[IPV6]: Support several new sockopt / ancillary data in Advanced API (RFC3542).
[linux-2.6] / net / ipv6 / udp.c
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  *      $Id: udp.c,v 1.65 2002/02/01 22:01:04 davem Exp $
11  *
12  *      Fixes:
13  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
14  *      YOSHIFUJI Hideaki @USAGI and:   Support IPV6_V6ONLY socket option, which
15  *      Alexey Kuznetsov                allow both IPv4 and IPv6 sockets to bind
16  *                                      a single port at the same time.
17  *      Kazunori MIYAZAWA @USAGI:       change process style to use ip6_append_data
18  *      YOSHIFUJI Hideaki @USAGI:       convert /proc/net/udp6 to seq_file.
19  *
20  *      This program is free software; you can redistribute it and/or
21  *      modify it under the terms of the GNU General Public License
22  *      as published by the Free Software Foundation; either version
23  *      2 of the License, or (at your option) any later version.
24  */
25
26 #include <linux/config.h>
27 #include <linux/errno.h>
28 #include <linux/types.h>
29 #include <linux/socket.h>
30 #include <linux/sockios.h>
31 #include <linux/sched.h>
32 #include <linux/net.h>
33 #include <linux/in6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/ipv6.h>
37 #include <linux/icmpv6.h>
38 #include <linux/init.h>
39 #include <asm/uaccess.h>
40
41 #include <net/sock.h>
42 #include <net/snmp.h>
43
44 #include <net/ipv6.h>
45 #include <net/ndisc.h>
46 #include <net/protocol.h>
47 #include <net/transp_v6.h>
48 #include <net/ip6_route.h>
49 #include <net/addrconf.h>
50 #include <net/ip.h>
51 #include <net/udp.h>
52 #include <net/raw.h>
53 #include <net/inet_common.h>
54 #include <net/tcp_states.h>
55
56 #include <net/ip6_checksum.h>
57 #include <net/xfrm.h>
58
59 #include <linux/proc_fs.h>
60 #include <linux/seq_file.h>
61
62 DEFINE_SNMP_STAT(struct udp_mib, udp_stats_in6) __read_mostly;
63
64 /* Grrr, addr_type already calculated by caller, but I don't want
65  * to add some silly "cookie" argument to this method just for that.
66  */
67 static int udp_v6_get_port(struct sock *sk, unsigned short snum)
68 {
69         struct sock *sk2;
70         struct hlist_node *node;
71
72         write_lock_bh(&udp_hash_lock);
73         if (snum == 0) {
74                 int best_size_so_far, best, result, i;
75
76                 if (udp_port_rover > sysctl_local_port_range[1] ||
77                     udp_port_rover < sysctl_local_port_range[0])
78                         udp_port_rover = sysctl_local_port_range[0];
79                 best_size_so_far = 32767;
80                 best = result = udp_port_rover;
81                 for (i = 0; i < UDP_HTABLE_SIZE; i++, result++) {
82                         int size;
83                         struct hlist_head *list;
84
85                         list = &udp_hash[result & (UDP_HTABLE_SIZE - 1)];
86                         if (hlist_empty(list)) {
87                                 if (result > sysctl_local_port_range[1])
88                                         result = sysctl_local_port_range[0] +
89                                                 ((result - sysctl_local_port_range[0]) &
90                                                  (UDP_HTABLE_SIZE - 1));
91                                 goto gotit;
92                         }
93                         size = 0;
94                         sk_for_each(sk2, node, list)
95                                 if (++size >= best_size_so_far)
96                                         goto next;
97                         best_size_so_far = size;
98                         best = result;
99                 next:;
100                 }
101                 result = best;
102                 for(;; result += UDP_HTABLE_SIZE) {
103                         if (result > sysctl_local_port_range[1])
104                                 result = sysctl_local_port_range[0]
105                                         + ((result - sysctl_local_port_range[0]) &
106                                            (UDP_HTABLE_SIZE - 1));
107                         if (!udp_lport_inuse(result))
108                                 break;
109                 }
110 gotit:
111                 udp_port_rover = snum = result;
112         } else {
113                 sk_for_each(sk2, node,
114                             &udp_hash[snum & (UDP_HTABLE_SIZE - 1)]) {
115                         if (inet_sk(sk2)->num == snum &&
116                             sk2 != sk &&
117                             (!sk2->sk_bound_dev_if ||
118                              !sk->sk_bound_dev_if ||
119                              sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
120                             (!sk2->sk_reuse || !sk->sk_reuse) &&
121                             ipv6_rcv_saddr_equal(sk, sk2))
122                                 goto fail;
123                 }
124         }
125
126         inet_sk(sk)->num = snum;
127         if (sk_unhashed(sk)) {
128                 sk_add_node(sk, &udp_hash[snum & (UDP_HTABLE_SIZE - 1)]);
129                 sock_prot_inc_use(sk->sk_prot);
130         }
131         write_unlock_bh(&udp_hash_lock);
132         return 0;
133
134 fail:
135         write_unlock_bh(&udp_hash_lock);
136         return 1;
137 }
138
139 static void udp_v6_hash(struct sock *sk)
140 {
141         BUG();
142 }
143
144 static void udp_v6_unhash(struct sock *sk)
145 {
146         write_lock_bh(&udp_hash_lock);
147         if (sk_del_node_init(sk)) {
148                 inet_sk(sk)->num = 0;
149                 sock_prot_dec_use(sk->sk_prot);
150         }
151         write_unlock_bh(&udp_hash_lock);
152 }
153
154 static struct sock *udp_v6_lookup(struct in6_addr *saddr, u16 sport,
155                                   struct in6_addr *daddr, u16 dport, int dif)
156 {
157         struct sock *sk, *result = NULL;
158         struct hlist_node *node;
159         unsigned short hnum = ntohs(dport);
160         int badness = -1;
161
162         read_lock(&udp_hash_lock);
163         sk_for_each(sk, node, &udp_hash[hnum & (UDP_HTABLE_SIZE - 1)]) {
164                 struct inet_sock *inet = inet_sk(sk);
165
166                 if (inet->num == hnum && sk->sk_family == PF_INET6) {
167                         struct ipv6_pinfo *np = inet6_sk(sk);
168                         int score = 0;
169                         if (inet->dport) {
170                                 if (inet->dport != sport)
171                                         continue;
172                                 score++;
173                         }
174                         if (!ipv6_addr_any(&np->rcv_saddr)) {
175                                 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
176                                         continue;
177                                 score++;
178                         }
179                         if (!ipv6_addr_any(&np->daddr)) {
180                                 if (!ipv6_addr_equal(&np->daddr, saddr))
181                                         continue;
182                                 score++;
183                         }
184                         if (sk->sk_bound_dev_if) {
185                                 if (sk->sk_bound_dev_if != dif)
186                                         continue;
187                                 score++;
188                         }
189                         if(score == 4) {
190                                 result = sk;
191                                 break;
192                         } else if(score > badness) {
193                                 result = sk;
194                                 badness = score;
195                         }
196                 }
197         }
198         if (result)
199                 sock_hold(result);
200         read_unlock(&udp_hash_lock);
201         return result;
202 }
203
204 /*
205  *
206  */
207
208 static void udpv6_close(struct sock *sk, long timeout)
209 {
210         sk_common_release(sk);
211 }
212
213 /*
214  *      This should be easy, if there is something there we
215  *      return it, otherwise we block.
216  */
217
218 static int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk, 
219                   struct msghdr *msg, size_t len,
220                   int noblock, int flags, int *addr_len)
221 {
222         struct ipv6_pinfo *np = inet6_sk(sk);
223         struct inet_sock *inet = inet_sk(sk);
224         struct sk_buff *skb;
225         size_t copied;
226         int err;
227
228         if (addr_len)
229                 *addr_len=sizeof(struct sockaddr_in6);
230   
231         if (flags & MSG_ERRQUEUE)
232                 return ipv6_recv_error(sk, msg, len);
233
234 try_again:
235         skb = skb_recv_datagram(sk, flags, noblock, &err);
236         if (!skb)
237                 goto out;
238
239         copied = skb->len - sizeof(struct udphdr);
240         if (copied > len) {
241                 copied = len;
242                 msg->msg_flags |= MSG_TRUNC;
243         }
244
245         if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
246                 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
247                                               copied);
248         } else if (msg->msg_flags&MSG_TRUNC) {
249                 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum)))
250                         goto csum_copy_err;
251                 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
252                                               copied);
253         } else {
254                 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
255                 if (err == -EINVAL)
256                         goto csum_copy_err;
257         }
258         if (err)
259                 goto out_free;
260
261         sock_recv_timestamp(msg, sk, skb);
262
263         /* Copy the address. */
264         if (msg->msg_name) {
265                 struct sockaddr_in6 *sin6;
266           
267                 sin6 = (struct sockaddr_in6 *) msg->msg_name;
268                 sin6->sin6_family = AF_INET6;
269                 sin6->sin6_port = skb->h.uh->source;
270                 sin6->sin6_flowinfo = 0;
271                 sin6->sin6_scope_id = 0;
272
273                 if (skb->protocol == htons(ETH_P_IP))
274                         ipv6_addr_set(&sin6->sin6_addr, 0, 0,
275                                       htonl(0xffff), skb->nh.iph->saddr);
276                 else {
277                         ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
278                         if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
279                                 sin6->sin6_scope_id = IP6CB(skb)->iif;
280                 }
281
282         }
283         if (skb->protocol == htons(ETH_P_IP)) {
284                 if (inet->cmsg_flags)
285                         ip_cmsg_recv(msg, skb);
286         } else {
287                 if (np->rxopt.all)
288                         datagram_recv_ctl(sk, msg, skb);
289         }
290
291         err = copied;
292         if (flags & MSG_TRUNC)
293                 err = skb->len - sizeof(struct udphdr);
294
295 out_free:
296         skb_free_datagram(sk, skb);
297 out:
298         return err;
299
300 csum_copy_err:
301         /* Clear queue. */
302         if (flags&MSG_PEEK) {
303                 int clear = 0;
304                 spin_lock_bh(&sk->sk_receive_queue.lock);
305                 if (skb == skb_peek(&sk->sk_receive_queue)) {
306                         __skb_unlink(skb, &sk->sk_receive_queue);
307                         clear = 1;
308                 }
309                 spin_unlock_bh(&sk->sk_receive_queue.lock);
310                 if (clear)
311                         kfree_skb(skb);
312         }
313
314         skb_free_datagram(sk, skb);
315
316         if (flags & MSG_DONTWAIT) {
317                 UDP6_INC_STATS_USER(UDP_MIB_INERRORS);
318                 return -EAGAIN;
319         }
320         goto try_again;
321 }
322
323 static void udpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
324                int type, int code, int offset, __u32 info)
325 {
326         struct ipv6_pinfo *np;
327         struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
328         struct net_device *dev = skb->dev;
329         struct in6_addr *saddr = &hdr->saddr;
330         struct in6_addr *daddr = &hdr->daddr;
331         struct udphdr *uh = (struct udphdr*)(skb->data+offset);
332         struct sock *sk;
333         int err;
334
335         sk = udp_v6_lookup(daddr, uh->dest, saddr, uh->source, dev->ifindex);
336    
337         if (sk == NULL)
338                 return;
339
340         np = inet6_sk(sk);
341
342         if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
343                 goto out;
344
345         if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
346                 goto out;
347
348         if (np->recverr)
349                 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
350
351         sk->sk_err = err;
352         sk->sk_error_report(sk);
353 out:
354         sock_put(sk);
355 }
356
357 static inline int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
358 {
359         if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
360                 kfree_skb(skb);
361                 return -1;
362         }
363
364         if (skb->ip_summed != CHECKSUM_UNNECESSARY) {
365                 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
366                         UDP6_INC_STATS_BH(UDP_MIB_INERRORS);
367                         kfree_skb(skb);
368                         return 0;
369                 }
370                 skb->ip_summed = CHECKSUM_UNNECESSARY;
371         }
372
373         if (sock_queue_rcv_skb(sk,skb)<0) {
374                 UDP6_INC_STATS_BH(UDP_MIB_INERRORS);
375                 kfree_skb(skb);
376                 return 0;
377         }
378         UDP6_INC_STATS_BH(UDP_MIB_INDATAGRAMS);
379         return 0;
380 }
381
382 static struct sock *udp_v6_mcast_next(struct sock *sk,
383                                       u16 loc_port, struct in6_addr *loc_addr,
384                                       u16 rmt_port, struct in6_addr *rmt_addr,
385                                       int dif)
386 {
387         struct hlist_node *node;
388         struct sock *s = sk;
389         unsigned short num = ntohs(loc_port);
390
391         sk_for_each_from(s, node) {
392                 struct inet_sock *inet = inet_sk(s);
393
394                 if (inet->num == num && s->sk_family == PF_INET6) {
395                         struct ipv6_pinfo *np = inet6_sk(s);
396                         if (inet->dport) {
397                                 if (inet->dport != rmt_port)
398                                         continue;
399                         }
400                         if (!ipv6_addr_any(&np->daddr) &&
401                             !ipv6_addr_equal(&np->daddr, rmt_addr))
402                                 continue;
403
404                         if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
405                                 continue;
406
407                         if (!ipv6_addr_any(&np->rcv_saddr)) {
408                                 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
409                                         return s;
410                                 continue;
411                         }
412                         if(!inet6_mc_check(s, loc_addr, rmt_addr))
413                                 continue;
414                         return s;
415                 }
416         }
417         return NULL;
418 }
419
420 /*
421  * Note: called only from the BH handler context,
422  * so we don't need to lock the hashes.
423  */
424 static void udpv6_mcast_deliver(struct udphdr *uh,
425                                 struct in6_addr *saddr, struct in6_addr *daddr,
426                                 struct sk_buff *skb)
427 {
428         struct sock *sk, *sk2;
429         int dif;
430
431         read_lock(&udp_hash_lock);
432         sk = sk_head(&udp_hash[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
433         dif = skb->dev->ifindex;
434         sk = udp_v6_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
435         if (!sk) {
436                 kfree_skb(skb);
437                 goto out;
438         }
439
440         sk2 = sk;
441         while ((sk2 = udp_v6_mcast_next(sk_next(sk2), uh->dest, daddr,
442                                         uh->source, saddr, dif))) {
443                 struct sk_buff *buff = skb_clone(skb, GFP_ATOMIC);
444                 if (buff)
445                         udpv6_queue_rcv_skb(sk2, buff);
446         }
447         udpv6_queue_rcv_skb(sk, skb);
448 out:
449         read_unlock(&udp_hash_lock);
450 }
451
452 static int udpv6_rcv(struct sk_buff **pskb, unsigned int *nhoffp)
453 {
454         struct sk_buff *skb = *pskb;
455         struct sock *sk;
456         struct udphdr *uh;
457         struct net_device *dev = skb->dev;
458         struct in6_addr *saddr, *daddr;
459         u32 ulen = 0;
460
461         if (!pskb_may_pull(skb, sizeof(struct udphdr)))
462                 goto short_packet;
463
464         saddr = &skb->nh.ipv6h->saddr;
465         daddr = &skb->nh.ipv6h->daddr;
466         uh = skb->h.uh;
467
468         ulen = ntohs(uh->len);
469
470         /* Check for jumbo payload */
471         if (ulen == 0)
472                 ulen = skb->len;
473
474         if (ulen > skb->len || ulen < sizeof(*uh))
475                 goto short_packet;
476
477         if (uh->check == 0) {
478                 /* RFC 2460 section 8.1 says that we SHOULD log
479                    this error. Well, it is reasonable.
480                  */
481                 LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n");
482                 goto discard;
483         }
484
485         if (ulen < skb->len) {
486                 if (__pskb_trim(skb, ulen))
487                         goto discard;
488                 saddr = &skb->nh.ipv6h->saddr;
489                 daddr = &skb->nh.ipv6h->daddr;
490                 uh = skb->h.uh;
491         }
492
493         if (skb->ip_summed==CHECKSUM_HW) {
494                 skb->ip_summed = CHECKSUM_UNNECESSARY;
495                 if (csum_ipv6_magic(saddr, daddr, ulen, IPPROTO_UDP, skb->csum)) {
496                         LIMIT_NETDEBUG(KERN_DEBUG "udp v6 hw csum failure.\n");
497                         skb->ip_summed = CHECKSUM_NONE;
498                 }
499         }
500         if (skb->ip_summed != CHECKSUM_UNNECESSARY)
501                 skb->csum = ~csum_ipv6_magic(saddr, daddr, ulen, IPPROTO_UDP, 0);
502
503         /* 
504          *      Multicast receive code 
505          */
506         if (ipv6_addr_is_multicast(daddr)) {
507                 udpv6_mcast_deliver(uh, saddr, daddr, skb);
508                 return 0;
509         }
510
511         /* Unicast */
512
513         /* 
514          * check socket cache ... must talk to Alan about his plans
515          * for sock caches... i'll skip this for now.
516          */
517         sk = udp_v6_lookup(saddr, uh->source, daddr, uh->dest, dev->ifindex);
518
519         if (sk == NULL) {
520                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
521                         goto discard;
522
523                 if (skb->ip_summed != CHECKSUM_UNNECESSARY &&
524                     (unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum)))
525                         goto discard;
526                 UDP6_INC_STATS_BH(UDP_MIB_NOPORTS);
527
528                 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0, dev);
529
530                 kfree_skb(skb);
531                 return(0);
532         }
533         
534         /* deliver */
535         
536         udpv6_queue_rcv_skb(sk, skb);
537         sock_put(sk);
538         return(0);
539
540 short_packet:   
541         if (net_ratelimit())
542                 printk(KERN_DEBUG "UDP: short packet: %d/%u\n", ulen, skb->len);
543
544 discard:
545         UDP6_INC_STATS_BH(UDP_MIB_INERRORS);
546         kfree_skb(skb);
547         return(0);      
548 }
549 /*
550  * Throw away all pending data and cancel the corking. Socket is locked.
551  */
552 static void udp_v6_flush_pending_frames(struct sock *sk)
553 {
554         struct udp_sock *up = udp_sk(sk);
555
556         if (up->pending) {
557                 up->len = 0;
558                 up->pending = 0;
559                 ip6_flush_pending_frames(sk);
560         }
561 }
562
563 /*
564  *      Sending
565  */
566
567 static int udp_v6_push_pending_frames(struct sock *sk, struct udp_sock *up)
568 {
569         struct sk_buff *skb;
570         struct udphdr *uh;
571         struct inet_sock *inet = inet_sk(sk);
572         struct flowi *fl = &inet->cork.fl;
573         int err = 0;
574
575         /* Grab the skbuff where UDP header space exists. */
576         if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
577                 goto out;
578
579         /*
580          * Create a UDP header
581          */
582         uh = skb->h.uh;
583         uh->source = fl->fl_ip_sport;
584         uh->dest = fl->fl_ip_dport;
585         uh->len = htons(up->len);
586         uh->check = 0;
587
588         if (sk->sk_no_check == UDP_CSUM_NOXMIT) {
589                 skb->ip_summed = CHECKSUM_NONE;
590                 goto send;
591         }
592
593         if (skb_queue_len(&sk->sk_write_queue) == 1) {
594                 skb->csum = csum_partial((char *)uh,
595                                 sizeof(struct udphdr), skb->csum);
596                 uh->check = csum_ipv6_magic(&fl->fl6_src,
597                                             &fl->fl6_dst,
598                                             up->len, fl->proto, skb->csum);
599         } else {
600                 u32 tmp_csum = 0;
601
602                 skb_queue_walk(&sk->sk_write_queue, skb) {
603                         tmp_csum = csum_add(tmp_csum, skb->csum);
604                 }
605                 tmp_csum = csum_partial((char *)uh,
606                                 sizeof(struct udphdr), tmp_csum);
607                 tmp_csum = csum_ipv6_magic(&fl->fl6_src,
608                                            &fl->fl6_dst,
609                                            up->len, fl->proto, tmp_csum);
610                 uh->check = tmp_csum;
611
612         }
613         if (uh->check == 0)
614                 uh->check = -1;
615
616 send:
617         err = ip6_push_pending_frames(sk);
618 out:
619         up->len = 0;
620         up->pending = 0;
621         return err;
622 }
623
624 static int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk, 
625                   struct msghdr *msg, size_t len)
626 {
627         struct ipv6_txoptions opt_space;
628         struct udp_sock *up = udp_sk(sk);
629         struct inet_sock *inet = inet_sk(sk);
630         struct ipv6_pinfo *np = inet6_sk(sk);
631         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
632         struct in6_addr *daddr, *final_p = NULL, final;
633         struct ipv6_txoptions *opt = NULL;
634         struct ip6_flowlabel *flowlabel = NULL;
635         struct flowi *fl = &inet->cork.fl;
636         struct dst_entry *dst;
637         int addr_len = msg->msg_namelen;
638         int ulen = len;
639         int hlimit = -1;
640         int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
641         int err;
642
643         /* destination address check */
644         if (sin6) {
645                 if (addr_len < offsetof(struct sockaddr, sa_data))
646                         return -EINVAL;
647
648                 switch (sin6->sin6_family) {
649                 case AF_INET6:
650                         if (addr_len < SIN6_LEN_RFC2133)
651                                 return -EINVAL;
652                         daddr = &sin6->sin6_addr;
653                         break;
654                 case AF_INET:
655                         goto do_udp_sendmsg;
656                 case AF_UNSPEC:
657                         msg->msg_name = sin6 = NULL;
658                         msg->msg_namelen = addr_len = 0;
659                         daddr = NULL;
660                         break;
661                 default:
662                         return -EINVAL;
663                 }
664         } else if (!up->pending) {
665                 if (sk->sk_state != TCP_ESTABLISHED)
666                         return -EDESTADDRREQ;
667                 daddr = &np->daddr;
668         } else 
669                 daddr = NULL;
670
671         if (daddr) {
672                 if (ipv6_addr_type(daddr) == IPV6_ADDR_MAPPED) {
673                         struct sockaddr_in sin;
674                         sin.sin_family = AF_INET;
675                         sin.sin_port = sin6 ? sin6->sin6_port : inet->dport;
676                         sin.sin_addr.s_addr = daddr->s6_addr32[3];
677                         msg->msg_name = &sin;
678                         msg->msg_namelen = sizeof(sin);
679 do_udp_sendmsg:
680                         if (__ipv6_only_sock(sk))
681                                 return -ENETUNREACH;
682                         return udp_sendmsg(iocb, sk, msg, len);
683                 }
684         }
685
686         if (up->pending == AF_INET)
687                 return udp_sendmsg(iocb, sk, msg, len);
688
689         /* Rough check on arithmetic overflow,
690            better check is made in ip6_build_xmit
691            */
692         if (len > INT_MAX - sizeof(struct udphdr))
693                 return -EMSGSIZE;
694         
695         if (up->pending) {
696                 /*
697                  * There are pending frames.
698                  * The socket lock must be held while it's corked.
699                  */
700                 lock_sock(sk);
701                 if (likely(up->pending)) {
702                         if (unlikely(up->pending != AF_INET6)) {
703                                 release_sock(sk);
704                                 return -EAFNOSUPPORT;
705                         }
706                         dst = NULL;
707                         goto do_append_data;
708                 }
709                 release_sock(sk);
710         }
711         ulen += sizeof(struct udphdr);
712
713         memset(fl, 0, sizeof(*fl));
714
715         if (sin6) {
716                 if (sin6->sin6_port == 0)
717                         return -EINVAL;
718
719                 fl->fl_ip_dport = sin6->sin6_port;
720                 daddr = &sin6->sin6_addr;
721
722                 if (np->sndflow) {
723                         fl->fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
724                         if (fl->fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
725                                 flowlabel = fl6_sock_lookup(sk, fl->fl6_flowlabel);
726                                 if (flowlabel == NULL)
727                                         return -EINVAL;
728                                 daddr = &flowlabel->dst;
729                         }
730                 }
731
732                 /*
733                  * Otherwise it will be difficult to maintain
734                  * sk->sk_dst_cache.
735                  */
736                 if (sk->sk_state == TCP_ESTABLISHED &&
737                     ipv6_addr_equal(daddr, &np->daddr))
738                         daddr = &np->daddr;
739
740                 if (addr_len >= sizeof(struct sockaddr_in6) &&
741                     sin6->sin6_scope_id &&
742                     ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
743                         fl->oif = sin6->sin6_scope_id;
744         } else {
745                 if (sk->sk_state != TCP_ESTABLISHED)
746                         return -EDESTADDRREQ;
747
748                 fl->fl_ip_dport = inet->dport;
749                 daddr = &np->daddr;
750                 fl->fl6_flowlabel = np->flow_label;
751         }
752
753         if (!fl->oif)
754                 fl->oif = sk->sk_bound_dev_if;
755
756         if (msg->msg_controllen) {
757                 opt = &opt_space;
758                 memset(opt, 0, sizeof(struct ipv6_txoptions));
759                 opt->tot_len = sizeof(*opt);
760
761                 err = datagram_send_ctl(msg, fl, opt, &hlimit);
762                 if (err < 0) {
763                         fl6_sock_release(flowlabel);
764                         return err;
765                 }
766                 if ((fl->fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
767                         flowlabel = fl6_sock_lookup(sk, fl->fl6_flowlabel);
768                         if (flowlabel == NULL)
769                                 return -EINVAL;
770                 }
771                 if (!(opt->opt_nflen|opt->opt_flen))
772                         opt = NULL;
773         }
774         if (opt == NULL)
775                 opt = np->opt;
776         opt = fl6_merge_options(&opt_space, flowlabel, opt);
777
778         fl->proto = IPPROTO_UDP;
779         ipv6_addr_copy(&fl->fl6_dst, daddr);
780         if (ipv6_addr_any(&fl->fl6_src) && !ipv6_addr_any(&np->saddr))
781                 ipv6_addr_copy(&fl->fl6_src, &np->saddr);
782         fl->fl_ip_sport = inet->sport;
783         
784         /* merge ip6_build_xmit from ip6_output */
785         if (opt && opt->srcrt) {
786                 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
787                 ipv6_addr_copy(&final, &fl->fl6_dst);
788                 ipv6_addr_copy(&fl->fl6_dst, rt0->addr);
789                 final_p = &final;
790         }
791
792         if (!fl->oif && ipv6_addr_is_multicast(&fl->fl6_dst))
793                 fl->oif = np->mcast_oif;
794
795         err = ip6_dst_lookup(sk, &dst, fl);
796         if (err)
797                 goto out;
798         if (final_p)
799                 ipv6_addr_copy(&fl->fl6_dst, final_p);
800
801         if ((err = xfrm_lookup(&dst, fl, sk, 0)) < 0) {
802                 dst_release(dst);
803                 goto out;
804         }
805
806         if (hlimit < 0) {
807                 if (ipv6_addr_is_multicast(&fl->fl6_dst))
808                         hlimit = np->mcast_hops;
809                 else
810                         hlimit = np->hop_limit;
811                 if (hlimit < 0)
812                         hlimit = dst_metric(dst, RTAX_HOPLIMIT);
813                 if (hlimit < 0)
814                         hlimit = ipv6_get_hoplimit(dst->dev);
815         }
816
817         if (msg->msg_flags&MSG_CONFIRM)
818                 goto do_confirm;
819 back_from_confirm:
820
821         lock_sock(sk);
822         if (unlikely(up->pending)) {
823                 /* The socket is already corked while preparing it. */
824                 /* ... which is an evident application bug. --ANK */
825                 release_sock(sk);
826
827                 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
828                 err = -EINVAL;
829                 goto out;
830         }
831
832         up->pending = AF_INET6;
833
834 do_append_data:
835         up->len += ulen;
836         err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, ulen, sizeof(struct udphdr),
837                               hlimit, opt, fl, (struct rt6_info*)dst,
838                               corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
839         if (err)
840                 udp_v6_flush_pending_frames(sk);
841         else if (!corkreq)
842                 err = udp_v6_push_pending_frames(sk, up);
843
844         if (dst)
845                 ip6_dst_store(sk, dst,
846                               ipv6_addr_equal(&fl->fl6_dst, &np->daddr) ?
847                               &np->daddr : NULL);
848         if (err > 0)
849                 err = np->recverr ? net_xmit_errno(err) : 0;
850         release_sock(sk);
851 out:
852         fl6_sock_release(flowlabel);
853         if (!err) {
854                 UDP6_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS);
855                 return len;
856         }
857         return err;
858
859 do_confirm:
860         dst_confirm(dst);
861         if (!(msg->msg_flags&MSG_PROBE) || len)
862                 goto back_from_confirm;
863         err = 0;
864         goto out;
865 }
866
867 static int udpv6_destroy_sock(struct sock *sk)
868 {
869         lock_sock(sk);
870         udp_v6_flush_pending_frames(sk);
871         release_sock(sk);
872
873         inet6_destroy_sock(sk);
874
875         return 0;
876 }
877
878 /*
879  *      Socket option code for UDP
880  */
881 static int udpv6_setsockopt(struct sock *sk, int level, int optname, 
882                           char __user *optval, int optlen)
883 {
884         struct udp_sock *up = udp_sk(sk);
885         int val;
886         int err = 0;
887
888         if (level != SOL_UDP)
889                 return ipv6_setsockopt(sk, level, optname, optval, optlen);
890
891         if(optlen<sizeof(int))
892                 return -EINVAL;
893
894         if (get_user(val, (int __user *)optval))
895                 return -EFAULT;
896
897         switch(optname) {
898         case UDP_CORK:
899                 if (val != 0) {
900                         up->corkflag = 1;
901                 } else {
902                         up->corkflag = 0;
903                         lock_sock(sk);
904                         udp_v6_push_pending_frames(sk, up);
905                         release_sock(sk);
906                 }
907                 break;
908                 
909         case UDP_ENCAP:
910                 switch (val) {
911                 case 0:
912                         up->encap_type = val;
913                         break;
914                 default:
915                         err = -ENOPROTOOPT;
916                         break;
917                 }
918                 break;
919
920         default:
921                 err = -ENOPROTOOPT;
922                 break;
923         };
924
925         return err;
926 }
927
928 static int udpv6_getsockopt(struct sock *sk, int level, int optname, 
929                           char __user *optval, int __user *optlen)
930 {
931         struct udp_sock *up = udp_sk(sk);
932         int val, len;
933
934         if (level != SOL_UDP)
935                 return ipv6_getsockopt(sk, level, optname, optval, optlen);
936
937         if(get_user(len,optlen))
938                 return -EFAULT;
939
940         len = min_t(unsigned int, len, sizeof(int));
941         
942         if(len < 0)
943                 return -EINVAL;
944
945         switch(optname) {
946         case UDP_CORK:
947                 val = up->corkflag;
948                 break;
949
950         case UDP_ENCAP:
951                 val = up->encap_type;
952                 break;
953
954         default:
955                 return -ENOPROTOOPT;
956         };
957
958         if(put_user(len, optlen))
959                 return -EFAULT;
960         if(copy_to_user(optval, &val,len))
961                 return -EFAULT;
962         return 0;
963 }
964
965 static struct inet6_protocol udpv6_protocol = {
966         .handler        =       udpv6_rcv,
967         .err_handler    =       udpv6_err,
968         .flags          =       INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
969 };
970
971 /* ------------------------------------------------------------------------ */
972 #ifdef CONFIG_PROC_FS
973
974 static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket)
975 {
976         struct inet_sock *inet = inet_sk(sp);
977         struct ipv6_pinfo *np = inet6_sk(sp);
978         struct in6_addr *dest, *src;
979         __u16 destp, srcp;
980
981         dest  = &np->daddr;
982         src   = &np->rcv_saddr;
983         destp = ntohs(inet->dport);
984         srcp  = ntohs(inet->sport);
985         seq_printf(seq,
986                    "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
987                    "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
988                    bucket,
989                    src->s6_addr32[0], src->s6_addr32[1],
990                    src->s6_addr32[2], src->s6_addr32[3], srcp,
991                    dest->s6_addr32[0], dest->s6_addr32[1],
992                    dest->s6_addr32[2], dest->s6_addr32[3], destp,
993                    sp->sk_state, 
994                    atomic_read(&sp->sk_wmem_alloc),
995                    atomic_read(&sp->sk_rmem_alloc),
996                    0, 0L, 0,
997                    sock_i_uid(sp), 0,
998                    sock_i_ino(sp),
999                    atomic_read(&sp->sk_refcnt), sp);
1000 }
1001
1002 static int udp6_seq_show(struct seq_file *seq, void *v)
1003 {
1004         if (v == SEQ_START_TOKEN)
1005                 seq_printf(seq,
1006                            "  sl  "
1007                            "local_address                         "
1008                            "remote_address                        "
1009                            "st tx_queue rx_queue tr tm->when retrnsmt"
1010                            "   uid  timeout inode\n");
1011         else
1012                 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket);
1013         return 0;
1014 }
1015
1016 static struct file_operations udp6_seq_fops;
1017 static struct udp_seq_afinfo udp6_seq_afinfo = {
1018         .owner          = THIS_MODULE,
1019         .name           = "udp6",
1020         .family         = AF_INET6,
1021         .seq_show       = udp6_seq_show,
1022         .seq_fops       = &udp6_seq_fops,
1023 };
1024
1025 int __init udp6_proc_init(void)
1026 {
1027         return udp_proc_register(&udp6_seq_afinfo);
1028 }
1029
1030 void udp6_proc_exit(void) {
1031         udp_proc_unregister(&udp6_seq_afinfo);
1032 }
1033 #endif /* CONFIG_PROC_FS */
1034
1035 /* ------------------------------------------------------------------------ */
1036
1037 struct proto udpv6_prot = {
1038         .name =         "UDPv6",
1039         .owner =        THIS_MODULE,
1040         .close =        udpv6_close,
1041         .connect =      ip6_datagram_connect,
1042         .disconnect =   udp_disconnect,
1043         .ioctl =        udp_ioctl,
1044         .destroy =      udpv6_destroy_sock,
1045         .setsockopt =   udpv6_setsockopt,
1046         .getsockopt =   udpv6_getsockopt,
1047         .sendmsg =      udpv6_sendmsg,
1048         .recvmsg =      udpv6_recvmsg,
1049         .backlog_rcv =  udpv6_queue_rcv_skb,
1050         .hash =         udp_v6_hash,
1051         .unhash =       udp_v6_unhash,
1052         .get_port =     udp_v6_get_port,
1053         .obj_size =     sizeof(struct udp6_sock),
1054 };
1055
1056 static struct inet_protosw udpv6_protosw = {
1057         .type =      SOCK_DGRAM,
1058         .protocol =  IPPROTO_UDP,
1059         .prot =      &udpv6_prot,
1060         .ops =       &inet6_dgram_ops,
1061         .capability =-1,
1062         .no_check =  UDP_CSUM_DEFAULT,
1063         .flags =     INET_PROTOSW_PERMANENT,
1064 };
1065
1066
1067 void __init udpv6_init(void)
1068 {
1069         if (inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP) < 0)
1070                 printk(KERN_ERR "udpv6_init: Could not register protocol\n");
1071         inet6_register_protosw(&udpv6_protosw);
1072 }