?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
driver core: Partially revert "driver core: correct device's shutdown order"
[linux-2.6/btrfs-unstable.git] / net / ipv6 / ipv6_sockglue.c
blob4d780c7f013060732dda2db760d7ba0474c812e3
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 #include <net/seg6.h>
56
57 #include <linux/uaccess.h>
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)->inet_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) {
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 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 }
114 opt = xchg((__force struct ipv6_txoptions **)&inet6_sk(sk)->opt,
115 opt);
116 sk_dst_reset(sk);
117
118 return opt;
119 }
120
121 static bool setsockopt_needs_rtnl(int optname)
122 {
123 switch (optname) {
124 case IPV6_ADDRFORM:
125 case IPV6_ADD_MEMBERSHIP:
126 case IPV6_DROP_MEMBERSHIP:
127 case IPV6_JOIN_ANYCAST:
128 case IPV6_LEAVE_ANYCAST:
129 case MCAST_JOIN_GROUP:
130 case MCAST_LEAVE_GROUP:
131 case MCAST_JOIN_SOURCE_GROUP:
132 case MCAST_LEAVE_SOURCE_GROUP:
133 case MCAST_BLOCK_SOURCE:
134 case MCAST_UNBLOCK_SOURCE:
135 case MCAST_MSFILTER:
136 return true;
137 }
138 return false;
139 }
140
141 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
142 char __user *optval, unsigned int optlen)
143 {
144 struct ipv6_pinfo *np = inet6_sk(sk);
145 struct net *net = sock_net(sk);
146 int val, valbool;
147 int retv = -ENOPROTOOPT;
148 bool needs_rtnl = setsockopt_needs_rtnl(optname);
149
150 if (!optval)
151 val = 0;
152 else {
153 if (optlen >= sizeof(int)) {
154 if (get_user(val, (int __user *) optval))
155 return -EFAULT;
156 } else
157 val = 0;
158 }
159
160 valbool = (val != 0);
161
162 if (ip6_mroute_opt(optname))
163 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
164
165 if (needs_rtnl)
166 rtnl_lock();
167 lock_sock(sk);
168
169 switch (optname) {
170
171 case IPV6_ADDRFORM:
172 if (optlen < sizeof(int))
173 goto e_inval;
174 if (val == PF_INET) {
175 struct ipv6_txoptions *opt;
176 struct sk_buff *pktopt;
177
178 if (sk->sk_type == SOCK_RAW)
179 break;
180
181 if (sk->sk_protocol == IPPROTO_UDP ||
182 sk->sk_protocol == IPPROTO_UDPLITE) {
183 struct udp_sock *up = udp_sk(sk);
184 if (up->pending == AF_INET6) {
185 retv = -EBUSY;
186 break;
187 }
188 } else if (sk->sk_protocol != IPPROTO_TCP)
189 break;
190
191 if (sk->sk_state != TCP_ESTABLISHED) {
192 retv = -ENOTCONN;
193 break;
194 }
195
196 if (ipv6_only_sock(sk) ||
197 !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
198 retv = -EADDRNOTAVAIL;
199 break;
200 }
201
202 fl6_free_socklist(sk);
203 __ipv6_sock_mc_close(sk);
204
205 /*
206 * Sock is moving from IPv6 to IPv4 (sk_prot), so
207 * remove it from the refcnt debug socks count in the
208 * original family...
209 */
210 sk_refcnt_debug_dec(sk);
211
212 if (sk->sk_protocol == IPPROTO_TCP) {
213 struct inet_connection_sock *icsk = inet_csk(sk);
214 local_bh_disable();
215 sock_prot_inuse_add(net, sk->sk_prot, -1);
216 sock_prot_inuse_add(net, &tcp_prot, 1);
217 local_bh_enable();
218 sk->sk_prot = &tcp_prot;
219 icsk->icsk_af_ops = &ipv4_specific;
220 sk->sk_socket->ops = &inet_stream_ops;
221 sk->sk_family = PF_INET;
222 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
223 } else {
224 struct proto *prot = &udp_prot;
225
226 if (sk->sk_protocol == IPPROTO_UDPLITE)
227 prot = &udplite_prot;
228 local_bh_disable();
229 sock_prot_inuse_add(net, sk->sk_prot, -1);
230 sock_prot_inuse_add(net, prot, 1);
231 local_bh_enable();
232 sk->sk_prot = prot;
233 sk->sk_socket->ops = &inet_dgram_ops;
234 sk->sk_family = PF_INET;
235 }
236 opt = xchg((__force struct ipv6_txoptions **)&np->opt,
237 NULL);
238 if (opt) {
239 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
240 txopt_put(opt);
241 }
242 pktopt = xchg(&np->pktoptions, NULL);
243 kfree_skb(pktopt);
244
245 /*
246 * ... and add it to the refcnt debug socks count
247 * in the new family. -acme
248 */
249 sk_refcnt_debug_inc(sk);
250 module_put(THIS_MODULE);
251 retv = 0;
252 break;
253 }
254 goto e_inval;
255
256 case IPV6_V6ONLY:
257 if (optlen < sizeof(int) ||
258 inet_sk(sk)->inet_num)
259 goto e_inval;
260 sk->sk_ipv6only = valbool;
261 retv = 0;
262 break;
263
264 case IPV6_RECVPKTINFO:
265 if (optlen < sizeof(int))
266 goto e_inval;
267 np->rxopt.bits.rxinfo = valbool;
268 retv = 0;
269 break;
270
271 case IPV6_2292PKTINFO:
272 if (optlen < sizeof(int))
273 goto e_inval;
274 np->rxopt.bits.rxoinfo = valbool;
275 retv = 0;
276 break;
277
278 case IPV6_RECVHOPLIMIT:
279 if (optlen < sizeof(int))
280 goto e_inval;
281 np->rxopt.bits.rxhlim = valbool;
282 retv = 0;
283 break;
284
285 case IPV6_2292HOPLIMIT:
286 if (optlen < sizeof(int))
287 goto e_inval;
288 np->rxopt.bits.rxohlim = valbool;
289 retv = 0;
290 break;
291
292 case IPV6_RECVRTHDR:
293 if (optlen < sizeof(int))
294 goto e_inval;
295 np->rxopt.bits.srcrt = valbool;
296 retv = 0;
297 break;
298
299 case IPV6_2292RTHDR:
300 if (optlen < sizeof(int))
301 goto e_inval;
302 np->rxopt.bits.osrcrt = valbool;
303 retv = 0;
304 break;
305
306 case IPV6_RECVHOPOPTS:
307 if (optlen < sizeof(int))
308 goto e_inval;
309 np->rxopt.bits.hopopts = valbool;
310 retv = 0;
311 break;
312
313 case IPV6_2292HOPOPTS:
314 if (optlen < sizeof(int))
315 goto e_inval;
316 np->rxopt.bits.ohopopts = valbool;
317 retv = 0;
318 break;
319
320 case IPV6_RECVDSTOPTS:
321 if (optlen < sizeof(int))
322 goto e_inval;
323 np->rxopt.bits.dstopts = valbool;
324 retv = 0;
325 break;
326
327 case IPV6_2292DSTOPTS:
328 if (optlen < sizeof(int))
329 goto e_inval;
330 np->rxopt.bits.odstopts = valbool;
331 retv = 0;
332 break;
333
334 case IPV6_TCLASS:
335 if (optlen < sizeof(int))
336 goto e_inval;
337 if (val < -1 || val > 0xff)
338 goto e_inval;
339 /* RFC 3542, 6.5: default traffic class of 0x0 */
340 if (val == -1)
341 val = 0;
342 np->tclass = val;
343 retv = 0;
344 break;
345
346 case IPV6_RECVTCLASS:
347 if (optlen < sizeof(int))
348 goto e_inval;
349 np->rxopt.bits.rxtclass = valbool;
350 retv = 0;
351 break;
352
353 case IPV6_FLOWINFO:
354 if (optlen < sizeof(int))
355 goto e_inval;
356 np->rxopt.bits.rxflow = valbool;
357 retv = 0;
358 break;
359
360 case IPV6_RECVPATHMTU:
361 if (optlen < sizeof(int))
362 goto e_inval;
363 np->rxopt.bits.rxpmtu = valbool;
364 retv = 0;
365 break;
366
367 case IPV6_TRANSPARENT:
368 if (valbool && !ns_capable(net->user_ns, CAP_NET_ADMIN) &&
369 !ns_capable(net->user_ns, CAP_NET_RAW)) {
370 retv = -EPERM;
371 break;
372 }
373 if (optlen < sizeof(int))
374 goto e_inval;
375 /* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
376 inet_sk(sk)->transparent = valbool;
377 retv = 0;
378 break;
379
380 case IPV6_FREEBIND:
381 if (optlen < sizeof(int))
382 goto e_inval;
383 /* we also don't have a separate freebind bit for IPV6 */
384 inet_sk(sk)->freebind = valbool;
385 retv = 0;
386 break;
387
388 case IPV6_RECVORIGDSTADDR:
389 if (optlen < sizeof(int))
390 goto e_inval;
391 np->rxopt.bits.rxorigdstaddr = valbool;
392 retv = 0;
393 break;
394
395 case IPV6_HOPOPTS:
396 case IPV6_RTHDRDSTOPTS:
397 case IPV6_RTHDR:
398 case IPV6_DSTOPTS:
399 {
400 struct ipv6_txoptions *opt;
401
402 /* remove any sticky options header with a zero option
403 * length, per RFC3542.
404 */
405 if (optlen == 0)
406 optval = NULL;
407 else if (!optval)
408 goto e_inval;
409 else if (optlen < sizeof(struct ipv6_opt_hdr) ||
410 optlen & 0x7 || optlen > 8 * 255)
411 goto e_inval;
412
413 /* hop-by-hop / destination options are privileged option */
414 retv = -EPERM;
415 if (optname != IPV6_RTHDR && !ns_capable(net->user_ns, CAP_NET_RAW))
416 break;
417
418 opt = rcu_dereference_protected(np->opt,
419 lockdep_sock_is_held(sk));
420 opt = ipv6_renew_options(sk, opt, optname,
421 (struct ipv6_opt_hdr __user *)optval,
422 optlen);
423 if (IS_ERR(opt)) {
424 retv = PTR_ERR(opt);
425 break;
426 }
427
428 /* routing header option needs extra check */
429 retv = -EINVAL;
430 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
431 struct ipv6_rt_hdr *rthdr = opt->srcrt;
432 switch (rthdr->type) {
433 #if IS_ENABLED(CONFIG_IPV6_MIP6)
434 case IPV6_SRCRT_TYPE_2:
435 if (rthdr->hdrlen != 2 ||
436 rthdr->segments_left != 1)
437 goto sticky_done;
438
439 break;
440 #endif
441 case IPV6_SRCRT_TYPE_4:
442 {
443 struct ipv6_sr_hdr *srh = (struct ipv6_sr_hdr *)
444 opt->srcrt;
445
446 if (!seg6_validate_srh(srh, optlen))
447 goto sticky_done;
448 break;
449 }
450 default:
451 goto sticky_done;
452 }
453 }
454
455 retv = 0;
456 opt = ipv6_update_options(sk, opt);
457 sticky_done:
458 if (opt) {
459 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
460 txopt_put(opt);
461 }
462 break;
463 }
464
465 case IPV6_PKTINFO:
466 {
467 struct in6_pktinfo pkt;
468
469 if (optlen == 0)
470 goto e_inval;
471 else if (optlen < sizeof(struct in6_pktinfo) || !optval)
472 goto e_inval;
473
474 if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
475 retv = -EFAULT;
476 break;
477 }
478 if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
479 goto e_inval;
480
481 np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
482 np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
483 retv = 0;
484 break;
485 }
486
487 case IPV6_2292PKTOPTIONS:
488 {
489 struct ipv6_txoptions *opt = NULL;
490 struct msghdr msg;
491 struct flowi6 fl6;
492 struct sockcm_cookie sockc_junk;
493 struct ipcm6_cookie ipc6;
494
495 memset(&fl6, 0, sizeof(fl6));
496 fl6.flowi6_oif = sk->sk_bound_dev_if;
497 fl6.flowi6_mark = sk->sk_mark;
498
499 if (optlen == 0)
500 goto update;
501
502 /* 1K is probably excessive
503 * 1K is surely not enough, 2K per standard header is 16K.
504 */
505 retv = -EINVAL;
506 if (optlen > 64*1024)
507 break;
508
509 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
510 retv = -ENOBUFS;
511 if (!opt)
512 break;
513
514 memset(opt, 0, sizeof(*opt));
515 refcount_set(&opt->refcnt, 1);
516 opt->tot_len = sizeof(*opt) + optlen;
517 retv = -EFAULT;
518 if (copy_from_user(opt+1, optval, optlen))
519 goto done;
520
521 msg.msg_controllen = optlen;
522 msg.msg_control = (void *)(opt+1);
523 ipc6.opt = opt;
524
525 retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6, &sockc_junk);
526 if (retv)
527 goto done;
528 update:
529 retv = 0;
530 opt = ipv6_update_options(sk, opt);
531 done:
532 if (opt) {
533 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
534 txopt_put(opt);
535 }
536 break;
537 }
538 case IPV6_UNICAST_HOPS:
539 if (optlen < sizeof(int))
540 goto e_inval;
541 if (val > 255 || val < -1)
542 goto e_inval;
543 np->hop_limit = val;
544 retv = 0;
545 break;
546
547 case IPV6_MULTICAST_HOPS:
548 if (sk->sk_type == SOCK_STREAM)
549 break;
550 if (optlen < sizeof(int))
551 goto e_inval;
552 if (val > 255 || val < -1)
553 goto e_inval;
554 np->mcast_hops = (val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
555 retv = 0;
556 break;
557
558 case IPV6_MULTICAST_LOOP:
559 if (optlen < sizeof(int))
560 goto e_inval;
561 if (val != valbool)
562 goto e_inval;
563 np->mc_loop = valbool;
564 retv = 0;
565 break;
566
567 case IPV6_UNICAST_IF:
568 {
569 struct net_device *dev = NULL;
570 int ifindex;
571
572 if (optlen != sizeof(int))
573 goto e_inval;
574
575 ifindex = (__force int)ntohl((__force __be32)val);
576 if (ifindex == 0) {
577 np->ucast_oif = 0;
578 retv = 0;
579 break;
580 }
581
582 dev = dev_get_by_index(net, ifindex);
583 retv = -EADDRNOTAVAIL;
584 if (!dev)
585 break;
586 dev_put(dev);
587
588 retv = -EINVAL;
589 if (sk->sk_bound_dev_if)
590 break;
591
592 np->ucast_oif = ifindex;
593 retv = 0;
594 break;
595 }
596
597 case IPV6_MULTICAST_IF:
598 if (sk->sk_type == SOCK_STREAM)
599 break;
600 if (optlen < sizeof(int))
601 goto e_inval;
602
603 if (val) {
604 struct net_device *dev;
605 int midx;
606
607 rcu_read_lock();
608
609 dev = dev_get_by_index_rcu(net, val);
610 if (!dev) {
611 rcu_read_unlock();
612 retv = -ENODEV;
613 break;
614 }
615 midx = l3mdev_master_ifindex_rcu(dev);
616
617 rcu_read_unlock();
618
619 if (sk->sk_bound_dev_if &&
620 sk->sk_bound_dev_if != val &&
621 (!midx || midx != sk->sk_bound_dev_if))
622 goto e_inval;
623 }
624 np->mcast_oif = val;
625 retv = 0;
626 break;
627 case IPV6_ADD_MEMBERSHIP:
628 case IPV6_DROP_MEMBERSHIP:
629 {
630 struct ipv6_mreq mreq;
631
632 if (optlen < sizeof(struct ipv6_mreq))
633 goto e_inval;
634
635 retv = -EPROTO;
636 if (inet_sk(sk)->is_icsk)
637 break;
638
639 retv = -EFAULT;
640 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
641 break;
642
643 if (optname == IPV6_ADD_MEMBERSHIP)
644 retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
645 else
646 retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
647 break;
648 }
649 case IPV6_JOIN_ANYCAST:
650 case IPV6_LEAVE_ANYCAST:
651 {
652 struct ipv6_mreq mreq;
653
654 if (optlen < sizeof(struct ipv6_mreq))
655 goto e_inval;
656
657 retv = -EFAULT;
658 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
659 break;
660
661 if (optname == IPV6_JOIN_ANYCAST)
662 retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
663 else
664 retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
665 break;
666 }
667 case MCAST_JOIN_GROUP:
668 case MCAST_LEAVE_GROUP:
669 {
670 struct group_req greq;
671 struct sockaddr_in6 *psin6;
672
673 if (optlen < sizeof(struct group_req))
674 goto e_inval;
675
676 retv = -EFAULT;
677 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
678 break;
679 if (greq.gr_group.ss_family != AF_INET6) {
680 retv = -EADDRNOTAVAIL;
681 break;
682 }
683 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
684 if (optname == MCAST_JOIN_GROUP)
685 retv = ipv6_sock_mc_join(sk, greq.gr_interface,
686 &psin6->sin6_addr);
687 else
688 retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
689 &psin6->sin6_addr);
690 break;
691 }
692 case MCAST_JOIN_SOURCE_GROUP:
693 case MCAST_LEAVE_SOURCE_GROUP:
694 case MCAST_BLOCK_SOURCE:
695 case MCAST_UNBLOCK_SOURCE:
696 {
697 struct group_source_req greqs;
698 int omode, add;
699
700 if (optlen < sizeof(struct group_source_req))
701 goto e_inval;
702 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
703 retv = -EFAULT;
704 break;
705 }
706 if (greqs.gsr_group.ss_family != AF_INET6 ||
707 greqs.gsr_source.ss_family != AF_INET6) {
708 retv = -EADDRNOTAVAIL;
709 break;
710 }
711 if (optname == MCAST_BLOCK_SOURCE) {
712 omode = MCAST_EXCLUDE;
713 add = 1;
714 } else if (optname == MCAST_UNBLOCK_SOURCE) {
715 omode = MCAST_EXCLUDE;
716 add = 0;
717 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
718 struct sockaddr_in6 *psin6;
719
720 psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
721 retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
722 &psin6->sin6_addr);
723 /* prior join w/ different source is ok */
724 if (retv && retv != -EADDRINUSE)
725 break;
726 omode = MCAST_INCLUDE;
727 add = 1;
728 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
729 omode = MCAST_INCLUDE;
730 add = 0;
731 }
732 retv = ip6_mc_source(add, omode, sk, &greqs);
733 break;
734 }
735 case MCAST_MSFILTER:
736 {
737 struct group_filter *gsf;
738
739 if (optlen < GROUP_FILTER_SIZE(0))
740 goto e_inval;
741 if (optlen > sysctl_optmem_max) {
742 retv = -ENOBUFS;
743 break;
744 }
745 gsf = memdup_user(optval, optlen);
746 if (IS_ERR(gsf)) {
747 retv = PTR_ERR(gsf);
748 break;
749 }
750 /* numsrc >= (4G-140)/128 overflow in 32 bits */
751 if (gsf->gf_numsrc >= 0x1ffffffU ||
752 gsf->gf_numsrc > sysctl_mld_max_msf) {
753 kfree(gsf);
754 retv = -ENOBUFS;
755 break;
756 }
757 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
758 kfree(gsf);
759 retv = -EINVAL;
760 break;
761 }
762 retv = ip6_mc_msfilter(sk, gsf);
763 kfree(gsf);
764
765 break;
766 }
767 case IPV6_ROUTER_ALERT:
768 if (optlen < sizeof(int))
769 goto e_inval;
770 retv = ip6_ra_control(sk, val);
771 break;
772 case IPV6_MTU_DISCOVER:
773 if (optlen < sizeof(int))
774 goto e_inval;
775 if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
776 goto e_inval;
777 np->pmtudisc = val;
778 retv = 0;
779 break;
780 case IPV6_MTU:
781 if (optlen < sizeof(int))
782 goto e_inval;
783 if (val && val < IPV6_MIN_MTU)
784 goto e_inval;
785 np->frag_size = val;
786 retv = 0;
787 break;
788 case IPV6_RECVERR:
789 if (optlen < sizeof(int))
790 goto e_inval;
791 np->recverr = valbool;
792 if (!val)
793 skb_queue_purge(&sk->sk_error_queue);
794 retv = 0;
795 break;
796 case IPV6_FLOWINFO_SEND:
797 if (optlen < sizeof(int))
798 goto e_inval;
799 np->sndflow = valbool;
800 retv = 0;
801 break;
802 case IPV6_FLOWLABEL_MGR:
803 retv = ipv6_flowlabel_opt(sk, optval, optlen);
804 break;
805 case IPV6_IPSEC_POLICY:
806 case IPV6_XFRM_POLICY:
807 retv = -EPERM;
808 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
809 break;
810 retv = xfrm_user_policy(sk, optname, optval, optlen);
811 break;
812
813 case IPV6_ADDR_PREFERENCES:
814 {
815 unsigned int pref = 0;
816 unsigned int prefmask = ~0;
817
818 if (optlen < sizeof(int))
819 goto e_inval;
820
821 retv = -EINVAL;
822
823 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
824 switch (val & (IPV6_PREFER_SRC_PUBLIC|
825 IPV6_PREFER_SRC_TMP|
826 IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
827 case IPV6_PREFER_SRC_PUBLIC:
828 pref |= IPV6_PREFER_SRC_PUBLIC;
829 break;
830 case IPV6_PREFER_SRC_TMP:
831 pref |= IPV6_PREFER_SRC_TMP;
832 break;
833 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
834 break;
835 case 0:
836 goto pref_skip_pubtmp;
837 default:
838 goto e_inval;
839 }
840
841 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
842 IPV6_PREFER_SRC_TMP);
843 pref_skip_pubtmp:
844
845 /* check HOME/COA conflicts */
846 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
847 case IPV6_PREFER_SRC_HOME:
848 break;
849 case IPV6_PREFER_SRC_COA:
850 pref |= IPV6_PREFER_SRC_COA;
851 case 0:
852 goto pref_skip_coa;
853 default:
854 goto e_inval;
855 }
856
857 prefmask &= ~IPV6_PREFER_SRC_COA;
858 pref_skip_coa:
859
860 /* check CGA/NONCGA conflicts */
861 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
862 case IPV6_PREFER_SRC_CGA:
863 case IPV6_PREFER_SRC_NONCGA:
864 case 0:
865 break;
866 default:
867 goto e_inval;
868 }
869
870 np->srcprefs = (np->srcprefs & prefmask) | pref;
871 retv = 0;
872
873 break;
874 }
875 case IPV6_MINHOPCOUNT:
876 if (optlen < sizeof(int))
877 goto e_inval;
878 if (val < 0 || val > 255)
879 goto e_inval;
880 np->min_hopcount = val;
881 retv = 0;
882 break;
883 case IPV6_DONTFRAG:
884 np->dontfrag = valbool;
885 retv = 0;
886 break;
887 case IPV6_AUTOFLOWLABEL:
888 np->autoflowlabel = valbool;
889 np->autoflowlabel_set = 1;
890 retv = 0;
891 break;
892 case IPV6_RECVFRAGSIZE:
893 np->rxopt.bits.recvfragsize = valbool;
894 retv = 0;
895 break;
896 }
897
898 release_sock(sk);
899 if (needs_rtnl)
900 rtnl_unlock();
901
902 return retv;
903
904 e_inval:
905 release_sock(sk);
906 if (needs_rtnl)
907 rtnl_unlock();
908 return -EINVAL;
909 }
910
911 int ipv6_setsockopt(struct sock *sk, int level, int optname,
912 char __user *optval, unsigned int optlen)
913 {
914 int err;
915
916 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
917 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
918
919 if (level != SOL_IPV6)
920 return -ENOPROTOOPT;
921
922 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
923 #ifdef CONFIG_NETFILTER
924 /* we need to exclude all possible ENOPROTOOPTs except default case */
925 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
926 optname != IPV6_XFRM_POLICY)
927 err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
928 #endif
929 return err;
930 }
931 EXPORT_SYMBOL(ipv6_setsockopt);
932
933 #ifdef CONFIG_COMPAT
934 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
935 char __user *optval, unsigned int optlen)
936 {
937 int err;
938
939 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
940 if (udp_prot.compat_setsockopt != NULL)
941 return udp_prot.compat_setsockopt(sk, level, optname,
942 optval, optlen);
943 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
944 }
945
946 if (level != SOL_IPV6)
947 return -ENOPROTOOPT;
948
949 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
950 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
951 ipv6_setsockopt);
952
953 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
954 #ifdef CONFIG_NETFILTER
955 /* we need to exclude all possible ENOPROTOOPTs except default case */
956 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
957 optname != IPV6_XFRM_POLICY)
958 err = compat_nf_setsockopt(sk, PF_INET6, optname, optval,
959 optlen);
960 #endif
961 return err;
962 }
963 EXPORT_SYMBOL(compat_ipv6_setsockopt);
964 #endif
965
966 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
967 int optname, char __user *optval, int len)
968 {
969 struct ipv6_opt_hdr *hdr;
970
971 if (!opt)
972 return 0;
973
974 switch (optname) {
975 case IPV6_HOPOPTS:
976 hdr = opt->hopopt;
977 break;
978 case IPV6_RTHDRDSTOPTS:
979 hdr = opt->dst0opt;
980 break;
981 case IPV6_RTHDR:
982 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
983 break;
984 case IPV6_DSTOPTS:
985 hdr = opt->dst1opt;
986 break;
987 default:
988 return -EINVAL; /* should not happen */
989 }
990
991 if (!hdr)
992 return 0;
993
994 len = min_t(unsigned int, len, ipv6_optlen(hdr));
995 if (copy_to_user(optval, hdr, len))
996 return -EFAULT;
997 return len;
998 }
999
1000 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1001 char __user *optval, int __user *optlen, unsigned int flags)
1002 {
1003 struct ipv6_pinfo *np = inet6_sk(sk);
1004 int len;
1005 int val;
1006
1007 if (ip6_mroute_opt(optname))
1008 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1009
1010 if (get_user(len, optlen))
1011 return -EFAULT;
1012 switch (optname) {
1013 case IPV6_ADDRFORM:
1014 if (sk->sk_protocol != IPPROTO_UDP &&
1015 sk->sk_protocol != IPPROTO_UDPLITE &&
1016 sk->sk_protocol != IPPROTO_TCP)
1017 return -ENOPROTOOPT;
1018 if (sk->sk_state != TCP_ESTABLISHED)
1019 return -ENOTCONN;
1020 val = sk->sk_family;
1021 break;
1022 case MCAST_MSFILTER:
1023 {
1024 struct group_filter gsf;
1025 int err;
1026
1027 if (len < GROUP_FILTER_SIZE(0))
1028 return -EINVAL;
1029 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
1030 return -EFAULT;
1031 if (gsf.gf_group.ss_family != AF_INET6)
1032 return -EADDRNOTAVAIL;
1033 lock_sock(sk);
1034 err = ip6_mc_msfget(sk, &gsf,
1035 (struct group_filter __user *)optval, optlen);
1036 release_sock(sk);
1037 return err;
1038 }
1039
1040 case IPV6_2292PKTOPTIONS:
1041 {
1042 struct msghdr msg;
1043 struct sk_buff *skb;
1044
1045 if (sk->sk_type != SOCK_STREAM)
1046 return -ENOPROTOOPT;
1047
1048 msg.msg_control = optval;
1049 msg.msg_controllen = len;
1050 msg.msg_flags = flags;
1051
1052 lock_sock(sk);
1053 skb = np->pktoptions;
1054 if (skb)
1055 ip6_datagram_recv_ctl(sk, &msg, skb);
1056 release_sock(sk);
1057 if (!skb) {
1058 if (np->rxopt.bits.rxinfo) {
1059 struct in6_pktinfo src_info;
1060 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1061 np->sticky_pktinfo.ipi6_ifindex;
1062 src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1063 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1064 }
1065 if (np->rxopt.bits.rxhlim) {
1066 int hlim = np->mcast_hops;
1067 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1068 }
1069 if (np->rxopt.bits.rxtclass) {
1070 int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1071
1072 put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1073 }
1074 if (np->rxopt.bits.rxoinfo) {
1075 struct in6_pktinfo src_info;
1076 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1077 np->sticky_pktinfo.ipi6_ifindex;
1078 src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
1079 np->sticky_pktinfo.ipi6_addr;
1080 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1081 }
1082 if (np->rxopt.bits.rxohlim) {
1083 int hlim = np->mcast_hops;
1084 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1085 }
1086 if (np->rxopt.bits.rxflow) {
1087 __be32 flowinfo = np->rcv_flowinfo;
1088
1089 put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1090 }
1091 }
1092 len -= msg.msg_controllen;
1093 return put_user(len, optlen);
1094 }
1095 case IPV6_MTU:
1096 {
1097 struct dst_entry *dst;
1098
1099 val = 0;
1100 rcu_read_lock();
1101 dst = __sk_dst_get(sk);
1102 if (dst)
1103 val = dst_mtu(dst);
1104 rcu_read_unlock();
1105 if (!val)
1106 return -ENOTCONN;
1107 break;
1108 }
1109
1110 case IPV6_V6ONLY:
1111 val = sk->sk_ipv6only;
1112 break;
1113
1114 case IPV6_RECVPKTINFO:
1115 val = np->rxopt.bits.rxinfo;
1116 break;
1117
1118 case IPV6_2292PKTINFO:
1119 val = np->rxopt.bits.rxoinfo;
1120 break;
1121
1122 case IPV6_RECVHOPLIMIT:
1123 val = np->rxopt.bits.rxhlim;
1124 break;
1125
1126 case IPV6_2292HOPLIMIT:
1127 val = np->rxopt.bits.rxohlim;
1128 break;
1129
1130 case IPV6_RECVRTHDR:
1131 val = np->rxopt.bits.srcrt;
1132 break;
1133
1134 case IPV6_2292RTHDR:
1135 val = np->rxopt.bits.osrcrt;
1136 break;
1137
1138 case IPV6_HOPOPTS:
1139 case IPV6_RTHDRDSTOPTS:
1140 case IPV6_RTHDR:
1141 case IPV6_DSTOPTS:
1142 {
1143 struct ipv6_txoptions *opt;
1144
1145 lock_sock(sk);
1146 opt = rcu_dereference_protected(np->opt,
1147 lockdep_sock_is_held(sk));
1148 len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1149 release_sock(sk);
1150 /* check if ipv6_getsockopt_sticky() returns err code */
1151 if (len < 0)
1152 return len;
1153 return put_user(len, optlen);
1154 }
1155
1156 case IPV6_RECVHOPOPTS:
1157 val = np->rxopt.bits.hopopts;
1158 break;
1159
1160 case IPV6_2292HOPOPTS:
1161 val = np->rxopt.bits.ohopopts;
1162 break;
1163
1164 case IPV6_RECVDSTOPTS:
1165 val = np->rxopt.bits.dstopts;
1166 break;
1167
1168 case IPV6_2292DSTOPTS:
1169 val = np->rxopt.bits.odstopts;
1170 break;
1171
1172 case IPV6_TCLASS:
1173 val = np->tclass;
1174 break;
1175
1176 case IPV6_RECVTCLASS:
1177 val = np->rxopt.bits.rxtclass;
1178 break;
1179
1180 case IPV6_FLOWINFO:
1181 val = np->rxopt.bits.rxflow;
1182 break;
1183
1184 case IPV6_RECVPATHMTU:
1185 val = np->rxopt.bits.rxpmtu;
1186 break;
1187
1188 case IPV6_PATHMTU:
1189 {
1190 struct dst_entry *dst;
1191 struct ip6_mtuinfo mtuinfo;
1192
1193 if (len < sizeof(mtuinfo))
1194 return -EINVAL;
1195
1196 len = sizeof(mtuinfo);
1197 memset(&mtuinfo, 0, sizeof(mtuinfo));
1198
1199 rcu_read_lock();
1200 dst = __sk_dst_get(sk);
1201 if (dst)
1202 mtuinfo.ip6m_mtu = dst_mtu(dst);
1203 rcu_read_unlock();
1204 if (!mtuinfo.ip6m_mtu)
1205 return -ENOTCONN;
1206
1207 if (put_user(len, optlen))
1208 return -EFAULT;
1209 if (copy_to_user(optval, &mtuinfo, len))
1210 return -EFAULT;
1211
1212 return 0;
1213 }
1214
1215 case IPV6_TRANSPARENT:
1216 val = inet_sk(sk)->transparent;
1217 break;
1218
1219 case IPV6_FREEBIND:
1220 val = inet_sk(sk)->freebind;
1221 break;
1222
1223 case IPV6_RECVORIGDSTADDR:
1224 val = np->rxopt.bits.rxorigdstaddr;
1225 break;
1226
1227 case IPV6_UNICAST_HOPS:
1228 case IPV6_MULTICAST_HOPS:
1229 {
1230 struct dst_entry *dst;
1231
1232 if (optname == IPV6_UNICAST_HOPS)
1233 val = np->hop_limit;
1234 else
1235 val = np->mcast_hops;
1236
1237 if (val < 0) {
1238 rcu_read_lock();
1239 dst = __sk_dst_get(sk);
1240 if (dst)
1241 val = ip6_dst_hoplimit(dst);
1242 rcu_read_unlock();
1243 }
1244
1245 if (val < 0)
1246 val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1247 break;
1248 }
1249
1250 case IPV6_MULTICAST_LOOP:
1251 val = np->mc_loop;
1252 break;
1253
1254 case IPV6_MULTICAST_IF:
1255 val = np->mcast_oif;
1256 break;
1257
1258 case IPV6_UNICAST_IF:
1259 val = (__force int)htonl((__u32) np->ucast_oif);
1260 break;
1261
1262 case IPV6_MTU_DISCOVER:
1263 val = np->pmtudisc;
1264 break;
1265
1266 case IPV6_RECVERR:
1267 val = np->recverr;
1268 break;
1269
1270 case IPV6_FLOWINFO_SEND:
1271 val = np->sndflow;
1272 break;
1273
1274 case IPV6_FLOWLABEL_MGR:
1275 {
1276 struct in6_flowlabel_req freq;
1277 int flags;
1278
1279 if (len < sizeof(freq))
1280 return -EINVAL;
1281
1282 if (copy_from_user(&freq, optval, sizeof(freq)))
1283 return -EFAULT;
1284
1285 if (freq.flr_action != IPV6_FL_A_GET)
1286 return -EINVAL;
1287
1288 len = sizeof(freq);
1289 flags = freq.flr_flags;
1290
1291 memset(&freq, 0, sizeof(freq));
1292
1293 val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1294 if (val < 0)
1295 return val;
1296
1297 if (put_user(len, optlen))
1298 return -EFAULT;
1299 if (copy_to_user(optval, &freq, len))
1300 return -EFAULT;
1301
1302 return 0;
1303 }
1304
1305 case IPV6_ADDR_PREFERENCES:
1306 val = 0;
1307
1308 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1309 val |= IPV6_PREFER_SRC_TMP;
1310 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1311 val |= IPV6_PREFER_SRC_PUBLIC;
1312 else {
1313 /* XXX: should we return system default? */
1314 val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1315 }
1316
1317 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1318 val |= IPV6_PREFER_SRC_COA;
1319 else
1320 val |= IPV6_PREFER_SRC_HOME;
1321 break;
1322
1323 case IPV6_MINHOPCOUNT:
1324 val = np->min_hopcount;
1325 break;
1326
1327 case IPV6_DONTFRAG:
1328 val = np->dontfrag;
1329 break;
1330
1331 case IPV6_AUTOFLOWLABEL:
1332 val = ip6_autoflowlabel(sock_net(sk), np);
1333 break;
1334
1335 case IPV6_RECVFRAGSIZE:
1336 val = np->rxopt.bits.recvfragsize;
1337 break;
1338
1339 default:
1340 return -ENOPROTOOPT;
1341 }
1342 len = min_t(unsigned int, sizeof(int), len);
1343 if (put_user(len, optlen))
1344 return -EFAULT;
1345 if (copy_to_user(optval, &val, len))
1346 return -EFAULT;
1347 return 0;
1348 }
1349
1350 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1351 char __user *optval, int __user *optlen)
1352 {
1353 int err;
1354
1355 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1356 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1357
1358 if (level != SOL_IPV6)
1359 return -ENOPROTOOPT;
1360
1361 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen, 0);
1362 #ifdef CONFIG_NETFILTER
1363 /* we need to exclude all possible ENOPROTOOPTs except default case */
1364 if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1365 int len;
1366
1367 if (get_user(len, optlen))
1368 return -EFAULT;
1369
1370 err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1371 if (err >= 0)
1372 err = put_user(len, optlen);
1373 }
1374 #endif
1375 return err;
1376 }
1377 EXPORT_SYMBOL(ipv6_getsockopt);
1378
1379 #ifdef CONFIG_COMPAT
1380 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1381 char __user *optval, int __user *optlen)
1382 {
1383 int err;
1384
1385 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1386 if (udp_prot.compat_getsockopt != NULL)
1387 return udp_prot.compat_getsockopt(sk, level, optname,
1388 optval, optlen);
1389 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1390 }
1391
1392 if (level != SOL_IPV6)
1393 return -ENOPROTOOPT;
1394
1395 if (optname == MCAST_MSFILTER)
1396 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1397 ipv6_getsockopt);
1398
1399 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen,
1400 MSG_CMSG_COMPAT);
1401 #ifdef CONFIG_NETFILTER
1402 /* we need to exclude all possible ENOPROTOOPTs except default case */
1403 if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1404 int len;
1405
1406 if (get_user(len, optlen))
1407 return -EFAULT;
1408
1409 err = compat_nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1410 if (err >= 0)
1411 err = put_user(len, optlen);
1412 }
1413 #endif
1414 return err;
1415 }
1416 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1417 #endif
(deprecated) Btrfs devel tree