3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Adapted from linux/net/ipv4/raw.c
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>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/ip6_route.h>
47 #include <net/ip6_checksum.h>
48 #include <net/addrconf.h>
49 #include <net/transp_v6.h>
51 #include <net/inet_common.h>
52 #include <net/tcp_states.h>
53 #if IS_ENABLED(CONFIG_IPV6_MIP6)
56 #include <linux/mroute6.h>
59 #include <net/rawv6.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 static struct raw_hashinfo raw_v6_hashinfo
= {
69 .lock
= __RW_LOCK_UNLOCKED(raw_v6_hashinfo
.lock
),
72 static struct sock
*__raw_v6_lookup(struct net
*net
, struct sock
*sk
,
73 unsigned short num
, const struct in6_addr
*loc_addr
,
74 const struct in6_addr
*rmt_addr
, int dif
)
76 bool is_multicast
= ipv6_addr_is_multicast(loc_addr
);
79 if (inet_sk(sk
)->inet_num
== num
) {
81 if (!net_eq(sock_net(sk
), net
))
84 if (!ipv6_addr_any(&sk
->sk_v6_daddr
) &&
85 !ipv6_addr_equal(&sk
->sk_v6_daddr
, rmt_addr
))
88 if (sk
->sk_bound_dev_if
&& sk
->sk_bound_dev_if
!= dif
)
91 if (!ipv6_addr_any(&sk
->sk_v6_rcv_saddr
)) {
92 if (ipv6_addr_equal(&sk
->sk_v6_rcv_saddr
, loc_addr
))
95 inet6_mc_check(sk
, loc_addr
, rmt_addr
))
110 static int icmpv6_filter(const struct sock
*sk
, const struct sk_buff
*skb
)
112 struct icmp6hdr _hdr
;
113 const struct icmp6hdr
*hdr
;
115 /* We require only the four bytes of the ICMPv6 header, not any
116 * additional bytes of message body in "struct icmp6hdr".
118 hdr
= skb_header_pointer(skb
, skb_transport_offset(skb
),
119 ICMPV6_HDRLEN
, &_hdr
);
121 const __u32
*data
= &raw6_sk(sk
)->filter
.data
[0];
122 unsigned int type
= hdr
->icmp6_type
;
124 return (data
[type
>> 5] & (1U << (type
& 31))) != 0;
129 #if IS_ENABLED(CONFIG_IPV6_MIP6)
130 typedef int mh_filter_t(struct sock
*sock
, struct sk_buff
*skb
);
132 static mh_filter_t __rcu
*mh_filter __read_mostly
;
134 int rawv6_mh_filter_register(mh_filter_t filter
)
136 rcu_assign_pointer(mh_filter
, filter
);
139 EXPORT_SYMBOL(rawv6_mh_filter_register
);
141 int rawv6_mh_filter_unregister(mh_filter_t filter
)
143 RCU_INIT_POINTER(mh_filter
, NULL
);
147 EXPORT_SYMBOL(rawv6_mh_filter_unregister
);
152 * demultiplex raw sockets.
153 * (should consider queueing the skb in the sock receive_queue
154 * without calling rawv6.c)
156 * Caller owns SKB so we must make clones.
158 static bool ipv6_raw_deliver(struct sk_buff
*skb
, int nexthdr
)
160 const struct in6_addr
*saddr
;
161 const struct in6_addr
*daddr
;
163 bool delivered
= false;
167 saddr
= &ipv6_hdr(skb
)->saddr
;
170 hash
= nexthdr
& (RAW_HTABLE_SIZE
- 1);
172 read_lock(&raw_v6_hashinfo
.lock
);
173 sk
= sk_head(&raw_v6_hashinfo
.ht
[hash
]);
178 net
= dev_net(skb
->dev
);
179 sk
= __raw_v6_lookup(net
, sk
, nexthdr
, daddr
, saddr
, IP6CB(skb
)->iif
);
187 filtered
= icmpv6_filter(sk
, skb
);
190 #if IS_ENABLED(CONFIG_IPV6_MIP6)
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.
201 filter
= rcu_dereference(mh_filter
);
202 filtered
= filter
? (*filter
)(sk
, skb
) : 0;
214 struct sk_buff
*clone
= skb_clone(skb
, GFP_ATOMIC
);
216 /* Not releasing hash table! */
219 rawv6_rcv(sk
, clone
);
222 sk
= __raw_v6_lookup(net
, sk_next(sk
), nexthdr
, daddr
, saddr
,
226 read_unlock(&raw_v6_hashinfo
.lock
);
230 bool raw6_local_deliver(struct sk_buff
*skb
, int nexthdr
)
234 raw_sk
= sk_head(&raw_v6_hashinfo
.ht
[nexthdr
& (RAW_HTABLE_SIZE
- 1)]);
235 if (raw_sk
&& !ipv6_raw_deliver(skb
, nexthdr
))
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
;
251 if (addr_len
< SIN6_LEN_RFC2133
)
253 addr_type
= ipv6_addr_type(&addr
->sin6_addr
);
255 /* Raw sockets are IPv6 only */
256 if (addr_type
== IPV6_ADDR_MAPPED
)
257 return -EADDRNOTAVAIL
;
262 if (sk
->sk_state
!= TCP_CLOSE
)
266 /* Check if the address belongs to the host. */
267 if (addr_type
!= IPV6_ADDR_ANY
) {
268 struct net_device
*dev
= NULL
;
270 if (__ipv6_addr_needs_scope_id(addr_type
)) {
271 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
272 addr
->sin6_scope_id
) {
273 /* Override any existing binding, if another
274 * one is supplied by user.
276 sk
->sk_bound_dev_if
= addr
->sin6_scope_id
;
279 /* Binding to link-local address requires an interface */
280 if (!sk
->sk_bound_dev_if
)
284 dev
= dev_get_by_index_rcu(sock_net(sk
),
285 sk
->sk_bound_dev_if
);
290 /* ipv4 addr of the socket is invalid. Only the
291 * unspecified and mapped address have a v4 equivalent.
293 v4addr
= LOOPBACK4_IPV6
;
294 if (!(addr_type
& IPV6_ADDR_MULTICAST
)) {
295 err
= -EADDRNOTAVAIL
;
296 if (!ipv6_chk_addr(sock_net(sk
), &addr
->sin6_addr
,
303 inet
->inet_rcv_saddr
= inet
->inet_saddr
= v4addr
;
304 sk
->sk_v6_rcv_saddr
= addr
->sin6_addr
;
305 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
306 np
->saddr
= addr
->sin6_addr
;
315 static void rawv6_err(struct sock
*sk
, struct sk_buff
*skb
,
316 struct inet6_skb_parm
*opt
,
317 u8 type
, u8 code
, int offset
, __be32 info
)
319 struct inet_sock
*inet
= inet_sk(sk
);
320 struct ipv6_pinfo
*np
= inet6_sk(sk
);
324 /* Report error on raw socket, if:
325 1. User requested recverr.
326 2. Socket is connected (otherwise the error indication
327 is useless without recverr and error is hard.
329 if (!np
->recverr
&& sk
->sk_state
!= TCP_ESTABLISHED
)
332 harderr
= icmpv6_err_convert(type
, code
, &err
);
333 if (type
== ICMPV6_PKT_TOOBIG
) {
334 ip6_sk_update_pmtu(skb
, sk
, info
);
335 harderr
= (np
->pmtudisc
== IPV6_PMTUDISC_DO
);
337 if (type
== NDISC_REDIRECT
) {
338 ip6_sk_redirect(skb
, sk
);
342 u8
*payload
= skb
->data
;
345 ipv6_icmp_error(sk
, skb
, err
, 0, ntohl(info
), payload
);
348 if (np
->recverr
|| harderr
) {
350 sk
->sk_error_report(sk
);
354 void raw6_icmp_error(struct sk_buff
*skb
, int nexthdr
,
355 u8 type
, u8 code
, int inner_offset
, __be32 info
)
359 const struct in6_addr
*saddr
, *daddr
;
362 hash
= nexthdr
& (RAW_HTABLE_SIZE
- 1);
364 read_lock(&raw_v6_hashinfo
.lock
);
365 sk
= sk_head(&raw_v6_hashinfo
.ht
[hash
]);
367 /* Note: ipv6_hdr(skb) != skb->data */
368 const struct ipv6hdr
*ip6h
= (const struct ipv6hdr
*)skb
->data
;
369 saddr
= &ip6h
->saddr
;
370 daddr
= &ip6h
->daddr
;
371 net
= dev_net(skb
->dev
);
373 while ((sk
= __raw_v6_lookup(net
, sk
, nexthdr
, saddr
, daddr
,
375 rawv6_err(sk
, skb
, NULL
, type
, code
,
380 read_unlock(&raw_v6_hashinfo
.lock
);
383 static inline int rawv6_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
)
385 if ((raw6_sk(sk
)->checksum
|| rcu_access_pointer(sk
->sk_filter
)) &&
386 skb_checksum_complete(skb
)) {
387 atomic_inc(&sk
->sk_drops
);
392 /* Charge it to the socket. */
394 if (sock_queue_rcv_skb(sk
, skb
) < 0) {
403 * This is next to useless...
404 * if we demultiplex in network layer we don't need the extra call
405 * just to queue the skb...
406 * maybe we could have the network decide upon a hint if it
407 * should call raw_rcv for demultiplexing
409 int rawv6_rcv(struct sock
*sk
, struct sk_buff
*skb
)
411 struct inet_sock
*inet
= inet_sk(sk
);
412 struct raw6_sock
*rp
= raw6_sk(sk
);
414 if (!xfrm6_policy_check(sk
, XFRM_POLICY_IN
, skb
)) {
415 atomic_inc(&sk
->sk_drops
);
421 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
423 if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
424 skb_postpull_rcsum(skb
, skb_network_header(skb
),
425 skb_network_header_len(skb
));
426 if (!csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
427 &ipv6_hdr(skb
)->daddr
,
428 skb
->len
, inet
->inet_num
, skb
->csum
))
429 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
431 if (!skb_csum_unnecessary(skb
))
432 skb
->csum
= ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
433 &ipv6_hdr(skb
)->daddr
,
438 if (skb_checksum_complete(skb
)) {
439 atomic_inc(&sk
->sk_drops
);
445 rawv6_rcv_skb(sk
, skb
);
451 * This should be easy, if there is something there
452 * we return it, otherwise we block.
455 static int rawv6_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
456 struct msghdr
*msg
, size_t len
,
457 int noblock
, int flags
, int *addr_len
)
459 struct ipv6_pinfo
*np
= inet6_sk(sk
);
460 struct sockaddr_in6
*sin6
= (struct sockaddr_in6
*)msg
->msg_name
;
468 if (flags
& MSG_ERRQUEUE
)
469 return ipv6_recv_error(sk
, msg
, len
, addr_len
);
471 if (np
->rxpmtu
&& np
->rxopt
.bits
.rxpmtu
)
472 return ipv6_recv_rxpmtu(sk
, msg
, len
, addr_len
);
474 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
481 msg
->msg_flags
|= MSG_TRUNC
;
484 if (skb_csum_unnecessary(skb
)) {
485 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
486 } else if (msg
->msg_flags
&MSG_TRUNC
) {
487 if (__skb_checksum_complete(skb
))
489 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
491 err
= skb_copy_and_csum_datagram_iovec(skb
, 0, msg
->msg_iov
);
498 /* Copy the address. */
500 sin6
->sin6_family
= AF_INET6
;
502 sin6
->sin6_addr
= ipv6_hdr(skb
)->saddr
;
503 sin6
->sin6_flowinfo
= 0;
504 sin6
->sin6_scope_id
= ipv6_iface_scope_id(&sin6
->sin6_addr
,
506 *addr_len
= sizeof(*sin6
);
509 sock_recv_ts_and_drops(msg
, sk
, skb
);
512 ip6_datagram_recv_ctl(sk
, msg
, skb
);
515 if (flags
& MSG_TRUNC
)
519 skb_free_datagram(sk
, skb
);
524 skb_kill_datagram(sk
, skb
, flags
);
526 /* Error for blocking case is chosen to masquerade
527 as some normal condition.
529 err
= (flags
&MSG_DONTWAIT
) ? -EAGAIN
: -EHOSTUNREACH
;
533 static int rawv6_push_pending_frames(struct sock
*sk
, struct flowi6
*fl6
,
534 struct raw6_sock
*rp
)
547 if ((skb
= skb_peek(&sk
->sk_write_queue
)) == NULL
)
551 total_len
= inet_sk(sk
)->cork
.base
.length
;
552 if (offset
>= total_len
- 1) {
554 ip6_flush_pending_frames(sk
);
558 /* should be check HW csum miyazawa */
559 if (skb_queue_len(&sk
->sk_write_queue
) == 1) {
561 * Only one fragment on the socket.
563 tmp_csum
= skb
->csum
;
565 struct sk_buff
*csum_skb
= NULL
;
568 skb_queue_walk(&sk
->sk_write_queue
, skb
) {
569 tmp_csum
= csum_add(tmp_csum
, skb
->csum
);
574 len
= skb
->len
- skb_transport_offset(skb
);
586 offset
+= skb_transport_offset(skb
);
587 if (skb_copy_bits(skb
, offset
, &csum
, 2))
590 /* in case cksum was not initialized */
592 tmp_csum
= csum_sub(tmp_csum
, csum_unfold(csum
));
594 csum
= csum_ipv6_magic(&fl6
->saddr
, &fl6
->daddr
,
595 total_len
, fl6
->flowi6_proto
, tmp_csum
);
597 if (csum
== 0 && fl6
->flowi6_proto
== IPPROTO_UDP
)
598 csum
= CSUM_MANGLED_0
;
600 if (skb_store_bits(skb
, offset
, &csum
, 2))
604 err
= ip6_push_pending_frames(sk
);
609 static int rawv6_send_hdrinc(struct sock
*sk
, void *from
, int length
,
610 struct flowi6
*fl6
, struct dst_entry
**dstp
,
613 struct ipv6_pinfo
*np
= inet6_sk(sk
);
617 struct rt6_info
*rt
= (struct rt6_info
*)*dstp
;
618 int hlen
= LL_RESERVED_SPACE(rt
->dst
.dev
);
619 int tlen
= rt
->dst
.dev
->needed_tailroom
;
621 if (length
> rt
->dst
.dev
->mtu
) {
622 ipv6_local_error(sk
, EMSGSIZE
, fl6
, rt
->dst
.dev
->mtu
);
628 skb
= sock_alloc_send_skb(sk
,
629 length
+ hlen
+ tlen
+ 15,
630 flags
& MSG_DONTWAIT
, &err
);
633 skb_reserve(skb
, hlen
);
635 skb
->protocol
= htons(ETH_P_IPV6
);
636 skb
->priority
= sk
->sk_priority
;
637 skb
->mark
= sk
->sk_mark
;
638 skb_dst_set(skb
, &rt
->dst
);
641 skb_put(skb
, length
);
642 skb_reset_network_header(skb
);
645 skb
->ip_summed
= CHECKSUM_NONE
;
647 skb
->transport_header
= skb
->network_header
;
648 err
= memcpy_fromiovecend((void *)iph
, from
, 0, length
);
652 IP6_UPD_PO_STATS(sock_net(sk
), rt
->rt6i_idev
, IPSTATS_MIB_OUT
, skb
->len
);
653 err
= NF_HOOK(NFPROTO_IPV6
, NF_INET_LOCAL_OUT
, skb
, NULL
,
654 rt
->dst
.dev
, dst_output
);
656 err
= net_xmit_errno(err
);
666 IP6_INC_STATS(sock_net(sk
), rt
->rt6i_idev
, IPSTATS_MIB_OUTDISCARDS
);
667 if (err
== -ENOBUFS
&& !np
->recverr
)
672 static int rawv6_probe_proto_opt(struct flowi6
*fl6
, struct msghdr
*msg
)
675 u8 __user
*type
= NULL
;
676 u8 __user
*code
= NULL
;
684 for (i
= 0; i
< msg
->msg_iovlen
; i
++) {
685 iov
= &msg
->msg_iov
[i
];
689 switch (fl6
->flowi6_proto
) {
691 /* check if one-byte field is readable or not. */
692 if (iov
->iov_base
&& iov
->iov_len
< 1)
696 type
= iov
->iov_base
;
697 /* check if code field is readable or not. */
698 if (iov
->iov_len
> 1)
701 code
= iov
->iov_base
;
704 if (get_user(fl6
->fl6_icmp_type
, type
) ||
705 get_user(fl6
->fl6_icmp_code
, code
))
711 if (iov
->iov_base
&& iov
->iov_len
< 1)
713 /* check if type field is readable or not. */
714 if (iov
->iov_len
> 2 - len
) {
715 u8 __user
*p
= iov
->iov_base
;
716 if (get_user(fl6
->fl6_mh_type
, &p
[2 - len
]))
733 static int rawv6_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
734 struct msghdr
*msg
, size_t len
)
736 struct ipv6_txoptions opt_space
;
737 struct sockaddr_in6
* sin6
= (struct sockaddr_in6
*) msg
->msg_name
;
738 struct in6_addr
*daddr
, *final_p
, final
;
739 struct inet_sock
*inet
= inet_sk(sk
);
740 struct ipv6_pinfo
*np
= inet6_sk(sk
);
741 struct raw6_sock
*rp
= raw6_sk(sk
);
742 struct ipv6_txoptions
*opt
= NULL
;
743 struct ip6_flowlabel
*flowlabel
= NULL
;
744 struct dst_entry
*dst
= NULL
;
746 int addr_len
= msg
->msg_namelen
;
753 /* Rough check on arithmetic overflow,
754 better check is made in ip6_append_data().
759 /* Mirror BSD error message compatibility */
760 if (msg
->msg_flags
& MSG_OOB
)
764 * Get and verify the address.
766 memset(&fl6
, 0, sizeof(fl6
));
768 fl6
.flowi6_mark
= sk
->sk_mark
;
771 if (addr_len
< SIN6_LEN_RFC2133
)
774 if (sin6
->sin6_family
&& sin6
->sin6_family
!= AF_INET6
)
775 return -EAFNOSUPPORT
;
777 /* port is the proto value [0..255] carried in nexthdr */
778 proto
= ntohs(sin6
->sin6_port
);
781 proto
= inet
->inet_num
;
782 else if (proto
!= inet
->inet_num
)
788 daddr
= &sin6
->sin6_addr
;
790 fl6
.flowlabel
= sin6
->sin6_flowinfo
&IPV6_FLOWINFO_MASK
;
791 if (fl6
.flowlabel
&IPV6_FLOWLABEL_MASK
) {
792 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
793 if (flowlabel
== NULL
)
799 * Otherwise it will be difficult to maintain
802 if (sk
->sk_state
== TCP_ESTABLISHED
&&
803 ipv6_addr_equal(daddr
, &sk
->sk_v6_daddr
))
804 daddr
= &sk
->sk_v6_daddr
;
806 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
807 sin6
->sin6_scope_id
&&
808 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr
)))
809 fl6
.flowi6_oif
= sin6
->sin6_scope_id
;
811 if (sk
->sk_state
!= TCP_ESTABLISHED
)
812 return -EDESTADDRREQ
;
814 proto
= inet
->inet_num
;
815 daddr
= &sk
->sk_v6_daddr
;
816 fl6
.flowlabel
= np
->flow_label
;
819 if (fl6
.flowi6_oif
== 0)
820 fl6
.flowi6_oif
= sk
->sk_bound_dev_if
;
822 if (msg
->msg_controllen
) {
824 memset(opt
, 0, sizeof(struct ipv6_txoptions
));
825 opt
->tot_len
= sizeof(struct ipv6_txoptions
);
827 err
= ip6_datagram_send_ctl(sock_net(sk
), sk
, msg
, &fl6
, opt
,
828 &hlimit
, &tclass
, &dontfrag
);
830 fl6_sock_release(flowlabel
);
833 if ((fl6
.flowlabel
&IPV6_FLOWLABEL_MASK
) && !flowlabel
) {
834 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
835 if (flowlabel
== NULL
)
838 if (!(opt
->opt_nflen
|opt
->opt_flen
))
844 opt
= fl6_merge_options(&opt_space
, flowlabel
, opt
);
845 opt
= ipv6_fixup_options(&opt_space
, opt
);
847 fl6
.flowi6_proto
= proto
;
848 err
= rawv6_probe_proto_opt(&fl6
, msg
);
852 if (!ipv6_addr_any(daddr
))
855 fl6
.daddr
.s6_addr
[15] = 0x1; /* :: means loopback (BSD'ism) */
856 if (ipv6_addr_any(&fl6
.saddr
) && !ipv6_addr_any(&np
->saddr
))
857 fl6
.saddr
= np
->saddr
;
859 final_p
= fl6_update_dst(&fl6
, opt
, &final
);
861 if (!fl6
.flowi6_oif
&& ipv6_addr_is_multicast(&fl6
.daddr
))
862 fl6
.flowi6_oif
= np
->mcast_oif
;
863 else if (!fl6
.flowi6_oif
)
864 fl6
.flowi6_oif
= np
->ucast_oif
;
865 security_sk_classify_flow(sk
, flowi6_to_flowi(&fl6
));
867 dst
= ip6_dst_lookup_flow(sk
, &fl6
, final_p
, true);
873 if (ipv6_addr_is_multicast(&fl6
.daddr
))
874 hlimit
= np
->mcast_hops
;
876 hlimit
= np
->hop_limit
;
878 hlimit
= ip6_dst_hoplimit(dst
);
885 dontfrag
= np
->dontfrag
;
887 if (msg
->msg_flags
&MSG_CONFIRM
)
892 err
= rawv6_send_hdrinc(sk
, msg
->msg_iov
, len
, &fl6
, &dst
, msg
->msg_flags
);
895 err
= ip6_append_data(sk
, ip_generic_getfrag
, msg
->msg_iov
,
896 len
, 0, hlimit
, tclass
, opt
, &fl6
, (struct rt6_info
*)dst
,
897 msg
->msg_flags
, dontfrag
);
900 ip6_flush_pending_frames(sk
);
901 else if (!(msg
->msg_flags
& MSG_MORE
))
902 err
= rawv6_push_pending_frames(sk
, &fl6
, rp
);
908 fl6_sock_release(flowlabel
);
909 return err
<0?err
:len
;
912 if (!(msg
->msg_flags
& MSG_PROBE
) || len
)
913 goto back_from_confirm
;
918 static int rawv6_seticmpfilter(struct sock
*sk
, int level
, int optname
,
919 char __user
*optval
, int optlen
)
923 if (optlen
> sizeof(struct icmp6_filter
))
924 optlen
= sizeof(struct icmp6_filter
);
925 if (copy_from_user(&raw6_sk(sk
)->filter
, optval
, optlen
))
935 static int rawv6_geticmpfilter(struct sock
*sk
, int level
, int optname
,
936 char __user
*optval
, int __user
*optlen
)
942 if (get_user(len
, optlen
))
946 if (len
> sizeof(struct icmp6_filter
))
947 len
= sizeof(struct icmp6_filter
);
948 if (put_user(len
, optlen
))
950 if (copy_to_user(optval
, &raw6_sk(sk
)->filter
, len
))
961 static int do_rawv6_setsockopt(struct sock
*sk
, int level
, int optname
,
962 char __user
*optval
, unsigned int optlen
)
964 struct raw6_sock
*rp
= raw6_sk(sk
);
967 if (get_user(val
, (int __user
*)optval
))
972 if (inet_sk(sk
)->inet_num
== IPPROTO_ICMPV6
&&
973 level
== IPPROTO_IPV6
) {
975 * RFC3542 tells that IPV6_CHECKSUM socket
976 * option in the IPPROTO_IPV6 level is not
977 * allowed on ICMPv6 sockets.
978 * If you want to set it, use IPPROTO_RAW
979 * level IPV6_CHECKSUM socket option
985 /* You may get strange result with a positive odd offset;
986 RFC2292bis agrees with me. */
987 if (val
> 0 && (val
&1))
1003 static int rawv6_setsockopt(struct sock
*sk
, int level
, int optname
,
1004 char __user
*optval
, unsigned int optlen
)
1011 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1013 return rawv6_seticmpfilter(sk
, level
, optname
, optval
, optlen
);
1015 if (optname
== IPV6_CHECKSUM
)
1018 return ipv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1021 return do_rawv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1024 #ifdef CONFIG_COMPAT
1025 static int compat_rawv6_setsockopt(struct sock
*sk
, int level
, int optname
,
1026 char __user
*optval
, unsigned int optlen
)
1032 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1034 return rawv6_seticmpfilter(sk
, level
, optname
, optval
, optlen
);
1036 if (optname
== IPV6_CHECKSUM
)
1039 return compat_ipv6_setsockopt(sk
, level
, optname
,
1042 return do_rawv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1046 static int do_rawv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1047 char __user
*optval
, int __user
*optlen
)
1049 struct raw6_sock
*rp
= raw6_sk(sk
);
1052 if (get_user(len
,optlen
))
1058 * We allow getsockopt() for IPPROTO_IPV6-level
1059 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1060 * since RFC3542 is silent about it.
1062 if (rp
->checksum
== 0)
1069 return -ENOPROTOOPT
;
1072 len
= min_t(unsigned int, sizeof(int), len
);
1074 if (put_user(len
, optlen
))
1076 if (copy_to_user(optval
,&val
,len
))
1081 static int rawv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1082 char __user
*optval
, int __user
*optlen
)
1089 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1091 return rawv6_geticmpfilter(sk
, level
, optname
, optval
, optlen
);
1093 if (optname
== IPV6_CHECKSUM
)
1096 return ipv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1099 return do_rawv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1102 #ifdef CONFIG_COMPAT
1103 static int compat_rawv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1104 char __user
*optval
, int __user
*optlen
)
1110 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1112 return rawv6_geticmpfilter(sk
, level
, optname
, optval
, optlen
);
1114 if (optname
== IPV6_CHECKSUM
)
1117 return compat_ipv6_getsockopt(sk
, level
, optname
,
1120 return do_rawv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1124 static int rawv6_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
1128 int amount
= sk_wmem_alloc_get(sk
);
1130 return put_user(amount
, (int __user
*)arg
);
1133 struct sk_buff
*skb
;
1136 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
1137 skb
= skb_peek(&sk
->sk_receive_queue
);
1139 amount
= skb_tail_pointer(skb
) -
1140 skb_transport_header(skb
);
1141 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
1142 return put_user(amount
, (int __user
*)arg
);
1146 #ifdef CONFIG_IPV6_MROUTE
1147 return ip6mr_ioctl(sk
, cmd
, (void __user
*)arg
);
1149 return -ENOIOCTLCMD
;
1154 #ifdef CONFIG_COMPAT
1155 static int compat_rawv6_ioctl(struct sock
*sk
, unsigned int cmd
, unsigned long arg
)
1160 return -ENOIOCTLCMD
;
1162 #ifdef CONFIG_IPV6_MROUTE
1163 return ip6mr_compat_ioctl(sk
, cmd
, compat_ptr(arg
));
1165 return -ENOIOCTLCMD
;
1171 static void rawv6_close(struct sock
*sk
, long timeout
)
1173 if (inet_sk(sk
)->inet_num
== IPPROTO_RAW
)
1174 ip6_ra_control(sk
, -1);
1176 sk_common_release(sk
);
1179 static void raw6_destroy(struct sock
*sk
)
1182 ip6_flush_pending_frames(sk
);
1185 inet6_destroy_sock(sk
);
1188 static int rawv6_init_sk(struct sock
*sk
)
1190 struct raw6_sock
*rp
= raw6_sk(sk
);
1192 switch (inet_sk(sk
)->inet_num
) {
1193 case IPPROTO_ICMPV6
:
1207 struct proto rawv6_prot
= {
1209 .owner
= THIS_MODULE
,
1210 .close
= rawv6_close
,
1211 .destroy
= raw6_destroy
,
1212 .connect
= ip6_datagram_connect
,
1213 .disconnect
= udp_disconnect
,
1214 .ioctl
= rawv6_ioctl
,
1215 .init
= rawv6_init_sk
,
1216 .setsockopt
= rawv6_setsockopt
,
1217 .getsockopt
= rawv6_getsockopt
,
1218 .sendmsg
= rawv6_sendmsg
,
1219 .recvmsg
= rawv6_recvmsg
,
1221 .backlog_rcv
= rawv6_rcv_skb
,
1222 .hash
= raw_hash_sk
,
1223 .unhash
= raw_unhash_sk
,
1224 .obj_size
= sizeof(struct raw6_sock
),
1225 .h
.raw_hash
= &raw_v6_hashinfo
,
1226 #ifdef CONFIG_COMPAT
1227 .compat_setsockopt
= compat_rawv6_setsockopt
,
1228 .compat_getsockopt
= compat_rawv6_getsockopt
,
1229 .compat_ioctl
= compat_rawv6_ioctl
,
1233 #ifdef CONFIG_PROC_FS
1234 static int raw6_seq_show(struct seq_file
*seq
, void *v
)
1236 if (v
== SEQ_START_TOKEN
) {
1237 seq_puts(seq
, IPV6_SEQ_DGRAM_HEADER
);
1239 struct sock
*sp
= v
;
1240 __u16 srcp
= inet_sk(sp
)->inet_num
;
1241 ip6_dgram_sock_seq_show(seq
, v
, srcp
, 0,
1242 raw_seq_private(seq
)->bucket
);
1247 static const struct seq_operations raw6_seq_ops
= {
1248 .start
= raw_seq_start
,
1249 .next
= raw_seq_next
,
1250 .stop
= raw_seq_stop
,
1251 .show
= raw6_seq_show
,
1254 static int raw6_seq_open(struct inode
*inode
, struct file
*file
)
1256 return raw_seq_open(inode
, file
, &raw_v6_hashinfo
, &raw6_seq_ops
);
1259 static const struct file_operations raw6_seq_fops
= {
1260 .owner
= THIS_MODULE
,
1261 .open
= raw6_seq_open
,
1263 .llseek
= seq_lseek
,
1264 .release
= seq_release_net
,
1267 static int __net_init
raw6_init_net(struct net
*net
)
1269 if (!proc_create("raw6", S_IRUGO
, net
->proc_net
, &raw6_seq_fops
))
1275 static void __net_exit
raw6_exit_net(struct net
*net
)
1277 remove_proc_entry("raw6", net
->proc_net
);
1280 static struct pernet_operations raw6_net_ops
= {
1281 .init
= raw6_init_net
,
1282 .exit
= raw6_exit_net
,
1285 int __init
raw6_proc_init(void)
1287 return register_pernet_subsys(&raw6_net_ops
);
1290 void raw6_proc_exit(void)
1292 unregister_pernet_subsys(&raw6_net_ops
);
1294 #endif /* CONFIG_PROC_FS */
1296 /* Same as inet6_dgram_ops, sans udp_poll. */
1297 static const struct proto_ops inet6_sockraw_ops
= {
1299 .owner
= THIS_MODULE
,
1300 .release
= inet6_release
,
1302 .connect
= inet_dgram_connect
, /* ok */
1303 .socketpair
= sock_no_socketpair
, /* a do nothing */
1304 .accept
= sock_no_accept
, /* a do nothing */
1305 .getname
= inet6_getname
,
1306 .poll
= datagram_poll
, /* ok */
1307 .ioctl
= inet6_ioctl
, /* must change */
1308 .listen
= sock_no_listen
, /* ok */
1309 .shutdown
= inet_shutdown
, /* ok */
1310 .setsockopt
= sock_common_setsockopt
, /* ok */
1311 .getsockopt
= sock_common_getsockopt
, /* ok */
1312 .sendmsg
= inet_sendmsg
, /* ok */
1313 .recvmsg
= sock_common_recvmsg
, /* ok */
1314 .mmap
= sock_no_mmap
,
1315 .sendpage
= sock_no_sendpage
,
1316 #ifdef CONFIG_COMPAT
1317 .compat_setsockopt
= compat_sock_common_setsockopt
,
1318 .compat_getsockopt
= compat_sock_common_getsockopt
,
1322 static struct inet_protosw rawv6_protosw
= {
1324 .protocol
= IPPROTO_IP
, /* wild card */
1325 .prot
= &rawv6_prot
,
1326 .ops
= &inet6_sockraw_ops
,
1327 .no_check
= UDP_CSUM_DEFAULT
,
1328 .flags
= INET_PROTOSW_REUSE
,
1331 int __init
rawv6_init(void)
1335 ret
= inet6_register_protosw(&rawv6_protosw
);
1342 void rawv6_exit(void)
1344 inet6_unregister_protosw(&rawv6_protosw
);