Linux 5.1.15
[linux/fpc-iii.git] / net / ipv6 / raw.c
blob5f8fe98b435b24137d643b4ff1f3a889c3990e22
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 <linux/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 IS_ENABLED(CONFIG_IPV6_MIP6)
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>
64 #include <linux/export.h>
66 #define ICMPV6_HDRLEN 4 /* ICMPv6 header, RFC 4443 Section 2.1 */
68 struct raw_hashinfo raw_v6_hashinfo = {
69 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
71 EXPORT_SYMBOL_GPL(raw_v6_hashinfo);
73 struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
74 unsigned short num, const struct in6_addr *loc_addr,
75 const struct in6_addr *rmt_addr, int dif, int sdif)
77 bool is_multicast = ipv6_addr_is_multicast(loc_addr);
79 sk_for_each_from(sk)
80 if (inet_sk(sk)->inet_num == num) {
82 if (!net_eq(sock_net(sk), net))
83 continue;
85 if (!ipv6_addr_any(&sk->sk_v6_daddr) &&
86 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr))
87 continue;
89 if (!raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if,
90 dif, sdif))
91 continue;
93 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
94 if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))
95 goto found;
96 if (is_multicast &&
97 inet6_mc_check(sk, loc_addr, rmt_addr))
98 goto found;
99 continue;
101 goto found;
103 sk = NULL;
104 found:
105 return sk;
107 EXPORT_SYMBOL_GPL(__raw_v6_lookup);
110 * 0 - deliver
111 * 1 - block
113 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
115 struct icmp6hdr _hdr;
116 const struct icmp6hdr *hdr;
118 /* We require only the four bytes of the ICMPv6 header, not any
119 * additional bytes of message body in "struct icmp6hdr".
121 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
122 ICMPV6_HDRLEN, &_hdr);
123 if (hdr) {
124 const __u32 *data = &raw6_sk(sk)->filter.data[0];
125 unsigned int type = hdr->icmp6_type;
127 return (data[type >> 5] & (1U << (type & 31))) != 0;
129 return 1;
132 #if IS_ENABLED(CONFIG_IPV6_MIP6)
133 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
135 static mh_filter_t __rcu *mh_filter __read_mostly;
137 int rawv6_mh_filter_register(mh_filter_t filter)
139 rcu_assign_pointer(mh_filter, filter);
140 return 0;
142 EXPORT_SYMBOL(rawv6_mh_filter_register);
144 int rawv6_mh_filter_unregister(mh_filter_t filter)
146 RCU_INIT_POINTER(mh_filter, NULL);
147 synchronize_rcu();
148 return 0;
150 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
152 #endif
155 * demultiplex raw sockets.
156 * (should consider queueing the skb in the sock receive_queue
157 * without calling rawv6.c)
159 * Caller owns SKB so we must make clones.
161 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
163 const struct in6_addr *saddr;
164 const struct in6_addr *daddr;
165 struct sock *sk;
166 bool delivered = false;
167 __u8 hash;
168 struct net *net;
170 saddr = &ipv6_hdr(skb)->saddr;
171 daddr = saddr + 1;
173 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
175 read_lock(&raw_v6_hashinfo.lock);
176 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
178 if (!sk)
179 goto out;
181 net = dev_net(skb->dev);
182 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr,
183 inet6_iif(skb), inet6_sdif(skb));
185 while (sk) {
186 int filtered;
188 delivered = true;
189 switch (nexthdr) {
190 case IPPROTO_ICMPV6:
191 filtered = icmpv6_filter(sk, skb);
192 break;
194 #if IS_ENABLED(CONFIG_IPV6_MIP6)
195 case IPPROTO_MH:
197 /* XXX: To validate MH only once for each packet,
198 * this is placed here. It should be after checking
199 * xfrm policy, however it doesn't. The checking xfrm
200 * policy is placed in rawv6_rcv() because it is
201 * required for each socket.
203 mh_filter_t *filter;
205 filter = rcu_dereference(mh_filter);
206 filtered = filter ? (*filter)(sk, skb) : 0;
207 break;
209 #endif
210 default:
211 filtered = 0;
212 break;
215 if (filtered < 0)
216 break;
217 if (filtered == 0) {
218 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
220 /* Not releasing hash table! */
221 if (clone) {
222 nf_reset(clone);
223 rawv6_rcv(sk, clone);
226 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
227 inet6_iif(skb), inet6_sdif(skb));
229 out:
230 read_unlock(&raw_v6_hashinfo.lock);
231 return delivered;
234 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
236 struct sock *raw_sk;
238 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]);
239 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
240 raw_sk = NULL;
242 return raw_sk != NULL;
245 /* This cleans up af_inet6 a bit. -DaveM */
246 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
248 struct inet_sock *inet = inet_sk(sk);
249 struct ipv6_pinfo *np = inet6_sk(sk);
250 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
251 __be32 v4addr = 0;
252 int addr_type;
253 int err;
255 if (addr_len < SIN6_LEN_RFC2133)
256 return -EINVAL;
258 if (addr->sin6_family != AF_INET6)
259 return -EINVAL;
261 addr_type = ipv6_addr_type(&addr->sin6_addr);
263 /* Raw sockets are IPv6 only */
264 if (addr_type == IPV6_ADDR_MAPPED)
265 return -EADDRNOTAVAIL;
267 lock_sock(sk);
269 err = -EINVAL;
270 if (sk->sk_state != TCP_CLOSE)
271 goto out;
273 rcu_read_lock();
274 /* Check if the address belongs to the host. */
275 if (addr_type != IPV6_ADDR_ANY) {
276 struct net_device *dev = NULL;
278 if (__ipv6_addr_needs_scope_id(addr_type)) {
279 if (addr_len >= sizeof(struct sockaddr_in6) &&
280 addr->sin6_scope_id) {
281 /* Override any existing binding, if another
282 * one is supplied by user.
284 sk->sk_bound_dev_if = addr->sin6_scope_id;
287 /* Binding to link-local address requires an interface */
288 if (!sk->sk_bound_dev_if)
289 goto out_unlock;
292 if (sk->sk_bound_dev_if) {
293 err = -ENODEV;
294 dev = dev_get_by_index_rcu(sock_net(sk),
295 sk->sk_bound_dev_if);
296 if (!dev)
297 goto out_unlock;
300 /* ipv4 addr of the socket is invalid. Only the
301 * unspecified and mapped address have a v4 equivalent.
303 v4addr = LOOPBACK4_IPV6;
304 if (!(addr_type & IPV6_ADDR_MULTICAST) &&
305 !sock_net(sk)->ipv6.sysctl.ip_nonlocal_bind) {
306 err = -EADDRNOTAVAIL;
307 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
308 dev, 0)) {
309 goto out_unlock;
314 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
315 sk->sk_v6_rcv_saddr = addr->sin6_addr;
316 if (!(addr_type & IPV6_ADDR_MULTICAST))
317 np->saddr = addr->sin6_addr;
318 err = 0;
319 out_unlock:
320 rcu_read_unlock();
321 out:
322 release_sock(sk);
323 return err;
326 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
327 struct inet6_skb_parm *opt,
328 u8 type, u8 code, int offset, __be32 info)
330 struct inet_sock *inet = inet_sk(sk);
331 struct ipv6_pinfo *np = inet6_sk(sk);
332 int err;
333 int harderr;
335 /* Report error on raw socket, if:
336 1. User requested recverr.
337 2. Socket is connected (otherwise the error indication
338 is useless without recverr and error is hard.
340 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
341 return;
343 harderr = icmpv6_err_convert(type, code, &err);
344 if (type == ICMPV6_PKT_TOOBIG) {
345 ip6_sk_update_pmtu(skb, sk, info);
346 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
348 if (type == NDISC_REDIRECT) {
349 ip6_sk_redirect(skb, sk);
350 return;
352 if (np->recverr) {
353 u8 *payload = skb->data;
354 if (!inet->hdrincl)
355 payload += offset;
356 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
359 if (np->recverr || harderr) {
360 sk->sk_err = err;
361 sk->sk_error_report(sk);
365 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
366 u8 type, u8 code, int inner_offset, __be32 info)
368 struct sock *sk;
369 int hash;
370 const struct in6_addr *saddr, *daddr;
371 struct net *net;
373 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
375 read_lock(&raw_v6_hashinfo.lock);
376 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
377 if (sk) {
378 /* Note: ipv6_hdr(skb) != skb->data */
379 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
380 saddr = &ip6h->saddr;
381 daddr = &ip6h->daddr;
382 net = dev_net(skb->dev);
384 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
385 inet6_iif(skb), inet6_iif(skb)))) {
386 rawv6_err(sk, skb, NULL, type, code,
387 inner_offset, info);
388 sk = sk_next(sk);
391 read_unlock(&raw_v6_hashinfo.lock);
394 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
396 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
397 skb_checksum_complete(skb)) {
398 atomic_inc(&sk->sk_drops);
399 kfree_skb(skb);
400 return NET_RX_DROP;
403 /* Charge it to the socket. */
404 skb_dst_drop(skb);
405 if (sock_queue_rcv_skb(sk, skb) < 0) {
406 kfree_skb(skb);
407 return NET_RX_DROP;
410 return 0;
414 * This is next to useless...
415 * if we demultiplex in network layer we don't need the extra call
416 * just to queue the skb...
417 * maybe we could have the network decide upon a hint if it
418 * should call raw_rcv for demultiplexing
420 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
422 struct inet_sock *inet = inet_sk(sk);
423 struct raw6_sock *rp = raw6_sk(sk);
425 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
426 atomic_inc(&sk->sk_drops);
427 kfree_skb(skb);
428 return NET_RX_DROP;
431 if (!rp->checksum)
432 skb->ip_summed = CHECKSUM_UNNECESSARY;
434 if (skb->ip_summed == CHECKSUM_COMPLETE) {
435 skb_postpull_rcsum(skb, skb_network_header(skb),
436 skb_network_header_len(skb));
437 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
438 &ipv6_hdr(skb)->daddr,
439 skb->len, inet->inet_num, skb->csum))
440 skb->ip_summed = CHECKSUM_UNNECESSARY;
442 if (!skb_csum_unnecessary(skb))
443 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
444 &ipv6_hdr(skb)->daddr,
445 skb->len,
446 inet->inet_num, 0));
448 if (inet->hdrincl) {
449 if (skb_checksum_complete(skb)) {
450 atomic_inc(&sk->sk_drops);
451 kfree_skb(skb);
452 return NET_RX_DROP;
456 rawv6_rcv_skb(sk, skb);
457 return 0;
462 * This should be easy, if there is something there
463 * we return it, otherwise we block.
466 static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
467 int noblock, int flags, int *addr_len)
469 struct ipv6_pinfo *np = inet6_sk(sk);
470 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
471 struct sk_buff *skb;
472 size_t copied;
473 int err;
475 if (flags & MSG_OOB)
476 return -EOPNOTSUPP;
478 if (flags & MSG_ERRQUEUE)
479 return ipv6_recv_error(sk, msg, len, addr_len);
481 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
482 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
484 skb = skb_recv_datagram(sk, flags, noblock, &err);
485 if (!skb)
486 goto out;
488 copied = skb->len;
489 if (copied > len) {
490 copied = len;
491 msg->msg_flags |= MSG_TRUNC;
494 if (skb_csum_unnecessary(skb)) {
495 err = skb_copy_datagram_msg(skb, 0, msg, copied);
496 } else if (msg->msg_flags&MSG_TRUNC) {
497 if (__skb_checksum_complete(skb))
498 goto csum_copy_err;
499 err = skb_copy_datagram_msg(skb, 0, msg, copied);
500 } else {
501 err = skb_copy_and_csum_datagram_msg(skb, 0, msg);
502 if (err == -EINVAL)
503 goto csum_copy_err;
505 if (err)
506 goto out_free;
508 /* Copy the address. */
509 if (sin6) {
510 sin6->sin6_family = AF_INET6;
511 sin6->sin6_port = 0;
512 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
513 sin6->sin6_flowinfo = 0;
514 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
515 inet6_iif(skb));
516 *addr_len = sizeof(*sin6);
519 sock_recv_ts_and_drops(msg, sk, skb);
521 if (np->rxopt.all)
522 ip6_datagram_recv_ctl(sk, msg, skb);
524 err = copied;
525 if (flags & MSG_TRUNC)
526 err = skb->len;
528 out_free:
529 skb_free_datagram(sk, skb);
530 out:
531 return err;
533 csum_copy_err:
534 skb_kill_datagram(sk, skb, flags);
536 /* Error for blocking case is chosen to masquerade
537 as some normal condition.
539 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
540 goto out;
543 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
544 struct raw6_sock *rp)
546 struct sk_buff *skb;
547 int err = 0;
548 int offset;
549 int len;
550 int total_len;
551 __wsum tmp_csum;
552 __sum16 csum;
554 if (!rp->checksum)
555 goto send;
557 skb = skb_peek(&sk->sk_write_queue);
558 if (!skb)
559 goto out;
561 offset = rp->offset;
562 total_len = inet_sk(sk)->cork.base.length;
563 if (offset >= total_len - 1) {
564 err = -EINVAL;
565 ip6_flush_pending_frames(sk);
566 goto out;
569 /* should be check HW csum miyazawa */
570 if (skb_queue_len(&sk->sk_write_queue) == 1) {
572 * Only one fragment on the socket.
574 tmp_csum = skb->csum;
575 } else {
576 struct sk_buff *csum_skb = NULL;
577 tmp_csum = 0;
579 skb_queue_walk(&sk->sk_write_queue, skb) {
580 tmp_csum = csum_add(tmp_csum, skb->csum);
582 if (csum_skb)
583 continue;
585 len = skb->len - skb_transport_offset(skb);
586 if (offset >= len) {
587 offset -= len;
588 continue;
591 csum_skb = skb;
594 skb = csum_skb;
597 offset += skb_transport_offset(skb);
598 err = skb_copy_bits(skb, offset, &csum, 2);
599 if (err < 0) {
600 ip6_flush_pending_frames(sk);
601 goto out;
604 /* in case cksum was not initialized */
605 if (unlikely(csum))
606 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
608 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
609 total_len, fl6->flowi6_proto, tmp_csum);
611 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
612 csum = CSUM_MANGLED_0;
614 BUG_ON(skb_store_bits(skb, offset, &csum, 2));
616 send:
617 err = ip6_push_pending_frames(sk);
618 out:
619 return err;
622 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length,
623 struct flowi6 *fl6, struct dst_entry **dstp,
624 unsigned int flags, const struct sockcm_cookie *sockc)
626 struct ipv6_pinfo *np = inet6_sk(sk);
627 struct net *net = sock_net(sk);
628 struct ipv6hdr *iph;
629 struct sk_buff *skb;
630 int err;
631 struct rt6_info *rt = (struct rt6_info *)*dstp;
632 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
633 int tlen = rt->dst.dev->needed_tailroom;
635 if (length > rt->dst.dev->mtu) {
636 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
637 return -EMSGSIZE;
639 if (length < sizeof(struct ipv6hdr))
640 return -EINVAL;
641 if (flags&MSG_PROBE)
642 goto out;
644 skb = sock_alloc_send_skb(sk,
645 length + hlen + tlen + 15,
646 flags & MSG_DONTWAIT, &err);
647 if (!skb)
648 goto error;
649 skb_reserve(skb, hlen);
651 skb->protocol = htons(ETH_P_IPV6);
652 skb->priority = sk->sk_priority;
653 skb->mark = sk->sk_mark;
654 skb->tstamp = sockc->transmit_time;
656 skb_put(skb, length);
657 skb_reset_network_header(skb);
658 iph = ipv6_hdr(skb);
660 skb->ip_summed = CHECKSUM_NONE;
662 skb_setup_tx_timestamp(skb, sockc->tsflags);
664 if (flags & MSG_CONFIRM)
665 skb_set_dst_pending_confirm(skb, 1);
667 skb->transport_header = skb->network_header;
668 err = memcpy_from_msg(iph, msg, length);
669 if (err) {
670 err = -EFAULT;
671 kfree_skb(skb);
672 goto error;
675 skb_dst_set(skb, &rt->dst);
676 *dstp = NULL;
678 /* if egress device is enslaved to an L3 master device pass the
679 * skb to its handler for processing
681 skb = l3mdev_ip6_out(sk, skb);
682 if (unlikely(!skb))
683 return 0;
685 /* Acquire rcu_read_lock() in case we need to use rt->rt6i_idev
686 * in the error path. Since skb has been freed, the dst could
687 * have been queued for deletion.
689 rcu_read_lock();
690 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
691 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
692 NULL, rt->dst.dev, dst_output);
693 if (err > 0)
694 err = net_xmit_errno(err);
695 if (err) {
696 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
697 rcu_read_unlock();
698 goto error_check;
700 rcu_read_unlock();
701 out:
702 return 0;
704 error:
705 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
706 error_check:
707 if (err == -ENOBUFS && !np->recverr)
708 err = 0;
709 return err;
712 struct raw6_frag_vec {
713 struct msghdr *msg;
714 int hlen;
715 char c[4];
718 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6)
720 int err = 0;
721 switch (fl6->flowi6_proto) {
722 case IPPROTO_ICMPV6:
723 rfv->hlen = 2;
724 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
725 if (!err) {
726 fl6->fl6_icmp_type = rfv->c[0];
727 fl6->fl6_icmp_code = rfv->c[1];
729 break;
730 case IPPROTO_MH:
731 rfv->hlen = 4;
732 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
733 if (!err)
734 fl6->fl6_mh_type = rfv->c[2];
736 return err;
739 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd,
740 struct sk_buff *skb)
742 struct raw6_frag_vec *rfv = from;
744 if (offset < rfv->hlen) {
745 int copy = min(rfv->hlen - offset, len);
747 if (skb->ip_summed == CHECKSUM_PARTIAL)
748 memcpy(to, rfv->c + offset, copy);
749 else
750 skb->csum = csum_block_add(
751 skb->csum,
752 csum_partial_copy_nocheck(rfv->c + offset,
753 to, copy, 0),
754 odd);
756 odd = 0;
757 offset += copy;
758 to += copy;
759 len -= copy;
761 if (!len)
762 return 0;
765 offset -= rfv->hlen;
767 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
770 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
772 struct ipv6_txoptions *opt_to_free = NULL;
773 struct ipv6_txoptions opt_space;
774 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
775 struct in6_addr *daddr, *final_p, final;
776 struct inet_sock *inet = inet_sk(sk);
777 struct ipv6_pinfo *np = inet6_sk(sk);
778 struct raw6_sock *rp = raw6_sk(sk);
779 struct ipv6_txoptions *opt = NULL;
780 struct ip6_flowlabel *flowlabel = NULL;
781 struct dst_entry *dst = NULL;
782 struct raw6_frag_vec rfv;
783 struct flowi6 fl6;
784 struct ipcm6_cookie ipc6;
785 int addr_len = msg->msg_namelen;
786 int hdrincl;
787 u16 proto;
788 int err;
790 /* Rough check on arithmetic overflow,
791 better check is made in ip6_append_data().
793 if (len > INT_MAX)
794 return -EMSGSIZE;
796 /* Mirror BSD error message compatibility */
797 if (msg->msg_flags & MSG_OOB)
798 return -EOPNOTSUPP;
800 /* hdrincl should be READ_ONCE(inet->hdrincl)
801 * but READ_ONCE() doesn't work with bit fields.
802 * Doing this indirectly yields the same result.
804 hdrincl = inet->hdrincl;
805 hdrincl = READ_ONCE(hdrincl);
808 * Get and verify the address.
810 memset(&fl6, 0, sizeof(fl6));
812 fl6.flowi6_mark = sk->sk_mark;
813 fl6.flowi6_uid = sk->sk_uid;
815 ipcm6_init(&ipc6);
816 ipc6.sockc.tsflags = sk->sk_tsflags;
818 if (sin6) {
819 if (addr_len < SIN6_LEN_RFC2133)
820 return -EINVAL;
822 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
823 return -EAFNOSUPPORT;
825 /* port is the proto value [0..255] carried in nexthdr */
826 proto = ntohs(sin6->sin6_port);
828 if (!proto)
829 proto = inet->inet_num;
830 else if (proto != inet->inet_num)
831 return -EINVAL;
833 if (proto > 255)
834 return -EINVAL;
836 daddr = &sin6->sin6_addr;
837 if (np->sndflow) {
838 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
839 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
840 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
841 if (!flowlabel)
842 return -EINVAL;
847 * Otherwise it will be difficult to maintain
848 * sk->sk_dst_cache.
850 if (sk->sk_state == TCP_ESTABLISHED &&
851 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
852 daddr = &sk->sk_v6_daddr;
854 if (addr_len >= sizeof(struct sockaddr_in6) &&
855 sin6->sin6_scope_id &&
856 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
857 fl6.flowi6_oif = sin6->sin6_scope_id;
858 } else {
859 if (sk->sk_state != TCP_ESTABLISHED)
860 return -EDESTADDRREQ;
862 proto = inet->inet_num;
863 daddr = &sk->sk_v6_daddr;
864 fl6.flowlabel = np->flow_label;
867 if (fl6.flowi6_oif == 0)
868 fl6.flowi6_oif = sk->sk_bound_dev_if;
870 if (msg->msg_controllen) {
871 opt = &opt_space;
872 memset(opt, 0, sizeof(struct ipv6_txoptions));
873 opt->tot_len = sizeof(struct ipv6_txoptions);
874 ipc6.opt = opt;
876 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
877 if (err < 0) {
878 fl6_sock_release(flowlabel);
879 return err;
881 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
882 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
883 if (!flowlabel)
884 return -EINVAL;
886 if (!(opt->opt_nflen|opt->opt_flen))
887 opt = NULL;
889 if (!opt) {
890 opt = txopt_get(np);
891 opt_to_free = opt;
893 if (flowlabel)
894 opt = fl6_merge_options(&opt_space, flowlabel, opt);
895 opt = ipv6_fixup_options(&opt_space, opt);
897 fl6.flowi6_proto = proto;
899 if (!hdrincl) {
900 rfv.msg = msg;
901 rfv.hlen = 0;
902 err = rawv6_probe_proto_opt(&rfv, &fl6);
903 if (err)
904 goto out;
907 if (!ipv6_addr_any(daddr))
908 fl6.daddr = *daddr;
909 else
910 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
911 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
912 fl6.saddr = np->saddr;
914 final_p = fl6_update_dst(&fl6, opt, &final);
916 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
917 fl6.flowi6_oif = np->mcast_oif;
918 else if (!fl6.flowi6_oif)
919 fl6.flowi6_oif = np->ucast_oif;
920 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
922 if (hdrincl)
923 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH;
925 if (ipc6.tclass < 0)
926 ipc6.tclass = np->tclass;
928 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
930 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
931 if (IS_ERR(dst)) {
932 err = PTR_ERR(dst);
933 goto out;
935 if (ipc6.hlimit < 0)
936 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
938 if (ipc6.dontfrag < 0)
939 ipc6.dontfrag = np->dontfrag;
941 if (msg->msg_flags&MSG_CONFIRM)
942 goto do_confirm;
944 back_from_confirm:
945 if (hdrincl)
946 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst,
947 msg->msg_flags, &ipc6.sockc);
948 else {
949 ipc6.opt = opt;
950 lock_sock(sk);
951 err = ip6_append_data(sk, raw6_getfrag, &rfv,
952 len, 0, &ipc6, &fl6, (struct rt6_info *)dst,
953 msg->msg_flags);
955 if (err)
956 ip6_flush_pending_frames(sk);
957 else if (!(msg->msg_flags & MSG_MORE))
958 err = rawv6_push_pending_frames(sk, &fl6, rp);
959 release_sock(sk);
961 done:
962 dst_release(dst);
963 out:
964 fl6_sock_release(flowlabel);
965 txopt_put(opt_to_free);
966 return err < 0 ? err : len;
967 do_confirm:
968 if (msg->msg_flags & MSG_PROBE)
969 dst_confirm_neigh(dst, &fl6.daddr);
970 if (!(msg->msg_flags & MSG_PROBE) || len)
971 goto back_from_confirm;
972 err = 0;
973 goto done;
976 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
977 char __user *optval, int optlen)
979 switch (optname) {
980 case ICMPV6_FILTER:
981 if (optlen > sizeof(struct icmp6_filter))
982 optlen = sizeof(struct icmp6_filter);
983 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
984 return -EFAULT;
985 return 0;
986 default:
987 return -ENOPROTOOPT;
990 return 0;
993 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
994 char __user *optval, int __user *optlen)
996 int len;
998 switch (optname) {
999 case ICMPV6_FILTER:
1000 if (get_user(len, optlen))
1001 return -EFAULT;
1002 if (len < 0)
1003 return -EINVAL;
1004 if (len > sizeof(struct icmp6_filter))
1005 len = sizeof(struct icmp6_filter);
1006 if (put_user(len, optlen))
1007 return -EFAULT;
1008 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
1009 return -EFAULT;
1010 return 0;
1011 default:
1012 return -ENOPROTOOPT;
1015 return 0;
1019 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
1020 char __user *optval, unsigned int optlen)
1022 struct raw6_sock *rp = raw6_sk(sk);
1023 int val;
1025 if (get_user(val, (int __user *)optval))
1026 return -EFAULT;
1028 switch (optname) {
1029 case IPV6_HDRINCL:
1030 if (sk->sk_type != SOCK_RAW)
1031 return -EINVAL;
1032 inet_sk(sk)->hdrincl = !!val;
1033 return 0;
1034 case IPV6_CHECKSUM:
1035 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
1036 level == IPPROTO_IPV6) {
1038 * RFC3542 tells that IPV6_CHECKSUM socket
1039 * option in the IPPROTO_IPV6 level is not
1040 * allowed on ICMPv6 sockets.
1041 * If you want to set it, use IPPROTO_RAW
1042 * level IPV6_CHECKSUM socket option
1043 * (Linux extension).
1045 return -EINVAL;
1048 /* You may get strange result with a positive odd offset;
1049 RFC2292bis agrees with me. */
1050 if (val > 0 && (val&1))
1051 return -EINVAL;
1052 if (val < 0) {
1053 rp->checksum = 0;
1054 } else {
1055 rp->checksum = 1;
1056 rp->offset = val;
1059 return 0;
1061 default:
1062 return -ENOPROTOOPT;
1066 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1067 char __user *optval, unsigned int optlen)
1069 switch (level) {
1070 case SOL_RAW:
1071 break;
1073 case SOL_ICMPV6:
1074 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1075 return -EOPNOTSUPP;
1076 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1077 case SOL_IPV6:
1078 if (optname == IPV6_CHECKSUM ||
1079 optname == IPV6_HDRINCL)
1080 break;
1081 /* fall through */
1082 default:
1083 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1086 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1089 #ifdef CONFIG_COMPAT
1090 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1091 char __user *optval, unsigned int optlen)
1093 switch (level) {
1094 case SOL_RAW:
1095 break;
1096 case SOL_ICMPV6:
1097 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1098 return -EOPNOTSUPP;
1099 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1100 case SOL_IPV6:
1101 if (optname == IPV6_CHECKSUM ||
1102 optname == IPV6_HDRINCL)
1103 break;
1104 /* fall through */
1105 default:
1106 return compat_ipv6_setsockopt(sk, level, optname,
1107 optval, optlen);
1109 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1111 #endif
1113 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1114 char __user *optval, int __user *optlen)
1116 struct raw6_sock *rp = raw6_sk(sk);
1117 int val, len;
1119 if (get_user(len, optlen))
1120 return -EFAULT;
1122 switch (optname) {
1123 case IPV6_HDRINCL:
1124 val = inet_sk(sk)->hdrincl;
1125 break;
1126 case IPV6_CHECKSUM:
1128 * We allow getsockopt() for IPPROTO_IPV6-level
1129 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1130 * since RFC3542 is silent about it.
1132 if (rp->checksum == 0)
1133 val = -1;
1134 else
1135 val = rp->offset;
1136 break;
1138 default:
1139 return -ENOPROTOOPT;
1142 len = min_t(unsigned int, sizeof(int), len);
1144 if (put_user(len, optlen))
1145 return -EFAULT;
1146 if (copy_to_user(optval, &val, len))
1147 return -EFAULT;
1148 return 0;
1151 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1152 char __user *optval, int __user *optlen)
1154 switch (level) {
1155 case SOL_RAW:
1156 break;
1158 case SOL_ICMPV6:
1159 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1160 return -EOPNOTSUPP;
1161 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1162 case SOL_IPV6:
1163 if (optname == IPV6_CHECKSUM ||
1164 optname == IPV6_HDRINCL)
1165 break;
1166 /* fall through */
1167 default:
1168 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1171 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1174 #ifdef CONFIG_COMPAT
1175 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1176 char __user *optval, int __user *optlen)
1178 switch (level) {
1179 case SOL_RAW:
1180 break;
1181 case SOL_ICMPV6:
1182 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1183 return -EOPNOTSUPP;
1184 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1185 case SOL_IPV6:
1186 if (optname == IPV6_CHECKSUM ||
1187 optname == IPV6_HDRINCL)
1188 break;
1189 /* fall through */
1190 default:
1191 return compat_ipv6_getsockopt(sk, level, optname,
1192 optval, optlen);
1194 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1196 #endif
1198 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1200 switch (cmd) {
1201 case SIOCOUTQ: {
1202 int amount = sk_wmem_alloc_get(sk);
1204 return put_user(amount, (int __user *)arg);
1206 case SIOCINQ: {
1207 struct sk_buff *skb;
1208 int amount = 0;
1210 spin_lock_bh(&sk->sk_receive_queue.lock);
1211 skb = skb_peek(&sk->sk_receive_queue);
1212 if (skb)
1213 amount = skb->len;
1214 spin_unlock_bh(&sk->sk_receive_queue.lock);
1215 return put_user(amount, (int __user *)arg);
1218 default:
1219 #ifdef CONFIG_IPV6_MROUTE
1220 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1221 #else
1222 return -ENOIOCTLCMD;
1223 #endif
1227 #ifdef CONFIG_COMPAT
1228 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1230 switch (cmd) {
1231 case SIOCOUTQ:
1232 case SIOCINQ:
1233 return -ENOIOCTLCMD;
1234 default:
1235 #ifdef CONFIG_IPV6_MROUTE
1236 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1237 #else
1238 return -ENOIOCTLCMD;
1239 #endif
1242 #endif
1244 static void rawv6_close(struct sock *sk, long timeout)
1246 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1247 ip6_ra_control(sk, -1);
1248 ip6mr_sk_done(sk);
1249 sk_common_release(sk);
1252 static void raw6_destroy(struct sock *sk)
1254 lock_sock(sk);
1255 ip6_flush_pending_frames(sk);
1256 release_sock(sk);
1258 inet6_destroy_sock(sk);
1261 static int rawv6_init_sk(struct sock *sk)
1263 struct raw6_sock *rp = raw6_sk(sk);
1265 switch (inet_sk(sk)->inet_num) {
1266 case IPPROTO_ICMPV6:
1267 rp->checksum = 1;
1268 rp->offset = 2;
1269 break;
1270 case IPPROTO_MH:
1271 rp->checksum = 1;
1272 rp->offset = 4;
1273 break;
1274 default:
1275 break;
1277 return 0;
1280 struct proto rawv6_prot = {
1281 .name = "RAWv6",
1282 .owner = THIS_MODULE,
1283 .close = rawv6_close,
1284 .destroy = raw6_destroy,
1285 .connect = ip6_datagram_connect_v6_only,
1286 .disconnect = __udp_disconnect,
1287 .ioctl = rawv6_ioctl,
1288 .init = rawv6_init_sk,
1289 .setsockopt = rawv6_setsockopt,
1290 .getsockopt = rawv6_getsockopt,
1291 .sendmsg = rawv6_sendmsg,
1292 .recvmsg = rawv6_recvmsg,
1293 .bind = rawv6_bind,
1294 .backlog_rcv = rawv6_rcv_skb,
1295 .hash = raw_hash_sk,
1296 .unhash = raw_unhash_sk,
1297 .obj_size = sizeof(struct raw6_sock),
1298 .useroffset = offsetof(struct raw6_sock, filter),
1299 .usersize = sizeof_field(struct raw6_sock, filter),
1300 .h.raw_hash = &raw_v6_hashinfo,
1301 #ifdef CONFIG_COMPAT
1302 .compat_setsockopt = compat_rawv6_setsockopt,
1303 .compat_getsockopt = compat_rawv6_getsockopt,
1304 .compat_ioctl = compat_rawv6_ioctl,
1305 #endif
1306 .diag_destroy = raw_abort,
1309 #ifdef CONFIG_PROC_FS
1310 static int raw6_seq_show(struct seq_file *seq, void *v)
1312 if (v == SEQ_START_TOKEN) {
1313 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1314 } else {
1315 struct sock *sp = v;
1316 __u16 srcp = inet_sk(sp)->inet_num;
1317 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1318 raw_seq_private(seq)->bucket);
1320 return 0;
1323 static const struct seq_operations raw6_seq_ops = {
1324 .start = raw_seq_start,
1325 .next = raw_seq_next,
1326 .stop = raw_seq_stop,
1327 .show = raw6_seq_show,
1330 static int __net_init raw6_init_net(struct net *net)
1332 if (!proc_create_net_data("raw6", 0444, net->proc_net, &raw6_seq_ops,
1333 sizeof(struct raw_iter_state), &raw_v6_hashinfo))
1334 return -ENOMEM;
1336 return 0;
1339 static void __net_exit raw6_exit_net(struct net *net)
1341 remove_proc_entry("raw6", net->proc_net);
1344 static struct pernet_operations raw6_net_ops = {
1345 .init = raw6_init_net,
1346 .exit = raw6_exit_net,
1349 int __init raw6_proc_init(void)
1351 return register_pernet_subsys(&raw6_net_ops);
1354 void raw6_proc_exit(void)
1356 unregister_pernet_subsys(&raw6_net_ops);
1358 #endif /* CONFIG_PROC_FS */
1360 /* Same as inet6_dgram_ops, sans udp_poll. */
1361 const struct proto_ops inet6_sockraw_ops = {
1362 .family = PF_INET6,
1363 .owner = THIS_MODULE,
1364 .release = inet6_release,
1365 .bind = inet6_bind,
1366 .connect = inet_dgram_connect, /* ok */
1367 .socketpair = sock_no_socketpair, /* a do nothing */
1368 .accept = sock_no_accept, /* a do nothing */
1369 .getname = inet6_getname,
1370 .poll = datagram_poll, /* ok */
1371 .ioctl = inet6_ioctl, /* must change */
1372 .listen = sock_no_listen, /* ok */
1373 .shutdown = inet_shutdown, /* ok */
1374 .setsockopt = sock_common_setsockopt, /* ok */
1375 .getsockopt = sock_common_getsockopt, /* ok */
1376 .sendmsg = inet_sendmsg, /* ok */
1377 .recvmsg = sock_common_recvmsg, /* ok */
1378 .mmap = sock_no_mmap,
1379 .sendpage = sock_no_sendpage,
1380 #ifdef CONFIG_COMPAT
1381 .compat_setsockopt = compat_sock_common_setsockopt,
1382 .compat_getsockopt = compat_sock_common_getsockopt,
1383 #endif
1386 static struct inet_protosw rawv6_protosw = {
1387 .type = SOCK_RAW,
1388 .protocol = IPPROTO_IP, /* wild card */
1389 .prot = &rawv6_prot,
1390 .ops = &inet6_sockraw_ops,
1391 .flags = INET_PROTOSW_REUSE,
1394 int __init rawv6_init(void)
1396 return inet6_register_protosw(&rawv6_protosw);
1399 void rawv6_exit(void)
1401 inet6_unregister_protosw(&rawv6_protosw);