netfilter: get rid of atomic ops in fast path
[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 #include <linux/slab.h>
40
41 #include <net/sock.h>
42 #include <net/snmp.h>
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/transp_v6.h>
47 #include <net/ip6_route.h>
48 #include <net/addrconf.h>
49 #include <net/inet_common.h>
50 #include <net/tcp.h>
51 #include <net/udp.h>
52 #include <net/udplite.h>
53 #include <net/xfrm.h>
54 #include <net/compat.h>
55
56 #include <asm/uaccess.h>
57
58 struct ip6_ra_chain *ip6_ra_chain;
59 DEFINE_RWLOCK(ip6_ra_lock);
60
61 int ip6_ra_control(struct sock *sk, int sel)
62 {
63         struct ip6_ra_chain *ra, *new_ra, **rap;
64
65         /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
66         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW)
67                 return -ENOPROTOOPT;
68
69         new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
70
71         write_lock_bh(&ip6_ra_lock);
72         for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
73                 if (ra->sk == sk) {
74                         if (sel>=0) {
75                                 write_unlock_bh(&ip6_ra_lock);
76                                 kfree(new_ra);
77                                 return -EADDRINUSE;
78                         }
79
80                         *rap = ra->next;
81                         write_unlock_bh(&ip6_ra_lock);
82
83                         sock_put(sk);
84                         kfree(ra);
85                         return 0;
86                 }
87         }
88         if (new_ra == NULL) {
89                 write_unlock_bh(&ip6_ra_lock);
90                 return -ENOBUFS;
91         }
92         new_ra->sk = sk;
93         new_ra->sel = sel;
94         new_ra->next = ra;
95         *rap = new_ra;
96         sock_hold(sk);
97         write_unlock_bh(&ip6_ra_lock);
98         return 0;
99 }
100
101 static
102 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
103                                            struct ipv6_txoptions *opt)
104 {
105         if (inet_sk(sk)->is_icsk) {
106                 if (opt &&
107                     !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
108                     inet_sk(sk)->inet_daddr != LOOPBACK4_IPV6) {
109                         struct inet_connection_sock *icsk = inet_csk(sk);
110                         icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
111                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
112                 }
113                 opt = xchg(&inet6_sk(sk)->opt, opt);
114         } else {
115                 spin_lock(&sk->sk_dst_lock);
116                 opt = xchg(&inet6_sk(sk)->opt, opt);
117                 spin_unlock(&sk->sk_dst_lock);
118         }
119         sk_dst_reset(sk);
120
121         return opt;
122 }
123
124 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
125                     char __user *optval, unsigned int optlen)
126 {
127         struct ipv6_pinfo *np = inet6_sk(sk);
128         struct net *net = sock_net(sk);
129         int val, valbool;
130         int retv = -ENOPROTOOPT;
131
132         if (optval == NULL)
133                 val=0;
134         else {
135                 if (optlen >= sizeof(int)) {
136                         if (get_user(val, (int __user *) optval))
137                                 return -EFAULT;
138                 } else
139                         val = 0;
140         }
141
142         valbool = (val!=0);
143
144         if (ip6_mroute_opt(optname))
145                 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
146
147         lock_sock(sk);
148
149         switch (optname) {
150
151         case IPV6_ADDRFORM:
152                 if (optlen < sizeof(int))
153                         goto e_inval;
154                 if (val == PF_INET) {
155                         struct ipv6_txoptions *opt;
156                         struct sk_buff *pktopt;
157
158                         if (sk->sk_type == SOCK_RAW)
159                                 break;
160
161                         if (sk->sk_protocol == IPPROTO_UDP ||
162                             sk->sk_protocol == IPPROTO_UDPLITE) {
163                                 struct udp_sock *up = udp_sk(sk);
164                                 if (up->pending == AF_INET6) {
165                                         retv = -EBUSY;
166                                         break;
167                                 }
168                         } else if (sk->sk_protocol != IPPROTO_TCP)
169                                 break;
170
171                         if (sk->sk_state != TCP_ESTABLISHED) {
172                                 retv = -ENOTCONN;
173                                 break;
174                         }
175
176                         if (ipv6_only_sock(sk) ||
177                             !ipv6_addr_v4mapped(&np->daddr)) {
178                                 retv = -EADDRNOTAVAIL;
179                                 break;
180                         }
181
182                         fl6_free_socklist(sk);
183                         ipv6_sock_mc_close(sk);
184
185                         /*
186                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
187                          * remove it from the refcnt debug socks count in the
188                          * original family...
189                          */
190                         sk_refcnt_debug_dec(sk);
191
192                         if (sk->sk_protocol == IPPROTO_TCP) {
193                                 struct inet_connection_sock *icsk = inet_csk(sk);
194                                 local_bh_disable();
195                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
196                                 sock_prot_inuse_add(net, &tcp_prot, 1);
197                                 local_bh_enable();
198                                 sk->sk_prot = &tcp_prot;
199                                 icsk->icsk_af_ops = &ipv4_specific;
200                                 sk->sk_socket->ops = &inet_stream_ops;
201                                 sk->sk_family = PF_INET;
202                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
203                         } else {
204                                 struct proto *prot = &udp_prot;
205
206                                 if (sk->sk_protocol == IPPROTO_UDPLITE)
207                                         prot = &udplite_prot;
208                                 local_bh_disable();
209                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
210                                 sock_prot_inuse_add(net, prot, 1);
211                                 local_bh_enable();
212                                 sk->sk_prot = prot;
213                                 sk->sk_socket->ops = &inet_dgram_ops;
214                                 sk->sk_family = PF_INET;
215                         }
216                         opt = xchg(&np->opt, NULL);
217                         if (opt)
218                                 sock_kfree_s(sk, opt, opt->tot_len);
219                         pktopt = xchg(&np->pktoptions, NULL);
220                         kfree_skb(pktopt);
221
222                         sk->sk_destruct = inet_sock_destruct;
223                         /*
224                          * ... and add it to the refcnt debug socks count
225                          * in the new family. -acme
226                          */
227                         sk_refcnt_debug_inc(sk);
228                         module_put(THIS_MODULE);
229                         retv = 0;
230                         break;
231                 }
232                 goto e_inval;
233
234         case IPV6_V6ONLY:
235                 if (optlen < sizeof(int) ||
236                     inet_sk(sk)->inet_num)
237                         goto e_inval;
238                 np->ipv6only = valbool;
239                 retv = 0;
240                 break;
241
242         case IPV6_RECVPKTINFO:
243                 if (optlen < sizeof(int))
244                         goto e_inval;
245                 np->rxopt.bits.rxinfo = valbool;
246                 retv = 0;
247                 break;
248
249         case IPV6_2292PKTINFO:
250                 if (optlen < sizeof(int))
251                         goto e_inval;
252                 np->rxopt.bits.rxoinfo = valbool;
253                 retv = 0;
254                 break;
255
256         case IPV6_RECVHOPLIMIT:
257                 if (optlen < sizeof(int))
258                         goto e_inval;
259                 np->rxopt.bits.rxhlim = valbool;
260                 retv = 0;
261                 break;
262
263         case IPV6_2292HOPLIMIT:
264                 if (optlen < sizeof(int))
265                         goto e_inval;
266                 np->rxopt.bits.rxohlim = valbool;
267                 retv = 0;
268                 break;
269
270         case IPV6_RECVRTHDR:
271                 if (optlen < sizeof(int))
272                         goto e_inval;
273                 np->rxopt.bits.srcrt = valbool;
274                 retv = 0;
275                 break;
276
277         case IPV6_2292RTHDR:
278                 if (optlen < sizeof(int))
279                         goto e_inval;
280                 np->rxopt.bits.osrcrt = valbool;
281                 retv = 0;
282                 break;
283
284         case IPV6_RECVHOPOPTS:
285                 if (optlen < sizeof(int))
286                         goto e_inval;
287                 np->rxopt.bits.hopopts = valbool;
288                 retv = 0;
289                 break;
290
291         case IPV6_2292HOPOPTS:
292                 if (optlen < sizeof(int))
293                         goto e_inval;
294                 np->rxopt.bits.ohopopts = valbool;
295                 retv = 0;
296                 break;
297
298         case IPV6_RECVDSTOPTS:
299                 if (optlen < sizeof(int))
300                         goto e_inval;
301                 np->rxopt.bits.dstopts = valbool;
302                 retv = 0;
303                 break;
304
305         case IPV6_2292DSTOPTS:
306                 if (optlen < sizeof(int))
307                         goto e_inval;
308                 np->rxopt.bits.odstopts = valbool;
309                 retv = 0;
310                 break;
311
312         case IPV6_TCLASS:
313                 if (optlen < sizeof(int))
314                         goto e_inval;
315                 if (val < -1 || val > 0xff)
316                         goto e_inval;
317                 /* RFC 3542, 6.5: default traffic class of 0x0 */
318                 if (val == -1)
319                         val = 0;
320                 np->tclass = val;
321                 retv = 0;
322                 break;
323
324         case IPV6_RECVTCLASS:
325                 if (optlen < sizeof(int))
326                         goto e_inval;
327                 np->rxopt.bits.rxtclass = valbool;
328                 retv = 0;
329                 break;
330
331         case IPV6_FLOWINFO:
332                 if (optlen < sizeof(int))
333                         goto e_inval;
334                 np->rxopt.bits.rxflow = valbool;
335                 retv = 0;
336                 break;
337
338         case IPV6_RECVPATHMTU:
339                 if (optlen < sizeof(int))
340                         goto e_inval;
341                 np->rxopt.bits.rxpmtu = valbool;
342                 retv = 0;
343                 break;
344
345         case IPV6_TRANSPARENT:
346                 if (!capable(CAP_NET_ADMIN)) {
347                         retv = -EPERM;
348                         break;
349                 }
350                 if (optlen < sizeof(int))
351                         goto e_inval;
352                 /* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
353                 inet_sk(sk)->transparent = valbool;
354                 retv = 0;
355                 break;
356
357         case IPV6_RECVORIGDSTADDR:
358                 if (optlen < sizeof(int))
359                         goto e_inval;
360                 np->rxopt.bits.rxorigdstaddr = valbool;
361                 retv = 0;
362                 break;
363
364         case IPV6_HOPOPTS:
365         case IPV6_RTHDRDSTOPTS:
366         case IPV6_RTHDR:
367         case IPV6_DSTOPTS:
368         {
369                 struct ipv6_txoptions *opt;
370
371                 /* remove any sticky options header with a zero option
372                  * length, per RFC3542.
373                  */
374                 if (optlen == 0)
375                         optval = NULL;
376                 else if (optval == NULL)
377                         goto e_inval;
378                 else if (optlen < sizeof(struct ipv6_opt_hdr) ||
379                          optlen & 0x7 || optlen > 8 * 255)
380                         goto e_inval;
381
382                 /* hop-by-hop / destination options are privileged option */
383                 retv = -EPERM;
384                 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
385                         break;
386
387                 opt = ipv6_renew_options(sk, np->opt, optname,
388                                          (struct ipv6_opt_hdr __user *)optval,
389                                          optlen);
390                 if (IS_ERR(opt)) {
391                         retv = PTR_ERR(opt);
392                         break;
393                 }
394
395                 /* routing header option needs extra check */
396                 retv = -EINVAL;
397                 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
398                         struct ipv6_rt_hdr *rthdr = opt->srcrt;
399                         switch (rthdr->type) {
400 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
401                         case IPV6_SRCRT_TYPE_2:
402                                 if (rthdr->hdrlen != 2 ||
403                                     rthdr->segments_left != 1)
404                                         goto sticky_done;
405
406                                 break;
407 #endif
408                         default:
409                                 goto sticky_done;
410                         }
411                 }
412
413                 retv = 0;
414                 opt = ipv6_update_options(sk, opt);
415 sticky_done:
416                 if (opt)
417                         sock_kfree_s(sk, opt, opt->tot_len);
418                 break;
419         }
420
421         case IPV6_PKTINFO:
422         {
423                 struct in6_pktinfo pkt;
424
425                 if (optlen == 0)
426                         goto e_inval;
427                 else if (optlen < sizeof(struct in6_pktinfo) || optval == NULL)
428                         goto e_inval;
429
430                 if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
431                                 retv = -EFAULT;
432                                 break;
433                 }
434                 if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
435                         goto e_inval;
436
437                 np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
438                 ipv6_addr_copy(&np->sticky_pktinfo.ipi6_addr, &pkt.ipi6_addr);
439                 retv = 0;
440                 break;
441         }
442
443         case IPV6_2292PKTOPTIONS:
444         {
445                 struct ipv6_txoptions *opt = NULL;
446                 struct msghdr msg;
447                 struct flowi6 fl6;
448                 int junk;
449
450                 memset(&fl6, 0, sizeof(fl6));
451                 fl6.flowi6_oif = sk->sk_bound_dev_if;
452                 fl6.flowi6_mark = sk->sk_mark;
453
454                 if (optlen == 0)
455                         goto update;
456
457                 /* 1K is probably excessive
458                  * 1K is surely not enough, 2K per standard header is 16K.
459                  */
460                 retv = -EINVAL;
461                 if (optlen > 64*1024)
462                         break;
463
464                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
465                 retv = -ENOBUFS;
466                 if (opt == NULL)
467                         break;
468
469                 memset(opt, 0, sizeof(*opt));
470                 opt->tot_len = sizeof(*opt) + optlen;
471                 retv = -EFAULT;
472                 if (copy_from_user(opt+1, optval, optlen))
473                         goto done;
474
475                 msg.msg_controllen = optlen;
476                 msg.msg_control = (void*)(opt+1);
477
478                 retv = datagram_send_ctl(net, &msg, &fl6, opt, &junk, &junk,
479                                          &junk);
480                 if (retv)
481                         goto done;
482 update:
483                 retv = 0;
484                 opt = ipv6_update_options(sk, opt);
485 done:
486                 if (opt)
487                         sock_kfree_s(sk, opt, opt->tot_len);
488                 break;
489         }
490         case IPV6_UNICAST_HOPS:
491                 if (optlen < sizeof(int))
492                         goto e_inval;
493                 if (val > 255 || val < -1)
494                         goto e_inval;
495                 np->hop_limit = val;
496                 retv = 0;
497                 break;
498
499         case IPV6_MULTICAST_HOPS:
500                 if (sk->sk_type == SOCK_STREAM)
501                         break;
502                 if (optlen < sizeof(int))
503                         goto e_inval;
504                 if (val > 255 || val < -1)
505                         goto e_inval;
506                 np->mcast_hops = val;
507                 retv = 0;
508                 break;
509
510         case IPV6_MULTICAST_LOOP:
511                 if (optlen < sizeof(int))
512                         goto e_inval;
513                 if (val != valbool)
514                         goto e_inval;
515                 np->mc_loop = valbool;
516                 retv = 0;
517                 break;
518
519         case IPV6_MULTICAST_IF:
520                 if (sk->sk_type == SOCK_STREAM)
521                         break;
522                 if (optlen < sizeof(int))
523                         goto e_inval;
524
525                 if (val) {
526                         struct net_device *dev;
527
528                         if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
529                                 goto e_inval;
530
531                         dev = dev_get_by_index(net, val);
532                         if (!dev) {
533                                 retv = -ENODEV;
534                                 break;
535                         }
536                         dev_put(dev);
537                 }
538                 np->mcast_oif = val;
539                 retv = 0;
540                 break;
541         case IPV6_ADD_MEMBERSHIP:
542         case IPV6_DROP_MEMBERSHIP:
543         {
544                 struct ipv6_mreq mreq;
545
546                 if (optlen < sizeof(struct ipv6_mreq))
547                         goto e_inval;
548
549                 retv = -EPROTO;
550                 if (inet_sk(sk)->is_icsk)
551                         break;
552
553                 retv = -EFAULT;
554                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
555                         break;
556
557                 if (optname == IPV6_ADD_MEMBERSHIP)
558                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
559                 else
560                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
561                 break;
562         }
563         case IPV6_JOIN_ANYCAST:
564         case IPV6_LEAVE_ANYCAST:
565         {
566                 struct ipv6_mreq mreq;
567
568                 if (optlen < sizeof(struct ipv6_mreq))
569                         goto e_inval;
570
571                 retv = -EFAULT;
572                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
573                         break;
574
575                 if (optname == IPV6_JOIN_ANYCAST)
576                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
577                 else
578                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
579                 break;
580         }
581         case MCAST_JOIN_GROUP:
582         case MCAST_LEAVE_GROUP:
583         {
584                 struct group_req greq;
585                 struct sockaddr_in6 *psin6;
586
587                 if (optlen < sizeof(struct group_req))
588                         goto e_inval;
589
590                 retv = -EFAULT;
591                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
592                         break;
593                 if (greq.gr_group.ss_family != AF_INET6) {
594                         retv = -EADDRNOTAVAIL;
595                         break;
596                 }
597                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
598                 if (optname == MCAST_JOIN_GROUP)
599                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
600                                 &psin6->sin6_addr);
601                 else
602                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
603                                 &psin6->sin6_addr);
604                 break;
605         }
606         case MCAST_JOIN_SOURCE_GROUP:
607         case MCAST_LEAVE_SOURCE_GROUP:
608         case MCAST_BLOCK_SOURCE:
609         case MCAST_UNBLOCK_SOURCE:
610         {
611                 struct group_source_req greqs;
612                 int omode, add;
613
614                 if (optlen < sizeof(struct group_source_req))
615                         goto e_inval;
616                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
617                         retv = -EFAULT;
618                         break;
619                 }
620                 if (greqs.gsr_group.ss_family != AF_INET6 ||
621                     greqs.gsr_source.ss_family != AF_INET6) {
622                         retv = -EADDRNOTAVAIL;
623                         break;
624                 }
625                 if (optname == MCAST_BLOCK_SOURCE) {
626                         omode = MCAST_EXCLUDE;
627                         add = 1;
628                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
629                         omode = MCAST_EXCLUDE;
630                         add = 0;
631                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
632                         struct sockaddr_in6 *psin6;
633
634                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
635                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
636                                 &psin6->sin6_addr);
637                         /* prior join w/ different source is ok */
638                         if (retv && retv != -EADDRINUSE)
639                                 break;
640                         omode = MCAST_INCLUDE;
641                         add = 1;
642                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
643                         omode = MCAST_INCLUDE;
644                         add = 0;
645                 }
646                 retv = ip6_mc_source(add, omode, sk, &greqs);
647                 break;
648         }
649         case MCAST_MSFILTER:
650         {
651                 extern int sysctl_mld_max_msf;
652                 struct group_filter *gsf;
653
654                 if (optlen < GROUP_FILTER_SIZE(0))
655                         goto e_inval;
656                 if (optlen > sysctl_optmem_max) {
657                         retv = -ENOBUFS;
658                         break;
659                 }
660                 gsf = kmalloc(optlen,GFP_KERNEL);
661                 if (!gsf) {
662                         retv = -ENOBUFS;
663                         break;
664                 }
665                 retv = -EFAULT;
666                 if (copy_from_user(gsf, optval, optlen)) {
667                         kfree(gsf);
668                         break;
669                 }
670                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
671                 if (gsf->gf_numsrc >= 0x1ffffffU ||
672                     gsf->gf_numsrc > sysctl_mld_max_msf) {
673                         kfree(gsf);
674                         retv = -ENOBUFS;
675                         break;
676                 }
677                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
678                         kfree(gsf);
679                         retv = -EINVAL;
680                         break;
681                 }
682                 retv = ip6_mc_msfilter(sk, gsf);
683                 kfree(gsf);
684
685                 break;
686         }
687         case IPV6_ROUTER_ALERT:
688                 if (optlen < sizeof(int))
689                         goto e_inval;
690                 retv = ip6_ra_control(sk, val);
691                 break;
692         case IPV6_MTU_DISCOVER:
693                 if (optlen < sizeof(int))
694                         goto e_inval;
695                 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
696                         goto e_inval;
697                 np->pmtudisc = val;
698                 retv = 0;
699                 break;
700         case IPV6_MTU:
701                 if (optlen < sizeof(int))
702                         goto e_inval;
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                 if (optlen < sizeof(int))
710                         goto e_inval;
711                 np->recverr = valbool;
712                 if (!val)
713                         skb_queue_purge(&sk->sk_error_queue);
714                 retv = 0;
715                 break;
716         case IPV6_FLOWINFO_SEND:
717                 if (optlen < sizeof(int))
718                         goto e_inval;
719                 np->sndflow = valbool;
720                 retv = 0;
721                 break;
722         case IPV6_FLOWLABEL_MGR:
723                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
724                 break;
725         case IPV6_IPSEC_POLICY:
726         case IPV6_XFRM_POLICY:
727                 retv = -EPERM;
728                 if (!capable(CAP_NET_ADMIN))
729                         break;
730                 retv = xfrm_user_policy(sk, optname, optval, optlen);
731                 break;
732
733         case IPV6_ADDR_PREFERENCES:
734             {
735                 unsigned int pref = 0;
736                 unsigned int prefmask = ~0;
737
738                 if (optlen < sizeof(int))
739                         goto e_inval;
740
741                 retv = -EINVAL;
742
743                 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
744                 switch (val & (IPV6_PREFER_SRC_PUBLIC|
745                                IPV6_PREFER_SRC_TMP|
746                                IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
747                 case IPV6_PREFER_SRC_PUBLIC:
748                         pref |= IPV6_PREFER_SRC_PUBLIC;
749                         break;
750                 case IPV6_PREFER_SRC_TMP:
751                         pref |= IPV6_PREFER_SRC_TMP;
752                         break;
753                 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
754                         break;
755                 case 0:
756                         goto pref_skip_pubtmp;
757                 default:
758                         goto e_inval;
759                 }
760
761                 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
762                               IPV6_PREFER_SRC_TMP);
763 pref_skip_pubtmp:
764
765                 /* check HOME/COA conflicts */
766                 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
767                 case IPV6_PREFER_SRC_HOME:
768                         break;
769                 case IPV6_PREFER_SRC_COA:
770                         pref |= IPV6_PREFER_SRC_COA;
771                 case 0:
772                         goto pref_skip_coa;
773                 default:
774                         goto e_inval;
775                 }
776
777                 prefmask &= ~IPV6_PREFER_SRC_COA;
778 pref_skip_coa:
779
780                 /* check CGA/NONCGA conflicts */
781                 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
782                 case IPV6_PREFER_SRC_CGA:
783                 case IPV6_PREFER_SRC_NONCGA:
784                 case 0:
785                         break;
786                 default:
787                         goto e_inval;
788                 }
789
790                 np->srcprefs = (np->srcprefs & prefmask) | pref;
791                 retv = 0;
792
793                 break;
794             }
795         case IPV6_MINHOPCOUNT:
796                 if (optlen < sizeof(int))
797                         goto e_inval;
798                 if (val < 0 || val > 255)
799                         goto e_inval;
800                 np->min_hopcount = val;
801                 break;
802         case IPV6_DONTFRAG:
803                 np->dontfrag = valbool;
804                 retv = 0;
805                 break;
806         }
807
808         release_sock(sk);
809
810         return retv;
811
812 e_inval:
813         release_sock(sk);
814         return -EINVAL;
815 }
816
817 int ipv6_setsockopt(struct sock *sk, int level, int optname,
818                     char __user *optval, unsigned int optlen)
819 {
820         int err;
821
822         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
823                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
824
825         if (level != SOL_IPV6)
826                 return -ENOPROTOOPT;
827
828         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
829 #ifdef CONFIG_NETFILTER
830         /* we need to exclude all possible ENOPROTOOPTs except default case */
831         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
832                         optname != IPV6_XFRM_POLICY) {
833                 lock_sock(sk);
834                 err = nf_setsockopt(sk, PF_INET6, optname, optval,
835                                 optlen);
836                 release_sock(sk);
837         }
838 #endif
839         return err;
840 }
841
842 EXPORT_SYMBOL(ipv6_setsockopt);
843
844 #ifdef CONFIG_COMPAT
845 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
846                            char __user *optval, unsigned int optlen)
847 {
848         int err;
849
850         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
851                 if (udp_prot.compat_setsockopt != NULL)
852                         return udp_prot.compat_setsockopt(sk, level, optname,
853                                                           optval, optlen);
854                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
855         }
856
857         if (level != SOL_IPV6)
858                 return -ENOPROTOOPT;
859
860         if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
861                 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
862                         ipv6_setsockopt);
863
864         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
865 #ifdef CONFIG_NETFILTER
866         /* we need to exclude all possible ENOPROTOOPTs except default case */
867         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
868             optname != IPV6_XFRM_POLICY) {
869                 lock_sock(sk);
870                 err = compat_nf_setsockopt(sk, PF_INET6, optname,
871                                            optval, optlen);
872                 release_sock(sk);
873         }
874 #endif
875         return err;
876 }
877
878 EXPORT_SYMBOL(compat_ipv6_setsockopt);
879 #endif
880
881 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
882                                   int optname, char __user *optval, int len)
883 {
884         struct ipv6_opt_hdr *hdr;
885
886         if (!opt)
887                 return 0;
888
889         switch(optname) {
890         case IPV6_HOPOPTS:
891                 hdr = opt->hopopt;
892                 break;
893         case IPV6_RTHDRDSTOPTS:
894                 hdr = opt->dst0opt;
895                 break;
896         case IPV6_RTHDR:
897                 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
898                 break;
899         case IPV6_DSTOPTS:
900                 hdr = opt->dst1opt;
901                 break;
902         default:
903                 return -EINVAL; /* should not happen */
904         }
905
906         if (!hdr)
907                 return 0;
908
909         len = min_t(unsigned int, len, ipv6_optlen(hdr));
910         if (copy_to_user(optval, hdr, len))
911                 return -EFAULT;
912         return len;
913 }
914
915 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
916                     char __user *optval, int __user *optlen)
917 {
918         struct ipv6_pinfo *np = inet6_sk(sk);
919         int len;
920         int val;
921
922         if (ip6_mroute_opt(optname))
923                 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
924
925         if (get_user(len, optlen))
926                 return -EFAULT;
927         switch (optname) {
928         case IPV6_ADDRFORM:
929                 if (sk->sk_protocol != IPPROTO_UDP &&
930                     sk->sk_protocol != IPPROTO_UDPLITE &&
931                     sk->sk_protocol != IPPROTO_TCP)
932                         return -ENOPROTOOPT;
933                 if (sk->sk_state != TCP_ESTABLISHED)
934                         return -ENOTCONN;
935                 val = sk->sk_family;
936                 break;
937         case MCAST_MSFILTER:
938         {
939                 struct group_filter gsf;
940                 int err;
941
942                 if (len < GROUP_FILTER_SIZE(0))
943                         return -EINVAL;
944                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
945                         return -EFAULT;
946                 if (gsf.gf_group.ss_family != AF_INET6)
947                         return -EADDRNOTAVAIL;
948                 lock_sock(sk);
949                 err = ip6_mc_msfget(sk, &gsf,
950                         (struct group_filter __user *)optval, optlen);
951                 release_sock(sk);
952                 return err;
953         }
954
955         case IPV6_2292PKTOPTIONS:
956         {
957                 struct msghdr msg;
958                 struct sk_buff *skb;
959
960                 if (sk->sk_type != SOCK_STREAM)
961                         return -ENOPROTOOPT;
962
963                 msg.msg_control = optval;
964                 msg.msg_controllen = len;
965                 msg.msg_flags = 0;
966
967                 lock_sock(sk);
968                 skb = np->pktoptions;
969                 if (skb)
970                         atomic_inc(&skb->users);
971                 release_sock(sk);
972
973                 if (skb) {
974                         int err = datagram_recv_ctl(sk, &msg, skb);
975                         kfree_skb(skb);
976                         if (err)
977                                 return err;
978                 } else {
979                         if (np->rxopt.bits.rxinfo) {
980                                 struct in6_pktinfo src_info;
981                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
982                                         np->sticky_pktinfo.ipi6_ifindex;
983                                 np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
984                                         ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
985                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
986                         }
987                         if (np->rxopt.bits.rxhlim) {
988                                 int hlim = np->mcast_hops;
989                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
990                         }
991                         if (np->rxopt.bits.rxoinfo) {
992                                 struct in6_pktinfo src_info;
993                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
994                                         np->sticky_pktinfo.ipi6_ifindex;
995                                 np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
996                                         ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
997                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
998                         }
999                         if (np->rxopt.bits.rxohlim) {
1000                                 int hlim = np->mcast_hops;
1001                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1002                         }
1003                 }
1004                 len -= msg.msg_controllen;
1005                 return put_user(len, optlen);
1006         }
1007         case IPV6_MTU:
1008         {
1009                 struct dst_entry *dst;
1010
1011                 val = 0;
1012                 rcu_read_lock();
1013                 dst = __sk_dst_get(sk);
1014                 if (dst)
1015                         val = dst_mtu(dst);
1016                 rcu_read_unlock();
1017                 if (!val)
1018                         return -ENOTCONN;
1019                 break;
1020         }
1021
1022         case IPV6_V6ONLY:
1023                 val = np->ipv6only;
1024                 break;
1025
1026         case IPV6_RECVPKTINFO:
1027                 val = np->rxopt.bits.rxinfo;
1028                 break;
1029
1030         case IPV6_2292PKTINFO:
1031                 val = np->rxopt.bits.rxoinfo;
1032                 break;
1033
1034         case IPV6_RECVHOPLIMIT:
1035                 val = np->rxopt.bits.rxhlim;
1036                 break;
1037
1038         case IPV6_2292HOPLIMIT:
1039                 val = np->rxopt.bits.rxohlim;
1040                 break;
1041
1042         case IPV6_RECVRTHDR:
1043                 val = np->rxopt.bits.srcrt;
1044                 break;
1045
1046         case IPV6_2292RTHDR:
1047                 val = np->rxopt.bits.osrcrt;
1048                 break;
1049
1050         case IPV6_HOPOPTS:
1051         case IPV6_RTHDRDSTOPTS:
1052         case IPV6_RTHDR:
1053         case IPV6_DSTOPTS:
1054         {
1055
1056                 lock_sock(sk);
1057                 len = ipv6_getsockopt_sticky(sk, np->opt,
1058                                              optname, optval, len);
1059                 release_sock(sk);
1060                 /* check if ipv6_getsockopt_sticky() returns err code */
1061                 if (len < 0)
1062                         return len;
1063                 return put_user(len, optlen);
1064         }
1065
1066         case IPV6_RECVHOPOPTS:
1067                 val = np->rxopt.bits.hopopts;
1068                 break;
1069
1070         case IPV6_2292HOPOPTS:
1071                 val = np->rxopt.bits.ohopopts;
1072                 break;
1073
1074         case IPV6_RECVDSTOPTS:
1075                 val = np->rxopt.bits.dstopts;
1076                 break;
1077
1078         case IPV6_2292DSTOPTS:
1079                 val = np->rxopt.bits.odstopts;
1080                 break;
1081
1082         case IPV6_TCLASS:
1083                 val = np->tclass;
1084                 break;
1085
1086         case IPV6_RECVTCLASS:
1087                 val = np->rxopt.bits.rxtclass;
1088                 break;
1089
1090         case IPV6_FLOWINFO:
1091                 val = np->rxopt.bits.rxflow;
1092                 break;
1093
1094         case IPV6_RECVPATHMTU:
1095                 val = np->rxopt.bits.rxpmtu;
1096                 break;
1097
1098         case IPV6_PATHMTU:
1099         {
1100                 struct dst_entry *dst;
1101                 struct ip6_mtuinfo mtuinfo;
1102
1103                 if (len < sizeof(mtuinfo))
1104                         return -EINVAL;
1105
1106                 len = sizeof(mtuinfo);
1107                 memset(&mtuinfo, 0, sizeof(mtuinfo));
1108
1109                 rcu_read_lock();
1110                 dst = __sk_dst_get(sk);
1111                 if (dst)
1112                         mtuinfo.ip6m_mtu = dst_mtu(dst);
1113                 rcu_read_unlock();
1114                 if (!mtuinfo.ip6m_mtu)
1115                         return -ENOTCONN;
1116
1117                 if (put_user(len, optlen))
1118                         return -EFAULT;
1119                 if (copy_to_user(optval, &mtuinfo, len))
1120                         return -EFAULT;
1121
1122                 return 0;
1123                 break;
1124         }
1125
1126         case IPV6_TRANSPARENT:
1127                 val = inet_sk(sk)->transparent;
1128                 break;
1129
1130         case IPV6_RECVORIGDSTADDR:
1131                 val = np->rxopt.bits.rxorigdstaddr;
1132                 break;
1133
1134         case IPV6_UNICAST_HOPS:
1135         case IPV6_MULTICAST_HOPS:
1136         {
1137                 struct dst_entry *dst;
1138
1139                 if (optname == IPV6_UNICAST_HOPS)
1140                         val = np->hop_limit;
1141                 else
1142                         val = np->mcast_hops;
1143
1144                 if (val < 0) {
1145                         rcu_read_lock();
1146                         dst = __sk_dst_get(sk);
1147                         if (dst)
1148                                 val = ip6_dst_hoplimit(dst);
1149                         rcu_read_unlock();
1150                 }
1151
1152                 if (val < 0)
1153                         val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1154                 break;
1155         }
1156
1157         case IPV6_MULTICAST_LOOP:
1158                 val = np->mc_loop;
1159                 break;
1160
1161         case IPV6_MULTICAST_IF:
1162                 val = np->mcast_oif;
1163                 break;
1164
1165         case IPV6_MTU_DISCOVER:
1166                 val = np->pmtudisc;
1167                 break;
1168
1169         case IPV6_RECVERR:
1170                 val = np->recverr;
1171                 break;
1172
1173         case IPV6_FLOWINFO_SEND:
1174                 val = np->sndflow;
1175                 break;
1176
1177         case IPV6_ADDR_PREFERENCES:
1178                 val = 0;
1179
1180                 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1181                         val |= IPV6_PREFER_SRC_TMP;
1182                 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1183                         val |= IPV6_PREFER_SRC_PUBLIC;
1184                 else {
1185                         /* XXX: should we return system default? */
1186                         val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1187                 }
1188
1189                 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1190                         val |= IPV6_PREFER_SRC_COA;
1191                 else
1192                         val |= IPV6_PREFER_SRC_HOME;
1193                 break;
1194
1195         case IPV6_MINHOPCOUNT:
1196                 val = np->min_hopcount;
1197                 break;
1198
1199         case IPV6_DONTFRAG:
1200                 val = np->dontfrag;
1201                 break;
1202
1203         default:
1204                 return -ENOPROTOOPT;
1205         }
1206         len = min_t(unsigned int, sizeof(int), len);
1207         if(put_user(len, optlen))
1208                 return -EFAULT;
1209         if(copy_to_user(optval,&val,len))
1210                 return -EFAULT;
1211         return 0;
1212 }
1213
1214 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1215                     char __user *optval, int __user *optlen)
1216 {
1217         int err;
1218
1219         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1220                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1221
1222         if(level != SOL_IPV6)
1223                 return -ENOPROTOOPT;
1224
1225         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1226 #ifdef CONFIG_NETFILTER
1227         /* we need to exclude all possible ENOPROTOOPTs except default case */
1228         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1229                 int len;
1230
1231                 if (get_user(len, optlen))
1232                         return -EFAULT;
1233
1234                 lock_sock(sk);
1235                 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1236                                 &len);
1237                 release_sock(sk);
1238                 if (err >= 0)
1239                         err = put_user(len, optlen);
1240         }
1241 #endif
1242         return err;
1243 }
1244
1245 EXPORT_SYMBOL(ipv6_getsockopt);
1246
1247 #ifdef CONFIG_COMPAT
1248 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1249                            char __user *optval, int __user *optlen)
1250 {
1251         int err;
1252
1253         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1254                 if (udp_prot.compat_getsockopt != NULL)
1255                         return udp_prot.compat_getsockopt(sk, level, optname,
1256                                                           optval, optlen);
1257                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1258         }
1259
1260         if (level != SOL_IPV6)
1261                 return -ENOPROTOOPT;
1262
1263         if (optname == MCAST_MSFILTER)
1264                 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1265                         ipv6_getsockopt);
1266
1267         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1268 #ifdef CONFIG_NETFILTER
1269         /* we need to exclude all possible ENOPROTOOPTs except default case */
1270         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1271                 int len;
1272
1273                 if (get_user(len, optlen))
1274                         return -EFAULT;
1275
1276                 lock_sock(sk);
1277                 err = compat_nf_getsockopt(sk, PF_INET6,
1278                                            optname, optval, &len);
1279                 release_sock(sk);
1280                 if (err >= 0)
1281                         err = put_user(len, optlen);
1282         }
1283 #endif
1284         return err;
1285 }
1286
1287 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1288 #endif
1289