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>
64 #include <linux/export.h>
66 static struct raw_hashinfo raw_v6_hashinfo
= {
67 .lock
= __RW_LOCK_UNLOCKED(raw_v6_hashinfo
.lock
),
70 static struct sock
*__raw_v6_lookup(struct net
*net
, struct sock
*sk
,
71 unsigned short num
, const struct in6_addr
*loc_addr
,
72 const struct in6_addr
*rmt_addr
, int dif
)
74 struct hlist_node
*node
;
75 bool is_multicast
= ipv6_addr_is_multicast(loc_addr
);
77 sk_for_each_from(sk
, node
)
78 if (inet_sk(sk
)->inet_num
== num
) {
79 struct ipv6_pinfo
*np
= inet6_sk(sk
);
81 if (!net_eq(sock_net(sk
), net
))
84 if (!ipv6_addr_any(&np
->daddr
) &&
85 !ipv6_addr_equal(&np
->daddr
, rmt_addr
))
88 if (sk
->sk_bound_dev_if
&& sk
->sk_bound_dev_if
!= dif
)
91 if (!ipv6_addr_any(&np
->rcv_saddr
)) {
92 if (ipv6_addr_equal(&np
->rcv_saddr
, loc_addr
))
95 inet6_mc_check(sk
, loc_addr
, rmt_addr
))
110 static __inline__
int icmpv6_filter(struct sock
*sk
, struct sk_buff
*skb
)
112 struct icmp6hdr
*icmph
;
113 struct raw6_sock
*rp
= raw6_sk(sk
);
115 if (pskb_may_pull(skb
, sizeof(struct icmp6hdr
))) {
116 __u32
*data
= &rp
->filter
.data
[0];
119 icmph
= (struct icmp6hdr
*) skb
->data
;
120 bit_nr
= icmph
->icmp6_type
;
122 return (data
[bit_nr
>> 5] & (1 << (bit_nr
& 31))) != 0;
127 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
128 typedef int mh_filter_t(struct sock
*sock
, struct sk_buff
*skb
);
130 static mh_filter_t __rcu
*mh_filter __read_mostly
;
132 int rawv6_mh_filter_register(mh_filter_t filter
)
134 rcu_assign_pointer(mh_filter
, filter
);
137 EXPORT_SYMBOL(rawv6_mh_filter_register
);
139 int rawv6_mh_filter_unregister(mh_filter_t filter
)
141 RCU_INIT_POINTER(mh_filter
, NULL
);
145 EXPORT_SYMBOL(rawv6_mh_filter_unregister
);
150 * demultiplex raw sockets.
151 * (should consider queueing the skb in the sock receive_queue
152 * without calling rawv6.c)
154 * Caller owns SKB so we must make clones.
156 static bool ipv6_raw_deliver(struct sk_buff
*skb
, int nexthdr
)
158 const struct in6_addr
*saddr
;
159 const struct in6_addr
*daddr
;
161 bool delivered
= false;
165 saddr
= &ipv6_hdr(skb
)->saddr
;
168 hash
= nexthdr
& (RAW_HTABLE_SIZE
- 1);
170 read_lock(&raw_v6_hashinfo
.lock
);
171 sk
= sk_head(&raw_v6_hashinfo
.ht
[hash
]);
176 net
= dev_net(skb
->dev
);
177 sk
= __raw_v6_lookup(net
, sk
, nexthdr
, daddr
, saddr
, IP6CB(skb
)->iif
);
185 filtered
= icmpv6_filter(sk
, skb
);
188 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
191 /* XXX: To validate MH only once for each packet,
192 * this is placed here. It should be after checking
193 * xfrm policy, however it doesn't. The checking xfrm
194 * policy is placed in rawv6_rcv() because it is
195 * required for each socket.
199 filter
= rcu_dereference(mh_filter
);
200 filtered
= filter
? (*filter
)(sk
, skb
) : 0;
212 struct sk_buff
*clone
= skb_clone(skb
, GFP_ATOMIC
);
214 /* Not releasing hash table! */
217 rawv6_rcv(sk
, clone
);
220 sk
= __raw_v6_lookup(net
, sk_next(sk
), nexthdr
, daddr
, saddr
,
224 read_unlock(&raw_v6_hashinfo
.lock
);
228 bool raw6_local_deliver(struct sk_buff
*skb
, int nexthdr
)
232 raw_sk
= sk_head(&raw_v6_hashinfo
.ht
[nexthdr
& (RAW_HTABLE_SIZE
- 1)]);
233 if (raw_sk
&& !ipv6_raw_deliver(skb
, nexthdr
))
236 return raw_sk
!= NULL
;
239 /* This cleans up af_inet6 a bit. -DaveM */
240 static int rawv6_bind(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
242 struct inet_sock
*inet
= inet_sk(sk
);
243 struct ipv6_pinfo
*np
= inet6_sk(sk
);
244 struct sockaddr_in6
*addr
= (struct sockaddr_in6
*) uaddr
;
249 if (addr_len
< SIN6_LEN_RFC2133
)
251 addr_type
= ipv6_addr_type(&addr
->sin6_addr
);
253 /* Raw sockets are IPv6 only */
254 if (addr_type
== IPV6_ADDR_MAPPED
)
255 return -EADDRNOTAVAIL
;
260 if (sk
->sk_state
!= TCP_CLOSE
)
264 /* Check if the address belongs to the host. */
265 if (addr_type
!= IPV6_ADDR_ANY
) {
266 struct net_device
*dev
= NULL
;
268 if (addr_type
& IPV6_ADDR_LINKLOCAL
) {
269 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
270 addr
->sin6_scope_id
) {
271 /* Override any existing binding, if another
272 * one is supplied by user.
274 sk
->sk_bound_dev_if
= addr
->sin6_scope_id
;
277 /* Binding to link-local address requires an interface */
278 if (!sk
->sk_bound_dev_if
)
282 dev
= dev_get_by_index_rcu(sock_net(sk
),
283 sk
->sk_bound_dev_if
);
288 /* ipv4 addr of the socket is invalid. Only the
289 * unspecified and mapped address have a v4 equivalent.
291 v4addr
= LOOPBACK4_IPV6
;
292 if (!(addr_type
& IPV6_ADDR_MULTICAST
)) {
293 err
= -EADDRNOTAVAIL
;
294 if (!ipv6_chk_addr(sock_net(sk
), &addr
->sin6_addr
,
301 inet
->inet_rcv_saddr
= inet
->inet_saddr
= v4addr
;
302 np
->rcv_saddr
= addr
->sin6_addr
;
303 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
304 np
->saddr
= addr
->sin6_addr
;
313 static void rawv6_err(struct sock
*sk
, struct sk_buff
*skb
,
314 struct inet6_skb_parm
*opt
,
315 u8 type
, u8 code
, int offset
, __be32 info
)
317 struct inet_sock
*inet
= inet_sk(sk
);
318 struct ipv6_pinfo
*np
= inet6_sk(sk
);
322 /* Report error on raw socket, if:
323 1. User requested recverr.
324 2. Socket is connected (otherwise the error indication
325 is useless without recverr and error is hard.
327 if (!np
->recverr
&& sk
->sk_state
!= TCP_ESTABLISHED
)
330 harderr
= icmpv6_err_convert(type
, code
, &err
);
331 if (type
== ICMPV6_PKT_TOOBIG
) {
332 ip6_sk_update_pmtu(skb
, sk
, info
);
333 harderr
= (np
->pmtudisc
== IPV6_PMTUDISC_DO
);
335 if (type
== NDISC_REDIRECT
)
336 ip6_sk_redirect(skb
, sk
);
338 u8
*payload
= skb
->data
;
341 ipv6_icmp_error(sk
, skb
, err
, 0, ntohl(info
), payload
);
344 if (np
->recverr
|| harderr
) {
346 sk
->sk_error_report(sk
);
350 void raw6_icmp_error(struct sk_buff
*skb
, int nexthdr
,
351 u8 type
, u8 code
, int inner_offset
, __be32 info
)
355 const struct in6_addr
*saddr
, *daddr
;
358 hash
= nexthdr
& (RAW_HTABLE_SIZE
- 1);
360 read_lock(&raw_v6_hashinfo
.lock
);
361 sk
= sk_head(&raw_v6_hashinfo
.ht
[hash
]);
363 /* Note: ipv6_hdr(skb) != skb->data */
364 const struct ipv6hdr
*ip6h
= (const struct ipv6hdr
*)skb
->data
;
365 saddr
= &ip6h
->saddr
;
366 daddr
= &ip6h
->daddr
;
367 net
= dev_net(skb
->dev
);
369 while ((sk
= __raw_v6_lookup(net
, sk
, nexthdr
, saddr
, daddr
,
371 rawv6_err(sk
, skb
, NULL
, type
, code
,
376 read_unlock(&raw_v6_hashinfo
.lock
);
379 static inline int rawv6_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
)
381 if ((raw6_sk(sk
)->checksum
|| rcu_access_pointer(sk
->sk_filter
)) &&
382 skb_checksum_complete(skb
)) {
383 atomic_inc(&sk
->sk_drops
);
388 /* Charge it to the socket. */
390 if (sock_queue_rcv_skb(sk
, skb
) < 0) {
399 * This is next to useless...
400 * if we demultiplex in network layer we don't need the extra call
401 * just to queue the skb...
402 * maybe we could have the network decide upon a hint if it
403 * should call raw_rcv for demultiplexing
405 int rawv6_rcv(struct sock
*sk
, struct sk_buff
*skb
)
407 struct inet_sock
*inet
= inet_sk(sk
);
408 struct raw6_sock
*rp
= raw6_sk(sk
);
410 if (!xfrm6_policy_check(sk
, XFRM_POLICY_IN
, skb
)) {
411 atomic_inc(&sk
->sk_drops
);
417 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
419 if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
420 skb_postpull_rcsum(skb
, skb_network_header(skb
),
421 skb_network_header_len(skb
));
422 if (!csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
423 &ipv6_hdr(skb
)->daddr
,
424 skb
->len
, inet
->inet_num
, skb
->csum
))
425 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
427 if (!skb_csum_unnecessary(skb
))
428 skb
->csum
= ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
429 &ipv6_hdr(skb
)->daddr
,
434 if (skb_checksum_complete(skb
)) {
435 atomic_inc(&sk
->sk_drops
);
441 rawv6_rcv_skb(sk
, skb
);
447 * This should be easy, if there is something there
448 * we return it, otherwise we block.
451 static int rawv6_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
452 struct msghdr
*msg
, size_t len
,
453 int noblock
, int flags
, int *addr_len
)
455 struct ipv6_pinfo
*np
= inet6_sk(sk
);
456 struct sockaddr_in6
*sin6
= (struct sockaddr_in6
*)msg
->msg_name
;
465 *addr_len
=sizeof(*sin6
);
467 if (flags
& MSG_ERRQUEUE
)
468 return ipv6_recv_error(sk
, msg
, len
);
470 if (np
->rxpmtu
&& np
->rxopt
.bits
.rxpmtu
)
471 return ipv6_recv_rxpmtu(sk
, msg
, len
);
473 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
480 msg
->msg_flags
|= MSG_TRUNC
;
483 if (skb_csum_unnecessary(skb
)) {
484 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
485 } else if (msg
->msg_flags
&MSG_TRUNC
) {
486 if (__skb_checksum_complete(skb
))
488 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
490 err
= skb_copy_and_csum_datagram_iovec(skb
, 0, msg
->msg_iov
);
497 /* Copy the address. */
499 sin6
->sin6_family
= AF_INET6
;
501 sin6
->sin6_addr
= ipv6_hdr(skb
)->saddr
;
502 sin6
->sin6_flowinfo
= 0;
503 sin6
->sin6_scope_id
= 0;
504 if (ipv6_addr_type(&sin6
->sin6_addr
) & IPV6_ADDR_LINKLOCAL
)
505 sin6
->sin6_scope_id
= IP6CB(skb
)->iif
;
508 sock_recv_ts_and_drops(msg
, sk
, skb
);
511 datagram_recv_ctl(sk
, msg
, skb
);
514 if (flags
& MSG_TRUNC
)
518 skb_free_datagram(sk
, skb
);
523 skb_kill_datagram(sk
, skb
, flags
);
525 /* Error for blocking case is chosen to masquerade
526 as some normal condition.
528 err
= (flags
&MSG_DONTWAIT
) ? -EAGAIN
: -EHOSTUNREACH
;
532 static int rawv6_push_pending_frames(struct sock
*sk
, struct flowi6
*fl6
,
533 struct raw6_sock
*rp
)
546 if ((skb
= skb_peek(&sk
->sk_write_queue
)) == NULL
)
550 total_len
= inet_sk(sk
)->cork
.base
.length
;
551 if (offset
>= total_len
- 1) {
553 ip6_flush_pending_frames(sk
);
557 /* should be check HW csum miyazawa */
558 if (skb_queue_len(&sk
->sk_write_queue
) == 1) {
560 * Only one fragment on the socket.
562 tmp_csum
= skb
->csum
;
564 struct sk_buff
*csum_skb
= NULL
;
567 skb_queue_walk(&sk
->sk_write_queue
, skb
) {
568 tmp_csum
= csum_add(tmp_csum
, skb
->csum
);
573 len
= skb
->len
- skb_transport_offset(skb
);
585 offset
+= skb_transport_offset(skb
);
586 if (skb_copy_bits(skb
, offset
, &csum
, 2))
589 /* in case cksum was not initialized */
591 tmp_csum
= csum_sub(tmp_csum
, csum_unfold(csum
));
593 csum
= csum_ipv6_magic(&fl6
->saddr
, &fl6
->daddr
,
594 total_len
, fl6
->flowi6_proto
, tmp_csum
);
596 if (csum
== 0 && fl6
->flowi6_proto
== IPPROTO_UDP
)
597 csum
= CSUM_MANGLED_0
;
599 if (skb_store_bits(skb
, offset
, &csum
, 2))
603 err
= ip6_push_pending_frames(sk
);
608 static int rawv6_send_hdrinc(struct sock
*sk
, void *from
, int length
,
609 struct flowi6
*fl6
, struct dst_entry
**dstp
,
612 struct ipv6_pinfo
*np
= inet6_sk(sk
);
616 struct rt6_info
*rt
= (struct rt6_info
*)*dstp
;
617 int hlen
= LL_RESERVED_SPACE(rt
->dst
.dev
);
618 int tlen
= rt
->dst
.dev
->needed_tailroom
;
620 if (length
> rt
->dst
.dev
->mtu
) {
621 ipv6_local_error(sk
, EMSGSIZE
, fl6
, rt
->dst
.dev
->mtu
);
627 skb
= sock_alloc_send_skb(sk
,
628 length
+ hlen
+ tlen
+ 15,
629 flags
& MSG_DONTWAIT
, &err
);
632 skb_reserve(skb
, hlen
);
634 skb
->priority
= sk
->sk_priority
;
635 skb
->mark
= sk
->sk_mark
;
636 skb_dst_set(skb
, &rt
->dst
);
639 skb_put(skb
, length
);
640 skb_reset_network_header(skb
);
643 skb
->ip_summed
= CHECKSUM_NONE
;
645 skb
->transport_header
= skb
->network_header
;
646 err
= memcpy_fromiovecend((void *)iph
, from
, 0, length
);
650 IP6_UPD_PO_STATS(sock_net(sk
), rt
->rt6i_idev
, IPSTATS_MIB_OUT
, skb
->len
);
651 err
= NF_HOOK(NFPROTO_IPV6
, NF_INET_LOCAL_OUT
, skb
, NULL
,
652 rt
->dst
.dev
, dst_output
);
654 err
= net_xmit_errno(err
);
664 IP6_INC_STATS(sock_net(sk
), rt
->rt6i_idev
, IPSTATS_MIB_OUTDISCARDS
);
665 if (err
== -ENOBUFS
&& !np
->recverr
)
670 static int rawv6_probe_proto_opt(struct flowi6
*fl6
, struct msghdr
*msg
)
673 u8 __user
*type
= NULL
;
674 u8 __user
*code
= NULL
;
682 for (i
= 0; i
< msg
->msg_iovlen
; i
++) {
683 iov
= &msg
->msg_iov
[i
];
687 switch (fl6
->flowi6_proto
) {
689 /* check if one-byte field is readable or not. */
690 if (iov
->iov_base
&& iov
->iov_len
< 1)
694 type
= iov
->iov_base
;
695 /* check if code field is readable or not. */
696 if (iov
->iov_len
> 1)
699 code
= iov
->iov_base
;
702 if (get_user(fl6
->fl6_icmp_type
, type
) ||
703 get_user(fl6
->fl6_icmp_code
, code
))
709 if (iov
->iov_base
&& iov
->iov_len
< 1)
711 /* check if type field is readable or not. */
712 if (iov
->iov_len
> 2 - len
) {
713 u8 __user
*p
= iov
->iov_base
;
714 if (get_user(fl6
->fl6_mh_type
, &p
[2 - len
]))
731 static int rawv6_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
732 struct msghdr
*msg
, size_t len
)
734 struct ipv6_txoptions opt_space
;
735 struct sockaddr_in6
* sin6
= (struct sockaddr_in6
*) msg
->msg_name
;
736 struct in6_addr
*daddr
, *final_p
, final
;
737 struct inet_sock
*inet
= inet_sk(sk
);
738 struct ipv6_pinfo
*np
= inet6_sk(sk
);
739 struct raw6_sock
*rp
= raw6_sk(sk
);
740 struct ipv6_txoptions
*opt
= NULL
;
741 struct ip6_flowlabel
*flowlabel
= NULL
;
742 struct dst_entry
*dst
= NULL
;
744 int addr_len
= msg
->msg_namelen
;
751 /* Rough check on arithmetic overflow,
752 better check is made in ip6_append_data().
757 /* Mirror BSD error message compatibility */
758 if (msg
->msg_flags
& MSG_OOB
)
762 * Get and verify the address.
764 memset(&fl6
, 0, sizeof(fl6
));
766 fl6
.flowi6_mark
= sk
->sk_mark
;
769 if (addr_len
< SIN6_LEN_RFC2133
)
772 if (sin6
->sin6_family
&& sin6
->sin6_family
!= AF_INET6
)
773 return -EAFNOSUPPORT
;
775 /* port is the proto value [0..255] carried in nexthdr */
776 proto
= ntohs(sin6
->sin6_port
);
779 proto
= inet
->inet_num
;
780 else if (proto
!= inet
->inet_num
)
786 daddr
= &sin6
->sin6_addr
;
788 fl6
.flowlabel
= sin6
->sin6_flowinfo
&IPV6_FLOWINFO_MASK
;
789 if (fl6
.flowlabel
&IPV6_FLOWLABEL_MASK
) {
790 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
791 if (flowlabel
== NULL
)
793 daddr
= &flowlabel
->dst
;
798 * Otherwise it will be difficult to maintain
801 if (sk
->sk_state
== TCP_ESTABLISHED
&&
802 ipv6_addr_equal(daddr
, &np
->daddr
))
805 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
806 sin6
->sin6_scope_id
&&
807 ipv6_addr_type(daddr
)&IPV6_ADDR_LINKLOCAL
)
808 fl6
.flowi6_oif
= sin6
->sin6_scope_id
;
810 if (sk
->sk_state
!= TCP_ESTABLISHED
)
811 return -EDESTADDRREQ
;
813 proto
= inet
->inet_num
;
815 fl6
.flowlabel
= np
->flow_label
;
818 if (fl6
.flowi6_oif
== 0)
819 fl6
.flowi6_oif
= sk
->sk_bound_dev_if
;
821 if (msg
->msg_controllen
) {
823 memset(opt
, 0, sizeof(struct ipv6_txoptions
));
824 opt
->tot_len
= sizeof(struct ipv6_txoptions
);
826 err
= datagram_send_ctl(sock_net(sk
), sk
, msg
, &fl6
, opt
,
827 &hlimit
, &tclass
, &dontfrag
);
829 fl6_sock_release(flowlabel
);
832 if ((fl6
.flowlabel
&IPV6_FLOWLABEL_MASK
) && !flowlabel
) {
833 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
834 if (flowlabel
== NULL
)
837 if (!(opt
->opt_nflen
|opt
->opt_flen
))
843 opt
= fl6_merge_options(&opt_space
, flowlabel
, opt
);
844 opt
= ipv6_fixup_options(&opt_space
, opt
);
846 fl6
.flowi6_proto
= proto
;
847 err
= rawv6_probe_proto_opt(&fl6
, msg
);
851 if (!ipv6_addr_any(daddr
))
854 fl6
.daddr
.s6_addr
[15] = 0x1; /* :: means loopback (BSD'ism) */
855 if (ipv6_addr_any(&fl6
.saddr
) && !ipv6_addr_any(&np
->saddr
))
856 fl6
.saddr
= np
->saddr
;
858 final_p
= fl6_update_dst(&fl6
, opt
, &final
);
860 if (!fl6
.flowi6_oif
&& ipv6_addr_is_multicast(&fl6
.daddr
))
861 fl6
.flowi6_oif
= np
->mcast_oif
;
862 else if (!fl6
.flowi6_oif
)
863 fl6
.flowi6_oif
= np
->ucast_oif
;
864 security_sk_classify_flow(sk
, flowi6_to_flowi(&fl6
));
866 dst
= ip6_dst_lookup_flow(sk
, &fl6
, final_p
, true);
872 if (ipv6_addr_is_multicast(&fl6
.daddr
))
873 hlimit
= np
->mcast_hops
;
875 hlimit
= np
->hop_limit
;
877 hlimit
= ip6_dst_hoplimit(dst
);
884 dontfrag
= np
->dontfrag
;
886 if (msg
->msg_flags
&MSG_CONFIRM
)
891 err
= rawv6_send_hdrinc(sk
, msg
->msg_iov
, len
, &fl6
, &dst
, msg
->msg_flags
);
894 err
= ip6_append_data(sk
, ip_generic_getfrag
, msg
->msg_iov
,
895 len
, 0, hlimit
, tclass
, opt
, &fl6
, (struct rt6_info
*)dst
,
896 msg
->msg_flags
, dontfrag
);
899 ip6_flush_pending_frames(sk
);
900 else if (!(msg
->msg_flags
& MSG_MORE
))
901 err
= rawv6_push_pending_frames(sk
, &fl6
, rp
);
907 fl6_sock_release(flowlabel
);
908 return err
<0?err
:len
;
911 if (!(msg
->msg_flags
& MSG_PROBE
) || len
)
912 goto back_from_confirm
;
917 static int rawv6_seticmpfilter(struct sock
*sk
, int level
, int optname
,
918 char __user
*optval
, int optlen
)
922 if (optlen
> sizeof(struct icmp6_filter
))
923 optlen
= sizeof(struct icmp6_filter
);
924 if (copy_from_user(&raw6_sk(sk
)->filter
, optval
, optlen
))
934 static int rawv6_geticmpfilter(struct sock
*sk
, int level
, int optname
,
935 char __user
*optval
, int __user
*optlen
)
941 if (get_user(len
, optlen
))
945 if (len
> sizeof(struct icmp6_filter
))
946 len
= sizeof(struct icmp6_filter
);
947 if (put_user(len
, optlen
))
949 if (copy_to_user(optval
, &raw6_sk(sk
)->filter
, len
))
960 static int do_rawv6_setsockopt(struct sock
*sk
, int level
, int optname
,
961 char __user
*optval
, unsigned int optlen
)
963 struct raw6_sock
*rp
= raw6_sk(sk
);
966 if (get_user(val
, (int __user
*)optval
))
971 if (inet_sk(sk
)->inet_num
== IPPROTO_ICMPV6
&&
972 level
== IPPROTO_IPV6
) {
974 * RFC3542 tells that IPV6_CHECKSUM socket
975 * option in the IPPROTO_IPV6 level is not
976 * allowed on ICMPv6 sockets.
977 * If you want to set it, use IPPROTO_RAW
978 * level IPV6_CHECKSUM socket option
984 /* You may get strange result with a positive odd offset;
985 RFC2292bis agrees with me. */
986 if (val
> 0 && (val
&1))
1002 static int rawv6_setsockopt(struct sock
*sk
, int level
, int optname
,
1003 char __user
*optval
, unsigned int optlen
)
1010 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1012 return rawv6_seticmpfilter(sk
, level
, optname
, optval
, optlen
);
1014 if (optname
== IPV6_CHECKSUM
)
1017 return ipv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1020 return do_rawv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1023 #ifdef CONFIG_COMPAT
1024 static int compat_rawv6_setsockopt(struct sock
*sk
, int level
, int optname
,
1025 char __user
*optval
, unsigned int optlen
)
1031 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1033 return rawv6_seticmpfilter(sk
, level
, optname
, optval
, optlen
);
1035 if (optname
== IPV6_CHECKSUM
)
1038 return compat_ipv6_setsockopt(sk
, level
, optname
,
1041 return do_rawv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1045 static int do_rawv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1046 char __user
*optval
, int __user
*optlen
)
1048 struct raw6_sock
*rp
= raw6_sk(sk
);
1051 if (get_user(len
,optlen
))
1057 * We allow getsockopt() for IPPROTO_IPV6-level
1058 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1059 * since RFC3542 is silent about it.
1061 if (rp
->checksum
== 0)
1068 return -ENOPROTOOPT
;
1071 len
= min_t(unsigned int, sizeof(int), len
);
1073 if (put_user(len
, optlen
))
1075 if (copy_to_user(optval
,&val
,len
))
1080 static int rawv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1081 char __user
*optval
, int __user
*optlen
)
1088 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1090 return rawv6_geticmpfilter(sk
, level
, optname
, optval
, optlen
);
1092 if (optname
== IPV6_CHECKSUM
)
1095 return ipv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1098 return do_rawv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1101 #ifdef CONFIG_COMPAT
1102 static int compat_rawv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1103 char __user
*optval
, int __user
*optlen
)
1109 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1111 return rawv6_geticmpfilter(sk
, level
, optname
, optval
, optlen
);
1113 if (optname
== IPV6_CHECKSUM
)
1116 return compat_ipv6_getsockopt(sk
, level
, optname
,
1119 return do_rawv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1123 static int rawv6_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
1127 int amount
= sk_wmem_alloc_get(sk
);
1129 return put_user(amount
, (int __user
*)arg
);
1132 struct sk_buff
*skb
;
1135 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
1136 skb
= skb_peek(&sk
->sk_receive_queue
);
1138 amount
= skb
->tail
- skb
->transport_header
;
1139 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
1140 return put_user(amount
, (int __user
*)arg
);
1144 #ifdef CONFIG_IPV6_MROUTE
1145 return ip6mr_ioctl(sk
, cmd
, (void __user
*)arg
);
1147 return -ENOIOCTLCMD
;
1152 #ifdef CONFIG_COMPAT
1153 static int compat_rawv6_ioctl(struct sock
*sk
, unsigned int cmd
, unsigned long arg
)
1158 return -ENOIOCTLCMD
;
1160 #ifdef CONFIG_IPV6_MROUTE
1161 return ip6mr_compat_ioctl(sk
, cmd
, compat_ptr(arg
));
1163 return -ENOIOCTLCMD
;
1169 static void rawv6_close(struct sock
*sk
, long timeout
)
1171 if (inet_sk(sk
)->inet_num
== IPPROTO_RAW
)
1172 ip6_ra_control(sk
, -1);
1174 sk_common_release(sk
);
1177 static void raw6_destroy(struct sock
*sk
)
1180 ip6_flush_pending_frames(sk
);
1183 inet6_destroy_sock(sk
);
1186 static int rawv6_init_sk(struct sock
*sk
)
1188 struct raw6_sock
*rp
= raw6_sk(sk
);
1190 switch (inet_sk(sk
)->inet_num
) {
1191 case IPPROTO_ICMPV6
:
1205 struct proto rawv6_prot
= {
1207 .owner
= THIS_MODULE
,
1208 .close
= rawv6_close
,
1209 .destroy
= raw6_destroy
,
1210 .connect
= ip6_datagram_connect
,
1211 .disconnect
= udp_disconnect
,
1212 .ioctl
= rawv6_ioctl
,
1213 .init
= rawv6_init_sk
,
1214 .setsockopt
= rawv6_setsockopt
,
1215 .getsockopt
= rawv6_getsockopt
,
1216 .sendmsg
= rawv6_sendmsg
,
1217 .recvmsg
= rawv6_recvmsg
,
1219 .backlog_rcv
= rawv6_rcv_skb
,
1220 .hash
= raw_hash_sk
,
1221 .unhash
= raw_unhash_sk
,
1222 .obj_size
= sizeof(struct raw6_sock
),
1223 .h
.raw_hash
= &raw_v6_hashinfo
,
1224 #ifdef CONFIG_COMPAT
1225 .compat_setsockopt
= compat_rawv6_setsockopt
,
1226 .compat_getsockopt
= compat_rawv6_getsockopt
,
1227 .compat_ioctl
= compat_rawv6_ioctl
,
1231 #ifdef CONFIG_PROC_FS
1232 static void raw6_sock_seq_show(struct seq_file
*seq
, struct sock
*sp
, int i
)
1234 struct ipv6_pinfo
*np
= inet6_sk(sp
);
1235 const struct in6_addr
*dest
, *src
;
1239 src
= &np
->rcv_saddr
;
1241 srcp
= inet_sk(sp
)->inet_num
;
1243 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1244 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1246 src
->s6_addr32
[0], src
->s6_addr32
[1],
1247 src
->s6_addr32
[2], src
->s6_addr32
[3], srcp
,
1248 dest
->s6_addr32
[0], dest
->s6_addr32
[1],
1249 dest
->s6_addr32
[2], dest
->s6_addr32
[3], destp
,
1251 sk_wmem_alloc_get(sp
),
1252 sk_rmem_alloc_get(sp
),
1256 atomic_read(&sp
->sk_refcnt
), sp
, atomic_read(&sp
->sk_drops
));
1259 static int raw6_seq_show(struct seq_file
*seq
, void *v
)
1261 if (v
== SEQ_START_TOKEN
)
1266 "st tx_queue rx_queue tr tm->when retrnsmt"
1267 " uid timeout inode ref pointer drops\n");
1269 raw6_sock_seq_show(seq
, v
, raw_seq_private(seq
)->bucket
);
1273 static const struct seq_operations raw6_seq_ops
= {
1274 .start
= raw_seq_start
,
1275 .next
= raw_seq_next
,
1276 .stop
= raw_seq_stop
,
1277 .show
= raw6_seq_show
,
1280 static int raw6_seq_open(struct inode
*inode
, struct file
*file
)
1282 return raw_seq_open(inode
, file
, &raw_v6_hashinfo
, &raw6_seq_ops
);
1285 static const struct file_operations raw6_seq_fops
= {
1286 .owner
= THIS_MODULE
,
1287 .open
= raw6_seq_open
,
1289 .llseek
= seq_lseek
,
1290 .release
= seq_release_net
,
1293 static int __net_init
raw6_init_net(struct net
*net
)
1295 if (!proc_net_fops_create(net
, "raw6", S_IRUGO
, &raw6_seq_fops
))
1301 static void __net_exit
raw6_exit_net(struct net
*net
)
1303 proc_net_remove(net
, "raw6");
1306 static struct pernet_operations raw6_net_ops
= {
1307 .init
= raw6_init_net
,
1308 .exit
= raw6_exit_net
,
1311 int __init
raw6_proc_init(void)
1313 return register_pernet_subsys(&raw6_net_ops
);
1316 void raw6_proc_exit(void)
1318 unregister_pernet_subsys(&raw6_net_ops
);
1320 #endif /* CONFIG_PROC_FS */
1322 /* Same as inet6_dgram_ops, sans udp_poll. */
1323 static const struct proto_ops inet6_sockraw_ops
= {
1325 .owner
= THIS_MODULE
,
1326 .release
= inet6_release
,
1328 .connect
= inet_dgram_connect
, /* ok */
1329 .socketpair
= sock_no_socketpair
, /* a do nothing */
1330 .accept
= sock_no_accept
, /* a do nothing */
1331 .getname
= inet6_getname
,
1332 .poll
= datagram_poll
, /* ok */
1333 .ioctl
= inet6_ioctl
, /* must change */
1334 .listen
= sock_no_listen
, /* ok */
1335 .shutdown
= inet_shutdown
, /* ok */
1336 .setsockopt
= sock_common_setsockopt
, /* ok */
1337 .getsockopt
= sock_common_getsockopt
, /* ok */
1338 .sendmsg
= inet_sendmsg
, /* ok */
1339 .recvmsg
= sock_common_recvmsg
, /* ok */
1340 .mmap
= sock_no_mmap
,
1341 .sendpage
= sock_no_sendpage
,
1342 #ifdef CONFIG_COMPAT
1343 .compat_setsockopt
= compat_sock_common_setsockopt
,
1344 .compat_getsockopt
= compat_sock_common_getsockopt
,
1348 static struct inet_protosw rawv6_protosw
= {
1350 .protocol
= IPPROTO_IP
, /* wild card */
1351 .prot
= &rawv6_prot
,
1352 .ops
= &inet6_sockraw_ops
,
1353 .no_check
= UDP_CSUM_DEFAULT
,
1354 .flags
= INET_PROTOSW_REUSE
,
1357 int __init
rawv6_init(void)
1361 ret
= inet6_register_protosw(&rawv6_protosw
);
1368 void rawv6_exit(void)
1370 inet6_unregister_protosw(&rawv6_protosw
);