?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
initial commit with v2.6.9
[linux-2.6.9-moxart.git] / net / ipv6 / raw.c
blobc73d720114fa9721608b68d02b51f7b0008aa1c6
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/sched.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/icmpv6.h>
33 #include <linux/netfilter.h>
34 #include <linux/netfilter_ipv6.h>
35 #include <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
52 #include <net/rawv6.h>
53 #include <net/xfrm.h>
54
55 #include <linux/proc_fs.h>
56 #include <linux/seq_file.h>
57
58 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
59 rwlock_t raw_v6_lock = RW_LOCK_UNLOCKED;
60
61 static void raw_v6_hash(struct sock *sk)
62 {
63 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num &
64 (RAWV6_HTABLE_SIZE - 1)];
65
66 write_lock_bh(&raw_v6_lock);
67 sk_add_node(sk, list);
68 sock_prot_inc_use(sk->sk_prot);
69 write_unlock_bh(&raw_v6_lock);
70 }
71
72 static void raw_v6_unhash(struct sock *sk)
73 {
74 write_lock_bh(&raw_v6_lock);
75 if (sk_del_node_init(sk))
76 sock_prot_dec_use(sk->sk_prot);
77 write_unlock_bh(&raw_v6_lock);
78 }
79
80
81 /* Grumble... icmp and ip_input want to get at this... */
82 struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
83 struct in6_addr *loc_addr, struct in6_addr *rmt_addr)
84 {
85 struct hlist_node *node;
86 int is_multicast = ipv6_addr_is_multicast(loc_addr);
87
88 sk_for_each_from(sk, node)
89 if (inet_sk(sk)->num == num) {
90 struct ipv6_pinfo *np = inet6_sk(sk);
91
92 if (!ipv6_addr_any(&np->daddr) &&
93 ipv6_addr_cmp(&np->daddr, rmt_addr))
94 continue;
95
96 if (!ipv6_addr_any(&np->rcv_saddr)) {
97 if (!ipv6_addr_cmp(&np->rcv_saddr, loc_addr))
98 goto found;
99 if (is_multicast &&
100 inet6_mc_check(sk, loc_addr, rmt_addr))
101 goto found;
102 continue;
103 }
104 goto found;
105 }
106 sk = NULL;
107 found:
108 return sk;
109 }
110
111 /*
112 * 0 - deliver
113 * 1 - block
114 */
115 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
116 {
117 struct icmp6hdr *icmph;
118 struct raw6_opt *opt = raw6_sk(sk);
119
120 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
121 __u32 *data = &opt->filter.data[0];
122 int bit_nr;
123
124 icmph = (struct icmp6hdr *) skb->data;
125 bit_nr = icmph->icmp6_type;
126
127 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
128 }
129 return 0;
130 }
131
132 /*
133 * demultiplex raw sockets.
134 * (should consider queueing the skb in the sock receive_queue
135 * without calling rawv6.c)
136 *
137 * Caller owns SKB so we must make clones.
138 */
139 void ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
140 {
141 struct in6_addr *saddr;
142 struct in6_addr *daddr;
143 struct sock *sk;
144 __u8 hash;
145
146 saddr = &skb->nh.ipv6h->saddr;
147 daddr = saddr + 1;
148
149 hash = nexthdr & (MAX_INET_PROTOS - 1);
150
151 read_lock(&raw_v6_lock);
152 sk = sk_head(&raw_v6_htable[hash]);
153
154 /*
155 * The first socket found will be delivered after
156 * delivery to transport protocols.
157 */
158
159 if (sk == NULL)
160 goto out;
161
162 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr);
163
164 while (sk) {
165 if (nexthdr != IPPROTO_ICMPV6 || !icmpv6_filter(sk, skb)) {
166 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
167
168 /* Not releasing hash table! */
169 if (clone)
170 rawv6_rcv(sk, clone);
171 }
172 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr);
173 }
174 out:
175 read_unlock(&raw_v6_lock);
176 }
177
178 /* This cleans up af_inet6 a bit. -DaveM */
179 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
180 {
181 struct inet_opt *inet = inet_sk(sk);
182 struct ipv6_pinfo *np = inet6_sk(sk);
183 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
184 __u32 v4addr = 0;
185 int addr_type;
186 int err;
187
188 if (addr_len < SIN6_LEN_RFC2133)
189 return -EINVAL;
190 addr_type = ipv6_addr_type(&addr->sin6_addr);
191
192 /* Raw sockets are IPv6 only */
193 if (addr_type == IPV6_ADDR_MAPPED)
194 return(-EADDRNOTAVAIL);
195
196 lock_sock(sk);
197
198 err = -EINVAL;
199 if (sk->sk_state != TCP_CLOSE)
200 goto out;
201
202 /* Check if the address belongs to the host. */
203 if (addr_type != IPV6_ADDR_ANY) {
204 struct net_device *dev = NULL;
205
206 if (addr_type & IPV6_ADDR_LINKLOCAL) {
207 if (addr_len >= sizeof(struct sockaddr_in6) &&
208 addr->sin6_scope_id) {
209 /* Override any existing binding, if another
210 * one is supplied by user.
211 */
212 sk->sk_bound_dev_if = addr->sin6_scope_id;
213 }
214
215 /* Binding to link-local address requires an interface */
216 if (!sk->sk_bound_dev_if)
217 goto out;
218
219 dev = dev_get_by_index(sk->sk_bound_dev_if);
220 if (!dev) {
221 err = -ENODEV;
222 goto out;
223 }
224 }
225
226 /* ipv4 addr of the socket is invalid. Only the
227 * unspecified and mapped address have a v4 equivalent.
228 */
229 v4addr = LOOPBACK4_IPV6;
230 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
231 err = -EADDRNOTAVAIL;
232 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
233 if (dev)
234 dev_put(dev);
235 goto out;
236 }
237 }
238 if (dev)
239 dev_put(dev);
240 }
241
242 inet->rcv_saddr = inet->saddr = v4addr;
243 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
244 if (!(addr_type & IPV6_ADDR_MULTICAST))
245 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
246 err = 0;
247 out:
248 release_sock(sk);
249 return err;
250 }
251
252 void rawv6_err(struct sock *sk, struct sk_buff *skb,
253 struct inet6_skb_parm *opt,
254 int type, int code, int offset, u32 info)
255 {
256 struct inet_opt *inet = inet_sk(sk);
257 struct ipv6_pinfo *np = inet6_sk(sk);
258 int err;
259 int harderr;
260
261 /* Report error on raw socket, if:
262 1. User requested recverr.
263 2. Socket is connected (otherwise the error indication
264 is useless without recverr and error is hard.
265 */
266 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
267 return;
268
269 harderr = icmpv6_err_convert(type, code, &err);
270 if (type == ICMPV6_PKT_TOOBIG)
271 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
272
273 if (np->recverr) {
274 u8 *payload = skb->data;
275 if (!inet->hdrincl)
276 payload += offset;
277 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
278 }
279
280 if (np->recverr || harderr) {
281 sk->sk_err = err;
282 sk->sk_error_report(sk);
283 }
284 }
285
286 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
287 {
288 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
289 skb->ip_summed != CHECKSUM_UNNECESSARY) {
290 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
291 /* FIXME: increment a raw6 drops counter here */
292 kfree_skb(skb);
293 return 0;
294 }
295 skb->ip_summed = CHECKSUM_UNNECESSARY;
296 }
297
298 /* Charge it to the socket. */
299 if (sock_queue_rcv_skb(sk,skb)<0) {
300 /* FIXME: increment a raw6 drops counter here */
301 kfree_skb(skb);
302 return 0;
303 }
304
305 return 0;
306 }
307
308 /*
309 * This is next to useless...
310 * if we demultiplex in network layer we don't need the extra call
311 * just to queue the skb...
312 * maybe we could have the network decide upon a hint if it
313 * should call raw_rcv for demultiplexing
314 */
315 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
316 {
317 struct inet_opt *inet = inet_sk(sk);
318 struct raw6_opt *raw_opt = raw6_sk(sk);
319
320 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
321 kfree_skb(skb);
322 return NET_RX_DROP;
323 }
324
325 if (!raw_opt->checksum)
326 skb->ip_summed = CHECKSUM_UNNECESSARY;
327
328 if (skb->ip_summed != CHECKSUM_UNNECESSARY) {
329 if (skb->ip_summed == CHECKSUM_HW) {
330 skb->ip_summed = CHECKSUM_UNNECESSARY;
331 if (csum_ipv6_magic(&skb->nh.ipv6h->saddr,
332 &skb->nh.ipv6h->daddr,
333 skb->len, inet->num, skb->csum)) {
334 LIMIT_NETDEBUG(
335 printk(KERN_DEBUG "raw v6 hw csum failure.\n"));
336 skb->ip_summed = CHECKSUM_NONE;
337 }
338 }
339 if (skb->ip_summed == CHECKSUM_NONE)
340 skb->csum = ~csum_ipv6_magic(&skb->nh.ipv6h->saddr,
341 &skb->nh.ipv6h->daddr,
342 skb->len, inet->num, 0);
343 }
344
345 if (inet->hdrincl) {
346 if (skb->ip_summed != CHECKSUM_UNNECESSARY &&
347 (unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
348 /* FIXME: increment a raw6 drops counter here */
349 kfree_skb(skb);
350 return 0;
351 }
352 skb->ip_summed = CHECKSUM_UNNECESSARY;
353 }
354
355 rawv6_rcv_skb(sk, skb);
356 return 0;
357 }
358
359
360 /*
361 * This should be easy, if there is something there
362 * we return it, otherwise we block.
363 */
364
365 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
366 struct msghdr *msg, size_t len,
367 int noblock, int flags, int *addr_len)
368 {
369 struct ipv6_pinfo *np = inet6_sk(sk);
370 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
371 struct sk_buff *skb;
372 size_t copied;
373 int err;
374
375 if (flags & MSG_OOB)
376 return -EOPNOTSUPP;
377
378 if (addr_len)
379 *addr_len=sizeof(*sin6);
380
381 if (flags & MSG_ERRQUEUE)
382 return ipv6_recv_error(sk, msg, len);
383
384 skb = skb_recv_datagram(sk, flags, noblock, &err);
385 if (!skb)
386 goto out;
387
388 copied = skb->len;
389 if (copied > len) {
390 copied = len;
391 msg->msg_flags |= MSG_TRUNC;
392 }
393
394 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
395 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
396 } else if (msg->msg_flags&MSG_TRUNC) {
397 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum)))
398 goto csum_copy_err;
399 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
400 } else {
401 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
402 if (err == -EINVAL)
403 goto csum_copy_err;
404 }
405 if (err)
406 goto out_free;
407
408 /* Copy the address. */
409 if (sin6) {
410 sin6->sin6_family = AF_INET6;
411 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
412 sin6->sin6_flowinfo = 0;
413 sin6->sin6_scope_id = 0;
414 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
415 sin6->sin6_scope_id = IP6CB(skb)->iif;
416 }
417
418 sock_recv_timestamp(msg, sk, skb);
419
420 if (np->rxopt.all)
421 datagram_recv_ctl(sk, msg, skb);
422
423 err = copied;
424 if (flags & MSG_TRUNC)
425 err = skb->len;
426
427 out_free:
428 skb_free_datagram(sk, skb);
429 out:
430 return err;
431
432 csum_copy_err:
433 /* Clear queue. */
434 if (flags&MSG_PEEK) {
435 int clear = 0;
436 spin_lock_irq(&sk->sk_receive_queue.lock);
437 if (skb == skb_peek(&sk->sk_receive_queue)) {
438 __skb_unlink(skb, &sk->sk_receive_queue);
439 clear = 1;
440 }
441 spin_unlock_irq(&sk->sk_receive_queue.lock);
442 if (clear)
443 kfree_skb(skb);
444 }
445
446 /* Error for blocking case is chosen to masquerade
447 as some normal condition.
448 */
449 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
450 /* FIXME: increment a raw6 drops counter here */
451 goto out_free;
452 }
453
454 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl, struct raw6_opt *opt, int len)
455 {
456 struct sk_buff *skb;
457 int err = 0;
458 u16 *csum;
459 u32 tmp_csum;
460
461 if (!opt->checksum)
462 goto send;
463
464 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
465 goto out;
466
467 if (opt->offset + 1 < len)
468 csum = (u16 *)(skb->h.raw + opt->offset);
469 else {
470 err = -EINVAL;
471 goto out;
472 }
473
474 /* should be check HW csum miyazawa */
475 if (skb_queue_len(&sk->sk_write_queue) == 1) {
476 /*
477 * Only one fragment on the socket.
478 */
479 tmp_csum = skb->csum;
480 } else {
481 tmp_csum = 0;
482
483 skb_queue_walk(&sk->sk_write_queue, skb) {
484 tmp_csum = csum_add(tmp_csum, skb->csum);
485 }
486 }
487
488 /* in case cksum was not initialized */
489 if (unlikely(*csum))
490 tmp_csum = csum_sub(tmp_csum, *csum);
491
492 *csum = csum_ipv6_magic(&fl->fl6_src,
493 &fl->fl6_dst,
494 len, fl->proto, tmp_csum);
495
496 if (*csum == 0)
497 *csum = -1;
498 send:
499 err = ip6_push_pending_frames(sk);
500 out:
501 return err;
502 }
503
504 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
505 struct flowi *fl, struct rt6_info *rt,
506 unsigned int flags)
507 {
508 struct inet_opt *inet = inet_sk(sk);
509 struct ipv6hdr *iph;
510 struct sk_buff *skb;
511 unsigned int hh_len;
512 int err;
513
514 if (length > rt->u.dst.dev->mtu) {
515 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
516 return -EMSGSIZE;
517 }
518 if (flags&MSG_PROBE)
519 goto out;
520
521 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
522
523 skb = sock_alloc_send_skb(sk, length+hh_len+15,
524 flags&MSG_DONTWAIT, &err);
525 if (skb == NULL)
526 goto error;
527 skb_reserve(skb, hh_len);
528
529 skb->priority = sk->sk_priority;
530 skb->dst = dst_clone(&rt->u.dst);
531
532 skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length);
533
534 skb->ip_summed = CHECKSUM_NONE;
535
536 skb->h.raw = skb->nh.raw;
537 err = memcpy_fromiovecend((void *)iph, from, 0, length);
538 if (err)
539 goto error_fault;
540
541 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
542 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
543 dst_output);
544 if (err > 0)
545 err = inet->recverr ? net_xmit_errno(err) : 0;
546 if (err)
547 goto error;
548 out:
549 return 0;
550
551 error_fault:
552 err = -EFAULT;
553 kfree_skb(skb);
554 error:
555 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
556 return err;
557 }
558
559 static void rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
560 {
561 struct iovec *iov;
562 u8 __user *type = NULL;
563 u8 __user *code = NULL;
564 int probed = 0;
565 int i;
566
567 if (!msg->msg_iov)
568 return;
569
570 for (i = 0; i < msg->msg_iovlen; i++) {
571 iov = &msg->msg_iov[i];
572 if (!iov)
573 continue;
574
575 switch (fl->proto) {
576 case IPPROTO_ICMPV6:
577 /* check if one-byte field is readable or not. */
578 if (iov->iov_base && iov->iov_len < 1)
579 break;
580
581 if (!type) {
582 type = iov->iov_base;
583 /* check if code field is readable or not. */
584 if (iov->iov_len > 1)
585 code = type + 1;
586 } else if (!code)
587 code = iov->iov_base;
588
589 if (type && code) {
590 get_user(fl->fl_icmp_type, type);
591 __get_user(fl->fl_icmp_code, code);
592 probed = 1;
593 }
594 break;
595 default:
596 probed = 1;
597 break;
598 }
599 if (probed)
600 break;
601 }
602 }
603
604 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
605 struct msghdr *msg, size_t len)
606 {
607 struct ipv6_txoptions opt_space;
608 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
609 struct in6_addr *daddr, *final_p = NULL, final;
610 struct inet_opt *inet = inet_sk(sk);
611 struct ipv6_pinfo *np = inet6_sk(sk);
612 struct raw6_opt *raw_opt = raw6_sk(sk);
613 struct ipv6_txoptions *opt = NULL;
614 struct ip6_flowlabel *flowlabel = NULL;
615 struct dst_entry *dst = NULL;
616 struct flowi fl;
617 int addr_len = msg->msg_namelen;
618 int hlimit = -1;
619 u16 proto;
620 int err;
621
622 /* Rough check on arithmetic overflow,
623 better check is made in ip6_build_xmit
624 */
625 if (len < 0)
626 return -EMSGSIZE;
627
628 /* Mirror BSD error message compatibility */
629 if (msg->msg_flags & MSG_OOB)
630 return -EOPNOTSUPP;
631
632 /*
633 * Get and verify the address.
634 */
635 memset(&fl, 0, sizeof(fl));
636
637 if (sin6) {
638 if (addr_len < SIN6_LEN_RFC2133)
639 return -EINVAL;
640
641 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
642 return(-EINVAL);
643
644 /* port is the proto value [0..255] carried in nexthdr */
645 proto = ntohs(sin6->sin6_port);
646
647 if (!proto)
648 proto = inet->num;
649 else if (proto != inet->num)
650 return(-EINVAL);
651
652 if (proto > 255)
653 return(-EINVAL);
654
655 daddr = &sin6->sin6_addr;
656 if (np->sndflow) {
657 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
658 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
659 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
660 if (flowlabel == NULL)
661 return -EINVAL;
662 daddr = &flowlabel->dst;
663 }
664 }
665
666 /*
667 * Otherwise it will be difficult to maintain
668 * sk->sk_dst_cache.
669 */
670 if (sk->sk_state == TCP_ESTABLISHED &&
671 !ipv6_addr_cmp(daddr, &np->daddr))
672 daddr = &np->daddr;
673
674 if (addr_len >= sizeof(struct sockaddr_in6) &&
675 sin6->sin6_scope_id &&
676 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
677 fl.oif = sin6->sin6_scope_id;
678 } else {
679 if (sk->sk_state != TCP_ESTABLISHED)
680 return -EDESTADDRREQ;
681
682 proto = inet->num;
683 daddr = &np->daddr;
684 fl.fl6_flowlabel = np->flow_label;
685 }
686
687 if (ipv6_addr_any(daddr)) {
688 /*
689 * unspecified destination address
690 * treated as error... is this correct ?
691 */
692 fl6_sock_release(flowlabel);
693 return(-EINVAL);
694 }
695
696 if (fl.oif == 0)
697 fl.oif = sk->sk_bound_dev_if;
698
699 if (msg->msg_controllen) {
700 opt = &opt_space;
701 memset(opt, 0, sizeof(struct ipv6_txoptions));
702 opt->tot_len = sizeof(struct ipv6_txoptions);
703
704 err = datagram_send_ctl(msg, &fl, opt, &hlimit);
705 if (err < 0) {
706 fl6_sock_release(flowlabel);
707 return err;
708 }
709 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
710 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
711 if (flowlabel == NULL)
712 return -EINVAL;
713 }
714 if (!(opt->opt_nflen|opt->opt_flen))
715 opt = NULL;
716 }
717 if (opt == NULL)
718 opt = np->opt;
719 if (flowlabel)
720 opt = fl6_merge_options(&opt_space, flowlabel, opt);
721
722 fl.proto = proto;
723 rawv6_probe_proto_opt(&fl, msg);
724
725 ipv6_addr_copy(&fl.fl6_dst, daddr);
726 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
727 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
728
729 /* merge ip6_build_xmit from ip6_output */
730 if (opt && opt->srcrt) {
731 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
732 ipv6_addr_copy(&final, &fl.fl6_dst);
733 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
734 final_p = &final;
735 }
736
737 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
738 fl.oif = np->mcast_oif;
739
740 err = ip6_dst_lookup(sk, &dst, &fl);
741 if (err)
742 goto out;
743 if (final_p)
744 ipv6_addr_copy(&fl.fl6_dst, final_p);
745
746 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
747 dst_release(dst);
748 goto out;
749 }
750
751 if (hlimit < 0) {
752 if (ipv6_addr_is_multicast(&fl.fl6_dst))
753 hlimit = np->mcast_hops;
754 else
755 hlimit = np->hop_limit;
756 if (hlimit < 0)
757 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
758 }
759
760 if (msg->msg_flags&MSG_CONFIRM)
761 goto do_confirm;
762
763 back_from_confirm:
764 if (inet->hdrincl) {
765 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
766 } else {
767 lock_sock(sk);
768 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
769 hlimit, opt, &fl, (struct rt6_info*)dst, msg->msg_flags);
770
771 if (err)
772 ip6_flush_pending_frames(sk);
773 else if (!(msg->msg_flags & MSG_MORE))
774 err = rawv6_push_pending_frames(sk, &fl, raw_opt, len);
775 }
776 done:
777 ip6_dst_store(sk, dst,
778 !ipv6_addr_cmp(&fl.fl6_dst, &np->daddr) ?
779 &np->daddr : NULL);
780 if (err > 0)
781 err = np->recverr ? net_xmit_errno(err) : 0;
782
783 release_sock(sk);
784 out:
785 fl6_sock_release(flowlabel);
786 return err<0?err:len;
787 do_confirm:
788 dst_confirm(dst);
789 if (!(msg->msg_flags & MSG_PROBE) || len)
790 goto back_from_confirm;
791 err = 0;
792 goto done;
793 }
794
795 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
796 char __user *optval, int optlen)
797 {
798 switch (optname) {
799 case ICMPV6_FILTER:
800 if (optlen > sizeof(struct icmp6_filter))
801 optlen = sizeof(struct icmp6_filter);
802 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
803 return -EFAULT;
804 return 0;
805 default:
806 return -ENOPROTOOPT;
807 };
808
809 return 0;
810 }
811
812 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
813 char __user *optval, int __user *optlen)
814 {
815 int len;
816
817 switch (optname) {
818 case ICMPV6_FILTER:
819 if (get_user(len, optlen))
820 return -EFAULT;
821 if (len < 0)
822 return -EINVAL;
823 if (len > sizeof(struct icmp6_filter))
824 len = sizeof(struct icmp6_filter);
825 if (put_user(len, optlen))
826 return -EFAULT;
827 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
828 return -EFAULT;
829 return 0;
830 default:
831 return -ENOPROTOOPT;
832 };
833
834 return 0;
835 }
836
837
838 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
839 char __user *optval, int optlen)
840 {
841 struct raw6_opt *opt = raw6_sk(sk);
842 int val;
843
844 switch(level) {
845 case SOL_RAW:
846 break;
847
848 case SOL_ICMPV6:
849 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
850 return -EOPNOTSUPP;
851 return rawv6_seticmpfilter(sk, level, optname, optval,
852 optlen);
853 case SOL_IPV6:
854 if (optname == IPV6_CHECKSUM)
855 break;
856 default:
857 return ipv6_setsockopt(sk, level, optname, optval,
858 optlen);
859 };
860
861 if (get_user(val, (int __user *)optval))
862 return -EFAULT;
863
864 switch (optname) {
865 case IPV6_CHECKSUM:
866 /* You may get strange result with a positive odd offset;
867 RFC2292bis agrees with me. */
868 if (val > 0 && (val&1))
869 return(-EINVAL);
870 if (val < 0) {
871 opt->checksum = 0;
872 } else {
873 opt->checksum = 1;
874 opt->offset = val;
875 }
876
877 return 0;
878 break;
879
880 default:
881 return(-ENOPROTOOPT);
882 }
883 }
884
885 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
886 char __user *optval, int __user *optlen)
887 {
888 struct raw6_opt *opt = raw6_sk(sk);
889 int val, len;
890
891 switch(level) {
892 case SOL_RAW:
893 break;
894
895 case SOL_ICMPV6:
896 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
897 return -EOPNOTSUPP;
898 return rawv6_geticmpfilter(sk, level, optname, optval,
899 optlen);
900 case SOL_IPV6:
901 if (optname == IPV6_CHECKSUM)
902 break;
903 default:
904 return ipv6_getsockopt(sk, level, optname, optval,
905 optlen);
906 };
907
908 if (get_user(len,optlen))
909 return -EFAULT;
910
911 switch (optname) {
912 case IPV6_CHECKSUM:
913 if (opt->checksum == 0)
914 val = -1;
915 else
916 val = opt->offset;
917 break;
918
919 default:
920 return -ENOPROTOOPT;
921 }
922
923 len = min_t(unsigned int, sizeof(int), len);
924
925 if (put_user(len, optlen))
926 return -EFAULT;
927 if (copy_to_user(optval,&val,len))
928 return -EFAULT;
929 return 0;
930 }
931
932 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
933 {
934 switch(cmd) {
935 case SIOCOUTQ:
936 {
937 int amount = atomic_read(&sk->sk_wmem_alloc);
938 return put_user(amount, (int __user *)arg);
939 }
940 case SIOCINQ:
941 {
942 struct sk_buff *skb;
943 int amount = 0;
944
945 spin_lock_irq(&sk->sk_receive_queue.lock);
946 skb = skb_peek(&sk->sk_receive_queue);
947 if (skb != NULL)
948 amount = skb->tail - skb->h.raw;
949 spin_unlock_irq(&sk->sk_receive_queue.lock);
950 return put_user(amount, (int __user *)arg);
951 }
952
953 default:
954 return -ENOIOCTLCMD;
955 }
956 }
957
958 static void rawv6_close(struct sock *sk, long timeout)
959 {
960 if (inet_sk(sk)->num == IPPROTO_RAW)
961 ip6_ra_control(sk, -1, NULL);
962
963 sk_common_release(sk);
964 }
965
966 static int rawv6_init_sk(struct sock *sk)
967 {
968 if (inet_sk(sk)->num == IPPROTO_ICMPV6) {
969 struct raw6_opt *opt = raw6_sk(sk);
970 opt->checksum = 1;
971 opt->offset = 2;
972 }
973 return(0);
974 }
975
976 struct proto rawv6_prot = {
977 .name = "RAW",
978 .close = rawv6_close,
979 .connect = ip6_datagram_connect,
980 .disconnect = udp_disconnect,
981 .ioctl = rawv6_ioctl,
982 .init = rawv6_init_sk,
983 .destroy = inet6_destroy_sock,
984 .setsockopt = rawv6_setsockopt,
985 .getsockopt = rawv6_getsockopt,
986 .sendmsg = rawv6_sendmsg,
987 .recvmsg = rawv6_recvmsg,
988 .bind = rawv6_bind,
989 .backlog_rcv = rawv6_rcv_skb,
990 .hash = raw_v6_hash,
991 .unhash = raw_v6_unhash,
992 .slab_obj_size = sizeof(struct raw6_sock),
993 };
994
995 #ifdef CONFIG_PROC_FS
996 struct raw6_iter_state {
997 int bucket;
998 };
999
1000 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1001
1002 static struct sock *raw6_get_first(struct seq_file *seq)
1003 {
1004 struct sock *sk;
1005 struct hlist_node *node;
1006 struct raw6_iter_state* state = raw6_seq_private(seq);
1007
1008 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1009 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1010 if (sk->sk_family == PF_INET6)
1011 goto out;
1012 sk = NULL;
1013 out:
1014 return sk;
1015 }
1016
1017 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1018 {
1019 struct raw6_iter_state* state = raw6_seq_private(seq);
1020
1021 do {
1022 sk = sk_next(sk);
1023 try_again:
1024 ;
1025 } while (sk && sk->sk_family != PF_INET6);
1026
1027 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1028 sk = sk_head(&raw_v6_htable[state->bucket]);
1029 goto try_again;
1030 }
1031 return sk;
1032 }
1033
1034 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1035 {
1036 struct sock *sk = raw6_get_first(seq);
1037 if (sk)
1038 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1039 --pos;
1040 return pos ? NULL : sk;
1041 }
1042
1043 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1044 {
1045 read_lock(&raw_v6_lock);
1046 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1047 }
1048
1049 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1050 {
1051 struct sock *sk;
1052
1053 if (v == SEQ_START_TOKEN)
1054 sk = raw6_get_first(seq);
1055 else
1056 sk = raw6_get_next(seq, v);
1057 ++*pos;
1058 return sk;
1059 }
1060
1061 static void raw6_seq_stop(struct seq_file *seq, void *v)
1062 {
1063 read_unlock(&raw_v6_lock);
1064 }
1065
1066 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1067 {
1068 struct ipv6_pinfo *np = inet6_sk(sp);
1069 struct in6_addr *dest, *src;
1070 __u16 destp, srcp;
1071
1072 dest = &np->daddr;
1073 src = &np->rcv_saddr;
1074 destp = 0;
1075 srcp = inet_sk(sp)->num;
1076 seq_printf(seq,
1077 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1078 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1079 i,
1080 src->s6_addr32[0], src->s6_addr32[1],
1081 src->s6_addr32[2], src->s6_addr32[3], srcp,
1082 dest->s6_addr32[0], dest->s6_addr32[1],
1083 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1084 sp->sk_state,
1085 atomic_read(&sp->sk_wmem_alloc),
1086 atomic_read(&sp->sk_rmem_alloc),
1087 0, 0L, 0,
1088 sock_i_uid(sp), 0,
1089 sock_i_ino(sp),
1090 atomic_read(&sp->sk_refcnt), sp);
1091 }
1092
1093 static int raw6_seq_show(struct seq_file *seq, void *v)
1094 {
1095 if (v == SEQ_START_TOKEN)
1096 seq_printf(seq,
1097 " sl "
1098 "local_address "
1099 "remote_address "
1100 "st tx_queue rx_queue tr tm->when retrnsmt"
1101 " uid timeout inode\n");
1102 else
1103 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1104 return 0;
1105 }
1106
1107 static struct seq_operations raw6_seq_ops = {
1108 .start = raw6_seq_start,
1109 .next = raw6_seq_next,
1110 .stop = raw6_seq_stop,
1111 .show = raw6_seq_show,
1112 };
1113
1114 static int raw6_seq_open(struct inode *inode, struct file *file)
1115 {
1116 struct seq_file *seq;
1117 int rc = -ENOMEM;
1118 struct raw6_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
1119 if (!s)
1120 goto out;
1121 rc = seq_open(file, &raw6_seq_ops);
1122 if (rc)
1123 goto out_kfree;
1124 seq = file->private_data;
1125 seq->private = s;
1126 memset(s, 0, sizeof(*s));
1127 out:
1128 return rc;
1129 out_kfree:
1130 kfree(s);
1131 goto out;
1132 }
1133
1134 static struct file_operations raw6_seq_fops = {
1135 .owner = THIS_MODULE,
1136 .open = raw6_seq_open,
1137 .read = seq_read,
1138 .llseek = seq_lseek,
1139 .release = seq_release_private,
1140 };
1141
1142 int __init raw6_proc_init(void)
1143 {
1144 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1145 return -ENOMEM;
1146 return 0;
1147 }
1148
1149 void raw6_proc_exit(void)
1150 {
1151 proc_net_remove("raw6");
1152 }
1153 #endif /* CONFIG_PROC_FS */
MOXA 2.6.9 from sdlinux-moxaart.tgz