 */
550 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
551
552 if (sk == NULL) {
553 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
554 goto discard;
555
556 if (udp_lib_checksum_complete(skb))
557 goto discard;
558 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS,
559 proto == IPPROTO_UDPLITE);
560
561 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0, dev);
562
563 kfree_skb(skb);
564 return 0;
565 }
566
567 /* deliver */
568
569 bh_lock_sock(sk);
570 if (!sock_owned_by_user(sk))
571 udpv6_queue_rcv_skb(sk, skb);
572 else
573 sk_add_backlog(sk, skb);
574 bh_unlock_sock(sk);
575 sock_put(sk);
576 return 0;
577
578 short_packet:
579 LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: %d/%u\n",
580 proto == IPPROTO_UDPLITE ? "-Lite" : "",
581 ulen, skb->len);
582
583 discard:
584 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
585 kfree_skb(skb);
586 return 0;
587 }
588
udpv6_rcv(struct sk_buff * skb)589 static __inline__ int udpv6_rcv(struct sk_buff *skb)
590 {
591 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
592 }
593
594 /*
595 * Throw away all pending data and cancel the corking. Socket is locked.
596 */
udp_v6_flush_pending_frames(struct sock * sk)597 static void udp_v6_flush_pending_frames(struct sock *sk)
598 {
599 struct udp_sock *up = udp_sk(sk);
600
601 if (up->pending == AF_INET)
602 udp_flush_pending_frames(sk);
603 else if (up->pending) {
604 up->len = 0;
605 up->pending = 0;
606 ip6_flush_pending_frames(sk);
607 }
608 }
609
610 /*
611 * Sending
612 */
613
udp_v6_push_pending_frames(struct sock * sk)614 static int udp_v6_push_pending_frames(struct sock *sk)
615 {
616 struct sk_buff *skb;
617 struct udphdr *uh;
618 struct udp_sock *up = udp_sk(sk);
619 struct inet_sock *inet = inet_sk(sk);
620 struct flowi *fl = &inet->cork.fl;
621 int err = 0;
622 int is_udplite = IS_UDPLITE(sk);
623 __wsum csum = 0;
624
625 /* Grab the skbuff where UDP header space exists. */
626 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
627 goto out;
628
629 /*
630 * Create a UDP header
631 */
632 uh = udp_hdr(skb);
633 uh->source = fl->fl_ip_sport;
634 uh->dest = fl->fl_ip_dport;
635 uh->len = htons(up->len);
636 uh->check = 0;
637
638 if (is_udplite)
639 csum = udplite_csum_outgoing(sk, skb);
640 else
641 csum = udp_csum_outgoing(sk, skb);
642
643 /* add protocol-dependent pseudo-header */
644 uh->check = csum_ipv6_magic(&fl->fl6_src, &fl->fl6_dst,
645 up->len, fl->proto, csum );
646 if (uh->check == 0)
647 uh->check = CSUM_MANGLED_0;
648
649 err = ip6_push_pending_frames(sk);
650 out:
651 up->len = 0;
652 up->pending = 0;
653 if (!err)
654 UDP6_INC_STATS_USER(sock_net(sk),
655 UDP_MIB_OUTDATAGRAMS, is_udplite);
656 return err;
657 }
658
udpv6_sendmsg(struct kiocb * iocb,struct sock * sk,struct msghdr * msg,size_t len)659 int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
660 struct msghdr *msg, size_t len)
661 {
662 struct ipv6_txoptions opt_space;
663 struct udp_sock *up = udp_sk(sk);
664 struct inet_sock *inet = inet_sk(sk);
665 struct ipv6_pinfo *np = inet6_sk(sk);
666 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
667 struct in6_addr *daddr, *final_p = NULL, final;
668 struct ipv6_txoptions *opt = NULL;
669 struct ip6_flowlabel *flowlabel = NULL;
670 struct flowi fl;
671 struct dst_entry *dst;
672 int addr_len = msg->msg_namelen;
673 int ulen = len;
674 int hlimit = -1;
675 int tclass = -1;
676 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
677 int err;
678 int connected = 0;
679 int is_udplite = IS_UDPLITE(sk);
680 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
681
682 /* destination address check */
683 if (sin6) {
684 if (addr_len < offsetof(struct sockaddr, sa_data))
685 return -EINVAL;
686
687 switch (sin6->sin6_family) {
688 case AF_INET6:
689 if (addr_len < SIN6_LEN_RFC2133)
690 return -EINVAL;
691 daddr = &sin6->sin6_addr;
692 break;
693 case AF_INET:
694 goto do_udp_sendmsg;
695 case AF_UNSPEC:
696 msg->msg_name = sin6 = NULL;
697 msg->msg_namelen = addr_len = 0;
698 daddr = NULL;
699 break;
700 default:
701 return -EINVAL;
702 }
703 } else if (!up->pending) {
704 if (sk->sk_state != TCP_ESTABLISHED)
705 return -EDESTADDRREQ;
706 daddr = &np->daddr;
707 } else
708 daddr = NULL;
709
710 if (daddr) {
711 if (ipv6_addr_v4mapped(daddr)) {
712 struct sockaddr_in sin;
713 sin.sin_family = AF_INET;
714 sin.sin_port = sin6 ? sin6->sin6_port : inet->dport;
715 sin.sin_addr.s_addr = daddr->s6_addr32[3];
716 msg->msg_name = &sin;
717 msg->msg_namelen = sizeof(sin);
718 do_udp_sendmsg:
719 if (__ipv6_only_sock(sk))
720 return -ENETUNREACH;
721 return udp_sendmsg(iocb, sk, msg, len);
722 }
723 }
724
725 if (up->pending == AF_INET)
726 return udp_sendmsg(iocb, sk, msg, len);
727
728 /* Rough check on arithmetic overflow,
729 better check is made in ip6_append_data().
730 */
731 if (len > INT_MAX - sizeof(struct udphdr))
732 return -EMSGSIZE;
733
734 if (up->pending) {
735 /*
736 * There are pending frames.
737 * The socket lock must be held while it's corked.
738 */
739 lock_sock(sk);
740 if (likely(up->pending)) {
741 if (unlikely(up->pending != AF_INET6)) {
742 release_sock(sk);
743 return -EAFNOSUPPORT;
744 }
745 dst = NULL;
746 goto do_append_data;
747 }
748 release_sock(sk);
749 }
750 ulen += sizeof(struct udphdr);
751
752 memset(&fl, 0, sizeof(fl));
753
754 if (sin6) {
755 if (sin6->sin6_port == 0)
756 return -EINVAL;
757
758 fl.fl_ip_dport = sin6->sin6_port;
759 daddr = &sin6->sin6_addr;
760
761 if (np->sndflow) {
762 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
763 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
764 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
765 if (flowlabel == NULL)
766 return -EINVAL;
767 daddr = &flowlabel->dst;
768 }
769 }
770
771 /*
772 * Otherwise it will be difficult to maintain
773 * sk->sk_dst_cache.
774 */
775 if (sk->sk_state == TCP_ESTABLISHED &&
776 ipv6_addr_equal(daddr, &np->daddr))
777 daddr = &np->daddr;
778
779 if (addr_len >= sizeof(struct sockaddr_in6) &&
780 sin6->sin6_scope_id &&
781 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
782 fl.oif = sin6->sin6_scope_id;
783 } else {
784 if (sk->sk_state != TCP_ESTABLISHED)
785 return -EDESTADDRREQ;
786
787 fl.fl_ip_dport = inet->dport;
788 daddr = &np->daddr;
789 fl.fl6_flowlabel = np->flow_label;
790 connected = 1;
791 }
792
793 if (!fl.oif)
794 fl.oif = sk->sk_bound_dev_if;
795
796 if (!fl.oif)
797 fl.oif = np->sticky_pktinfo.ipi6_ifindex;
798
799 if (msg->msg_controllen) {
800 opt = &opt_space;
801 memset(opt, 0, sizeof(struct ipv6_txoptions));
802 opt->tot_len = sizeof(*opt);
803
804 err = datagram_send_ctl(sock_net(sk), msg, &fl, opt, &hlimit, &tclass);
805 if (err < 0) {
806 fl6_sock_release(flowlabel);
807 return err;
808 }
809 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
810 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
811 if (flowlabel == NULL)
812 return -EINVAL;
813 }
814 if (!(opt->opt_nflen|opt->opt_flen))
815 opt = NULL;
816 connected = 0;
817 }
818 if (opt == NULL)
819 opt = np->opt;
820 if (flowlabel)
821 opt = fl6_merge_options(&opt_space, flowlabel, opt);
822 opt = ipv6_fixup_options(&opt_space, opt);
823
824 fl.proto = sk->sk_protocol;
825 if (!ipv6_addr_any(daddr))
826 ipv6_addr_copy(&fl.fl6_dst, daddr);
827 else
828 fl.fl6_dst.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
829 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
830 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
831 fl.fl_ip_sport = inet->sport;
832
833 /* merge ip6_build_xmit from ip6_output */
834 if (opt && opt->srcrt) {
835 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
836 ipv6_addr_copy(&final, &fl.fl6_dst);
837 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
838 final_p = &final;
839 connected = 0;
840 }
841
842 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) {
843 fl.oif = np->mcast_oif;
844 connected = 0;
845 }
846
847 security_sk_classify_flow(sk, &fl);
848
849 err = ip6_sk_dst_lookup(sk, &dst, &fl);
850 if (err)
851 goto out;
852 if (final_p)
853 ipv6_addr_copy(&fl.fl6_dst, final_p);
854
855 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
856 if (err < 0) {
857 if (err == -EREMOTE)
858 err = ip6_dst_blackhole(sk, &dst, &fl);
859 if (err < 0)
860 goto out;
861 }
862
863 if (hlimit < 0) {
864 if (ipv6_addr_is_multicast(&fl.fl6_dst))
865 hlimit = np->mcast_hops;
866 else
867 hlimit = np->hop_limit;
868 if (hlimit < 0)
869 hlimit = ip6_dst_hoplimit(dst);
870 }
871
872 if (tclass < 0) {
873 tclass = np->tclass;
874 if (tclass < 0)
875 tclass = 0;
876 }
877
878 if (msg->msg_flags&MSG_CONFIRM)
879 goto do_confirm;
880 back_from_confirm:
881
882 lock_sock(sk);
883 if (unlikely(up->pending)) {
884 /* The socket is already corked while preparing it. */
885 /* ... which is an evident application bug. --ANK */
886 release_sock(sk);
887
888 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
889 err = -EINVAL;
890 goto out;
891 }
892
893 up->pending = AF_INET6;
894
895 do_append_data:
896 up->len += ulen;
897 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
898 err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
899 sizeof(struct udphdr), hlimit, tclass, opt, &fl,
900 (struct rt6_info*)dst,
901 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
902 if (err)
903 udp_v6_flush_pending_frames(sk);
904 else if (!corkreq)
905 err = udp_v6_push_pending_frames(sk);
906 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
907 up->pending = 0;
908
909 if (dst) {
910 if (connected) {
911 ip6_dst_store(sk, dst,
912 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
913 &np->daddr : NULL,
914 #ifdef CONFIG_IPV6_SUBTREES
915 ipv6_addr_equal(&fl.fl6_src, &np->saddr) ?
916 &np->saddr :
917 #endif
918 NULL);
919 } else {
920 dst_release(dst);
921 }
922 dst = NULL;
923 }
924
925 if (err > 0)
926 err = np->recverr ? net_xmit_errno(err) : 0;
927 release_sock(sk);
928 out:
929 dst_release(dst);
930 fl6_sock_release(flowlabel);
931 if (!err)
932 return len;
933 /*
934 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
935 * ENOBUFS might not be good (it's not tunable per se), but otherwise
936 * we don't have a good statistic (IpOutDiscards but it can be too many
937 * things). We could add another new stat but at least for now that
938 * seems like overkill.
939 */
940 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
941 UDP6_INC_STATS_USER(sock_net(sk),
942 UDP_MIB_SNDBUFERRORS, is_udplite);
943 }
944 return err;
945
946 do_confirm:
947 dst_confirm(dst);
948 if (!(msg->msg_flags&MSG_PROBE) || len)
949 goto back_from_confirm;
950 err = 0;
951 goto out;
952 }
953
udpv6_destroy_sock(struct sock * sk)954 void udpv6_destroy_sock(struct sock *sk)
955 {
956 lock_sock(sk);
957 udp_v6_flush_pending_frames(sk);
958 release_sock(sk);
959
960 inet6_destroy_sock(sk);
961 }
962
963 /*
964 * Socket option code for UDP
965 */
udpv6_setsockopt(struct sock * sk,int level,int optname,char __user * optval,int optlen)966 int udpv6_setsockopt(struct sock *sk, int level, int optname,
967 char __user *optval, int optlen)
968 {
969 if (level == SOL_UDP || level == SOL_UDPLITE)
970 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
971 udp_v6_push_pending_frames);
972 return ipv6_setsockopt(sk, level, optname, optval, optlen);
973 }
974
975 #ifdef CONFIG_COMPAT
compat_udpv6_setsockopt(struct sock * sk,int level,int optname,char __user * optval,int optlen)976 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
977 char __user *optval, int optlen)
978 {
979 if (level == SOL_UDP || level == SOL_UDPLITE)
980 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
981 udp_v6_push_pending_frames);
982 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
983 }
984 #endif
985
udpv6_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)986 int udpv6_getsockopt(struct sock *sk, int level, int optname,
987 char __user *optval, int __user *optlen)
988 {
989 if (level == SOL_UDP || level == SOL_UDPLITE)
990 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
991 return ipv6_getsockopt(sk, level, optname, optval, optlen);
992 }
993
994 #ifdef CONFIG_COMPAT
compat_udpv6_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)995 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
996 char __user *optval, int __user *optlen)
997 {
998 if (level == SOL_UDP || level == SOL_UDPLITE)
999 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1000 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1001 }
1002 #endif
1003
1004 static struct inet6_protocol udpv6_protocol = {
1005 .handler = udpv6_rcv,
1006 .err_handler = udpv6_err,
1007 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1008 };
1009
1010 /* ------------------------------------------------------------------------ */
1011 #ifdef CONFIG_PROC_FS
1012
udp6_sock_seq_show(struct seq_file * seq,struct sock * sp,int bucket)1013 static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket)
1014 {
1015 struct inet_sock *inet = inet_sk(sp);
1016 struct ipv6_pinfo *np = inet6_sk(sp);
1017 struct in6_addr *dest, *src;
1018 __u16 destp, srcp;
1019
1020 dest = &np->daddr;
1021 src = &np->rcv_saddr;
1022 destp = ntohs(inet->dport);
1023 srcp = ntohs(inet->sport);
1024 seq_printf(seq,
1025 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1026 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1027 bucket,
1028 src->s6_addr32[0], src->s6_addr32[1],
1029 src->s6_addr32[2], src->s6_addr32[3], srcp,
1030 dest->s6_addr32[0], dest->s6_addr32[1],
1031 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1032 sp->sk_state,
1033 atomic_read(&sp->sk_wmem_alloc),
1034 atomic_read(&sp->sk_rmem_alloc),
1035 0, 0L, 0,
1036 sock_i_uid(sp), 0,
1037 sock_i_ino(sp),
1038 atomic_read(&sp->sk_refcnt), sp,
1039 atomic_read(&sp->sk_drops));
1040 }
1041
udp6_seq_show(struct seq_file * seq,void * v)1042 int udp6_seq_show(struct seq_file *seq, void *v)
1043 {
1044 if (v == SEQ_START_TOKEN)
1045 seq_printf(seq,
1046 " sl "
1047 "local_address "
1048 "remote_address "
1049 "st tx_queue rx_queue tr tm->when retrnsmt"
1050 " uid timeout inode ref pointer drops\n");
1051 else
1052 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket);
1053 return 0;
1054 }
1055
1056 static struct udp_seq_afinfo udp6_seq_afinfo = {
1057 .name = "udp6",
1058 .family = AF_INET6,
1059 .udp_table = &udp_table,
1060 .seq_fops = {
1061 .owner = THIS_MODULE,
1062 },
1063 .seq_ops = {
1064 .show = udp6_seq_show,
1065 },
1066 };
1067
udp6_proc_init(struct net * net)1068 int udp6_proc_init(struct net *net)
1069 {
1070 return udp_proc_register(net, &udp6_seq_afinfo);
1071 }
1072
udp6_proc_exit(struct net * net)1073 void udp6_proc_exit(struct net *net) {
1074 udp_proc_unregister(net, &udp6_seq_afinfo);
1075 }
1076 #endif /* CONFIG_PROC_FS */
1077
1078 /* ------------------------------------------------------------------------ */
1079
1080 struct proto udpv6_prot = {
1081 .name = "UDPv6",
1082 .owner = THIS_MODULE,
1083 .close = udp_lib_close,
1084 .connect = ip6_datagram_connect,
1085 .disconnect = udp_disconnect,
1086 .ioctl = udp_ioctl,
1087 .destroy = udpv6_destroy_sock,
1088 .setsockopt = udpv6_setsockopt,
1089 .getsockopt = udpv6_getsockopt,
1090 .sendmsg = udpv6_sendmsg,
1091 .recvmsg = udpv6_recvmsg,
1092 .backlog_rcv = udpv6_queue_rcv_skb,
1093 .hash = udp_lib_hash,
1094 .unhash = udp_lib_unhash,
1095 .get_port = udp_v6_get_port,
1096 .memory_allocated = &udp_memory_allocated,
1097 .sysctl_mem = sysctl_udp_mem,
1098 .sysctl_wmem = &sysctl_udp_wmem_min,
1099 .sysctl_rmem = &sysctl_udp_rmem_min,
1100 .obj_size = sizeof(struct udp6_sock),
1101 .slab_flags = SLAB_DESTROY_BY_RCU,
1102 .h.udp_table = &udp_table,
1103 #ifdef CONFIG_COMPAT
1104 .compat_setsockopt = compat_udpv6_setsockopt,
1105 .compat_getsockopt = compat_udpv6_getsockopt,
1106 #endif
1107 };
1108
1109 static struct inet_protosw udpv6_protosw = {
1110 .type = SOCK_DGRAM,
1111 .protocol = IPPROTO_UDP,
1112 .prot = &udpv6_prot,
1113 .ops = &inet6_dgram_ops,
1114 .capability =-1,
1115 .no_check = UDP_CSUM_DEFAULT,
1116 .flags = INET_PROTOSW_PERMANENT,
1117 };
1118
1119
udpv6_init(void)1120 int __init udpv6_init(void)
1121 {
1122 int ret;
1123
1124 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1125 if (ret)
1126 goto out;
1127
1128 ret = inet6_register_protosw(&udpv6_protosw);
1129 if (ret)
1130 goto out_udpv6_protocol;
1131 out:
1132 return ret;
1133
1134 out_udpv6_protocol:
1135 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1136 goto out;
1137 }
1138
udpv6_exit(void)1139 void udpv6_exit(void)
1140 {
1141 inet6_unregister_protosw(&udpv6_protosw);
1142 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1143 }
1144