Merge branch 'for-linus' of git://git390.osdl.marist.edu/pub/scm/linux-2.6
[linux-2.6] / net / ipv6 / ipv6_sockglue.c
1 /*
2  *      IPv6 BSD socket options interface
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *
8  *      Based on linux/net/ipv4/ip_sockglue.c
9  *
10  *      $Id: ipv6_sockglue.c,v 1.41 2002/02/01 22:01:04 davem Exp $
11  *
12  *      This program is free software; you can redistribute it and/or
13  *      modify it under the terms of the GNU General Public License
14  *      as published by the Free Software Foundation; either version
15  *      2 of the License, or (at your option) any later version.
16  *
17  *      FIXME: Make the setsockopt code POSIX compliant: That is
18  *
19  *      o       Return -EINVAL for setsockopt of short lengths
20  *      o       Truncate getsockopt returns
21  *      o       Return an optlen of the truncated length if need be
22  *
23  *      Changes:
24  *      David L Stevens <dlstevens@us.ibm.com>:
25  *              - added multicast source filtering API for MLDv2
26  */
27
28 #include <linux/module.h>
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/types.h>
32 #include <linux/socket.h>
33 #include <linux/sockios.h>
34 #include <linux/net.h>
35 #include <linux/in6.h>
36 #include <linux/netdevice.h>
37 #include <linux/if_arp.h>
38 #include <linux/init.h>
39 #include <linux/sysctl.h>
40 #include <linux/netfilter.h>
41
42 #include <net/sock.h>
43 #include <net/snmp.h>
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/inet_common.h>
51 #include <net/tcp.h>
52 #include <net/udp.h>
53 #include <net/udplite.h>
54 #include <net/xfrm.h>
55
56 #include <asm/uaccess.h>
57
58 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
59
60 static struct inet6_protocol *ipv6_gso_pull_exthdrs(struct sk_buff *skb,
61                                                     int proto)
62 {
63         struct inet6_protocol *ops = NULL;
64
65         for (;;) {
66                 struct ipv6_opt_hdr *opth;
67                 int len;
68
69                 if (proto != NEXTHDR_HOP) {
70                         ops = rcu_dereference(inet6_protos[proto]);
71
72                         if (unlikely(!ops))
73                                 break;
74
75                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
76                                 break;
77                 }
78
79                 if (unlikely(!pskb_may_pull(skb, 8)))
80                         break;
81
82                 opth = (void *)skb->data;
83                 len = opth->hdrlen * 8 + 8;
84
85                 if (unlikely(!pskb_may_pull(skb, len)))
86                         break;
87
88                 proto = opth->nexthdr;
89                 __skb_pull(skb, len);
90         }
91
92         return ops;
93 }
94
95 static int ipv6_gso_send_check(struct sk_buff *skb)
96 {
97         struct ipv6hdr *ipv6h;
98         struct inet6_protocol *ops;
99         int err = -EINVAL;
100
101         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
102                 goto out;
103
104         ipv6h = ipv6_hdr(skb);
105         __skb_pull(skb, sizeof(*ipv6h));
106         err = -EPROTONOSUPPORT;
107
108         rcu_read_lock();
109         ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
110         if (likely(ops && ops->gso_send_check)) {
111                 skb_reset_transport_header(skb);
112                 err = ops->gso_send_check(skb);
113         }
114         rcu_read_unlock();
115
116 out:
117         return err;
118 }
119
120 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
121 {
122         struct sk_buff *segs = ERR_PTR(-EINVAL);
123         struct ipv6hdr *ipv6h;
124         struct inet6_protocol *ops;
125
126         if (!(features & NETIF_F_HW_CSUM))
127                 features &= ~NETIF_F_SG;
128
129         if (unlikely(skb_shinfo(skb)->gso_type &
130                      ~(SKB_GSO_UDP |
131                        SKB_GSO_DODGY |
132                        SKB_GSO_TCP_ECN |
133                        SKB_GSO_TCPV6 |
134                        0)))
135                 goto out;
136
137         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
138                 goto out;
139
140         ipv6h = ipv6_hdr(skb);
141         __skb_pull(skb, sizeof(*ipv6h));
142         segs = ERR_PTR(-EPROTONOSUPPORT);
143
144         rcu_read_lock();
145         ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
146         if (likely(ops && ops->gso_segment)) {
147                 skb_reset_transport_header(skb);
148                 segs = ops->gso_segment(skb, features);
149         }
150         rcu_read_unlock();
151
152         if (unlikely(IS_ERR(segs)))
153                 goto out;
154
155         for (skb = segs; skb; skb = skb->next) {
156                 ipv6h = ipv6_hdr(skb);
157                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
158                                            sizeof(*ipv6h));
159         }
160
161 out:
162         return segs;
163 }
164
165 static struct packet_type ipv6_packet_type = {
166         .type = __constant_htons(ETH_P_IPV6),
167         .func = ipv6_rcv,
168         .gso_send_check = ipv6_gso_send_check,
169         .gso_segment = ipv6_gso_segment,
170 };
171
172 struct ip6_ra_chain *ip6_ra_chain;
173 DEFINE_RWLOCK(ip6_ra_lock);
174
175 int ip6_ra_control(struct sock *sk, int sel, void (*destructor)(struct sock *))
176 {
177         struct ip6_ra_chain *ra, *new_ra, **rap;
178
179         /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
180         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
181                 return -EINVAL;
182
183         new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
184
185         write_lock_bh(&ip6_ra_lock);
186         for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
187                 if (ra->sk == sk) {
188                         if (sel>=0) {
189                                 write_unlock_bh(&ip6_ra_lock);
190                                 kfree(new_ra);
191                                 return -EADDRINUSE;
192                         }
193
194                         *rap = ra->next;
195                         write_unlock_bh(&ip6_ra_lock);
196
197                         if (ra->destructor)
198                                 ra->destructor(sk);
199                         sock_put(sk);
200                         kfree(ra);
201                         return 0;
202                 }
203         }
204         if (new_ra == NULL) {
205                 write_unlock_bh(&ip6_ra_lock);
206                 return -ENOBUFS;
207         }
208         new_ra->sk = sk;
209         new_ra->sel = sel;
210         new_ra->destructor = destructor;
211         new_ra->next = ra;
212         *rap = new_ra;
213         sock_hold(sk);
214         write_unlock_bh(&ip6_ra_lock);
215         return 0;
216 }
217
218 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
219                     char __user *optval, int optlen)
220 {
221         struct ipv6_pinfo *np = inet6_sk(sk);
222         int val, valbool;
223         int retv = -ENOPROTOOPT;
224
225         if (optval == NULL)
226                 val=0;
227         else if (get_user(val, (int __user *) optval))
228                 return -EFAULT;
229
230         valbool = (val!=0);
231
232         lock_sock(sk);
233
234         switch (optname) {
235
236         case IPV6_ADDRFORM:
237                 if (val == PF_INET) {
238                         struct ipv6_txoptions *opt;
239                         struct sk_buff *pktopt;
240
241                         if (sk->sk_protocol != IPPROTO_UDP &&
242                             sk->sk_protocol != IPPROTO_UDPLITE &&
243                             sk->sk_protocol != IPPROTO_TCP)
244                                 break;
245
246                         if (sk->sk_state != TCP_ESTABLISHED) {
247                                 retv = -ENOTCONN;
248                                 break;
249                         }
250
251                         if (ipv6_only_sock(sk) ||
252                             !(ipv6_addr_type(&np->daddr) & IPV6_ADDR_MAPPED)) {
253                                 retv = -EADDRNOTAVAIL;
254                                 break;
255                         }
256
257                         fl6_free_socklist(sk);
258                         ipv6_sock_mc_close(sk);
259
260                         /*
261                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
262                          * remove it from the refcnt debug socks count in the
263                          * original family...
264                          */
265                         sk_refcnt_debug_dec(sk);
266
267                         if (sk->sk_protocol == IPPROTO_TCP) {
268                                 struct inet_connection_sock *icsk = inet_csk(sk);
269
270                                 local_bh_disable();
271                                 sock_prot_dec_use(sk->sk_prot);
272                                 sock_prot_inc_use(&tcp_prot);
273                                 local_bh_enable();
274                                 sk->sk_prot = &tcp_prot;
275                                 icsk->icsk_af_ops = &ipv4_specific;
276                                 sk->sk_socket->ops = &inet_stream_ops;
277                                 sk->sk_family = PF_INET;
278                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
279                         } else {
280                                 struct proto *prot = &udp_prot;
281
282                                 if (sk->sk_protocol == IPPROTO_UDPLITE)
283                                         prot = &udplite_prot;
284                                 local_bh_disable();
285                                 sock_prot_dec_use(sk->sk_prot);
286                                 sock_prot_inc_use(prot);
287                                 local_bh_enable();
288                                 sk->sk_prot = prot;
289                                 sk->sk_socket->ops = &inet_dgram_ops;
290                                 sk->sk_family = PF_INET;
291                         }
292                         opt = xchg(&np->opt, NULL);
293                         if (opt)
294                                 sock_kfree_s(sk, opt, opt->tot_len);
295                         pktopt = xchg(&np->pktoptions, NULL);
296                         if (pktopt)
297                                 kfree_skb(pktopt);
298
299                         sk->sk_destruct = inet_sock_destruct;
300                         /*
301                          * ... and add it to the refcnt debug socks count
302                          * in the new family. -acme
303                          */
304                         sk_refcnt_debug_inc(sk);
305                         module_put(THIS_MODULE);
306                         retv = 0;
307                         break;
308                 }
309                 goto e_inval;
310
311         case IPV6_V6ONLY:
312                 if (inet_sk(sk)->num)
313                         goto e_inval;
314                 np->ipv6only = valbool;
315                 retv = 0;
316                 break;
317
318         case IPV6_RECVPKTINFO:
319                 np->rxopt.bits.rxinfo = valbool;
320                 retv = 0;
321                 break;
322
323         case IPV6_2292PKTINFO:
324                 np->rxopt.bits.rxoinfo = valbool;
325                 retv = 0;
326                 break;
327
328         case IPV6_RECVHOPLIMIT:
329                 np->rxopt.bits.rxhlim = valbool;
330                 retv = 0;
331                 break;
332
333         case IPV6_2292HOPLIMIT:
334                 np->rxopt.bits.rxohlim = valbool;
335                 retv = 0;
336                 break;
337
338         case IPV6_RECVRTHDR:
339                 if (val < 0 || val > 2)
340                         goto e_inval;
341                 np->rxopt.bits.srcrt = val;
342                 retv = 0;
343                 break;
344
345         case IPV6_2292RTHDR:
346                 if (val < 0 || val > 2)
347                         goto e_inval;
348                 np->rxopt.bits.osrcrt = val;
349                 retv = 0;
350                 break;
351
352         case IPV6_RECVHOPOPTS:
353                 np->rxopt.bits.hopopts = valbool;
354                 retv = 0;
355                 break;
356
357         case IPV6_2292HOPOPTS:
358                 np->rxopt.bits.ohopopts = valbool;
359                 retv = 0;
360                 break;
361
362         case IPV6_RECVDSTOPTS:
363                 np->rxopt.bits.dstopts = valbool;
364                 retv = 0;
365                 break;
366
367         case IPV6_2292DSTOPTS:
368                 np->rxopt.bits.odstopts = valbool;
369                 retv = 0;
370                 break;
371
372         case IPV6_TCLASS:
373                 if (val < -1 || val > 0xff)
374                         goto e_inval;
375                 np->tclass = val;
376                 retv = 0;
377                 break;
378
379         case IPV6_RECVTCLASS:
380                 np->rxopt.bits.rxtclass = valbool;
381                 retv = 0;
382                 break;
383
384         case IPV6_FLOWINFO:
385                 np->rxopt.bits.rxflow = valbool;
386                 retv = 0;
387                 break;
388
389         case IPV6_HOPOPTS:
390         case IPV6_RTHDRDSTOPTS:
391         case IPV6_RTHDR:
392         case IPV6_DSTOPTS:
393         {
394                 struct ipv6_txoptions *opt;
395                 if (optlen == 0)
396                         optval = NULL;
397
398                 /* hop-by-hop / destination options are privileged option */
399                 retv = -EPERM;
400                 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
401                         break;
402
403                 retv = -EINVAL;
404                 if (optlen & 0x7 || optlen > 8 * 255)
405                         break;
406
407                 opt = ipv6_renew_options(sk, np->opt, optname,
408                                          (struct ipv6_opt_hdr __user *)optval,
409                                          optlen);
410                 if (IS_ERR(opt)) {
411                         retv = PTR_ERR(opt);
412                         break;
413                 }
414
415                 /* routing header option needs extra check */
416                 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
417                         struct ipv6_rt_hdr *rthdr = opt->srcrt;
418                         switch (rthdr->type) {
419                         case IPV6_SRCRT_TYPE_0:
420 #ifdef CONFIG_IPV6_MIP6
421                         case IPV6_SRCRT_TYPE_2:
422 #endif
423                                 break;
424                         default:
425                                 goto sticky_done;
426                         }
427
428                         if ((rthdr->hdrlen & 1) ||
429                             (rthdr->hdrlen >> 1) != rthdr->segments_left)
430                                 goto sticky_done;
431                 }
432
433                 retv = 0;
434                 if (inet_sk(sk)->is_icsk) {
435                         if (opt) {
436                                 struct inet_connection_sock *icsk = inet_csk(sk);
437                                 if (!((1 << sk->sk_state) &
438                                       (TCPF_LISTEN | TCPF_CLOSE))
439                                     && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
440                                         icsk->icsk_ext_hdr_len =
441                                                 opt->opt_flen + opt->opt_nflen;
442                                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
443                                 }
444                         }
445                         opt = xchg(&np->opt, opt);
446                         sk_dst_reset(sk);
447                 } else {
448                         write_lock(&sk->sk_dst_lock);
449                         opt = xchg(&np->opt, opt);
450                         write_unlock(&sk->sk_dst_lock);
451                         sk_dst_reset(sk);
452                 }
453 sticky_done:
454                 if (opt)
455                         sock_kfree_s(sk, opt, opt->tot_len);
456                 break;
457         }
458
459         case IPV6_2292PKTOPTIONS:
460         {
461                 struct ipv6_txoptions *opt = NULL;
462                 struct msghdr msg;
463                 struct flowi fl;
464                 int junk;
465
466                 fl.fl6_flowlabel = 0;
467                 fl.oif = sk->sk_bound_dev_if;
468
469                 if (optlen == 0)
470                         goto update;
471
472                 /* 1K is probably excessive
473                  * 1K is surely not enough, 2K per standard header is 16K.
474                  */
475                 retv = -EINVAL;
476                 if (optlen > 64*1024)
477                         break;
478
479                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
480                 retv = -ENOBUFS;
481                 if (opt == NULL)
482                         break;
483
484                 memset(opt, 0, sizeof(*opt));
485                 opt->tot_len = sizeof(*opt) + optlen;
486                 retv = -EFAULT;
487                 if (copy_from_user(opt+1, optval, optlen))
488                         goto done;
489
490                 msg.msg_controllen = optlen;
491                 msg.msg_control = (void*)(opt+1);
492
493                 retv = datagram_send_ctl(&msg, &fl, opt, &junk, &junk);
494                 if (retv)
495                         goto done;
496 update:
497                 retv = 0;
498                 if (inet_sk(sk)->is_icsk) {
499                         if (opt) {
500                                 struct inet_connection_sock *icsk = inet_csk(sk);
501                                 if (!((1 << sk->sk_state) &
502                                       (TCPF_LISTEN | TCPF_CLOSE))
503                                     && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
504                                         icsk->icsk_ext_hdr_len =
505                                                 opt->opt_flen + opt->opt_nflen;
506                                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
507                                 }
508                         }
509                         opt = xchg(&np->opt, opt);
510                         sk_dst_reset(sk);
511                 } else {
512                         write_lock(&sk->sk_dst_lock);
513                         opt = xchg(&np->opt, opt);
514                         write_unlock(&sk->sk_dst_lock);
515                         sk_dst_reset(sk);
516                 }
517
518 done:
519                 if (opt)
520                         sock_kfree_s(sk, opt, opt->tot_len);
521                 break;
522         }
523         case IPV6_UNICAST_HOPS:
524                 if (val > 255 || val < -1)
525                         goto e_inval;
526                 np->hop_limit = val;
527                 retv = 0;
528                 break;
529
530         case IPV6_MULTICAST_HOPS:
531                 if (sk->sk_type == SOCK_STREAM)
532                         goto e_inval;
533                 if (val > 255 || val < -1)
534                         goto e_inval;
535                 np->mcast_hops = val;
536                 retv = 0;
537                 break;
538
539         case IPV6_MULTICAST_LOOP:
540                 np->mc_loop = valbool;
541                 retv = 0;
542                 break;
543
544         case IPV6_MULTICAST_IF:
545                 if (sk->sk_type == SOCK_STREAM)
546                         goto e_inval;
547                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
548                         goto e_inval;
549
550                 if (__dev_get_by_index(val) == NULL) {
551                         retv = -ENODEV;
552                         break;
553                 }
554                 np->mcast_oif = val;
555                 retv = 0;
556                 break;
557         case IPV6_ADD_MEMBERSHIP:
558         case IPV6_DROP_MEMBERSHIP:
559         {
560                 struct ipv6_mreq mreq;
561
562                 retv = -EFAULT;
563                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
564                         break;
565
566                 if (optname == IPV6_ADD_MEMBERSHIP)
567                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
568                 else
569                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
570                 break;
571         }
572         case IPV6_JOIN_ANYCAST:
573         case IPV6_LEAVE_ANYCAST:
574         {
575                 struct ipv6_mreq mreq;
576
577                 if (optlen != sizeof(struct ipv6_mreq))
578                         goto e_inval;
579
580                 retv = -EFAULT;
581                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
582                         break;
583
584                 if (optname == IPV6_JOIN_ANYCAST)
585                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
586                 else
587                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
588                 break;
589         }
590         case MCAST_JOIN_GROUP:
591         case MCAST_LEAVE_GROUP:
592         {
593                 struct group_req greq;
594                 struct sockaddr_in6 *psin6;
595
596                 retv = -EFAULT;
597                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
598                         break;
599                 if (greq.gr_group.ss_family != AF_INET6) {
600                         retv = -EADDRNOTAVAIL;
601                         break;
602                 }
603                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
604                 if (optname == MCAST_JOIN_GROUP)
605                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
606                                 &psin6->sin6_addr);
607                 else
608                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
609                                 &psin6->sin6_addr);
610                 break;
611         }
612         case MCAST_JOIN_SOURCE_GROUP:
613         case MCAST_LEAVE_SOURCE_GROUP:
614         case MCAST_BLOCK_SOURCE:
615         case MCAST_UNBLOCK_SOURCE:
616         {
617                 struct group_source_req greqs;
618                 int omode, add;
619
620                 if (optlen != sizeof(struct group_source_req))
621                         goto e_inval;
622                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
623                         retv = -EFAULT;
624                         break;
625                 }
626                 if (greqs.gsr_group.ss_family != AF_INET6 ||
627                     greqs.gsr_source.ss_family != AF_INET6) {
628                         retv = -EADDRNOTAVAIL;
629                         break;
630                 }
631                 if (optname == MCAST_BLOCK_SOURCE) {
632                         omode = MCAST_EXCLUDE;
633                         add = 1;
634                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
635                         omode = MCAST_EXCLUDE;
636                         add = 0;
637                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
638                         struct sockaddr_in6 *psin6;
639
640                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
641                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
642                                 &psin6->sin6_addr);
643                         /* prior join w/ different source is ok */
644                         if (retv && retv != -EADDRINUSE)
645                                 break;
646                         omode = MCAST_INCLUDE;
647                         add = 1;
648                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
649                         omode = MCAST_INCLUDE;
650                         add = 0;
651                 }
652                 retv = ip6_mc_source(add, omode, sk, &greqs);
653                 break;
654         }
655         case MCAST_MSFILTER:
656         {
657                 extern int sysctl_mld_max_msf;
658                 struct group_filter *gsf;
659
660                 if (optlen < GROUP_FILTER_SIZE(0))
661                         goto e_inval;
662                 if (optlen > sysctl_optmem_max) {
663                         retv = -ENOBUFS;
664                         break;
665                 }
666                 gsf = kmalloc(optlen,GFP_KERNEL);
667                 if (gsf == 0) {
668                         retv = -ENOBUFS;
669                         break;
670                 }
671                 retv = -EFAULT;
672                 if (copy_from_user(gsf, optval, optlen)) {
673                         kfree(gsf);
674                         break;
675                 }
676                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
677                 if (gsf->gf_numsrc >= 0x1ffffffU ||
678                     gsf->gf_numsrc > sysctl_mld_max_msf) {
679                         kfree(gsf);
680                         retv = -ENOBUFS;
681                         break;
682                 }
683                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
684                         kfree(gsf);
685                         retv = -EINVAL;
686                         break;
687                 }
688                 retv = ip6_mc_msfilter(sk, gsf);
689                 kfree(gsf);
690
691                 break;
692         }
693         case IPV6_ROUTER_ALERT:
694                 retv = ip6_ra_control(sk, val, NULL);
695                 break;
696         case IPV6_MTU_DISCOVER:
697                 if (val<0 || val>3)
698                         goto e_inval;
699                 np->pmtudisc = val;
700                 retv = 0;
701                 break;
702         case IPV6_MTU:
703                 if (val && val < IPV6_MIN_MTU)
704                         goto e_inval;
705                 np->frag_size = val;
706                 retv = 0;
707                 break;
708         case IPV6_RECVERR:
709                 np->recverr = valbool;
710                 if (!val)
711                         skb_queue_purge(&sk->sk_error_queue);
712                 retv = 0;
713                 break;
714         case IPV6_FLOWINFO_SEND:
715                 np->sndflow = valbool;
716                 retv = 0;
717                 break;
718         case IPV6_FLOWLABEL_MGR:
719                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
720                 break;
721         case IPV6_IPSEC_POLICY:
722         case IPV6_XFRM_POLICY:
723                 retv = -EPERM;
724                 if (!capable(CAP_NET_ADMIN))
725                         break;
726                 retv = xfrm_user_policy(sk, optname, optval, optlen);
727                 break;
728
729         }
730         release_sock(sk);
731
732         return retv;
733
734 e_inval:
735         release_sock(sk);
736         return -EINVAL;
737 }
738
739 int ipv6_setsockopt(struct sock *sk, int level, int optname,
740                     char __user *optval, int optlen)
741 {
742         int err;
743
744         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
745                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
746
747         if (level != SOL_IPV6)
748                 return -ENOPROTOOPT;
749
750         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
751 #ifdef CONFIG_NETFILTER
752         /* we need to exclude all possible ENOPROTOOPTs except default case */
753         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
754                         optname != IPV6_XFRM_POLICY) {
755                 lock_sock(sk);
756                 err = nf_setsockopt(sk, PF_INET6, optname, optval,
757                                 optlen);
758                 release_sock(sk);
759         }
760 #endif
761         return err;
762 }
763
764 EXPORT_SYMBOL(ipv6_setsockopt);
765
766 #ifdef CONFIG_COMPAT
767 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
768                            char __user *optval, int optlen)
769 {
770         int err;
771
772         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
773                 if (udp_prot.compat_setsockopt != NULL)
774                         return udp_prot.compat_setsockopt(sk, level, optname,
775                                                           optval, optlen);
776                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
777         }
778
779         if (level != SOL_IPV6)
780                 return -ENOPROTOOPT;
781
782         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
783 #ifdef CONFIG_NETFILTER
784         /* we need to exclude all possible ENOPROTOOPTs except default case */
785         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
786             optname != IPV6_XFRM_POLICY) {
787                 lock_sock(sk);
788                 err = compat_nf_setsockopt(sk, PF_INET6, optname,
789                                            optval, optlen);
790                 release_sock(sk);
791         }
792 #endif
793         return err;
794 }
795
796 EXPORT_SYMBOL(compat_ipv6_setsockopt);
797 #endif
798
799 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
800                                   int optname, char __user *optval, int len)
801 {
802         struct ipv6_opt_hdr *hdr;
803
804         if (!opt)
805                 return 0;
806
807         switch(optname) {
808         case IPV6_HOPOPTS:
809                 hdr = opt->hopopt;
810                 break;
811         case IPV6_RTHDRDSTOPTS:
812                 hdr = opt->dst0opt;
813                 break;
814         case IPV6_RTHDR:
815                 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
816                 break;
817         case IPV6_DSTOPTS:
818                 hdr = opt->dst1opt;
819                 break;
820         default:
821                 return -EINVAL; /* should not happen */
822         }
823
824         if (!hdr)
825                 return 0;
826
827         len = min_t(unsigned int, len, ipv6_optlen(hdr));
828         if (copy_to_user(optval, hdr, len));
829                 return -EFAULT;
830         return ipv6_optlen(hdr);
831 }
832
833 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
834                     char __user *optval, int __user *optlen)
835 {
836         struct ipv6_pinfo *np = inet6_sk(sk);
837         int len;
838         int val;
839
840         if (get_user(len, optlen))
841                 return -EFAULT;
842         switch (optname) {
843         case IPV6_ADDRFORM:
844                 if (sk->sk_protocol != IPPROTO_UDP &&
845                     sk->sk_protocol != IPPROTO_UDPLITE &&
846                     sk->sk_protocol != IPPROTO_TCP)
847                         return -EINVAL;
848                 if (sk->sk_state != TCP_ESTABLISHED)
849                         return -ENOTCONN;
850                 val = sk->sk_family;
851                 break;
852         case MCAST_MSFILTER:
853         {
854                 struct group_filter gsf;
855                 int err;
856
857                 if (len < GROUP_FILTER_SIZE(0))
858                         return -EINVAL;
859                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
860                         return -EFAULT;
861                 lock_sock(sk);
862                 err = ip6_mc_msfget(sk, &gsf,
863                         (struct group_filter __user *)optval, optlen);
864                 release_sock(sk);
865                 return err;
866         }
867
868         case IPV6_2292PKTOPTIONS:
869         {
870                 struct msghdr msg;
871                 struct sk_buff *skb;
872
873                 if (sk->sk_type != SOCK_STREAM)
874                         return -ENOPROTOOPT;
875
876                 msg.msg_control = optval;
877                 msg.msg_controllen = len;
878                 msg.msg_flags = 0;
879
880                 lock_sock(sk);
881                 skb = np->pktoptions;
882                 if (skb)
883                         atomic_inc(&skb->users);
884                 release_sock(sk);
885
886                 if (skb) {
887                         int err = datagram_recv_ctl(sk, &msg, skb);
888                         kfree_skb(skb);
889                         if (err)
890                                 return err;
891                 } else {
892                         if (np->rxopt.bits.rxinfo) {
893                                 struct in6_pktinfo src_info;
894                                 src_info.ipi6_ifindex = np->mcast_oif;
895                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
896                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
897                         }
898                         if (np->rxopt.bits.rxhlim) {
899                                 int hlim = np->mcast_hops;
900                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
901                         }
902                         if (np->rxopt.bits.rxoinfo) {
903                                 struct in6_pktinfo src_info;
904                                 src_info.ipi6_ifindex = np->mcast_oif;
905                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
906                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
907                         }
908                         if (np->rxopt.bits.rxohlim) {
909                                 int hlim = np->mcast_hops;
910                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
911                         }
912                 }
913                 len -= msg.msg_controllen;
914                 return put_user(len, optlen);
915         }
916         case IPV6_MTU:
917         {
918                 struct dst_entry *dst;
919                 val = 0;
920                 lock_sock(sk);
921                 dst = sk_dst_get(sk);
922                 if (dst) {
923                         val = dst_mtu(dst);
924                         dst_release(dst);
925                 }
926                 release_sock(sk);
927                 if (!val)
928                         return -ENOTCONN;
929                 break;
930         }
931
932         case IPV6_V6ONLY:
933                 val = np->ipv6only;
934                 break;
935
936         case IPV6_RECVPKTINFO:
937                 val = np->rxopt.bits.rxinfo;
938                 break;
939
940         case IPV6_2292PKTINFO:
941                 val = np->rxopt.bits.rxoinfo;
942                 break;
943
944         case IPV6_RECVHOPLIMIT:
945                 val = np->rxopt.bits.rxhlim;
946                 break;
947
948         case IPV6_2292HOPLIMIT:
949                 val = np->rxopt.bits.rxohlim;
950                 break;
951
952         case IPV6_RECVRTHDR:
953                 val = np->rxopt.bits.srcrt;
954                 break;
955
956         case IPV6_2292RTHDR:
957                 val = np->rxopt.bits.osrcrt;
958                 break;
959
960         case IPV6_HOPOPTS:
961         case IPV6_RTHDRDSTOPTS:
962         case IPV6_RTHDR:
963         case IPV6_DSTOPTS:
964         {
965
966                 lock_sock(sk);
967                 len = ipv6_getsockopt_sticky(sk, np->opt,
968                                              optname, optval, len);
969                 release_sock(sk);
970                 return put_user(len, optlen);
971         }
972
973         case IPV6_RECVHOPOPTS:
974                 val = np->rxopt.bits.hopopts;
975                 break;
976
977         case IPV6_2292HOPOPTS:
978                 val = np->rxopt.bits.ohopopts;
979                 break;
980
981         case IPV6_RECVDSTOPTS:
982                 val = np->rxopt.bits.dstopts;
983                 break;
984
985         case IPV6_2292DSTOPTS:
986                 val = np->rxopt.bits.odstopts;
987                 break;
988
989         case IPV6_TCLASS:
990                 val = np->tclass;
991                 if (val < 0)
992                         val = 0;
993                 break;
994
995         case IPV6_RECVTCLASS:
996                 val = np->rxopt.bits.rxtclass;
997                 break;
998
999         case IPV6_FLOWINFO:
1000                 val = np->rxopt.bits.rxflow;
1001                 break;
1002
1003         case IPV6_UNICAST_HOPS:
1004         case IPV6_MULTICAST_HOPS:
1005         {
1006                 struct dst_entry *dst;
1007
1008                 if (optname == IPV6_UNICAST_HOPS)
1009                         val = np->hop_limit;
1010                 else
1011                         val = np->mcast_hops;
1012
1013                 dst = sk_dst_get(sk);
1014                 if (dst) {
1015                         if (val < 0)
1016                                 val = dst_metric(dst, RTAX_HOPLIMIT);
1017                         if (val < 0)
1018                                 val = ipv6_get_hoplimit(dst->dev);
1019                         dst_release(dst);
1020                 }
1021                 if (val < 0)
1022                         val = ipv6_devconf.hop_limit;
1023                 break;
1024         }
1025
1026         case IPV6_MULTICAST_LOOP:
1027                 val = np->mc_loop;
1028                 break;
1029
1030         case IPV6_MULTICAST_IF:
1031                 val = np->mcast_oif;
1032                 break;
1033
1034         case IPV6_MTU_DISCOVER:
1035                 val = np->pmtudisc;
1036                 break;
1037
1038         case IPV6_RECVERR:
1039                 val = np->recverr;
1040                 break;
1041
1042         case IPV6_FLOWINFO_SEND:
1043                 val = np->sndflow;
1044                 break;
1045
1046         default:
1047                 return -EINVAL;
1048         }
1049         len = min_t(unsigned int, sizeof(int), len);
1050         if(put_user(len, optlen))
1051                 return -EFAULT;
1052         if(copy_to_user(optval,&val,len))
1053                 return -EFAULT;
1054         return 0;
1055 }
1056
1057 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1058                     char __user *optval, int __user *optlen)
1059 {
1060         int err;
1061
1062         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1063                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1064
1065         if(level != SOL_IPV6)
1066                 return -ENOPROTOOPT;
1067
1068         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1069 #ifdef CONFIG_NETFILTER
1070         /* we need to exclude all possible EINVALs except default case */
1071         if (err == -EINVAL && optname != IPV6_ADDRFORM &&
1072                         optname != MCAST_MSFILTER) {
1073                 int len;
1074
1075                 if (get_user(len, optlen))
1076                         return -EFAULT;
1077
1078                 lock_sock(sk);
1079                 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1080                                 &len);
1081                 release_sock(sk);
1082                 if (err >= 0)
1083                         err = put_user(len, optlen);
1084         }
1085 #endif
1086         return err;
1087 }
1088
1089 EXPORT_SYMBOL(ipv6_getsockopt);
1090
1091 #ifdef CONFIG_COMPAT
1092 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1093                            char __user *optval, int __user *optlen)
1094 {
1095         int err;
1096
1097         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1098                 if (udp_prot.compat_getsockopt != NULL)
1099                         return udp_prot.compat_getsockopt(sk, level, optname,
1100                                                           optval, optlen);
1101                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1102         }
1103
1104         if (level != SOL_IPV6)
1105                 return -ENOPROTOOPT;
1106
1107         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1108 #ifdef CONFIG_NETFILTER
1109         /* we need to exclude all possible EINVALs except default case */
1110         if (err == -EINVAL && optname != IPV6_ADDRFORM &&
1111                         optname != MCAST_MSFILTER) {
1112                 int len;
1113
1114                 if (get_user(len, optlen))
1115                         return -EFAULT;
1116
1117                 lock_sock(sk);
1118                 err = compat_nf_getsockopt(sk, PF_INET6,
1119                                            optname, optval, &len);
1120                 release_sock(sk);
1121                 if (err >= 0)
1122                         err = put_user(len, optlen);
1123         }
1124 #endif
1125         return err;
1126 }
1127
1128 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1129 #endif
1130
1131 void __init ipv6_packet_init(void)
1132 {
1133         dev_add_pack(&ipv6_packet_type);
1134 }
1135
1136 void ipv6_packet_cleanup(void)
1137 {
1138         dev_remove_pack(&ipv6_packet_type);
1139 }