3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
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 <linux/compat.h>
35 #include <asm/uaccess.h>
36 #include <asm/ioctls.h>
38 #include <net/net_namespace.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/ip6_route.h>
47 #include <net/ip6_checksum.h>
48 #include <net/addrconf.h>
49 #include <net/transp_v6.h>
51 #include <net/inet_common.h>
52 #include <net/tcp_states.h>
53 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
56 #include <linux/mroute6.h>
59 #include <net/rawv6.h>
62 #include <linux/proc_fs.h>
63 #include <linux/seq_file.h>
64 #include <linux/export.h>
66 static struct raw_hashinfo raw_v6_hashinfo = {
67 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
70 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
71 unsigned short num, const struct in6_addr *loc_addr,
72 const struct in6_addr *rmt_addr, int dif)
74 struct hlist_node *node;
75 bool is_multicast = ipv6_addr_is_multicast(loc_addr);
77 sk_for_each_from(sk, node)
78 if (inet_sk(sk)->inet_num == num) {
79 struct ipv6_pinfo *np = inet6_sk(sk);
81 if (!net_eq(sock_net(sk), net))
84 if (!ipv6_addr_any(&np->daddr) &&
85 !ipv6_addr_equal(&np->daddr, rmt_addr))
88 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
91 if (!ipv6_addr_any(&np->rcv_saddr)) {
92 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
95 inet6_mc_check(sk, loc_addr, rmt_addr))
110 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
112 struct icmp6hdr *icmph;
113 struct raw6_sock *rp = raw6_sk(sk);
115 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
116 __u32 *data = &rp->filter.data[0];
119 icmph = (struct icmp6hdr *) skb->data;
120 bit_nr = icmph->icmp6_type;
122 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
127 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
128 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
130 static mh_filter_t __rcu *mh_filter __read_mostly;
132 int rawv6_mh_filter_register(mh_filter_t filter)
134 rcu_assign_pointer(mh_filter, filter);
137 EXPORT_SYMBOL(rawv6_mh_filter_register);
139 int rawv6_mh_filter_unregister(mh_filter_t filter)
141 RCU_INIT_POINTER(mh_filter, NULL);
145 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
150 * demultiplex raw sockets.
151 * (should consider queueing the skb in the sock receive_queue
152 * without calling rawv6.c)
154 * Caller owns SKB so we must make clones.
156 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
158 const struct in6_addr *saddr;
159 const struct in6_addr *daddr;
161 bool delivered = false;
165 saddr = &ipv6_hdr(skb)->saddr;
168 hash = nexthdr & (MAX_INET_PROTOS - 1);
170 read_lock(&raw_v6_hashinfo.lock);
171 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
176 net = dev_net(skb->dev);
177 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
185 filtered = icmpv6_filter(sk, skb);
188 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
191 /* XXX: To validate MH only once for each packet,
192 * this is placed here. It should be after checking
193 * xfrm policy, however it doesn't. The checking xfrm
194 * policy is placed in rawv6_rcv() because it is
195 * required for each socket.
199 filter = rcu_dereference(mh_filter);
200 filtered = filter ? (*filter)(sk, skb) : 0;
212 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
214 /* Not releasing hash table! */
217 rawv6_rcv(sk, clone);
220 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
224 read_unlock(&raw_v6_hashinfo.lock);
228 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
232 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
233 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
236 return raw_sk != NULL;
239 /* This cleans up af_inet6 a bit. -DaveM */
240 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
242 struct inet_sock *inet = inet_sk(sk);
243 struct ipv6_pinfo *np = inet6_sk(sk);
244 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
249 if (addr_len < SIN6_LEN_RFC2133)
251 addr_type = ipv6_addr_type(&addr->sin6_addr);
253 /* Raw sockets are IPv6 only */
254 if (addr_type == IPV6_ADDR_MAPPED)
255 return -EADDRNOTAVAIL;
260 if (sk->sk_state != TCP_CLOSE)
264 /* Check if the address belongs to the host. */
265 if (addr_type != IPV6_ADDR_ANY) {
266 struct net_device *dev = NULL;
268 if (addr_type & IPV6_ADDR_LINKLOCAL) {
269 if (addr_len >= sizeof(struct sockaddr_in6) &&
270 addr->sin6_scope_id) {
271 /* Override any existing binding, if another
272 * one is supplied by user.
274 sk->sk_bound_dev_if = addr->sin6_scope_id;
277 /* Binding to link-local address requires an interface */
278 if (!sk->sk_bound_dev_if)
282 dev = dev_get_by_index_rcu(sock_net(sk),
283 sk->sk_bound_dev_if);
288 /* ipv4 addr of the socket is invalid. Only the
289 * unspecified and mapped address have a v4 equivalent.
291 v4addr = LOOPBACK4_IPV6;
292 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
293 err = -EADDRNOTAVAIL;
294 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
301 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
302 np->rcv_saddr = addr->sin6_addr;
303 if (!(addr_type & IPV6_ADDR_MULTICAST))
304 np->saddr = addr->sin6_addr;
313 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
314 struct inet6_skb_parm *opt,
315 u8 type, u8 code, int offset, __be32 info)
317 struct inet_sock *inet = inet_sk(sk);
318 struct ipv6_pinfo *np = inet6_sk(sk);
322 /* Report error on raw socket, if:
323 1. User requested recverr.
324 2. Socket is connected (otherwise the error indication
325 is useless without recverr and error is hard.
327 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
330 harderr = icmpv6_err_convert(type, code, &err);
331 if (type == ICMPV6_PKT_TOOBIG) {
332 ip6_sk_update_pmtu(skb, sk, info);
333 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
336 u8 *payload = skb->data;
339 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
342 if (np->recverr || harderr) {
344 sk->sk_error_report(sk);
348 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
349 u8 type, u8 code, int inner_offset, __be32 info)
353 const struct in6_addr *saddr, *daddr;
356 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
358 read_lock(&raw_v6_hashinfo.lock);
359 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
361 /* Note: ipv6_hdr(skb) != skb->data */
362 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
363 saddr = &ip6h->saddr;
364 daddr = &ip6h->daddr;
365 net = dev_net(skb->dev);
367 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
369 rawv6_err(sk, skb, NULL, type, code,
374 read_unlock(&raw_v6_hashinfo.lock);
377 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
379 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
380 skb_checksum_complete(skb)) {
381 atomic_inc(&sk->sk_drops);
386 /* Charge it to the socket. */
388 if (sock_queue_rcv_skb(sk, skb) < 0) {
397 * This is next to useless...
398 * if we demultiplex in network layer we don't need the extra call
399 * just to queue the skb...
400 * maybe we could have the network decide upon a hint if it
401 * should call raw_rcv for demultiplexing
403 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
405 struct inet_sock *inet = inet_sk(sk);
406 struct raw6_sock *rp = raw6_sk(sk);
408 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
409 atomic_inc(&sk->sk_drops);
415 skb->ip_summed = CHECKSUM_UNNECESSARY;
417 if (skb->ip_summed == CHECKSUM_COMPLETE) {
418 skb_postpull_rcsum(skb, skb_network_header(skb),
419 skb_network_header_len(skb));
420 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
421 &ipv6_hdr(skb)->daddr,
422 skb->len, inet->inet_num, skb->csum))
423 skb->ip_summed = CHECKSUM_UNNECESSARY;
425 if (!skb_csum_unnecessary(skb))
426 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
427 &ipv6_hdr(skb)->daddr,
432 if (skb_checksum_complete(skb)) {
433 atomic_inc(&sk->sk_drops);
439 rawv6_rcv_skb(sk, skb);
445 * This should be easy, if there is something there
446 * we return it, otherwise we block.
449 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
450 struct msghdr *msg, size_t len,
451 int noblock, int flags, int *addr_len)
453 struct ipv6_pinfo *np = inet6_sk(sk);
454 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
463 *addr_len=sizeof(*sin6);
465 if (flags & MSG_ERRQUEUE)
466 return ipv6_recv_error(sk, msg, len);
468 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
469 return ipv6_recv_rxpmtu(sk, msg, len);
471 skb = skb_recv_datagram(sk, flags, noblock, &err);
478 msg->msg_flags |= MSG_TRUNC;
481 if (skb_csum_unnecessary(skb)) {
482 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
483 } else if (msg->msg_flags&MSG_TRUNC) {
484 if (__skb_checksum_complete(skb))
486 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
488 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
495 /* Copy the address. */
497 sin6->sin6_family = AF_INET6;
499 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
500 sin6->sin6_flowinfo = 0;
501 sin6->sin6_scope_id = 0;
502 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
503 sin6->sin6_scope_id = IP6CB(skb)->iif;
506 sock_recv_ts_and_drops(msg, sk, skb);
509 datagram_recv_ctl(sk, msg, skb);
512 if (flags & MSG_TRUNC)
516 skb_free_datagram(sk, skb);
521 skb_kill_datagram(sk, skb, flags);
523 /* Error for blocking case is chosen to masquerade
524 as some normal condition.
526 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
530 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
531 struct raw6_sock *rp)
544 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
548 total_len = inet_sk(sk)->cork.base.length;
549 if (offset >= total_len - 1) {
551 ip6_flush_pending_frames(sk);
555 /* should be check HW csum miyazawa */
556 if (skb_queue_len(&sk->sk_write_queue) == 1) {
558 * Only one fragment on the socket.
560 tmp_csum = skb->csum;
562 struct sk_buff *csum_skb = NULL;
565 skb_queue_walk(&sk->sk_write_queue, skb) {
566 tmp_csum = csum_add(tmp_csum, skb->csum);
571 len = skb->len - skb_transport_offset(skb);
583 offset += skb_transport_offset(skb);
584 if (skb_copy_bits(skb, offset, &csum, 2))
587 /* in case cksum was not initialized */
589 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
591 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
592 total_len, fl6->flowi6_proto, tmp_csum);
594 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
595 csum = CSUM_MANGLED_0;
597 if (skb_store_bits(skb, offset, &csum, 2))
601 err = ip6_push_pending_frames(sk);
606 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
607 struct flowi6 *fl6, struct dst_entry **dstp,
610 struct ipv6_pinfo *np = inet6_sk(sk);
614 struct rt6_info *rt = (struct rt6_info *)*dstp;
615 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
616 int tlen = rt->dst.dev->needed_tailroom;
618 if (length > rt->dst.dev->mtu) {
619 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
625 skb = sock_alloc_send_skb(sk,
626 length + hlen + tlen + 15,
627 flags & MSG_DONTWAIT, &err);
630 skb_reserve(skb, hlen);
632 skb->priority = sk->sk_priority;
633 skb->mark = sk->sk_mark;
634 skb_dst_set(skb, &rt->dst);
637 skb_put(skb, length);
638 skb_reset_network_header(skb);
641 skb->ip_summed = CHECKSUM_NONE;
643 skb->transport_header = skb->network_header;
644 err = memcpy_fromiovecend((void *)iph, from, 0, length);
648 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
649 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
650 rt->dst.dev, dst_output);
652 err = net_xmit_errno(err);
662 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
663 if (err == -ENOBUFS && !np->recverr)
668 static int rawv6_probe_proto_opt(struct flowi6 *fl6, struct msghdr *msg)
671 u8 __user *type = NULL;
672 u8 __user *code = NULL;
680 for (i = 0; i < msg->msg_iovlen; i++) {
681 iov = &msg->msg_iov[i];
685 switch (fl6->flowi6_proto) {
687 /* check if one-byte field is readable or not. */
688 if (iov->iov_base && iov->iov_len < 1)
692 type = iov->iov_base;
693 /* check if code field is readable or not. */
694 if (iov->iov_len > 1)
697 code = iov->iov_base;
700 if (get_user(fl6->fl6_icmp_type, type) ||
701 get_user(fl6->fl6_icmp_code, code))
707 if (iov->iov_base && iov->iov_len < 1)
709 /* check if type field is readable or not. */
710 if (iov->iov_len > 2 - len) {
711 u8 __user *p = iov->iov_base;
712 if (get_user(fl6->fl6_mh_type, &p[2 - len]))
729 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
730 struct msghdr *msg, size_t len)
732 struct ipv6_txoptions opt_space;
733 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
734 struct in6_addr *daddr, *final_p, final;
735 struct inet_sock *inet = inet_sk(sk);
736 struct ipv6_pinfo *np = inet6_sk(sk);
737 struct raw6_sock *rp = raw6_sk(sk);
738 struct ipv6_txoptions *opt = NULL;
739 struct ip6_flowlabel *flowlabel = NULL;
740 struct dst_entry *dst = NULL;
742 int addr_len = msg->msg_namelen;
749 /* Rough check on arithmetic overflow,
750 better check is made in ip6_append_data().
755 /* Mirror BSD error message compatibility */
756 if (msg->msg_flags & MSG_OOB)
760 * Get and verify the address.
762 memset(&fl6, 0, sizeof(fl6));
764 fl6.flowi6_mark = sk->sk_mark;
767 if (addr_len < SIN6_LEN_RFC2133)
770 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
771 return -EAFNOSUPPORT;
773 /* port is the proto value [0..255] carried in nexthdr */
774 proto = ntohs(sin6->sin6_port);
777 proto = inet->inet_num;
778 else if (proto != inet->inet_num)
784 daddr = &sin6->sin6_addr;
786 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
787 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
788 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
789 if (flowlabel == NULL)
791 daddr = &flowlabel->dst;
796 * Otherwise it will be difficult to maintain
799 if (sk->sk_state == TCP_ESTABLISHED &&
800 ipv6_addr_equal(daddr, &np->daddr))
803 if (addr_len >= sizeof(struct sockaddr_in6) &&
804 sin6->sin6_scope_id &&
805 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
806 fl6.flowi6_oif = sin6->sin6_scope_id;
808 if (sk->sk_state != TCP_ESTABLISHED)
809 return -EDESTADDRREQ;
811 proto = inet->inet_num;
813 fl6.flowlabel = np->flow_label;
816 if (fl6.flowi6_oif == 0)
817 fl6.flowi6_oif = sk->sk_bound_dev_if;
819 if (msg->msg_controllen) {
821 memset(opt, 0, sizeof(struct ipv6_txoptions));
822 opt->tot_len = sizeof(struct ipv6_txoptions);
824 err = datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
825 &hlimit, &tclass, &dontfrag);
827 fl6_sock_release(flowlabel);
830 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
831 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
832 if (flowlabel == NULL)
835 if (!(opt->opt_nflen|opt->opt_flen))
841 opt = fl6_merge_options(&opt_space, flowlabel, opt);
842 opt = ipv6_fixup_options(&opt_space, opt);
844 fl6.flowi6_proto = proto;
845 err = rawv6_probe_proto_opt(&fl6, msg);
849 if (!ipv6_addr_any(daddr))
852 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
853 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
854 fl6.saddr = np->saddr;
856 final_p = fl6_update_dst(&fl6, opt, &final);
858 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
859 fl6.flowi6_oif = np->mcast_oif;
860 else if (!fl6.flowi6_oif)
861 fl6.flowi6_oif = np->ucast_oif;
862 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
864 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
870 if (ipv6_addr_is_multicast(&fl6.daddr))
871 hlimit = np->mcast_hops;
873 hlimit = np->hop_limit;
875 hlimit = ip6_dst_hoplimit(dst);
882 dontfrag = np->dontfrag;
884 if (msg->msg_flags&MSG_CONFIRM)
889 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl6, &dst, msg->msg_flags);
892 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
893 len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst,
894 msg->msg_flags, dontfrag);
897 ip6_flush_pending_frames(sk);
898 else if (!(msg->msg_flags & MSG_MORE))
899 err = rawv6_push_pending_frames(sk, &fl6, rp);
905 fl6_sock_release(flowlabel);
906 return err<0?err:len;
909 if (!(msg->msg_flags & MSG_PROBE) || len)
910 goto back_from_confirm;
915 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
916 char __user *optval, int optlen)
920 if (optlen > sizeof(struct icmp6_filter))
921 optlen = sizeof(struct icmp6_filter);
922 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
932 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
933 char __user *optval, int __user *optlen)
939 if (get_user(len, optlen))
943 if (len > sizeof(struct icmp6_filter))
944 len = sizeof(struct icmp6_filter);
945 if (put_user(len, optlen))
947 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
958 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
959 char __user *optval, unsigned int optlen)
961 struct raw6_sock *rp = raw6_sk(sk);
964 if (get_user(val, (int __user *)optval))
969 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
970 level == IPPROTO_IPV6) {
972 * RFC3542 tells that IPV6_CHECKSUM socket
973 * option in the IPPROTO_IPV6 level is not
974 * allowed on ICMPv6 sockets.
975 * If you want to set it, use IPPROTO_RAW
976 * level IPV6_CHECKSUM socket option
982 /* You may get strange result with a positive odd offset;
983 RFC2292bis agrees with me. */
984 if (val > 0 && (val&1))
1000 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1001 char __user *optval, unsigned int optlen)
1008 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1010 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1012 if (optname == IPV6_CHECKSUM)
1015 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1018 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1021 #ifdef CONFIG_COMPAT
1022 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1023 char __user *optval, unsigned int optlen)
1029 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1031 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1033 if (optname == IPV6_CHECKSUM)
1036 return compat_ipv6_setsockopt(sk, level, optname,
1039 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1043 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1044 char __user *optval, int __user *optlen)
1046 struct raw6_sock *rp = raw6_sk(sk);
1049 if (get_user(len,optlen))
1055 * We allow getsockopt() for IPPROTO_IPV6-level
1056 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1057 * since RFC3542 is silent about it.
1059 if (rp->checksum == 0)
1066 return -ENOPROTOOPT;
1069 len = min_t(unsigned int, sizeof(int), len);
1071 if (put_user(len, optlen))
1073 if (copy_to_user(optval,&val,len))
1078 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1079 char __user *optval, int __user *optlen)
1086 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1088 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1090 if (optname == IPV6_CHECKSUM)
1093 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1096 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1099 #ifdef CONFIG_COMPAT
1100 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1101 char __user *optval, int __user *optlen)
1107 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1109 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1111 if (optname == IPV6_CHECKSUM)
1114 return compat_ipv6_getsockopt(sk, level, optname,
1117 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1121 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1125 int amount = sk_wmem_alloc_get(sk);
1127 return put_user(amount, (int __user *)arg);
1130 struct sk_buff *skb;
1133 spin_lock_bh(&sk->sk_receive_queue.lock);
1134 skb = skb_peek(&sk->sk_receive_queue);
1136 amount = skb->tail - skb->transport_header;
1137 spin_unlock_bh(&sk->sk_receive_queue.lock);
1138 return put_user(amount, (int __user *)arg);
1142 #ifdef CONFIG_IPV6_MROUTE
1143 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1145 return -ENOIOCTLCMD;
1150 #ifdef CONFIG_COMPAT
1151 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1156 return -ENOIOCTLCMD;
1158 #ifdef CONFIG_IPV6_MROUTE
1159 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1161 return -ENOIOCTLCMD;
1167 static void rawv6_close(struct sock *sk, long timeout)
1169 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1170 ip6_ra_control(sk, -1);
1172 sk_common_release(sk);
1175 static void raw6_destroy(struct sock *sk)
1178 ip6_flush_pending_frames(sk);
1181 inet6_destroy_sock(sk);
1184 static int rawv6_init_sk(struct sock *sk)
1186 struct raw6_sock *rp = raw6_sk(sk);
1188 switch (inet_sk(sk)->inet_num) {
1189 case IPPROTO_ICMPV6:
1203 struct proto rawv6_prot = {
1205 .owner = THIS_MODULE,
1206 .close = rawv6_close,
1207 .destroy = raw6_destroy,
1208 .connect = ip6_datagram_connect,
1209 .disconnect = udp_disconnect,
1210 .ioctl = rawv6_ioctl,
1211 .init = rawv6_init_sk,
1212 .setsockopt = rawv6_setsockopt,
1213 .getsockopt = rawv6_getsockopt,
1214 .sendmsg = rawv6_sendmsg,
1215 .recvmsg = rawv6_recvmsg,
1217 .backlog_rcv = rawv6_rcv_skb,
1218 .hash = raw_hash_sk,
1219 .unhash = raw_unhash_sk,
1220 .obj_size = sizeof(struct raw6_sock),
1221 .h.raw_hash = &raw_v6_hashinfo,
1222 #ifdef CONFIG_COMPAT
1223 .compat_setsockopt = compat_rawv6_setsockopt,
1224 .compat_getsockopt = compat_rawv6_getsockopt,
1225 .compat_ioctl = compat_rawv6_ioctl,
1229 #ifdef CONFIG_PROC_FS
1230 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1232 struct ipv6_pinfo *np = inet6_sk(sp);
1233 const struct in6_addr *dest, *src;
1237 src = &np->rcv_saddr;
1239 srcp = inet_sk(sp)->inet_num;
1241 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1242 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1244 src->s6_addr32[0], src->s6_addr32[1],
1245 src->s6_addr32[2], src->s6_addr32[3], srcp,
1246 dest->s6_addr32[0], dest->s6_addr32[1],
1247 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1249 sk_wmem_alloc_get(sp),
1250 sk_rmem_alloc_get(sp),
1254 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1257 static int raw6_seq_show(struct seq_file *seq, void *v)
1259 if (v == SEQ_START_TOKEN)
1264 "st tx_queue rx_queue tr tm->when retrnsmt"
1265 " uid timeout inode ref pointer drops\n");
1267 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1271 static const struct seq_operations raw6_seq_ops = {
1272 .start = raw_seq_start,
1273 .next = raw_seq_next,
1274 .stop = raw_seq_stop,
1275 .show = raw6_seq_show,
1278 static int raw6_seq_open(struct inode *inode, struct file *file)
1280 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1283 static const struct file_operations raw6_seq_fops = {
1284 .owner = THIS_MODULE,
1285 .open = raw6_seq_open,
1287 .llseek = seq_lseek,
1288 .release = seq_release_net,
1291 static int __net_init raw6_init_net(struct net *net)
1293 if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1299 static void __net_exit raw6_exit_net(struct net *net)
1301 proc_net_remove(net, "raw6");
1304 static struct pernet_operations raw6_net_ops = {
1305 .init = raw6_init_net,
1306 .exit = raw6_exit_net,
1309 int __init raw6_proc_init(void)
1311 return register_pernet_subsys(&raw6_net_ops);
1314 void raw6_proc_exit(void)
1316 unregister_pernet_subsys(&raw6_net_ops);
1318 #endif /* CONFIG_PROC_FS */
1320 /* Same as inet6_dgram_ops, sans udp_poll. */
1321 static const struct proto_ops inet6_sockraw_ops = {
1323 .owner = THIS_MODULE,
1324 .release = inet6_release,
1326 .connect = inet_dgram_connect, /* ok */
1327 .socketpair = sock_no_socketpair, /* a do nothing */
1328 .accept = sock_no_accept, /* a do nothing */
1329 .getname = inet6_getname,
1330 .poll = datagram_poll, /* ok */
1331 .ioctl = inet6_ioctl, /* must change */
1332 .listen = sock_no_listen, /* ok */
1333 .shutdown = inet_shutdown, /* ok */
1334 .setsockopt = sock_common_setsockopt, /* ok */
1335 .getsockopt = sock_common_getsockopt, /* ok */
1336 .sendmsg = inet_sendmsg, /* ok */
1337 .recvmsg = sock_common_recvmsg, /* ok */
1338 .mmap = sock_no_mmap,
1339 .sendpage = sock_no_sendpage,
1340 #ifdef CONFIG_COMPAT
1341 .compat_setsockopt = compat_sock_common_setsockopt,
1342 .compat_getsockopt = compat_sock_common_getsockopt,
1346 static struct inet_protosw rawv6_protosw = {
1348 .protocol = IPPROTO_IP, /* wild card */
1349 .prot = &rawv6_prot,
1350 .ops = &inet6_sockraw_ops,
1351 .no_check = UDP_CSUM_DEFAULT,
1352 .flags = INET_PROTOSW_REUSE,
1355 int __init rawv6_init(void)
1359 ret = inet6_register_protosw(&rawv6_protosw);
1366 void rawv6_exit(void)
1368 inet6_unregister_protosw(&rawv6_protosw);