3 * Linux INET6 implementation
8 * Adapted from linux/net/ipv4/raw.c
11 * Hideaki YOSHIFUJI : sin6_scope_id support
12 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
13 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
15 * This program is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU General Public License
17 * as published by the Free Software Foundation; either version
18 * 2 of the License, or (at your option) any later version.
21 #include <linux/errno.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/slab.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmpv6.h>
31 #include <linux/netfilter.h>
32 #include <linux/netfilter_ipv6.h>
33 #include <linux/skbuff.h>
34 #include <asm/uaccess.h>
35 #include <asm/ioctls.h>
37 #include <net/net_namespace.h>
43 #include <net/ndisc.h>
44 #include <net/protocol.h>
45 #include <net/ip6_route.h>
46 #include <net/ip6_checksum.h>
47 #include <net/addrconf.h>
48 #include <net/transp_v6.h>
50 #include <net/inet_common.h>
51 #include <net/tcp_states.h>
52 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
55 #include <linux/mroute6.h>
58 #include <net/rawv6.h>
61 #include <linux/proc_fs.h>
62 #include <linux/seq_file.h>
64 static struct raw_hashinfo raw_v6_hashinfo = {
65 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
68 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
69 unsigned short num, struct in6_addr *loc_addr,
70 struct in6_addr *rmt_addr, int dif)
72 struct hlist_node *node;
73 int is_multicast = ipv6_addr_is_multicast(loc_addr);
75 sk_for_each_from(sk, node)
76 if (inet_sk(sk)->inet_num == num) {
77 struct ipv6_pinfo *np = inet6_sk(sk);
79 if (!net_eq(sock_net(sk), net))
82 if (!ipv6_addr_any(&np->daddr) &&
83 !ipv6_addr_equal(&np->daddr, rmt_addr))
86 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
89 if (!ipv6_addr_any(&np->rcv_saddr)) {
90 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
93 inet6_mc_check(sk, loc_addr, rmt_addr))
108 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
110 struct icmp6hdr *icmph;
111 struct raw6_sock *rp = raw6_sk(sk);
113 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
114 __u32 *data = &rp->filter.data[0];
117 icmph = (struct icmp6hdr *) skb->data;
118 bit_nr = icmph->icmp6_type;
120 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
125 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
126 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb);
128 int rawv6_mh_filter_register(int (*filter)(struct sock *sock,
129 struct sk_buff *skb))
131 rcu_assign_pointer(mh_filter, filter);
134 EXPORT_SYMBOL(rawv6_mh_filter_register);
136 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock,
137 struct sk_buff *skb))
139 rcu_assign_pointer(mh_filter, NULL);
143 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
148 * demultiplex raw sockets.
149 * (should consider queueing the skb in the sock receive_queue
150 * without calling rawv6.c)
152 * Caller owns SKB so we must make clones.
154 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
156 struct in6_addr *saddr;
157 struct in6_addr *daddr;
163 saddr = &ipv6_hdr(skb)->saddr;
166 hash = nexthdr & (MAX_INET_PROTOS - 1);
168 read_lock(&raw_v6_hashinfo.lock);
169 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
174 net = dev_net(skb->dev);
175 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
183 filtered = icmpv6_filter(sk, skb);
186 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
189 /* XXX: To validate MH only once for each packet,
190 * this is placed here. It should be after checking
191 * xfrm policy, however it doesn't. The checking xfrm
192 * policy is placed in rawv6_rcv() because it is
193 * required for each socket.
195 int (*filter)(struct sock *sock, struct sk_buff *skb);
197 filter = rcu_dereference(mh_filter);
198 filtered = filter ? filter(sk, skb) : 0;
210 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
212 /* Not releasing hash table! */
215 rawv6_rcv(sk, clone);
218 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
222 read_unlock(&raw_v6_hashinfo.lock);
226 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
230 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
231 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
234 return raw_sk != NULL;
237 /* This cleans up af_inet6 a bit. -DaveM */
238 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
240 struct inet_sock *inet = inet_sk(sk);
241 struct ipv6_pinfo *np = inet6_sk(sk);
242 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
247 if (addr_len < SIN6_LEN_RFC2133)
249 addr_type = ipv6_addr_type(&addr->sin6_addr);
251 /* Raw sockets are IPv6 only */
252 if (addr_type == IPV6_ADDR_MAPPED)
253 return -EADDRNOTAVAIL;
258 if (sk->sk_state != TCP_CLOSE)
262 /* Check if the address belongs to the host. */
263 if (addr_type != IPV6_ADDR_ANY) {
264 struct net_device *dev = NULL;
266 if (addr_type & IPV6_ADDR_LINKLOCAL) {
267 if (addr_len >= sizeof(struct sockaddr_in6) &&
268 addr->sin6_scope_id) {
269 /* Override any existing binding, if another
270 * one is supplied by user.
272 sk->sk_bound_dev_if = addr->sin6_scope_id;
275 /* Binding to link-local address requires an interface */
276 if (!sk->sk_bound_dev_if)
280 dev = dev_get_by_index_rcu(sock_net(sk),
281 sk->sk_bound_dev_if);
286 /* ipv4 addr of the socket is invalid. Only the
287 * unspecified and mapped address have a v4 equivalent.
289 v4addr = LOOPBACK4_IPV6;
290 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
291 err = -EADDRNOTAVAIL;
292 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
299 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
300 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
301 if (!(addr_type & IPV6_ADDR_MULTICAST))
302 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
311 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
312 struct inet6_skb_parm *opt,
313 u8 type, u8 code, int offset, __be32 info)
315 struct inet_sock *inet = inet_sk(sk);
316 struct ipv6_pinfo *np = inet6_sk(sk);
320 /* Report error on raw socket, if:
321 1. User requested recverr.
322 2. Socket is connected (otherwise the error indication
323 is useless without recverr and error is hard.
325 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
328 harderr = icmpv6_err_convert(type, code, &err);
329 if (type == ICMPV6_PKT_TOOBIG)
330 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
333 u8 *payload = skb->data;
336 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
339 if (np->recverr || harderr) {
341 sk->sk_error_report(sk);
345 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
346 u8 type, u8 code, int inner_offset, __be32 info)
350 struct in6_addr *saddr, *daddr;
353 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
355 read_lock(&raw_v6_hashinfo.lock);
356 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
358 /* Note: ipv6_hdr(skb) != skb->data */
359 struct ipv6hdr *ip6h = (struct ipv6hdr *)skb->data;
360 saddr = &ip6h->saddr;
361 daddr = &ip6h->daddr;
362 net = dev_net(skb->dev);
364 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
366 rawv6_err(sk, skb, NULL, type, code,
371 read_unlock(&raw_v6_hashinfo.lock);
374 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
376 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
377 skb_checksum_complete(skb)) {
378 atomic_inc(&sk->sk_drops);
383 /* Charge it to the socket. */
384 if (ip_queue_rcv_skb(sk, skb) < 0) {
393 * This is next to useless...
394 * if we demultiplex in network layer we don't need the extra call
395 * just to queue the skb...
396 * maybe we could have the network decide upon a hint if it
397 * should call raw_rcv for demultiplexing
399 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
401 struct inet_sock *inet = inet_sk(sk);
402 struct raw6_sock *rp = raw6_sk(sk);
404 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
405 atomic_inc(&sk->sk_drops);
411 skb->ip_summed = CHECKSUM_UNNECESSARY;
413 if (skb->ip_summed == CHECKSUM_COMPLETE) {
414 skb_postpull_rcsum(skb, skb_network_header(skb),
415 skb_network_header_len(skb));
416 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
417 &ipv6_hdr(skb)->daddr,
418 skb->len, inet->inet_num, skb->csum))
419 skb->ip_summed = CHECKSUM_UNNECESSARY;
421 if (!skb_csum_unnecessary(skb))
422 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
423 &ipv6_hdr(skb)->daddr,
428 if (skb_checksum_complete(skb)) {
429 atomic_inc(&sk->sk_drops);
435 rawv6_rcv_skb(sk, skb);
441 * This should be easy, if there is something there
442 * we return it, otherwise we block.
445 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
446 struct msghdr *msg, size_t len,
447 int noblock, int flags, int *addr_len)
449 struct ipv6_pinfo *np = inet6_sk(sk);
450 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
459 *addr_len=sizeof(*sin6);
461 if (flags & MSG_ERRQUEUE)
462 return ipv6_recv_error(sk, msg, len);
464 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
465 return ipv6_recv_rxpmtu(sk, msg, len);
467 skb = skb_recv_datagram(sk, flags, noblock, &err);
474 msg->msg_flags |= MSG_TRUNC;
477 if (skb_csum_unnecessary(skb)) {
478 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
479 } else if (msg->msg_flags&MSG_TRUNC) {
480 if (__skb_checksum_complete(skb))
482 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
484 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
491 /* Copy the address. */
493 sin6->sin6_family = AF_INET6;
495 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
496 sin6->sin6_flowinfo = 0;
497 sin6->sin6_scope_id = 0;
498 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
499 sin6->sin6_scope_id = IP6CB(skb)->iif;
502 sock_recv_ts_and_drops(msg, sk, skb);
505 datagram_recv_ctl(sk, msg, skb);
508 if (flags & MSG_TRUNC)
512 skb_free_datagram(sk, skb);
517 skb_kill_datagram(sk, skb, flags);
519 /* Error for blocking case is chosen to masquerade
520 as some normal condition.
522 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
526 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
527 struct raw6_sock *rp)
540 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
544 total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
546 if (offset >= total_len - 1) {
548 ip6_flush_pending_frames(sk);
552 /* should be check HW csum miyazawa */
553 if (skb_queue_len(&sk->sk_write_queue) == 1) {
555 * Only one fragment on the socket.
557 tmp_csum = skb->csum;
559 struct sk_buff *csum_skb = NULL;
562 skb_queue_walk(&sk->sk_write_queue, skb) {
563 tmp_csum = csum_add(tmp_csum, skb->csum);
568 len = skb->len - skb_transport_offset(skb);
580 offset += skb_transport_offset(skb);
581 if (skb_copy_bits(skb, offset, &csum, 2))
584 /* in case cksum was not initialized */
586 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
588 csum = csum_ipv6_magic(&fl->fl6_src,
590 total_len, fl->proto, tmp_csum);
592 if (csum == 0 && fl->proto == IPPROTO_UDP)
593 csum = CSUM_MANGLED_0;
595 if (skb_store_bits(skb, offset, &csum, 2))
599 err = ip6_push_pending_frames(sk);
604 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
605 struct flowi *fl, struct rt6_info *rt,
608 struct ipv6_pinfo *np = inet6_sk(sk);
613 if (length > rt->u.dst.dev->mtu) {
614 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
620 skb = sock_alloc_send_skb(sk,
621 length + LL_ALLOCATED_SPACE(rt->u.dst.dev) + 15,
622 flags & MSG_DONTWAIT, &err);
625 skb_reserve(skb, LL_RESERVED_SPACE(rt->u.dst.dev));
627 skb->priority = sk->sk_priority;
628 skb->mark = sk->sk_mark;
629 skb_dst_set(skb, dst_clone(&rt->u.dst));
631 skb_put(skb, length);
632 skb_reset_network_header(skb);
635 skb->ip_summed = CHECKSUM_NONE;
637 skb->transport_header = skb->network_header;
638 err = memcpy_fromiovecend((void *)iph, from, 0, length);
642 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
643 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
644 rt->u.dst.dev, dst_output);
646 err = net_xmit_errno(err);
656 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
657 if (err == -ENOBUFS && !np->recverr)
662 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
665 u8 __user *type = NULL;
666 u8 __user *code = NULL;
674 for (i = 0; i < msg->msg_iovlen; i++) {
675 iov = &msg->msg_iov[i];
681 /* check if one-byte field is readable or not. */
682 if (iov->iov_base && iov->iov_len < 1)
686 type = iov->iov_base;
687 /* check if code field is readable or not. */
688 if (iov->iov_len > 1)
691 code = iov->iov_base;
694 if (get_user(fl->fl_icmp_type, type) ||
695 get_user(fl->fl_icmp_code, code))
701 if (iov->iov_base && iov->iov_len < 1)
703 /* check if type field is readable or not. */
704 if (iov->iov_len > 2 - len) {
705 u8 __user *p = iov->iov_base;
706 if (get_user(fl->fl_mh_type, &p[2 - len]))
723 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
724 struct msghdr *msg, size_t len)
726 struct ipv6_txoptions opt_space;
727 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
728 struct in6_addr *daddr, *final_p = NULL, final;
729 struct inet_sock *inet = inet_sk(sk);
730 struct ipv6_pinfo *np = inet6_sk(sk);
731 struct raw6_sock *rp = raw6_sk(sk);
732 struct ipv6_txoptions *opt = NULL;
733 struct ip6_flowlabel *flowlabel = NULL;
734 struct dst_entry *dst = NULL;
736 int addr_len = msg->msg_namelen;
743 /* Rough check on arithmetic overflow,
744 better check is made in ip6_append_data().
749 /* Mirror BSD error message compatibility */
750 if (msg->msg_flags & MSG_OOB)
754 * Get and verify the address.
756 memset(&fl, 0, sizeof(fl));
758 fl.mark = sk->sk_mark;
761 if (addr_len < SIN6_LEN_RFC2133)
764 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
765 return(-EAFNOSUPPORT);
767 /* port is the proto value [0..255] carried in nexthdr */
768 proto = ntohs(sin6->sin6_port);
771 proto = inet->inet_num;
772 else if (proto != inet->inet_num)
778 daddr = &sin6->sin6_addr;
780 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
781 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
782 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
783 if (flowlabel == NULL)
785 daddr = &flowlabel->dst;
790 * Otherwise it will be difficult to maintain
793 if (sk->sk_state == TCP_ESTABLISHED &&
794 ipv6_addr_equal(daddr, &np->daddr))
797 if (addr_len >= sizeof(struct sockaddr_in6) &&
798 sin6->sin6_scope_id &&
799 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
800 fl.oif = sin6->sin6_scope_id;
802 if (sk->sk_state != TCP_ESTABLISHED)
803 return -EDESTADDRREQ;
805 proto = inet->inet_num;
807 fl.fl6_flowlabel = np->flow_label;
811 fl.oif = sk->sk_bound_dev_if;
813 if (msg->msg_controllen) {
815 memset(opt, 0, sizeof(struct ipv6_txoptions));
816 opt->tot_len = sizeof(struct ipv6_txoptions);
818 err = datagram_send_ctl(sock_net(sk), msg, &fl, opt, &hlimit,
821 fl6_sock_release(flowlabel);
824 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
825 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
826 if (flowlabel == NULL)
829 if (!(opt->opt_nflen|opt->opt_flen))
835 opt = fl6_merge_options(&opt_space, flowlabel, opt);
836 opt = ipv6_fixup_options(&opt_space, opt);
839 err = rawv6_probe_proto_opt(&fl, msg);
843 if (!ipv6_addr_any(daddr))
844 ipv6_addr_copy(&fl.fl6_dst, daddr);
846 fl.fl6_dst.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
847 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
848 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
850 /* merge ip6_build_xmit from ip6_output */
851 if (opt && opt->srcrt) {
852 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
853 ipv6_addr_copy(&final, &fl.fl6_dst);
854 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
858 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
859 fl.oif = np->mcast_oif;
860 security_sk_classify_flow(sk, &fl);
862 err = ip6_dst_lookup(sk, &dst, &fl);
866 ipv6_addr_copy(&fl.fl6_dst, final_p);
868 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
871 err = ip6_dst_blackhole(sk, &dst, &fl);
877 if (ipv6_addr_is_multicast(&fl.fl6_dst))
878 hlimit = np->mcast_hops;
880 hlimit = np->hop_limit;
882 hlimit = ip6_dst_hoplimit(dst);
889 dontfrag = np->dontfrag;
891 if (msg->msg_flags&MSG_CONFIRM)
896 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
899 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
900 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
901 msg->msg_flags, dontfrag);
904 ip6_flush_pending_frames(sk);
905 else if (!(msg->msg_flags & MSG_MORE))
906 err = rawv6_push_pending_frames(sk, &fl, rp);
912 fl6_sock_release(flowlabel);
913 return err<0?err:len;
916 if (!(msg->msg_flags & MSG_PROBE) || len)
917 goto back_from_confirm;
922 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
923 char __user *optval, int optlen)
927 if (optlen > sizeof(struct icmp6_filter))
928 optlen = sizeof(struct icmp6_filter);
929 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
939 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
940 char __user *optval, int __user *optlen)
946 if (get_user(len, optlen))
950 if (len > sizeof(struct icmp6_filter))
951 len = sizeof(struct icmp6_filter);
952 if (put_user(len, optlen))
954 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
965 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
966 char __user *optval, unsigned int optlen)
968 struct raw6_sock *rp = raw6_sk(sk);
971 if (get_user(val, (int __user *)optval))
976 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
977 level == IPPROTO_IPV6) {
979 * RFC3542 tells that IPV6_CHECKSUM socket
980 * option in the IPPROTO_IPV6 level is not
981 * allowed on ICMPv6 sockets.
982 * If you want to set it, use IPPROTO_RAW
983 * level IPV6_CHECKSUM socket option
989 /* You may get strange result with a positive odd offset;
990 RFC2292bis agrees with me. */
991 if (val > 0 && (val&1))
1004 return(-ENOPROTOOPT);
1008 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1009 char __user *optval, unsigned int optlen)
1016 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1018 return rawv6_seticmpfilter(sk, level, optname, optval,
1021 if (optname == IPV6_CHECKSUM)
1024 return ipv6_setsockopt(sk, level, optname, optval,
1028 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1031 #ifdef CONFIG_COMPAT
1032 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1033 char __user *optval, unsigned int optlen)
1039 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1041 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1043 if (optname == IPV6_CHECKSUM)
1046 return compat_ipv6_setsockopt(sk, level, optname,
1049 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1053 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1054 char __user *optval, int __user *optlen)
1056 struct raw6_sock *rp = raw6_sk(sk);
1059 if (get_user(len,optlen))
1065 * We allow getsockopt() for IPPROTO_IPV6-level
1066 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1067 * since RFC3542 is silent about it.
1069 if (rp->checksum == 0)
1076 return -ENOPROTOOPT;
1079 len = min_t(unsigned int, sizeof(int), len);
1081 if (put_user(len, optlen))
1083 if (copy_to_user(optval,&val,len))
1088 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1089 char __user *optval, int __user *optlen)
1096 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1098 return rawv6_geticmpfilter(sk, level, optname, optval,
1101 if (optname == IPV6_CHECKSUM)
1104 return ipv6_getsockopt(sk, level, optname, optval,
1108 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1111 #ifdef CONFIG_COMPAT
1112 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1113 char __user *optval, int __user *optlen)
1119 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1121 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1123 if (optname == IPV6_CHECKSUM)
1126 return compat_ipv6_getsockopt(sk, level, optname,
1129 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1133 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1138 int amount = sk_wmem_alloc_get(sk);
1140 return put_user(amount, (int __user *)arg);
1144 struct sk_buff *skb;
1147 spin_lock_bh(&sk->sk_receive_queue.lock);
1148 skb = skb_peek(&sk->sk_receive_queue);
1150 amount = skb->tail - skb->transport_header;
1151 spin_unlock_bh(&sk->sk_receive_queue.lock);
1152 return put_user(amount, (int __user *)arg);
1156 #ifdef CONFIG_IPV6_MROUTE
1157 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1159 return -ENOIOCTLCMD;
1164 static void rawv6_close(struct sock *sk, long timeout)
1166 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1167 ip6_ra_control(sk, -1);
1169 sk_common_release(sk);
1172 static void raw6_destroy(struct sock *sk)
1175 ip6_flush_pending_frames(sk);
1178 inet6_destroy_sock(sk);
1181 static int rawv6_init_sk(struct sock *sk)
1183 struct raw6_sock *rp = raw6_sk(sk);
1185 switch (inet_sk(sk)->inet_num) {
1186 case IPPROTO_ICMPV6:
1200 struct proto rawv6_prot = {
1202 .owner = THIS_MODULE,
1203 .close = rawv6_close,
1204 .destroy = raw6_destroy,
1205 .connect = ip6_datagram_connect,
1206 .disconnect = udp_disconnect,
1207 .ioctl = rawv6_ioctl,
1208 .init = rawv6_init_sk,
1209 .setsockopt = rawv6_setsockopt,
1210 .getsockopt = rawv6_getsockopt,
1211 .sendmsg = rawv6_sendmsg,
1212 .recvmsg = rawv6_recvmsg,
1214 .backlog_rcv = rawv6_rcv_skb,
1215 .hash = raw_hash_sk,
1216 .unhash = raw_unhash_sk,
1217 .obj_size = sizeof(struct raw6_sock),
1218 .h.raw_hash = &raw_v6_hashinfo,
1219 #ifdef CONFIG_COMPAT
1220 .compat_setsockopt = compat_rawv6_setsockopt,
1221 .compat_getsockopt = compat_rawv6_getsockopt,
1225 #ifdef CONFIG_PROC_FS
1226 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1228 struct ipv6_pinfo *np = inet6_sk(sp);
1229 struct in6_addr *dest, *src;
1233 src = &np->rcv_saddr;
1235 srcp = inet_sk(sp)->inet_num;
1237 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1238 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1240 src->s6_addr32[0], src->s6_addr32[1],
1241 src->s6_addr32[2], src->s6_addr32[3], srcp,
1242 dest->s6_addr32[0], dest->s6_addr32[1],
1243 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1245 sk_wmem_alloc_get(sp),
1246 sk_rmem_alloc_get(sp),
1250 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1253 static int raw6_seq_show(struct seq_file *seq, void *v)
1255 if (v == SEQ_START_TOKEN)
1260 "st tx_queue rx_queue tr tm->when retrnsmt"
1261 " uid timeout inode ref pointer drops\n");
1263 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1267 static const struct seq_operations raw6_seq_ops = {
1268 .start = raw_seq_start,
1269 .next = raw_seq_next,
1270 .stop = raw_seq_stop,
1271 .show = raw6_seq_show,
1274 static int raw6_seq_open(struct inode *inode, struct file *file)
1276 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1279 static const struct file_operations raw6_seq_fops = {
1280 .owner = THIS_MODULE,
1281 .open = raw6_seq_open,
1283 .llseek = seq_lseek,
1284 .release = seq_release_net,
1287 static int __net_init raw6_init_net(struct net *net)
1289 if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1295 static void __net_exit raw6_exit_net(struct net *net)
1297 proc_net_remove(net, "raw6");
1300 static struct pernet_operations raw6_net_ops = {
1301 .init = raw6_init_net,
1302 .exit = raw6_exit_net,
1305 int __init raw6_proc_init(void)
1307 return register_pernet_subsys(&raw6_net_ops);
1310 void raw6_proc_exit(void)
1312 unregister_pernet_subsys(&raw6_net_ops);
1314 #endif /* CONFIG_PROC_FS */
1316 /* Same as inet6_dgram_ops, sans udp_poll. */
1317 static const struct proto_ops inet6_sockraw_ops = {
1319 .owner = THIS_MODULE,
1320 .release = inet6_release,
1322 .connect = inet_dgram_connect, /* ok */
1323 .socketpair = sock_no_socketpair, /* a do nothing */
1324 .accept = sock_no_accept, /* a do nothing */
1325 .getname = inet6_getname,
1326 .poll = datagram_poll, /* ok */
1327 .ioctl = inet6_ioctl, /* must change */
1328 .listen = sock_no_listen, /* ok */
1329 .shutdown = inet_shutdown, /* ok */
1330 .setsockopt = sock_common_setsockopt, /* ok */
1331 .getsockopt = sock_common_getsockopt, /* ok */
1332 .sendmsg = inet_sendmsg, /* ok */
1333 .recvmsg = sock_common_recvmsg, /* ok */
1334 .mmap = sock_no_mmap,
1335 .sendpage = sock_no_sendpage,
1336 #ifdef CONFIG_COMPAT
1337 .compat_setsockopt = compat_sock_common_setsockopt,
1338 .compat_getsockopt = compat_sock_common_getsockopt,
1342 static struct inet_protosw rawv6_protosw = {
1344 .protocol = IPPROTO_IP, /* wild card */
1345 .prot = &rawv6_prot,
1346 .ops = &inet6_sockraw_ops,
1347 .no_check = UDP_CSUM_DEFAULT,
1348 .flags = INET_PROTOSW_REUSE,
1351 int __init rawv6_init(void)
1355 ret = inet6_register_protosw(&rawv6_protosw);
1362 void rawv6_exit(void)
1364 inet6_unregister_protosw(&rawv6_protosw);