btrfs: clear __GFP_FS flag in the space cache inode
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / raw.c
blobc5b0915d106bfe788cdf6d691e1c6c371831c86c
1 /*
2 * RAW sockets for IPv6
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Adapted from linux/net/ipv4/raw.c
10 * Fixes:
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>
39 #include <net/ip.h>
40 #include <net/sock.h>
41 #include <net/snmp.h>
43 #include <net/ipv6.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>
50 #include <net/udp.h>
51 #include <net/inet_common.h>
52 #include <net/tcp_states.h>
53 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
54 #include <net/mip6.h>
55 #endif
56 #include <linux/mroute6.h>
58 #include <net/raw.h>
59 #include <net/rawv6.h>
60 #include <net/xfrm.h>
62 #include <linux/proc_fs.h>
63 #include <linux/seq_file.h>
65 static struct raw_hashinfo raw_v6_hashinfo = {
66 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
69 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
70 unsigned short num, struct in6_addr *loc_addr,
71 struct in6_addr *rmt_addr, int dif)
73 struct hlist_node *node;
74 int is_multicast = ipv6_addr_is_multicast(loc_addr);
76 sk_for_each_from(sk, node)
77 if (inet_sk(sk)->inet_num == num) {
78 struct ipv6_pinfo *np = inet6_sk(sk);
80 if (!net_eq(sock_net(sk), net))
81 continue;
83 if (!ipv6_addr_any(&np->daddr) &&
84 !ipv6_addr_equal(&np->daddr, rmt_addr))
85 continue;
87 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
88 continue;
90 if (!ipv6_addr_any(&np->rcv_saddr)) {
91 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
92 goto found;
93 if (is_multicast &&
94 inet6_mc_check(sk, loc_addr, rmt_addr))
95 goto found;
96 continue;
98 goto found;
100 sk = NULL;
101 found:
102 return sk;
106 * 0 - deliver
107 * 1 - block
109 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
111 struct icmp6hdr *icmph;
112 struct raw6_sock *rp = raw6_sk(sk);
114 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
115 __u32 *data = &rp->filter.data[0];
116 int bit_nr;
118 icmph = (struct icmp6hdr *) skb->data;
119 bit_nr = icmph->icmp6_type;
121 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
123 return 0;
126 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
127 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb);
129 int rawv6_mh_filter_register(int (*filter)(struct sock *sock,
130 struct sk_buff *skb))
132 rcu_assign_pointer(mh_filter, filter);
133 return 0;
135 EXPORT_SYMBOL(rawv6_mh_filter_register);
137 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock,
138 struct sk_buff *skb))
140 rcu_assign_pointer(mh_filter, NULL);
141 synchronize_rcu();
142 return 0;
144 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
146 #endif
149 * demultiplex raw sockets.
150 * (should consider queueing the skb in the sock receive_queue
151 * without calling rawv6.c)
153 * Caller owns SKB so we must make clones.
155 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
157 struct in6_addr *saddr;
158 struct in6_addr *daddr;
159 struct sock *sk;
160 int delivered = 0;
161 __u8 hash;
162 struct net *net;
164 saddr = &ipv6_hdr(skb)->saddr;
165 daddr = saddr + 1;
167 hash = nexthdr & (MAX_INET_PROTOS - 1);
169 read_lock(&raw_v6_hashinfo.lock);
170 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
172 if (sk == NULL)
173 goto out;
175 net = dev_net(skb->dev);
176 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
178 while (sk) {
179 int filtered;
181 delivered = 1;
182 switch (nexthdr) {
183 case IPPROTO_ICMPV6:
184 filtered = icmpv6_filter(sk, skb);
185 break;
187 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
188 case IPPROTO_MH:
190 /* XXX: To validate MH only once for each packet,
191 * this is placed here. It should be after checking
192 * xfrm policy, however it doesn't. The checking xfrm
193 * policy is placed in rawv6_rcv() because it is
194 * required for each socket.
196 int (*filter)(struct sock *sock, struct sk_buff *skb);
198 filter = rcu_dereference(mh_filter);
199 filtered = filter ? filter(sk, skb) : 0;
200 break;
202 #endif
203 default:
204 filtered = 0;
205 break;
208 if (filtered < 0)
209 break;
210 if (filtered == 0) {
211 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
213 /* Not releasing hash table! */
214 if (clone) {
215 nf_reset(clone);
216 rawv6_rcv(sk, clone);
219 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
220 IP6CB(skb)->iif);
222 out:
223 read_unlock(&raw_v6_hashinfo.lock);
224 return delivered;
227 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
229 struct sock *raw_sk;
231 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
232 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
233 raw_sk = NULL;
235 return raw_sk != NULL;
238 /* This cleans up af_inet6 a bit. -DaveM */
239 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
241 struct inet_sock *inet = inet_sk(sk);
242 struct ipv6_pinfo *np = inet6_sk(sk);
243 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
244 __be32 v4addr = 0;
245 int addr_type;
246 int err;
248 if (addr_len < SIN6_LEN_RFC2133)
249 return -EINVAL;
250 addr_type = ipv6_addr_type(&addr->sin6_addr);
252 /* Raw sockets are IPv6 only */
253 if (addr_type == IPV6_ADDR_MAPPED)
254 return -EADDRNOTAVAIL;
256 lock_sock(sk);
258 err = -EINVAL;
259 if (sk->sk_state != TCP_CLOSE)
260 goto out;
262 rcu_read_lock();
263 /* Check if the address belongs to the host. */
264 if (addr_type != IPV6_ADDR_ANY) {
265 struct net_device *dev = NULL;
267 if (addr_type & IPV6_ADDR_LINKLOCAL) {
268 if (addr_len >= sizeof(struct sockaddr_in6) &&
269 addr->sin6_scope_id) {
270 /* Override any existing binding, if another
271 * one is supplied by user.
273 sk->sk_bound_dev_if = addr->sin6_scope_id;
276 /* Binding to link-local address requires an interface */
277 if (!sk->sk_bound_dev_if)
278 goto out_unlock;
280 err = -ENODEV;
281 dev = dev_get_by_index_rcu(sock_net(sk),
282 sk->sk_bound_dev_if);
283 if (!dev)
284 goto out_unlock;
287 /* ipv4 addr of the socket is invalid. Only the
288 * unspecified and mapped address have a v4 equivalent.
290 v4addr = LOOPBACK4_IPV6;
291 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
292 err = -EADDRNOTAVAIL;
293 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
294 dev, 0)) {
295 goto out_unlock;
300 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
301 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
302 if (!(addr_type & IPV6_ADDR_MULTICAST))
303 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
304 err = 0;
305 out_unlock:
306 rcu_read_unlock();
307 out:
308 release_sock(sk);
309 return err;
312 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
313 struct inet6_skb_parm *opt,
314 u8 type, u8 code, int offset, __be32 info)
316 struct inet_sock *inet = inet_sk(sk);
317 struct ipv6_pinfo *np = inet6_sk(sk);
318 int err;
319 int harderr;
321 /* Report error on raw socket, if:
322 1. User requested recverr.
323 2. Socket is connected (otherwise the error indication
324 is useless without recverr and error is hard.
326 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
327 return;
329 harderr = icmpv6_err_convert(type, code, &err);
330 if (type == ICMPV6_PKT_TOOBIG)
331 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
333 if (np->recverr) {
334 u8 *payload = skb->data;
335 if (!inet->hdrincl)
336 payload += offset;
337 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
340 if (np->recverr || harderr) {
341 sk->sk_err = err;
342 sk->sk_error_report(sk);
346 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
347 u8 type, u8 code, int inner_offset, __be32 info)
349 struct sock *sk;
350 int hash;
351 struct in6_addr *saddr, *daddr;
352 struct net *net;
354 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
356 read_lock(&raw_v6_hashinfo.lock);
357 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
358 if (sk != NULL) {
359 /* Note: ipv6_hdr(skb) != skb->data */
360 struct ipv6hdr *ip6h = (struct ipv6hdr *)skb->data;
361 saddr = &ip6h->saddr;
362 daddr = &ip6h->daddr;
363 net = dev_net(skb->dev);
365 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
366 IP6CB(skb)->iif))) {
367 rawv6_err(sk, skb, NULL, type, code,
368 inner_offset, info);
369 sk = sk_next(sk);
372 read_unlock(&raw_v6_hashinfo.lock);
375 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
377 if ((raw6_sk(sk)->checksum || rcu_dereference_raw(sk->sk_filter)) &&
378 skb_checksum_complete(skb)) {
379 atomic_inc(&sk->sk_drops);
380 kfree_skb(skb);
381 return NET_RX_DROP;
384 /* Charge it to the socket. */
385 if (ip_queue_rcv_skb(sk, skb) < 0) {
386 kfree_skb(skb);
387 return NET_RX_DROP;
390 return 0;
394 * This is next to useless...
395 * if we demultiplex in network layer we don't need the extra call
396 * just to queue the skb...
397 * maybe we could have the network decide upon a hint if it
398 * should call raw_rcv for demultiplexing
400 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
402 struct inet_sock *inet = inet_sk(sk);
403 struct raw6_sock *rp = raw6_sk(sk);
405 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
406 atomic_inc(&sk->sk_drops);
407 kfree_skb(skb);
408 return NET_RX_DROP;
411 if (!rp->checksum)
412 skb->ip_summed = CHECKSUM_UNNECESSARY;
414 if (skb->ip_summed == CHECKSUM_COMPLETE) {
415 skb_postpull_rcsum(skb, skb_network_header(skb),
416 skb_network_header_len(skb));
417 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
418 &ipv6_hdr(skb)->daddr,
419 skb->len, inet->inet_num, skb->csum))
420 skb->ip_summed = CHECKSUM_UNNECESSARY;
422 if (!skb_csum_unnecessary(skb))
423 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
424 &ipv6_hdr(skb)->daddr,
425 skb->len,
426 inet->inet_num, 0));
428 if (inet->hdrincl) {
429 if (skb_checksum_complete(skb)) {
430 atomic_inc(&sk->sk_drops);
431 kfree_skb(skb);
432 return NET_RX_DROP;
436 rawv6_rcv_skb(sk, skb);
437 return 0;
442 * This should be easy, if there is something there
443 * we return it, otherwise we block.
446 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
447 struct msghdr *msg, size_t len,
448 int noblock, int flags, int *addr_len)
450 struct ipv6_pinfo *np = inet6_sk(sk);
451 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
452 struct sk_buff *skb;
453 size_t copied;
454 int err;
456 if (flags & MSG_OOB)
457 return -EOPNOTSUPP;
459 if (addr_len)
460 *addr_len=sizeof(*sin6);
462 if (flags & MSG_ERRQUEUE)
463 return ipv6_recv_error(sk, msg, len);
465 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
466 return ipv6_recv_rxpmtu(sk, msg, len);
468 skb = skb_recv_datagram(sk, flags, noblock, &err);
469 if (!skb)
470 goto out;
472 copied = skb->len;
473 if (copied > len) {
474 copied = len;
475 msg->msg_flags |= MSG_TRUNC;
478 if (skb_csum_unnecessary(skb)) {
479 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
480 } else if (msg->msg_flags&MSG_TRUNC) {
481 if (__skb_checksum_complete(skb))
482 goto csum_copy_err;
483 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
484 } else {
485 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
486 if (err == -EINVAL)
487 goto csum_copy_err;
489 if (err)
490 goto out_free;
492 /* Copy the address. */
493 if (sin6) {
494 sin6->sin6_family = AF_INET6;
495 sin6->sin6_port = 0;
496 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
497 sin6->sin6_flowinfo = 0;
498 sin6->sin6_scope_id = 0;
499 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
500 sin6->sin6_scope_id = IP6CB(skb)->iif;
503 sock_recv_ts_and_drops(msg, sk, skb);
505 if (np->rxopt.all)
506 datagram_recv_ctl(sk, msg, skb);
508 err = copied;
509 if (flags & MSG_TRUNC)
510 err = skb->len;
512 out_free:
513 skb_free_datagram(sk, skb);
514 out:
515 return err;
517 csum_copy_err:
518 skb_kill_datagram(sk, skb, flags);
520 /* Error for blocking case is chosen to masquerade
521 as some normal condition.
523 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
524 goto out;
527 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
528 struct raw6_sock *rp)
530 struct sk_buff *skb;
531 int err = 0;
532 int offset;
533 int len;
534 int total_len;
535 __wsum tmp_csum;
536 __sum16 csum;
538 if (!rp->checksum)
539 goto send;
541 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
542 goto out;
544 offset = rp->offset;
545 total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
546 skb->data);
547 if (offset >= total_len - 1) {
548 err = -EINVAL;
549 ip6_flush_pending_frames(sk);
550 goto out;
553 /* should be check HW csum miyazawa */
554 if (skb_queue_len(&sk->sk_write_queue) == 1) {
556 * Only one fragment on the socket.
558 tmp_csum = skb->csum;
559 } else {
560 struct sk_buff *csum_skb = NULL;
561 tmp_csum = 0;
563 skb_queue_walk(&sk->sk_write_queue, skb) {
564 tmp_csum = csum_add(tmp_csum, skb->csum);
566 if (csum_skb)
567 continue;
569 len = skb->len - skb_transport_offset(skb);
570 if (offset >= len) {
571 offset -= len;
572 continue;
575 csum_skb = skb;
578 skb = csum_skb;
581 offset += skb_transport_offset(skb);
582 if (skb_copy_bits(skb, offset, &csum, 2))
583 BUG();
585 /* in case cksum was not initialized */
586 if (unlikely(csum))
587 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
589 csum = csum_ipv6_magic(&fl->fl6_src,
590 &fl->fl6_dst,
591 total_len, fl->proto, tmp_csum);
593 if (csum == 0 && fl->proto == IPPROTO_UDP)
594 csum = CSUM_MANGLED_0;
596 if (skb_store_bits(skb, offset, &csum, 2))
597 BUG();
599 send:
600 err = ip6_push_pending_frames(sk);
601 out:
602 return err;
605 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
606 struct flowi *fl, struct dst_entry **dstp,
607 unsigned int flags)
609 struct ipv6_pinfo *np = inet6_sk(sk);
610 struct ipv6hdr *iph;
611 struct sk_buff *skb;
612 int err;
613 struct rt6_info *rt = (struct rt6_info *)*dstp;
615 if (length > rt->dst.dev->mtu) {
616 ipv6_local_error(sk, EMSGSIZE, fl, rt->dst.dev->mtu);
617 return -EMSGSIZE;
619 if (flags&MSG_PROBE)
620 goto out;
622 skb = sock_alloc_send_skb(sk,
623 length + LL_ALLOCATED_SPACE(rt->dst.dev) + 15,
624 flags & MSG_DONTWAIT, &err);
625 if (skb == NULL)
626 goto error;
627 skb_reserve(skb, LL_RESERVED_SPACE(rt->dst.dev));
629 skb->priority = sk->sk_priority;
630 skb->mark = sk->sk_mark;
631 skb_dst_set(skb, &rt->dst);
632 *dstp = NULL;
634 skb_put(skb, length);
635 skb_reset_network_header(skb);
636 iph = ipv6_hdr(skb);
638 skb->ip_summed = CHECKSUM_NONE;
640 skb->transport_header = skb->network_header;
641 err = memcpy_fromiovecend((void *)iph, from, 0, length);
642 if (err)
643 goto error_fault;
645 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
646 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
647 rt->dst.dev, dst_output);
648 if (err > 0)
649 err = net_xmit_errno(err);
650 if (err)
651 goto error;
652 out:
653 return 0;
655 error_fault:
656 err = -EFAULT;
657 kfree_skb(skb);
658 error:
659 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
660 if (err == -ENOBUFS && !np->recverr)
661 err = 0;
662 return err;
665 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
667 struct iovec *iov;
668 u8 __user *type = NULL;
669 u8 __user *code = NULL;
670 u8 len = 0;
671 int probed = 0;
672 int i;
674 if (!msg->msg_iov)
675 return 0;
677 for (i = 0; i < msg->msg_iovlen; i++) {
678 iov = &msg->msg_iov[i];
679 if (!iov)
680 continue;
682 switch (fl->proto) {
683 case IPPROTO_ICMPV6:
684 /* check if one-byte field is readable or not. */
685 if (iov->iov_base && iov->iov_len < 1)
686 break;
688 if (!type) {
689 type = iov->iov_base;
690 /* check if code field is readable or not. */
691 if (iov->iov_len > 1)
692 code = type + 1;
693 } else if (!code)
694 code = iov->iov_base;
696 if (type && code) {
697 if (get_user(fl->fl_icmp_type, type) ||
698 get_user(fl->fl_icmp_code, code))
699 return -EFAULT;
700 probed = 1;
702 break;
703 case IPPROTO_MH:
704 if (iov->iov_base && iov->iov_len < 1)
705 break;
706 /* check if type field is readable or not. */
707 if (iov->iov_len > 2 - len) {
708 u8 __user *p = iov->iov_base;
709 if (get_user(fl->fl_mh_type, &p[2 - len]))
710 return -EFAULT;
711 probed = 1;
712 } else
713 len += iov->iov_len;
715 break;
716 default:
717 probed = 1;
718 break;
720 if (probed)
721 break;
723 return 0;
726 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
727 struct msghdr *msg, size_t len)
729 struct ipv6_txoptions opt_space;
730 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
731 struct in6_addr *daddr, *final_p, final;
732 struct inet_sock *inet = inet_sk(sk);
733 struct ipv6_pinfo *np = inet6_sk(sk);
734 struct raw6_sock *rp = raw6_sk(sk);
735 struct ipv6_txoptions *opt = NULL;
736 struct ip6_flowlabel *flowlabel = NULL;
737 struct dst_entry *dst = NULL;
738 struct flowi fl;
739 int addr_len = msg->msg_namelen;
740 int hlimit = -1;
741 int tclass = -1;
742 int dontfrag = -1;
743 u16 proto;
744 int err;
746 /* Rough check on arithmetic overflow,
747 better check is made in ip6_append_data().
749 if (len > INT_MAX)
750 return -EMSGSIZE;
752 /* Mirror BSD error message compatibility */
753 if (msg->msg_flags & MSG_OOB)
754 return -EOPNOTSUPP;
757 * Get and verify the address.
759 memset(&fl, 0, sizeof(fl));
761 fl.mark = sk->sk_mark;
763 if (sin6) {
764 if (addr_len < SIN6_LEN_RFC2133)
765 return -EINVAL;
767 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
768 return -EAFNOSUPPORT;
770 /* port is the proto value [0..255] carried in nexthdr */
771 proto = ntohs(sin6->sin6_port);
773 if (!proto)
774 proto = inet->inet_num;
775 else if (proto != inet->inet_num)
776 return -EINVAL;
778 if (proto > 255)
779 return -EINVAL;
781 daddr = &sin6->sin6_addr;
782 if (np->sndflow) {
783 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
784 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
785 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
786 if (flowlabel == NULL)
787 return -EINVAL;
788 daddr = &flowlabel->dst;
793 * Otherwise it will be difficult to maintain
794 * sk->sk_dst_cache.
796 if (sk->sk_state == TCP_ESTABLISHED &&
797 ipv6_addr_equal(daddr, &np->daddr))
798 daddr = &np->daddr;
800 if (addr_len >= sizeof(struct sockaddr_in6) &&
801 sin6->sin6_scope_id &&
802 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
803 fl.oif = sin6->sin6_scope_id;
804 } else {
805 if (sk->sk_state != TCP_ESTABLISHED)
806 return -EDESTADDRREQ;
808 proto = inet->inet_num;
809 daddr = &np->daddr;
810 fl.fl6_flowlabel = np->flow_label;
813 if (fl.oif == 0)
814 fl.oif = sk->sk_bound_dev_if;
816 if (msg->msg_controllen) {
817 opt = &opt_space;
818 memset(opt, 0, sizeof(struct ipv6_txoptions));
819 opt->tot_len = sizeof(struct ipv6_txoptions);
821 err = datagram_send_ctl(sock_net(sk), msg, &fl, opt, &hlimit,
822 &tclass, &dontfrag);
823 if (err < 0) {
824 fl6_sock_release(flowlabel);
825 return err;
827 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
828 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
829 if (flowlabel == NULL)
830 return -EINVAL;
832 if (!(opt->opt_nflen|opt->opt_flen))
833 opt = NULL;
835 if (opt == NULL)
836 opt = np->opt;
837 if (flowlabel)
838 opt = fl6_merge_options(&opt_space, flowlabel, opt);
839 opt = ipv6_fixup_options(&opt_space, opt);
841 fl.proto = proto;
842 err = rawv6_probe_proto_opt(&fl, msg);
843 if (err)
844 goto out;
846 if (!ipv6_addr_any(daddr))
847 ipv6_addr_copy(&fl.fl6_dst, daddr);
848 else
849 fl.fl6_dst.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
850 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
851 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
853 final_p = fl6_update_dst(&fl, opt, &final);
855 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
856 fl.oif = np->mcast_oif;
857 security_sk_classify_flow(sk, &fl);
859 err = ip6_dst_lookup(sk, &dst, &fl);
860 if (err)
861 goto out;
862 if (final_p)
863 ipv6_addr_copy(&fl.fl6_dst, final_p);
865 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
866 if (err < 0) {
867 if (err == -EREMOTE)
868 err = ip6_dst_blackhole(sk, &dst, &fl);
869 if (err < 0)
870 goto out;
873 if (hlimit < 0) {
874 if (ipv6_addr_is_multicast(&fl.fl6_dst))
875 hlimit = np->mcast_hops;
876 else
877 hlimit = np->hop_limit;
878 if (hlimit < 0)
879 hlimit = ip6_dst_hoplimit(dst);
882 if (tclass < 0)
883 tclass = np->tclass;
885 if (dontfrag < 0)
886 dontfrag = np->dontfrag;
888 if (msg->msg_flags&MSG_CONFIRM)
889 goto do_confirm;
891 back_from_confirm:
892 if (inet->hdrincl)
893 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, &dst, msg->msg_flags);
894 else {
895 lock_sock(sk);
896 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
897 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
898 msg->msg_flags, dontfrag);
900 if (err)
901 ip6_flush_pending_frames(sk);
902 else if (!(msg->msg_flags & MSG_MORE))
903 err = rawv6_push_pending_frames(sk, &fl, rp);
904 release_sock(sk);
906 done:
907 dst_release(dst);
908 out:
909 fl6_sock_release(flowlabel);
910 return err<0?err:len;
911 do_confirm:
912 dst_confirm(dst);
913 if (!(msg->msg_flags & MSG_PROBE) || len)
914 goto back_from_confirm;
915 err = 0;
916 goto done;
919 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
920 char __user *optval, int optlen)
922 switch (optname) {
923 case ICMPV6_FILTER:
924 if (optlen > sizeof(struct icmp6_filter))
925 optlen = sizeof(struct icmp6_filter);
926 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
927 return -EFAULT;
928 return 0;
929 default:
930 return -ENOPROTOOPT;
933 return 0;
936 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
937 char __user *optval, int __user *optlen)
939 int len;
941 switch (optname) {
942 case ICMPV6_FILTER:
943 if (get_user(len, optlen))
944 return -EFAULT;
945 if (len < 0)
946 return -EINVAL;
947 if (len > sizeof(struct icmp6_filter))
948 len = sizeof(struct icmp6_filter);
949 if (put_user(len, optlen))
950 return -EFAULT;
951 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
952 return -EFAULT;
953 return 0;
954 default:
955 return -ENOPROTOOPT;
958 return 0;
962 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
963 char __user *optval, unsigned int optlen)
965 struct raw6_sock *rp = raw6_sk(sk);
966 int val;
968 if (get_user(val, (int __user *)optval))
969 return -EFAULT;
971 switch (optname) {
972 case IPV6_CHECKSUM:
973 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
974 level == IPPROTO_IPV6) {
976 * RFC3542 tells that IPV6_CHECKSUM socket
977 * option in the IPPROTO_IPV6 level is not
978 * allowed on ICMPv6 sockets.
979 * If you want to set it, use IPPROTO_RAW
980 * level IPV6_CHECKSUM socket option
981 * (Linux extension).
983 return -EINVAL;
986 /* You may get strange result with a positive odd offset;
987 RFC2292bis agrees with me. */
988 if (val > 0 && (val&1))
989 return -EINVAL;
990 if (val < 0) {
991 rp->checksum = 0;
992 } else {
993 rp->checksum = 1;
994 rp->offset = val;
997 return 0;
998 break;
1000 default:
1001 return -ENOPROTOOPT;
1005 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1006 char __user *optval, unsigned int optlen)
1008 switch(level) {
1009 case SOL_RAW:
1010 break;
1012 case SOL_ICMPV6:
1013 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1014 return -EOPNOTSUPP;
1015 return rawv6_seticmpfilter(sk, level, optname, optval,
1016 optlen);
1017 case SOL_IPV6:
1018 if (optname == IPV6_CHECKSUM)
1019 break;
1020 default:
1021 return ipv6_setsockopt(sk, level, optname, optval,
1022 optlen);
1025 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1028 #ifdef CONFIG_COMPAT
1029 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1030 char __user *optval, unsigned int optlen)
1032 switch (level) {
1033 case SOL_RAW:
1034 break;
1035 case SOL_ICMPV6:
1036 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1037 return -EOPNOTSUPP;
1038 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1039 case SOL_IPV6:
1040 if (optname == IPV6_CHECKSUM)
1041 break;
1042 default:
1043 return compat_ipv6_setsockopt(sk, level, optname,
1044 optval, optlen);
1046 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1048 #endif
1050 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1051 char __user *optval, int __user *optlen)
1053 struct raw6_sock *rp = raw6_sk(sk);
1054 int val, len;
1056 if (get_user(len,optlen))
1057 return -EFAULT;
1059 switch (optname) {
1060 case IPV6_CHECKSUM:
1062 * We allow getsockopt() for IPPROTO_IPV6-level
1063 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1064 * since RFC3542 is silent about it.
1066 if (rp->checksum == 0)
1067 val = -1;
1068 else
1069 val = rp->offset;
1070 break;
1072 default:
1073 return -ENOPROTOOPT;
1076 len = min_t(unsigned int, sizeof(int), len);
1078 if (put_user(len, optlen))
1079 return -EFAULT;
1080 if (copy_to_user(optval,&val,len))
1081 return -EFAULT;
1082 return 0;
1085 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1086 char __user *optval, int __user *optlen)
1088 switch(level) {
1089 case SOL_RAW:
1090 break;
1092 case SOL_ICMPV6:
1093 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1094 return -EOPNOTSUPP;
1095 return rawv6_geticmpfilter(sk, level, optname, optval,
1096 optlen);
1097 case SOL_IPV6:
1098 if (optname == IPV6_CHECKSUM)
1099 break;
1100 default:
1101 return ipv6_getsockopt(sk, level, optname, optval,
1102 optlen);
1105 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1108 #ifdef CONFIG_COMPAT
1109 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1110 char __user *optval, int __user *optlen)
1112 switch (level) {
1113 case SOL_RAW:
1114 break;
1115 case SOL_ICMPV6:
1116 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1117 return -EOPNOTSUPP;
1118 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1119 case SOL_IPV6:
1120 if (optname == IPV6_CHECKSUM)
1121 break;
1122 default:
1123 return compat_ipv6_getsockopt(sk, level, optname,
1124 optval, optlen);
1126 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1128 #endif
1130 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1132 switch(cmd) {
1133 case SIOCOUTQ:
1135 int amount = sk_wmem_alloc_get(sk);
1137 return put_user(amount, (int __user *)arg);
1139 case SIOCINQ:
1141 struct sk_buff *skb;
1142 int amount = 0;
1144 spin_lock_bh(&sk->sk_receive_queue.lock);
1145 skb = skb_peek(&sk->sk_receive_queue);
1146 if (skb != NULL)
1147 amount = skb->tail - skb->transport_header;
1148 spin_unlock_bh(&sk->sk_receive_queue.lock);
1149 return put_user(amount, (int __user *)arg);
1152 default:
1153 #ifdef CONFIG_IPV6_MROUTE
1154 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1155 #else
1156 return -ENOIOCTLCMD;
1157 #endif
1161 #ifdef CONFIG_COMPAT
1162 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1164 switch (cmd) {
1165 case SIOCOUTQ:
1166 case SIOCINQ:
1167 return -ENOIOCTLCMD;
1168 default:
1169 #ifdef CONFIG_IPV6_MROUTE
1170 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1171 #else
1172 return -ENOIOCTLCMD;
1173 #endif
1176 #endif
1178 static void rawv6_close(struct sock *sk, long timeout)
1180 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1181 ip6_ra_control(sk, -1);
1182 ip6mr_sk_done(sk);
1183 sk_common_release(sk);
1186 static void raw6_destroy(struct sock *sk)
1188 lock_sock(sk);
1189 ip6_flush_pending_frames(sk);
1190 release_sock(sk);
1192 inet6_destroy_sock(sk);
1195 static int rawv6_init_sk(struct sock *sk)
1197 struct raw6_sock *rp = raw6_sk(sk);
1199 switch (inet_sk(sk)->inet_num) {
1200 case IPPROTO_ICMPV6:
1201 rp->checksum = 1;
1202 rp->offset = 2;
1203 break;
1204 case IPPROTO_MH:
1205 rp->checksum = 1;
1206 rp->offset = 4;
1207 break;
1208 default:
1209 break;
1211 return 0;
1214 struct proto rawv6_prot = {
1215 .name = "RAWv6",
1216 .owner = THIS_MODULE,
1217 .close = rawv6_close,
1218 .destroy = raw6_destroy,
1219 .connect = ip6_datagram_connect,
1220 .disconnect = udp_disconnect,
1221 .ioctl = rawv6_ioctl,
1222 .init = rawv6_init_sk,
1223 .setsockopt = rawv6_setsockopt,
1224 .getsockopt = rawv6_getsockopt,
1225 .sendmsg = rawv6_sendmsg,
1226 .recvmsg = rawv6_recvmsg,
1227 .bind = rawv6_bind,
1228 .backlog_rcv = rawv6_rcv_skb,
1229 .hash = raw_hash_sk,
1230 .unhash = raw_unhash_sk,
1231 .obj_size = sizeof(struct raw6_sock),
1232 .h.raw_hash = &raw_v6_hashinfo,
1233 #ifdef CONFIG_COMPAT
1234 .compat_setsockopt = compat_rawv6_setsockopt,
1235 .compat_getsockopt = compat_rawv6_getsockopt,
1236 .compat_ioctl = compat_rawv6_ioctl,
1237 #endif
1240 #ifdef CONFIG_PROC_FS
1241 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1243 struct ipv6_pinfo *np = inet6_sk(sp);
1244 struct in6_addr *dest, *src;
1245 __u16 destp, srcp;
1247 dest = &np->daddr;
1248 src = &np->rcv_saddr;
1249 destp = 0;
1250 srcp = inet_sk(sp)->inet_num;
1251 seq_printf(seq,
1252 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1253 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1255 src->s6_addr32[0], src->s6_addr32[1],
1256 src->s6_addr32[2], src->s6_addr32[3], srcp,
1257 dest->s6_addr32[0], dest->s6_addr32[1],
1258 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1259 sp->sk_state,
1260 sk_wmem_alloc_get(sp),
1261 sk_rmem_alloc_get(sp),
1262 0, 0L, 0,
1263 sock_i_uid(sp), 0,
1264 sock_i_ino(sp),
1265 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1268 static int raw6_seq_show(struct seq_file *seq, void *v)
1270 if (v == SEQ_START_TOKEN)
1271 seq_printf(seq,
1272 " sl "
1273 "local_address "
1274 "remote_address "
1275 "st tx_queue rx_queue tr tm->when retrnsmt"
1276 " uid timeout inode ref pointer drops\n");
1277 else
1278 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1279 return 0;
1282 static const struct seq_operations raw6_seq_ops = {
1283 .start = raw_seq_start,
1284 .next = raw_seq_next,
1285 .stop = raw_seq_stop,
1286 .show = raw6_seq_show,
1289 static int raw6_seq_open(struct inode *inode, struct file *file)
1291 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1294 static const struct file_operations raw6_seq_fops = {
1295 .owner = THIS_MODULE,
1296 .open = raw6_seq_open,
1297 .read = seq_read,
1298 .llseek = seq_lseek,
1299 .release = seq_release_net,
1302 static int __net_init raw6_init_net(struct net *net)
1304 if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1305 return -ENOMEM;
1307 return 0;
1310 static void __net_exit raw6_exit_net(struct net *net)
1312 proc_net_remove(net, "raw6");
1315 static struct pernet_operations raw6_net_ops = {
1316 .init = raw6_init_net,
1317 .exit = raw6_exit_net,
1320 int __init raw6_proc_init(void)
1322 return register_pernet_subsys(&raw6_net_ops);
1325 void raw6_proc_exit(void)
1327 unregister_pernet_subsys(&raw6_net_ops);
1329 #endif /* CONFIG_PROC_FS */
1331 /* Same as inet6_dgram_ops, sans udp_poll. */
1332 static const struct proto_ops inet6_sockraw_ops = {
1333 .family = PF_INET6,
1334 .owner = THIS_MODULE,
1335 .release = inet6_release,
1336 .bind = inet6_bind,
1337 .connect = inet_dgram_connect, /* ok */
1338 .socketpair = sock_no_socketpair, /* a do nothing */
1339 .accept = sock_no_accept, /* a do nothing */
1340 .getname = inet6_getname,
1341 .poll = datagram_poll, /* ok */
1342 .ioctl = inet6_ioctl, /* must change */
1343 .listen = sock_no_listen, /* ok */
1344 .shutdown = inet_shutdown, /* ok */
1345 .setsockopt = sock_common_setsockopt, /* ok */
1346 .getsockopt = sock_common_getsockopt, /* ok */
1347 .sendmsg = inet_sendmsg, /* ok */
1348 .recvmsg = sock_common_recvmsg, /* ok */
1349 .mmap = sock_no_mmap,
1350 .sendpage = sock_no_sendpage,
1351 #ifdef CONFIG_COMPAT
1352 .compat_setsockopt = compat_sock_common_setsockopt,
1353 .compat_getsockopt = compat_sock_common_getsockopt,
1354 #endif
1357 static struct inet_protosw rawv6_protosw = {
1358 .type = SOCK_RAW,
1359 .protocol = IPPROTO_IP, /* wild card */
1360 .prot = &rawv6_prot,
1361 .ops = &inet6_sockraw_ops,
1362 .no_check = UDP_CSUM_DEFAULT,
1363 .flags = INET_PROTOSW_REUSE,
1366 int __init rawv6_init(void)
1368 int ret;
1370 ret = inet6_register_protosw(&rawv6_protosw);
1371 if (ret)
1372 goto out;
1373 out:
1374 return ret;
1377 void rawv6_exit(void)
1379 inet6_unregister_protosw(&rawv6_protosw);