Linux 5.6-rc4
[linux/fpc-iii.git] / net / ipv6 / raw.c
blobdfe5e603ffe16b61d2c58088e52e71d736e5d89f
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * RAW sockets for IPv6
4 * Linux INET6 implementation
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
9 * Adapted from linux/net/ipv4/raw.c
11 * Fixes:
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>
35 #include <net/ip.h>
36 #include <net/sock.h>
37 #include <net/snmp.h>
39 #include <net/ipv6.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>
46 #include <net/udp.h>
47 #include <net/inet_common.h>
48 #include <net/tcp_states.h>
49 #if IS_ENABLED(CONFIG_IPV6_MIP6)
50 #include <net/mip6.h>
51 #endif
52 #include <linux/mroute6.h>
54 #include <net/raw.h>
55 #include <net/rawv6.h>
56 #include <net/xfrm.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 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
67 EXPORT_SYMBOL_GPL(raw_v6_hashinfo);
69 struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
70 unsigned short num, const struct in6_addr *loc_addr,
71 const struct in6_addr *rmt_addr, int dif, int sdif)
73 bool is_multicast = ipv6_addr_is_multicast(loc_addr);
75 sk_for_each_from(sk)
76 if (inet_sk(sk)->inet_num == num) {
78 if (!net_eq(sock_net(sk), net))
79 continue;
81 if (!ipv6_addr_any(&sk->sk_v6_daddr) &&
82 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr))
83 continue;
85 if (!raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if,
86 dif, sdif))
87 continue;
89 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
90 if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))
91 goto found;
92 if (is_multicast &&
93 inet6_mc_check(sk, loc_addr, rmt_addr))
94 goto found;
95 continue;
97 goto found;
99 sk = NULL;
100 found:
101 return sk;
103 EXPORT_SYMBOL_GPL(__raw_v6_lookup);
106 * 0 - deliver
107 * 1 - block
109 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
111 struct icmp6hdr _hdr;
112 const struct icmp6hdr *hdr;
114 /* We require only the four bytes of the ICMPv6 header, not any
115 * additional bytes of message body in "struct icmp6hdr".
117 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
118 ICMPV6_HDRLEN, &_hdr);
119 if (hdr) {
120 const __u32 *data = &raw6_sk(sk)->filter.data[0];
121 unsigned int type = hdr->icmp6_type;
123 return (data[type >> 5] & (1U << (type & 31))) != 0;
125 return 1;
128 #if IS_ENABLED(CONFIG_IPV6_MIP6)
129 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
131 static mh_filter_t __rcu *mh_filter __read_mostly;
133 int rawv6_mh_filter_register(mh_filter_t filter)
135 rcu_assign_pointer(mh_filter, filter);
136 return 0;
138 EXPORT_SYMBOL(rawv6_mh_filter_register);
140 int rawv6_mh_filter_unregister(mh_filter_t filter)
142 RCU_INIT_POINTER(mh_filter, NULL);
143 synchronize_rcu();
144 return 0;
146 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
148 #endif
151 * demultiplex raw sockets.
152 * (should consider queueing the skb in the sock receive_queue
153 * without calling rawv6.c)
155 * Caller owns SKB so we must make clones.
157 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
159 const struct in6_addr *saddr;
160 const struct in6_addr *daddr;
161 struct sock *sk;
162 bool delivered = false;
163 __u8 hash;
164 struct net *net;
166 saddr = &ipv6_hdr(skb)->saddr;
167 daddr = saddr + 1;
169 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
171 read_lock(&raw_v6_hashinfo.lock);
172 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
174 if (!sk)
175 goto out;
177 net = dev_net(skb->dev);
178 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr,
179 inet6_iif(skb), inet6_sdif(skb));
181 while (sk) {
182 int filtered;
184 delivered = true;
185 switch (nexthdr) {
186 case IPPROTO_ICMPV6:
187 filtered = icmpv6_filter(sk, skb);
188 break;
190 #if IS_ENABLED(CONFIG_IPV6_MIP6)
191 case IPPROTO_MH:
193 /* XXX: To validate MH only once for each packet,
194 * this is placed here. It should be after checking
195 * xfrm policy, however it doesn't. The checking xfrm
196 * policy is placed in rawv6_rcv() because it is
197 * required for each socket.
199 mh_filter_t *filter;
201 filter = rcu_dereference(mh_filter);
202 filtered = filter ? (*filter)(sk, skb) : 0;
203 break;
205 #endif
206 default:
207 filtered = 0;
208 break;
211 if (filtered < 0)
212 break;
213 if (filtered == 0) {
214 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
216 /* Not releasing hash table! */
217 if (clone) {
218 nf_reset_ct(clone);
219 rawv6_rcv(sk, clone);
222 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
223 inet6_iif(skb), inet6_sdif(skb));
225 out:
226 read_unlock(&raw_v6_hashinfo.lock);
227 return delivered;
230 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
232 struct sock *raw_sk;
234 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]);
235 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
236 raw_sk = NULL;
238 return raw_sk != NULL;
241 /* This cleans up af_inet6 a bit. -DaveM */
242 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
244 struct inet_sock *inet = inet_sk(sk);
245 struct ipv6_pinfo *np = inet6_sk(sk);
246 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
247 __be32 v4addr = 0;
248 int addr_type;
249 int err;
251 if (addr_len < SIN6_LEN_RFC2133)
252 return -EINVAL;
254 if (addr->sin6_family != AF_INET6)
255 return -EINVAL;
257 addr_type = ipv6_addr_type(&addr->sin6_addr);
259 /* Raw sockets are IPv6 only */
260 if (addr_type == IPV6_ADDR_MAPPED)
261 return -EADDRNOTAVAIL;
263 lock_sock(sk);
265 err = -EINVAL;
266 if (sk->sk_state != TCP_CLOSE)
267 goto out;
269 rcu_read_lock();
270 /* Check if the address belongs to the host. */
271 if (addr_type != IPV6_ADDR_ANY) {
272 struct net_device *dev = NULL;
274 if (__ipv6_addr_needs_scope_id(addr_type)) {
275 if (addr_len >= sizeof(struct sockaddr_in6) &&
276 addr->sin6_scope_id) {
277 /* Override any existing binding, if another
278 * one is supplied by user.
280 sk->sk_bound_dev_if = addr->sin6_scope_id;
283 /* Binding to link-local address requires an interface */
284 if (!sk->sk_bound_dev_if)
285 goto out_unlock;
288 if (sk->sk_bound_dev_if) {
289 err = -ENODEV;
290 dev = dev_get_by_index_rcu(sock_net(sk),
291 sk->sk_bound_dev_if);
292 if (!dev)
293 goto out_unlock;
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,
304 dev, 0)) {
305 goto out_unlock;
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;
314 err = 0;
315 out_unlock:
316 rcu_read_unlock();
317 out:
318 release_sock(sk);
319 return err;
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);
328 int err;
329 int harderr;
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)
337 return;
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);
346 return;
348 if (np->recverr) {
349 u8 *payload = skb->data;
350 if (!inet->hdrincl)
351 payload += offset;
352 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
355 if (np->recverr || harderr) {
356 sk->sk_err = err;
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)
364 struct sock *sk;
365 int hash;
366 const struct in6_addr *saddr, *daddr;
367 struct net *net;
369 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
371 read_lock(&raw_v6_hashinfo.lock);
372 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
373 if (sk) {
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,
381 inet6_iif(skb), inet6_iif(skb)))) {
382 rawv6_err(sk, skb, NULL, type, code,
383 inner_offset, info);
384 sk = sk_next(sk);
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);
395 kfree_skb(skb);
396 return NET_RX_DROP;
399 /* Charge it to the socket. */
400 skb_dst_drop(skb);
401 if (sock_queue_rcv_skb(sk, skb) < 0) {
402 kfree_skb(skb);
403 return NET_RX_DROP;
406 return 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);
423 kfree_skb(skb);
424 return NET_RX_DROP;
427 if (!rp->checksum)
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,
441 skb->len,
442 inet->inet_num, 0));
444 if (inet->hdrincl) {
445 if (skb_checksum_complete(skb)) {
446 atomic_inc(&sk->sk_drops);
447 kfree_skb(skb);
448 return NET_RX_DROP;
452 rawv6_rcv_skb(sk, skb);
453 return 0;
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);
467 struct sk_buff *skb;
468 size_t copied;
469 int err;
471 if (flags & MSG_OOB)
472 return -EOPNOTSUPP;
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);
481 if (!skb)
482 goto out;
484 copied = skb->len;
485 if (copied > len) {
486 copied = len;
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))
494 goto csum_copy_err;
495 err = skb_copy_datagram_msg(skb, 0, msg, copied);
496 } else {
497 err = skb_copy_and_csum_datagram_msg(skb, 0, msg);
498 if (err == -EINVAL)
499 goto csum_copy_err;
501 if (err)
502 goto out_free;
504 /* Copy the address. */
505 if (sin6) {
506 sin6->sin6_family = AF_INET6;
507 sin6->sin6_port = 0;
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,
511 inet6_iif(skb));
512 *addr_len = sizeof(*sin6);
515 sock_recv_ts_and_drops(msg, sk, skb);
517 if (np->rxopt.all)
518 ip6_datagram_recv_ctl(sk, msg, skb);
520 err = copied;
521 if (flags & MSG_TRUNC)
522 err = skb->len;
524 out_free:
525 skb_free_datagram(sk, skb);
526 out:
527 return err;
529 csum_copy_err:
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;
536 goto out;
539 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
540 struct raw6_sock *rp)
542 struct sk_buff *skb;
543 int err = 0;
544 int offset;
545 int len;
546 int total_len;
547 __wsum tmp_csum;
548 __sum16 csum;
550 if (!rp->checksum)
551 goto send;
553 skb = skb_peek(&sk->sk_write_queue);
554 if (!skb)
555 goto out;
557 offset = rp->offset;
558 total_len = inet_sk(sk)->cork.base.length;
559 if (offset >= total_len - 1) {
560 err = -EINVAL;
561 ip6_flush_pending_frames(sk);
562 goto out;
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;
571 } else {
572 struct sk_buff *csum_skb = NULL;
573 tmp_csum = 0;
575 skb_queue_walk(&sk->sk_write_queue, skb) {
576 tmp_csum = csum_add(tmp_csum, skb->csum);
578 if (csum_skb)
579 continue;
581 len = skb->len - skb_transport_offset(skb);
582 if (offset >= len) {
583 offset -= len;
584 continue;
587 csum_skb = skb;
590 skb = csum_skb;
593 offset += skb_transport_offset(skb);
594 err = skb_copy_bits(skb, offset, &csum, 2);
595 if (err < 0) {
596 ip6_flush_pending_frames(sk);
597 goto out;
600 /* in case cksum was not initialized */
601 if (unlikely(csum))
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));
612 send:
613 err = ip6_push_pending_frames(sk);
614 out:
615 return err;
618 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length,
619 struct flowi6 *fl6, struct dst_entry **dstp,
620 unsigned int flags, const struct sockcm_cookie *sockc)
622 struct ipv6_pinfo *np = inet6_sk(sk);
623 struct net *net = sock_net(sk);
624 struct ipv6hdr *iph;
625 struct sk_buff *skb;
626 int err;
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);
633 return -EMSGSIZE;
635 if (length < sizeof(struct ipv6hdr))
636 return -EINVAL;
637 if (flags&MSG_PROBE)
638 goto out;
640 skb = sock_alloc_send_skb(sk,
641 length + hlen + tlen + 15,
642 flags & MSG_DONTWAIT, &err);
643 if (!skb)
644 goto error;
645 skb_reserve(skb, hlen);
647 skb->protocol = htons(ETH_P_IPV6);
648 skb->priority = sk->sk_priority;
649 skb->mark = sockc->mark;
650 skb->tstamp = sockc->transmit_time;
652 skb_put(skb, length);
653 skb_reset_network_header(skb);
654 iph = ipv6_hdr(skb);
656 skb->ip_summed = CHECKSUM_NONE;
658 skb_setup_tx_timestamp(skb, sockc->tsflags);
660 if (flags & MSG_CONFIRM)
661 skb_set_dst_pending_confirm(skb, 1);
663 skb->transport_header = skb->network_header;
664 err = memcpy_from_msg(iph, msg, length);
665 if (err) {
666 err = -EFAULT;
667 kfree_skb(skb);
668 goto error;
671 skb_dst_set(skb, &rt->dst);
672 *dstp = NULL;
674 /* if egress device is enslaved to an L3 master device pass the
675 * skb to its handler for processing
677 skb = l3mdev_ip6_out(sk, skb);
678 if (unlikely(!skb))
679 return 0;
681 /* Acquire rcu_read_lock() in case we need to use rt->rt6i_idev
682 * in the error path. Since skb has been freed, the dst could
683 * have been queued for deletion.
685 rcu_read_lock();
686 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
687 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
688 NULL, rt->dst.dev, dst_output);
689 if (err > 0)
690 err = net_xmit_errno(err);
691 if (err) {
692 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
693 rcu_read_unlock();
694 goto error_check;
696 rcu_read_unlock();
697 out:
698 return 0;
700 error:
701 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
702 error_check:
703 if (err == -ENOBUFS && !np->recverr)
704 err = 0;
705 return err;
708 struct raw6_frag_vec {
709 struct msghdr *msg;
710 int hlen;
711 char c[4];
714 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6)
716 int err = 0;
717 switch (fl6->flowi6_proto) {
718 case IPPROTO_ICMPV6:
719 rfv->hlen = 2;
720 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
721 if (!err) {
722 fl6->fl6_icmp_type = rfv->c[0];
723 fl6->fl6_icmp_code = rfv->c[1];
725 break;
726 case IPPROTO_MH:
727 rfv->hlen = 4;
728 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
729 if (!err)
730 fl6->fl6_mh_type = rfv->c[2];
732 return err;
735 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd,
736 struct sk_buff *skb)
738 struct raw6_frag_vec *rfv = from;
740 if (offset < rfv->hlen) {
741 int copy = min(rfv->hlen - offset, len);
743 if (skb->ip_summed == CHECKSUM_PARTIAL)
744 memcpy(to, rfv->c + offset, copy);
745 else
746 skb->csum = csum_block_add(
747 skb->csum,
748 csum_partial_copy_nocheck(rfv->c + offset,
749 to, copy, 0),
750 odd);
752 odd = 0;
753 offset += copy;
754 to += copy;
755 len -= copy;
757 if (!len)
758 return 0;
761 offset -= rfv->hlen;
763 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
766 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
768 struct ipv6_txoptions *opt_to_free = NULL;
769 struct ipv6_txoptions opt_space;
770 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
771 struct in6_addr *daddr, *final_p, final;
772 struct inet_sock *inet = inet_sk(sk);
773 struct ipv6_pinfo *np = inet6_sk(sk);
774 struct raw6_sock *rp = raw6_sk(sk);
775 struct ipv6_txoptions *opt = NULL;
776 struct ip6_flowlabel *flowlabel = NULL;
777 struct dst_entry *dst = NULL;
778 struct raw6_frag_vec rfv;
779 struct flowi6 fl6;
780 struct ipcm6_cookie ipc6;
781 int addr_len = msg->msg_namelen;
782 int hdrincl;
783 u16 proto;
784 int err;
786 /* Rough check on arithmetic overflow,
787 better check is made in ip6_append_data().
789 if (len > INT_MAX)
790 return -EMSGSIZE;
792 /* Mirror BSD error message compatibility */
793 if (msg->msg_flags & MSG_OOB)
794 return -EOPNOTSUPP;
796 /* hdrincl should be READ_ONCE(inet->hdrincl)
797 * but READ_ONCE() doesn't work with bit fields.
798 * Doing this indirectly yields the same result.
800 hdrincl = inet->hdrincl;
801 hdrincl = READ_ONCE(hdrincl);
804 * Get and verify the address.
806 memset(&fl6, 0, sizeof(fl6));
808 fl6.flowi6_mark = sk->sk_mark;
809 fl6.flowi6_uid = sk->sk_uid;
811 ipcm6_init(&ipc6);
812 ipc6.sockc.tsflags = sk->sk_tsflags;
813 ipc6.sockc.mark = sk->sk_mark;
815 if (sin6) {
816 if (addr_len < SIN6_LEN_RFC2133)
817 return -EINVAL;
819 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
820 return -EAFNOSUPPORT;
822 /* port is the proto value [0..255] carried in nexthdr */
823 proto = ntohs(sin6->sin6_port);
825 if (!proto)
826 proto = inet->inet_num;
827 else if (proto != inet->inet_num)
828 return -EINVAL;
830 if (proto > 255)
831 return -EINVAL;
833 daddr = &sin6->sin6_addr;
834 if (np->sndflow) {
835 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
836 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
837 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
838 if (IS_ERR(flowlabel))
839 return -EINVAL;
844 * Otherwise it will be difficult to maintain
845 * sk->sk_dst_cache.
847 if (sk->sk_state == TCP_ESTABLISHED &&
848 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
849 daddr = &sk->sk_v6_daddr;
851 if (addr_len >= sizeof(struct sockaddr_in6) &&
852 sin6->sin6_scope_id &&
853 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
854 fl6.flowi6_oif = sin6->sin6_scope_id;
855 } else {
856 if (sk->sk_state != TCP_ESTABLISHED)
857 return -EDESTADDRREQ;
859 proto = inet->inet_num;
860 daddr = &sk->sk_v6_daddr;
861 fl6.flowlabel = np->flow_label;
864 if (fl6.flowi6_oif == 0)
865 fl6.flowi6_oif = sk->sk_bound_dev_if;
867 if (msg->msg_controllen) {
868 opt = &opt_space;
869 memset(opt, 0, sizeof(struct ipv6_txoptions));
870 opt->tot_len = sizeof(struct ipv6_txoptions);
871 ipc6.opt = opt;
873 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
874 if (err < 0) {
875 fl6_sock_release(flowlabel);
876 return err;
878 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
879 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
880 if (IS_ERR(flowlabel))
881 return -EINVAL;
883 if (!(opt->opt_nflen|opt->opt_flen))
884 opt = NULL;
886 if (!opt) {
887 opt = txopt_get(np);
888 opt_to_free = opt;
890 if (flowlabel)
891 opt = fl6_merge_options(&opt_space, flowlabel, opt);
892 opt = ipv6_fixup_options(&opt_space, opt);
894 fl6.flowi6_proto = proto;
895 fl6.flowi6_mark = ipc6.sockc.mark;
897 if (!hdrincl) {
898 rfv.msg = msg;
899 rfv.hlen = 0;
900 err = rawv6_probe_proto_opt(&rfv, &fl6);
901 if (err)
902 goto out;
905 if (!ipv6_addr_any(daddr))
906 fl6.daddr = *daddr;
907 else
908 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
909 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
910 fl6.saddr = np->saddr;
912 final_p = fl6_update_dst(&fl6, opt, &final);
914 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
915 fl6.flowi6_oif = np->mcast_oif;
916 else if (!fl6.flowi6_oif)
917 fl6.flowi6_oif = np->ucast_oif;
918 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
920 if (hdrincl)
921 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH;
923 if (ipc6.tclass < 0)
924 ipc6.tclass = np->tclass;
926 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
928 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
929 if (IS_ERR(dst)) {
930 err = PTR_ERR(dst);
931 goto out;
933 if (ipc6.hlimit < 0)
934 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
936 if (ipc6.dontfrag < 0)
937 ipc6.dontfrag = np->dontfrag;
939 if (msg->msg_flags&MSG_CONFIRM)
940 goto do_confirm;
942 back_from_confirm:
943 if (hdrincl)
944 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst,
945 msg->msg_flags, &ipc6.sockc);
946 else {
947 ipc6.opt = opt;
948 lock_sock(sk);
949 err = ip6_append_data(sk, raw6_getfrag, &rfv,
950 len, 0, &ipc6, &fl6, (struct rt6_info *)dst,
951 msg->msg_flags);
953 if (err)
954 ip6_flush_pending_frames(sk);
955 else if (!(msg->msg_flags & MSG_MORE))
956 err = rawv6_push_pending_frames(sk, &fl6, rp);
957 release_sock(sk);
959 done:
960 dst_release(dst);
961 out:
962 fl6_sock_release(flowlabel);
963 txopt_put(opt_to_free);
964 return err < 0 ? err : len;
965 do_confirm:
966 if (msg->msg_flags & MSG_PROBE)
967 dst_confirm_neigh(dst, &fl6.daddr);
968 if (!(msg->msg_flags & MSG_PROBE) || len)
969 goto back_from_confirm;
970 err = 0;
971 goto done;
974 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
975 char __user *optval, int optlen)
977 switch (optname) {
978 case ICMPV6_FILTER:
979 if (optlen > sizeof(struct icmp6_filter))
980 optlen = sizeof(struct icmp6_filter);
981 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
982 return -EFAULT;
983 return 0;
984 default:
985 return -ENOPROTOOPT;
988 return 0;
991 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
992 char __user *optval, int __user *optlen)
994 int len;
996 switch (optname) {
997 case ICMPV6_FILTER:
998 if (get_user(len, optlen))
999 return -EFAULT;
1000 if (len < 0)
1001 return -EINVAL;
1002 if (len > sizeof(struct icmp6_filter))
1003 len = sizeof(struct icmp6_filter);
1004 if (put_user(len, optlen))
1005 return -EFAULT;
1006 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
1007 return -EFAULT;
1008 return 0;
1009 default:
1010 return -ENOPROTOOPT;
1013 return 0;
1017 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
1018 char __user *optval, unsigned int optlen)
1020 struct raw6_sock *rp = raw6_sk(sk);
1021 int val;
1023 if (get_user(val, (int __user *)optval))
1024 return -EFAULT;
1026 switch (optname) {
1027 case IPV6_HDRINCL:
1028 if (sk->sk_type != SOCK_RAW)
1029 return -EINVAL;
1030 inet_sk(sk)->hdrincl = !!val;
1031 return 0;
1032 case IPV6_CHECKSUM:
1033 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
1034 level == IPPROTO_IPV6) {
1036 * RFC3542 tells that IPV6_CHECKSUM socket
1037 * option in the IPPROTO_IPV6 level is not
1038 * allowed on ICMPv6 sockets.
1039 * If you want to set it, use IPPROTO_RAW
1040 * level IPV6_CHECKSUM socket option
1041 * (Linux extension).
1043 return -EINVAL;
1046 /* You may get strange result with a positive odd offset;
1047 RFC2292bis agrees with me. */
1048 if (val > 0 && (val&1))
1049 return -EINVAL;
1050 if (val < 0) {
1051 rp->checksum = 0;
1052 } else {
1053 rp->checksum = 1;
1054 rp->offset = val;
1057 return 0;
1059 default:
1060 return -ENOPROTOOPT;
1064 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1065 char __user *optval, unsigned int optlen)
1067 switch (level) {
1068 case SOL_RAW:
1069 break;
1071 case SOL_ICMPV6:
1072 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1073 return -EOPNOTSUPP;
1074 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1075 case SOL_IPV6:
1076 if (optname == IPV6_CHECKSUM ||
1077 optname == IPV6_HDRINCL)
1078 break;
1079 /* fall through */
1080 default:
1081 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1084 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1087 #ifdef CONFIG_COMPAT
1088 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1089 char __user *optval, unsigned int optlen)
1091 switch (level) {
1092 case SOL_RAW:
1093 break;
1094 case SOL_ICMPV6:
1095 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1096 return -EOPNOTSUPP;
1097 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1098 case SOL_IPV6:
1099 if (optname == IPV6_CHECKSUM ||
1100 optname == IPV6_HDRINCL)
1101 break;
1102 /* fall through */
1103 default:
1104 return compat_ipv6_setsockopt(sk, level, optname,
1105 optval, optlen);
1107 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1109 #endif
1111 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1112 char __user *optval, int __user *optlen)
1114 struct raw6_sock *rp = raw6_sk(sk);
1115 int val, len;
1117 if (get_user(len, optlen))
1118 return -EFAULT;
1120 switch (optname) {
1121 case IPV6_HDRINCL:
1122 val = inet_sk(sk)->hdrincl;
1123 break;
1124 case IPV6_CHECKSUM:
1126 * We allow getsockopt() for IPPROTO_IPV6-level
1127 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1128 * since RFC3542 is silent about it.
1130 if (rp->checksum == 0)
1131 val = -1;
1132 else
1133 val = rp->offset;
1134 break;
1136 default:
1137 return -ENOPROTOOPT;
1140 len = min_t(unsigned int, sizeof(int), len);
1142 if (put_user(len, optlen))
1143 return -EFAULT;
1144 if (copy_to_user(optval, &val, len))
1145 return -EFAULT;
1146 return 0;
1149 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1150 char __user *optval, int __user *optlen)
1152 switch (level) {
1153 case SOL_RAW:
1154 break;
1156 case SOL_ICMPV6:
1157 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1158 return -EOPNOTSUPP;
1159 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1160 case SOL_IPV6:
1161 if (optname == IPV6_CHECKSUM ||
1162 optname == IPV6_HDRINCL)
1163 break;
1164 /* fall through */
1165 default:
1166 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1169 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1172 #ifdef CONFIG_COMPAT
1173 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1174 char __user *optval, int __user *optlen)
1176 switch (level) {
1177 case SOL_RAW:
1178 break;
1179 case SOL_ICMPV6:
1180 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1181 return -EOPNOTSUPP;
1182 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1183 case SOL_IPV6:
1184 if (optname == IPV6_CHECKSUM ||
1185 optname == IPV6_HDRINCL)
1186 break;
1187 /* fall through */
1188 default:
1189 return compat_ipv6_getsockopt(sk, level, optname,
1190 optval, optlen);
1192 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1194 #endif
1196 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1198 switch (cmd) {
1199 case SIOCOUTQ: {
1200 int amount = sk_wmem_alloc_get(sk);
1202 return put_user(amount, (int __user *)arg);
1204 case SIOCINQ: {
1205 struct sk_buff *skb;
1206 int amount = 0;
1208 spin_lock_bh(&sk->sk_receive_queue.lock);
1209 skb = skb_peek(&sk->sk_receive_queue);
1210 if (skb)
1211 amount = skb->len;
1212 spin_unlock_bh(&sk->sk_receive_queue.lock);
1213 return put_user(amount, (int __user *)arg);
1216 default:
1217 #ifdef CONFIG_IPV6_MROUTE
1218 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1219 #else
1220 return -ENOIOCTLCMD;
1221 #endif
1225 #ifdef CONFIG_COMPAT
1226 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1228 switch (cmd) {
1229 case SIOCOUTQ:
1230 case SIOCINQ:
1231 return -ENOIOCTLCMD;
1232 default:
1233 #ifdef CONFIG_IPV6_MROUTE
1234 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1235 #else
1236 return -ENOIOCTLCMD;
1237 #endif
1240 #endif
1242 static void rawv6_close(struct sock *sk, long timeout)
1244 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1245 ip6_ra_control(sk, -1);
1246 ip6mr_sk_done(sk);
1247 sk_common_release(sk);
1250 static void raw6_destroy(struct sock *sk)
1252 lock_sock(sk);
1253 ip6_flush_pending_frames(sk);
1254 release_sock(sk);
1256 inet6_destroy_sock(sk);
1259 static int rawv6_init_sk(struct sock *sk)
1261 struct raw6_sock *rp = raw6_sk(sk);
1263 switch (inet_sk(sk)->inet_num) {
1264 case IPPROTO_ICMPV6:
1265 rp->checksum = 1;
1266 rp->offset = 2;
1267 break;
1268 case IPPROTO_MH:
1269 rp->checksum = 1;
1270 rp->offset = 4;
1271 break;
1272 default:
1273 break;
1275 return 0;
1278 struct proto rawv6_prot = {
1279 .name = "RAWv6",
1280 .owner = THIS_MODULE,
1281 .close = rawv6_close,
1282 .destroy = raw6_destroy,
1283 .connect = ip6_datagram_connect_v6_only,
1284 .disconnect = __udp_disconnect,
1285 .ioctl = rawv6_ioctl,
1286 .init = rawv6_init_sk,
1287 .setsockopt = rawv6_setsockopt,
1288 .getsockopt = rawv6_getsockopt,
1289 .sendmsg = rawv6_sendmsg,
1290 .recvmsg = rawv6_recvmsg,
1291 .bind = rawv6_bind,
1292 .backlog_rcv = rawv6_rcv_skb,
1293 .hash = raw_hash_sk,
1294 .unhash = raw_unhash_sk,
1295 .obj_size = sizeof(struct raw6_sock),
1296 .useroffset = offsetof(struct raw6_sock, filter),
1297 .usersize = sizeof_field(struct raw6_sock, filter),
1298 .h.raw_hash = &raw_v6_hashinfo,
1299 #ifdef CONFIG_COMPAT
1300 .compat_setsockopt = compat_rawv6_setsockopt,
1301 .compat_getsockopt = compat_rawv6_getsockopt,
1302 .compat_ioctl = compat_rawv6_ioctl,
1303 #endif
1304 .diag_destroy = raw_abort,
1307 #ifdef CONFIG_PROC_FS
1308 static int raw6_seq_show(struct seq_file *seq, void *v)
1310 if (v == SEQ_START_TOKEN) {
1311 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1312 } else {
1313 struct sock *sp = v;
1314 __u16 srcp = inet_sk(sp)->inet_num;
1315 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1316 raw_seq_private(seq)->bucket);
1318 return 0;
1321 static const struct seq_operations raw6_seq_ops = {
1322 .start = raw_seq_start,
1323 .next = raw_seq_next,
1324 .stop = raw_seq_stop,
1325 .show = raw6_seq_show,
1328 static int __net_init raw6_init_net(struct net *net)
1330 if (!proc_create_net_data("raw6", 0444, net->proc_net, &raw6_seq_ops,
1331 sizeof(struct raw_iter_state), &raw_v6_hashinfo))
1332 return -ENOMEM;
1334 return 0;
1337 static void __net_exit raw6_exit_net(struct net *net)
1339 remove_proc_entry("raw6", net->proc_net);
1342 static struct pernet_operations raw6_net_ops = {
1343 .init = raw6_init_net,
1344 .exit = raw6_exit_net,
1347 int __init raw6_proc_init(void)
1349 return register_pernet_subsys(&raw6_net_ops);
1352 void raw6_proc_exit(void)
1354 unregister_pernet_subsys(&raw6_net_ops);
1356 #endif /* CONFIG_PROC_FS */
1358 /* Same as inet6_dgram_ops, sans udp_poll. */
1359 const struct proto_ops inet6_sockraw_ops = {
1360 .family = PF_INET6,
1361 .owner = THIS_MODULE,
1362 .release = inet6_release,
1363 .bind = inet6_bind,
1364 .connect = inet_dgram_connect, /* ok */
1365 .socketpair = sock_no_socketpair, /* a do nothing */
1366 .accept = sock_no_accept, /* a do nothing */
1367 .getname = inet6_getname,
1368 .poll = datagram_poll, /* ok */
1369 .ioctl = inet6_ioctl, /* must change */
1370 .gettstamp = sock_gettstamp,
1371 .listen = sock_no_listen, /* ok */
1372 .shutdown = inet_shutdown, /* ok */
1373 .setsockopt = sock_common_setsockopt, /* ok */
1374 .getsockopt = sock_common_getsockopt, /* ok */
1375 .sendmsg = inet_sendmsg, /* ok */
1376 .recvmsg = sock_common_recvmsg, /* ok */
1377 .mmap = sock_no_mmap,
1378 .sendpage = sock_no_sendpage,
1379 #ifdef CONFIG_COMPAT
1380 .compat_setsockopt = compat_sock_common_setsockopt,
1381 .compat_getsockopt = compat_sock_common_getsockopt,
1382 #endif
1385 static struct inet_protosw rawv6_protosw = {
1386 .type = SOCK_RAW,
1387 .protocol = IPPROTO_IP, /* wild card */
1388 .prot = &rawv6_prot,
1389 .ops = &inet6_sockraw_ops,
1390 .flags = INET_PROTOSW_REUSE,
1393 int __init rawv6_init(void)
1395 return inet6_register_protosw(&rawv6_protosw);
1398 void rawv6_exit(void)
1400 inet6_unregister_protosw(&rawv6_protosw);