1 // SPDX-License-Identifier: GPL-2.0-or-later
4 * Linux INET6 implementation
9 * Adapted from linux/net/ipv4/raw.c
12 * Hideaki YOSHIFUJI : sin6_scope_id support
13 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
14 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
17 #include <linux/errno.h>
18 #include <linux/types.h>
19 #include <linux/socket.h>
20 #include <linux/slab.h>
21 #include <linux/sockios.h>
22 #include <linux/net.h>
23 #include <linux/in6.h>
24 #include <linux/netdevice.h>
25 #include <linux/if_arp.h>
26 #include <linux/icmpv6.h>
27 #include <linux/netfilter.h>
28 #include <linux/netfilter_ipv6.h>
29 #include <linux/skbuff.h>
30 #include <linux/compat.h>
31 #include <linux/uaccess.h>
32 #include <asm/ioctls.h>
34 #include <net/net_namespace.h>
40 #include <net/ndisc.h>
41 #include <net/protocol.h>
42 #include <net/ip6_route.h>
43 #include <net/ip6_checksum.h>
44 #include <net/addrconf.h>
45 #include <net/transp_v6.h>
47 #include <net/inet_common.h>
48 #include <net/tcp_states.h>
49 #if IS_ENABLED(CONFIG_IPV6_MIP6)
52 #include <linux/mroute6.h>
55 #include <net/rawv6.h>
58 #include <linux/proc_fs.h>
59 #include <linux/seq_file.h>
60 #include <linux/export.h>
62 #define ICMPV6_HDRLEN 4 /* ICMPv6 header, RFC 4443 Section 2.1 */
64 struct raw_hashinfo raw_v6_hashinfo;
65 EXPORT_SYMBOL_GPL(raw_v6_hashinfo);
67 bool raw_v6_match(struct net *net, struct sock *sk, unsigned short num,
68 const struct in6_addr *loc_addr,
69 const struct in6_addr *rmt_addr, int dif, int sdif)
71 if (inet_sk(sk)->inet_num != num ||
72 !net_eq(sock_net(sk), net) ||
73 (!ipv6_addr_any(&sk->sk_v6_daddr) &&
74 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
75 !raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if,
79 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) ||
80 ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr) ||
81 (ipv6_addr_is_multicast(loc_addr) &&
82 inet6_mc_check(sk, loc_addr, rmt_addr)))
87 EXPORT_SYMBOL_GPL(raw_v6_match);
93 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
96 const struct icmp6hdr *hdr;
98 /* We require only the four bytes of the ICMPv6 header, not any
99 * additional bytes of message body in "struct icmp6hdr".
101 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
102 ICMPV6_HDRLEN, &_hdr);
104 const __u32 *data = &raw6_sk(sk)->filter.data[0];
105 unsigned int type = hdr->icmp6_type;
107 return (data[type >> 5] & (1U << (type & 31))) != 0;
112 #if IS_ENABLED(CONFIG_IPV6_MIP6)
113 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
115 static mh_filter_t __rcu *mh_filter __read_mostly;
117 int rawv6_mh_filter_register(mh_filter_t filter)
119 rcu_assign_pointer(mh_filter, filter);
122 EXPORT_SYMBOL(rawv6_mh_filter_register);
124 int rawv6_mh_filter_unregister(mh_filter_t filter)
126 RCU_INIT_POINTER(mh_filter, NULL);
130 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
135 * demultiplex raw sockets.
136 * (should consider queueing the skb in the sock receive_queue
137 * without calling rawv6.c)
139 * Caller owns SKB so we must make clones.
141 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
143 struct net *net = dev_net(skb->dev);
144 struct hlist_nulls_head *hlist;
145 struct hlist_nulls_node *hnode;
146 const struct in6_addr *saddr;
147 const struct in6_addr *daddr;
149 bool delivered = false;
152 saddr = &ipv6_hdr(skb)->saddr;
155 hash = raw_hashfunc(net, nexthdr);
156 hlist = &raw_v6_hashinfo.ht[hash];
158 sk_nulls_for_each(sk, hnode, hlist) {
161 if (!raw_v6_match(net, sk, nexthdr, daddr, saddr,
162 inet6_iif(skb), inet6_sdif(skb)))
167 filtered = icmpv6_filter(sk, skb);
170 #if IS_ENABLED(CONFIG_IPV6_MIP6)
173 /* XXX: To validate MH only once for each packet,
174 * this is placed here. It should be after checking
175 * xfrm policy, however it doesn't. The checking xfrm
176 * policy is placed in rawv6_rcv() because it is
177 * required for each socket.
181 filter = rcu_dereference(mh_filter);
182 filtered = filter ? (*filter)(sk, skb) : 0;
194 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
196 /* Not releasing hash table! */
199 rawv6_rcv(sk, clone);
207 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
209 return ipv6_raw_deliver(skb, nexthdr);
212 /* This cleans up af_inet6 a bit. -DaveM */
213 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
215 struct inet_sock *inet = inet_sk(sk);
216 struct ipv6_pinfo *np = inet6_sk(sk);
217 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
222 if (addr_len < SIN6_LEN_RFC2133)
225 if (addr->sin6_family != AF_INET6)
228 addr_type = ipv6_addr_type(&addr->sin6_addr);
230 /* Raw sockets are IPv6 only */
231 if (addr_type == IPV6_ADDR_MAPPED)
232 return -EADDRNOTAVAIL;
237 if (sk->sk_state != TCP_CLOSE)
241 /* Check if the address belongs to the host. */
242 if (addr_type != IPV6_ADDR_ANY) {
243 struct net_device *dev = NULL;
245 if (__ipv6_addr_needs_scope_id(addr_type)) {
246 if (addr_len >= sizeof(struct sockaddr_in6) &&
247 addr->sin6_scope_id) {
248 /* Override any existing binding, if another
249 * one is supplied by user.
251 sk->sk_bound_dev_if = addr->sin6_scope_id;
254 /* Binding to link-local address requires an interface */
255 if (!sk->sk_bound_dev_if)
259 if (sk->sk_bound_dev_if) {
261 dev = dev_get_by_index_rcu(sock_net(sk),
262 sk->sk_bound_dev_if);
267 /* ipv4 addr of the socket is invalid. Only the
268 * unspecified and mapped address have a v4 equivalent.
270 v4addr = LOOPBACK4_IPV6;
271 if (!(addr_type & IPV6_ADDR_MULTICAST) &&
272 !ipv6_can_nonlocal_bind(sock_net(sk), inet)) {
273 err = -EADDRNOTAVAIL;
274 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
281 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
282 sk->sk_v6_rcv_saddr = addr->sin6_addr;
283 if (!(addr_type & IPV6_ADDR_MULTICAST))
284 np->saddr = addr->sin6_addr;
293 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
294 struct inet6_skb_parm *opt,
295 u8 type, u8 code, int offset, __be32 info)
297 struct inet_sock *inet = inet_sk(sk);
298 struct ipv6_pinfo *np = inet6_sk(sk);
302 /* Report error on raw socket, if:
303 1. User requested recverr.
304 2. Socket is connected (otherwise the error indication
305 is useless without recverr and error is hard.
307 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
310 harderr = icmpv6_err_convert(type, code, &err);
311 if (type == ICMPV6_PKT_TOOBIG) {
312 ip6_sk_update_pmtu(skb, sk, info);
313 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
315 if (type == NDISC_REDIRECT) {
316 ip6_sk_redirect(skb, sk);
320 u8 *payload = skb->data;
323 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
326 if (np->recverr || harderr) {
332 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
333 u8 type, u8 code, int inner_offset, __be32 info)
335 struct net *net = dev_net(skb->dev);
336 struct hlist_nulls_head *hlist;
337 struct hlist_nulls_node *hnode;
341 hash = raw_hashfunc(net, nexthdr);
342 hlist = &raw_v6_hashinfo.ht[hash];
344 sk_nulls_for_each(sk, hnode, hlist) {
345 /* Note: ipv6_hdr(skb) != skb->data */
346 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
348 if (!raw_v6_match(net, sk, nexthdr, &ip6h->saddr, &ip6h->daddr,
349 inet6_iif(skb), inet6_iif(skb)))
351 rawv6_err(sk, skb, NULL, type, code, inner_offset, info);
356 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
358 enum skb_drop_reason reason;
360 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
361 skb_checksum_complete(skb)) {
362 atomic_inc(&sk->sk_drops);
363 kfree_skb_reason(skb, SKB_DROP_REASON_SKB_CSUM);
367 /* Charge it to the socket. */
369 if (sock_queue_rcv_skb_reason(sk, skb, &reason) < 0) {
370 kfree_skb_reason(skb, reason);
378 * This is next to useless...
379 * if we demultiplex in network layer we don't need the extra call
380 * just to queue the skb...
381 * maybe we could have the network decide upon a hint if it
382 * should call raw_rcv for demultiplexing
384 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
386 struct inet_sock *inet = inet_sk(sk);
387 struct raw6_sock *rp = raw6_sk(sk);
389 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
390 atomic_inc(&sk->sk_drops);
391 kfree_skb_reason(skb, SKB_DROP_REASON_XFRM_POLICY);
396 skb->ip_summed = CHECKSUM_UNNECESSARY;
398 if (skb->ip_summed == CHECKSUM_COMPLETE) {
399 skb_postpull_rcsum(skb, skb_network_header(skb),
400 skb_network_header_len(skb));
401 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
402 &ipv6_hdr(skb)->daddr,
403 skb->len, inet->inet_num, skb->csum))
404 skb->ip_summed = CHECKSUM_UNNECESSARY;
406 if (!skb_csum_unnecessary(skb))
407 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
408 &ipv6_hdr(skb)->daddr,
413 if (skb_checksum_complete(skb)) {
414 atomic_inc(&sk->sk_drops);
415 kfree_skb_reason(skb, SKB_DROP_REASON_SKB_CSUM);
420 rawv6_rcv_skb(sk, skb);
426 * This should be easy, if there is something there
427 * we return it, otherwise we block.
430 static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
431 int flags, int *addr_len)
433 struct ipv6_pinfo *np = inet6_sk(sk);
434 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
442 if (flags & MSG_ERRQUEUE)
443 return ipv6_recv_error(sk, msg, len, addr_len);
445 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
446 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
448 skb = skb_recv_datagram(sk, flags, &err);
455 msg->msg_flags |= MSG_TRUNC;
458 if (skb_csum_unnecessary(skb)) {
459 err = skb_copy_datagram_msg(skb, 0, msg, copied);
460 } else if (msg->msg_flags&MSG_TRUNC) {
461 if (__skb_checksum_complete(skb))
463 err = skb_copy_datagram_msg(skb, 0, msg, copied);
465 err = skb_copy_and_csum_datagram_msg(skb, 0, msg);
472 /* Copy the address. */
474 sin6->sin6_family = AF_INET6;
476 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
477 sin6->sin6_flowinfo = 0;
478 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
480 *addr_len = sizeof(*sin6);
483 sock_recv_cmsgs(msg, sk, skb);
486 ip6_datagram_recv_ctl(sk, msg, skb);
489 if (flags & MSG_TRUNC)
493 skb_free_datagram(sk, skb);
498 skb_kill_datagram(sk, skb, flags);
500 /* Error for blocking case is chosen to masquerade
501 as some normal condition.
503 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
507 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
508 struct raw6_sock *rp)
510 struct ipv6_txoptions *opt;
522 skb = skb_peek(&sk->sk_write_queue);
527 total_len = inet_sk(sk)->cork.base.length;
528 opt = inet6_sk(sk)->cork.opt;
529 total_len -= opt ? opt->opt_flen : 0;
531 if (offset >= total_len - 1) {
533 ip6_flush_pending_frames(sk);
537 /* should be check HW csum miyazawa */
538 if (skb_queue_len(&sk->sk_write_queue) == 1) {
540 * Only one fragment on the socket.
542 tmp_csum = skb->csum;
544 struct sk_buff *csum_skb = NULL;
547 skb_queue_walk(&sk->sk_write_queue, skb) {
548 tmp_csum = csum_add(tmp_csum, skb->csum);
553 len = skb->len - skb_transport_offset(skb);
565 offset += skb_transport_offset(skb);
566 err = skb_copy_bits(skb, offset, &csum, 2);
568 ip6_flush_pending_frames(sk);
572 /* in case cksum was not initialized */
574 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
576 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
577 total_len, fl6->flowi6_proto, tmp_csum);
579 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
580 csum = CSUM_MANGLED_0;
582 BUG_ON(skb_store_bits(skb, offset, &csum, 2));
585 err = ip6_push_pending_frames(sk);
590 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length,
591 struct flowi6 *fl6, struct dst_entry **dstp,
592 unsigned int flags, const struct sockcm_cookie *sockc)
594 struct ipv6_pinfo *np = inet6_sk(sk);
595 struct net *net = sock_net(sk);
599 struct rt6_info *rt = (struct rt6_info *)*dstp;
600 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
601 int tlen = rt->dst.dev->needed_tailroom;
603 if (length > rt->dst.dev->mtu) {
604 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
607 if (length < sizeof(struct ipv6hdr))
612 skb = sock_alloc_send_skb(sk,
613 length + hlen + tlen + 15,
614 flags & MSG_DONTWAIT, &err);
617 skb_reserve(skb, hlen);
619 skb->protocol = htons(ETH_P_IPV6);
620 skb->priority = sk->sk_priority;
621 skb->mark = sockc->mark;
622 skb->tstamp = sockc->transmit_time;
624 skb_put(skb, length);
625 skb_reset_network_header(skb);
628 skb->ip_summed = CHECKSUM_NONE;
630 skb_setup_tx_timestamp(skb, sockc->tsflags);
632 if (flags & MSG_CONFIRM)
633 skb_set_dst_pending_confirm(skb, 1);
635 skb->transport_header = skb->network_header;
636 err = memcpy_from_msg(iph, msg, length);
643 skb_dst_set(skb, &rt->dst);
646 /* if egress device is enslaved to an L3 master device pass the
647 * skb to its handler for processing
649 skb = l3mdev_ip6_out(sk, skb);
653 /* Acquire rcu_read_lock() in case we need to use rt->rt6i_idev
654 * in the error path. Since skb has been freed, the dst could
655 * have been queued for deletion.
658 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
659 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
660 NULL, rt->dst.dev, dst_output);
662 err = net_xmit_errno(err);
664 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
673 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
675 if (err == -ENOBUFS && !np->recverr)
680 struct raw6_frag_vec {
686 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6)
689 switch (fl6->flowi6_proto) {
692 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
694 fl6->fl6_icmp_type = rfv->c[0];
695 fl6->fl6_icmp_code = rfv->c[1];
700 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
702 fl6->fl6_mh_type = rfv->c[2];
707 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd,
710 struct raw6_frag_vec *rfv = from;
712 if (offset < rfv->hlen) {
713 int copy = min(rfv->hlen - offset, len);
715 if (skb->ip_summed == CHECKSUM_PARTIAL)
716 memcpy(to, rfv->c + offset, copy);
718 skb->csum = csum_block_add(
720 csum_partial_copy_nocheck(rfv->c + offset,
735 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
738 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
740 struct ipv6_txoptions *opt_to_free = NULL;
741 struct ipv6_txoptions opt_space;
742 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
743 struct in6_addr *daddr, *final_p, final;
744 struct inet_sock *inet = inet_sk(sk);
745 struct ipv6_pinfo *np = inet6_sk(sk);
746 struct raw6_sock *rp = raw6_sk(sk);
747 struct ipv6_txoptions *opt = NULL;
748 struct ip6_flowlabel *flowlabel = NULL;
749 struct dst_entry *dst = NULL;
750 struct raw6_frag_vec rfv;
752 struct ipcm6_cookie ipc6;
753 int addr_len = msg->msg_namelen;
758 /* Rough check on arithmetic overflow,
759 better check is made in ip6_append_data().
764 /* Mirror BSD error message compatibility */
765 if (msg->msg_flags & MSG_OOB)
768 /* hdrincl should be READ_ONCE(inet->hdrincl)
769 * but READ_ONCE() doesn't work with bit fields.
770 * Doing this indirectly yields the same result.
772 hdrincl = inet->hdrincl;
773 hdrincl = READ_ONCE(hdrincl);
776 * Get and verify the address.
778 memset(&fl6, 0, sizeof(fl6));
780 fl6.flowi6_mark = sk->sk_mark;
781 fl6.flowi6_uid = sk->sk_uid;
784 ipc6.sockc.tsflags = sk->sk_tsflags;
785 ipc6.sockc.mark = sk->sk_mark;
788 if (addr_len < SIN6_LEN_RFC2133)
791 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
792 return -EAFNOSUPPORT;
794 /* port is the proto value [0..255] carried in nexthdr */
795 proto = ntohs(sin6->sin6_port);
798 proto = inet->inet_num;
799 else if (proto != inet->inet_num)
805 daddr = &sin6->sin6_addr;
807 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
808 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
809 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
810 if (IS_ERR(flowlabel))
816 * Otherwise it will be difficult to maintain
819 if (sk->sk_state == TCP_ESTABLISHED &&
820 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
821 daddr = &sk->sk_v6_daddr;
823 if (addr_len >= sizeof(struct sockaddr_in6) &&
824 sin6->sin6_scope_id &&
825 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
826 fl6.flowi6_oif = sin6->sin6_scope_id;
828 if (sk->sk_state != TCP_ESTABLISHED)
829 return -EDESTADDRREQ;
831 proto = inet->inet_num;
832 daddr = &sk->sk_v6_daddr;
833 fl6.flowlabel = np->flow_label;
836 if (fl6.flowi6_oif == 0)
837 fl6.flowi6_oif = sk->sk_bound_dev_if;
839 if (msg->msg_controllen) {
841 memset(opt, 0, sizeof(struct ipv6_txoptions));
842 opt->tot_len = sizeof(struct ipv6_txoptions);
845 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
847 fl6_sock_release(flowlabel);
850 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
851 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
852 if (IS_ERR(flowlabel))
855 if (!(opt->opt_nflen|opt->opt_flen))
863 opt = fl6_merge_options(&opt_space, flowlabel, opt);
864 opt = ipv6_fixup_options(&opt_space, opt);
866 fl6.flowi6_proto = proto;
867 fl6.flowi6_mark = ipc6.sockc.mark;
872 err = rawv6_probe_proto_opt(&rfv, &fl6);
877 if (!ipv6_addr_any(daddr))
880 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
881 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
882 fl6.saddr = np->saddr;
884 final_p = fl6_update_dst(&fl6, opt, &final);
886 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
887 fl6.flowi6_oif = np->mcast_oif;
888 else if (!fl6.flowi6_oif)
889 fl6.flowi6_oif = np->ucast_oif;
890 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
893 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH;
896 ipc6.tclass = np->tclass;
898 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
900 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
906 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
908 if (ipc6.dontfrag < 0)
909 ipc6.dontfrag = np->dontfrag;
911 if (msg->msg_flags&MSG_CONFIRM)
916 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst,
917 msg->msg_flags, &ipc6.sockc);
921 err = ip6_append_data(sk, raw6_getfrag, &rfv,
922 len, 0, &ipc6, &fl6, (struct rt6_info *)dst,
926 ip6_flush_pending_frames(sk);
927 else if (!(msg->msg_flags & MSG_MORE))
928 err = rawv6_push_pending_frames(sk, &fl6, rp);
934 fl6_sock_release(flowlabel);
935 txopt_put(opt_to_free);
936 return err < 0 ? err : len;
938 if (msg->msg_flags & MSG_PROBE)
939 dst_confirm_neigh(dst, &fl6.daddr);
940 if (!(msg->msg_flags & MSG_PROBE) || len)
941 goto back_from_confirm;
946 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
947 sockptr_t optval, int optlen)
951 if (optlen > sizeof(struct icmp6_filter))
952 optlen = sizeof(struct icmp6_filter);
953 if (copy_from_sockptr(&raw6_sk(sk)->filter, optval, optlen))
963 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
964 char __user *optval, int __user *optlen)
970 if (get_user(len, optlen))
974 if (len > sizeof(struct icmp6_filter))
975 len = sizeof(struct icmp6_filter);
976 if (put_user(len, optlen))
978 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
989 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
990 sockptr_t optval, unsigned int optlen)
992 struct raw6_sock *rp = raw6_sk(sk);
995 if (optlen < sizeof(val))
998 if (copy_from_sockptr(&val, optval, sizeof(val)))
1003 if (sk->sk_type != SOCK_RAW)
1005 inet_sk(sk)->hdrincl = !!val;
1008 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
1009 level == IPPROTO_IPV6) {
1011 * RFC3542 tells that IPV6_CHECKSUM socket
1012 * option in the IPPROTO_IPV6 level is not
1013 * allowed on ICMPv6 sockets.
1014 * If you want to set it, use IPPROTO_RAW
1015 * level IPV6_CHECKSUM socket option
1016 * (Linux extension).
1021 /* You may get strange result with a positive odd offset;
1022 RFC2292bis agrees with me. */
1023 if (val > 0 && (val&1))
1035 return -ENOPROTOOPT;
1039 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1040 sockptr_t optval, unsigned int optlen)
1047 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1049 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1051 if (optname == IPV6_CHECKSUM ||
1052 optname == IPV6_HDRINCL)
1056 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1059 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1062 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1063 char __user *optval, int __user *optlen)
1065 struct raw6_sock *rp = raw6_sk(sk);
1068 if (get_user(len, optlen))
1073 val = inet_sk(sk)->hdrincl;
1077 * We allow getsockopt() for IPPROTO_IPV6-level
1078 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1079 * since RFC3542 is silent about it.
1081 if (rp->checksum == 0)
1088 return -ENOPROTOOPT;
1091 len = min_t(unsigned int, sizeof(int), len);
1093 if (put_user(len, optlen))
1095 if (copy_to_user(optval, &val, len))
1100 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1101 char __user *optval, int __user *optlen)
1108 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1110 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1112 if (optname == IPV6_CHECKSUM ||
1113 optname == IPV6_HDRINCL)
1117 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1120 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1123 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1127 int amount = sk_wmem_alloc_get(sk);
1129 return put_user(amount, (int __user *)arg);
1132 struct sk_buff *skb;
1135 spin_lock_bh(&sk->sk_receive_queue.lock);
1136 skb = skb_peek(&sk->sk_receive_queue);
1139 spin_unlock_bh(&sk->sk_receive_queue.lock);
1140 return put_user(amount, (int __user *)arg);
1144 #ifdef CONFIG_IPV6_MROUTE
1145 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1147 return -ENOIOCTLCMD;
1152 #ifdef CONFIG_COMPAT
1153 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1158 return -ENOIOCTLCMD;
1160 #ifdef CONFIG_IPV6_MROUTE
1161 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1163 return -ENOIOCTLCMD;
1169 static void rawv6_close(struct sock *sk, long timeout)
1171 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1172 ip6_ra_control(sk, -1);
1174 sk_common_release(sk);
1177 static void raw6_destroy(struct sock *sk)
1180 ip6_flush_pending_frames(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_v6_only,
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 .useroffset = offsetof(struct raw6_sock, filter),
1222 .usersize = sizeof_field(struct raw6_sock, filter),
1223 .h.raw_hash = &raw_v6_hashinfo,
1224 #ifdef CONFIG_COMPAT
1225 .compat_ioctl = compat_rawv6_ioctl,
1227 .diag_destroy = raw_abort,
1230 #ifdef CONFIG_PROC_FS
1231 static int raw6_seq_show(struct seq_file *seq, void *v)
1233 if (v == SEQ_START_TOKEN) {
1234 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1236 struct sock *sp = v;
1237 __u16 srcp = inet_sk(sp)->inet_num;
1238 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1239 raw_seq_private(seq)->bucket);
1244 static const struct seq_operations raw6_seq_ops = {
1245 .start = raw_seq_start,
1246 .next = raw_seq_next,
1247 .stop = raw_seq_stop,
1248 .show = raw6_seq_show,
1251 static int __net_init raw6_init_net(struct net *net)
1253 if (!proc_create_net_data("raw6", 0444, net->proc_net, &raw6_seq_ops,
1254 sizeof(struct raw_iter_state), &raw_v6_hashinfo))
1260 static void __net_exit raw6_exit_net(struct net *net)
1262 remove_proc_entry("raw6", net->proc_net);
1265 static struct pernet_operations raw6_net_ops = {
1266 .init = raw6_init_net,
1267 .exit = raw6_exit_net,
1270 int __init raw6_proc_init(void)
1272 return register_pernet_subsys(&raw6_net_ops);
1275 void raw6_proc_exit(void)
1277 unregister_pernet_subsys(&raw6_net_ops);
1279 #endif /* CONFIG_PROC_FS */
1281 /* Same as inet6_dgram_ops, sans udp_poll. */
1282 const struct proto_ops inet6_sockraw_ops = {
1284 .owner = THIS_MODULE,
1285 .release = inet6_release,
1287 .connect = inet_dgram_connect, /* ok */
1288 .socketpair = sock_no_socketpair, /* a do nothing */
1289 .accept = sock_no_accept, /* a do nothing */
1290 .getname = inet6_getname,
1291 .poll = datagram_poll, /* ok */
1292 .ioctl = inet6_ioctl, /* must change */
1293 .gettstamp = sock_gettstamp,
1294 .listen = sock_no_listen, /* ok */
1295 .shutdown = inet_shutdown, /* ok */
1296 .setsockopt = sock_common_setsockopt, /* ok */
1297 .getsockopt = sock_common_getsockopt, /* ok */
1298 .sendmsg = inet_sendmsg, /* ok */
1299 .recvmsg = sock_common_recvmsg, /* ok */
1300 .mmap = sock_no_mmap,
1301 .sendpage = sock_no_sendpage,
1302 #ifdef CONFIG_COMPAT
1303 .compat_ioctl = inet6_compat_ioctl,
1307 static struct inet_protosw rawv6_protosw = {
1309 .protocol = IPPROTO_IP, /* wild card */
1310 .prot = &rawv6_prot,
1311 .ops = &inet6_sockraw_ops,
1312 .flags = INET_PROTOSW_REUSE,
1315 int __init rawv6_init(void)
1317 return inet6_register_protosw(&rawv6_protosw);
1320 void rawv6_exit(void)
1322 inet6_unregister_protosw(&rawv6_protosw);