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 defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
56 #include <linux/mroute6.h>
59 #include <net/rawv6.h>
62 #include <linux/proc_fs.h>
63 #include <linux/seq_file.h>
65 static struct raw_hashinfo raw_v6_hashinfo
= {
66 .lock
= __RW_LOCK_UNLOCKED(raw_v6_hashinfo
.lock
),
69 static struct sock
*__raw_v6_lookup(struct net
*net
, struct sock
*sk
,
70 unsigned short num
, const struct in6_addr
*loc_addr
,
71 const struct in6_addr
*rmt_addr
, int dif
)
73 struct hlist_node
*node
;
74 int is_multicast
= ipv6_addr_is_multicast(loc_addr
);
76 sk_for_each_from(sk
, node
)
77 if (inet_sk(sk
)->inet_num
== num
) {
78 struct ipv6_pinfo
*np
= inet6_sk(sk
);
80 if (!net_eq(sock_net(sk
), net
))
83 if (!ipv6_addr_any(&np
->daddr
) &&
84 !ipv6_addr_equal(&np
->daddr
, rmt_addr
))
87 if (sk
->sk_bound_dev_if
&& sk
->sk_bound_dev_if
!= dif
)
90 if (!ipv6_addr_any(&np
->rcv_saddr
)) {
91 if (ipv6_addr_equal(&np
->rcv_saddr
, loc_addr
))
94 inet6_mc_check(sk
, loc_addr
, rmt_addr
))
109 static __inline__
int icmpv6_filter(struct sock
*sk
, struct sk_buff
*skb
)
111 struct icmp6hdr
*icmph
;
112 struct raw6_sock
*rp
= raw6_sk(sk
);
114 if (pskb_may_pull(skb
, sizeof(struct icmp6hdr
))) {
115 __u32
*data
= &rp
->filter
.data
[0];
118 icmph
= (struct icmp6hdr
*) skb
->data
;
119 bit_nr
= icmph
->icmp6_type
;
121 return (data
[bit_nr
>> 5] & (1 << (bit_nr
& 31))) != 0;
126 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
127 typedef int mh_filter_t(struct sock
*sock
, struct sk_buff
*skb
);
129 static mh_filter_t __rcu
*mh_filter __read_mostly
;
131 int rawv6_mh_filter_register(mh_filter_t filter
)
133 rcu_assign_pointer(mh_filter
, filter
);
136 EXPORT_SYMBOL(rawv6_mh_filter_register
);
138 int rawv6_mh_filter_unregister(mh_filter_t filter
)
140 rcu_assign_pointer(mh_filter
, NULL
);
144 EXPORT_SYMBOL(rawv6_mh_filter_unregister
);
149 * demultiplex raw sockets.
150 * (should consider queueing the skb in the sock receive_queue
151 * without calling rawv6.c)
153 * Caller owns SKB so we must make clones.
155 static int ipv6_raw_deliver(struct sk_buff
*skb
, int nexthdr
)
157 const struct in6_addr
*saddr
;
158 const struct in6_addr
*daddr
;
164 saddr
= &ipv6_hdr(skb
)->saddr
;
167 hash
= nexthdr
& (MAX_INET_PROTOS
- 1);
169 read_lock(&raw_v6_hashinfo
.lock
);
170 sk
= sk_head(&raw_v6_hashinfo
.ht
[hash
]);
175 net
= dev_net(skb
->dev
);
176 sk
= __raw_v6_lookup(net
, sk
, nexthdr
, daddr
, saddr
, IP6CB(skb
)->iif
);
184 filtered
= icmpv6_filter(sk
, skb
);
187 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
190 /* XXX: To validate MH only once for each packet,
191 * this is placed here. It should be after checking
192 * xfrm policy, however it doesn't. The checking xfrm
193 * policy is placed in rawv6_rcv() because it is
194 * required for each socket.
198 filter
= rcu_dereference(mh_filter
);
199 filtered
= filter
? (*filter
)(sk
, skb
) : 0;
211 struct sk_buff
*clone
= skb_clone(skb
, GFP_ATOMIC
);
213 /* Not releasing hash table! */
216 rawv6_rcv(sk
, clone
);
219 sk
= __raw_v6_lookup(net
, sk_next(sk
), nexthdr
, daddr
, saddr
,
223 read_unlock(&raw_v6_hashinfo
.lock
);
227 int raw6_local_deliver(struct sk_buff
*skb
, int nexthdr
)
231 raw_sk
= sk_head(&raw_v6_hashinfo
.ht
[nexthdr
& (MAX_INET_PROTOS
- 1)]);
232 if (raw_sk
&& !ipv6_raw_deliver(skb
, nexthdr
))
235 return raw_sk
!= NULL
;
238 /* This cleans up af_inet6 a bit. -DaveM */
239 static int rawv6_bind(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
241 struct inet_sock
*inet
= inet_sk(sk
);
242 struct ipv6_pinfo
*np
= inet6_sk(sk
);
243 struct sockaddr_in6
*addr
= (struct sockaddr_in6
*) uaddr
;
248 if (addr_len
< SIN6_LEN_RFC2133
)
250 addr_type
= ipv6_addr_type(&addr
->sin6_addr
);
252 /* Raw sockets are IPv6 only */
253 if (addr_type
== IPV6_ADDR_MAPPED
)
254 return -EADDRNOTAVAIL
;
259 if (sk
->sk_state
!= TCP_CLOSE
)
263 /* Check if the address belongs to the host. */
264 if (addr_type
!= IPV6_ADDR_ANY
) {
265 struct net_device
*dev
= NULL
;
267 if (addr_type
& IPV6_ADDR_LINKLOCAL
) {
268 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
269 addr
->sin6_scope_id
) {
270 /* Override any existing binding, if another
271 * one is supplied by user.
273 sk
->sk_bound_dev_if
= addr
->sin6_scope_id
;
276 /* Binding to link-local address requires an interface */
277 if (!sk
->sk_bound_dev_if
)
281 dev
= dev_get_by_index_rcu(sock_net(sk
),
282 sk
->sk_bound_dev_if
);
287 /* ipv4 addr of the socket is invalid. Only the
288 * unspecified and mapped address have a v4 equivalent.
290 v4addr
= LOOPBACK4_IPV6
;
291 if (!(addr_type
& IPV6_ADDR_MULTICAST
)) {
292 err
= -EADDRNOTAVAIL
;
293 if (!ipv6_chk_addr(sock_net(sk
), &addr
->sin6_addr
,
300 inet
->inet_rcv_saddr
= inet
->inet_saddr
= v4addr
;
301 ipv6_addr_copy(&np
->rcv_saddr
, &addr
->sin6_addr
);
302 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
303 ipv6_addr_copy(&np
->saddr
, &addr
->sin6_addr
);
312 static void rawv6_err(struct sock
*sk
, struct sk_buff
*skb
,
313 struct inet6_skb_parm
*opt
,
314 u8 type
, u8 code
, int offset
, __be32 info
)
316 struct inet_sock
*inet
= inet_sk(sk
);
317 struct ipv6_pinfo
*np
= inet6_sk(sk
);
321 /* Report error on raw socket, if:
322 1. User requested recverr.
323 2. Socket is connected (otherwise the error indication
324 is useless without recverr and error is hard.
326 if (!np
->recverr
&& sk
->sk_state
!= TCP_ESTABLISHED
)
329 harderr
= icmpv6_err_convert(type
, code
, &err
);
330 if (type
== ICMPV6_PKT_TOOBIG
)
331 harderr
= (np
->pmtudisc
== IPV6_PMTUDISC_DO
);
334 u8
*payload
= skb
->data
;
337 ipv6_icmp_error(sk
, skb
, err
, 0, ntohl(info
), payload
);
340 if (np
->recverr
|| harderr
) {
342 sk
->sk_error_report(sk
);
346 void raw6_icmp_error(struct sk_buff
*skb
, int nexthdr
,
347 u8 type
, u8 code
, int inner_offset
, __be32 info
)
351 const struct in6_addr
*saddr
, *daddr
;
354 hash
= nexthdr
& (RAW_HTABLE_SIZE
- 1);
356 read_lock(&raw_v6_hashinfo
.lock
);
357 sk
= sk_head(&raw_v6_hashinfo
.ht
[hash
]);
359 /* Note: ipv6_hdr(skb) != skb->data */
360 const struct ipv6hdr
*ip6h
= (const struct ipv6hdr
*)skb
->data
;
361 saddr
= &ip6h
->saddr
;
362 daddr
= &ip6h
->daddr
;
363 net
= dev_net(skb
->dev
);
365 while ((sk
= __raw_v6_lookup(net
, sk
, nexthdr
, saddr
, daddr
,
367 rawv6_err(sk
, skb
, NULL
, type
, code
,
372 read_unlock(&raw_v6_hashinfo
.lock
);
375 static inline int rawv6_rcv_skb(struct sock
* sk
, struct sk_buff
* skb
)
377 if ((raw6_sk(sk
)->checksum
|| rcu_dereference_raw(sk
->sk_filter
)) &&
378 skb_checksum_complete(skb
)) {
379 atomic_inc(&sk
->sk_drops
);
384 /* Charge it to the socket. */
385 if (ip_queue_rcv_skb(sk
, skb
) < 0) {
394 * This is next to useless...
395 * if we demultiplex in network layer we don't need the extra call
396 * just to queue the skb...
397 * maybe we could have the network decide upon a hint if it
398 * should call raw_rcv for demultiplexing
400 int rawv6_rcv(struct sock
*sk
, struct sk_buff
*skb
)
402 struct inet_sock
*inet
= inet_sk(sk
);
403 struct raw6_sock
*rp
= raw6_sk(sk
);
405 if (!xfrm6_policy_check(sk
, XFRM_POLICY_IN
, skb
)) {
406 atomic_inc(&sk
->sk_drops
);
412 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
414 if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
415 skb_postpull_rcsum(skb
, skb_network_header(skb
),
416 skb_network_header_len(skb
));
417 if (!csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
418 &ipv6_hdr(skb
)->daddr
,
419 skb
->len
, inet
->inet_num
, skb
->csum
))
420 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
422 if (!skb_csum_unnecessary(skb
))
423 skb
->csum
= ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
424 &ipv6_hdr(skb
)->daddr
,
429 if (skb_checksum_complete(skb
)) {
430 atomic_inc(&sk
->sk_drops
);
436 rawv6_rcv_skb(sk
, skb
);
442 * This should be easy, if there is something there
443 * we return it, otherwise we block.
446 static int rawv6_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
447 struct msghdr
*msg
, size_t len
,
448 int noblock
, int flags
, int *addr_len
)
450 struct ipv6_pinfo
*np
= inet6_sk(sk
);
451 struct sockaddr_in6
*sin6
= (struct sockaddr_in6
*)msg
->msg_name
;
460 *addr_len
=sizeof(*sin6
);
462 if (flags
& MSG_ERRQUEUE
)
463 return ipv6_recv_error(sk
, msg
, len
);
465 if (np
->rxpmtu
&& np
->rxopt
.bits
.rxpmtu
)
466 return ipv6_recv_rxpmtu(sk
, msg
, len
);
468 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
475 msg
->msg_flags
|= MSG_TRUNC
;
478 if (skb_csum_unnecessary(skb
)) {
479 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
480 } else if (msg
->msg_flags
&MSG_TRUNC
) {
481 if (__skb_checksum_complete(skb
))
483 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
485 err
= skb_copy_and_csum_datagram_iovec(skb
, 0, msg
->msg_iov
);
492 /* Copy the address. */
494 sin6
->sin6_family
= AF_INET6
;
496 ipv6_addr_copy(&sin6
->sin6_addr
, &ipv6_hdr(skb
)->saddr
);
497 sin6
->sin6_flowinfo
= 0;
498 sin6
->sin6_scope_id
= 0;
499 if (ipv6_addr_type(&sin6
->sin6_addr
) & IPV6_ADDR_LINKLOCAL
)
500 sin6
->sin6_scope_id
= IP6CB(skb
)->iif
;
503 sock_recv_ts_and_drops(msg
, sk
, skb
);
506 datagram_recv_ctl(sk
, msg
, skb
);
509 if (flags
& MSG_TRUNC
)
513 skb_free_datagram(sk
, skb
);
518 skb_kill_datagram(sk
, skb
, flags
);
520 /* Error for blocking case is chosen to masquerade
521 as some normal condition.
523 err
= (flags
&MSG_DONTWAIT
) ? -EAGAIN
: -EHOSTUNREACH
;
527 static int rawv6_push_pending_frames(struct sock
*sk
, struct flowi6
*fl6
,
528 struct raw6_sock
*rp
)
541 if ((skb
= skb_peek(&sk
->sk_write_queue
)) == NULL
)
545 total_len
= inet_sk(sk
)->cork
.base
.length
- (skb_network_header(skb
) -
547 if (offset
>= total_len
- 1) {
549 ip6_flush_pending_frames(sk
);
553 /* should be check HW csum miyazawa */
554 if (skb_queue_len(&sk
->sk_write_queue
) == 1) {
556 * Only one fragment on the socket.
558 tmp_csum
= skb
->csum
;
560 struct sk_buff
*csum_skb
= NULL
;
563 skb_queue_walk(&sk
->sk_write_queue
, skb
) {
564 tmp_csum
= csum_add(tmp_csum
, skb
->csum
);
569 len
= skb
->len
- skb_transport_offset(skb
);
581 offset
+= skb_transport_offset(skb
);
582 if (skb_copy_bits(skb
, offset
, &csum
, 2))
585 /* in case cksum was not initialized */
587 tmp_csum
= csum_sub(tmp_csum
, csum_unfold(csum
));
589 csum
= csum_ipv6_magic(&fl6
->saddr
, &fl6
->daddr
,
590 total_len
, fl6
->flowi6_proto
, tmp_csum
);
592 if (csum
== 0 && fl6
->flowi6_proto
== IPPROTO_UDP
)
593 csum
= CSUM_MANGLED_0
;
595 if (skb_store_bits(skb
, offset
, &csum
, 2))
599 err
= ip6_push_pending_frames(sk
);
604 static int rawv6_send_hdrinc(struct sock
*sk
, void *from
, int length
,
605 struct flowi6
*fl6
, struct dst_entry
**dstp
,
608 struct ipv6_pinfo
*np
= inet6_sk(sk
);
612 struct rt6_info
*rt
= (struct rt6_info
*)*dstp
;
614 if (length
> rt
->dst
.dev
->mtu
) {
615 ipv6_local_error(sk
, EMSGSIZE
, fl6
, rt
->dst
.dev
->mtu
);
621 skb
= sock_alloc_send_skb(sk
,
622 length
+ LL_ALLOCATED_SPACE(rt
->dst
.dev
) + 15,
623 flags
& MSG_DONTWAIT
, &err
);
626 skb_reserve(skb
, LL_RESERVED_SPACE(rt
->dst
.dev
));
628 skb
->priority
= sk
->sk_priority
;
629 skb
->mark
= sk
->sk_mark
;
630 skb_dst_set(skb
, &rt
->dst
);
633 skb_put(skb
, length
);
634 skb_reset_network_header(skb
);
637 skb
->ip_summed
= CHECKSUM_NONE
;
639 skb
->transport_header
= skb
->network_header
;
640 err
= memcpy_fromiovecend((void *)iph
, from
, 0, length
);
644 IP6_UPD_PO_STATS(sock_net(sk
), rt
->rt6i_idev
, IPSTATS_MIB_OUT
, skb
->len
);
645 err
= NF_HOOK(NFPROTO_IPV6
, NF_INET_LOCAL_OUT
, skb
, NULL
,
646 rt
->dst
.dev
, dst_output
);
648 err
= net_xmit_errno(err
);
658 IP6_INC_STATS(sock_net(sk
), rt
->rt6i_idev
, IPSTATS_MIB_OUTDISCARDS
);
659 if (err
== -ENOBUFS
&& !np
->recverr
)
664 static int rawv6_probe_proto_opt(struct flowi6
*fl6
, struct msghdr
*msg
)
667 u8 __user
*type
= NULL
;
668 u8 __user
*code
= NULL
;
676 for (i
= 0; i
< msg
->msg_iovlen
; i
++) {
677 iov
= &msg
->msg_iov
[i
];
681 switch (fl6
->flowi6_proto
) {
683 /* check if one-byte field is readable or not. */
684 if (iov
->iov_base
&& iov
->iov_len
< 1)
688 type
= iov
->iov_base
;
689 /* check if code field is readable or not. */
690 if (iov
->iov_len
> 1)
693 code
= iov
->iov_base
;
696 if (get_user(fl6
->fl6_icmp_type
, type
) ||
697 get_user(fl6
->fl6_icmp_code
, code
))
703 if (iov
->iov_base
&& iov
->iov_len
< 1)
705 /* check if type field is readable or not. */
706 if (iov
->iov_len
> 2 - len
) {
707 u8 __user
*p
= iov
->iov_base
;
708 if (get_user(fl6
->fl6_mh_type
, &p
[2 - len
]))
725 static int rawv6_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
726 struct msghdr
*msg
, size_t len
)
728 struct ipv6_txoptions opt_space
;
729 struct sockaddr_in6
* sin6
= (struct sockaddr_in6
*) msg
->msg_name
;
730 struct in6_addr
*daddr
, *final_p
, final
;
731 struct inet_sock
*inet
= inet_sk(sk
);
732 struct ipv6_pinfo
*np
= inet6_sk(sk
);
733 struct raw6_sock
*rp
= raw6_sk(sk
);
734 struct ipv6_txoptions
*opt
= NULL
;
735 struct ip6_flowlabel
*flowlabel
= NULL
;
736 struct dst_entry
*dst
= NULL
;
738 int addr_len
= msg
->msg_namelen
;
745 /* Rough check on arithmetic overflow,
746 better check is made in ip6_append_data().
751 /* Mirror BSD error message compatibility */
752 if (msg
->msg_flags
& MSG_OOB
)
756 * Get and verify the address.
758 memset(&fl6
, 0, sizeof(fl6
));
760 fl6
.flowi6_mark
= sk
->sk_mark
;
763 if (addr_len
< SIN6_LEN_RFC2133
)
766 if (sin6
->sin6_family
&& sin6
->sin6_family
!= AF_INET6
)
767 return -EAFNOSUPPORT
;
769 /* port is the proto value [0..255] carried in nexthdr */
770 proto
= ntohs(sin6
->sin6_port
);
773 proto
= inet
->inet_num
;
774 else if (proto
!= inet
->inet_num
)
780 daddr
= &sin6
->sin6_addr
;
782 fl6
.flowlabel
= sin6
->sin6_flowinfo
&IPV6_FLOWINFO_MASK
;
783 if (fl6
.flowlabel
&IPV6_FLOWLABEL_MASK
) {
784 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
785 if (flowlabel
== NULL
)
787 daddr
= &flowlabel
->dst
;
792 * Otherwise it will be difficult to maintain
795 if (sk
->sk_state
== TCP_ESTABLISHED
&&
796 ipv6_addr_equal(daddr
, &np
->daddr
))
799 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
800 sin6
->sin6_scope_id
&&
801 ipv6_addr_type(daddr
)&IPV6_ADDR_LINKLOCAL
)
802 fl6
.flowi6_oif
= sin6
->sin6_scope_id
;
804 if (sk
->sk_state
!= TCP_ESTABLISHED
)
805 return -EDESTADDRREQ
;
807 proto
= inet
->inet_num
;
809 fl6
.flowlabel
= np
->flow_label
;
812 if (fl6
.flowi6_oif
== 0)
813 fl6
.flowi6_oif
= sk
->sk_bound_dev_if
;
815 if (msg
->msg_controllen
) {
817 memset(opt
, 0, sizeof(struct ipv6_txoptions
));
818 opt
->tot_len
= sizeof(struct ipv6_txoptions
);
820 err
= datagram_send_ctl(sock_net(sk
), sk
, msg
, &fl6
, opt
,
821 &hlimit
, &tclass
, &dontfrag
);
823 fl6_sock_release(flowlabel
);
826 if ((fl6
.flowlabel
&IPV6_FLOWLABEL_MASK
) && !flowlabel
) {
827 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
828 if (flowlabel
== NULL
)
831 if (!(opt
->opt_nflen
|opt
->opt_flen
))
837 opt
= fl6_merge_options(&opt_space
, flowlabel
, opt
);
838 opt
= ipv6_fixup_options(&opt_space
, opt
);
840 fl6
.flowi6_proto
= proto
;
841 err
= rawv6_probe_proto_opt(&fl6
, msg
);
845 if (!ipv6_addr_any(daddr
))
846 ipv6_addr_copy(&fl6
.daddr
, daddr
);
848 fl6
.daddr
.s6_addr
[15] = 0x1; /* :: means loopback (BSD'ism) */
849 if (ipv6_addr_any(&fl6
.saddr
) && !ipv6_addr_any(&np
->saddr
))
850 ipv6_addr_copy(&fl6
.saddr
, &np
->saddr
);
852 final_p
= fl6_update_dst(&fl6
, opt
, &final
);
854 if (!fl6
.flowi6_oif
&& ipv6_addr_is_multicast(&fl6
.daddr
))
855 fl6
.flowi6_oif
= np
->mcast_oif
;
856 security_sk_classify_flow(sk
, flowi6_to_flowi(&fl6
));
858 dst
= ip6_dst_lookup_flow(sk
, &fl6
, final_p
, true);
864 if (ipv6_addr_is_multicast(&fl6
.daddr
))
865 hlimit
= np
->mcast_hops
;
867 hlimit
= np
->hop_limit
;
869 hlimit
= ip6_dst_hoplimit(dst
);
876 dontfrag
= np
->dontfrag
;
878 if (msg
->msg_flags
&MSG_CONFIRM
)
883 err
= rawv6_send_hdrinc(sk
, msg
->msg_iov
, len
, &fl6
, &dst
, msg
->msg_flags
);
886 err
= ip6_append_data(sk
, ip_generic_getfrag
, msg
->msg_iov
,
887 len
, 0, hlimit
, tclass
, opt
, &fl6
, (struct rt6_info
*)dst
,
888 msg
->msg_flags
, dontfrag
);
891 ip6_flush_pending_frames(sk
);
892 else if (!(msg
->msg_flags
& MSG_MORE
))
893 err
= rawv6_push_pending_frames(sk
, &fl6
, rp
);
899 fl6_sock_release(flowlabel
);
900 return err
<0?err
:len
;
903 if (!(msg
->msg_flags
& MSG_PROBE
) || len
)
904 goto back_from_confirm
;
909 static int rawv6_seticmpfilter(struct sock
*sk
, int level
, int optname
,
910 char __user
*optval
, int optlen
)
914 if (optlen
> sizeof(struct icmp6_filter
))
915 optlen
= sizeof(struct icmp6_filter
);
916 if (copy_from_user(&raw6_sk(sk
)->filter
, optval
, optlen
))
926 static int rawv6_geticmpfilter(struct sock
*sk
, int level
, int optname
,
927 char __user
*optval
, int __user
*optlen
)
933 if (get_user(len
, optlen
))
937 if (len
> sizeof(struct icmp6_filter
))
938 len
= sizeof(struct icmp6_filter
);
939 if (put_user(len
, optlen
))
941 if (copy_to_user(optval
, &raw6_sk(sk
)->filter
, len
))
952 static int do_rawv6_setsockopt(struct sock
*sk
, int level
, int optname
,
953 char __user
*optval
, unsigned int optlen
)
955 struct raw6_sock
*rp
= raw6_sk(sk
);
958 if (get_user(val
, (int __user
*)optval
))
963 if (inet_sk(sk
)->inet_num
== IPPROTO_ICMPV6
&&
964 level
== IPPROTO_IPV6
) {
966 * RFC3542 tells that IPV6_CHECKSUM socket
967 * option in the IPPROTO_IPV6 level is not
968 * allowed on ICMPv6 sockets.
969 * If you want to set it, use IPPROTO_RAW
970 * level IPV6_CHECKSUM socket option
976 /* You may get strange result with a positive odd offset;
977 RFC2292bis agrees with me. */
978 if (val
> 0 && (val
&1))
994 static int rawv6_setsockopt(struct sock
*sk
, int level
, int optname
,
995 char __user
*optval
, unsigned int optlen
)
1002 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1004 return rawv6_seticmpfilter(sk
, level
, optname
, optval
, optlen
);
1006 if (optname
== IPV6_CHECKSUM
)
1009 return ipv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1012 return do_rawv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1015 #ifdef CONFIG_COMPAT
1016 static int compat_rawv6_setsockopt(struct sock
*sk
, int level
, int optname
,
1017 char __user
*optval
, unsigned int optlen
)
1023 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1025 return rawv6_seticmpfilter(sk
, level
, optname
, optval
, optlen
);
1027 if (optname
== IPV6_CHECKSUM
)
1030 return compat_ipv6_setsockopt(sk
, level
, optname
,
1033 return do_rawv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1037 static int do_rawv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1038 char __user
*optval
, int __user
*optlen
)
1040 struct raw6_sock
*rp
= raw6_sk(sk
);
1043 if (get_user(len
,optlen
))
1049 * We allow getsockopt() for IPPROTO_IPV6-level
1050 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1051 * since RFC3542 is silent about it.
1053 if (rp
->checksum
== 0)
1060 return -ENOPROTOOPT
;
1063 len
= min_t(unsigned int, sizeof(int), len
);
1065 if (put_user(len
, optlen
))
1067 if (copy_to_user(optval
,&val
,len
))
1072 static int rawv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1073 char __user
*optval
, int __user
*optlen
)
1080 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1082 return rawv6_geticmpfilter(sk
, level
, optname
, optval
, optlen
);
1084 if (optname
== IPV6_CHECKSUM
)
1087 return ipv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1090 return do_rawv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1093 #ifdef CONFIG_COMPAT
1094 static int compat_rawv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1095 char __user
*optval
, int __user
*optlen
)
1101 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1103 return rawv6_geticmpfilter(sk
, level
, optname
, optval
, optlen
);
1105 if (optname
== IPV6_CHECKSUM
)
1108 return compat_ipv6_getsockopt(sk
, level
, optname
,
1111 return do_rawv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1115 static int rawv6_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
1119 int amount
= sk_wmem_alloc_get(sk
);
1121 return put_user(amount
, (int __user
*)arg
);
1124 struct sk_buff
*skb
;
1127 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
1128 skb
= skb_peek(&sk
->sk_receive_queue
);
1130 amount
= skb
->tail
- skb
->transport_header
;
1131 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
1132 return put_user(amount
, (int __user
*)arg
);
1136 #ifdef CONFIG_IPV6_MROUTE
1137 return ip6mr_ioctl(sk
, cmd
, (void __user
*)arg
);
1139 return -ENOIOCTLCMD
;
1144 #ifdef CONFIG_COMPAT
1145 static int compat_rawv6_ioctl(struct sock
*sk
, unsigned int cmd
, unsigned long arg
)
1150 return -ENOIOCTLCMD
;
1152 #ifdef CONFIG_IPV6_MROUTE
1153 return ip6mr_compat_ioctl(sk
, cmd
, compat_ptr(arg
));
1155 return -ENOIOCTLCMD
;
1161 static void rawv6_close(struct sock
*sk
, long timeout
)
1163 if (inet_sk(sk
)->inet_num
== IPPROTO_RAW
)
1164 ip6_ra_control(sk
, -1);
1166 sk_common_release(sk
);
1169 static void raw6_destroy(struct sock
*sk
)
1172 ip6_flush_pending_frames(sk
);
1175 inet6_destroy_sock(sk
);
1178 static int rawv6_init_sk(struct sock
*sk
)
1180 struct raw6_sock
*rp
= raw6_sk(sk
);
1182 switch (inet_sk(sk
)->inet_num
) {
1183 case IPPROTO_ICMPV6
:
1197 struct proto rawv6_prot
= {
1199 .owner
= THIS_MODULE
,
1200 .close
= rawv6_close
,
1201 .destroy
= raw6_destroy
,
1202 .connect
= ip6_datagram_connect
,
1203 .disconnect
= udp_disconnect
,
1204 .ioctl
= rawv6_ioctl
,
1205 .init
= rawv6_init_sk
,
1206 .setsockopt
= rawv6_setsockopt
,
1207 .getsockopt
= rawv6_getsockopt
,
1208 .sendmsg
= rawv6_sendmsg
,
1209 .recvmsg
= rawv6_recvmsg
,
1211 .backlog_rcv
= rawv6_rcv_skb
,
1212 .hash
= raw_hash_sk
,
1213 .unhash
= raw_unhash_sk
,
1214 .obj_size
= sizeof(struct raw6_sock
),
1215 .h
.raw_hash
= &raw_v6_hashinfo
,
1216 #ifdef CONFIG_COMPAT
1217 .compat_setsockopt
= compat_rawv6_setsockopt
,
1218 .compat_getsockopt
= compat_rawv6_getsockopt
,
1219 .compat_ioctl
= compat_rawv6_ioctl
,
1223 #ifdef CONFIG_PROC_FS
1224 static void raw6_sock_seq_show(struct seq_file
*seq
, struct sock
*sp
, int i
)
1226 struct ipv6_pinfo
*np
= inet6_sk(sp
);
1227 const struct in6_addr
*dest
, *src
;
1231 src
= &np
->rcv_saddr
;
1233 srcp
= inet_sk(sp
)->inet_num
;
1235 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1236 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1238 src
->s6_addr32
[0], src
->s6_addr32
[1],
1239 src
->s6_addr32
[2], src
->s6_addr32
[3], srcp
,
1240 dest
->s6_addr32
[0], dest
->s6_addr32
[1],
1241 dest
->s6_addr32
[2], dest
->s6_addr32
[3], destp
,
1243 sk_wmem_alloc_get(sp
),
1244 sk_rmem_alloc_get(sp
),
1248 atomic_read(&sp
->sk_refcnt
), sp
, atomic_read(&sp
->sk_drops
));
1251 static int raw6_seq_show(struct seq_file
*seq
, void *v
)
1253 if (v
== SEQ_START_TOKEN
)
1258 "st tx_queue rx_queue tr tm->when retrnsmt"
1259 " uid timeout inode ref pointer drops\n");
1261 raw6_sock_seq_show(seq
, v
, raw_seq_private(seq
)->bucket
);
1265 static const struct seq_operations raw6_seq_ops
= {
1266 .start
= raw_seq_start
,
1267 .next
= raw_seq_next
,
1268 .stop
= raw_seq_stop
,
1269 .show
= raw6_seq_show
,
1272 static int raw6_seq_open(struct inode
*inode
, struct file
*file
)
1274 return raw_seq_open(inode
, file
, &raw_v6_hashinfo
, &raw6_seq_ops
);
1277 static const struct file_operations raw6_seq_fops
= {
1278 .owner
= THIS_MODULE
,
1279 .open
= raw6_seq_open
,
1281 .llseek
= seq_lseek
,
1282 .release
= seq_release_net
,
1285 static int __net_init
raw6_init_net(struct net
*net
)
1287 if (!proc_net_fops_create(net
, "raw6", S_IRUGO
, &raw6_seq_fops
))
1293 static void __net_exit
raw6_exit_net(struct net
*net
)
1295 proc_net_remove(net
, "raw6");
1298 static struct pernet_operations raw6_net_ops
= {
1299 .init
= raw6_init_net
,
1300 .exit
= raw6_exit_net
,
1303 int __init
raw6_proc_init(void)
1305 return register_pernet_subsys(&raw6_net_ops
);
1308 void raw6_proc_exit(void)
1310 unregister_pernet_subsys(&raw6_net_ops
);
1312 #endif /* CONFIG_PROC_FS */
1314 /* Same as inet6_dgram_ops, sans udp_poll. */
1315 static const struct proto_ops inet6_sockraw_ops
= {
1317 .owner
= THIS_MODULE
,
1318 .release
= inet6_release
,
1320 .connect
= inet_dgram_connect
, /* ok */
1321 .socketpair
= sock_no_socketpair
, /* a do nothing */
1322 .accept
= sock_no_accept
, /* a do nothing */
1323 .getname
= inet6_getname
,
1324 .poll
= datagram_poll
, /* ok */
1325 .ioctl
= inet6_ioctl
, /* must change */
1326 .listen
= sock_no_listen
, /* ok */
1327 .shutdown
= inet_shutdown
, /* ok */
1328 .setsockopt
= sock_common_setsockopt
, /* ok */
1329 .getsockopt
= sock_common_getsockopt
, /* ok */
1330 .sendmsg
= inet_sendmsg
, /* ok */
1331 .recvmsg
= sock_common_recvmsg
, /* ok */
1332 .mmap
= sock_no_mmap
,
1333 .sendpage
= sock_no_sendpage
,
1334 #ifdef CONFIG_COMPAT
1335 .compat_setsockopt
= compat_sock_common_setsockopt
,
1336 .compat_getsockopt
= compat_sock_common_getsockopt
,
1340 static struct inet_protosw rawv6_protosw
= {
1342 .protocol
= IPPROTO_IP
, /* wild card */
1343 .prot
= &rawv6_prot
,
1344 .ops
= &inet6_sockraw_ops
,
1345 .no_check
= UDP_CSUM_DEFAULT
,
1346 .flags
= INET_PROTOSW_REUSE
,
1349 int __init
rawv6_init(void)
1353 ret
= inet6_register_protosw(&rawv6_protosw
);
1360 void rawv6_exit(void)
1362 inet6_unregister_protosw(&rawv6_protosw
);