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 <linux/compat.h>
35 #include <linux/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 IS_ENABLED(CONFIG_IPV6_MIP6)
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 #define ICMPV6_HDRLEN 4 /* ICMPv6 header, RFC 4443 Section 2.1 */
68 struct raw_hashinfo raw_v6_hashinfo = {
69 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
71 EXPORT_SYMBOL_GPL(raw_v6_hashinfo);
73 struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
74 unsigned short num, const struct in6_addr *loc_addr,
75 const struct in6_addr *rmt_addr, int dif)
77 bool is_multicast = ipv6_addr_is_multicast(loc_addr);
80 if (inet_sk(sk)->inet_num == num) {
82 if (!net_eq(sock_net(sk), net))
85 if (!ipv6_addr_any(&sk->sk_v6_daddr) &&
86 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr))
89 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
92 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
93 if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))
96 inet6_mc_check(sk, loc_addr, rmt_addr))
106 EXPORT_SYMBOL_GPL(__raw_v6_lookup);
112 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
114 struct icmp6hdr _hdr;
115 const struct icmp6hdr *hdr;
117 /* We require only the four bytes of the ICMPv6 header, not any
118 * additional bytes of message body in "struct icmp6hdr".
120 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
121 ICMPV6_HDRLEN, &_hdr);
123 const __u32 *data = &raw6_sk(sk)->filter.data[0];
124 unsigned int type = hdr->icmp6_type;
126 return (data[type >> 5] & (1U << (type & 31))) != 0;
131 #if IS_ENABLED(CONFIG_IPV6_MIP6)
132 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
134 static mh_filter_t __rcu *mh_filter __read_mostly;
136 int rawv6_mh_filter_register(mh_filter_t filter)
138 rcu_assign_pointer(mh_filter, filter);
141 EXPORT_SYMBOL(rawv6_mh_filter_register);
143 int rawv6_mh_filter_unregister(mh_filter_t filter)
145 RCU_INIT_POINTER(mh_filter, NULL);
149 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
154 * demultiplex raw sockets.
155 * (should consider queueing the skb in the sock receive_queue
156 * without calling rawv6.c)
158 * Caller owns SKB so we must make clones.
160 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
162 const struct in6_addr *saddr;
163 const struct in6_addr *daddr;
165 bool delivered = false;
169 saddr = &ipv6_hdr(skb)->saddr;
172 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
174 read_lock(&raw_v6_hashinfo.lock);
175 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
180 net = dev_net(skb->dev);
181 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, inet6_iif(skb));
189 filtered = icmpv6_filter(sk, skb);
192 #if IS_ENABLED(CONFIG_IPV6_MIP6)
195 /* XXX: To validate MH only once for each packet,
196 * this is placed here. It should be after checking
197 * xfrm policy, however it doesn't. The checking xfrm
198 * policy is placed in rawv6_rcv() because it is
199 * required for each socket.
203 filter = rcu_dereference(mh_filter);
204 filtered = filter ? (*filter)(sk, skb) : 0;
216 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
218 /* Not releasing hash table! */
221 rawv6_rcv(sk, clone);
224 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
228 read_unlock(&raw_v6_hashinfo.lock);
232 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
236 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]);
237 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
240 return raw_sk != NULL;
243 /* This cleans up af_inet6 a bit. -DaveM */
244 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
246 struct inet_sock *inet = inet_sk(sk);
247 struct ipv6_pinfo *np = inet6_sk(sk);
248 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
253 if (addr_len < SIN6_LEN_RFC2133)
256 if (addr->sin6_family != AF_INET6)
259 addr_type = ipv6_addr_type(&addr->sin6_addr);
261 /* Raw sockets are IPv6 only */
262 if (addr_type == IPV6_ADDR_MAPPED)
263 return -EADDRNOTAVAIL;
268 if (sk->sk_state != TCP_CLOSE)
272 /* Check if the address belongs to the host. */
273 if (addr_type != IPV6_ADDR_ANY) {
274 struct net_device *dev = NULL;
276 if (__ipv6_addr_needs_scope_id(addr_type)) {
277 if (addr_len >= sizeof(struct sockaddr_in6) &&
278 addr->sin6_scope_id) {
279 /* Override any existing binding, if another
280 * one is supplied by user.
282 sk->sk_bound_dev_if = addr->sin6_scope_id;
285 /* Binding to link-local address requires an interface */
286 if (!sk->sk_bound_dev_if)
290 dev = dev_get_by_index_rcu(sock_net(sk),
291 sk->sk_bound_dev_if);
296 /* ipv4 addr of the socket is invalid. Only the
297 * unspecified and mapped address have a v4 equivalent.
299 v4addr = LOOPBACK4_IPV6;
300 if (!(addr_type & IPV6_ADDR_MULTICAST) &&
301 !sock_net(sk)->ipv6.sysctl.ip_nonlocal_bind) {
302 err = -EADDRNOTAVAIL;
303 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
310 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
311 sk->sk_v6_rcv_saddr = addr->sin6_addr;
312 if (!(addr_type & IPV6_ADDR_MULTICAST))
313 np->saddr = addr->sin6_addr;
322 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
323 struct inet6_skb_parm *opt,
324 u8 type, u8 code, int offset, __be32 info)
326 struct inet_sock *inet = inet_sk(sk);
327 struct ipv6_pinfo *np = inet6_sk(sk);
331 /* Report error on raw socket, if:
332 1. User requested recverr.
333 2. Socket is connected (otherwise the error indication
334 is useless without recverr and error is hard.
336 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
339 harderr = icmpv6_err_convert(type, code, &err);
340 if (type == ICMPV6_PKT_TOOBIG) {
341 ip6_sk_update_pmtu(skb, sk, info);
342 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
344 if (type == NDISC_REDIRECT) {
345 ip6_sk_redirect(skb, sk);
349 u8 *payload = skb->data;
352 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
355 if (np->recverr || harderr) {
357 sk->sk_error_report(sk);
361 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
362 u8 type, u8 code, int inner_offset, __be32 info)
366 const struct in6_addr *saddr, *daddr;
369 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
371 read_lock(&raw_v6_hashinfo.lock);
372 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
374 /* Note: ipv6_hdr(skb) != skb->data */
375 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
376 saddr = &ip6h->saddr;
377 daddr = &ip6h->daddr;
378 net = dev_net(skb->dev);
380 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
382 rawv6_err(sk, skb, NULL, type, code,
387 read_unlock(&raw_v6_hashinfo.lock);
390 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
392 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
393 skb_checksum_complete(skb)) {
394 atomic_inc(&sk->sk_drops);
399 /* Charge it to the socket. */
401 if (sock_queue_rcv_skb(sk, skb) < 0) {
410 * This is next to useless...
411 * if we demultiplex in network layer we don't need the extra call
412 * just to queue the skb...
413 * maybe we could have the network decide upon a hint if it
414 * should call raw_rcv for demultiplexing
416 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
418 struct inet_sock *inet = inet_sk(sk);
419 struct raw6_sock *rp = raw6_sk(sk);
421 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
422 atomic_inc(&sk->sk_drops);
428 skb->ip_summed = CHECKSUM_UNNECESSARY;
430 if (skb->ip_summed == CHECKSUM_COMPLETE) {
431 skb_postpull_rcsum(skb, skb_network_header(skb),
432 skb_network_header_len(skb));
433 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
434 &ipv6_hdr(skb)->daddr,
435 skb->len, inet->inet_num, skb->csum))
436 skb->ip_summed = CHECKSUM_UNNECESSARY;
438 if (!skb_csum_unnecessary(skb))
439 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
440 &ipv6_hdr(skb)->daddr,
445 if (skb_checksum_complete(skb)) {
446 atomic_inc(&sk->sk_drops);
452 rawv6_rcv_skb(sk, skb);
458 * This should be easy, if there is something there
459 * we return it, otherwise we block.
462 static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
463 int noblock, int flags, int *addr_len)
465 struct ipv6_pinfo *np = inet6_sk(sk);
466 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
474 if (flags & MSG_ERRQUEUE)
475 return ipv6_recv_error(sk, msg, len, addr_len);
477 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
478 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
480 skb = skb_recv_datagram(sk, flags, noblock, &err);
487 msg->msg_flags |= MSG_TRUNC;
490 if (skb_csum_unnecessary(skb)) {
491 err = skb_copy_datagram_msg(skb, 0, msg, copied);
492 } else if (msg->msg_flags&MSG_TRUNC) {
493 if (__skb_checksum_complete(skb))
495 err = skb_copy_datagram_msg(skb, 0, msg, copied);
497 err = skb_copy_and_csum_datagram_msg(skb, 0, msg);
504 /* Copy the address. */
506 sin6->sin6_family = AF_INET6;
508 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
509 sin6->sin6_flowinfo = 0;
510 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
512 *addr_len = sizeof(*sin6);
515 sock_recv_ts_and_drops(msg, sk, skb);
518 ip6_datagram_recv_ctl(sk, msg, skb);
521 if (flags & MSG_TRUNC)
525 skb_free_datagram(sk, skb);
530 skb_kill_datagram(sk, skb, flags);
532 /* Error for blocking case is chosen to masquerade
533 as some normal condition.
535 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
539 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
540 struct raw6_sock *rp)
553 skb = skb_peek(&sk->sk_write_queue);
558 total_len = inet_sk(sk)->cork.base.length;
559 if (offset >= total_len - 1) {
561 ip6_flush_pending_frames(sk);
565 /* should be check HW csum miyazawa */
566 if (skb_queue_len(&sk->sk_write_queue) == 1) {
568 * Only one fragment on the socket.
570 tmp_csum = skb->csum;
572 struct sk_buff *csum_skb = NULL;
575 skb_queue_walk(&sk->sk_write_queue, skb) {
576 tmp_csum = csum_add(tmp_csum, skb->csum);
581 len = skb->len - skb_transport_offset(skb);
593 offset += skb_transport_offset(skb);
594 err = skb_copy_bits(skb, offset, &csum, 2);
596 ip6_flush_pending_frames(sk);
600 /* in case cksum was not initialized */
602 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
604 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
605 total_len, fl6->flowi6_proto, tmp_csum);
607 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
608 csum = CSUM_MANGLED_0;
610 BUG_ON(skb_store_bits(skb, offset, &csum, 2));
613 err = ip6_push_pending_frames(sk);
618 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length,
619 struct flowi6 *fl6, struct dst_entry **dstp,
622 struct ipv6_pinfo *np = inet6_sk(sk);
623 struct net *net = sock_net(sk);
627 struct rt6_info *rt = (struct rt6_info *)*dstp;
628 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
629 int tlen = rt->dst.dev->needed_tailroom;
631 if (length > rt->dst.dev->mtu) {
632 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
638 skb = sock_alloc_send_skb(sk,
639 length + hlen + tlen + 15,
640 flags & MSG_DONTWAIT, &err);
643 skb_reserve(skb, hlen);
645 skb->protocol = htons(ETH_P_IPV6);
646 skb->priority = sk->sk_priority;
647 skb->mark = sk->sk_mark;
648 skb_dst_set(skb, &rt->dst);
651 skb_put(skb, length);
652 skb_reset_network_header(skb);
655 skb->ip_summed = CHECKSUM_NONE;
657 if (flags & MSG_CONFIRM)
658 skb_set_dst_pending_confirm(skb, 1);
660 skb->transport_header = skb->network_header;
661 err = memcpy_from_msg(iph, msg, length);
665 /* if egress device is enslaved to an L3 master device pass the
666 * skb to its handler for processing
668 skb = l3mdev_ip6_out(sk, skb);
672 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
673 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
674 NULL, rt->dst.dev, dst_output);
676 err = net_xmit_errno(err);
686 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
687 if (err == -ENOBUFS && !np->recverr)
692 struct raw6_frag_vec {
698 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6)
701 switch (fl6->flowi6_proto) {
704 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
706 fl6->fl6_icmp_type = rfv->c[0];
707 fl6->fl6_icmp_code = rfv->c[1];
712 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
714 fl6->fl6_mh_type = rfv->c[2];
719 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd,
722 struct raw6_frag_vec *rfv = from;
724 if (offset < rfv->hlen) {
725 int copy = min(rfv->hlen - offset, len);
727 if (skb->ip_summed == CHECKSUM_PARTIAL)
728 memcpy(to, rfv->c + offset, copy);
730 skb->csum = csum_block_add(
732 csum_partial_copy_nocheck(rfv->c + offset,
747 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
750 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
752 struct ipv6_txoptions *opt_to_free = NULL;
753 struct ipv6_txoptions opt_space;
754 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
755 struct in6_addr *daddr, *final_p, final;
756 struct inet_sock *inet = inet_sk(sk);
757 struct ipv6_pinfo *np = inet6_sk(sk);
758 struct raw6_sock *rp = raw6_sk(sk);
759 struct ipv6_txoptions *opt = NULL;
760 struct ip6_flowlabel *flowlabel = NULL;
761 struct dst_entry *dst = NULL;
762 struct raw6_frag_vec rfv;
764 struct sockcm_cookie sockc;
765 struct ipcm6_cookie ipc6;
766 int addr_len = msg->msg_namelen;
770 /* Rough check on arithmetic overflow,
771 better check is made in ip6_append_data().
776 /* Mirror BSD error message compatibility */
777 if (msg->msg_flags & MSG_OOB)
781 * Get and verify the address.
783 memset(&fl6, 0, sizeof(fl6));
785 fl6.flowi6_mark = sk->sk_mark;
786 fl6.flowi6_uid = sk->sk_uid;
794 if (addr_len < SIN6_LEN_RFC2133)
797 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
798 return -EAFNOSUPPORT;
800 /* port is the proto value [0..255] carried in nexthdr */
801 proto = ntohs(sin6->sin6_port);
804 proto = inet->inet_num;
805 else if (proto != inet->inet_num)
811 daddr = &sin6->sin6_addr;
813 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
814 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
815 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
822 * Otherwise it will be difficult to maintain
825 if (sk->sk_state == TCP_ESTABLISHED &&
826 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
827 daddr = &sk->sk_v6_daddr;
829 if (addr_len >= sizeof(struct sockaddr_in6) &&
830 sin6->sin6_scope_id &&
831 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
832 fl6.flowi6_oif = sin6->sin6_scope_id;
834 if (sk->sk_state != TCP_ESTABLISHED)
835 return -EDESTADDRREQ;
837 proto = inet->inet_num;
838 daddr = &sk->sk_v6_daddr;
839 fl6.flowlabel = np->flow_label;
842 if (fl6.flowi6_oif == 0)
843 fl6.flowi6_oif = sk->sk_bound_dev_if;
845 sockc.tsflags = sk->sk_tsflags;
846 if (msg->msg_controllen) {
848 memset(opt, 0, sizeof(struct ipv6_txoptions));
849 opt->tot_len = sizeof(struct ipv6_txoptions);
852 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6, &sockc);
854 fl6_sock_release(flowlabel);
857 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
858 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
862 if (!(opt->opt_nflen|opt->opt_flen))
870 opt = fl6_merge_options(&opt_space, flowlabel, opt);
871 opt = ipv6_fixup_options(&opt_space, opt);
873 fl6.flowi6_proto = proto;
876 err = rawv6_probe_proto_opt(&rfv, &fl6);
880 if (!ipv6_addr_any(daddr))
883 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
884 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
885 fl6.saddr = np->saddr;
887 final_p = fl6_update_dst(&fl6, opt, &final);
889 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
890 fl6.flowi6_oif = np->mcast_oif;
891 else if (!fl6.flowi6_oif)
892 fl6.flowi6_oif = np->ucast_oif;
893 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
896 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH;
899 ipc6.tclass = np->tclass;
901 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
903 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
909 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
911 if (ipc6.dontfrag < 0)
912 ipc6.dontfrag = np->dontfrag;
914 if (msg->msg_flags&MSG_CONFIRM)
919 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst, msg->msg_flags);
923 err = ip6_append_data(sk, raw6_getfrag, &rfv,
924 len, 0, &ipc6, &fl6, (struct rt6_info *)dst,
925 msg->msg_flags, &sockc);
928 ip6_flush_pending_frames(sk);
929 else if (!(msg->msg_flags & MSG_MORE))
930 err = rawv6_push_pending_frames(sk, &fl6, rp);
936 fl6_sock_release(flowlabel);
937 txopt_put(opt_to_free);
938 return err < 0 ? err : len;
940 if (msg->msg_flags & MSG_PROBE)
941 dst_confirm_neigh(dst, &fl6.daddr);
942 if (!(msg->msg_flags & MSG_PROBE) || len)
943 goto back_from_confirm;
948 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
949 char __user *optval, int optlen)
953 if (optlen > sizeof(struct icmp6_filter))
954 optlen = sizeof(struct icmp6_filter);
955 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
965 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
966 char __user *optval, int __user *optlen)
972 if (get_user(len, optlen))
976 if (len > sizeof(struct icmp6_filter))
977 len = sizeof(struct icmp6_filter);
978 if (put_user(len, optlen))
980 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
991 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
992 char __user *optval, unsigned int optlen)
994 struct raw6_sock *rp = raw6_sk(sk);
997 if (get_user(val, (int __user *)optval))
1002 if (sk->sk_type != SOCK_RAW)
1004 inet_sk(sk)->hdrincl = !!val;
1007 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
1008 level == IPPROTO_IPV6) {
1010 * RFC3542 tells that IPV6_CHECKSUM socket
1011 * option in the IPPROTO_IPV6 level is not
1012 * allowed on ICMPv6 sockets.
1013 * If you want to set it, use IPPROTO_RAW
1014 * level IPV6_CHECKSUM socket option
1015 * (Linux extension).
1020 /* You may get strange result with a positive odd offset;
1021 RFC2292bis agrees with me. */
1022 if (val > 0 && (val&1))
1034 return -ENOPROTOOPT;
1038 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1039 char __user *optval, unsigned int optlen)
1046 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1048 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1050 if (optname == IPV6_CHECKSUM ||
1051 optname == IPV6_HDRINCL)
1054 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1057 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1060 #ifdef CONFIG_COMPAT
1061 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1062 char __user *optval, unsigned int optlen)
1068 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1070 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1072 if (optname == IPV6_CHECKSUM ||
1073 optname == IPV6_HDRINCL)
1076 return compat_ipv6_setsockopt(sk, level, optname,
1079 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1083 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1084 char __user *optval, int __user *optlen)
1086 struct raw6_sock *rp = raw6_sk(sk);
1089 if (get_user(len, optlen))
1094 val = inet_sk(sk)->hdrincl;
1098 * We allow getsockopt() for IPPROTO_IPV6-level
1099 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1100 * since RFC3542 is silent about it.
1102 if (rp->checksum == 0)
1109 return -ENOPROTOOPT;
1112 len = min_t(unsigned int, sizeof(int), len);
1114 if (put_user(len, optlen))
1116 if (copy_to_user(optval, &val, len))
1121 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1122 char __user *optval, int __user *optlen)
1129 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1131 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1133 if (optname == IPV6_CHECKSUM ||
1134 optname == IPV6_HDRINCL)
1137 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1140 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1143 #ifdef CONFIG_COMPAT
1144 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1145 char __user *optval, int __user *optlen)
1151 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1153 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1155 if (optname == IPV6_CHECKSUM ||
1156 optname == IPV6_HDRINCL)
1159 return compat_ipv6_getsockopt(sk, level, optname,
1162 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1166 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1170 int amount = sk_wmem_alloc_get(sk);
1172 return put_user(amount, (int __user *)arg);
1175 struct sk_buff *skb;
1178 spin_lock_bh(&sk->sk_receive_queue.lock);
1179 skb = skb_peek(&sk->sk_receive_queue);
1181 amount = skb_tail_pointer(skb) -
1182 skb_transport_header(skb);
1183 spin_unlock_bh(&sk->sk_receive_queue.lock);
1184 return put_user(amount, (int __user *)arg);
1188 #ifdef CONFIG_IPV6_MROUTE
1189 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1191 return -ENOIOCTLCMD;
1196 #ifdef CONFIG_COMPAT
1197 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1202 return -ENOIOCTLCMD;
1204 #ifdef CONFIG_IPV6_MROUTE
1205 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1207 return -ENOIOCTLCMD;
1213 static void rawv6_close(struct sock *sk, long timeout)
1215 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1216 ip6_ra_control(sk, -1);
1218 sk_common_release(sk);
1221 static void raw6_destroy(struct sock *sk)
1224 ip6_flush_pending_frames(sk);
1227 inet6_destroy_sock(sk);
1230 static int rawv6_init_sk(struct sock *sk)
1232 struct raw6_sock *rp = raw6_sk(sk);
1234 switch (inet_sk(sk)->inet_num) {
1235 case IPPROTO_ICMPV6:
1249 struct proto rawv6_prot = {
1251 .owner = THIS_MODULE,
1252 .close = rawv6_close,
1253 .destroy = raw6_destroy,
1254 .connect = ip6_datagram_connect_v6_only,
1255 .disconnect = __udp_disconnect,
1256 .ioctl = rawv6_ioctl,
1257 .init = rawv6_init_sk,
1258 .setsockopt = rawv6_setsockopt,
1259 .getsockopt = rawv6_getsockopt,
1260 .sendmsg = rawv6_sendmsg,
1261 .recvmsg = rawv6_recvmsg,
1263 .backlog_rcv = rawv6_rcv_skb,
1264 .hash = raw_hash_sk,
1265 .unhash = raw_unhash_sk,
1266 .obj_size = sizeof(struct raw6_sock),
1267 .h.raw_hash = &raw_v6_hashinfo,
1268 #ifdef CONFIG_COMPAT
1269 .compat_setsockopt = compat_rawv6_setsockopt,
1270 .compat_getsockopt = compat_rawv6_getsockopt,
1271 .compat_ioctl = compat_rawv6_ioctl,
1273 .diag_destroy = raw_abort,
1276 #ifdef CONFIG_PROC_FS
1277 static int raw6_seq_show(struct seq_file *seq, void *v)
1279 if (v == SEQ_START_TOKEN) {
1280 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1282 struct sock *sp = v;
1283 __u16 srcp = inet_sk(sp)->inet_num;
1284 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1285 raw_seq_private(seq)->bucket);
1290 static const struct seq_operations raw6_seq_ops = {
1291 .start = raw_seq_start,
1292 .next = raw_seq_next,
1293 .stop = raw_seq_stop,
1294 .show = raw6_seq_show,
1297 static int raw6_seq_open(struct inode *inode, struct file *file)
1299 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1302 static const struct file_operations raw6_seq_fops = {
1303 .owner = THIS_MODULE,
1304 .open = raw6_seq_open,
1306 .llseek = seq_lseek,
1307 .release = seq_release_net,
1310 static int __net_init raw6_init_net(struct net *net)
1312 if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops))
1318 static void __net_exit raw6_exit_net(struct net *net)
1320 remove_proc_entry("raw6", net->proc_net);
1323 static struct pernet_operations raw6_net_ops = {
1324 .init = raw6_init_net,
1325 .exit = raw6_exit_net,
1328 int __init raw6_proc_init(void)
1330 return register_pernet_subsys(&raw6_net_ops);
1333 void raw6_proc_exit(void)
1335 unregister_pernet_subsys(&raw6_net_ops);
1337 #endif /* CONFIG_PROC_FS */
1339 /* Same as inet6_dgram_ops, sans udp_poll. */
1340 static const struct proto_ops inet6_sockraw_ops = {
1342 .owner = THIS_MODULE,
1343 .release = inet6_release,
1345 .connect = inet_dgram_connect, /* ok */
1346 .socketpair = sock_no_socketpair, /* a do nothing */
1347 .accept = sock_no_accept, /* a do nothing */
1348 .getname = inet6_getname,
1349 .poll = datagram_poll, /* ok */
1350 .ioctl = inet6_ioctl, /* must change */
1351 .listen = sock_no_listen, /* ok */
1352 .shutdown = inet_shutdown, /* ok */
1353 .setsockopt = sock_common_setsockopt, /* ok */
1354 .getsockopt = sock_common_getsockopt, /* ok */
1355 .sendmsg = inet_sendmsg, /* ok */
1356 .recvmsg = sock_common_recvmsg, /* ok */
1357 .mmap = sock_no_mmap,
1358 .sendpage = sock_no_sendpage,
1359 #ifdef CONFIG_COMPAT
1360 .compat_setsockopt = compat_sock_common_setsockopt,
1361 .compat_getsockopt = compat_sock_common_getsockopt,
1365 static struct inet_protosw rawv6_protosw = {
1367 .protocol = IPPROTO_IP, /* wild card */
1368 .prot = &rawv6_prot,
1369 .ops = &inet6_sockraw_ops,
1370 .flags = INET_PROTOSW_REUSE,
1373 int __init rawv6_init(void)
1375 return inet6_register_protosw(&rawv6_protosw);
1378 void rawv6_exit(void)
1380 inet6_unregister_protosw(&rawv6_protosw);