?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[NET_SCHED]: Remove old estimator implementation
[hh.org.git] / net / ipv6 / raw.c
blobd09329ca32670bdb76639826bfdfa53be41f1ae0
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 <linux/skbuff.h>
36 #include <asm/uaccess.h>
37 #include <asm/ioctls.h>
38
39 #include <net/ip.h>
40 #include <net/sock.h>
41 #include <net/snmp.h>
42
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/ip6_route.h>
47 #include <net/ip6_checksum.h>
48 #include <net/addrconf.h>
49 #include <net/transp_v6.h>
50 #include <net/udp.h>
51 #include <net/inet_common.h>
52 #include <net/tcp_states.h>
53 #ifdef CONFIG_IPV6_MIP6
54 #include <net/mip6.h>
55 #endif
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 = &skb->nh.ipv6h->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 __u32 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, u32 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->nh.raw,
366 skb->h.raw - skb->nh.raw);
367 if (!csum_ipv6_magic(&skb->nh.ipv6h->saddr,
368 &skb->nh.ipv6h->daddr,
369 skb->len, inet->num, skb->csum))
370 skb->ip_summed = CHECKSUM_UNNECESSARY;
371 }
372 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
373 skb->csum = ~csum_ipv6_magic(&skb->nh.ipv6h->saddr,
374 &skb->nh.ipv6h->daddr,
375 skb->len, inet->num, 0);
376
377 if (inet->hdrincl) {
378 if (skb_checksum_complete(skb)) {
379 /* FIXME: increment a raw6 drops counter here */
380 kfree_skb(skb);
381 return 0;
382 }
383 }
384
385 rawv6_rcv_skb(sk, skb);
386 return 0;
387 }
388
389
390 /*
391 * This should be easy, if there is something there
392 * we return it, otherwise we block.
393 */
394
395 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
396 struct msghdr *msg, size_t len,
397 int noblock, int flags, int *addr_len)
398 {
399 struct ipv6_pinfo *np = inet6_sk(sk);
400 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
401 struct sk_buff *skb;
402 size_t copied;
403 int err;
404
405 if (flags & MSG_OOB)
406 return -EOPNOTSUPP;
407
408 if (addr_len)
409 *addr_len=sizeof(*sin6);
410
411 if (flags & MSG_ERRQUEUE)
412 return ipv6_recv_error(sk, msg, len);
413
414 skb = skb_recv_datagram(sk, flags, noblock, &err);
415 if (!skb)
416 goto out;
417
418 copied = skb->len;
419 if (copied > len) {
420 copied = len;
421 msg->msg_flags |= MSG_TRUNC;
422 }
423
424 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
425 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
426 } else if (msg->msg_flags&MSG_TRUNC) {
427 if (__skb_checksum_complete(skb))
428 goto csum_copy_err;
429 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
430 } else {
431 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
432 if (err == -EINVAL)
433 goto csum_copy_err;
434 }
435 if (err)
436 goto out_free;
437
438 /* Copy the address. */
439 if (sin6) {
440 sin6->sin6_family = AF_INET6;
441 sin6->sin6_port = 0;
442 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
443 sin6->sin6_flowinfo = 0;
444 sin6->sin6_scope_id = 0;
445 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
446 sin6->sin6_scope_id = IP6CB(skb)->iif;
447 }
448
449 sock_recv_timestamp(msg, sk, skb);
450
451 if (np->rxopt.all)
452 datagram_recv_ctl(sk, msg, skb);
453
454 err = copied;
455 if (flags & MSG_TRUNC)
456 err = skb->len;
457
458 out_free:
459 skb_free_datagram(sk, skb);
460 out:
461 return err;
462
463 csum_copy_err:
464 skb_kill_datagram(sk, skb, flags);
465
466 /* Error for blocking case is chosen to masquerade
467 as some normal condition.
468 */
469 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
470 /* FIXME: increment a raw6 drops counter here */
471 goto out;
472 }
473
474 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
475 struct raw6_sock *rp)
476 {
477 struct sk_buff *skb;
478 int err = 0;
479 int offset;
480 int len;
481 int total_len;
482 u32 tmp_csum;
483 u16 csum;
484
485 if (!rp->checksum)
486 goto send;
487
488 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
489 goto out;
490
491 offset = rp->offset;
492 total_len = inet_sk(sk)->cork.length - (skb->nh.raw - skb->data);
493 if (offset >= total_len - 1) {
494 err = -EINVAL;
495 ip6_flush_pending_frames(sk);
496 goto out;
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);
534
535 tmp_csum = csum_ipv6_magic(&fl->fl6_src,
536 &fl->fl6_dst,
537 total_len, fl->proto, tmp_csum);
538
539 if (tmp_csum == 0)
540 tmp_csum = -1;
541
542 csum = tmp_csum;
543 if (skb_store_bits(skb, offset, &csum, 2))
544 BUG();
545
546 send:
547 err = ip6_push_pending_frames(sk);
548 out:
549 return err;
550 }
551
552 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
553 struct flowi *fl, struct rt6_info *rt,
554 unsigned int flags)
555 {
556 struct ipv6_pinfo *np = inet6_sk(sk);
557 struct ipv6hdr *iph;
558 struct sk_buff *skb;
559 unsigned int hh_len;
560 int err;
561
562 if (length > rt->u.dst.dev->mtu) {
563 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
564 return -EMSGSIZE;
565 }
566 if (flags&MSG_PROBE)
567 goto out;
568
569 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
570
571 skb = sock_alloc_send_skb(sk, length+hh_len+15,
572 flags&MSG_DONTWAIT, &err);
573 if (skb == NULL)
574 goto error;
575 skb_reserve(skb, hh_len);
576
577 skb->priority = sk->sk_priority;
578 skb->dst = dst_clone(&rt->u.dst);
579
580 skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length);
581
582 skb->ip_summed = CHECKSUM_NONE;
583
584 skb->h.raw = skb->nh.raw;
585 err = memcpy_fromiovecend((void *)iph, from, 0, length);
586 if (err)
587 goto error_fault;
588
589 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
590 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
591 dst_output);
592 if (err > 0)
593 err = np->recverr ? net_xmit_errno(err) : 0;
594 if (err)
595 goto error;
596 out:
597 return 0;
598
599 error_fault:
600 err = -EFAULT;
601 kfree_skb(skb);
602 error:
603 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
604 return err;
605 }
606
607 static void rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
608 {
609 struct iovec *iov;
610 u8 __user *type = NULL;
611 u8 __user *code = NULL;
612 #ifdef CONFIG_IPV6_MIP6
613 u8 len = 0;
614 #endif
615 int probed = 0;
616 int i;
617
618 if (!msg->msg_iov)
619 return;
620
621 for (i = 0; i < msg->msg_iovlen; i++) {
622 iov = &msg->msg_iov[i];
623 if (!iov)
624 continue;
625
626 switch (fl->proto) {
627 case IPPROTO_ICMPV6:
628 /* check if one-byte field is readable or not. */
629 if (iov->iov_base && iov->iov_len < 1)
630 break;
631
632 if (!type) {
633 type = iov->iov_base;
634 /* check if code field is readable or not. */
635 if (iov->iov_len > 1)
636 code = type + 1;
637 } else if (!code)
638 code = iov->iov_base;
639
640 if (type && code) {
641 get_user(fl->fl_icmp_type, type);
642 get_user(fl->fl_icmp_code, code);
643 probed = 1;
644 }
645 break;
646 #ifdef CONFIG_IPV6_MIP6
647 case IPPROTO_MH:
648 if (iov->iov_base && iov->iov_len < 1)
649 break;
650 /* check if type field is readable or not. */
651 if (iov->iov_len > 2 - len) {
652 u8 __user *p = iov->iov_base;
653 get_user(fl->fl_mh_type, &p[2 - len]);
654 probed = 1;
655 } else
656 len += iov->iov_len;
657
658 break;
659 #endif
660 default:
661 probed = 1;
662 break;
663 }
664 if (probed)
665 break;
666 }
667 }
668
669 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
670 struct msghdr *msg, size_t len)
671 {
672 struct ipv6_txoptions opt_space;
673 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
674 struct in6_addr *daddr, *final_p = NULL, final;
675 struct inet_sock *inet = inet_sk(sk);
676 struct ipv6_pinfo *np = inet6_sk(sk);
677 struct raw6_sock *rp = raw6_sk(sk);
678 struct ipv6_txoptions *opt = NULL;
679 struct ip6_flowlabel *flowlabel = NULL;
680 struct dst_entry *dst = NULL;
681 struct flowi fl;
682 int addr_len = msg->msg_namelen;
683 int hlimit = -1;
684 int tclass = -1;
685 u16 proto;
686 int err;
687
688 /* Rough check on arithmetic overflow,
689 better check is made in ip6_build_xmit
690 */
691 if (len < 0)
692 return -EMSGSIZE;
693
694 /* Mirror BSD error message compatibility */
695 if (msg->msg_flags & MSG_OOB)
696 return -EOPNOTSUPP;
697
698 /*
699 * Get and verify the address.
700 */
701 memset(&fl, 0, sizeof(fl));
702
703 if (sin6) {
704 if (addr_len < SIN6_LEN_RFC2133)
705 return -EINVAL;
706
707 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
708 return(-EAFNOSUPPORT);
709
710 /* port is the proto value [0..255] carried in nexthdr */
711 proto = ntohs(sin6->sin6_port);
712
713 if (!proto)
714 proto = inet->num;
715 else if (proto != inet->num)
716 return(-EINVAL);
717
718 if (proto > 255)
719 return(-EINVAL);
720
721 daddr = &sin6->sin6_addr;
722 if (np->sndflow) {
723 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
724 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
725 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
726 if (flowlabel == NULL)
727 return -EINVAL;
728 daddr = &flowlabel->dst;
729 }
730 }
731
732 /*
733 * Otherwise it will be difficult to maintain
734 * sk->sk_dst_cache.
735 */
736 if (sk->sk_state == TCP_ESTABLISHED &&
737 ipv6_addr_equal(daddr, &np->daddr))
738 daddr = &np->daddr;
739
740 if (addr_len >= sizeof(struct sockaddr_in6) &&
741 sin6->sin6_scope_id &&
742 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
743 fl.oif = sin6->sin6_scope_id;
744 } else {
745 if (sk->sk_state != TCP_ESTABLISHED)
746 return -EDESTADDRREQ;
747
748 proto = inet->num;
749 daddr = &np->daddr;
750 fl.fl6_flowlabel = np->flow_label;
751 }
752
753 if (ipv6_addr_any(daddr)) {
754 /*
755 * unspecified destination address
756 * treated as error... is this correct ?
757 */
758 fl6_sock_release(flowlabel);
759 return(-EINVAL);
760 }
761
762 if (fl.oif == 0)
763 fl.oif = sk->sk_bound_dev_if;
764
765 if (msg->msg_controllen) {
766 opt = &opt_space;
767 memset(opt, 0, sizeof(struct ipv6_txoptions));
768 opt->tot_len = sizeof(struct ipv6_txoptions);
769
770 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
771 if (err < 0) {
772 fl6_sock_release(flowlabel);
773 return err;
774 }
775 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
776 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
777 if (flowlabel == NULL)
778 return -EINVAL;
779 }
780 if (!(opt->opt_nflen|opt->opt_flen))
781 opt = NULL;
782 }
783 if (opt == NULL)
784 opt = np->opt;
785 if (flowlabel)
786 opt = fl6_merge_options(&opt_space, flowlabel, opt);
787 opt = ipv6_fixup_options(&opt_space, opt);
788
789 fl.proto = proto;
790 rawv6_probe_proto_opt(&fl, msg);
791
792 ipv6_addr_copy(&fl.fl6_dst, daddr);
793 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
794 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
795
796 /* merge ip6_build_xmit from ip6_output */
797 if (opt && opt->srcrt) {
798 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
799 ipv6_addr_copy(&final, &fl.fl6_dst);
800 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
801 final_p = &final;
802 }
803
804 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
805 fl.oif = np->mcast_oif;
806 security_sk_classify_flow(sk, &fl);
807
808 err = ip6_dst_lookup(sk, &dst, &fl);
809 if (err)
810 goto out;
811 if (final_p)
812 ipv6_addr_copy(&fl.fl6_dst, final_p);
813
814 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
815 goto out;
816
817 if (hlimit < 0) {
818 if (ipv6_addr_is_multicast(&fl.fl6_dst))
819 hlimit = np->mcast_hops;
820 else
821 hlimit = np->hop_limit;
822 if (hlimit < 0)
823 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
824 if (hlimit < 0)
825 hlimit = ipv6_get_hoplimit(dst->dev);
826 }
827
828 if (tclass < 0) {
829 tclass = np->tclass;
830 if (tclass < 0)
831 tclass = 0;
832 }
833
834 if (msg->msg_flags&MSG_CONFIRM)
835 goto do_confirm;
836
837 back_from_confirm:
838 if (inet->hdrincl) {
839 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
840 } else {
841 lock_sock(sk);
842 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
843 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
844 msg->msg_flags);
845
846 if (err)
847 ip6_flush_pending_frames(sk);
848 else if (!(msg->msg_flags & MSG_MORE))
849 err = rawv6_push_pending_frames(sk, &fl, rp);
850 }
851 done:
852 dst_release(dst);
853 release_sock(sk);
854 out:
855 fl6_sock_release(flowlabel);
856 return err<0?err:len;
857 do_confirm:
858 dst_confirm(dst);
859 if (!(msg->msg_flags & MSG_PROBE) || len)
860 goto back_from_confirm;
861 err = 0;
862 goto done;
863 }
864
865 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
866 char __user *optval, int optlen)
867 {
868 switch (optname) {
869 case ICMPV6_FILTER:
870 if (optlen > sizeof(struct icmp6_filter))
871 optlen = sizeof(struct icmp6_filter);
872 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
873 return -EFAULT;
874 return 0;
875 default:
876 return -ENOPROTOOPT;
877 };
878
879 return 0;
880 }
881
882 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
883 char __user *optval, int __user *optlen)
884 {
885 int len;
886
887 switch (optname) {
888 case ICMPV6_FILTER:
889 if (get_user(len, optlen))
890 return -EFAULT;
891 if (len < 0)
892 return -EINVAL;
893 if (len > sizeof(struct icmp6_filter))
894 len = sizeof(struct icmp6_filter);
895 if (put_user(len, optlen))
896 return -EFAULT;
897 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
898 return -EFAULT;
899 return 0;
900 default:
901 return -ENOPROTOOPT;
902 };
903
904 return 0;
905 }
906
907
908 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
909 char __user *optval, int optlen)
910 {
911 struct raw6_sock *rp = raw6_sk(sk);
912 int val;
913
914 if (get_user(val, (int __user *)optval))
915 return -EFAULT;
916
917 switch (optname) {
918 case IPV6_CHECKSUM:
919 /* You may get strange result with a positive odd offset;
920 RFC2292bis agrees with me. */
921 if (val > 0 && (val&1))
922 return(-EINVAL);
923 if (val < 0) {
924 rp->checksum = 0;
925 } else {
926 rp->checksum = 1;
927 rp->offset = val;
928 }
929
930 return 0;
931 break;
932
933 default:
934 return(-ENOPROTOOPT);
935 }
936 }
937
938 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
939 char __user *optval, int optlen)
940 {
941 switch(level) {
942 case SOL_RAW:
943 break;
944
945 case SOL_ICMPV6:
946 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
947 return -EOPNOTSUPP;
948 return rawv6_seticmpfilter(sk, level, optname, optval,
949 optlen);
950 case SOL_IPV6:
951 if (optname == IPV6_CHECKSUM)
952 break;
953 default:
954 return ipv6_setsockopt(sk, level, optname, optval,
955 optlen);
956 };
957 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
958 }
959
960 #ifdef CONFIG_COMPAT
961 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
962 char __user *optval, int optlen)
963 {
964 switch (level) {
965 case SOL_RAW:
966 break;
967 case SOL_ICMPV6:
968 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
969 return -EOPNOTSUPP;
970 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
971 case SOL_IPV6:
972 if (optname == IPV6_CHECKSUM)
973 break;
974 default:
975 return compat_ipv6_setsockopt(sk, level, optname,
976 optval, optlen);
977 };
978 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
979 }
980 #endif
981
982 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
983 char __user *optval, int __user *optlen)
984 {
985 struct raw6_sock *rp = raw6_sk(sk);
986 int val, len;
987
988 if (get_user(len,optlen))
989 return -EFAULT;
990
991 switch (optname) {
992 case IPV6_CHECKSUM:
993 if (rp->checksum == 0)
994 val = -1;
995 else
996 val = rp->offset;
997 break;
998
999 default:
1000 return -ENOPROTOOPT;
1001 }
1002
1003 len = min_t(unsigned int, sizeof(int), len);
1004
1005 if (put_user(len, optlen))
1006 return -EFAULT;
1007 if (copy_to_user(optval,&val,len))
1008 return -EFAULT;
1009 return 0;
1010 }
1011
1012 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1013 char __user *optval, int __user *optlen)
1014 {
1015 switch(level) {
1016 case SOL_RAW:
1017 break;
1018
1019 case SOL_ICMPV6:
1020 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1021 return -EOPNOTSUPP;
1022 return rawv6_geticmpfilter(sk, level, optname, optval,
1023 optlen);
1024 case SOL_IPV6:
1025 if (optname == IPV6_CHECKSUM)
1026 break;
1027 default:
1028 return ipv6_getsockopt(sk, level, optname, optval,
1029 optlen);
1030 };
1031 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1032 }
1033
1034 #ifdef CONFIG_COMPAT
1035 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1036 char __user *optval, int __user *optlen)
1037 {
1038 switch (level) {
1039 case SOL_RAW:
1040 break;
1041 case SOL_ICMPV6:
1042 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1043 return -EOPNOTSUPP;
1044 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1045 case SOL_IPV6:
1046 if (optname == IPV6_CHECKSUM)
1047 break;
1048 default:
1049 return compat_ipv6_getsockopt(sk, level, optname,
1050 optval, optlen);
1051 };
1052 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1053 }
1054 #endif
1055
1056 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1057 {
1058 switch(cmd) {
1059 case SIOCOUTQ:
1060 {
1061 int amount = atomic_read(&sk->sk_wmem_alloc);
1062 return put_user(amount, (int __user *)arg);
1063 }
1064 case SIOCINQ:
1065 {
1066 struct sk_buff *skb;
1067 int amount = 0;
1068
1069 spin_lock_bh(&sk->sk_receive_queue.lock);
1070 skb = skb_peek(&sk->sk_receive_queue);
1071 if (skb != NULL)
1072 amount = skb->tail - skb->h.raw;
1073 spin_unlock_bh(&sk->sk_receive_queue.lock);
1074 return put_user(amount, (int __user *)arg);
1075 }
1076
1077 default:
1078 return -ENOIOCTLCMD;
1079 }
1080 }
1081
1082 static void rawv6_close(struct sock *sk, long timeout)
1083 {
1084 if (inet_sk(sk)->num == IPPROTO_RAW)
1085 ip6_ra_control(sk, -1, NULL);
1086
1087 sk_common_release(sk);
1088 }
1089
1090 static int rawv6_init_sk(struct sock *sk)
1091 {
1092 if (inet_sk(sk)->num == IPPROTO_ICMPV6) {
1093 struct raw6_sock *rp = raw6_sk(sk);
1094 rp->checksum = 1;
1095 rp->offset = 2;
1096 }
1097 return(0);
1098 }
1099
1100 struct proto rawv6_prot = {
1101 .name = "RAWv6",
1102 .owner = THIS_MODULE,
1103 .close = rawv6_close,
1104 .connect = ip6_datagram_connect,
1105 .disconnect = udp_disconnect,
1106 .ioctl = rawv6_ioctl,
1107 .init = rawv6_init_sk,
1108 .destroy = inet6_destroy_sock,
1109 .setsockopt = rawv6_setsockopt,
1110 .getsockopt = rawv6_getsockopt,
1111 .sendmsg = rawv6_sendmsg,
1112 .recvmsg = rawv6_recvmsg,
1113 .bind = rawv6_bind,
1114 .backlog_rcv = rawv6_rcv_skb,
1115 .hash = raw_v6_hash,
1116 .unhash = raw_v6_unhash,
1117 .obj_size = sizeof(struct raw6_sock),
1118 #ifdef CONFIG_COMPAT
1119 .compat_setsockopt = compat_rawv6_setsockopt,
1120 .compat_getsockopt = compat_rawv6_getsockopt,
1121 #endif
1122 };
1123
1124 #ifdef CONFIG_PROC_FS
1125 struct raw6_iter_state {
1126 int bucket;
1127 };
1128
1129 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1130
1131 static struct sock *raw6_get_first(struct seq_file *seq)
1132 {
1133 struct sock *sk;
1134 struct hlist_node *node;
1135 struct raw6_iter_state* state = raw6_seq_private(seq);
1136
1137 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1138 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1139 if (sk->sk_family == PF_INET6)
1140 goto out;
1141 sk = NULL;
1142 out:
1143 return sk;
1144 }
1145
1146 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1147 {
1148 struct raw6_iter_state* state = raw6_seq_private(seq);
1149
1150 do {
1151 sk = sk_next(sk);
1152 try_again:
1153 ;
1154 } while (sk && sk->sk_family != PF_INET6);
1155
1156 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1157 sk = sk_head(&raw_v6_htable[state->bucket]);
1158 goto try_again;
1159 }
1160 return sk;
1161 }
1162
1163 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1164 {
1165 struct sock *sk = raw6_get_first(seq);
1166 if (sk)
1167 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1168 --pos;
1169 return pos ? NULL : sk;
1170 }
1171
1172 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1173 {
1174 read_lock(&raw_v6_lock);
1175 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1176 }
1177
1178 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1179 {
1180 struct sock *sk;
1181
1182 if (v == SEQ_START_TOKEN)
1183 sk = raw6_get_first(seq);
1184 else
1185 sk = raw6_get_next(seq, v);
1186 ++*pos;
1187 return sk;
1188 }
1189
1190 static void raw6_seq_stop(struct seq_file *seq, void *v)
1191 {
1192 read_unlock(&raw_v6_lock);
1193 }
1194
1195 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1196 {
1197 struct ipv6_pinfo *np = inet6_sk(sp);
1198 struct in6_addr *dest, *src;
1199 __u16 destp, srcp;
1200
1201 dest = &np->daddr;
1202 src = &np->rcv_saddr;
1203 destp = 0;
1204 srcp = inet_sk(sp)->num;
1205 seq_printf(seq,
1206 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1207 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1208 i,
1209 src->s6_addr32[0], src->s6_addr32[1],
1210 src->s6_addr32[2], src->s6_addr32[3], srcp,
1211 dest->s6_addr32[0], dest->s6_addr32[1],
1212 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1213 sp->sk_state,
1214 atomic_read(&sp->sk_wmem_alloc),
1215 atomic_read(&sp->sk_rmem_alloc),
1216 0, 0L, 0,
1217 sock_i_uid(sp), 0,
1218 sock_i_ino(sp),
1219 atomic_read(&sp->sk_refcnt), sp);
1220 }
1221
1222 static int raw6_seq_show(struct seq_file *seq, void *v)
1223 {
1224 if (v == SEQ_START_TOKEN)
1225 seq_printf(seq,
1226 " sl "
1227 "local_address "
1228 "remote_address "
1229 "st tx_queue rx_queue tr tm->when retrnsmt"
1230 " uid timeout inode\n");
1231 else
1232 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1233 return 0;
1234 }
1235
1236 static struct seq_operations raw6_seq_ops = {
1237 .start = raw6_seq_start,
1238 .next = raw6_seq_next,
1239 .stop = raw6_seq_stop,
1240 .show = raw6_seq_show,
1241 };
1242
1243 static int raw6_seq_open(struct inode *inode, struct file *file)
1244 {
1245 struct seq_file *seq;
1246 int rc = -ENOMEM;
1247 struct raw6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
1248 if (!s)
1249 goto out;
1250 rc = seq_open(file, &raw6_seq_ops);
1251 if (rc)
1252 goto out_kfree;
1253 seq = file->private_data;
1254 seq->private = s;
1255 out:
1256 return rc;
1257 out_kfree:
1258 kfree(s);
1259 goto out;
1260 }
1261
1262 static struct file_operations raw6_seq_fops = {
1263 .owner = THIS_MODULE,
1264 .open = raw6_seq_open,
1265 .read = seq_read,
1266 .llseek = seq_lseek,
1267 .release = seq_release_private,
1268 };
1269
1270 int __init raw6_proc_init(void)
1271 {
1272 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1273 return -ENOMEM;
1274 return 0;
1275 }
1276
1277 void raw6_proc_exit(void)
1278 {
1279 proc_net_remove("raw6");
1280 }
1281 #endif /* CONFIG_PROC_FS */
hh.org GIT migration