3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Adapted from linux/net/ipv4/raw.c
10 * $Id: raw.c,v 1.51 2002/02/01 22:01:04 davem Exp $
13 * Hideaki YOSHIFUJI : sin6_scope_id support
14 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
17 * This program is free software; you can redistribute it and/or
18 * modify it under the terms of the GNU General Public License
19 * as published by the Free Software Foundation; either version
20 * 2 of the License, or (at your option) any later version.
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/sched.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/icmpv6.h>
33 #include <linux/netfilter.h>
34 #include <linux/netfilter_ipv6.h>
35 #include <linux/skbuff.h>
36 #include <asm/uaccess.h>
37 #include <asm/ioctls.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 #ifdef CONFIG_IPV6_MIP6
57 #include <net/rawv6.h>
60 #include <linux/proc_fs.h>
61 #include <linux/seq_file.h>
63 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
64 DEFINE_RWLOCK(raw_v6_lock);
66 static void raw_v6_hash(struct sock *sk)
68 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num &
69 (RAWV6_HTABLE_SIZE - 1)];
71 write_lock_bh(&raw_v6_lock);
72 sk_add_node(sk, list);
73 sock_prot_inc_use(sk->sk_prot);
74 write_unlock_bh(&raw_v6_lock);
77 static void raw_v6_unhash(struct sock *sk)
79 write_lock_bh(&raw_v6_lock);
80 if (sk_del_node_init(sk))
81 sock_prot_dec_use(sk->sk_prot);
82 write_unlock_bh(&raw_v6_lock);
86 /* Grumble... icmp and ip_input want to get at this... */
87 struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
88 struct in6_addr *loc_addr, struct in6_addr *rmt_addr,
91 struct hlist_node *node;
92 int is_multicast = ipv6_addr_is_multicast(loc_addr);
94 sk_for_each_from(sk, node)
95 if (inet_sk(sk)->num == num) {
96 struct ipv6_pinfo *np = inet6_sk(sk);
98 if (!ipv6_addr_any(&np->daddr) &&
99 !ipv6_addr_equal(&np->daddr, rmt_addr))
102 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
105 if (!ipv6_addr_any(&np->rcv_saddr)) {
106 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr) &&
107 addr6_in_nx_info(sk->sk_nx_info, loc_addr))
110 inet6_mc_check(sk, loc_addr, rmt_addr))
125 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
127 struct icmp6hdr *icmph;
128 struct raw6_sock *rp = raw6_sk(sk);
130 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
131 __u32 *data = &rp->filter.data[0];
134 icmph = (struct icmp6hdr *) skb->data;
135 bit_nr = icmph->icmp6_type;
137 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
143 * demultiplex raw sockets.
144 * (should consider queueing the skb in the sock receive_queue
145 * without calling rawv6.c)
147 * Caller owns SKB so we must make clones.
149 int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
151 struct in6_addr *saddr;
152 struct in6_addr *daddr;
157 saddr = &skb->nh.ipv6h->saddr;
160 hash = nexthdr & (MAX_INET_PROTOS - 1);
162 read_lock(&raw_v6_lock);
163 sk = sk_head(&raw_v6_htable[hash]);
166 * The first socket found will be delivered after
167 * delivery to transport protocols.
173 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
181 filtered = icmpv6_filter(sk, skb);
183 #ifdef CONFIG_IPV6_MIP6
185 /* XXX: To validate MH only once for each packet,
186 * this is placed here. It should be after checking
187 * xfrm policy, however it doesn't. The checking xfrm
188 * policy is placed in rawv6_rcv() because it is
189 * required for each socket.
191 filtered = mip6_mh_filter(sk, skb);
202 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
204 /* Not releasing hash table! */
207 rawv6_rcv(sk, clone);
210 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr,
214 read_unlock(&raw_v6_lock);
218 /* This cleans up af_inet6 a bit. -DaveM */
219 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
221 struct inet_sock *inet = inet_sk(sk);
222 struct ipv6_pinfo *np = inet6_sk(sk);
223 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
228 if (addr_len < SIN6_LEN_RFC2133)
230 addr_type = ipv6_addr_type(&addr->sin6_addr);
232 /* Raw sockets are IPv6 only */
233 if (addr_type == IPV6_ADDR_MAPPED)
234 return(-EADDRNOTAVAIL);
239 if (sk->sk_state != TCP_CLOSE)
242 /* Check if the address belongs to the host. */
243 if (addr_type != IPV6_ADDR_ANY) {
244 struct net_device *dev = NULL;
246 if (addr_type & IPV6_ADDR_LINKLOCAL) {
247 if (addr_len >= sizeof(struct sockaddr_in6) &&
248 addr->sin6_scope_id) {
249 /* Override any existing binding, if another
250 * one is supplied by user.
252 sk->sk_bound_dev_if = addr->sin6_scope_id;
255 /* Binding to link-local address requires an interface */
256 if (!sk->sk_bound_dev_if)
259 dev = dev_get_by_index(sk->sk_bound_dev_if);
266 /* VServer: Only accept if we have the address available */
267 if (sk->sk_nx_info && !addr6_in_nx_info(sk->sk_nx_info, &addr->sin6_addr)) {
268 vxdprintk(VXD_CBIT(net, 3), "rawv6_bind(" NIP6_FMT ", %d): EADDRNOTAVAIL",
269 NIP6(addr->sin6_addr), sk->sk_nx_info->nx_id);
270 err = -EADDRNOTAVAIL;
275 vxdprintk(VXD_CBIT(net, 3), "rawv6_bind(" NIP6_FMT ", %d): PASS",
276 NIP6(addr->sin6_addr), sk->sk_nx_info ? sk->sk_nx_info->nx_id : 0);
278 /* ipv4 addr of the socket is invalid. Only the
279 * unspecified and mapped address have a v4 equivalent.
281 v4addr = LOOPBACK4_IPV6;
282 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
283 err = -EADDRNOTAVAIL;
284 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
294 inet->rcv_saddr = inet->saddr = v4addr;
295 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
296 if (!(addr_type & IPV6_ADDR_MULTICAST))
297 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
304 void rawv6_err(struct sock *sk, struct sk_buff *skb,
305 struct inet6_skb_parm *opt,
306 int type, int code, int offset, __be32 info)
308 struct inet_sock *inet = inet_sk(sk);
309 struct ipv6_pinfo *np = inet6_sk(sk);
313 /* Report error on raw socket, if:
314 1. User requested recverr.
315 2. Socket is connected (otherwise the error indication
316 is useless without recverr and error is hard.
318 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
321 harderr = icmpv6_err_convert(type, code, &err);
322 if (type == ICMPV6_PKT_TOOBIG)
323 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
326 u8 *payload = skb->data;
329 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
332 if (np->recverr || harderr) {
334 sk->sk_error_report(sk);
338 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
340 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
341 skb_checksum_complete(skb)) {
342 /* FIXME: increment a raw6 drops counter here */
347 /* Charge it to the socket. */
348 if (sock_queue_rcv_skb(sk,skb)<0) {
349 /* FIXME: increment a raw6 drops counter here */
358 * This is next to useless...
359 * if we demultiplex in network layer we don't need the extra call
360 * just to queue the skb...
361 * maybe we could have the network decide upon a hint if it
362 * should call raw_rcv for demultiplexing
364 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
366 struct inet_sock *inet = inet_sk(sk);
367 struct raw6_sock *rp = raw6_sk(sk);
369 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
375 skb->ip_summed = CHECKSUM_UNNECESSARY;
377 if (skb->ip_summed == CHECKSUM_COMPLETE) {
378 skb_postpull_rcsum(skb, skb->nh.raw,
379 skb->h.raw - skb->nh.raw);
380 if (!csum_ipv6_magic(&skb->nh.ipv6h->saddr,
381 &skb->nh.ipv6h->daddr,
382 skb->len, inet->num, skb->csum))
383 skb->ip_summed = CHECKSUM_UNNECESSARY;
385 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
386 skb->csum = ~csum_unfold(csum_ipv6_magic(&skb->nh.ipv6h->saddr,
387 &skb->nh.ipv6h->daddr,
388 skb->len, inet->num, 0));
391 if (skb_checksum_complete(skb)) {
392 /* FIXME: increment a raw6 drops counter here */
398 rawv6_rcv_skb(sk, skb);
404 * This should be easy, if there is something there
405 * we return it, otherwise we block.
408 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
409 struct msghdr *msg, size_t len,
410 int noblock, int flags, int *addr_len)
412 struct ipv6_pinfo *np = inet6_sk(sk);
413 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
422 *addr_len=sizeof(*sin6);
424 if (flags & MSG_ERRQUEUE)
425 return ipv6_recv_error(sk, msg, len);
427 skb = skb_recv_datagram(sk, flags, noblock, &err);
434 msg->msg_flags |= MSG_TRUNC;
437 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
438 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
439 } else if (msg->msg_flags&MSG_TRUNC) {
440 if (__skb_checksum_complete(skb))
442 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
444 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
451 /* Copy the address. */
453 sin6->sin6_family = AF_INET6;
455 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
456 sin6->sin6_flowinfo = 0;
457 sin6->sin6_scope_id = 0;
458 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
459 sin6->sin6_scope_id = IP6CB(skb)->iif;
462 sock_recv_timestamp(msg, sk, skb);
465 datagram_recv_ctl(sk, msg, skb);
468 if (flags & MSG_TRUNC)
472 skb_free_datagram(sk, skb);
477 skb_kill_datagram(sk, skb, flags);
479 /* Error for blocking case is chosen to masquerade
480 as some normal condition.
482 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
483 /* FIXME: increment a raw6 drops counter here */
487 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
488 struct raw6_sock *rp)
501 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
505 total_len = inet_sk(sk)->cork.length - (skb->nh.raw - skb->data);
506 if (offset >= total_len - 1) {
508 ip6_flush_pending_frames(sk);
512 /* should be check HW csum miyazawa */
513 if (skb_queue_len(&sk->sk_write_queue) == 1) {
515 * Only one fragment on the socket.
517 tmp_csum = skb->csum;
519 struct sk_buff *csum_skb = NULL;
522 skb_queue_walk(&sk->sk_write_queue, skb) {
523 tmp_csum = csum_add(tmp_csum, skb->csum);
528 len = skb->len - (skb->h.raw - skb->data);
540 offset += skb->h.raw - skb->data;
541 if (skb_copy_bits(skb, offset, &csum, 2))
544 /* in case cksum was not initialized */
546 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
548 csum = csum_ipv6_magic(&fl->fl6_src,
550 total_len, fl->proto, tmp_csum);
552 if (csum == 0 && fl->proto == IPPROTO_UDP)
553 csum = CSUM_MANGLED_0;
555 if (skb_store_bits(skb, offset, &csum, 2))
559 err = ip6_push_pending_frames(sk);
564 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
565 struct flowi *fl, struct rt6_info *rt,
568 struct ipv6_pinfo *np = inet6_sk(sk);
574 if (length > rt->u.dst.dev->mtu) {
575 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
581 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
583 skb = sock_alloc_send_skb(sk, length+hh_len+15,
584 flags&MSG_DONTWAIT, &err);
587 skb_reserve(skb, hh_len);
589 skb->priority = sk->sk_priority;
590 skb->dst = dst_clone(&rt->u.dst);
592 skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length);
594 skb->ip_summed = CHECKSUM_NONE;
596 skb->h.raw = skb->nh.raw;
597 err = memcpy_fromiovecend((void *)iph, from, 0, length);
602 if (!nx_check(0, VS_ADMIN) && !capable(CAP_NET_RAW)
603 && (!addr6_in_nx_info(sk->sk_nx_info, &iph->saddr)))
606 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
607 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
610 err = np->recverr ? net_xmit_errno(err) : 0;
621 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
625 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
628 u8 __user *type = NULL;
629 u8 __user *code = NULL;
630 #ifdef CONFIG_IPV6_MIP6
639 for (i = 0; i < msg->msg_iovlen; i++) {
640 iov = &msg->msg_iov[i];
646 /* check if one-byte field is readable or not. */
647 if (iov->iov_base && iov->iov_len < 1)
651 type = iov->iov_base;
652 /* check if code field is readable or not. */
653 if (iov->iov_len > 1)
656 code = iov->iov_base;
659 if (get_user(fl->fl_icmp_type, type) ||
660 get_user(fl->fl_icmp_code, code))
665 #ifdef CONFIG_IPV6_MIP6
667 if (iov->iov_base && iov->iov_len < 1)
669 /* check if type field is readable or not. */
670 if (iov->iov_len > 2 - len) {
671 u8 __user *p = iov->iov_base;
672 if (get_user(fl->fl_mh_type, &p[2 - len]))
690 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
691 struct msghdr *msg, size_t len)
693 struct ipv6_txoptions opt_space;
694 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
695 struct in6_addr *daddr, *final_p = NULL, final;
696 struct inet_sock *inet = inet_sk(sk);
697 struct ipv6_pinfo *np = inet6_sk(sk);
698 struct raw6_sock *rp = raw6_sk(sk);
699 struct ipv6_txoptions *opt = NULL;
700 struct ip6_flowlabel *flowlabel = NULL;
701 struct dst_entry *dst = NULL;
703 int addr_len = msg->msg_namelen;
709 /* Rough check on arithmetic overflow,
710 better check is made in ip6_append_data().
715 /* Mirror BSD error message compatibility */
716 if (msg->msg_flags & MSG_OOB)
720 * Get and verify the address.
722 memset(&fl, 0, sizeof(fl));
725 if (addr_len < SIN6_LEN_RFC2133)
728 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
729 return(-EAFNOSUPPORT);
731 /* port is the proto value [0..255] carried in nexthdr */
732 proto = ntohs(sin6->sin6_port);
736 else if (proto != inet->num)
742 daddr = &sin6->sin6_addr;
744 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
745 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
746 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
747 if (flowlabel == NULL)
749 daddr = &flowlabel->dst;
754 * Otherwise it will be difficult to maintain
757 if (sk->sk_state == TCP_ESTABLISHED &&
758 ipv6_addr_equal(daddr, &np->daddr))
761 if (addr_len >= sizeof(struct sockaddr_in6) &&
762 sin6->sin6_scope_id &&
763 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
764 fl.oif = sin6->sin6_scope_id;
766 if (sk->sk_state != TCP_ESTABLISHED)
767 return -EDESTADDRREQ;
771 fl.fl6_flowlabel = np->flow_label;
774 if (ipv6_addr_any(daddr)) {
776 * unspecified destination address
777 * treated as error... is this correct ?
779 fl6_sock_release(flowlabel);
784 fl.oif = sk->sk_bound_dev_if;
786 if (msg->msg_controllen) {
788 memset(opt, 0, sizeof(struct ipv6_txoptions));
789 opt->tot_len = sizeof(struct ipv6_txoptions);
791 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
793 fl6_sock_release(flowlabel);
796 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
797 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
798 if (flowlabel == NULL)
801 if (!(opt->opt_nflen|opt->opt_flen))
807 opt = fl6_merge_options(&opt_space, flowlabel, opt);
808 opt = ipv6_fixup_options(&opt_space, opt);
811 err = rawv6_probe_proto_opt(&fl, msg);
815 ipv6_addr_copy(&fl.fl6_dst, daddr);
816 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
817 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
819 /* merge ip6_build_xmit from ip6_output */
820 if (opt && opt->srcrt) {
821 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
822 ipv6_addr_copy(&final, &fl.fl6_dst);
823 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
827 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
828 fl.oif = np->mcast_oif;
829 security_sk_classify_flow(sk, &fl);
831 err = ip6_dst_lookup(sk, &dst, &fl);
835 ipv6_addr_copy(&fl.fl6_dst, final_p);
837 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
841 if (ipv6_addr_is_multicast(&fl.fl6_dst))
842 hlimit = np->mcast_hops;
844 hlimit = np->hop_limit;
846 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
848 hlimit = ipv6_get_hoplimit(dst->dev);
857 if (msg->msg_flags&MSG_CONFIRM)
862 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
865 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
866 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
870 ip6_flush_pending_frames(sk);
871 else if (!(msg->msg_flags & MSG_MORE))
872 err = rawv6_push_pending_frames(sk, &fl, rp);
879 fl6_sock_release(flowlabel);
880 return err<0?err:len;
883 if (!(msg->msg_flags & MSG_PROBE) || len)
884 goto back_from_confirm;
889 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
890 char __user *optval, int optlen)
894 if (optlen > sizeof(struct icmp6_filter))
895 optlen = sizeof(struct icmp6_filter);
896 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
906 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
907 char __user *optval, int __user *optlen)
913 if (get_user(len, optlen))
917 if (len > sizeof(struct icmp6_filter))
918 len = sizeof(struct icmp6_filter);
919 if (put_user(len, optlen))
921 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
932 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
933 char __user *optval, int optlen)
935 struct raw6_sock *rp = raw6_sk(sk);
938 if (get_user(val, (int __user *)optval))
943 /* You may get strange result with a positive odd offset;
944 RFC2292bis agrees with me. */
945 if (val > 0 && (val&1))
958 return(-ENOPROTOOPT);
962 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
963 char __user *optval, int optlen)
970 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
972 return rawv6_seticmpfilter(sk, level, optname, optval,
975 if (optname == IPV6_CHECKSUM)
978 return ipv6_setsockopt(sk, level, optname, optval,
981 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
985 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
986 char __user *optval, int optlen)
992 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
994 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
996 if (optname == IPV6_CHECKSUM)
999 return compat_ipv6_setsockopt(sk, level, optname,
1002 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1006 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1007 char __user *optval, int __user *optlen)
1009 struct raw6_sock *rp = raw6_sk(sk);
1012 if (get_user(len,optlen))
1017 if (rp->checksum == 0)
1024 return -ENOPROTOOPT;
1027 len = min_t(unsigned int, sizeof(int), len);
1029 if (put_user(len, optlen))
1031 if (copy_to_user(optval,&val,len))
1036 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1037 char __user *optval, int __user *optlen)
1044 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1046 return rawv6_geticmpfilter(sk, level, optname, optval,
1049 if (optname == IPV6_CHECKSUM)
1052 return ipv6_getsockopt(sk, level, optname, optval,
1055 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1058 #ifdef CONFIG_COMPAT
1059 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1060 char __user *optval, int __user *optlen)
1066 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1068 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1070 if (optname == IPV6_CHECKSUM)
1073 return compat_ipv6_getsockopt(sk, level, optname,
1076 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1080 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1085 int amount = atomic_read(&sk->sk_wmem_alloc);
1086 return put_user(amount, (int __user *)arg);
1090 struct sk_buff *skb;
1093 spin_lock_bh(&sk->sk_receive_queue.lock);
1094 skb = skb_peek(&sk->sk_receive_queue);
1096 amount = skb->tail - skb->h.raw;
1097 spin_unlock_bh(&sk->sk_receive_queue.lock);
1098 return put_user(amount, (int __user *)arg);
1102 return -ENOIOCTLCMD;
1106 static void rawv6_close(struct sock *sk, long timeout)
1108 if (inet_sk(sk)->num == IPPROTO_RAW)
1109 ip6_ra_control(sk, -1, NULL);
1111 sk_common_release(sk);
1114 static int rawv6_init_sk(struct sock *sk)
1116 if (inet_sk(sk)->num == IPPROTO_ICMPV6) {
1117 struct raw6_sock *rp = raw6_sk(sk);
1124 struct proto rawv6_prot = {
1126 .owner = THIS_MODULE,
1127 .close = rawv6_close,
1128 .connect = ip6_datagram_connect,
1129 .disconnect = udp_disconnect,
1130 .ioctl = rawv6_ioctl,
1131 .init = rawv6_init_sk,
1132 .destroy = inet6_destroy_sock,
1133 .setsockopt = rawv6_setsockopt,
1134 .getsockopt = rawv6_getsockopt,
1135 .sendmsg = rawv6_sendmsg,
1136 .recvmsg = rawv6_recvmsg,
1138 .backlog_rcv = rawv6_rcv_skb,
1139 .hash = raw_v6_hash,
1140 .unhash = raw_v6_unhash,
1141 .obj_size = sizeof(struct raw6_sock),
1142 #ifdef CONFIG_COMPAT
1143 .compat_setsockopt = compat_rawv6_setsockopt,
1144 .compat_getsockopt = compat_rawv6_getsockopt,
1148 #ifdef CONFIG_PROC_FS
1149 struct raw6_iter_state {
1153 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1155 static struct sock *raw6_get_first(struct seq_file *seq)
1158 struct hlist_node *node;
1159 struct raw6_iter_state* state = raw6_seq_private(seq);
1161 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1162 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1163 if (sk->sk_family == PF_INET6)
1170 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1172 struct raw6_iter_state* state = raw6_seq_private(seq);
1178 } while (sk && sk->sk_family != PF_INET6);
1180 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1181 sk = sk_head(&raw_v6_htable[state->bucket]);
1187 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1189 struct sock *sk = raw6_get_first(seq);
1191 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1193 return pos ? NULL : sk;
1196 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1198 read_lock(&raw_v6_lock);
1199 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1202 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1206 if (v == SEQ_START_TOKEN)
1207 sk = raw6_get_first(seq);
1209 sk = raw6_get_next(seq, v);
1214 static void raw6_seq_stop(struct seq_file *seq, void *v)
1216 read_unlock(&raw_v6_lock);
1219 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1221 struct ipv6_pinfo *np = inet6_sk(sp);
1222 struct in6_addr *dest, *src;
1226 src = &np->rcv_saddr;
1228 srcp = inet_sk(sp)->num;
1230 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1231 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1233 src->s6_addr32[0], src->s6_addr32[1],
1234 src->s6_addr32[2], src->s6_addr32[3], srcp,
1235 dest->s6_addr32[0], dest->s6_addr32[1],
1236 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1238 atomic_read(&sp->sk_wmem_alloc),
1239 atomic_read(&sp->sk_rmem_alloc),
1243 atomic_read(&sp->sk_refcnt), sp);
1246 static int raw6_seq_show(struct seq_file *seq, void *v)
1248 if (v == SEQ_START_TOKEN)
1253 "st tx_queue rx_queue tr tm->when retrnsmt"
1254 " uid timeout inode\n");
1256 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1260 static struct seq_operations raw6_seq_ops = {
1261 .start = raw6_seq_start,
1262 .next = raw6_seq_next,
1263 .stop = raw6_seq_stop,
1264 .show = raw6_seq_show,
1267 static int raw6_seq_open(struct inode *inode, struct file *file)
1269 struct seq_file *seq;
1271 struct raw6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
1274 rc = seq_open(file, &raw6_seq_ops);
1277 seq = file->private_data;
1286 static struct file_operations raw6_seq_fops = {
1287 .owner = THIS_MODULE,
1288 .open = raw6_seq_open,
1290 .llseek = seq_lseek,
1291 .release = seq_release_private,
1294 int __init raw6_proc_init(void)
1296 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1301 void raw6_proc_exit(void)
1303 proc_net_remove("raw6");
1305 #endif /* CONFIG_PROC_FS */