?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
RT-AC66 3.0.0.4.374.130 core
[tomato.git] / release / src-rt-6.x / linux / linux-2.6 / net / ipv6 / raw.c
blobde2418f23693eadd0ea284f6b78bd2c052334dc6
1 /*
2 * RAW sockets for IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Adapted from linux/net/ipv4/raw.c
9 *
10 * $Id: raw.c,v 1.51 2002/02/01 22:01:04 davem Exp $
11 *
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
16 *
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.
21 */
22
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>
37
38 #include <net/ip.h>
39 #include <net/sock.h>
40 #include <net/snmp.h>
41
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 #ifdef CONFIG_IPV6_MIP6
53 #include <net/mip6.h>
54 #endif
55 #include <linux/mroute6.h>
56
57 #include <net/rawv6.h>
58 #include <net/xfrm.h>
59
60 #include <linux/proc_fs.h>
61 #include <linux/seq_file.h>
62
63 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
64 DEFINE_RWLOCK(raw_v6_lock);
65
66 static void raw_v6_hash(struct sock *sk)
67 {
68 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num &
69 (RAWV6_HTABLE_SIZE - 1)];
70
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);
75 }
76
77 static void raw_v6_unhash(struct sock *sk)
78 {
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);
83 }
84
85
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)
90 {
91 struct hlist_node *node;
92 int is_multicast = ipv6_addr_is_multicast(loc_addr);
93
94 sk_for_each_from(sk, node)
95 if (inet_sk(sk)->num == num) {
96 struct ipv6_pinfo *np = inet6_sk(sk);
97
98 if (!ipv6_addr_any(&np->daddr) &&
99 !ipv6_addr_equal(&np->daddr, rmt_addr))
100 continue;
101
102 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
103 continue;
104
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;
112 }
113 goto found;
114 }
115 sk = NULL;
116 found:
117 return sk;
118 }
119
120 /*
121 * 0 - deliver
122 * 1 - block
123 */
124 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
125 {
126 struct icmp6hdr *icmph;
127 struct raw6_sock *rp = raw6_sk(sk);
128
129 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
130 __u32 *data = &rp->filter.data[0];
131 int bit_nr;
132
133 icmph = (struct icmp6hdr *) skb->data;
134 bit_nr = icmph->icmp6_type;
135
136 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
137 }
138 return 0;
139 }
140
141 /*
142 * demultiplex raw sockets.
143 * (should consider queueing the skb in the sock receive_queue
144 * without calling rawv6.c)
145 *
146 * Caller owns SKB so we must make clones.
147 */
148 int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
149 {
150 struct in6_addr *saddr;
151 struct in6_addr *daddr;
152 struct sock *sk;
153 int delivered = 0;
154 __u8 hash;
155
156 saddr = &ipv6_hdr(skb)->saddr;
157 daddr = saddr + 1;
158
159 hash = nexthdr & (MAX_INET_PROTOS - 1);
160
161 read_lock(&raw_v6_lock);
162 sk = sk_head(&raw_v6_htable[hash]);
163
164 /*
165 * The first socket found will be delivered after
166 * delivery to transport protocols.
167 */
168
169 if (sk == NULL)
170 goto out;
171
172 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
173
174 while (sk) {
175 int filtered;
176
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.
189 */
190 filtered = mip6_mh_filter(sk, skb);
191 break;
192 #endif
193 default:
194 filtered = 0;
195 break;
196 }
197
198 if (filtered < 0)
199 break;
200 if (filtered == 0) {
201 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
202
203 /* Not releasing hash table! */
204 if (clone) {
205 nf_reset(clone);
206 rawv6_rcv(sk, clone);
207 }
208 }
209 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr,
210 IP6CB(skb)->iif);
211 }
212 out:
213 read_unlock(&raw_v6_lock);
214 return delivered;
215 }
216
217 /* This cleans up af_inet6 a bit. -DaveM */
218 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
219 {
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;
226
227 if (addr_len < SIN6_LEN_RFC2133)
228 return -EINVAL;
229 addr_type = ipv6_addr_type(&addr->sin6_addr);
230
231 /* Raw sockets are IPv6 only */
232 if (addr_type == IPV6_ADDR_MAPPED)
233 return(-EADDRNOTAVAIL);
234
235 lock_sock(sk);
236
237 err = -EINVAL;
238 if (sk->sk_state != TCP_CLOSE)
239 goto out;
240
241 /* Check if the address belongs to the host. */
242 if (addr_type != IPV6_ADDR_ANY) {
243 struct net_device *dev = NULL;
244
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.
250 */
251 sk->sk_bound_dev_if = addr->sin6_scope_id;
252 }
253
254 /* Binding to link-local address requires an interface */
255 if (!sk->sk_bound_dev_if)
256 goto out;
257
258 dev = dev_get_by_index(sk->sk_bound_dev_if);
259 if (!dev) {
260 err = -ENODEV;
261 goto out;
262 }
263 }
264
265 /* ipv4 addr of the socket is invalid. Only the
266 * unspecified and mapped address have a v4 equivalent.
267 */
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;
275 }
276 }
277 if (dev)
278 dev_put(dev);
279 }
280
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;
289 }
290
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)
294 {
295 struct inet_sock *inet = inet_sk(sk);
296 struct ipv6_pinfo *np = inet6_sk(sk);
297 int err;
298 int harderr;
299
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.
304 */
305 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
306 return;
307
308 harderr = icmpv6_err_convert(type, code, &err);
309 if (type == ICMPV6_PKT_TOOBIG)
310 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
311
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);
317 }
318
319 if (np->recverr || harderr) {
320 sk->sk_err = err;
321 sk->sk_error_report(sk);
322 }
323 }
324
325 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
326 {
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;
332 }
333
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;
339 }
340
341 return 0;
342 }
343
344 /*
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
350 */
351 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
352 {
353 struct inet_sock *inet = inet_sk(sk);
354 struct raw6_sock *rp = raw6_sk(sk);
355
356 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
357 kfree_skb(skb);
358 return NET_RX_DROP;
359 }
360
361 if (!rp->checksum)
362 skb->ip_summed = CHECKSUM_UNNECESSARY;
363
364 if (skb->ip_summed == CHECKSUM_COMPLETE) {
365 skb_postpull_rcsum(skb, skb_network_header(skb),
366 skb_network_header_len(skb));
367 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
368 &ipv6_hdr(skb)->daddr,
369 skb->len, inet->num, skb->csum))
370 skb->ip_summed = CHECKSUM_UNNECESSARY;
371 }
372 if (!skb_csum_unnecessary(skb))
373 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
374 &ipv6_hdr(skb)->daddr,
375 skb->len,
376 inet->num, 0));
377
378 if (inet->hdrincl) {
379 if (skb_checksum_complete(skb)) {
380 /* FIXME: increment a raw6 drops counter here */
381 kfree_skb(skb);
382 return 0;
383 }
384 }
385
386 rawv6_rcv_skb(sk, skb);
387 return 0;
388 }
389
390
391 /*
392 * This should be easy, if there is something there
393 * we return it, otherwise we block.
394 */
395
396 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
397 struct msghdr *msg, size_t len,
398 int noblock, int flags, int *addr_len)
399 {
400 struct ipv6_pinfo *np = inet6_sk(sk);
401 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
402 struct sk_buff *skb;
403 size_t copied;
404 int err;
405
406 if (flags & MSG_OOB)
407 return -EOPNOTSUPP;
408
409 if (addr_len)
410 *addr_len=sizeof(*sin6);
411
412 if (flags & MSG_ERRQUEUE)
413 return ipv6_recv_error(sk, msg, len);
414
415 skb = skb_recv_datagram(sk, flags, noblock, &err);
416 if (!skb)
417 goto out;
418
419 copied = skb->len;
420 if (copied > len) {
421 copied = len;
422 msg->msg_flags |= MSG_TRUNC;
423 }
424
425 if (skb_csum_unnecessary(skb)) {
426 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
427 } else if (msg->msg_flags&MSG_TRUNC) {
428 if (__skb_checksum_complete(skb))
429 goto csum_copy_err;
430 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
431 } else {
432 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
433 if (err == -EINVAL)
434 goto csum_copy_err;
435 }
436 if (err)
437 goto out_free;
438
439 /* Copy the address. */
440 if (sin6) {
441 sin6->sin6_family = AF_INET6;
442 sin6->sin6_port = 0;
443 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
444 sin6->sin6_flowinfo = 0;
445 sin6->sin6_scope_id = 0;
446 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
447 sin6->sin6_scope_id = IP6CB(skb)->iif;
448 }
449
450 sock_recv_timestamp(msg, sk, skb);
451
452 if (np->rxopt.all)
453 datagram_recv_ctl(sk, msg, skb);
454
455 err = copied;
456 if (flags & MSG_TRUNC)
457 err = skb->len;
458
459 out_free:
460 skb_free_datagram(sk, skb);
461 out:
462 return err;
463
464 csum_copy_err:
465 skb_kill_datagram(sk, skb, flags);
466
467 /* Error for blocking case is chosen to masquerade
468 as some normal condition.
469 */
470 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
471 /* FIXME: increment a raw6 drops counter here */
472 goto out;
473 }
474
475 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
476 struct raw6_sock *rp)
477 {
478 struct sk_buff *skb;
479 int err = 0;
480 int offset;
481 int len;
482 int total_len;
483 __wsum tmp_csum;
484 __sum16 csum;
485
486 if (!rp->checksum)
487 goto send;
488
489 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
490 goto out;
491
492 offset = rp->offset;
493 total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
494 skb->data);
495 if (offset >= total_len - 1) {
496 err = -EINVAL;
497 ip6_flush_pending_frames(sk);
498 goto out;
499 }
500
501 /* should be check HW csum miyazawa */
502 if (skb_queue_len(&sk->sk_write_queue) == 1) {
503 /*
504 * Only one fragment on the socket.
505 */
506 tmp_csum = skb->csum;
507 } else {
508 struct sk_buff *csum_skb = NULL;
509 tmp_csum = 0;
510
511 skb_queue_walk(&sk->sk_write_queue, skb) {
512 tmp_csum = csum_add(tmp_csum, skb->csum);
513
514 if (csum_skb)
515 continue;
516
517 len = skb->len - skb_transport_offset(skb);
518 if (offset >= len) {
519 offset -= len;
520 continue;
521 }
522
523 csum_skb = skb;
524 }
525
526 skb = csum_skb;
527 }
528
529 offset += skb_transport_offset(skb);
530 if (skb_copy_bits(skb, offset, &csum, 2))
531 BUG();
532
533 /* in case cksum was not initialized */
534 if (unlikely(csum))
535 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
536
537 csum = csum_ipv6_magic(&fl->fl6_src,
538 &fl->fl6_dst,
539 total_len, fl->proto, tmp_csum);
540
541 if (csum == 0 && fl->proto == IPPROTO_UDP)
542 csum = CSUM_MANGLED_0;
543
544 if (skb_store_bits(skb, offset, &csum, 2))
545 BUG();
546
547 send:
548 err = ip6_push_pending_frames(sk);
549 out:
550 return err;
551 }
552
553 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
554 struct flowi *fl, struct rt6_info *rt,
555 unsigned int flags)
556 {
557 struct ipv6_pinfo *np = inet6_sk(sk);
558 struct ipv6hdr *iph;
559 struct sk_buff *skb;
560 unsigned int hh_len;
561 int err;
562
563 if (length > rt->u.dst.dev->mtu) {
564 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
565 return -EMSGSIZE;
566 }
567 if (flags&MSG_PROBE)
568 goto out;
569
570 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
571
572 skb = sock_alloc_send_skb(sk, length+hh_len+15,
573 flags&MSG_DONTWAIT, &err);
574 if (skb == NULL)
575 goto error;
576 skb_reserve(skb, hh_len);
577
578 skb->priority = sk->sk_priority;
579 skb->dst = dst_clone(&rt->u.dst);
580
581 skb_put(skb, length);
582 skb_reset_network_header(skb);
583 iph = ipv6_hdr(skb);
584
585 skb->ip_summed = CHECKSUM_NONE;
586
587 skb->transport_header = skb->network_header;
588 err = memcpy_fromiovecend((void *)iph, from, 0, length);
589 if (err)
590 goto error_fault;
591
592 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
593 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
594 dst_output);
595 if (err > 0)
596 err = np->recverr ? net_xmit_errno(err) : 0;
597 if (err)
598 goto error;
599 out:
600 return 0;
601
602 error_fault:
603 err = -EFAULT;
604 kfree_skb(skb);
605 error:
606 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
607 return err;
608 }
609
610 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
611 {
612 struct iovec *iov;
613 u8 __user *type = NULL;
614 u8 __user *code = NULL;
615 #ifdef CONFIG_IPV6_MIP6
616 u8 len = 0;
617 #endif
618 int probed = 0;
619 int i;
620
621 if (!msg->msg_iov)
622 return 0;
623
624 for (i = 0; i < msg->msg_iovlen; i++) {
625 iov = &msg->msg_iov[i];
626 if (!iov)
627 continue;
628
629 switch (fl->proto) {
630 case IPPROTO_ICMPV6:
631 /* check if one-byte field is readable or not. */
632 if (iov->iov_base && iov->iov_len < 1)
633 break;
634
635 if (!type) {
636 type = iov->iov_base;
637 /* check if code field is readable or not. */
638 if (iov->iov_len > 1)
639 code = type + 1;
640 } else if (!code)
641 code = iov->iov_base;
642
643 if (type && code) {
644 if (get_user(fl->fl_icmp_type, type) ||
645 get_user(fl->fl_icmp_code, code))
646 return -EFAULT;
647 probed = 1;
648 }
649 break;
650 #ifdef CONFIG_IPV6_MIP6
651 case IPPROTO_MH:
652 if (iov->iov_base && iov->iov_len < 1)
653 break;
654 /* check if type field is readable or not. */
655 if (iov->iov_len > 2 - len) {
656 u8 __user *p = iov->iov_base;
657 if (get_user(fl->fl_mh_type, &p[2 - len]))
658 return -EFAULT;
659 probed = 1;
660 } else
661 len += iov->iov_len;
662
663 break;
664 #endif
665 default:
666 probed = 1;
667 break;
668 }
669 if (probed)
670 break;
671 }
672 return 0;
673 }
674
675 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
676 struct msghdr *msg, size_t len)
677 {
678 struct ipv6_txoptions opt_space;
679 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
680 struct in6_addr *daddr, *final_p = NULL, final;
681 struct inet_sock *inet = inet_sk(sk);
682 struct ipv6_pinfo *np = inet6_sk(sk);
683 struct raw6_sock *rp = raw6_sk(sk);
684 struct ipv6_txoptions *opt = NULL;
685 struct ip6_flowlabel *flowlabel = NULL;
686 struct dst_entry *dst = NULL;
687 struct flowi fl;
688 int addr_len = msg->msg_namelen;
689 int hlimit = -1;
690 int tclass = -1;
691 u16 proto;
692 int err;
693
694 /* Rough check on arithmetic overflow,
695 better check is made in ip6_append_data().
696 */
697 if (len > INT_MAX)
698 return -EMSGSIZE;
699
700 /* Mirror BSD error message compatibility */
701 if (msg->msg_flags & MSG_OOB)
702 return -EOPNOTSUPP;
703
704 /*
705 * Get and verify the address.
706 */
707 memset(&fl, 0, sizeof(fl));
708
709 if (sin6) {
710 if (addr_len < SIN6_LEN_RFC2133)
711 return -EINVAL;
712
713 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
714 return(-EAFNOSUPPORT);
715
716 /* port is the proto value [0..255] carried in nexthdr */
717 proto = ntohs(sin6->sin6_port);
718
719 if (!proto)
720 proto = inet->num;
721 else if (proto != inet->num)
722 return(-EINVAL);
723
724 if (proto > 255)
725 return(-EINVAL);
726
727 daddr = &sin6->sin6_addr;
728 if (np->sndflow) {
729 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
730 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
731 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
732 if (flowlabel == NULL)
733 return -EINVAL;
734 daddr = &flowlabel->dst;
735 }
736 }
737
738 /*
739 * Otherwise it will be difficult to maintain
740 * sk->sk_dst_cache.
741 */
742 if (sk->sk_state == TCP_ESTABLISHED &&
743 ipv6_addr_equal(daddr, &np->daddr))
744 daddr = &np->daddr;
745
746 if (addr_len >= sizeof(struct sockaddr_in6) &&
747 sin6->sin6_scope_id &&
748 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
749 fl.oif = sin6->sin6_scope_id;
750 } else {
751 if (sk->sk_state != TCP_ESTABLISHED)
752 return -EDESTADDRREQ;
753
754 proto = inet->num;
755 daddr = &np->daddr;
756 fl.fl6_flowlabel = np->flow_label;
757 }
758
759 if (ipv6_addr_any(daddr)) {
760 /*
761 * unspecified destination address
762 * treated as error... is this correct ?
763 */
764 fl6_sock_release(flowlabel);
765 return(-EINVAL);
766 }
767
768 if (fl.oif == 0)
769 fl.oif = sk->sk_bound_dev_if;
770
771 if (msg->msg_controllen) {
772 opt = &opt_space;
773 memset(opt, 0, sizeof(struct ipv6_txoptions));
774 opt->tot_len = sizeof(struct ipv6_txoptions);
775
776 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
777 if (err < 0) {
778 fl6_sock_release(flowlabel);
779 return err;
780 }
781 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
782 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
783 if (flowlabel == NULL)
784 return -EINVAL;
785 }
786 if (!(opt->opt_nflen|opt->opt_flen))
787 opt = NULL;
788 }
789 if (opt == NULL)
790 opt = np->opt;
791 if (flowlabel)
792 opt = fl6_merge_options(&opt_space, flowlabel, opt);
793 opt = ipv6_fixup_options(&opt_space, opt);
794
795 fl.proto = proto;
796 err = rawv6_probe_proto_opt(&fl, msg);
797 if (err)
798 goto out;
799
800 ipv6_addr_copy(&fl.fl6_dst, daddr);
801 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
802 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
803
804 /* merge ip6_build_xmit from ip6_output */
805 if (opt && opt->srcrt) {
806 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
807 ipv6_addr_copy(&final, &fl.fl6_dst);
808 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
809 final_p = &final;
810 }
811
812 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
813 fl.oif = np->mcast_oif;
814 security_sk_classify_flow(sk, &fl);
815
816 err = ip6_dst_lookup(sk, &dst, &fl);
817 if (err)
818 goto out;
819 if (final_p)
820 ipv6_addr_copy(&fl.fl6_dst, final_p);
821
822 if ((err = __xfrm_lookup(&dst, &fl, sk, 1)) < 0) {
823 if (err == -EREMOTE)
824 err = ip6_dst_blackhole(sk, &dst, &fl);
825 if (err < 0)
826 goto out;
827 }
828
829 if (hlimit < 0) {
830 if (ipv6_addr_is_multicast(&fl.fl6_dst))
831 hlimit = np->mcast_hops;
832 else
833 hlimit = np->hop_limit;
834 if (hlimit < 0)
835 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
836 if (hlimit < 0)
837 hlimit = ipv6_get_hoplimit(dst->dev);
838 }
839
840 if (tclass < 0) {
841 tclass = np->tclass;
842 if (tclass < 0)
843 tclass = 0;
844 }
845
846 if (msg->msg_flags&MSG_CONFIRM)
847 goto do_confirm;
848
849 back_from_confirm:
850 if (inet->hdrincl) {
851 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
852 } else {
853 lock_sock(sk);
854 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
855 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
856 msg->msg_flags);
857
858 if (err)
859 ip6_flush_pending_frames(sk);
860 else if (!(msg->msg_flags & MSG_MORE))
861 err = rawv6_push_pending_frames(sk, &fl, rp);
862 release_sock(sk);
863 }
864 done:
865 dst_release(dst);
866 out:
867 fl6_sock_release(flowlabel);
868 return err<0?err:len;
869 do_confirm:
870 dst_confirm(dst);
871 if (!(msg->msg_flags & MSG_PROBE) || len)
872 goto back_from_confirm;
873 err = 0;
874 goto done;
875 }
876
877 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
878 char __user *optval, int optlen)
879 {
880 switch (optname) {
881 case ICMPV6_FILTER:
882 if (optlen > sizeof(struct icmp6_filter))
883 optlen = sizeof(struct icmp6_filter);
884 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
885 return -EFAULT;
886 return 0;
887 default:
888 return -ENOPROTOOPT;
889 }
890
891 return 0;
892 }
893
894 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
895 char __user *optval, int __user *optlen)
896 {
897 int len;
898
899 switch (optname) {
900 case ICMPV6_FILTER:
901 if (get_user(len, optlen))
902 return -EFAULT;
903 if (len < 0)
904 return -EINVAL;
905 if (len > sizeof(struct icmp6_filter))
906 len = sizeof(struct icmp6_filter);
907 if (put_user(len, optlen))
908 return -EFAULT;
909 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
910 return -EFAULT;
911 return 0;
912 default:
913 return -ENOPROTOOPT;
914 }
915
916 return 0;
917 }
918
919
920 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
921 char __user *optval, int optlen)
922 {
923 struct raw6_sock *rp = raw6_sk(sk);
924 int val;
925
926 if (get_user(val, (int __user *)optval))
927 return -EFAULT;
928
929 switch (optname) {
930 case IPV6_CHECKSUM:
931 /* You may get strange result with a positive odd offset;
932 RFC2292bis agrees with me. */
933 if (val > 0 && (val&1))
934 return(-EINVAL);
935 if (val < 0) {
936 rp->checksum = 0;
937 } else {
938 rp->checksum = 1;
939 rp->offset = val;
940 }
941
942 return 0;
943 break;
944
945 default:
946 return(-ENOPROTOOPT);
947 }
948 }
949
950 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
951 char __user *optval, int optlen)
952 {
953 switch(level) {
954 case SOL_RAW:
955 break;
956
957 case SOL_ICMPV6:
958 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
959 return -EOPNOTSUPP;
960 return rawv6_seticmpfilter(sk, level, optname, optval,
961 optlen);
962 case SOL_IPV6:
963 if (optname == IPV6_CHECKSUM)
964 break;
965 default:
966 return ipv6_setsockopt(sk, level, optname, optval,
967 optlen);
968 }
969
970 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
971 }
972
973 #ifdef CONFIG_COMPAT
974 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
975 char __user *optval, int optlen)
976 {
977 switch (level) {
978 case SOL_RAW:
979 break;
980 case SOL_ICMPV6:
981 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
982 return -EOPNOTSUPP;
983 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
984 case SOL_IPV6:
985 if (optname == IPV6_CHECKSUM)
986 break;
987 default:
988 return compat_ipv6_setsockopt(sk, level, optname,
989 optval, optlen);
990 }
991 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
992 }
993 #endif
994
995 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
996 char __user *optval, int __user *optlen)
997 {
998 struct raw6_sock *rp = raw6_sk(sk);
999 int val, len;
1000
1001 if (get_user(len,optlen))
1002 return -EFAULT;
1003
1004 switch (optname) {
1005 case IPV6_CHECKSUM:
1006 if (rp->checksum == 0)
1007 val = -1;
1008 else
1009 val = rp->offset;
1010 break;
1011
1012 default:
1013 return -ENOPROTOOPT;
1014 }
1015
1016 len = min_t(unsigned int, sizeof(int), len);
1017
1018 if (put_user(len, optlen))
1019 return -EFAULT;
1020 if (copy_to_user(optval,&val,len))
1021 return -EFAULT;
1022 return 0;
1023 }
1024
1025 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1026 char __user *optval, int __user *optlen)
1027 {
1028 switch(level) {
1029 case SOL_RAW:
1030 break;
1031
1032 case SOL_ICMPV6:
1033 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1034 return -EOPNOTSUPP;
1035 return rawv6_geticmpfilter(sk, level, optname, optval,
1036 optlen);
1037 case SOL_IPV6:
1038 if (optname == IPV6_CHECKSUM)
1039 break;
1040 default:
1041 return ipv6_getsockopt(sk, level, optname, optval,
1042 optlen);
1043 }
1044
1045 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1046 }
1047
1048 #ifdef CONFIG_COMPAT
1049 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1050 char __user *optval, int __user *optlen)
1051 {
1052 switch (level) {
1053 case SOL_RAW:
1054 break;
1055 case SOL_ICMPV6:
1056 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1057 return -EOPNOTSUPP;
1058 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1059 case SOL_IPV6:
1060 if (optname == IPV6_CHECKSUM)
1061 break;
1062 default:
1063 return compat_ipv6_getsockopt(sk, level, optname,
1064 optval, optlen);
1065 }
1066 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1067 }
1068 #endif
1069
1070 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1071 {
1072 switch(cmd) {
1073 case SIOCOUTQ:
1074 {
1075 int amount = sk_wmem_alloc_get(sk);
1076
1077 return put_user(amount, (int __user *)arg);
1078 }
1079 case SIOCINQ:
1080 {
1081 struct sk_buff *skb;
1082 int amount = 0;
1083
1084 spin_lock_bh(&sk->sk_receive_queue.lock);
1085 skb = skb_peek(&sk->sk_receive_queue);
1086 if (skb != NULL)
1087 amount = skb->tail - skb->transport_header;
1088 spin_unlock_bh(&sk->sk_receive_queue.lock);
1089 return put_user(amount, (int __user *)arg);
1090 }
1091
1092 default:
1093 #ifdef CONFIG_IPV6_MROUTE
1094 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1095 #else
1096 return -ENOIOCTLCMD;
1097 #endif
1098 }
1099 }
1100
1101 static void rawv6_close(struct sock *sk, long timeout)
1102 {
1103 if (inet_sk(sk)->num == IPPROTO_RAW)
1104 ip6_ra_control(sk, -1, NULL);
1105 ip6mr_sk_done(sk);
1106 sk_common_release(sk);
1107 }
1108
1109 static int raw6_destroy(struct sock *sk)
1110 {
1111 lock_sock(sk);
1112 ip6_flush_pending_frames(sk);
1113 release_sock(sk);
1114 return 0;
1115 }
1116
1117 static int rawv6_init_sk(struct sock *sk)
1118 {
1119 struct raw6_sock *rp = raw6_sk(sk);
1120
1121 switch (inet_sk(sk)->num) {
1122 case IPPROTO_ICMPV6:
1123 rp->checksum = 1;
1124 rp->offset = 2;
1125 break;
1126 case IPPROTO_MH:
1127 rp->checksum = 1;
1128 rp->offset = 4;
1129 break;
1130 default:
1131 break;
1132 }
1133 return(0);
1134 }
1135
1136 struct proto rawv6_prot = {
1137 .name = "RAWv6",
1138 .owner = THIS_MODULE,
1139 .close = rawv6_close,
1140 .destroy = raw6_destroy,
1141 .connect = ip6_datagram_connect,
1142 .disconnect = udp_disconnect,
1143 .ioctl = rawv6_ioctl,
1144 .init = rawv6_init_sk,
1145 .destroy = inet6_destroy_sock,
1146 .setsockopt = rawv6_setsockopt,
1147 .getsockopt = rawv6_getsockopt,
1148 .sendmsg = rawv6_sendmsg,
1149 .recvmsg = rawv6_recvmsg,
1150 .bind = rawv6_bind,
1151 .backlog_rcv = rawv6_rcv_skb,
1152 .hash = raw_v6_hash,
1153 .unhash = raw_v6_unhash,
1154 .obj_size = sizeof(struct raw6_sock),
1155 #ifdef CONFIG_COMPAT
1156 .compat_setsockopt = compat_rawv6_setsockopt,
1157 .compat_getsockopt = compat_rawv6_getsockopt,
1158 #endif
1159 };
1160
1161 #ifdef CONFIG_PROC_FS
1162 struct raw6_iter_state {
1163 int bucket;
1164 };
1165
1166 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1167
1168 static struct sock *raw6_get_first(struct seq_file *seq)
1169 {
1170 struct sock *sk;
1171 struct hlist_node *node;
1172 struct raw6_iter_state* state = raw6_seq_private(seq);
1173
1174 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1175 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1176 if (sk->sk_family == PF_INET6)
1177 goto out;
1178 sk = NULL;
1179 out:
1180 return sk;
1181 }
1182
1183 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1184 {
1185 struct raw6_iter_state* state = raw6_seq_private(seq);
1186
1187 do {
1188 sk = sk_next(sk);
1189 try_again:
1190 ;
1191 } while (sk && sk->sk_family != PF_INET6);
1192
1193 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1194 sk = sk_head(&raw_v6_htable[state->bucket]);
1195 goto try_again;
1196 }
1197 return sk;
1198 }
1199
1200 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1201 {
1202 struct sock *sk = raw6_get_first(seq);
1203 if (sk)
1204 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1205 --pos;
1206 return pos ? NULL : sk;
1207 }
1208
1209 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1210 {
1211 read_lock(&raw_v6_lock);
1212 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1213 }
1214
1215 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1216 {
1217 struct sock *sk;
1218
1219 if (v == SEQ_START_TOKEN)
1220 sk = raw6_get_first(seq);
1221 else
1222 sk = raw6_get_next(seq, v);
1223 ++*pos;
1224 return sk;
1225 }
1226
1227 static void raw6_seq_stop(struct seq_file *seq, void *v)
1228 {
1229 read_unlock(&raw_v6_lock);
1230 }
1231
1232 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1233 {
1234 struct ipv6_pinfo *np = inet6_sk(sp);
1235 struct in6_addr *dest, *src;
1236 __u16 destp, srcp;
1237
1238 dest = &np->daddr;
1239 src = &np->rcv_saddr;
1240 destp = 0;
1241 srcp = inet_sk(sp)->num;
1242 seq_printf(seq,
1243 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1244 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1245 i,
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,
1250 sp->sk_state,
1251 sk_wmem_alloc_get(sp),
1252 sk_rmem_alloc_get(sp),
1253 0, 0L, 0,
1254 sock_i_uid(sp), 0,
1255 sock_i_ino(sp),
1256 atomic_read(&sp->sk_refcnt), sp);
1257 }
1258
1259 static int raw6_seq_show(struct seq_file *seq, void *v)
1260 {
1261 if (v == SEQ_START_TOKEN)
1262 seq_printf(seq,
1263 " sl "
1264 "local_address "
1265 "remote_address "
1266 "st tx_queue rx_queue tr tm->when retrnsmt"
1267 " uid timeout inode\n");
1268 else
1269 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1270 return 0;
1271 }
1272
1273 static struct seq_operations raw6_seq_ops = {
1274 .start = raw6_seq_start,
1275 .next = raw6_seq_next,
1276 .stop = raw6_seq_stop,
1277 .show = raw6_seq_show,
1278 };
1279
1280 static int raw6_seq_open(struct inode *inode, struct file *file)
1281 {
1282 return seq_open_private(file, &raw6_seq_ops,
1283 sizeof(struct raw6_iter_state));
1284 }
1285
1286 static const struct file_operations raw6_seq_fops = {
1287 .owner = THIS_MODULE,
1288 .open = raw6_seq_open,
1289 .read = seq_read,
1290 .llseek = seq_lseek,
1291 .release = seq_release_private,
1292 };
1293
1294 int __init raw6_proc_init(void)
1295 {
1296 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1297 return -ENOMEM;
1298 return 0;
1299 }
1300
1301 void raw6_proc_exit(void)
1302 {
1303 proc_net_remove("raw6");
1304 }
1305 #endif /* CONFIG_PROC_FS */
Tomato firmware and modifications.