[NETFILTER] fix leak of fragment queue at unloading nf_conntrack_ipv6
[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(i = 0; i < (1 << 16) / UDP_HTABLE_SIZE; i++, 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                 if (i >= (1 << 16) / UDP_HTABLE_SIZE)
111                         goto fail;
112 gotit:
113                 udp_port_rover = snum = result;
114         } else {
115                 sk_for_each(sk2, node,
116                             &udp_hash[snum & (UDP_HTABLE_SIZE - 1)]) {
117                         if (inet_sk(sk2)->num == snum &&
118                             sk2 != sk &&
119                             (!sk2->sk_bound_dev_if ||
120                              !sk->sk_bound_dev_if ||
121                              sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
122                             (!sk2->sk_reuse || !sk->sk_reuse) &&
123                             ipv6_rcv_saddr_equal(sk, sk2))
124                                 goto fail;
125                 }
126         }
127
128         inet_sk(sk)->num = snum;
129         if (sk_unhashed(sk)) {
130                 sk_add_node(sk, &udp_hash[snum & (UDP_HTABLE_SIZE - 1)]);
131                 sock_prot_inc_use(sk->sk_prot);
132         }
133         write_unlock_bh(&udp_hash_lock);
134         return 0;
135
136 fail:
137         write_unlock_bh(&udp_hash_lock);
138         return 1;
139 }
140
141 static void udp_v6_hash(struct sock *sk)
142 {
143         BUG();
144 }
145
146 static void udp_v6_unhash(struct sock *sk)
147 {
148         write_lock_bh(&udp_hash_lock);
149         if (sk_del_node_init(sk)) {
150                 inet_sk(sk)->num = 0;
151                 sock_prot_dec_use(sk->sk_prot);
152         }
153         write_unlock_bh(&udp_hash_lock);
154 }
155
156 static struct sock *udp_v6_lookup(struct in6_addr *saddr, u16 sport,
157                                   struct in6_addr *daddr, u16 dport, int dif)
158 {
159         struct sock *sk, *result = NULL;
160         struct hlist_node *node;
161         unsigned short hnum = ntohs(dport);
162         int badness = -1;
163
164         read_lock(&udp_hash_lock);
165         sk_for_each(sk, node, &udp_hash[hnum & (UDP_HTABLE_SIZE - 1)]) {
166                 struct inet_sock *inet = inet_sk(sk);
167
168                 if (inet->num == hnum && sk->sk_family == PF_INET6) {
169                         struct ipv6_pinfo *np = inet6_sk(sk);
170                         int score = 0;
171                         if (inet->dport) {
172                                 if (inet->dport != sport)
173                                         continue;
174                                 score++;
175                         }
176                         if (!ipv6_addr_any(&np->rcv_saddr)) {
177                                 if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
178                                         continue;
179                                 score++;
180                         }
181                         if (!ipv6_addr_any(&np->daddr)) {
182                                 if (!ipv6_addr_equal(&np->daddr, saddr))
183                                         continue;
184                                 score++;
185                         }
186                         if (sk->sk_bound_dev_if) {
187                                 if (sk->sk_bound_dev_if != dif)
188                                         continue;
189                                 score++;
190                         }
191                         if(score == 4) {
192                                 result = sk;
193                                 break;
194                         } else if(score > badness) {
195                                 result = sk;
196                                 badness = score;
197                         }
198                 }
199         }
200         if (result)
201                 sock_hold(result);
202         read_unlock(&udp_hash_lock);
203         return result;
204 }
205
206 /*
207  *
208  */
209
210 static void udpv6_close(struct sock *sk, long timeout)
211 {
212         sk_common_release(sk);
213 }
214
215 /*
216  *      This should be easy, if there is something there we
217  *      return it, otherwise we block.
218  */
219
220 static int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk, 
221                   struct msghdr *msg, size_t len,
222                   int noblock, int flags, int *addr_len)
223 {
224         struct ipv6_pinfo *np = inet6_sk(sk);
225         struct inet_sock *inet = inet_sk(sk);
226         struct sk_buff *skb;
227         size_t copied;
228         int err;
229
230         if (addr_len)
231                 *addr_len=sizeof(struct sockaddr_in6);
232   
233         if (flags & MSG_ERRQUEUE)
234                 return ipv6_recv_error(sk, msg, len);
235
236 try_again:
237         skb = skb_recv_datagram(sk, flags, noblock, &err);
238         if (!skb)
239                 goto out;
240
241         copied = skb->len - sizeof(struct udphdr);
242         if (copied > len) {
243                 copied = len;
244                 msg->msg_flags |= MSG_TRUNC;
245         }
246
247         if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
248                 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
249                                               copied);
250         } else if (msg->msg_flags&MSG_TRUNC) {
251                 if (__skb_checksum_complete(skb))
252                         goto csum_copy_err;
253                 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
254                                               copied);
255         } else {
256                 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
257                 if (err == -EINVAL)
258                         goto csum_copy_err;
259         }
260         if (err)
261                 goto out_free;
262
263         sock_recv_timestamp(msg, sk, skb);
264
265         /* Copy the address. */
266         if (msg->msg_name) {
267                 struct sockaddr_in6 *sin6;
268           
269                 sin6 = (struct sockaddr_in6 *) msg->msg_name;
270                 sin6->sin6_family = AF_INET6;
271                 sin6->sin6_port = skb->h.uh->source;
272                 sin6->sin6_flowinfo = 0;
273                 sin6->sin6_scope_id = 0;
274
275                 if (skb->protocol == htons(ETH_P_IP))
276                         ipv6_addr_set(&sin6->sin6_addr, 0, 0,
277                                       htonl(0xffff), skb->nh.iph->saddr);
278                 else {
279                         ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
280                         if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
281                                 sin6->sin6_scope_id = IP6CB(skb)->iif;
282                 }
283
284         }
285         if (skb->protocol == htons(ETH_P_IP)) {
286                 if (inet->cmsg_flags)
287                         ip_cmsg_recv(msg, skb);
288         } else {
289                 if (np->rxopt.all)
290                         datagram_recv_ctl(sk, msg, skb);
291         }
292
293         err = copied;
294         if (flags & MSG_TRUNC)
295                 err = skb->len - sizeof(struct udphdr);
296
297 out_free:
298         skb_free_datagram(sk, skb);
299 out:
300         return err;
301
302 csum_copy_err:
303         /* Clear queue. */
304         if (flags&MSG_PEEK) {
305                 int clear = 0;
306                 spin_lock_bh(&sk->sk_receive_queue.lock);
307                 if (skb == skb_peek(&sk->sk_receive_queue)) {
308                         __skb_unlink(skb, &sk->sk_receive_queue);
309                         clear = 1;
310                 }
311                 spin_unlock_bh(&sk->sk_receive_queue.lock);
312                 if (clear)
313                         kfree_skb(skb);
314         }
315
316         skb_free_datagram(sk, skb);
317
318         if (flags & MSG_DONTWAIT) {
319                 UDP6_INC_STATS_USER(UDP_MIB_INERRORS);
320                 return -EAGAIN;
321         }
322         goto try_again;
323 }
324
325 static void udpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
326                int type, int code, int offset, __u32 info)
327 {
328         struct ipv6_pinfo *np;
329         struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
330         struct net_device *dev = skb->dev;
331         struct in6_addr *saddr = &hdr->saddr;
332         struct in6_addr *daddr = &hdr->daddr;
333         struct udphdr *uh = (struct udphdr*)(skb->data+offset);
334         struct sock *sk;
335         int err;
336
337         sk = udp_v6_lookup(daddr, uh->dest, saddr, uh->source, dev->ifindex);
338    
339         if (sk == NULL)
340                 return;
341
342         np = inet6_sk(sk);
343
344         if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
345                 goto out;
346
347         if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
348                 goto out;
349
350         if (np->recverr)
351                 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
352
353         sk->sk_err = err;
354         sk->sk_error_report(sk);
355 out:
356         sock_put(sk);
357 }
358
359 static inline int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
360 {
361         if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
362                 kfree_skb(skb);
363                 return -1;
364         }
365
366         if (skb_checksum_complete(skb)) {
367                 UDP6_INC_STATS_BH(UDP_MIB_INERRORS);
368                 kfree_skb(skb);
369                 return 0;
370         }
371
372         if (sock_queue_rcv_skb(sk,skb)<0) {
373                 UDP6_INC_STATS_BH(UDP_MIB_INERRORS);
374                 kfree_skb(skb);
375                 return 0;
376         }
377         UDP6_INC_STATS_BH(UDP_MIB_INDATAGRAMS);
378         return 0;
379 }
380
381 static struct sock *udp_v6_mcast_next(struct sock *sk,
382                                       u16 loc_port, struct in6_addr *loc_addr,
383                                       u16 rmt_port, struct in6_addr *rmt_addr,
384                                       int dif)
385 {
386         struct hlist_node *node;
387         struct sock *s = sk;
388         unsigned short num = ntohs(loc_port);
389
390         sk_for_each_from(s, node) {
391                 struct inet_sock *inet = inet_sk(s);
392
393                 if (inet->num == num && s->sk_family == PF_INET6) {
394                         struct ipv6_pinfo *np = inet6_sk(s);
395                         if (inet->dport) {
396                                 if (inet->dport != rmt_port)
397                                         continue;
398                         }
399                         if (!ipv6_addr_any(&np->daddr) &&
400                             !ipv6_addr_equal(&np->daddr, rmt_addr))
401                                 continue;
402
403                         if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
404                                 continue;
405
406                         if (!ipv6_addr_any(&np->rcv_saddr)) {
407                                 if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
408                                         continue;
409                         }
410                         if(!inet6_mc_check(s, loc_addr, rmt_addr))
411                                 continue;
412                         return s;
413                 }
414         }
415         return NULL;
416 }
417
418 /*
419  * Note: called only from the BH handler context,
420  * so we don't need to lock the hashes.
421  */
422 static void udpv6_mcast_deliver(struct udphdr *uh,
423                                 struct in6_addr *saddr, struct in6_addr *daddr,
424                                 struct sk_buff *skb)
425 {
426         struct sock *sk, *sk2;
427         int dif;
428
429         read_lock(&udp_hash_lock);
430         sk = sk_head(&udp_hash[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
431         dif = skb->dev->ifindex;
432         sk = udp_v6_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
433         if (!sk) {
434                 kfree_skb(skb);
435                 goto out;
436         }
437
438         sk2 = sk;
439         while ((sk2 = udp_v6_mcast_next(sk_next(sk2), uh->dest, daddr,
440                                         uh->source, saddr, dif))) {
441                 struct sk_buff *buff = skb_clone(skb, GFP_ATOMIC);
442                 if (buff)
443                         udpv6_queue_rcv_skb(sk2, buff);
444         }
445         udpv6_queue_rcv_skb(sk, skb);
446 out:
447         read_unlock(&udp_hash_lock);
448 }
449
450 static int udpv6_rcv(struct sk_buff **pskb, unsigned int *nhoffp)
451 {
452         struct sk_buff *skb = *pskb;
453         struct sock *sk;
454         struct udphdr *uh;
455         struct net_device *dev = skb->dev;
456         struct in6_addr *saddr, *daddr;
457         u32 ulen = 0;
458
459         if (!pskb_may_pull(skb, sizeof(struct udphdr)))
460                 goto short_packet;
461
462         saddr = &skb->nh.ipv6h->saddr;
463         daddr = &skb->nh.ipv6h->daddr;
464         uh = skb->h.uh;
465
466         ulen = ntohs(uh->len);
467
468         /* Check for jumbo payload */
469         if (ulen == 0)
470                 ulen = skb->len;
471
472         if (ulen > skb->len || ulen < sizeof(*uh))
473                 goto short_packet;
474
475         if (uh->check == 0) {
476                 /* RFC 2460 section 8.1 says that we SHOULD log
477                    this error. Well, it is reasonable.
478                  */
479                 LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n");
480                 goto discard;
481         }
482
483         if (ulen < skb->len) {
484                 if (pskb_trim_rcsum(skb, ulen))
485                         goto discard;
486                 saddr = &skb->nh.ipv6h->saddr;
487                 daddr = &skb->nh.ipv6h->daddr;
488                 uh = skb->h.uh;
489         }
490
491         if (skb->ip_summed == CHECKSUM_HW &&
492             !csum_ipv6_magic(saddr, daddr, ulen, IPPROTO_UDP, skb->csum))
493                 skb->ip_summed = CHECKSUM_UNNECESSARY;
494
495         if (skb->ip_summed != CHECKSUM_UNNECESSARY)
496                 skb->csum = ~csum_ipv6_magic(saddr, daddr, ulen, IPPROTO_UDP, 0);
497
498         /* 
499          *      Multicast receive code 
500          */
501         if (ipv6_addr_is_multicast(daddr)) {
502                 udpv6_mcast_deliver(uh, saddr, daddr, skb);
503                 return 0;
504         }
505
506         /* Unicast */
507
508         /* 
509          * check socket cache ... must talk to Alan about his plans
510          * for sock caches... i'll skip this for now.
511          */
512         sk = udp_v6_lookup(saddr, uh->source, daddr, uh->dest, dev->ifindex);
513
514         if (sk == NULL) {
515                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
516                         goto discard;
517
518                 if (skb_checksum_complete(skb))
519                         goto discard;
520                 UDP6_INC_STATS_BH(UDP_MIB_NOPORTS);
521
522                 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0, dev);
523
524                 kfree_skb(skb);
525                 return(0);
526         }
527         
528         /* deliver */
529         
530         udpv6_queue_rcv_skb(sk, skb);
531         sock_put(sk);
532         return(0);
533
534 short_packet:   
535         if (net_ratelimit())
536                 printk(KERN_DEBUG "UDP: short packet: %d/%u\n", ulen, skb->len);
537
538 discard:
539         UDP6_INC_STATS_BH(UDP_MIB_INERRORS);
540         kfree_skb(skb);
541         return(0);      
542 }
543 /*
544  * Throw away all pending data and cancel the corking. Socket is locked.
545  */
546 static void udp_v6_flush_pending_frames(struct sock *sk)
547 {
548         struct udp_sock *up = udp_sk(sk);
549
550         if (up->pending) {
551                 up->len = 0;
552                 up->pending = 0;
553                 ip6_flush_pending_frames(sk);
554         }
555 }
556
557 /*
558  *      Sending
559  */
560
561 static int udp_v6_push_pending_frames(struct sock *sk, struct udp_sock *up)
562 {
563         struct sk_buff *skb;
564         struct udphdr *uh;
565         struct inet_sock *inet = inet_sk(sk);
566         struct flowi *fl = &inet->cork.fl;
567         int err = 0;
568
569         /* Grab the skbuff where UDP header space exists. */
570         if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
571                 goto out;
572
573         /*
574          * Create a UDP header
575          */
576         uh = skb->h.uh;
577         uh->source = fl->fl_ip_sport;
578         uh->dest = fl->fl_ip_dport;
579         uh->len = htons(up->len);
580         uh->check = 0;
581
582         if (sk->sk_no_check == UDP_CSUM_NOXMIT) {
583                 skb->ip_summed = CHECKSUM_NONE;
584                 goto send;
585         }
586
587         if (skb_queue_len(&sk->sk_write_queue) == 1) {
588                 skb->csum = csum_partial((char *)uh,
589                                 sizeof(struct udphdr), skb->csum);
590                 uh->check = csum_ipv6_magic(&fl->fl6_src,
591                                             &fl->fl6_dst,
592                                             up->len, fl->proto, skb->csum);
593         } else {
594                 u32 tmp_csum = 0;
595
596                 skb_queue_walk(&sk->sk_write_queue, skb) {
597                         tmp_csum = csum_add(tmp_csum, skb->csum);
598                 }
599                 tmp_csum = csum_partial((char *)uh,
600                                 sizeof(struct udphdr), tmp_csum);
601                 tmp_csum = csum_ipv6_magic(&fl->fl6_src,
602                                            &fl->fl6_dst,
603                                            up->len, fl->proto, tmp_csum);
604                 uh->check = tmp_csum;
605
606         }
607         if (uh->check == 0)
608                 uh->check = -1;
609
610 send:
611         err = ip6_push_pending_frames(sk);
612 out:
613         up->len = 0;
614         up->pending = 0;
615         return err;
616 }
617
618 static int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk, 
619                   struct msghdr *msg, size_t len)
620 {
621         struct ipv6_txoptions opt_space;
622         struct udp_sock *up = udp_sk(sk);
623         struct inet_sock *inet = inet_sk(sk);
624         struct ipv6_pinfo *np = inet6_sk(sk);
625         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
626         struct in6_addr *daddr, *final_p = NULL, final;
627         struct ipv6_txoptions *opt = NULL;
628         struct ip6_flowlabel *flowlabel = NULL;
629         struct flowi *fl = &inet->cork.fl;
630         struct dst_entry *dst;
631         int addr_len = msg->msg_namelen;
632         int ulen = len;
633         int hlimit = -1;
634         int tclass = -1;
635         int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
636         int err;
637         int connected = 0;
638
639         /* destination address check */
640         if (sin6) {
641                 if (addr_len < offsetof(struct sockaddr, sa_data))
642                         return -EINVAL;
643
644                 switch (sin6->sin6_family) {
645                 case AF_INET6:
646                         if (addr_len < SIN6_LEN_RFC2133)
647                                 return -EINVAL;
648                         daddr = &sin6->sin6_addr;
649                         break;
650                 case AF_INET:
651                         goto do_udp_sendmsg;
652                 case AF_UNSPEC:
653                         msg->msg_name = sin6 = NULL;
654                         msg->msg_namelen = addr_len = 0;
655                         daddr = NULL;
656                         break;
657                 default:
658                         return -EINVAL;
659                 }
660         } else if (!up->pending) {
661                 if (sk->sk_state != TCP_ESTABLISHED)
662                         return -EDESTADDRREQ;
663                 daddr = &np->daddr;
664         } else 
665                 daddr = NULL;
666
667         if (daddr) {
668                 if (ipv6_addr_type(daddr) == IPV6_ADDR_MAPPED) {
669                         struct sockaddr_in sin;
670                         sin.sin_family = AF_INET;
671                         sin.sin_port = sin6 ? sin6->sin6_port : inet->dport;
672                         sin.sin_addr.s_addr = daddr->s6_addr32[3];
673                         msg->msg_name = &sin;
674                         msg->msg_namelen = sizeof(sin);
675 do_udp_sendmsg:
676                         if (__ipv6_only_sock(sk))
677                                 return -ENETUNREACH;
678                         return udp_sendmsg(iocb, sk, msg, len);
679                 }
680         }
681
682         if (up->pending == AF_INET)
683                 return udp_sendmsg(iocb, sk, msg, len);
684
685         /* Rough check on arithmetic overflow,
686            better check is made in ip6_build_xmit
687            */
688         if (len > INT_MAX - sizeof(struct udphdr))
689                 return -EMSGSIZE;
690         
691         if (up->pending) {
692                 /*
693                  * There are pending frames.
694                  * The socket lock must be held while it's corked.
695                  */
696                 lock_sock(sk);
697                 if (likely(up->pending)) {
698                         if (unlikely(up->pending != AF_INET6)) {
699                                 release_sock(sk);
700                                 return -EAFNOSUPPORT;
701                         }
702                         dst = NULL;
703                         goto do_append_data;
704                 }
705                 release_sock(sk);
706         }
707         ulen += sizeof(struct udphdr);
708
709         memset(fl, 0, sizeof(*fl));
710
711         if (sin6) {
712                 if (sin6->sin6_port == 0)
713                         return -EINVAL;
714
715                 fl->fl_ip_dport = sin6->sin6_port;
716                 daddr = &sin6->sin6_addr;
717
718                 if (np->sndflow) {
719                         fl->fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
720                         if (fl->fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
721                                 flowlabel = fl6_sock_lookup(sk, fl->fl6_flowlabel);
722                                 if (flowlabel == NULL)
723                                         return -EINVAL;
724                                 daddr = &flowlabel->dst;
725                         }
726                 }
727
728                 /*
729                  * Otherwise it will be difficult to maintain
730                  * sk->sk_dst_cache.
731                  */
732                 if (sk->sk_state == TCP_ESTABLISHED &&
733                     ipv6_addr_equal(daddr, &np->daddr))
734                         daddr = &np->daddr;
735
736                 if (addr_len >= sizeof(struct sockaddr_in6) &&
737                     sin6->sin6_scope_id &&
738                     ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
739                         fl->oif = sin6->sin6_scope_id;
740         } else {
741                 if (sk->sk_state != TCP_ESTABLISHED)
742                         return -EDESTADDRREQ;
743
744                 fl->fl_ip_dport = inet->dport;
745                 daddr = &np->daddr;
746                 fl->fl6_flowlabel = np->flow_label;
747                 connected = 1;
748         }
749
750         if (!fl->oif)
751                 fl->oif = sk->sk_bound_dev_if;
752
753         if (msg->msg_controllen) {
754                 opt = &opt_space;
755                 memset(opt, 0, sizeof(struct ipv6_txoptions));
756                 opt->tot_len = sizeof(*opt);
757
758                 err = datagram_send_ctl(msg, fl, opt, &hlimit, &tclass);
759                 if (err < 0) {
760                         fl6_sock_release(flowlabel);
761                         return err;
762                 }
763                 if ((fl->fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
764                         flowlabel = fl6_sock_lookup(sk, fl->fl6_flowlabel);
765                         if (flowlabel == NULL)
766                                 return -EINVAL;
767                 }
768                 if (!(opt->opt_nflen|opt->opt_flen))
769                         opt = NULL;
770                 connected = 0;
771         }
772         if (opt == NULL)
773                 opt = np->opt;
774         opt = fl6_merge_options(&opt_space, flowlabel, opt);
775
776         fl->proto = IPPROTO_UDP;
777         ipv6_addr_copy(&fl->fl6_dst, daddr);
778         if (ipv6_addr_any(&fl->fl6_src) && !ipv6_addr_any(&np->saddr))
779                 ipv6_addr_copy(&fl->fl6_src, &np->saddr);
780         fl->fl_ip_sport = inet->sport;
781         
782         /* merge ip6_build_xmit from ip6_output */
783         if (opt && opt->srcrt) {
784                 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
785                 ipv6_addr_copy(&final, &fl->fl6_dst);
786                 ipv6_addr_copy(&fl->fl6_dst, rt0->addr);
787                 final_p = &final;
788                 connected = 0;
789         }
790
791         if (!fl->oif && ipv6_addr_is_multicast(&fl->fl6_dst)) {
792                 fl->oif = np->mcast_oif;
793                 connected = 0;
794         }
795
796         err = ip6_dst_lookup(sk, &dst, fl);
797         if (err)
798                 goto out;
799         if (final_p)
800                 ipv6_addr_copy(&fl->fl6_dst, final_p);
801
802         if ((err = xfrm_lookup(&dst, fl, sk, 0)) < 0)
803                 goto out;
804
805         if (hlimit < 0) {
806                 if (ipv6_addr_is_multicast(&fl->fl6_dst))
807                         hlimit = np->mcast_hops;
808                 else
809                         hlimit = np->hop_limit;
810                 if (hlimit < 0)
811                         hlimit = dst_metric(dst, RTAX_HOPLIMIT);
812                 if (hlimit < 0)
813                         hlimit = ipv6_get_hoplimit(dst->dev);
814         }
815
816         if (tclass < 0) {
817                 tclass = np->tclass;
818                 if (tclass < 0)
819                         tclass = 0;
820         }
821
822         if (msg->msg_flags&MSG_CONFIRM)
823                 goto do_confirm;
824 back_from_confirm:
825
826         lock_sock(sk);
827         if (unlikely(up->pending)) {
828                 /* The socket is already corked while preparing it. */
829                 /* ... which is an evident application bug. --ANK */
830                 release_sock(sk);
831
832                 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
833                 err = -EINVAL;
834                 goto out;
835         }
836
837         up->pending = AF_INET6;
838
839 do_append_data:
840         up->len += ulen;
841         err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, ulen,
842                 sizeof(struct udphdr), hlimit, tclass, opt, fl,
843                 (struct rt6_info*)dst,
844                 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
845         if (err)
846                 udp_v6_flush_pending_frames(sk);
847         else if (!corkreq)
848                 err = udp_v6_push_pending_frames(sk, up);
849
850         if (dst) {
851                 if (connected) {
852                         ip6_dst_store(sk, dst,
853                                       ipv6_addr_equal(&fl->fl6_dst, &np->daddr) ?
854                                       &np->daddr : NULL);
855                 } else {
856                         dst_release(dst);
857                 }
858         }
859
860         if (err > 0)
861                 err = np->recverr ? net_xmit_errno(err) : 0;
862         release_sock(sk);
863 out:
864         fl6_sock_release(flowlabel);
865         if (!err) {
866                 UDP6_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS);
867                 return len;
868         }
869         return err;
870
871 do_confirm:
872         dst_confirm(dst);
873         if (!(msg->msg_flags&MSG_PROBE) || len)
874                 goto back_from_confirm;
875         err = 0;
876         goto out;
877 }
878
879 static int udpv6_destroy_sock(struct sock *sk)
880 {
881         lock_sock(sk);
882         udp_v6_flush_pending_frames(sk);
883         release_sock(sk);
884
885         inet6_destroy_sock(sk);
886
887         return 0;
888 }
889
890 /*
891  *      Socket option code for UDP
892  */
893 static int udpv6_setsockopt(struct sock *sk, int level, int optname, 
894                           char __user *optval, int optlen)
895 {
896         struct udp_sock *up = udp_sk(sk);
897         int val;
898         int err = 0;
899
900         if (level != SOL_UDP)
901                 return ipv6_setsockopt(sk, level, optname, optval, optlen);
902
903         if(optlen<sizeof(int))
904                 return -EINVAL;
905
906         if (get_user(val, (int __user *)optval))
907                 return -EFAULT;
908
909         switch(optname) {
910         case UDP_CORK:
911                 if (val != 0) {
912                         up->corkflag = 1;
913                 } else {
914                         up->corkflag = 0;
915                         lock_sock(sk);
916                         udp_v6_push_pending_frames(sk, up);
917                         release_sock(sk);
918                 }
919                 break;
920                 
921         case UDP_ENCAP:
922                 switch (val) {
923                 case 0:
924                         up->encap_type = val;
925                         break;
926                 default:
927                         err = -ENOPROTOOPT;
928                         break;
929                 }
930                 break;
931
932         default:
933                 err = -ENOPROTOOPT;
934                 break;
935         };
936
937         return err;
938 }
939
940 static int udpv6_getsockopt(struct sock *sk, int level, int optname, 
941                           char __user *optval, int __user *optlen)
942 {
943         struct udp_sock *up = udp_sk(sk);
944         int val, len;
945
946         if (level != SOL_UDP)
947                 return ipv6_getsockopt(sk, level, optname, optval, optlen);
948
949         if(get_user(len,optlen))
950                 return -EFAULT;
951
952         len = min_t(unsigned int, len, sizeof(int));
953         
954         if(len < 0)
955                 return -EINVAL;
956
957         switch(optname) {
958         case UDP_CORK:
959                 val = up->corkflag;
960                 break;
961
962         case UDP_ENCAP:
963                 val = up->encap_type;
964                 break;
965
966         default:
967                 return -ENOPROTOOPT;
968         };
969
970         if(put_user(len, optlen))
971                 return -EFAULT;
972         if(copy_to_user(optval, &val,len))
973                 return -EFAULT;
974         return 0;
975 }
976
977 static struct inet6_protocol udpv6_protocol = {
978         .handler        =       udpv6_rcv,
979         .err_handler    =       udpv6_err,
980         .flags          =       INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
981 };
982
983 /* ------------------------------------------------------------------------ */
984 #ifdef CONFIG_PROC_FS
985
986 static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket)
987 {
988         struct inet_sock *inet = inet_sk(sp);
989         struct ipv6_pinfo *np = inet6_sk(sp);
990         struct in6_addr *dest, *src;
991         __u16 destp, srcp;
992
993         dest  = &np->daddr;
994         src   = &np->rcv_saddr;
995         destp = ntohs(inet->dport);
996         srcp  = ntohs(inet->sport);
997         seq_printf(seq,
998                    "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
999                    "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1000                    bucket,
1001                    src->s6_addr32[0], src->s6_addr32[1],
1002                    src->s6_addr32[2], src->s6_addr32[3], srcp,
1003                    dest->s6_addr32[0], dest->s6_addr32[1],
1004                    dest->s6_addr32[2], dest->s6_addr32[3], destp,
1005                    sp->sk_state, 
1006                    atomic_read(&sp->sk_wmem_alloc),
1007                    atomic_read(&sp->sk_rmem_alloc),
1008                    0, 0L, 0,
1009                    sock_i_uid(sp), 0,
1010                    sock_i_ino(sp),
1011                    atomic_read(&sp->sk_refcnt), sp);
1012 }
1013
1014 static int udp6_seq_show(struct seq_file *seq, void *v)
1015 {
1016         if (v == SEQ_START_TOKEN)
1017                 seq_printf(seq,
1018                            "  sl  "
1019                            "local_address                         "
1020                            "remote_address                        "
1021                            "st tx_queue rx_queue tr tm->when retrnsmt"
1022                            "   uid  timeout inode\n");
1023         else
1024                 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket);
1025         return 0;
1026 }
1027
1028 static struct file_operations udp6_seq_fops;
1029 static struct udp_seq_afinfo udp6_seq_afinfo = {
1030         .owner          = THIS_MODULE,
1031         .name           = "udp6",
1032         .family         = AF_INET6,
1033         .seq_show       = udp6_seq_show,
1034         .seq_fops       = &udp6_seq_fops,
1035 };
1036
1037 int __init udp6_proc_init(void)
1038 {
1039         return udp_proc_register(&udp6_seq_afinfo);
1040 }
1041
1042 void udp6_proc_exit(void) {
1043         udp_proc_unregister(&udp6_seq_afinfo);
1044 }
1045 #endif /* CONFIG_PROC_FS */
1046
1047 /* ------------------------------------------------------------------------ */
1048
1049 struct proto udpv6_prot = {
1050         .name =         "UDPv6",
1051         .owner =        THIS_MODULE,
1052         .close =        udpv6_close,
1053         .connect =      ip6_datagram_connect,
1054         .disconnect =   udp_disconnect,
1055         .ioctl =        udp_ioctl,
1056         .destroy =      udpv6_destroy_sock,
1057         .setsockopt =   udpv6_setsockopt,
1058         .getsockopt =   udpv6_getsockopt,
1059         .sendmsg =      udpv6_sendmsg,
1060         .recvmsg =      udpv6_recvmsg,
1061         .backlog_rcv =  udpv6_queue_rcv_skb,
1062         .hash =         udp_v6_hash,
1063         .unhash =       udp_v6_unhash,
1064         .get_port =     udp_v6_get_port,
1065         .obj_size =     sizeof(struct udp6_sock),
1066 };
1067
1068 static struct inet_protosw udpv6_protosw = {
1069         .type =      SOCK_DGRAM,
1070         .protocol =  IPPROTO_UDP,
1071         .prot =      &udpv6_prot,
1072         .ops =       &inet6_dgram_ops,
1073         .capability =-1,
1074         .no_check =  UDP_CSUM_DEFAULT,
1075         .flags =     INET_PROTOSW_PERMANENT,
1076 };
1077
1078
1079 void __init udpv6_init(void)
1080 {
1081         if (inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP) < 0)
1082                 printk(KERN_ERR "udpv6_init: Could not register protocol\n");
1083         inet6_register_protosw(&udpv6_protosw);
1084 }