[x86 setup] Handle case of improperly terminated E820 chain
[wrt350n-kernel.git] / net / ipv6 / raw.c
blob77167afa3455a162fae611ce7fbbdeec764d949a
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/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/netfilter.h>
33 #include <linux/netfilter_ipv6.h>
34 #include <linux/skbuff.h>
35 #include <asm/uaccess.h>
36 #include <asm/ioctls.h>
38 #include <net/ip.h>
39 #include <net/sock.h>
40 #include <net/snmp.h>
42 #include <net/ipv6.h>
43 #include <net/ndisc.h>
44 #include <net/protocol.h>
45 #include <net/ip6_route.h>
46 #include <net/ip6_checksum.h>
47 #include <net/addrconf.h>
48 #include <net/transp_v6.h>
49 #include <net/udp.h>
50 #include <net/inet_common.h>
51 #include <net/tcp_states.h>
52 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
53 #include <net/mip6.h>
54 #endif
56 #include <net/rawv6.h>
57 #include <net/xfrm.h>
59 #include <linux/proc_fs.h>
60 #include <linux/seq_file.h>
62 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
63 DEFINE_RWLOCK(raw_v6_lock);
65 static void raw_v6_hash(struct sock *sk)
67 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num &
68 (RAWV6_HTABLE_SIZE - 1)];
70 write_lock_bh(&raw_v6_lock);
71 sk_add_node(sk, list);
72 sock_prot_inc_use(sk->sk_prot);
73 write_unlock_bh(&raw_v6_lock);
76 static void raw_v6_unhash(struct sock *sk)
78 write_lock_bh(&raw_v6_lock);
79 if (sk_del_node_init(sk))
80 sock_prot_dec_use(sk->sk_prot);
81 write_unlock_bh(&raw_v6_lock);
85 /* Grumble... icmp and ip_input want to get at this... */
86 struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
87 struct in6_addr *loc_addr, struct in6_addr *rmt_addr,
88 int dif)
90 struct hlist_node *node;
91 int is_multicast = ipv6_addr_is_multicast(loc_addr);
93 sk_for_each_from(sk, node)
94 if (inet_sk(sk)->num == num) {
95 struct ipv6_pinfo *np = inet6_sk(sk);
97 if (!ipv6_addr_any(&np->daddr) &&
98 !ipv6_addr_equal(&np->daddr, rmt_addr))
99 continue;
101 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
102 continue;
104 if (!ipv6_addr_any(&np->rcv_saddr)) {
105 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
106 goto found;
107 if (is_multicast &&
108 inet6_mc_check(sk, loc_addr, rmt_addr))
109 goto found;
110 continue;
112 goto found;
114 sk = NULL;
115 found:
116 return sk;
120 * 0 - deliver
121 * 1 - block
123 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
125 struct icmp6hdr *icmph;
126 struct raw6_sock *rp = raw6_sk(sk);
128 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
129 __u32 *data = &rp->filter.data[0];
130 int bit_nr;
132 icmph = (struct icmp6hdr *) skb->data;
133 bit_nr = icmph->icmp6_type;
135 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
137 return 0;
140 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
141 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb);
143 int rawv6_mh_filter_register(int (*filter)(struct sock *sock,
144 struct sk_buff *skb))
146 rcu_assign_pointer(mh_filter, filter);
147 return 0;
149 EXPORT_SYMBOL(rawv6_mh_filter_register);
151 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock,
152 struct sk_buff *skb))
154 rcu_assign_pointer(mh_filter, NULL);
155 synchronize_rcu();
156 return 0;
158 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
160 #endif
163 * demultiplex raw sockets.
164 * (should consider queueing the skb in the sock receive_queue
165 * without calling rawv6.c)
167 * Caller owns SKB so we must make clones.
169 int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
171 struct in6_addr *saddr;
172 struct in6_addr *daddr;
173 struct sock *sk;
174 int delivered = 0;
175 __u8 hash;
177 saddr = &ipv6_hdr(skb)->saddr;
178 daddr = saddr + 1;
180 hash = nexthdr & (MAX_INET_PROTOS - 1);
182 read_lock(&raw_v6_lock);
183 sk = sk_head(&raw_v6_htable[hash]);
186 * The first socket found will be delivered after
187 * delivery to transport protocols.
190 if (sk == NULL)
191 goto out;
193 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
195 while (sk) {
196 int filtered;
198 delivered = 1;
199 switch (nexthdr) {
200 case IPPROTO_ICMPV6:
201 filtered = icmpv6_filter(sk, skb);
202 break;
204 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
205 case IPPROTO_MH:
207 /* XXX: To validate MH only once for each packet,
208 * this is placed here. It should be after checking
209 * xfrm policy, however it doesn't. The checking xfrm
210 * policy is placed in rawv6_rcv() because it is
211 * required for each socket.
213 int (*filter)(struct sock *sock, struct sk_buff *skb);
215 filter = rcu_dereference(mh_filter);
216 filtered = filter ? filter(sk, skb) : 0;
217 break;
219 #endif
220 default:
221 filtered = 0;
222 break;
225 if (filtered < 0)
226 break;
227 if (filtered == 0) {
228 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
230 /* Not releasing hash table! */
231 if (clone) {
232 nf_reset(clone);
233 rawv6_rcv(sk, clone);
236 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr,
237 IP6CB(skb)->iif);
239 out:
240 read_unlock(&raw_v6_lock);
241 return delivered;
244 /* This cleans up af_inet6 a bit. -DaveM */
245 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
247 struct inet_sock *inet = inet_sk(sk);
248 struct ipv6_pinfo *np = inet6_sk(sk);
249 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
250 __be32 v4addr = 0;
251 int addr_type;
252 int err;
254 if (addr_len < SIN6_LEN_RFC2133)
255 return -EINVAL;
256 addr_type = ipv6_addr_type(&addr->sin6_addr);
258 /* Raw sockets are IPv6 only */
259 if (addr_type == IPV6_ADDR_MAPPED)
260 return(-EADDRNOTAVAIL);
262 lock_sock(sk);
264 err = -EINVAL;
265 if (sk->sk_state != TCP_CLOSE)
266 goto out;
268 /* Check if the address belongs to the host. */
269 if (addr_type != IPV6_ADDR_ANY) {
270 struct net_device *dev = NULL;
272 if (addr_type & IPV6_ADDR_LINKLOCAL) {
273 if (addr_len >= sizeof(struct sockaddr_in6) &&
274 addr->sin6_scope_id) {
275 /* Override any existing binding, if another
276 * one is supplied by user.
278 sk->sk_bound_dev_if = addr->sin6_scope_id;
281 /* Binding to link-local address requires an interface */
282 if (!sk->sk_bound_dev_if)
283 goto out;
285 dev = dev_get_by_index(sk->sk_bound_dev_if);
286 if (!dev) {
287 err = -ENODEV;
288 goto out;
292 /* ipv4 addr of the socket is invalid. Only the
293 * unspecified and mapped address have a v4 equivalent.
295 v4addr = LOOPBACK4_IPV6;
296 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
297 err = -EADDRNOTAVAIL;
298 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
299 if (dev)
300 dev_put(dev);
301 goto out;
304 if (dev)
305 dev_put(dev);
308 inet->rcv_saddr = inet->saddr = v4addr;
309 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
310 if (!(addr_type & IPV6_ADDR_MULTICAST))
311 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
312 err = 0;
313 out:
314 release_sock(sk);
315 return err;
318 void rawv6_err(struct sock *sk, struct sk_buff *skb,
319 struct inet6_skb_parm *opt,
320 int type, int code, int offset, __be32 info)
322 struct inet_sock *inet = inet_sk(sk);
323 struct ipv6_pinfo *np = inet6_sk(sk);
324 int err;
325 int harderr;
327 /* Report error on raw socket, if:
328 1. User requested recverr.
329 2. Socket is connected (otherwise the error indication
330 is useless without recverr and error is hard.
332 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
333 return;
335 harderr = icmpv6_err_convert(type, code, &err);
336 if (type == ICMPV6_PKT_TOOBIG)
337 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
339 if (np->recverr) {
340 u8 *payload = skb->data;
341 if (!inet->hdrincl)
342 payload += offset;
343 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
346 if (np->recverr || harderr) {
347 sk->sk_err = err;
348 sk->sk_error_report(sk);
352 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
354 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
355 skb_checksum_complete(skb)) {
356 /* FIXME: increment a raw6 drops counter here */
357 kfree_skb(skb);
358 return 0;
361 /* Charge it to the socket. */
362 if (sock_queue_rcv_skb(sk,skb)<0) {
363 /* FIXME: increment a raw6 drops counter here */
364 kfree_skb(skb);
365 return 0;
368 return 0;
372 * This is next to useless...
373 * if we demultiplex in network layer we don't need the extra call
374 * just to queue the skb...
375 * maybe we could have the network decide upon a hint if it
376 * should call raw_rcv for demultiplexing
378 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
380 struct inet_sock *inet = inet_sk(sk);
381 struct raw6_sock *rp = raw6_sk(sk);
383 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
384 kfree_skb(skb);
385 return NET_RX_DROP;
388 if (!rp->checksum)
389 skb->ip_summed = CHECKSUM_UNNECESSARY;
391 if (skb->ip_summed == CHECKSUM_COMPLETE) {
392 skb_postpull_rcsum(skb, skb_network_header(skb),
393 skb_network_header_len(skb));
394 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
395 &ipv6_hdr(skb)->daddr,
396 skb->len, inet->num, skb->csum))
397 skb->ip_summed = CHECKSUM_UNNECESSARY;
399 if (!skb_csum_unnecessary(skb))
400 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
401 &ipv6_hdr(skb)->daddr,
402 skb->len,
403 inet->num, 0));
405 if (inet->hdrincl) {
406 if (skb_checksum_complete(skb)) {
407 /* FIXME: increment a raw6 drops counter here */
408 kfree_skb(skb);
409 return 0;
413 rawv6_rcv_skb(sk, skb);
414 return 0;
419 * This should be easy, if there is something there
420 * we return it, otherwise we block.
423 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
424 struct msghdr *msg, size_t len,
425 int noblock, int flags, int *addr_len)
427 struct ipv6_pinfo *np = inet6_sk(sk);
428 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
429 struct sk_buff *skb;
430 size_t copied;
431 int err;
433 if (flags & MSG_OOB)
434 return -EOPNOTSUPP;
436 if (addr_len)
437 *addr_len=sizeof(*sin6);
439 if (flags & MSG_ERRQUEUE)
440 return ipv6_recv_error(sk, msg, len);
442 skb = skb_recv_datagram(sk, flags, noblock, &err);
443 if (!skb)
444 goto out;
446 copied = skb->len;
447 if (copied > len) {
448 copied = len;
449 msg->msg_flags |= MSG_TRUNC;
452 if (skb_csum_unnecessary(skb)) {
453 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
454 } else if (msg->msg_flags&MSG_TRUNC) {
455 if (__skb_checksum_complete(skb))
456 goto csum_copy_err;
457 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
458 } else {
459 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
460 if (err == -EINVAL)
461 goto csum_copy_err;
463 if (err)
464 goto out_free;
466 /* Copy the address. */
467 if (sin6) {
468 sin6->sin6_family = AF_INET6;
469 sin6->sin6_port = 0;
470 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
471 sin6->sin6_flowinfo = 0;
472 sin6->sin6_scope_id = 0;
473 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
474 sin6->sin6_scope_id = IP6CB(skb)->iif;
477 sock_recv_timestamp(msg, sk, skb);
479 if (np->rxopt.all)
480 datagram_recv_ctl(sk, msg, skb);
482 err = copied;
483 if (flags & MSG_TRUNC)
484 err = skb->len;
486 out_free:
487 skb_free_datagram(sk, skb);
488 out:
489 return err;
491 csum_copy_err:
492 skb_kill_datagram(sk, skb, flags);
494 /* Error for blocking case is chosen to masquerade
495 as some normal condition.
497 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
498 /* FIXME: increment a raw6 drops counter here */
499 goto out;
502 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
503 struct raw6_sock *rp)
505 struct sk_buff *skb;
506 int err = 0;
507 int offset;
508 int len;
509 int total_len;
510 __wsum tmp_csum;
511 __sum16 csum;
513 if (!rp->checksum)
514 goto send;
516 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
517 goto out;
519 offset = rp->offset;
520 total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
521 skb->data);
522 if (offset >= total_len - 1) {
523 err = -EINVAL;
524 ip6_flush_pending_frames(sk);
525 goto out;
528 /* should be check HW csum miyazawa */
529 if (skb_queue_len(&sk->sk_write_queue) == 1) {
531 * Only one fragment on the socket.
533 tmp_csum = skb->csum;
534 } else {
535 struct sk_buff *csum_skb = NULL;
536 tmp_csum = 0;
538 skb_queue_walk(&sk->sk_write_queue, skb) {
539 tmp_csum = csum_add(tmp_csum, skb->csum);
541 if (csum_skb)
542 continue;
544 len = skb->len - skb_transport_offset(skb);
545 if (offset >= len) {
546 offset -= len;
547 continue;
550 csum_skb = skb;
553 skb = csum_skb;
556 offset += skb_transport_offset(skb);
557 if (skb_copy_bits(skb, offset, &csum, 2))
558 BUG();
560 /* in case cksum was not initialized */
561 if (unlikely(csum))
562 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
564 csum = csum_ipv6_magic(&fl->fl6_src,
565 &fl->fl6_dst,
566 total_len, fl->proto, tmp_csum);
568 if (csum == 0 && fl->proto == IPPROTO_UDP)
569 csum = CSUM_MANGLED_0;
571 if (skb_store_bits(skb, offset, &csum, 2))
572 BUG();
574 send:
575 err = ip6_push_pending_frames(sk);
576 out:
577 return err;
580 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
581 struct flowi *fl, struct rt6_info *rt,
582 unsigned int flags)
584 struct ipv6_pinfo *np = inet6_sk(sk);
585 struct ipv6hdr *iph;
586 struct sk_buff *skb;
587 unsigned int hh_len;
588 int err;
590 if (length > rt->u.dst.dev->mtu) {
591 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
592 return -EMSGSIZE;
594 if (flags&MSG_PROBE)
595 goto out;
597 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
599 skb = sock_alloc_send_skb(sk, length+hh_len+15,
600 flags&MSG_DONTWAIT, &err);
601 if (skb == NULL)
602 goto error;
603 skb_reserve(skb, hh_len);
605 skb->priority = sk->sk_priority;
606 skb->dst = dst_clone(&rt->u.dst);
608 skb_put(skb, length);
609 skb_reset_network_header(skb);
610 iph = ipv6_hdr(skb);
612 skb->ip_summed = CHECKSUM_NONE;
614 skb->transport_header = skb->network_header;
615 err = memcpy_fromiovecend((void *)iph, from, 0, length);
616 if (err)
617 goto error_fault;
619 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
620 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
621 dst_output);
622 if (err > 0)
623 err = np->recverr ? net_xmit_errno(err) : 0;
624 if (err)
625 goto error;
626 out:
627 return 0;
629 error_fault:
630 err = -EFAULT;
631 kfree_skb(skb);
632 error:
633 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
634 return err;
637 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
639 struct iovec *iov;
640 u8 __user *type = NULL;
641 u8 __user *code = NULL;
642 u8 len = 0;
643 int probed = 0;
644 int i;
646 if (!msg->msg_iov)
647 return 0;
649 for (i = 0; i < msg->msg_iovlen; i++) {
650 iov = &msg->msg_iov[i];
651 if (!iov)
652 continue;
654 switch (fl->proto) {
655 case IPPROTO_ICMPV6:
656 /* check if one-byte field is readable or not. */
657 if (iov->iov_base && iov->iov_len < 1)
658 break;
660 if (!type) {
661 type = iov->iov_base;
662 /* check if code field is readable or not. */
663 if (iov->iov_len > 1)
664 code = type + 1;
665 } else if (!code)
666 code = iov->iov_base;
668 if (type && code) {
669 if (get_user(fl->fl_icmp_type, type) ||
670 get_user(fl->fl_icmp_code, code))
671 return -EFAULT;
672 probed = 1;
674 break;
675 case IPPROTO_MH:
676 if (iov->iov_base && iov->iov_len < 1)
677 break;
678 /* check if type field is readable or not. */
679 if (iov->iov_len > 2 - len) {
680 u8 __user *p = iov->iov_base;
681 if (get_user(fl->fl_mh_type, &p[2 - len]))
682 return -EFAULT;
683 probed = 1;
684 } else
685 len += iov->iov_len;
687 break;
688 default:
689 probed = 1;
690 break;
692 if (probed)
693 break;
695 return 0;
698 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
699 struct msghdr *msg, size_t len)
701 struct ipv6_txoptions opt_space;
702 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
703 struct in6_addr *daddr, *final_p = NULL, final;
704 struct inet_sock *inet = inet_sk(sk);
705 struct ipv6_pinfo *np = inet6_sk(sk);
706 struct raw6_sock *rp = raw6_sk(sk);
707 struct ipv6_txoptions *opt = NULL;
708 struct ip6_flowlabel *flowlabel = NULL;
709 struct dst_entry *dst = NULL;
710 struct flowi fl;
711 int addr_len = msg->msg_namelen;
712 int hlimit = -1;
713 int tclass = -1;
714 u16 proto;
715 int err;
717 /* Rough check on arithmetic overflow,
718 better check is made in ip6_append_data().
720 if (len > INT_MAX)
721 return -EMSGSIZE;
723 /* Mirror BSD error message compatibility */
724 if (msg->msg_flags & MSG_OOB)
725 return -EOPNOTSUPP;
728 * Get and verify the address.
730 memset(&fl, 0, sizeof(fl));
732 if (sin6) {
733 if (addr_len < SIN6_LEN_RFC2133)
734 return -EINVAL;
736 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
737 return(-EAFNOSUPPORT);
739 /* port is the proto value [0..255] carried in nexthdr */
740 proto = ntohs(sin6->sin6_port);
742 if (!proto)
743 proto = inet->num;
744 else if (proto != inet->num)
745 return(-EINVAL);
747 if (proto > 255)
748 return(-EINVAL);
750 daddr = &sin6->sin6_addr;
751 if (np->sndflow) {
752 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
753 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
754 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
755 if (flowlabel == NULL)
756 return -EINVAL;
757 daddr = &flowlabel->dst;
762 * Otherwise it will be difficult to maintain
763 * sk->sk_dst_cache.
765 if (sk->sk_state == TCP_ESTABLISHED &&
766 ipv6_addr_equal(daddr, &np->daddr))
767 daddr = &np->daddr;
769 if (addr_len >= sizeof(struct sockaddr_in6) &&
770 sin6->sin6_scope_id &&
771 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
772 fl.oif = sin6->sin6_scope_id;
773 } else {
774 if (sk->sk_state != TCP_ESTABLISHED)
775 return -EDESTADDRREQ;
777 proto = inet->num;
778 daddr = &np->daddr;
779 fl.fl6_flowlabel = np->flow_label;
782 if (ipv6_addr_any(daddr)) {
784 * unspecified destination address
785 * treated as error... is this correct ?
787 fl6_sock_release(flowlabel);
788 return(-EINVAL);
791 if (fl.oif == 0)
792 fl.oif = sk->sk_bound_dev_if;
794 if (msg->msg_controllen) {
795 opt = &opt_space;
796 memset(opt, 0, sizeof(struct ipv6_txoptions));
797 opt->tot_len = sizeof(struct ipv6_txoptions);
799 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
800 if (err < 0) {
801 fl6_sock_release(flowlabel);
802 return err;
804 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
805 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
806 if (flowlabel == NULL)
807 return -EINVAL;
809 if (!(opt->opt_nflen|opt->opt_flen))
810 opt = NULL;
812 if (opt == NULL)
813 opt = np->opt;
814 if (flowlabel)
815 opt = fl6_merge_options(&opt_space, flowlabel, opt);
816 opt = ipv6_fixup_options(&opt_space, opt);
818 fl.proto = proto;
819 err = rawv6_probe_proto_opt(&fl, msg);
820 if (err)
821 goto out;
823 ipv6_addr_copy(&fl.fl6_dst, daddr);
824 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
825 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
827 /* merge ip6_build_xmit from ip6_output */
828 if (opt && opt->srcrt) {
829 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
830 ipv6_addr_copy(&final, &fl.fl6_dst);
831 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
832 final_p = &final;
835 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
836 fl.oif = np->mcast_oif;
837 security_sk_classify_flow(sk, &fl);
839 err = ip6_dst_lookup(sk, &dst, &fl);
840 if (err)
841 goto out;
842 if (final_p)
843 ipv6_addr_copy(&fl.fl6_dst, final_p);
845 if ((err = __xfrm_lookup(&dst, &fl, sk, 1)) < 0) {
846 if (err == -EREMOTE)
847 err = ip6_dst_blackhole(sk, &dst, &fl);
848 if (err < 0)
849 goto out;
852 if (hlimit < 0) {
853 if (ipv6_addr_is_multicast(&fl.fl6_dst))
854 hlimit = np->mcast_hops;
855 else
856 hlimit = np->hop_limit;
857 if (hlimit < 0)
858 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
859 if (hlimit < 0)
860 hlimit = ipv6_get_hoplimit(dst->dev);
863 if (tclass < 0) {
864 tclass = np->tclass;
865 if (tclass < 0)
866 tclass = 0;
869 if (msg->msg_flags&MSG_CONFIRM)
870 goto do_confirm;
872 back_from_confirm:
873 if (inet->hdrincl) {
874 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
875 } else {
876 lock_sock(sk);
877 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
878 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
879 msg->msg_flags);
881 if (err)
882 ip6_flush_pending_frames(sk);
883 else if (!(msg->msg_flags & MSG_MORE))
884 err = rawv6_push_pending_frames(sk, &fl, rp);
885 release_sock(sk);
887 done:
888 dst_release(dst);
889 out:
890 fl6_sock_release(flowlabel);
891 return err<0?err:len;
892 do_confirm:
893 dst_confirm(dst);
894 if (!(msg->msg_flags & MSG_PROBE) || len)
895 goto back_from_confirm;
896 err = 0;
897 goto done;
900 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
901 char __user *optval, int optlen)
903 switch (optname) {
904 case ICMPV6_FILTER:
905 if (optlen > sizeof(struct icmp6_filter))
906 optlen = sizeof(struct icmp6_filter);
907 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
908 return -EFAULT;
909 return 0;
910 default:
911 return -ENOPROTOOPT;
914 return 0;
917 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
918 char __user *optval, int __user *optlen)
920 int len;
922 switch (optname) {
923 case ICMPV6_FILTER:
924 if (get_user(len, optlen))
925 return -EFAULT;
926 if (len < 0)
927 return -EINVAL;
928 if (len > sizeof(struct icmp6_filter))
929 len = sizeof(struct icmp6_filter);
930 if (put_user(len, optlen))
931 return -EFAULT;
932 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
933 return -EFAULT;
934 return 0;
935 default:
936 return -ENOPROTOOPT;
939 return 0;
943 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
944 char __user *optval, int optlen)
946 struct raw6_sock *rp = raw6_sk(sk);
947 int val;
949 if (get_user(val, (int __user *)optval))
950 return -EFAULT;
952 switch (optname) {
953 case IPV6_CHECKSUM:
954 /* You may get strange result with a positive odd offset;
955 RFC2292bis agrees with me. */
956 if (val > 0 && (val&1))
957 return(-EINVAL);
958 if (val < 0) {
959 rp->checksum = 0;
960 } else {
961 rp->checksum = 1;
962 rp->offset = val;
965 return 0;
966 break;
968 default:
969 return(-ENOPROTOOPT);
973 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
974 char __user *optval, int optlen)
976 switch(level) {
977 case SOL_RAW:
978 break;
980 case SOL_ICMPV6:
981 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
982 return -EOPNOTSUPP;
983 return rawv6_seticmpfilter(sk, level, optname, optval,
984 optlen);
985 case SOL_IPV6:
986 if (optname == IPV6_CHECKSUM)
987 break;
988 default:
989 return ipv6_setsockopt(sk, level, optname, optval,
990 optlen);
993 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
996 #ifdef CONFIG_COMPAT
997 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
998 char __user *optval, int optlen)
1000 switch (level) {
1001 case SOL_RAW:
1002 break;
1003 case SOL_ICMPV6:
1004 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1005 return -EOPNOTSUPP;
1006 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1007 case SOL_IPV6:
1008 if (optname == IPV6_CHECKSUM)
1009 break;
1010 default:
1011 return compat_ipv6_setsockopt(sk, level, optname,
1012 optval, optlen);
1014 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1016 #endif
1018 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1019 char __user *optval, int __user *optlen)
1021 struct raw6_sock *rp = raw6_sk(sk);
1022 int val, len;
1024 if (get_user(len,optlen))
1025 return -EFAULT;
1027 switch (optname) {
1028 case IPV6_CHECKSUM:
1029 if (rp->checksum == 0)
1030 val = -1;
1031 else
1032 val = rp->offset;
1033 break;
1035 default:
1036 return -ENOPROTOOPT;
1039 len = min_t(unsigned int, sizeof(int), len);
1041 if (put_user(len, optlen))
1042 return -EFAULT;
1043 if (copy_to_user(optval,&val,len))
1044 return -EFAULT;
1045 return 0;
1048 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1049 char __user *optval, int __user *optlen)
1051 switch(level) {
1052 case SOL_RAW:
1053 break;
1055 case SOL_ICMPV6:
1056 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1057 return -EOPNOTSUPP;
1058 return rawv6_geticmpfilter(sk, level, optname, optval,
1059 optlen);
1060 case SOL_IPV6:
1061 if (optname == IPV6_CHECKSUM)
1062 break;
1063 default:
1064 return ipv6_getsockopt(sk, level, optname, optval,
1065 optlen);
1068 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1071 #ifdef CONFIG_COMPAT
1072 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1073 char __user *optval, int __user *optlen)
1075 switch (level) {
1076 case SOL_RAW:
1077 break;
1078 case SOL_ICMPV6:
1079 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1080 return -EOPNOTSUPP;
1081 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1082 case SOL_IPV6:
1083 if (optname == IPV6_CHECKSUM)
1084 break;
1085 default:
1086 return compat_ipv6_getsockopt(sk, level, optname,
1087 optval, optlen);
1089 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1091 #endif
1093 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1095 switch(cmd) {
1096 case SIOCOUTQ:
1098 int amount = atomic_read(&sk->sk_wmem_alloc);
1099 return put_user(amount, (int __user *)arg);
1101 case SIOCINQ:
1103 struct sk_buff *skb;
1104 int amount = 0;
1106 spin_lock_bh(&sk->sk_receive_queue.lock);
1107 skb = skb_peek(&sk->sk_receive_queue);
1108 if (skb != NULL)
1109 amount = skb->tail - skb->transport_header;
1110 spin_unlock_bh(&sk->sk_receive_queue.lock);
1111 return put_user(amount, (int __user *)arg);
1114 default:
1115 return -ENOIOCTLCMD;
1119 static void rawv6_close(struct sock *sk, long timeout)
1121 if (inet_sk(sk)->num == IPPROTO_RAW)
1122 ip6_ra_control(sk, -1, NULL);
1124 sk_common_release(sk);
1127 static int rawv6_init_sk(struct sock *sk)
1129 struct raw6_sock *rp = raw6_sk(sk);
1131 switch (inet_sk(sk)->num) {
1132 case IPPROTO_ICMPV6:
1133 rp->checksum = 1;
1134 rp->offset = 2;
1135 break;
1136 case IPPROTO_MH:
1137 rp->checksum = 1;
1138 rp->offset = 4;
1139 break;
1140 default:
1141 break;
1143 return(0);
1146 struct proto rawv6_prot = {
1147 .name = "RAWv6",
1148 .owner = THIS_MODULE,
1149 .close = rawv6_close,
1150 .connect = ip6_datagram_connect,
1151 .disconnect = udp_disconnect,
1152 .ioctl = rawv6_ioctl,
1153 .init = rawv6_init_sk,
1154 .destroy = inet6_destroy_sock,
1155 .setsockopt = rawv6_setsockopt,
1156 .getsockopt = rawv6_getsockopt,
1157 .sendmsg = rawv6_sendmsg,
1158 .recvmsg = rawv6_recvmsg,
1159 .bind = rawv6_bind,
1160 .backlog_rcv = rawv6_rcv_skb,
1161 .hash = raw_v6_hash,
1162 .unhash = raw_v6_unhash,
1163 .obj_size = sizeof(struct raw6_sock),
1164 #ifdef CONFIG_COMPAT
1165 .compat_setsockopt = compat_rawv6_setsockopt,
1166 .compat_getsockopt = compat_rawv6_getsockopt,
1167 #endif
1170 #ifdef CONFIG_PROC_FS
1171 struct raw6_iter_state {
1172 int bucket;
1175 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1177 static struct sock *raw6_get_first(struct seq_file *seq)
1179 struct sock *sk;
1180 struct hlist_node *node;
1181 struct raw6_iter_state* state = raw6_seq_private(seq);
1183 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1184 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1185 if (sk->sk_family == PF_INET6)
1186 goto out;
1187 sk = NULL;
1188 out:
1189 return sk;
1192 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1194 struct raw6_iter_state* state = raw6_seq_private(seq);
1196 do {
1197 sk = sk_next(sk);
1198 try_again:
1200 } while (sk && sk->sk_family != PF_INET6);
1202 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1203 sk = sk_head(&raw_v6_htable[state->bucket]);
1204 goto try_again;
1206 return sk;
1209 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1211 struct sock *sk = raw6_get_first(seq);
1212 if (sk)
1213 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1214 --pos;
1215 return pos ? NULL : sk;
1218 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1220 read_lock(&raw_v6_lock);
1221 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1224 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1226 struct sock *sk;
1228 if (v == SEQ_START_TOKEN)
1229 sk = raw6_get_first(seq);
1230 else
1231 sk = raw6_get_next(seq, v);
1232 ++*pos;
1233 return sk;
1236 static void raw6_seq_stop(struct seq_file *seq, void *v)
1238 read_unlock(&raw_v6_lock);
1241 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1243 struct ipv6_pinfo *np = inet6_sk(sp);
1244 struct in6_addr *dest, *src;
1245 __u16 destp, srcp;
1247 dest = &np->daddr;
1248 src = &np->rcv_saddr;
1249 destp = 0;
1250 srcp = inet_sk(sp)->num;
1251 seq_printf(seq,
1252 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1253 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1255 src->s6_addr32[0], src->s6_addr32[1],
1256 src->s6_addr32[2], src->s6_addr32[3], srcp,
1257 dest->s6_addr32[0], dest->s6_addr32[1],
1258 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1259 sp->sk_state,
1260 atomic_read(&sp->sk_wmem_alloc),
1261 atomic_read(&sp->sk_rmem_alloc),
1262 0, 0L, 0,
1263 sock_i_uid(sp), 0,
1264 sock_i_ino(sp),
1265 atomic_read(&sp->sk_refcnt), sp);
1268 static int raw6_seq_show(struct seq_file *seq, void *v)
1270 if (v == SEQ_START_TOKEN)
1271 seq_printf(seq,
1272 " sl "
1273 "local_address "
1274 "remote_address "
1275 "st tx_queue rx_queue tr tm->when retrnsmt"
1276 " uid timeout inode\n");
1277 else
1278 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1279 return 0;
1282 static const struct seq_operations raw6_seq_ops = {
1283 .start = raw6_seq_start,
1284 .next = raw6_seq_next,
1285 .stop = raw6_seq_stop,
1286 .show = raw6_seq_show,
1289 static int raw6_seq_open(struct inode *inode, struct file *file)
1291 struct seq_file *seq;
1292 int rc = -ENOMEM;
1293 struct raw6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
1294 if (!s)
1295 goto out;
1296 rc = seq_open(file, &raw6_seq_ops);
1297 if (rc)
1298 goto out_kfree;
1299 seq = file->private_data;
1300 seq->private = s;
1301 out:
1302 return rc;
1303 out_kfree:
1304 kfree(s);
1305 goto out;
1308 static const struct file_operations raw6_seq_fops = {
1309 .owner = THIS_MODULE,
1310 .open = raw6_seq_open,
1311 .read = seq_read,
1312 .llseek = seq_lseek,
1313 .release = seq_release_private,
1316 int __init raw6_proc_init(void)
1318 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1319 return -ENOMEM;
1320 return 0;
1323 void raw6_proc_exit(void)
1325 proc_net_remove("raw6");
1327 #endif /* CONFIG_PROC_FS */