ipv6: When we droped a packet, we should return NET_RX_DROP instead of 0
[linux-2.6.git] / 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  *      This program is free software; you can redistribute it and/or
11  *      modify it under the terms of the GNU General Public License
12  *      as published by the Free Software Foundation; either version
13  *      2 of the License, or (at your option) any later version.
14  *
15  *      FIXME: Make the setsockopt code POSIX compliant: That is
16  *
17  *      o       Truncate getsockopt returns
18  *      o       Return an optlen of the truncated length if need be
19  *
20  *      Changes:
21  *      David L Stevens <dlstevens@us.ibm.com>:
22  *              - added multicast source filtering API for MLDv2
23  */
24
25 #include <linux/module.h>
26 #include <linux/capability.h>
27 #include <linux/errno.h>
28 #include <linux/types.h>
29 #include <linux/socket.h>
30 #include <linux/sockios.h>
31 #include <linux/net.h>
32 #include <linux/in6.h>
33 #include <linux/mroute6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/init.h>
37 #include <linux/sysctl.h>
38 #include <linux/netfilter.h>
39
40 #include <net/sock.h>
41 #include <net/snmp.h>
42 #include <net/ipv6.h>
43 #include <net/ndisc.h>
44 #include <net/protocol.h>
45 #include <net/transp_v6.h>
46 #include <net/ip6_route.h>
47 #include <net/addrconf.h>
48 #include <net/inet_common.h>
49 #include <net/tcp.h>
50 #include <net/udp.h>
51 #include <net/udplite.h>
52 #include <net/xfrm.h>
53 #include <net/compat.h>
54
55 #include <asm/uaccess.h>
56
57 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
58
59 struct ip6_ra_chain *ip6_ra_chain;
60 DEFINE_RWLOCK(ip6_ra_lock);
61
62 int ip6_ra_control(struct sock *sk, int sel)
63 {
64         struct ip6_ra_chain *ra, *new_ra, **rap;
65
66         /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
67         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
68                 return -ENOPROTOOPT;
69
70         new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
71
72         write_lock_bh(&ip6_ra_lock);
73         for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
74                 if (ra->sk == sk) {
75                         if (sel>=0) {
76                                 write_unlock_bh(&ip6_ra_lock);
77                                 kfree(new_ra);
78                                 return -EADDRINUSE;
79                         }
80
81                         *rap = ra->next;
82                         write_unlock_bh(&ip6_ra_lock);
83
84                         sock_put(sk);
85                         kfree(ra);
86                         return 0;
87                 }
88         }
89         if (new_ra == NULL) {
90                 write_unlock_bh(&ip6_ra_lock);
91                 return -ENOBUFS;
92         }
93         new_ra->sk = sk;
94         new_ra->sel = sel;
95         new_ra->next = ra;
96         *rap = new_ra;
97         sock_hold(sk);
98         write_unlock_bh(&ip6_ra_lock);
99         return 0;
100 }
101
102 static
103 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
104                                            struct ipv6_txoptions *opt)
105 {
106         if (inet_sk(sk)->is_icsk) {
107                 if (opt &&
108                     !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
109                     inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
110                         struct inet_connection_sock *icsk = inet_csk(sk);
111                         icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
112                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
113                 }
114                 opt = xchg(&inet6_sk(sk)->opt, opt);
115         } else {
116                 write_lock(&sk->sk_dst_lock);
117                 opt = xchg(&inet6_sk(sk)->opt, opt);
118                 write_unlock(&sk->sk_dst_lock);
119         }
120         sk_dst_reset(sk);
121
122         return opt;
123 }
124
125 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
126                     char __user *optval, int optlen)
127 {
128         struct ipv6_pinfo *np = inet6_sk(sk);
129         struct net *net = sock_net(sk);
130         int val, valbool;
131         int retv = -ENOPROTOOPT;
132
133         if (optval == NULL)
134                 val=0;
135         else {
136                 if (optlen >= sizeof(int)) {
137                         if (get_user(val, (int __user *) optval))
138                                 return -EFAULT;
139                 } else
140                         val = 0;
141         }
142
143         valbool = (val!=0);
144
145         if (ip6_mroute_opt(optname))
146                 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
147
148         lock_sock(sk);
149
150         switch (optname) {
151
152         case IPV6_ADDRFORM:
153                 if (optlen < sizeof(int))
154                         goto e_inval;
155                 if (val == PF_INET) {
156                         struct ipv6_txoptions *opt;
157                         struct sk_buff *pktopt;
158
159                         if (sk->sk_type == SOCK_RAW)
160                                 break;
161
162                         if (sk->sk_protocol == IPPROTO_UDP ||
163                             sk->sk_protocol == IPPROTO_UDPLITE) {
164                                 struct udp_sock *up = udp_sk(sk);
165                                 if (up->pending == AF_INET6) {
166                                         retv = -EBUSY;
167                                         break;
168                                 }
169                         } else if (sk->sk_protocol != IPPROTO_TCP)
170                                 break;
171
172                         if (sk->sk_state != TCP_ESTABLISHED) {
173                                 retv = -ENOTCONN;
174                                 break;
175                         }
176
177                         if (ipv6_only_sock(sk) ||
178                             !ipv6_addr_v4mapped(&np->daddr)) {
179                                 retv = -EADDRNOTAVAIL;
180                                 break;
181                         }
182
183                         fl6_free_socklist(sk);
184                         ipv6_sock_mc_close(sk);
185
186                         /*
187                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
188                          * remove it from the refcnt debug socks count in the
189                          * original family...
190                          */
191                         sk_refcnt_debug_dec(sk);
192
193                         if (sk->sk_protocol == IPPROTO_TCP) {
194                                 struct inet_connection_sock *icsk = inet_csk(sk);
195                                 local_bh_disable();
196                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
197                                 sock_prot_inuse_add(net, &tcp_prot, 1);
198                                 local_bh_enable();
199                                 sk->sk_prot = &tcp_prot;
200                                 icsk->icsk_af_ops = &ipv4_specific;
201                                 sk->sk_socket->ops = &inet_stream_ops;
202                                 sk->sk_family = PF_INET;
203                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
204                         } else {
205                                 struct proto *prot = &udp_prot;
206
207                                 if (sk->sk_protocol == IPPROTO_UDPLITE)
208                                         prot = &udplite_prot;
209                                 local_bh_disable();
210                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
211                                 sock_prot_inuse_add(net, prot, 1);
212                                 local_bh_enable();
213                                 sk->sk_prot = prot;
214                                 sk->sk_socket->ops = &inet_dgram_ops;
215                                 sk->sk_family = PF_INET;
216                         }
217                         opt = xchg(&np->opt, NULL);
218                         if (opt)
219                                 sock_kfree_s(sk, opt, opt->tot_len);
220                         pktopt = xchg(&np->pktoptions, NULL);
221                         if (pktopt)
222                                 kfree_skb(pktopt);
223
224                         sk->sk_destruct = inet_sock_destruct;
225                         /*
226                          * ... and add it to the refcnt debug socks count
227                          * in the new family. -acme
228                          */
229                         sk_refcnt_debug_inc(sk);
230                         module_put(THIS_MODULE);
231                         retv = 0;
232                         break;
233                 }
234                 goto e_inval;
235
236         case IPV6_V6ONLY:
237                 if (optlen < sizeof(int) ||
238                     inet_sk(sk)->num)
239                         goto e_inval;
240                 np->ipv6only = valbool;
241                 retv = 0;
242                 break;
243
244         case IPV6_RECVPKTINFO:
245                 if (optlen < sizeof(int))
246                         goto e_inval;
247                 np->rxopt.bits.rxinfo = valbool;
248                 retv = 0;
249                 break;
250
251         case IPV6_2292PKTINFO:
252                 if (optlen < sizeof(int))
253                         goto e_inval;
254                 np->rxopt.bits.rxoinfo = valbool;
255                 retv = 0;
256                 break;
257
258         case IPV6_RECVHOPLIMIT:
259                 if (optlen < sizeof(int))
260                         goto e_inval;
261                 np->rxopt.bits.rxhlim = valbool;
262                 retv = 0;
263                 break;
264
265         case IPV6_2292HOPLIMIT:
266                 if (optlen < sizeof(int))
267                         goto e_inval;
268                 np->rxopt.bits.rxohlim = valbool;
269                 retv = 0;
270                 break;
271
272         case IPV6_RECVRTHDR:
273                 if (optlen < sizeof(int))
274                         goto e_inval;
275                 np->rxopt.bits.srcrt = valbool;
276                 retv = 0;
277                 break;
278
279         case IPV6_2292RTHDR:
280                 if (optlen < sizeof(int))
281                         goto e_inval;
282                 np->rxopt.bits.osrcrt = valbool;
283                 retv = 0;
284                 break;
285
286         case IPV6_RECVHOPOPTS:
287                 if (optlen < sizeof(int))
288                         goto e_inval;
289                 np->rxopt.bits.hopopts = valbool;
290                 retv = 0;
291                 break;
292
293         case IPV6_2292HOPOPTS:
294                 if (optlen < sizeof(int))
295                         goto e_inval;
296                 np->rxopt.bits.ohopopts = valbool;
297                 retv = 0;
298                 break;
299
300         case IPV6_RECVDSTOPTS:
301                 if (optlen < sizeof(int))
302                         goto e_inval;
303                 np->rxopt.bits.dstopts = valbool;
304                 retv = 0;
305                 break;
306
307         case IPV6_2292DSTOPTS:
308                 if (optlen < sizeof(int))
309                         goto e_inval;
310                 np->rxopt.bits.odstopts = valbool;
311                 retv = 0;
312                 break;
313
314         case IPV6_TCLASS:
315                 if (optlen < sizeof(int))
316                         goto e_inval;
317                 if (val < -1 || val > 0xff)
318                         goto e_inval;
319                 np->tclass = val;
320                 retv = 0;
321                 break;
322
323         case IPV6_RECVTCLASS:
324                 if (optlen < sizeof(int))
325                         goto e_inval;
326                 np->rxopt.bits.rxtclass = valbool;
327                 retv = 0;
328                 break;
329
330         case IPV6_FLOWINFO:
331                 if (optlen < sizeof(int))
332                         goto e_inval;
333                 np->rxopt.bits.rxflow = valbool;
334                 retv = 0;
335                 break;
336
337         case IPV6_HOPOPTS:
338         case IPV6_RTHDRDSTOPTS:
339         case IPV6_RTHDR:
340         case IPV6_DSTOPTS:
341         {
342                 struct ipv6_txoptions *opt;
343
344                 /* remove any sticky options header with a zero option
345                  * length, per RFC3542.
346                  */
347                 if (optlen == 0)
348                         optval = NULL;
349                 else if (optval == NULL)
350                         goto e_inval;
351                 else if (optlen < sizeof(struct ipv6_opt_hdr) ||
352                          optlen & 0x7 || optlen > 8 * 255)
353                         goto e_inval;
354
355                 /* hop-by-hop / destination options are privileged option */
356                 retv = -EPERM;
357                 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
358                         break;
359
360                 opt = ipv6_renew_options(sk, np->opt, optname,
361                                          (struct ipv6_opt_hdr __user *)optval,
362                                          optlen);
363                 if (IS_ERR(opt)) {
364                         retv = PTR_ERR(opt);
365                         break;
366                 }
367
368                 /* routing header option needs extra check */
369                 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
370                         struct ipv6_rt_hdr *rthdr = opt->srcrt;
371                         switch (rthdr->type) {
372 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
373                         case IPV6_SRCRT_TYPE_2:
374                                 break;
375 #endif
376                         default:
377                                 goto sticky_done;
378                         }
379
380                         if ((rthdr->hdrlen & 1) ||
381                             (rthdr->hdrlen >> 1) != rthdr->segments_left)
382                                 goto sticky_done;
383                 }
384
385                 retv = 0;
386                 opt = ipv6_update_options(sk, opt);
387 sticky_done:
388                 if (opt)
389                         sock_kfree_s(sk, opt, opt->tot_len);
390                 break;
391         }
392
393         case IPV6_2292PKTOPTIONS:
394         {
395                 struct ipv6_txoptions *opt = NULL;
396                 struct msghdr msg;
397                 struct flowi fl;
398                 int junk;
399
400                 fl.fl6_flowlabel = 0;
401                 fl.oif = sk->sk_bound_dev_if;
402
403                 if (optlen == 0)
404                         goto update;
405
406                 /* 1K is probably excessive
407                  * 1K is surely not enough, 2K per standard header is 16K.
408                  */
409                 retv = -EINVAL;
410                 if (optlen > 64*1024)
411                         break;
412
413                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
414                 retv = -ENOBUFS;
415                 if (opt == NULL)
416                         break;
417
418                 memset(opt, 0, sizeof(*opt));
419                 opt->tot_len = sizeof(*opt) + optlen;
420                 retv = -EFAULT;
421                 if (copy_from_user(opt+1, optval, optlen))
422                         goto done;
423
424                 msg.msg_controllen = optlen;
425                 msg.msg_control = (void*)(opt+1);
426
427                 retv = datagram_send_ctl(net, &msg, &fl, opt, &junk, &junk);
428                 if (retv)
429                         goto done;
430 update:
431                 retv = 0;
432                 opt = ipv6_update_options(sk, opt);
433 done:
434                 if (opt)
435                         sock_kfree_s(sk, opt, opt->tot_len);
436                 break;
437         }
438         case IPV6_UNICAST_HOPS:
439                 if (optlen < sizeof(int))
440                         goto e_inval;
441                 if (val > 255 || val < -1)
442                         goto e_inval;
443                 np->hop_limit = val;
444                 retv = 0;
445                 break;
446
447         case IPV6_MULTICAST_HOPS:
448                 if (sk->sk_type == SOCK_STREAM)
449                         break;
450                 if (optlen < sizeof(int))
451                         goto e_inval;
452                 if (val > 255 || val < -1)
453                         goto e_inval;
454                 np->mcast_hops = val;
455                 retv = 0;
456                 break;
457
458         case IPV6_MULTICAST_LOOP:
459                 if (optlen < sizeof(int))
460                         goto e_inval;
461                 if (val != valbool)
462                         goto e_inval;
463                 np->mc_loop = valbool;
464                 retv = 0;
465                 break;
466
467         case IPV6_MULTICAST_IF:
468                 if (sk->sk_type == SOCK_STREAM)
469                         break;
470                 if (optlen < sizeof(int))
471                         goto e_inval;
472
473                 if (val) {
474                         if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
475                                 goto e_inval;
476
477                         if (__dev_get_by_index(net, val) == NULL) {
478                                 retv = -ENODEV;
479                                 break;
480                         }
481                 }
482                 np->mcast_oif = val;
483                 retv = 0;
484                 break;
485         case IPV6_ADD_MEMBERSHIP:
486         case IPV6_DROP_MEMBERSHIP:
487         {
488                 struct ipv6_mreq mreq;
489
490                 if (optlen < sizeof(struct ipv6_mreq))
491                         goto e_inval;
492
493                 retv = -EPROTO;
494                 if (inet_sk(sk)->is_icsk)
495                         break;
496
497                 retv = -EFAULT;
498                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
499                         break;
500
501                 if (optname == IPV6_ADD_MEMBERSHIP)
502                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
503                 else
504                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
505                 break;
506         }
507         case IPV6_JOIN_ANYCAST:
508         case IPV6_LEAVE_ANYCAST:
509         {
510                 struct ipv6_mreq mreq;
511
512                 if (optlen < sizeof(struct ipv6_mreq))
513                         goto e_inval;
514
515                 retv = -EFAULT;
516                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
517                         break;
518
519                 if (optname == IPV6_JOIN_ANYCAST)
520                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
521                 else
522                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
523                 break;
524         }
525         case MCAST_JOIN_GROUP:
526         case MCAST_LEAVE_GROUP:
527         {
528                 struct group_req greq;
529                 struct sockaddr_in6 *psin6;
530
531                 if (optlen < sizeof(struct group_req))
532                         goto e_inval;
533
534                 retv = -EFAULT;
535                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
536                         break;
537                 if (greq.gr_group.ss_family != AF_INET6) {
538                         retv = -EADDRNOTAVAIL;
539                         break;
540                 }
541                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
542                 if (optname == MCAST_JOIN_GROUP)
543                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
544                                 &psin6->sin6_addr);
545                 else
546                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
547                                 &psin6->sin6_addr);
548                 break;
549         }
550         case MCAST_JOIN_SOURCE_GROUP:
551         case MCAST_LEAVE_SOURCE_GROUP:
552         case MCAST_BLOCK_SOURCE:
553         case MCAST_UNBLOCK_SOURCE:
554         {
555                 struct group_source_req greqs;
556                 int omode, add;
557
558                 if (optlen < sizeof(struct group_source_req))
559                         goto e_inval;
560                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
561                         retv = -EFAULT;
562                         break;
563                 }
564                 if (greqs.gsr_group.ss_family != AF_INET6 ||
565                     greqs.gsr_source.ss_family != AF_INET6) {
566                         retv = -EADDRNOTAVAIL;
567                         break;
568                 }
569                 if (optname == MCAST_BLOCK_SOURCE) {
570                         omode = MCAST_EXCLUDE;
571                         add = 1;
572                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
573                         omode = MCAST_EXCLUDE;
574                         add = 0;
575                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
576                         struct sockaddr_in6 *psin6;
577
578                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
579                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
580                                 &psin6->sin6_addr);
581                         /* prior join w/ different source is ok */
582                         if (retv && retv != -EADDRINUSE)
583                                 break;
584                         omode = MCAST_INCLUDE;
585                         add = 1;
586                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
587                         omode = MCAST_INCLUDE;
588                         add = 0;
589                 }
590                 retv = ip6_mc_source(add, omode, sk, &greqs);
591                 break;
592         }
593         case MCAST_MSFILTER:
594         {
595                 extern int sysctl_mld_max_msf;
596                 struct group_filter *gsf;
597
598                 if (optlen < GROUP_FILTER_SIZE(0))
599                         goto e_inval;
600                 if (optlen > sysctl_optmem_max) {
601                         retv = -ENOBUFS;
602                         break;
603                 }
604                 gsf = kmalloc(optlen,GFP_KERNEL);
605                 if (!gsf) {
606                         retv = -ENOBUFS;
607                         break;
608                 }
609                 retv = -EFAULT;
610                 if (copy_from_user(gsf, optval, optlen)) {
611                         kfree(gsf);
612                         break;
613                 }
614                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
615                 if (gsf->gf_numsrc >= 0x1ffffffU ||
616                     gsf->gf_numsrc > sysctl_mld_max_msf) {
617                         kfree(gsf);
618                         retv = -ENOBUFS;
619                         break;
620                 }
621                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
622                         kfree(gsf);
623                         retv = -EINVAL;
624                         break;
625                 }
626                 retv = ip6_mc_msfilter(sk, gsf);
627                 kfree(gsf);
628
629                 break;
630         }
631         case IPV6_ROUTER_ALERT:
632                 if (optlen < sizeof(int))
633                         goto e_inval;
634                 retv = ip6_ra_control(sk, val);
635                 break;
636         case IPV6_MTU_DISCOVER:
637                 if (optlen < sizeof(int))
638                         goto e_inval;
639                 if (val<0 || val>3)
640                         goto e_inval;
641                 np->pmtudisc = val;
642                 retv = 0;
643                 break;
644         case IPV6_MTU:
645                 if (optlen < sizeof(int))
646                         goto e_inval;
647                 if (val && val < IPV6_MIN_MTU)
648                         goto e_inval;
649                 np->frag_size = val;
650                 retv = 0;
651                 break;
652         case IPV6_RECVERR:
653                 if (optlen < sizeof(int))
654                         goto e_inval;
655                 np->recverr = valbool;
656                 if (!val)
657                         skb_queue_purge(&sk->sk_error_queue);
658                 retv = 0;
659                 break;
660         case IPV6_FLOWINFO_SEND:
661                 if (optlen < sizeof(int))
662                         goto e_inval;
663                 np->sndflow = valbool;
664                 retv = 0;
665                 break;
666         case IPV6_FLOWLABEL_MGR:
667                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
668                 break;
669         case IPV6_IPSEC_POLICY:
670         case IPV6_XFRM_POLICY:
671                 retv = -EPERM;
672                 if (!capable(CAP_NET_ADMIN))
673                         break;
674                 retv = xfrm_user_policy(sk, optname, optval, optlen);
675                 break;
676
677         case IPV6_ADDR_PREFERENCES:
678             {
679                 unsigned int pref = 0;
680                 unsigned int prefmask = ~0;
681
682                 if (optlen < sizeof(int))
683                         goto e_inval;
684
685                 retv = -EINVAL;
686
687                 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
688                 switch (val & (IPV6_PREFER_SRC_PUBLIC|
689                                IPV6_PREFER_SRC_TMP|
690                                IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
691                 case IPV6_PREFER_SRC_PUBLIC:
692                         pref |= IPV6_PREFER_SRC_PUBLIC;
693                         break;
694                 case IPV6_PREFER_SRC_TMP:
695                         pref |= IPV6_PREFER_SRC_TMP;
696                         break;
697                 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
698                         break;
699                 case 0:
700                         goto pref_skip_pubtmp;
701                 default:
702                         goto e_inval;
703                 }
704
705                 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
706                               IPV6_PREFER_SRC_TMP);
707 pref_skip_pubtmp:
708
709                 /* check HOME/COA conflicts */
710                 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
711                 case IPV6_PREFER_SRC_HOME:
712                         break;
713                 case IPV6_PREFER_SRC_COA:
714                         pref |= IPV6_PREFER_SRC_COA;
715                 case 0:
716                         goto pref_skip_coa;
717                 default:
718                         goto e_inval;
719                 }
720
721                 prefmask &= ~IPV6_PREFER_SRC_COA;
722 pref_skip_coa:
723
724                 /* check CGA/NONCGA conflicts */
725                 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
726                 case IPV6_PREFER_SRC_CGA:
727                 case IPV6_PREFER_SRC_NONCGA:
728                 case 0:
729                         break;
730                 default:
731                         goto e_inval;
732                 }
733
734                 np->srcprefs = (np->srcprefs & prefmask) | pref;
735                 retv = 0;
736
737                 break;
738             }
739         }
740
741         release_sock(sk);
742
743         return retv;
744
745 e_inval:
746         release_sock(sk);
747         return -EINVAL;
748 }
749
750 int ipv6_setsockopt(struct sock *sk, int level, int optname,
751                     char __user *optval, int optlen)
752 {
753         int err;
754
755         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
756                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
757
758         if (level != SOL_IPV6)
759                 return -ENOPROTOOPT;
760
761         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
762 #ifdef CONFIG_NETFILTER
763         /* we need to exclude all possible ENOPROTOOPTs except default case */
764         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
765                         optname != IPV6_XFRM_POLICY) {
766                 lock_sock(sk);
767                 err = nf_setsockopt(sk, PF_INET6, optname, optval,
768                                 optlen);
769                 release_sock(sk);
770         }
771 #endif
772         return err;
773 }
774
775 EXPORT_SYMBOL(ipv6_setsockopt);
776
777 #ifdef CONFIG_COMPAT
778 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
779                            char __user *optval, int optlen)
780 {
781         int err;
782
783         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
784                 if (udp_prot.compat_setsockopt != NULL)
785                         return udp_prot.compat_setsockopt(sk, level, optname,
786                                                           optval, optlen);
787                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
788         }
789
790         if (level != SOL_IPV6)
791                 return -ENOPROTOOPT;
792
793         if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
794                 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
795                         ipv6_setsockopt);
796
797         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
798 #ifdef CONFIG_NETFILTER
799         /* we need to exclude all possible ENOPROTOOPTs except default case */
800         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
801             optname != IPV6_XFRM_POLICY) {
802                 lock_sock(sk);
803                 err = compat_nf_setsockopt(sk, PF_INET6, optname,
804                                            optval, optlen);
805                 release_sock(sk);
806         }
807 #endif
808         return err;
809 }
810
811 EXPORT_SYMBOL(compat_ipv6_setsockopt);
812 #endif
813
814 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
815                                   int optname, char __user *optval, int len)
816 {
817         struct ipv6_opt_hdr *hdr;
818
819         if (!opt)
820                 return 0;
821
822         switch(optname) {
823         case IPV6_HOPOPTS:
824                 hdr = opt->hopopt;
825                 break;
826         case IPV6_RTHDRDSTOPTS:
827                 hdr = opt->dst0opt;
828                 break;
829         case IPV6_RTHDR:
830                 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
831                 break;
832         case IPV6_DSTOPTS:
833                 hdr = opt->dst1opt;
834                 break;
835         default:
836                 return -EINVAL; /* should not happen */
837         }
838
839         if (!hdr)
840                 return 0;
841
842         len = min_t(unsigned int, len, ipv6_optlen(hdr));
843         if (copy_to_user(optval, hdr, len))
844                 return -EFAULT;
845         return len;
846 }
847
848 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
849                     char __user *optval, int __user *optlen)
850 {
851         struct ipv6_pinfo *np = inet6_sk(sk);
852         int len;
853         int val;
854
855         if (ip6_mroute_opt(optname))
856                 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
857
858         if (get_user(len, optlen))
859                 return -EFAULT;
860         switch (optname) {
861         case IPV6_ADDRFORM:
862                 if (sk->sk_protocol != IPPROTO_UDP &&
863                     sk->sk_protocol != IPPROTO_UDPLITE &&
864                     sk->sk_protocol != IPPROTO_TCP)
865                         return -ENOPROTOOPT;
866                 if (sk->sk_state != TCP_ESTABLISHED)
867                         return -ENOTCONN;
868                 val = sk->sk_family;
869                 break;
870         case MCAST_MSFILTER:
871         {
872                 struct group_filter gsf;
873                 int err;
874
875                 if (len < GROUP_FILTER_SIZE(0))
876                         return -EINVAL;
877                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
878                         return -EFAULT;
879                 if (gsf.gf_group.ss_family != AF_INET6)
880                         return -EADDRNOTAVAIL;
881                 lock_sock(sk);
882                 err = ip6_mc_msfget(sk, &gsf,
883                         (struct group_filter __user *)optval, optlen);
884                 release_sock(sk);
885                 return err;
886         }
887
888         case IPV6_2292PKTOPTIONS:
889         {
890                 struct msghdr msg;
891                 struct sk_buff *skb;
892
893                 if (sk->sk_type != SOCK_STREAM)
894                         return -ENOPROTOOPT;
895
896                 msg.msg_control = optval;
897                 msg.msg_controllen = len;
898                 msg.msg_flags = 0;
899
900                 lock_sock(sk);
901                 skb = np->pktoptions;
902                 if (skb)
903                         atomic_inc(&skb->users);
904                 release_sock(sk);
905
906                 if (skb) {
907                         int err = datagram_recv_ctl(sk, &msg, skb);
908                         kfree_skb(skb);
909                         if (err)
910                                 return err;
911                 } else {
912                         if (np->rxopt.bits.rxinfo) {
913                                 struct in6_pktinfo src_info;
914                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif : sk->sk_bound_dev_if;
915                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
916                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
917                         }
918                         if (np->rxopt.bits.rxhlim) {
919                                 int hlim = np->mcast_hops;
920                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
921                         }
922                         if (np->rxopt.bits.rxoinfo) {
923                                 struct in6_pktinfo src_info;
924                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif : sk->sk_bound_dev_if;
925                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
926                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
927                         }
928                         if (np->rxopt.bits.rxohlim) {
929                                 int hlim = np->mcast_hops;
930                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
931                         }
932                 }
933                 len -= msg.msg_controllen;
934                 return put_user(len, optlen);
935         }
936         case IPV6_MTU:
937         {
938                 struct dst_entry *dst;
939                 val = 0;
940                 lock_sock(sk);
941                 dst = sk_dst_get(sk);
942                 if (dst) {
943                         val = dst_mtu(dst);
944                         dst_release(dst);
945                 }
946                 release_sock(sk);
947                 if (!val)
948                         return -ENOTCONN;
949                 break;
950         }
951
952         case IPV6_V6ONLY:
953                 val = np->ipv6only;
954                 break;
955
956         case IPV6_RECVPKTINFO:
957                 val = np->rxopt.bits.rxinfo;
958                 break;
959
960         case IPV6_2292PKTINFO:
961                 val = np->rxopt.bits.rxoinfo;
962                 break;
963
964         case IPV6_RECVHOPLIMIT:
965                 val = np->rxopt.bits.rxhlim;
966                 break;
967
968         case IPV6_2292HOPLIMIT:
969                 val = np->rxopt.bits.rxohlim;
970                 break;
971
972         case IPV6_RECVRTHDR:
973                 val = np->rxopt.bits.srcrt;
974                 break;
975
976         case IPV6_2292RTHDR:
977                 val = np->rxopt.bits.osrcrt;
978                 break;
979
980         case IPV6_HOPOPTS:
981         case IPV6_RTHDRDSTOPTS:
982         case IPV6_RTHDR:
983         case IPV6_DSTOPTS:
984         {
985
986                 lock_sock(sk);
987                 len = ipv6_getsockopt_sticky(sk, np->opt,
988                                              optname, optval, len);
989                 release_sock(sk);
990                 /* check if ipv6_getsockopt_sticky() returns err code */
991                 if (len < 0)
992                         return len;
993                 return put_user(len, optlen);
994         }
995
996         case IPV6_RECVHOPOPTS:
997                 val = np->rxopt.bits.hopopts;
998                 break;
999
1000         case IPV6_2292HOPOPTS:
1001                 val = np->rxopt.bits.ohopopts;
1002                 break;
1003
1004         case IPV6_RECVDSTOPTS:
1005                 val = np->rxopt.bits.dstopts;
1006                 break;
1007
1008         case IPV6_2292DSTOPTS:
1009                 val = np->rxopt.bits.odstopts;
1010                 break;
1011
1012         case IPV6_TCLASS:
1013                 val = np->tclass;
1014                 if (val < 0)
1015                         val = 0;
1016                 break;
1017
1018         case IPV6_RECVTCLASS:
1019                 val = np->rxopt.bits.rxtclass;
1020                 break;
1021
1022         case IPV6_FLOWINFO:
1023                 val = np->rxopt.bits.rxflow;
1024                 break;
1025
1026         case IPV6_UNICAST_HOPS:
1027         case IPV6_MULTICAST_HOPS:
1028         {
1029                 struct dst_entry *dst;
1030
1031                 if (optname == IPV6_UNICAST_HOPS)
1032                         val = np->hop_limit;
1033                 else
1034                         val = np->mcast_hops;
1035
1036                 dst = sk_dst_get(sk);
1037                 if (dst) {
1038                         if (val < 0)
1039                                 val = ip6_dst_hoplimit(dst);
1040                         dst_release(dst);
1041                 }
1042                 if (val < 0)
1043                         val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1044                 break;
1045         }
1046
1047         case IPV6_MULTICAST_LOOP:
1048                 val = np->mc_loop;
1049                 break;
1050
1051         case IPV6_MULTICAST_IF:
1052                 val = np->mcast_oif;
1053                 break;
1054
1055         case IPV6_MTU_DISCOVER:
1056                 val = np->pmtudisc;
1057                 break;
1058
1059         case IPV6_RECVERR:
1060                 val = np->recverr;
1061                 break;
1062
1063         case IPV6_FLOWINFO_SEND:
1064                 val = np->sndflow;
1065                 break;
1066
1067         case IPV6_ADDR_PREFERENCES:
1068                 val = 0;
1069
1070                 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1071                         val |= IPV6_PREFER_SRC_TMP;
1072                 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1073                         val |= IPV6_PREFER_SRC_PUBLIC;
1074                 else {
1075                         /* XXX: should we return system default? */
1076                         val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1077                 }
1078
1079                 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1080                         val |= IPV6_PREFER_SRC_COA;
1081                 else
1082                         val |= IPV6_PREFER_SRC_HOME;
1083                 break;
1084
1085         default:
1086                 return -ENOPROTOOPT;
1087         }
1088         len = min_t(unsigned int, sizeof(int), len);
1089         if(put_user(len, optlen))
1090                 return -EFAULT;
1091         if(copy_to_user(optval,&val,len))
1092                 return -EFAULT;
1093         return 0;
1094 }
1095
1096 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1097                     char __user *optval, int __user *optlen)
1098 {
1099         int err;
1100
1101         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1102                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
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 ENOPROTOOPTs except default case */
1110         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1111                 int len;
1112
1113                 if (get_user(len, optlen))
1114                         return -EFAULT;
1115
1116                 lock_sock(sk);
1117                 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1118                                 &len);
1119                 release_sock(sk);
1120                 if (err >= 0)
1121                         err = put_user(len, optlen);
1122         }
1123 #endif
1124         return err;
1125 }
1126
1127 EXPORT_SYMBOL(ipv6_getsockopt);
1128
1129 #ifdef CONFIG_COMPAT
1130 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1131                            char __user *optval, int __user *optlen)
1132 {
1133         int err;
1134
1135         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1136                 if (udp_prot.compat_getsockopt != NULL)
1137                         return udp_prot.compat_getsockopt(sk, level, optname,
1138                                                           optval, optlen);
1139                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1140         }
1141
1142         if (level != SOL_IPV6)
1143                 return -ENOPROTOOPT;
1144
1145         if (optname == MCAST_MSFILTER)
1146                 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1147                         ipv6_getsockopt);
1148
1149         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1150 #ifdef CONFIG_NETFILTER
1151         /* we need to exclude all possible ENOPROTOOPTs except default case */
1152         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1153                 int len;
1154
1155                 if (get_user(len, optlen))
1156                         return -EFAULT;
1157
1158                 lock_sock(sk);
1159                 err = compat_nf_getsockopt(sk, PF_INET6,
1160                                            optname, optval, &len);
1161                 release_sock(sk);
1162                 if (err >= 0)
1163                         err = put_user(len, optlen);
1164         }
1165 #endif
1166         return err;
1167 }
1168
1169 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1170 #endif
1171