[CONNECTOR]: Replace delayed work with usual work queue.
[linux-2.6/verdex.git] / net / ipv6 / raw.c
blob4ae1b19ada5d3eb9e3c3c7f0b1b7a8030002eacb
1 /*
2 * RAW sockets for IPv6
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Adapted from linux/net/ipv4/raw.c
10 * $Id: raw.c,v 1.51 2002/02/01 22:01:04 davem Exp $
12 * Fixes:
13 * Hideaki YOSHIFUJI : sin6_scope_id support
14 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
17 * This program is free software; you can redistribute it and/or
18 * modify it under the terms of the GNU General Public License
19 * as published by the Free Software Foundation; either version
20 * 2 of the License, or (at your option) any later version.
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/sched.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/icmpv6.h>
33 #include <linux/netfilter.h>
34 #include <linux/netfilter_ipv6.h>
35 #include <linux/skbuff.h>
36 #include <asm/uaccess.h>
37 #include <asm/ioctls.h>
39 #include <net/ip.h>
40 #include <net/sock.h>
41 #include <net/snmp.h>
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/ip6_route.h>
47 #include <net/ip6_checksum.h>
48 #include <net/addrconf.h>
49 #include <net/transp_v6.h>
50 #include <net/udp.h>
51 #include <net/inet_common.h>
52 #include <net/tcp_states.h>
53 #ifdef CONFIG_IPV6_MIP6
54 #include <net/mip6.h>
55 #endif
57 #include <net/rawv6.h>
58 #include <net/xfrm.h>
60 #include <linux/proc_fs.h>
61 #include <linux/seq_file.h>
63 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
64 DEFINE_RWLOCK(raw_v6_lock);
66 static void raw_v6_hash(struct sock *sk)
68 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num &
69 (RAWV6_HTABLE_SIZE - 1)];
71 write_lock_bh(&raw_v6_lock);
72 sk_add_node(sk, list);
73 sock_prot_inc_use(sk->sk_prot);
74 write_unlock_bh(&raw_v6_lock);
77 static void raw_v6_unhash(struct sock *sk)
79 write_lock_bh(&raw_v6_lock);
80 if (sk_del_node_init(sk))
81 sock_prot_dec_use(sk->sk_prot);
82 write_unlock_bh(&raw_v6_lock);
86 /* Grumble... icmp and ip_input want to get at this... */
87 struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
88 struct in6_addr *loc_addr, struct in6_addr *rmt_addr,
89 int dif)
91 struct hlist_node *node;
92 int is_multicast = ipv6_addr_is_multicast(loc_addr);
94 sk_for_each_from(sk, node)
95 if (inet_sk(sk)->num == num) {
96 struct ipv6_pinfo *np = inet6_sk(sk);
98 if (!ipv6_addr_any(&np->daddr) &&
99 !ipv6_addr_equal(&np->daddr, rmt_addr))
100 continue;
102 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
103 continue;
105 if (!ipv6_addr_any(&np->rcv_saddr)) {
106 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
107 goto found;
108 if (is_multicast &&
109 inet6_mc_check(sk, loc_addr, rmt_addr))
110 goto found;
111 continue;
113 goto found;
115 sk = NULL;
116 found:
117 return sk;
121 * 0 - deliver
122 * 1 - block
124 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
126 struct icmp6hdr *icmph;
127 struct raw6_sock *rp = raw6_sk(sk);
129 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
130 __u32 *data = &rp->filter.data[0];
131 int bit_nr;
133 icmph = (struct icmp6hdr *) skb->data;
134 bit_nr = icmph->icmp6_type;
136 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
138 return 0;
142 * demultiplex raw sockets.
143 * (should consider queueing the skb in the sock receive_queue
144 * without calling rawv6.c)
146 * Caller owns SKB so we must make clones.
148 int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
150 struct in6_addr *saddr;
151 struct in6_addr *daddr;
152 struct sock *sk;
153 int delivered = 0;
154 __u8 hash;
156 saddr = &skb->nh.ipv6h->saddr;
157 daddr = saddr + 1;
159 hash = nexthdr & (MAX_INET_PROTOS - 1);
161 read_lock(&raw_v6_lock);
162 sk = sk_head(&raw_v6_htable[hash]);
165 * The first socket found will be delivered after
166 * delivery to transport protocols.
169 if (sk == NULL)
170 goto out;
172 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
174 while (sk) {
175 int filtered;
177 delivered = 1;
178 switch (nexthdr) {
179 case IPPROTO_ICMPV6:
180 filtered = icmpv6_filter(sk, skb);
181 break;
182 #ifdef CONFIG_IPV6_MIP6
183 case IPPROTO_MH:
184 /* XXX: To validate MH only once for each packet,
185 * this is placed here. It should be after checking
186 * xfrm policy, however it doesn't. The checking xfrm
187 * policy is placed in rawv6_rcv() because it is
188 * required for each socket.
190 filtered = mip6_mh_filter(sk, skb);
191 break;
192 #endif
193 default:
194 filtered = 0;
195 break;
198 if (filtered < 0)
199 break;
200 if (filtered == 0) {
201 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
203 /* Not releasing hash table! */
204 if (clone) {
205 nf_reset(clone);
206 rawv6_rcv(sk, clone);
209 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr,
210 IP6CB(skb)->iif);
212 out:
213 read_unlock(&raw_v6_lock);
214 return delivered;
217 /* This cleans up af_inet6 a bit. -DaveM */
218 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
220 struct inet_sock *inet = inet_sk(sk);
221 struct ipv6_pinfo *np = inet6_sk(sk);
222 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
223 __be32 v4addr = 0;
224 int addr_type;
225 int err;
227 if (addr_len < SIN6_LEN_RFC2133)
228 return -EINVAL;
229 addr_type = ipv6_addr_type(&addr->sin6_addr);
231 /* Raw sockets are IPv6 only */
232 if (addr_type == IPV6_ADDR_MAPPED)
233 return(-EADDRNOTAVAIL);
235 lock_sock(sk);
237 err = -EINVAL;
238 if (sk->sk_state != TCP_CLOSE)
239 goto out;
241 /* Check if the address belongs to the host. */
242 if (addr_type != IPV6_ADDR_ANY) {
243 struct net_device *dev = NULL;
245 if (addr_type & IPV6_ADDR_LINKLOCAL) {
246 if (addr_len >= sizeof(struct sockaddr_in6) &&
247 addr->sin6_scope_id) {
248 /* Override any existing binding, if another
249 * one is supplied by user.
251 sk->sk_bound_dev_if = addr->sin6_scope_id;
254 /* Binding to link-local address requires an interface */
255 if (!sk->sk_bound_dev_if)
256 goto out;
258 dev = dev_get_by_index(sk->sk_bound_dev_if);
259 if (!dev) {
260 err = -ENODEV;
261 goto out;
265 /* ipv4 addr of the socket is invalid. Only the
266 * unspecified and mapped address have a v4 equivalent.
268 v4addr = LOOPBACK4_IPV6;
269 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
270 err = -EADDRNOTAVAIL;
271 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
272 if (dev)
273 dev_put(dev);
274 goto out;
277 if (dev)
278 dev_put(dev);
281 inet->rcv_saddr = inet->saddr = v4addr;
282 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
283 if (!(addr_type & IPV6_ADDR_MULTICAST))
284 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
285 err = 0;
286 out:
287 release_sock(sk);
288 return err;
291 void rawv6_err(struct sock *sk, struct sk_buff *skb,
292 struct inet6_skb_parm *opt,
293 int type, int code, int offset, __be32 info)
295 struct inet_sock *inet = inet_sk(sk);
296 struct ipv6_pinfo *np = inet6_sk(sk);
297 int err;
298 int harderr;
300 /* Report error on raw socket, if:
301 1. User requested recverr.
302 2. Socket is connected (otherwise the error indication
303 is useless without recverr and error is hard.
305 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
306 return;
308 harderr = icmpv6_err_convert(type, code, &err);
309 if (type == ICMPV6_PKT_TOOBIG)
310 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
312 if (np->recverr) {
313 u8 *payload = skb->data;
314 if (!inet->hdrincl)
315 payload += offset;
316 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
319 if (np->recverr || harderr) {
320 sk->sk_err = err;
321 sk->sk_error_report(sk);
325 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
327 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
328 skb_checksum_complete(skb)) {
329 /* FIXME: increment a raw6 drops counter here */
330 kfree_skb(skb);
331 return 0;
334 /* Charge it to the socket. */
335 if (sock_queue_rcv_skb(sk,skb)<0) {
336 /* FIXME: increment a raw6 drops counter here */
337 kfree_skb(skb);
338 return 0;
341 return 0;
345 * This is next to useless...
346 * if we demultiplex in network layer we don't need the extra call
347 * just to queue the skb...
348 * maybe we could have the network decide upon a hint if it
349 * should call raw_rcv for demultiplexing
351 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
353 struct inet_sock *inet = inet_sk(sk);
354 struct raw6_sock *rp = raw6_sk(sk);
356 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
357 kfree_skb(skb);
358 return NET_RX_DROP;
361 if (!rp->checksum)
362 skb->ip_summed = CHECKSUM_UNNECESSARY;
364 if (skb->ip_summed == CHECKSUM_COMPLETE) {
365 skb_postpull_rcsum(skb, skb->nh.raw,
366 skb->h.raw - skb->nh.raw);
367 if (!csum_ipv6_magic(&skb->nh.ipv6h->saddr,
368 &skb->nh.ipv6h->daddr,
369 skb->len, inet->num, skb->csum))
370 skb->ip_summed = CHECKSUM_UNNECESSARY;
372 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
373 skb->csum = ~csum_unfold(csum_ipv6_magic(&skb->nh.ipv6h->saddr,
374 &skb->nh.ipv6h->daddr,
375 skb->len, inet->num, 0));
377 if (inet->hdrincl) {
378 if (skb_checksum_complete(skb)) {
379 /* FIXME: increment a raw6 drops counter here */
380 kfree_skb(skb);
381 return 0;
385 rawv6_rcv_skb(sk, skb);
386 return 0;
391 * This should be easy, if there is something there
392 * we return it, otherwise we block.
395 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
396 struct msghdr *msg, size_t len,
397 int noblock, int flags, int *addr_len)
399 struct ipv6_pinfo *np = inet6_sk(sk);
400 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
401 struct sk_buff *skb;
402 size_t copied;
403 int err;
405 if (flags & MSG_OOB)
406 return -EOPNOTSUPP;
408 if (addr_len)
409 *addr_len=sizeof(*sin6);
411 if (flags & MSG_ERRQUEUE)
412 return ipv6_recv_error(sk, msg, len);
414 skb = skb_recv_datagram(sk, flags, noblock, &err);
415 if (!skb)
416 goto out;
418 copied = skb->len;
419 if (copied > len) {
420 copied = len;
421 msg->msg_flags |= MSG_TRUNC;
424 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
425 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
426 } else if (msg->msg_flags&MSG_TRUNC) {
427 if (__skb_checksum_complete(skb))
428 goto csum_copy_err;
429 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
430 } else {
431 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
432 if (err == -EINVAL)
433 goto csum_copy_err;
435 if (err)
436 goto out_free;
438 /* Copy the address. */
439 if (sin6) {
440 sin6->sin6_family = AF_INET6;
441 sin6->sin6_port = 0;
442 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
443 sin6->sin6_flowinfo = 0;
444 sin6->sin6_scope_id = 0;
445 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
446 sin6->sin6_scope_id = IP6CB(skb)->iif;
449 sock_recv_timestamp(msg, sk, skb);
451 if (np->rxopt.all)
452 datagram_recv_ctl(sk, msg, skb);
454 err = copied;
455 if (flags & MSG_TRUNC)
456 err = skb->len;
458 out_free:
459 skb_free_datagram(sk, skb);
460 out:
461 return err;
463 csum_copy_err:
464 skb_kill_datagram(sk, skb, flags);
466 /* Error for blocking case is chosen to masquerade
467 as some normal condition.
469 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
470 /* FIXME: increment a raw6 drops counter here */
471 goto out;
474 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
475 struct raw6_sock *rp)
477 struct sk_buff *skb;
478 int err = 0;
479 int offset;
480 int len;
481 int total_len;
482 __wsum tmp_csum;
483 __sum16 csum;
485 if (!rp->checksum)
486 goto send;
488 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
489 goto out;
491 offset = rp->offset;
492 total_len = inet_sk(sk)->cork.length - (skb->nh.raw - skb->data);
493 if (offset >= total_len - 1) {
494 err = -EINVAL;
495 ip6_flush_pending_frames(sk);
496 goto out;
499 /* should be check HW csum miyazawa */
500 if (skb_queue_len(&sk->sk_write_queue) == 1) {
502 * Only one fragment on the socket.
504 tmp_csum = skb->csum;
505 } else {
506 struct sk_buff *csum_skb = NULL;
507 tmp_csum = 0;
509 skb_queue_walk(&sk->sk_write_queue, skb) {
510 tmp_csum = csum_add(tmp_csum, skb->csum);
512 if (csum_skb)
513 continue;
515 len = skb->len - (skb->h.raw - skb->data);
516 if (offset >= len) {
517 offset -= len;
518 continue;
521 csum_skb = skb;
524 skb = csum_skb;
527 offset += skb->h.raw - skb->data;
528 if (skb_copy_bits(skb, offset, &csum, 2))
529 BUG();
531 /* in case cksum was not initialized */
532 if (unlikely(csum))
533 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
535 csum = csum_ipv6_magic(&fl->fl6_src,
536 &fl->fl6_dst,
537 total_len, fl->proto, tmp_csum);
539 if (csum == 0 && fl->proto == IPPROTO_UDP)
540 csum = CSUM_MANGLED_0;
542 if (skb_store_bits(skb, offset, &csum, 2))
543 BUG();
545 send:
546 err = ip6_push_pending_frames(sk);
547 out:
548 return err;
551 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
552 struct flowi *fl, struct rt6_info *rt,
553 unsigned int flags)
555 struct ipv6_pinfo *np = inet6_sk(sk);
556 struct ipv6hdr *iph;
557 struct sk_buff *skb;
558 unsigned int hh_len;
559 int err;
561 if (length > rt->u.dst.dev->mtu) {
562 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
563 return -EMSGSIZE;
565 if (flags&MSG_PROBE)
566 goto out;
568 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
570 skb = sock_alloc_send_skb(sk, length+hh_len+15,
571 flags&MSG_DONTWAIT, &err);
572 if (skb == NULL)
573 goto error;
574 skb_reserve(skb, hh_len);
576 skb->priority = sk->sk_priority;
577 skb->dst = dst_clone(&rt->u.dst);
579 skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length);
581 skb->ip_summed = CHECKSUM_NONE;
583 skb->h.raw = skb->nh.raw;
584 err = memcpy_fromiovecend((void *)iph, from, 0, length);
585 if (err)
586 goto error_fault;
588 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
589 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
590 dst_output);
591 if (err > 0)
592 err = np->recverr ? net_xmit_errno(err) : 0;
593 if (err)
594 goto error;
595 out:
596 return 0;
598 error_fault:
599 err = -EFAULT;
600 kfree_skb(skb);
601 error:
602 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
603 return err;
606 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
608 struct iovec *iov;
609 u8 __user *type = NULL;
610 u8 __user *code = NULL;
611 #ifdef CONFIG_IPV6_MIP6
612 u8 len = 0;
613 #endif
614 int probed = 0;
615 int i;
617 if (!msg->msg_iov)
618 return 0;
620 for (i = 0; i < msg->msg_iovlen; i++) {
621 iov = &msg->msg_iov[i];
622 if (!iov)
623 continue;
625 switch (fl->proto) {
626 case IPPROTO_ICMPV6:
627 /* check if one-byte field is readable or not. */
628 if (iov->iov_base && iov->iov_len < 1)
629 break;
631 if (!type) {
632 type = iov->iov_base;
633 /* check if code field is readable or not. */
634 if (iov->iov_len > 1)
635 code = type + 1;
636 } else if (!code)
637 code = iov->iov_base;
639 if (type && code) {
640 if (get_user(fl->fl_icmp_type, type) ||
641 get_user(fl->fl_icmp_code, code))
642 return -EFAULT;
643 probed = 1;
645 break;
646 #ifdef CONFIG_IPV6_MIP6
647 case IPPROTO_MH:
648 if (iov->iov_base && iov->iov_len < 1)
649 break;
650 /* check if type field is readable or not. */
651 if (iov->iov_len > 2 - len) {
652 u8 __user *p = iov->iov_base;
653 if (get_user(fl->fl_mh_type, &p[2 - len]))
654 return -EFAULT;
655 probed = 1;
656 } else
657 len += iov->iov_len;
659 break;
660 #endif
661 default:
662 probed = 1;
663 break;
665 if (probed)
666 break;
668 return 0;
671 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
672 struct msghdr *msg, size_t len)
674 struct ipv6_txoptions opt_space;
675 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
676 struct in6_addr *daddr, *final_p = NULL, final;
677 struct inet_sock *inet = inet_sk(sk);
678 struct ipv6_pinfo *np = inet6_sk(sk);
679 struct raw6_sock *rp = raw6_sk(sk);
680 struct ipv6_txoptions *opt = NULL;
681 struct ip6_flowlabel *flowlabel = NULL;
682 struct dst_entry *dst = NULL;
683 struct flowi fl;
684 int addr_len = msg->msg_namelen;
685 int hlimit = -1;
686 int tclass = -1;
687 u16 proto;
688 int err;
690 /* Rough check on arithmetic overflow,
691 better check is made in ip6_build_xmit
693 if (len < 0)
694 return -EMSGSIZE;
696 /* Mirror BSD error message compatibility */
697 if (msg->msg_flags & MSG_OOB)
698 return -EOPNOTSUPP;
701 * Get and verify the address.
703 memset(&fl, 0, sizeof(fl));
705 if (sin6) {
706 if (addr_len < SIN6_LEN_RFC2133)
707 return -EINVAL;
709 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
710 return(-EAFNOSUPPORT);
712 /* port is the proto value [0..255] carried in nexthdr */
713 proto = ntohs(sin6->sin6_port);
715 if (!proto)
716 proto = inet->num;
717 else if (proto != inet->num)
718 return(-EINVAL);
720 if (proto > 255)
721 return(-EINVAL);
723 daddr = &sin6->sin6_addr;
724 if (np->sndflow) {
725 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
726 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
727 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
728 if (flowlabel == NULL)
729 return -EINVAL;
730 daddr = &flowlabel->dst;
735 * Otherwise it will be difficult to maintain
736 * sk->sk_dst_cache.
738 if (sk->sk_state == TCP_ESTABLISHED &&
739 ipv6_addr_equal(daddr, &np->daddr))
740 daddr = &np->daddr;
742 if (addr_len >= sizeof(struct sockaddr_in6) &&
743 sin6->sin6_scope_id &&
744 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
745 fl.oif = sin6->sin6_scope_id;
746 } else {
747 if (sk->sk_state != TCP_ESTABLISHED)
748 return -EDESTADDRREQ;
750 proto = inet->num;
751 daddr = &np->daddr;
752 fl.fl6_flowlabel = np->flow_label;
755 if (ipv6_addr_any(daddr)) {
757 * unspecified destination address
758 * treated as error... is this correct ?
760 fl6_sock_release(flowlabel);
761 return(-EINVAL);
764 if (fl.oif == 0)
765 fl.oif = sk->sk_bound_dev_if;
767 if (msg->msg_controllen) {
768 opt = &opt_space;
769 memset(opt, 0, sizeof(struct ipv6_txoptions));
770 opt->tot_len = sizeof(struct ipv6_txoptions);
772 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
773 if (err < 0) {
774 fl6_sock_release(flowlabel);
775 return err;
777 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
778 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
779 if (flowlabel == NULL)
780 return -EINVAL;
782 if (!(opt->opt_nflen|opt->opt_flen))
783 opt = NULL;
785 if (opt == NULL)
786 opt = np->opt;
787 if (flowlabel)
788 opt = fl6_merge_options(&opt_space, flowlabel, opt);
789 opt = ipv6_fixup_options(&opt_space, opt);
791 fl.proto = proto;
792 err = rawv6_probe_proto_opt(&fl, msg);
793 if (err)
794 goto out;
796 ipv6_addr_copy(&fl.fl6_dst, daddr);
797 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
798 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
800 /* merge ip6_build_xmit from ip6_output */
801 if (opt && opt->srcrt) {
802 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
803 ipv6_addr_copy(&final, &fl.fl6_dst);
804 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
805 final_p = &final;
808 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
809 fl.oif = np->mcast_oif;
810 security_sk_classify_flow(sk, &fl);
812 err = ip6_dst_lookup(sk, &dst, &fl);
813 if (err)
814 goto out;
815 if (final_p)
816 ipv6_addr_copy(&fl.fl6_dst, final_p);
818 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
819 goto out;
821 if (hlimit < 0) {
822 if (ipv6_addr_is_multicast(&fl.fl6_dst))
823 hlimit = np->mcast_hops;
824 else
825 hlimit = np->hop_limit;
826 if (hlimit < 0)
827 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
828 if (hlimit < 0)
829 hlimit = ipv6_get_hoplimit(dst->dev);
832 if (tclass < 0) {
833 tclass = np->tclass;
834 if (tclass < 0)
835 tclass = 0;
838 if (msg->msg_flags&MSG_CONFIRM)
839 goto do_confirm;
841 back_from_confirm:
842 if (inet->hdrincl) {
843 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
844 } else {
845 lock_sock(sk);
846 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
847 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
848 msg->msg_flags);
850 if (err)
851 ip6_flush_pending_frames(sk);
852 else if (!(msg->msg_flags & MSG_MORE))
853 err = rawv6_push_pending_frames(sk, &fl, rp);
855 done:
856 dst_release(dst);
857 if (!inet->hdrincl)
858 release_sock(sk);
859 out:
860 fl6_sock_release(flowlabel);
861 return err<0?err:len;
862 do_confirm:
863 dst_confirm(dst);
864 if (!(msg->msg_flags & MSG_PROBE) || len)
865 goto back_from_confirm;
866 err = 0;
867 goto done;
870 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
871 char __user *optval, int optlen)
873 switch (optname) {
874 case ICMPV6_FILTER:
875 if (optlen > sizeof(struct icmp6_filter))
876 optlen = sizeof(struct icmp6_filter);
877 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
878 return -EFAULT;
879 return 0;
880 default:
881 return -ENOPROTOOPT;
884 return 0;
887 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
888 char __user *optval, int __user *optlen)
890 int len;
892 switch (optname) {
893 case ICMPV6_FILTER:
894 if (get_user(len, optlen))
895 return -EFAULT;
896 if (len < 0)
897 return -EINVAL;
898 if (len > sizeof(struct icmp6_filter))
899 len = sizeof(struct icmp6_filter);
900 if (put_user(len, optlen))
901 return -EFAULT;
902 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
903 return -EFAULT;
904 return 0;
905 default:
906 return -ENOPROTOOPT;
909 return 0;
913 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
914 char __user *optval, int optlen)
916 struct raw6_sock *rp = raw6_sk(sk);
917 int val;
919 if (get_user(val, (int __user *)optval))
920 return -EFAULT;
922 switch (optname) {
923 case IPV6_CHECKSUM:
924 /* You may get strange result with a positive odd offset;
925 RFC2292bis agrees with me. */
926 if (val > 0 && (val&1))
927 return(-EINVAL);
928 if (val < 0) {
929 rp->checksum = 0;
930 } else {
931 rp->checksum = 1;
932 rp->offset = val;
935 return 0;
936 break;
938 default:
939 return(-ENOPROTOOPT);
943 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
944 char __user *optval, int optlen)
946 switch(level) {
947 case SOL_RAW:
948 break;
950 case SOL_ICMPV6:
951 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
952 return -EOPNOTSUPP;
953 return rawv6_seticmpfilter(sk, level, optname, optval,
954 optlen);
955 case SOL_IPV6:
956 if (optname == IPV6_CHECKSUM)
957 break;
958 default:
959 return ipv6_setsockopt(sk, level, optname, optval,
960 optlen);
962 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
965 #ifdef CONFIG_COMPAT
966 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
967 char __user *optval, int optlen)
969 switch (level) {
970 case SOL_RAW:
971 break;
972 case SOL_ICMPV6:
973 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
974 return -EOPNOTSUPP;
975 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
976 case SOL_IPV6:
977 if (optname == IPV6_CHECKSUM)
978 break;
979 default:
980 return compat_ipv6_setsockopt(sk, level, optname,
981 optval, optlen);
983 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
985 #endif
987 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
988 char __user *optval, int __user *optlen)
990 struct raw6_sock *rp = raw6_sk(sk);
991 int val, len;
993 if (get_user(len,optlen))
994 return -EFAULT;
996 switch (optname) {
997 case IPV6_CHECKSUM:
998 if (rp->checksum == 0)
999 val = -1;
1000 else
1001 val = rp->offset;
1002 break;
1004 default:
1005 return -ENOPROTOOPT;
1008 len = min_t(unsigned int, sizeof(int), len);
1010 if (put_user(len, optlen))
1011 return -EFAULT;
1012 if (copy_to_user(optval,&val,len))
1013 return -EFAULT;
1014 return 0;
1017 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1018 char __user *optval, int __user *optlen)
1020 switch(level) {
1021 case SOL_RAW:
1022 break;
1024 case SOL_ICMPV6:
1025 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1026 return -EOPNOTSUPP;
1027 return rawv6_geticmpfilter(sk, level, optname, optval,
1028 optlen);
1029 case SOL_IPV6:
1030 if (optname == IPV6_CHECKSUM)
1031 break;
1032 default:
1033 return ipv6_getsockopt(sk, level, optname, optval,
1034 optlen);
1036 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1039 #ifdef CONFIG_COMPAT
1040 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1041 char __user *optval, int __user *optlen)
1043 switch (level) {
1044 case SOL_RAW:
1045 break;
1046 case SOL_ICMPV6:
1047 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1048 return -EOPNOTSUPP;
1049 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1050 case SOL_IPV6:
1051 if (optname == IPV6_CHECKSUM)
1052 break;
1053 default:
1054 return compat_ipv6_getsockopt(sk, level, optname,
1055 optval, optlen);
1057 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1059 #endif
1061 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1063 switch(cmd) {
1064 case SIOCOUTQ:
1066 int amount = atomic_read(&sk->sk_wmem_alloc);
1067 return put_user(amount, (int __user *)arg);
1069 case SIOCINQ:
1071 struct sk_buff *skb;
1072 int amount = 0;
1074 spin_lock_bh(&sk->sk_receive_queue.lock);
1075 skb = skb_peek(&sk->sk_receive_queue);
1076 if (skb != NULL)
1077 amount = skb->tail - skb->h.raw;
1078 spin_unlock_bh(&sk->sk_receive_queue.lock);
1079 return put_user(amount, (int __user *)arg);
1082 default:
1083 return -ENOIOCTLCMD;
1087 static void rawv6_close(struct sock *sk, long timeout)
1089 if (inet_sk(sk)->num == IPPROTO_RAW)
1090 ip6_ra_control(sk, -1, NULL);
1092 sk_common_release(sk);
1095 static int rawv6_init_sk(struct sock *sk)
1097 if (inet_sk(sk)->num == IPPROTO_ICMPV6) {
1098 struct raw6_sock *rp = raw6_sk(sk);
1099 rp->checksum = 1;
1100 rp->offset = 2;
1102 return(0);
1105 struct proto rawv6_prot = {
1106 .name = "RAWv6",
1107 .owner = THIS_MODULE,
1108 .close = rawv6_close,
1109 .connect = ip6_datagram_connect,
1110 .disconnect = udp_disconnect,
1111 .ioctl = rawv6_ioctl,
1112 .init = rawv6_init_sk,
1113 .destroy = inet6_destroy_sock,
1114 .setsockopt = rawv6_setsockopt,
1115 .getsockopt = rawv6_getsockopt,
1116 .sendmsg = rawv6_sendmsg,
1117 .recvmsg = rawv6_recvmsg,
1118 .bind = rawv6_bind,
1119 .backlog_rcv = rawv6_rcv_skb,
1120 .hash = raw_v6_hash,
1121 .unhash = raw_v6_unhash,
1122 .obj_size = sizeof(struct raw6_sock),
1123 #ifdef CONFIG_COMPAT
1124 .compat_setsockopt = compat_rawv6_setsockopt,
1125 .compat_getsockopt = compat_rawv6_getsockopt,
1126 #endif
1129 #ifdef CONFIG_PROC_FS
1130 struct raw6_iter_state {
1131 int bucket;
1134 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1136 static struct sock *raw6_get_first(struct seq_file *seq)
1138 struct sock *sk;
1139 struct hlist_node *node;
1140 struct raw6_iter_state* state = raw6_seq_private(seq);
1142 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1143 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1144 if (sk->sk_family == PF_INET6)
1145 goto out;
1146 sk = NULL;
1147 out:
1148 return sk;
1151 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1153 struct raw6_iter_state* state = raw6_seq_private(seq);
1155 do {
1156 sk = sk_next(sk);
1157 try_again:
1159 } while (sk && sk->sk_family != PF_INET6);
1161 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1162 sk = sk_head(&raw_v6_htable[state->bucket]);
1163 goto try_again;
1165 return sk;
1168 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1170 struct sock *sk = raw6_get_first(seq);
1171 if (sk)
1172 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1173 --pos;
1174 return pos ? NULL : sk;
1177 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1179 read_lock(&raw_v6_lock);
1180 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1183 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1185 struct sock *sk;
1187 if (v == SEQ_START_TOKEN)
1188 sk = raw6_get_first(seq);
1189 else
1190 sk = raw6_get_next(seq, v);
1191 ++*pos;
1192 return sk;
1195 static void raw6_seq_stop(struct seq_file *seq, void *v)
1197 read_unlock(&raw_v6_lock);
1200 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1202 struct ipv6_pinfo *np = inet6_sk(sp);
1203 struct in6_addr *dest, *src;
1204 __u16 destp, srcp;
1206 dest = &np->daddr;
1207 src = &np->rcv_saddr;
1208 destp = 0;
1209 srcp = inet_sk(sp)->num;
1210 seq_printf(seq,
1211 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1212 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1214 src->s6_addr32[0], src->s6_addr32[1],
1215 src->s6_addr32[2], src->s6_addr32[3], srcp,
1216 dest->s6_addr32[0], dest->s6_addr32[1],
1217 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1218 sp->sk_state,
1219 atomic_read(&sp->sk_wmem_alloc),
1220 atomic_read(&sp->sk_rmem_alloc),
1221 0, 0L, 0,
1222 sock_i_uid(sp), 0,
1223 sock_i_ino(sp),
1224 atomic_read(&sp->sk_refcnt), sp);
1227 static int raw6_seq_show(struct seq_file *seq, void *v)
1229 if (v == SEQ_START_TOKEN)
1230 seq_printf(seq,
1231 " sl "
1232 "local_address "
1233 "remote_address "
1234 "st tx_queue rx_queue tr tm->when retrnsmt"
1235 " uid timeout inode\n");
1236 else
1237 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1238 return 0;
1241 static struct seq_operations raw6_seq_ops = {
1242 .start = raw6_seq_start,
1243 .next = raw6_seq_next,
1244 .stop = raw6_seq_stop,
1245 .show = raw6_seq_show,
1248 static int raw6_seq_open(struct inode *inode, struct file *file)
1250 struct seq_file *seq;
1251 int rc = -ENOMEM;
1252 struct raw6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
1253 if (!s)
1254 goto out;
1255 rc = seq_open(file, &raw6_seq_ops);
1256 if (rc)
1257 goto out_kfree;
1258 seq = file->private_data;
1259 seq->private = s;
1260 out:
1261 return rc;
1262 out_kfree:
1263 kfree(s);
1264 goto out;
1267 static struct file_operations raw6_seq_fops = {
1268 .owner = THIS_MODULE,
1269 .open = raw6_seq_open,
1270 .read = seq_read,
1271 .llseek = seq_lseek,
1272 .release = seq_release_private,
1275 int __init raw6_proc_init(void)
1277 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1278 return -ENOMEM;
1279 return 0;
1282 void raw6_proc_exit(void)
1284 proc_net_remove("raw6");
1286 #endif /* CONFIG_PROC_FS */