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
)
254 if (addr
->sin6_family
!= AF_INET6
)
257 addr_type
= ipv6_addr_type(&addr
->sin6_addr
);
259 /* Raw sockets are IPv6 only */
260 if (addr_type
== IPV6_ADDR_MAPPED
)
261 return -EADDRNOTAVAIL
;
266 if (sk
->sk_state
!= TCP_CLOSE
)
270 /* Check if the address belongs to the host. */
271 if (addr_type
!= IPV6_ADDR_ANY
) {
272 struct net_device
*dev
= NULL
;
274 if (__ipv6_addr_needs_scope_id(addr_type
)) {
275 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
276 addr
->sin6_scope_id
) {
277 /* Override any existing binding, if another
278 * one is supplied by user.
280 sk
->sk_bound_dev_if
= addr
->sin6_scope_id
;
283 /* Binding to link-local address requires an interface */
284 if (!sk
->sk_bound_dev_if
)
288 dev
= dev_get_by_index_rcu(sock_net(sk
),
289 sk
->sk_bound_dev_if
);
294 /* ipv4 addr of the socket is invalid. Only the
295 * unspecified and mapped address have a v4 equivalent.
297 v4addr
= LOOPBACK4_IPV6
;
298 if (!(addr_type
& IPV6_ADDR_MULTICAST
)) {
299 err
= -EADDRNOTAVAIL
;
300 if (!ipv6_chk_addr(sock_net(sk
), &addr
->sin6_addr
,
307 inet
->inet_rcv_saddr
= inet
->inet_saddr
= v4addr
;
308 sk
->sk_v6_rcv_saddr
= addr
->sin6_addr
;
309 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
310 np
->saddr
= addr
->sin6_addr
;
319 static void rawv6_err(struct sock
*sk
, struct sk_buff
*skb
,
320 struct inet6_skb_parm
*opt
,
321 u8 type
, u8 code
, int offset
, __be32 info
)
323 struct inet_sock
*inet
= inet_sk(sk
);
324 struct ipv6_pinfo
*np
= inet6_sk(sk
);
328 /* Report error on raw socket, if:
329 1. User requested recverr.
330 2. Socket is connected (otherwise the error indication
331 is useless without recverr and error is hard.
333 if (!np
->recverr
&& sk
->sk_state
!= TCP_ESTABLISHED
)
336 harderr
= icmpv6_err_convert(type
, code
, &err
);
337 if (type
== ICMPV6_PKT_TOOBIG
) {
338 ip6_sk_update_pmtu(skb
, sk
, info
);
339 harderr
= (np
->pmtudisc
== IPV6_PMTUDISC_DO
);
341 if (type
== NDISC_REDIRECT
) {
342 ip6_sk_redirect(skb
, sk
);
346 u8
*payload
= skb
->data
;
349 ipv6_icmp_error(sk
, skb
, err
, 0, ntohl(info
), payload
);
352 if (np
->recverr
|| harderr
) {
354 sk
->sk_error_report(sk
);
358 void raw6_icmp_error(struct sk_buff
*skb
, int nexthdr
,
359 u8 type
, u8 code
, int inner_offset
, __be32 info
)
363 const struct in6_addr
*saddr
, *daddr
;
366 hash
= nexthdr
& (RAW_HTABLE_SIZE
- 1);
368 read_lock(&raw_v6_hashinfo
.lock
);
369 sk
= sk_head(&raw_v6_hashinfo
.ht
[hash
]);
371 /* Note: ipv6_hdr(skb) != skb->data */
372 const struct ipv6hdr
*ip6h
= (const struct ipv6hdr
*)skb
->data
;
373 saddr
= &ip6h
->saddr
;
374 daddr
= &ip6h
->daddr
;
375 net
= dev_net(skb
->dev
);
377 while ((sk
= __raw_v6_lookup(net
, sk
, nexthdr
, saddr
, daddr
,
379 rawv6_err(sk
, skb
, NULL
, type
, code
,
384 read_unlock(&raw_v6_hashinfo
.lock
);
387 static inline int rawv6_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
)
389 if ((raw6_sk(sk
)->checksum
|| rcu_access_pointer(sk
->sk_filter
)) &&
390 skb_checksum_complete(skb
)) {
391 atomic_inc(&sk
->sk_drops
);
396 /* Charge it to the socket. */
398 if (sock_queue_rcv_skb(sk
, skb
) < 0) {
407 * This is next to useless...
408 * if we demultiplex in network layer we don't need the extra call
409 * just to queue the skb...
410 * maybe we could have the network decide upon a hint if it
411 * should call raw_rcv for demultiplexing
413 int rawv6_rcv(struct sock
*sk
, struct sk_buff
*skb
)
415 struct inet_sock
*inet
= inet_sk(sk
);
416 struct raw6_sock
*rp
= raw6_sk(sk
);
418 if (!xfrm6_policy_check(sk
, XFRM_POLICY_IN
, skb
)) {
419 atomic_inc(&sk
->sk_drops
);
425 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
427 if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
428 skb_postpull_rcsum(skb
, skb_network_header(skb
),
429 skb_network_header_len(skb
));
430 if (!csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
431 &ipv6_hdr(skb
)->daddr
,
432 skb
->len
, inet
->inet_num
, skb
->csum
))
433 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
435 if (!skb_csum_unnecessary(skb
))
436 skb
->csum
= ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
437 &ipv6_hdr(skb
)->daddr
,
442 if (skb_checksum_complete(skb
)) {
443 atomic_inc(&sk
->sk_drops
);
449 rawv6_rcv_skb(sk
, skb
);
455 * This should be easy, if there is something there
456 * we return it, otherwise we block.
459 static int rawv6_recvmsg(struct kiocb
*iocb
, struct sock
*sk
,
460 struct msghdr
*msg
, size_t len
,
461 int noblock
, int flags
, int *addr_len
)
463 struct ipv6_pinfo
*np
= inet6_sk(sk
);
464 DECLARE_SOCKADDR(struct sockaddr_in6
*, sin6
, msg
->msg_name
);
472 if (flags
& MSG_ERRQUEUE
)
473 return ipv6_recv_error(sk
, msg
, len
, addr_len
);
475 if (np
->rxpmtu
&& np
->rxopt
.bits
.rxpmtu
)
476 return ipv6_recv_rxpmtu(sk
, msg
, len
, addr_len
);
478 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
485 msg
->msg_flags
|= MSG_TRUNC
;
488 if (skb_csum_unnecessary(skb
)) {
489 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
490 } else if (msg
->msg_flags
&MSG_TRUNC
) {
491 if (__skb_checksum_complete(skb
))
493 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
495 err
= skb_copy_and_csum_datagram_iovec(skb
, 0, msg
->msg_iov
);
502 /* Copy the address. */
504 sin6
->sin6_family
= AF_INET6
;
506 sin6
->sin6_addr
= ipv6_hdr(skb
)->saddr
;
507 sin6
->sin6_flowinfo
= 0;
508 sin6
->sin6_scope_id
= ipv6_iface_scope_id(&sin6
->sin6_addr
,
510 *addr_len
= sizeof(*sin6
);
513 sock_recv_ts_and_drops(msg
, sk
, skb
);
516 ip6_datagram_recv_ctl(sk
, msg
, skb
);
519 if (flags
& MSG_TRUNC
)
523 skb_free_datagram(sk
, skb
);
528 skb_kill_datagram(sk
, skb
, flags
);
530 /* Error for blocking case is chosen to masquerade
531 as some normal condition.
533 err
= (flags
&MSG_DONTWAIT
) ? -EAGAIN
: -EHOSTUNREACH
;
537 static int rawv6_push_pending_frames(struct sock
*sk
, struct flowi6
*fl6
,
538 struct raw6_sock
*rp
)
551 if ((skb
= skb_peek(&sk
->sk_write_queue
)) == NULL
)
555 total_len
= inet_sk(sk
)->cork
.base
.length
;
556 if (offset
>= total_len
- 1) {
558 ip6_flush_pending_frames(sk
);
562 /* should be check HW csum miyazawa */
563 if (skb_queue_len(&sk
->sk_write_queue
) == 1) {
565 * Only one fragment on the socket.
567 tmp_csum
= skb
->csum
;
569 struct sk_buff
*csum_skb
= NULL
;
572 skb_queue_walk(&sk
->sk_write_queue
, skb
) {
573 tmp_csum
= csum_add(tmp_csum
, skb
->csum
);
578 len
= skb
->len
- skb_transport_offset(skb
);
590 offset
+= skb_transport_offset(skb
);
591 if (skb_copy_bits(skb
, offset
, &csum
, 2))
594 /* in case cksum was not initialized */
596 tmp_csum
= csum_sub(tmp_csum
, csum_unfold(csum
));
598 csum
= csum_ipv6_magic(&fl6
->saddr
, &fl6
->daddr
,
599 total_len
, fl6
->flowi6_proto
, tmp_csum
);
601 if (csum
== 0 && fl6
->flowi6_proto
== IPPROTO_UDP
)
602 csum
= CSUM_MANGLED_0
;
604 if (skb_store_bits(skb
, offset
, &csum
, 2))
608 err
= ip6_push_pending_frames(sk
);
613 static int rawv6_send_hdrinc(struct sock
*sk
, void *from
, int length
,
614 struct flowi6
*fl6
, struct dst_entry
**dstp
,
617 struct ipv6_pinfo
*np
= inet6_sk(sk
);
621 struct rt6_info
*rt
= (struct rt6_info
*)*dstp
;
622 int hlen
= LL_RESERVED_SPACE(rt
->dst
.dev
);
623 int tlen
= rt
->dst
.dev
->needed_tailroom
;
625 if (length
> rt
->dst
.dev
->mtu
) {
626 ipv6_local_error(sk
, EMSGSIZE
, fl6
, rt
->dst
.dev
->mtu
);
632 skb
= sock_alloc_send_skb(sk
,
633 length
+ hlen
+ tlen
+ 15,
634 flags
& MSG_DONTWAIT
, &err
);
637 skb_reserve(skb
, hlen
);
639 skb
->protocol
= htons(ETH_P_IPV6
);
640 skb
->priority
= sk
->sk_priority
;
641 skb
->mark
= sk
->sk_mark
;
642 skb_dst_set(skb
, &rt
->dst
);
645 skb_put(skb
, length
);
646 skb_reset_network_header(skb
);
649 skb
->ip_summed
= CHECKSUM_NONE
;
651 skb
->transport_header
= skb
->network_header
;
652 err
= memcpy_fromiovecend((void *)iph
, from
, 0, length
);
656 IP6_UPD_PO_STATS(sock_net(sk
), rt
->rt6i_idev
, IPSTATS_MIB_OUT
, skb
->len
);
657 err
= NF_HOOK(NFPROTO_IPV6
, NF_INET_LOCAL_OUT
, skb
, NULL
,
658 rt
->dst
.dev
, dst_output
);
660 err
= net_xmit_errno(err
);
670 IP6_INC_STATS(sock_net(sk
), rt
->rt6i_idev
, IPSTATS_MIB_OUTDISCARDS
);
671 if (err
== -ENOBUFS
&& !np
->recverr
)
676 static int rawv6_probe_proto_opt(struct flowi6
*fl6
, struct msghdr
*msg
)
679 u8 __user
*type
= NULL
;
680 u8 __user
*code
= NULL
;
688 for (i
= 0; i
< msg
->msg_iovlen
; i
++) {
689 iov
= &msg
->msg_iov
[i
];
693 switch (fl6
->flowi6_proto
) {
695 /* check if one-byte field is readable or not. */
696 if (iov
->iov_base
&& iov
->iov_len
< 1)
700 type
= iov
->iov_base
;
701 /* check if code field is readable or not. */
702 if (iov
->iov_len
> 1)
705 code
= iov
->iov_base
;
708 if (get_user(fl6
->fl6_icmp_type
, type
) ||
709 get_user(fl6
->fl6_icmp_code
, code
))
715 if (iov
->iov_base
&& iov
->iov_len
< 1)
717 /* check if type field is readable or not. */
718 if (iov
->iov_len
> 2 - len
) {
719 u8 __user
*p
= iov
->iov_base
;
720 if (get_user(fl6
->fl6_mh_type
, &p
[2 - len
]))
737 static int rawv6_sendmsg(struct kiocb
*iocb
, struct sock
*sk
,
738 struct msghdr
*msg
, size_t len
)
740 struct ipv6_txoptions opt_space
;
741 DECLARE_SOCKADDR(struct sockaddr_in6
*, sin6
, msg
->msg_name
);
742 struct in6_addr
*daddr
, *final_p
, final
;
743 struct inet_sock
*inet
= inet_sk(sk
);
744 struct ipv6_pinfo
*np
= inet6_sk(sk
);
745 struct raw6_sock
*rp
= raw6_sk(sk
);
746 struct ipv6_txoptions
*opt
= NULL
;
747 struct ip6_flowlabel
*flowlabel
= NULL
;
748 struct dst_entry
*dst
= NULL
;
750 int addr_len
= msg
->msg_namelen
;
757 /* Rough check on arithmetic overflow,
758 better check is made in ip6_append_data().
763 /* Mirror BSD error message compatibility */
764 if (msg
->msg_flags
& MSG_OOB
)
768 * Get and verify the address.
770 memset(&fl6
, 0, sizeof(fl6
));
772 fl6
.flowi6_mark
= sk
->sk_mark
;
775 if (addr_len
< SIN6_LEN_RFC2133
)
778 if (sin6
->sin6_family
&& sin6
->sin6_family
!= AF_INET6
)
779 return -EAFNOSUPPORT
;
781 /* port is the proto value [0..255] carried in nexthdr */
782 proto
= ntohs(sin6
->sin6_port
);
785 proto
= inet
->inet_num
;
786 else if (proto
!= inet
->inet_num
)
792 daddr
= &sin6
->sin6_addr
;
794 fl6
.flowlabel
= sin6
->sin6_flowinfo
&IPV6_FLOWINFO_MASK
;
795 if (fl6
.flowlabel
&IPV6_FLOWLABEL_MASK
) {
796 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
797 if (flowlabel
== NULL
)
803 * Otherwise it will be difficult to maintain
806 if (sk
->sk_state
== TCP_ESTABLISHED
&&
807 ipv6_addr_equal(daddr
, &sk
->sk_v6_daddr
))
808 daddr
= &sk
->sk_v6_daddr
;
810 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
811 sin6
->sin6_scope_id
&&
812 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr
)))
813 fl6
.flowi6_oif
= sin6
->sin6_scope_id
;
815 if (sk
->sk_state
!= TCP_ESTABLISHED
)
816 return -EDESTADDRREQ
;
818 proto
= inet
->inet_num
;
819 daddr
= &sk
->sk_v6_daddr
;
820 fl6
.flowlabel
= np
->flow_label
;
823 if (fl6
.flowi6_oif
== 0)
824 fl6
.flowi6_oif
= sk
->sk_bound_dev_if
;
826 if (msg
->msg_controllen
) {
828 memset(opt
, 0, sizeof(struct ipv6_txoptions
));
829 opt
->tot_len
= sizeof(struct ipv6_txoptions
);
831 err
= ip6_datagram_send_ctl(sock_net(sk
), sk
, msg
, &fl6
, opt
,
832 &hlimit
, &tclass
, &dontfrag
);
834 fl6_sock_release(flowlabel
);
837 if ((fl6
.flowlabel
&IPV6_FLOWLABEL_MASK
) && !flowlabel
) {
838 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
839 if (flowlabel
== NULL
)
842 if (!(opt
->opt_nflen
|opt
->opt_flen
))
848 opt
= fl6_merge_options(&opt_space
, flowlabel
, opt
);
849 opt
= ipv6_fixup_options(&opt_space
, opt
);
851 fl6
.flowi6_proto
= proto
;
852 err
= rawv6_probe_proto_opt(&fl6
, msg
);
856 if (!ipv6_addr_any(daddr
))
859 fl6
.daddr
.s6_addr
[15] = 0x1; /* :: means loopback (BSD'ism) */
860 if (ipv6_addr_any(&fl6
.saddr
) && !ipv6_addr_any(&np
->saddr
))
861 fl6
.saddr
= np
->saddr
;
863 final_p
= fl6_update_dst(&fl6
, opt
, &final
);
865 if (!fl6
.flowi6_oif
&& ipv6_addr_is_multicast(&fl6
.daddr
))
866 fl6
.flowi6_oif
= np
->mcast_oif
;
867 else if (!fl6
.flowi6_oif
)
868 fl6
.flowi6_oif
= np
->ucast_oif
;
869 security_sk_classify_flow(sk
, flowi6_to_flowi(&fl6
));
871 dst
= ip6_dst_lookup_flow(sk
, &fl6
, final_p
);
877 hlimit
= ip6_sk_dst_hoplimit(np
, &fl6
, dst
);
883 dontfrag
= np
->dontfrag
;
885 if (msg
->msg_flags
&MSG_CONFIRM
)
890 err
= rawv6_send_hdrinc(sk
, msg
->msg_iov
, len
, &fl6
, &dst
, msg
->msg_flags
);
893 err
= ip6_append_data(sk
, ip_generic_getfrag
, msg
->msg_iov
,
894 len
, 0, hlimit
, tclass
, opt
, &fl6
, (struct rt6_info
*)dst
,
895 msg
->msg_flags
, dontfrag
);
898 ip6_flush_pending_frames(sk
);
899 else if (!(msg
->msg_flags
& MSG_MORE
))
900 err
= rawv6_push_pending_frames(sk
, &fl6
, rp
);
906 fl6_sock_release(flowlabel
);
907 return err
<0?err
:len
;
910 if (!(msg
->msg_flags
& MSG_PROBE
) || len
)
911 goto back_from_confirm
;
916 static int rawv6_seticmpfilter(struct sock
*sk
, int level
, int optname
,
917 char __user
*optval
, int optlen
)
921 if (optlen
> sizeof(struct icmp6_filter
))
922 optlen
= sizeof(struct icmp6_filter
);
923 if (copy_from_user(&raw6_sk(sk
)->filter
, optval
, optlen
))
933 static int rawv6_geticmpfilter(struct sock
*sk
, int level
, int optname
,
934 char __user
*optval
, int __user
*optlen
)
940 if (get_user(len
, optlen
))
944 if (len
> sizeof(struct icmp6_filter
))
945 len
= sizeof(struct icmp6_filter
);
946 if (put_user(len
, optlen
))
948 if (copy_to_user(optval
, &raw6_sk(sk
)->filter
, len
))
959 static int do_rawv6_setsockopt(struct sock
*sk
, int level
, int optname
,
960 char __user
*optval
, unsigned int optlen
)
962 struct raw6_sock
*rp
= raw6_sk(sk
);
965 if (get_user(val
, (int __user
*)optval
))
970 if (inet_sk(sk
)->inet_num
== IPPROTO_ICMPV6
&&
971 level
== IPPROTO_IPV6
) {
973 * RFC3542 tells that IPV6_CHECKSUM socket
974 * option in the IPPROTO_IPV6 level is not
975 * allowed on ICMPv6 sockets.
976 * If you want to set it, use IPPROTO_RAW
977 * level IPV6_CHECKSUM socket option
983 /* You may get strange result with a positive odd offset;
984 RFC2292bis agrees with me. */
985 if (val
> 0 && (val
&1))
1001 static int rawv6_setsockopt(struct sock
*sk
, int level
, int optname
,
1002 char __user
*optval
, unsigned int optlen
)
1009 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1011 return rawv6_seticmpfilter(sk
, level
, optname
, optval
, optlen
);
1013 if (optname
== IPV6_CHECKSUM
)
1016 return ipv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1019 return do_rawv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1022 #ifdef CONFIG_COMPAT
1023 static int compat_rawv6_setsockopt(struct sock
*sk
, int level
, int optname
,
1024 char __user
*optval
, unsigned int optlen
)
1030 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1032 return rawv6_seticmpfilter(sk
, level
, optname
, optval
, optlen
);
1034 if (optname
== IPV6_CHECKSUM
)
1037 return compat_ipv6_setsockopt(sk
, level
, optname
,
1040 return do_rawv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1044 static int do_rawv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1045 char __user
*optval
, int __user
*optlen
)
1047 struct raw6_sock
*rp
= raw6_sk(sk
);
1050 if (get_user(len
,optlen
))
1056 * We allow getsockopt() for IPPROTO_IPV6-level
1057 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1058 * since RFC3542 is silent about it.
1060 if (rp
->checksum
== 0)
1067 return -ENOPROTOOPT
;
1070 len
= min_t(unsigned int, sizeof(int), len
);
1072 if (put_user(len
, optlen
))
1074 if (copy_to_user(optval
,&val
,len
))
1079 static int rawv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1080 char __user
*optval
, int __user
*optlen
)
1087 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1089 return rawv6_geticmpfilter(sk
, level
, optname
, optval
, optlen
);
1091 if (optname
== IPV6_CHECKSUM
)
1094 return ipv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1097 return do_rawv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1100 #ifdef CONFIG_COMPAT
1101 static int compat_rawv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1102 char __user
*optval
, int __user
*optlen
)
1108 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMPV6
)
1110 return rawv6_geticmpfilter(sk
, level
, optname
, optval
, optlen
);
1112 if (optname
== IPV6_CHECKSUM
)
1115 return compat_ipv6_getsockopt(sk
, level
, optname
,
1118 return do_rawv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1122 static int rawv6_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
1126 int amount
= sk_wmem_alloc_get(sk
);
1128 return put_user(amount
, (int __user
*)arg
);
1131 struct sk_buff
*skb
;
1134 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
1135 skb
= skb_peek(&sk
->sk_receive_queue
);
1137 amount
= skb_tail_pointer(skb
) -
1138 skb_transport_header(skb
);
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_v6_only
,
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 int raw6_seq_show(struct seq_file
*seq
, void *v
)
1234 if (v
== SEQ_START_TOKEN
) {
1235 seq_puts(seq
, IPV6_SEQ_DGRAM_HEADER
);
1237 struct sock
*sp
= v
;
1238 __u16 srcp
= inet_sk(sp
)->inet_num
;
1239 ip6_dgram_sock_seq_show(seq
, v
, srcp
, 0,
1240 raw_seq_private(seq
)->bucket
);
1245 static const struct seq_operations raw6_seq_ops
= {
1246 .start
= raw_seq_start
,
1247 .next
= raw_seq_next
,
1248 .stop
= raw_seq_stop
,
1249 .show
= raw6_seq_show
,
1252 static int raw6_seq_open(struct inode
*inode
, struct file
*file
)
1254 return raw_seq_open(inode
, file
, &raw_v6_hashinfo
, &raw6_seq_ops
);
1257 static const struct file_operations raw6_seq_fops
= {
1258 .owner
= THIS_MODULE
,
1259 .open
= raw6_seq_open
,
1261 .llseek
= seq_lseek
,
1262 .release
= seq_release_net
,
1265 static int __net_init
raw6_init_net(struct net
*net
)
1267 if (!proc_create("raw6", S_IRUGO
, net
->proc_net
, &raw6_seq_fops
))
1273 static void __net_exit
raw6_exit_net(struct net
*net
)
1275 remove_proc_entry("raw6", net
->proc_net
);
1278 static struct pernet_operations raw6_net_ops
= {
1279 .init
= raw6_init_net
,
1280 .exit
= raw6_exit_net
,
1283 int __init
raw6_proc_init(void)
1285 return register_pernet_subsys(&raw6_net_ops
);
1288 void raw6_proc_exit(void)
1290 unregister_pernet_subsys(&raw6_net_ops
);
1292 #endif /* CONFIG_PROC_FS */
1294 /* Same as inet6_dgram_ops, sans udp_poll. */
1295 static const struct proto_ops inet6_sockraw_ops
= {
1297 .owner
= THIS_MODULE
,
1298 .release
= inet6_release
,
1300 .connect
= inet_dgram_connect
, /* ok */
1301 .socketpair
= sock_no_socketpair
, /* a do nothing */
1302 .accept
= sock_no_accept
, /* a do nothing */
1303 .getname
= inet6_getname
,
1304 .poll
= datagram_poll
, /* ok */
1305 .ioctl
= inet6_ioctl
, /* must change */
1306 .listen
= sock_no_listen
, /* ok */
1307 .shutdown
= inet_shutdown
, /* ok */
1308 .setsockopt
= sock_common_setsockopt
, /* ok */
1309 .getsockopt
= sock_common_getsockopt
, /* ok */
1310 .sendmsg
= inet_sendmsg
, /* ok */
1311 .recvmsg
= sock_common_recvmsg
, /* ok */
1312 .mmap
= sock_no_mmap
,
1313 .sendpage
= sock_no_sendpage
,
1314 #ifdef CONFIG_COMPAT
1315 .compat_setsockopt
= compat_sock_common_setsockopt
,
1316 .compat_getsockopt
= compat_sock_common_getsockopt
,
1320 static struct inet_protosw rawv6_protosw
= {
1322 .protocol
= IPPROTO_IP
, /* wild card */
1323 .prot
= &rawv6_prot
,
1324 .ops
= &inet6_sockraw_ops
,
1325 .flags
= INET_PROTOSW_REUSE
,
1328 int __init
rawv6_init(void)
1332 ret
= inet6_register_protosw(&rawv6_protosw
);
1339 void rawv6_exit(void)
1341 inet6_unregister_protosw(&rawv6_protosw
);