?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[SK_BUFF]: Introduce skb_network_header()
[linux-2.6/openmoko-kernel/knife-kernel.git] / net / ipv6 / raw.c
blob9b2bcde73f1969eb521a0d89d45989022924d95b
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
56 #include <net/rawv6.h>
57 #include <net/xfrm.h>
58
59 #include <linux/proc_fs.h>
60 #include <linux/seq_file.h>
61
62 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
63 DEFINE_RWLOCK(raw_v6_lock);
64
65 static void raw_v6_hash(struct sock *sk)
66 {
67 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num &
68 (RAWV6_HTABLE_SIZE - 1)];
69
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);
74 }
75
76 static void raw_v6_unhash(struct sock *sk)
77 {
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);
82 }
83
84
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)
89 {
90 struct hlist_node *node;
91 int is_multicast = ipv6_addr_is_multicast(loc_addr);
92
93 sk_for_each_from(sk, node)
94 if (inet_sk(sk)->num == num) {
95 struct ipv6_pinfo *np = inet6_sk(sk);
96
97 if (!ipv6_addr_any(&np->daddr) &&
98 !ipv6_addr_equal(&np->daddr, rmt_addr))
99 continue;
100
101 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
102 continue;
103
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;
111 }
112 goto found;
113 }
114 sk = NULL;
115 found:
116 return sk;
117 }
118
119 /*
120 * 0 - deliver
121 * 1 - block
122 */
123 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
124 {
125 struct icmp6hdr *icmph;
126 struct raw6_sock *rp = raw6_sk(sk);
127
128 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
129 __u32 *data = &rp->filter.data[0];
130 int bit_nr;
131
132 icmph = (struct icmp6hdr *) skb->data;
133 bit_nr = icmph->icmp6_type;
134
135 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
136 }
137 return 0;
138 }
139
140 /*
141 * demultiplex raw sockets.
142 * (should consider queueing the skb in the sock receive_queue
143 * without calling rawv6.c)
144 *
145 * Caller owns SKB so we must make clones.
146 */
147 int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
148 {
149 struct in6_addr *saddr;
150 struct in6_addr *daddr;
151 struct sock *sk;
152 int delivered = 0;
153 __u8 hash;
154
155 saddr = &skb->nh.ipv6h->saddr;
156 daddr = saddr + 1;
157
158 hash = nexthdr & (MAX_INET_PROTOS - 1);
159
160 read_lock(&raw_v6_lock);
161 sk = sk_head(&raw_v6_htable[hash]);
162
163 /*
164 * The first socket found will be delivered after
165 * delivery to transport protocols.
166 */
167
168 if (sk == NULL)
169 goto out;
170
171 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
172
173 while (sk) {
174 int filtered;
175
176 delivered = 1;
177 switch (nexthdr) {
178 case IPPROTO_ICMPV6:
179 filtered = icmpv6_filter(sk, skb);
180 break;
181 #ifdef CONFIG_IPV6_MIP6
182 case IPPROTO_MH:
183 /* XXX: To validate MH only once for each packet,
184 * this is placed here. It should be after checking
185 * xfrm policy, however it doesn't. The checking xfrm
186 * policy is placed in rawv6_rcv() because it is
187 * required for each socket.
188 */
189 filtered = mip6_mh_filter(sk, skb);
190 break;
191 #endif
192 default:
193 filtered = 0;
194 break;
195 }
196
197 if (filtered < 0)
198 break;
199 if (filtered == 0) {
200 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
201
202 /* Not releasing hash table! */
203 if (clone) {
204 nf_reset(clone);
205 rawv6_rcv(sk, clone);
206 }
207 }
208 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr,
209 IP6CB(skb)->iif);
210 }
211 out:
212 read_unlock(&raw_v6_lock);
213 return delivered;
214 }
215
216 /* This cleans up af_inet6 a bit. -DaveM */
217 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
218 {
219 struct inet_sock *inet = inet_sk(sk);
220 struct ipv6_pinfo *np = inet6_sk(sk);
221 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
222 __be32 v4addr = 0;
223 int addr_type;
224 int err;
225
226 if (addr_len < SIN6_LEN_RFC2133)
227 return -EINVAL;
228 addr_type = ipv6_addr_type(&addr->sin6_addr);
229
230 /* Raw sockets are IPv6 only */
231 if (addr_type == IPV6_ADDR_MAPPED)
232 return(-EADDRNOTAVAIL);
233
234 lock_sock(sk);
235
236 err = -EINVAL;
237 if (sk->sk_state != TCP_CLOSE)
238 goto out;
239
240 /* Check if the address belongs to the host. */
241 if (addr_type != IPV6_ADDR_ANY) {
242 struct net_device *dev = NULL;
243
244 if (addr_type & IPV6_ADDR_LINKLOCAL) {
245 if (addr_len >= sizeof(struct sockaddr_in6) &&
246 addr->sin6_scope_id) {
247 /* Override any existing binding, if another
248 * one is supplied by user.
249 */
250 sk->sk_bound_dev_if = addr->sin6_scope_id;
251 }
252
253 /* Binding to link-local address requires an interface */
254 if (!sk->sk_bound_dev_if)
255 goto out;
256
257 dev = dev_get_by_index(sk->sk_bound_dev_if);
258 if (!dev) {
259 err = -ENODEV;
260 goto out;
261 }
262 }
263
264 /* ipv4 addr of the socket is invalid. Only the
265 * unspecified and mapped address have a v4 equivalent.
266 */
267 v4addr = LOOPBACK4_IPV6;
268 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
269 err = -EADDRNOTAVAIL;
270 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
271 if (dev)
272 dev_put(dev);
273 goto out;
274 }
275 }
276 if (dev)
277 dev_put(dev);
278 }
279
280 inet->rcv_saddr = inet->saddr = v4addr;
281 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
282 if (!(addr_type & IPV6_ADDR_MULTICAST))
283 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
284 err = 0;
285 out:
286 release_sock(sk);
287 return err;
288 }
289
290 void rawv6_err(struct sock *sk, struct sk_buff *skb,
291 struct inet6_skb_parm *opt,
292 int type, int code, int offset, __be32 info)
293 {
294 struct inet_sock *inet = inet_sk(sk);
295 struct ipv6_pinfo *np = inet6_sk(sk);
296 int err;
297 int harderr;
298
299 /* Report error on raw socket, if:
300 1. User requested recverr.
301 2. Socket is connected (otherwise the error indication
302 is useless without recverr and error is hard.
303 */
304 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
305 return;
306
307 harderr = icmpv6_err_convert(type, code, &err);
308 if (type == ICMPV6_PKT_TOOBIG)
309 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
310
311 if (np->recverr) {
312 u8 *payload = skb->data;
313 if (!inet->hdrincl)
314 payload += offset;
315 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
316 }
317
318 if (np->recverr || harderr) {
319 sk->sk_err = err;
320 sk->sk_error_report(sk);
321 }
322 }
323
324 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
325 {
326 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
327 skb_checksum_complete(skb)) {
328 /* FIXME: increment a raw6 drops counter here */
329 kfree_skb(skb);
330 return 0;
331 }
332
333 /* Charge it to the socket. */
334 if (sock_queue_rcv_skb(sk,skb)<0) {
335 /* FIXME: increment a raw6 drops counter here */
336 kfree_skb(skb);
337 return 0;
338 }
339
340 return 0;
341 }
342
343 /*
344 * This is next to useless...
345 * if we demultiplex in network layer we don't need the extra call
346 * just to queue the skb...
347 * maybe we could have the network decide upon a hint if it
348 * should call raw_rcv for demultiplexing
349 */
350 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
351 {
352 struct inet_sock *inet = inet_sk(sk);
353 struct raw6_sock *rp = raw6_sk(sk);
354
355 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
356 kfree_skb(skb);
357 return NET_RX_DROP;
358 }
359
360 if (!rp->checksum)
361 skb->ip_summed = CHECKSUM_UNNECESSARY;
362
363 if (skb->ip_summed == CHECKSUM_COMPLETE) {
364 skb_postpull_rcsum(skb, skb_network_header(skb),
365 skb->h.raw - skb->nh.raw);
366 if (!csum_ipv6_magic(&skb->nh.ipv6h->saddr,
367 &skb->nh.ipv6h->daddr,
368 skb->len, inet->num, skb->csum))
369 skb->ip_summed = CHECKSUM_UNNECESSARY;
370 }
371 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
372 skb->csum = ~csum_unfold(csum_ipv6_magic(&skb->nh.ipv6h->saddr,
373 &skb->nh.ipv6h->daddr,
374 skb->len, inet->num, 0));
375
376 if (inet->hdrincl) {
377 if (skb_checksum_complete(skb)) {
378 /* FIXME: increment a raw6 drops counter here */
379 kfree_skb(skb);
380 return 0;
381 }
382 }
383
384 rawv6_rcv_skb(sk, skb);
385 return 0;
386 }
387
388
389 /*
390 * This should be easy, if there is something there
391 * we return it, otherwise we block.
392 */
393
394 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
395 struct msghdr *msg, size_t len,
396 int noblock, int flags, int *addr_len)
397 {
398 struct ipv6_pinfo *np = inet6_sk(sk);
399 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
400 struct sk_buff *skb;
401 size_t copied;
402 int err;
403
404 if (flags & MSG_OOB)
405 return -EOPNOTSUPP;
406
407 if (addr_len)
408 *addr_len=sizeof(*sin6);
409
410 if (flags & MSG_ERRQUEUE)
411 return ipv6_recv_error(sk, msg, len);
412
413 skb = skb_recv_datagram(sk, flags, noblock, &err);
414 if (!skb)
415 goto out;
416
417 copied = skb->len;
418 if (copied > len) {
419 copied = len;
420 msg->msg_flags |= MSG_TRUNC;
421 }
422
423 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
424 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
425 } else if (msg->msg_flags&MSG_TRUNC) {
426 if (__skb_checksum_complete(skb))
427 goto csum_copy_err;
428 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
429 } else {
430 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
431 if (err == -EINVAL)
432 goto csum_copy_err;
433 }
434 if (err)
435 goto out_free;
436
437 /* Copy the address. */
438 if (sin6) {
439 sin6->sin6_family = AF_INET6;
440 sin6->sin6_port = 0;
441 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
442 sin6->sin6_flowinfo = 0;
443 sin6->sin6_scope_id = 0;
444 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
445 sin6->sin6_scope_id = IP6CB(skb)->iif;
446 }
447
448 sock_recv_timestamp(msg, sk, skb);
449
450 if (np->rxopt.all)
451 datagram_recv_ctl(sk, msg, skb);
452
453 err = copied;
454 if (flags & MSG_TRUNC)
455 err = skb->len;
456
457 out_free:
458 skb_free_datagram(sk, skb);
459 out:
460 return err;
461
462 csum_copy_err:
463 skb_kill_datagram(sk, skb, flags);
464
465 /* Error for blocking case is chosen to masquerade
466 as some normal condition.
467 */
468 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
469 /* FIXME: increment a raw6 drops counter here */
470 goto out;
471 }
472
473 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
474 struct raw6_sock *rp)
475 {
476 struct sk_buff *skb;
477 int err = 0;
478 int offset;
479 int len;
480 int total_len;
481 __wsum tmp_csum;
482 __sum16 csum;
483
484 if (!rp->checksum)
485 goto send;
486
487 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
488 goto out;
489
490 offset = rp->offset;
491 total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
492 skb->data);
493 if (offset >= total_len - 1) {
494 err = -EINVAL;
495 ip6_flush_pending_frames(sk);
496 goto out;
497 }
498
499 /* should be check HW csum miyazawa */
500 if (skb_queue_len(&sk->sk_write_queue) == 1) {
501 /*
502 * Only one fragment on the socket.
503 */
504 tmp_csum = skb->csum;
505 } else {
506 struct sk_buff *csum_skb = NULL;
507 tmp_csum = 0;
508
509 skb_queue_walk(&sk->sk_write_queue, skb) {
510 tmp_csum = csum_add(tmp_csum, skb->csum);
511
512 if (csum_skb)
513 continue;
514
515 len = skb->len - (skb->h.raw - skb->data);
516 if (offset >= len) {
517 offset -= len;
518 continue;
519 }
520
521 csum_skb = skb;
522 }
523
524 skb = csum_skb;
525 }
526
527 offset += skb->h.raw - skb->data;
528 if (skb_copy_bits(skb, offset, &csum, 2))
529 BUG();
530
531 /* in case cksum was not initialized */
532 if (unlikely(csum))
533 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
534
535 csum = csum_ipv6_magic(&fl->fl6_src,
536 &fl->fl6_dst,
537 total_len, fl->proto, tmp_csum);
538
539 if (csum == 0 && fl->proto == IPPROTO_UDP)
540 csum = CSUM_MANGLED_0;
541
542 if (skb_store_bits(skb, offset, &csum, 2))
543 BUG();
544
545 send:
546 err = ip6_push_pending_frames(sk);
547 out:
548 return err;
549 }
550
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)
554 {
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;
560
561 if (length > rt->u.dst.dev->mtu) {
562 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
563 return -EMSGSIZE;
564 }
565 if (flags&MSG_PROBE)
566 goto out;
567
568 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
569
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);
575
576 skb->priority = sk->sk_priority;
577 skb->dst = dst_clone(&rt->u.dst);
578
579 skb_put(skb, length);
580 skb_reset_network_header(skb);
581 iph = skb->nh.ipv6h;
582
583 skb->ip_summed = CHECKSUM_NONE;
584
585 skb->h.raw = skb->nh.raw;
586 err = memcpy_fromiovecend((void *)iph, from, 0, length);
587 if (err)
588 goto error_fault;
589
590 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
591 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
592 dst_output);
593 if (err > 0)
594 err = np->recverr ? net_xmit_errno(err) : 0;
595 if (err)
596 goto error;
597 out:
598 return 0;
599
600 error_fault:
601 err = -EFAULT;
602 kfree_skb(skb);
603 error:
604 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
605 return err;
606 }
607
608 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
609 {
610 struct iovec *iov;
611 u8 __user *type = NULL;
612 u8 __user *code = NULL;
613 #ifdef CONFIG_IPV6_MIP6
614 u8 len = 0;
615 #endif
616 int probed = 0;
617 int i;
618
619 if (!msg->msg_iov)
620 return 0;
621
622 for (i = 0; i < msg->msg_iovlen; i++) {
623 iov = &msg->msg_iov[i];
624 if (!iov)
625 continue;
626
627 switch (fl->proto) {
628 case IPPROTO_ICMPV6:
629 /* check if one-byte field is readable or not. */
630 if (iov->iov_base && iov->iov_len < 1)
631 break;
632
633 if (!type) {
634 type = iov->iov_base;
635 /* check if code field is readable or not. */
636 if (iov->iov_len > 1)
637 code = type + 1;
638 } else if (!code)
639 code = iov->iov_base;
640
641 if (type && code) {
642 if (get_user(fl->fl_icmp_type, type) ||
643 get_user(fl->fl_icmp_code, code))
644 return -EFAULT;
645 probed = 1;
646 }
647 break;
648 #ifdef CONFIG_IPV6_MIP6
649 case IPPROTO_MH:
650 if (iov->iov_base && iov->iov_len < 1)
651 break;
652 /* check if type field is readable or not. */
653 if (iov->iov_len > 2 - len) {
654 u8 __user *p = iov->iov_base;
655 if (get_user(fl->fl_mh_type, &p[2 - len]))
656 return -EFAULT;
657 probed = 1;
658 } else
659 len += iov->iov_len;
660
661 break;
662 #endif
663 default:
664 probed = 1;
665 break;
666 }
667 if (probed)
668 break;
669 }
670 return 0;
671 }
672
673 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
674 struct msghdr *msg, size_t len)
675 {
676 struct ipv6_txoptions opt_space;
677 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
678 struct in6_addr *daddr, *final_p = NULL, final;
679 struct inet_sock *inet = inet_sk(sk);
680 struct ipv6_pinfo *np = inet6_sk(sk);
681 struct raw6_sock *rp = raw6_sk(sk);
682 struct ipv6_txoptions *opt = NULL;
683 struct ip6_flowlabel *flowlabel = NULL;
684 struct dst_entry *dst = NULL;
685 struct flowi fl;
686 int addr_len = msg->msg_namelen;
687 int hlimit = -1;
688 int tclass = -1;
689 u16 proto;
690 int err;
691
692 /* Rough check on arithmetic overflow,
693 better check is made in ip6_append_data().
694 */
695 if (len > INT_MAX)
696 return -EMSGSIZE;
697
698 /* Mirror BSD error message compatibility */
699 if (msg->msg_flags & MSG_OOB)
700 return -EOPNOTSUPP;
701
702 /*
703 * Get and verify the address.
704 */
705 memset(&fl, 0, sizeof(fl));
706
707 if (sin6) {
708 if (addr_len < SIN6_LEN_RFC2133)
709 return -EINVAL;
710
711 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
712 return(-EAFNOSUPPORT);
713
714 /* port is the proto value [0..255] carried in nexthdr */
715 proto = ntohs(sin6->sin6_port);
716
717 if (!proto)
718 proto = inet->num;
719 else if (proto != inet->num)
720 return(-EINVAL);
721
722 if (proto > 255)
723 return(-EINVAL);
724
725 daddr = &sin6->sin6_addr;
726 if (np->sndflow) {
727 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
728 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
729 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
730 if (flowlabel == NULL)
731 return -EINVAL;
732 daddr = &flowlabel->dst;
733 }
734 }
735
736 /*
737 * Otherwise it will be difficult to maintain
738 * sk->sk_dst_cache.
739 */
740 if (sk->sk_state == TCP_ESTABLISHED &&
741 ipv6_addr_equal(daddr, &np->daddr))
742 daddr = &np->daddr;
743
744 if (addr_len >= sizeof(struct sockaddr_in6) &&
745 sin6->sin6_scope_id &&
746 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
747 fl.oif = sin6->sin6_scope_id;
748 } else {
749 if (sk->sk_state != TCP_ESTABLISHED)
750 return -EDESTADDRREQ;
751
752 proto = inet->num;
753 daddr = &np->daddr;
754 fl.fl6_flowlabel = np->flow_label;
755 }
756
757 if (ipv6_addr_any(daddr)) {
758 /*
759 * unspecified destination address
760 * treated as error... is this correct ?
761 */
762 fl6_sock_release(flowlabel);
763 return(-EINVAL);
764 }
765
766 if (fl.oif == 0)
767 fl.oif = sk->sk_bound_dev_if;
768
769 if (msg->msg_controllen) {
770 opt = &opt_space;
771 memset(opt, 0, sizeof(struct ipv6_txoptions));
772 opt->tot_len = sizeof(struct ipv6_txoptions);
773
774 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
775 if (err < 0) {
776 fl6_sock_release(flowlabel);
777 return err;
778 }
779 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
780 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
781 if (flowlabel == NULL)
782 return -EINVAL;
783 }
784 if (!(opt->opt_nflen|opt->opt_flen))
785 opt = NULL;
786 }
787 if (opt == NULL)
788 opt = np->opt;
789 if (flowlabel)
790 opt = fl6_merge_options(&opt_space, flowlabel, opt);
791 opt = ipv6_fixup_options(&opt_space, opt);
792
793 fl.proto = proto;
794 err = rawv6_probe_proto_opt(&fl, msg);
795 if (err)
796 goto out;
797
798 ipv6_addr_copy(&fl.fl6_dst, daddr);
799 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
800 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
801
802 /* merge ip6_build_xmit from ip6_output */
803 if (opt && opt->srcrt) {
804 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
805 ipv6_addr_copy(&final, &fl.fl6_dst);
806 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
807 final_p = &final;
808 }
809
810 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
811 fl.oif = np->mcast_oif;
812 security_sk_classify_flow(sk, &fl);
813
814 err = ip6_dst_lookup(sk, &dst, &fl);
815 if (err)
816 goto out;
817 if (final_p)
818 ipv6_addr_copy(&fl.fl6_dst, final_p);
819
820 if ((err = xfrm_lookup(&dst, &fl, sk, 1)) < 0)
821 goto out;
822
823 if (hlimit < 0) {
824 if (ipv6_addr_is_multicast(&fl.fl6_dst))
825 hlimit = np->mcast_hops;
826 else
827 hlimit = np->hop_limit;
828 if (hlimit < 0)
829 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
830 if (hlimit < 0)
831 hlimit = ipv6_get_hoplimit(dst->dev);
832 }
833
834 if (tclass < 0) {
835 tclass = np->tclass;
836 if (tclass < 0)
837 tclass = 0;
838 }
839
840 if (msg->msg_flags&MSG_CONFIRM)
841 goto do_confirm;
842
843 back_from_confirm:
844 if (inet->hdrincl) {
845 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
846 } else {
847 lock_sock(sk);
848 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
849 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
850 msg->msg_flags);
851
852 if (err)
853 ip6_flush_pending_frames(sk);
854 else if (!(msg->msg_flags & MSG_MORE))
855 err = rawv6_push_pending_frames(sk, &fl, rp);
856 }
857 done:
858 dst_release(dst);
859 if (!inet->hdrincl)
860 release_sock(sk);
861 out:
862 fl6_sock_release(flowlabel);
863 return err<0?err:len;
864 do_confirm:
865 dst_confirm(dst);
866 if (!(msg->msg_flags & MSG_PROBE) || len)
867 goto back_from_confirm;
868 err = 0;
869 goto done;
870 }
871
872 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
873 char __user *optval, int optlen)
874 {
875 switch (optname) {
876 case ICMPV6_FILTER:
877 if (optlen > sizeof(struct icmp6_filter))
878 optlen = sizeof(struct icmp6_filter);
879 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
880 return -EFAULT;
881 return 0;
882 default:
883 return -ENOPROTOOPT;
884 };
885
886 return 0;
887 }
888
889 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
890 char __user *optval, int __user *optlen)
891 {
892 int len;
893
894 switch (optname) {
895 case ICMPV6_FILTER:
896 if (get_user(len, optlen))
897 return -EFAULT;
898 if (len < 0)
899 return -EINVAL;
900 if (len > sizeof(struct icmp6_filter))
901 len = sizeof(struct icmp6_filter);
902 if (put_user(len, optlen))
903 return -EFAULT;
904 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
905 return -EFAULT;
906 return 0;
907 default:
908 return -ENOPROTOOPT;
909 };
910
911 return 0;
912 }
913
914
915 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
916 char __user *optval, int optlen)
917 {
918 struct raw6_sock *rp = raw6_sk(sk);
919 int val;
920
921 if (get_user(val, (int __user *)optval))
922 return -EFAULT;
923
924 switch (optname) {
925 case IPV6_CHECKSUM:
926 /* You may get strange result with a positive odd offset;
927 RFC2292bis agrees with me. */
928 if (val > 0 && (val&1))
929 return(-EINVAL);
930 if (val < 0) {
931 rp->checksum = 0;
932 } else {
933 rp->checksum = 1;
934 rp->offset = val;
935 }
936
937 return 0;
938 break;
939
940 default:
941 return(-ENOPROTOOPT);
942 }
943 }
944
945 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
946 char __user *optval, int optlen)
947 {
948 switch(level) {
949 case SOL_RAW:
950 break;
951
952 case SOL_ICMPV6:
953 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
954 return -EOPNOTSUPP;
955 return rawv6_seticmpfilter(sk, level, optname, optval,
956 optlen);
957 case SOL_IPV6:
958 if (optname == IPV6_CHECKSUM)
959 break;
960 default:
961 return ipv6_setsockopt(sk, level, optname, optval,
962 optlen);
963 };
964 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
965 }
966
967 #ifdef CONFIG_COMPAT
968 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
969 char __user *optval, int optlen)
970 {
971 switch (level) {
972 case SOL_RAW:
973 break;
974 case SOL_ICMPV6:
975 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
976 return -EOPNOTSUPP;
977 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
978 case SOL_IPV6:
979 if (optname == IPV6_CHECKSUM)
980 break;
981 default:
982 return compat_ipv6_setsockopt(sk, level, optname,
983 optval, optlen);
984 };
985 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
986 }
987 #endif
988
989 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
990 char __user *optval, int __user *optlen)
991 {
992 struct raw6_sock *rp = raw6_sk(sk);
993 int val, len;
994
995 if (get_user(len,optlen))
996 return -EFAULT;
997
998 switch (optname) {
999 case IPV6_CHECKSUM:
1000 if (rp->checksum == 0)
1001 val = -1;
1002 else
1003 val = rp->offset;
1004 break;
1005
1006 default:
1007 return -ENOPROTOOPT;
1008 }
1009
1010 len = min_t(unsigned int, sizeof(int), len);
1011
1012 if (put_user(len, optlen))
1013 return -EFAULT;
1014 if (copy_to_user(optval,&val,len))
1015 return -EFAULT;
1016 return 0;
1017 }
1018
1019 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1020 char __user *optval, int __user *optlen)
1021 {
1022 switch(level) {
1023 case SOL_RAW:
1024 break;
1025
1026 case SOL_ICMPV6:
1027 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1028 return -EOPNOTSUPP;
1029 return rawv6_geticmpfilter(sk, level, optname, optval,
1030 optlen);
1031 case SOL_IPV6:
1032 if (optname == IPV6_CHECKSUM)
1033 break;
1034 default:
1035 return ipv6_getsockopt(sk, level, optname, optval,
1036 optlen);
1037 };
1038 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1039 }
1040
1041 #ifdef CONFIG_COMPAT
1042 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1043 char __user *optval, int __user *optlen)
1044 {
1045 switch (level) {
1046 case SOL_RAW:
1047 break;
1048 case SOL_ICMPV6:
1049 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1050 return -EOPNOTSUPP;
1051 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1052 case SOL_IPV6:
1053 if (optname == IPV6_CHECKSUM)
1054 break;
1055 default:
1056 return compat_ipv6_getsockopt(sk, level, optname,
1057 optval, optlen);
1058 };
1059 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1060 }
1061 #endif
1062
1063 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1064 {
1065 switch(cmd) {
1066 case SIOCOUTQ:
1067 {
1068 int amount = atomic_read(&sk->sk_wmem_alloc);
1069 return put_user(amount, (int __user *)arg);
1070 }
1071 case SIOCINQ:
1072 {
1073 struct sk_buff *skb;
1074 int amount = 0;
1075
1076 spin_lock_bh(&sk->sk_receive_queue.lock);
1077 skb = skb_peek(&sk->sk_receive_queue);
1078 if (skb != NULL)
1079 amount = skb->tail - skb->h.raw;
1080 spin_unlock_bh(&sk->sk_receive_queue.lock);
1081 return put_user(amount, (int __user *)arg);
1082 }
1083
1084 default:
1085 return -ENOIOCTLCMD;
1086 }
1087 }
1088
1089 static void rawv6_close(struct sock *sk, long timeout)
1090 {
1091 if (inet_sk(sk)->num == IPPROTO_RAW)
1092 ip6_ra_control(sk, -1, NULL);
1093
1094 sk_common_release(sk);
1095 }
1096
1097 static int rawv6_init_sk(struct sock *sk)
1098 {
1099 struct raw6_sock *rp = raw6_sk(sk);
1100
1101 switch (inet_sk(sk)->num) {
1102 case IPPROTO_ICMPV6:
1103 rp->checksum = 1;
1104 rp->offset = 2;
1105 break;
1106 case IPPROTO_MH:
1107 rp->checksum = 1;
1108 rp->offset = 4;
1109 break;
1110 default:
1111 break;
1112 }
1113 return(0);
1114 }
1115
1116 struct proto rawv6_prot = {
1117 .name = "RAWv6",
1118 .owner = THIS_MODULE,
1119 .close = rawv6_close,
1120 .connect = ip6_datagram_connect,
1121 .disconnect = udp_disconnect,
1122 .ioctl = rawv6_ioctl,
1123 .init = rawv6_init_sk,
1124 .destroy = inet6_destroy_sock,
1125 .setsockopt = rawv6_setsockopt,
1126 .getsockopt = rawv6_getsockopt,
1127 .sendmsg = rawv6_sendmsg,
1128 .recvmsg = rawv6_recvmsg,
1129 .bind = rawv6_bind,
1130 .backlog_rcv = rawv6_rcv_skb,
1131 .hash = raw_v6_hash,
1132 .unhash = raw_v6_unhash,
1133 .obj_size = sizeof(struct raw6_sock),
1134 #ifdef CONFIG_COMPAT
1135 .compat_setsockopt = compat_rawv6_setsockopt,
1136 .compat_getsockopt = compat_rawv6_getsockopt,
1137 #endif
1138 };
1139
1140 #ifdef CONFIG_PROC_FS
1141 struct raw6_iter_state {
1142 int bucket;
1143 };
1144
1145 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1146
1147 static struct sock *raw6_get_first(struct seq_file *seq)
1148 {
1149 struct sock *sk;
1150 struct hlist_node *node;
1151 struct raw6_iter_state* state = raw6_seq_private(seq);
1152
1153 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1154 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1155 if (sk->sk_family == PF_INET6)
1156 goto out;
1157 sk = NULL;
1158 out:
1159 return sk;
1160 }
1161
1162 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1163 {
1164 struct raw6_iter_state* state = raw6_seq_private(seq);
1165
1166 do {
1167 sk = sk_next(sk);
1168 try_again:
1169 ;
1170 } while (sk && sk->sk_family != PF_INET6);
1171
1172 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1173 sk = sk_head(&raw_v6_htable[state->bucket]);
1174 goto try_again;
1175 }
1176 return sk;
1177 }
1178
1179 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1180 {
1181 struct sock *sk = raw6_get_first(seq);
1182 if (sk)
1183 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1184 --pos;
1185 return pos ? NULL : sk;
1186 }
1187
1188 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1189 {
1190 read_lock(&raw_v6_lock);
1191 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1192 }
1193
1194 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1195 {
1196 struct sock *sk;
1197
1198 if (v == SEQ_START_TOKEN)
1199 sk = raw6_get_first(seq);
1200 else
1201 sk = raw6_get_next(seq, v);
1202 ++*pos;
1203 return sk;
1204 }
1205
1206 static void raw6_seq_stop(struct seq_file *seq, void *v)
1207 {
1208 read_unlock(&raw_v6_lock);
1209 }
1210
1211 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1212 {
1213 struct ipv6_pinfo *np = inet6_sk(sp);
1214 struct in6_addr *dest, *src;
1215 __u16 destp, srcp;
1216
1217 dest = &np->daddr;
1218 src = &np->rcv_saddr;
1219 destp = 0;
1220 srcp = inet_sk(sp)->num;
1221 seq_printf(seq,
1222 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1223 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1224 i,
1225 src->s6_addr32[0], src->s6_addr32[1],
1226 src->s6_addr32[2], src->s6_addr32[3], srcp,
1227 dest->s6_addr32[0], dest->s6_addr32[1],
1228 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1229 sp->sk_state,
1230 atomic_read(&sp->sk_wmem_alloc),
1231 atomic_read(&sp->sk_rmem_alloc),
1232 0, 0L, 0,
1233 sock_i_uid(sp), 0,
1234 sock_i_ino(sp),
1235 atomic_read(&sp->sk_refcnt), sp);
1236 }
1237
1238 static int raw6_seq_show(struct seq_file *seq, void *v)
1239 {
1240 if (v == SEQ_START_TOKEN)
1241 seq_printf(seq,
1242 " sl "
1243 "local_address "
1244 "remote_address "
1245 "st tx_queue rx_queue tr tm->when retrnsmt"
1246 " uid timeout inode\n");
1247 else
1248 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1249 return 0;
1250 }
1251
1252 static struct seq_operations raw6_seq_ops = {
1253 .start = raw6_seq_start,
1254 .next = raw6_seq_next,
1255 .stop = raw6_seq_stop,
1256 .show = raw6_seq_show,
1257 };
1258
1259 static int raw6_seq_open(struct inode *inode, struct file *file)
1260 {
1261 struct seq_file *seq;
1262 int rc = -ENOMEM;
1263 struct raw6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
1264 if (!s)
1265 goto out;
1266 rc = seq_open(file, &raw6_seq_ops);
1267 if (rc)
1268 goto out_kfree;
1269 seq = file->private_data;
1270 seq->private = s;
1271 out:
1272 return rc;
1273 out_kfree:
1274 kfree(s);
1275 goto out;
1276 }
1277
1278 static const struct file_operations raw6_seq_fops = {
1279 .owner = THIS_MODULE,
1280 .open = raw6_seq_open,
1281 .read = seq_read,
1282 .llseek = seq_lseek,
1283 .release = seq_release_private,
1284 };
1285
1286 int __init raw6_proc_init(void)
1287 {
1288 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1289 return -ENOMEM;
1290 return 0;
1291 }
1292
1293 void raw6_proc_exit(void)
1294 {
1295 proc_net_remove("raw6");
1296 }
1297 #endif /* CONFIG_PROC_FS */
Community contributions to the Openmoko Kernel