projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
[deliverable/linux.git] / net / ipv6 / raw.c
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 #include <asm/bug.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
53 #include <net/rawv6.h>
54 #include <net/xfrm.h>
55
56 #include <linux/proc_fs.h>
57 #include <linux/seq_file.h>
58
59 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
60 DEFINE_RWLOCK(raw_v6_lock);
61
62 static void raw_v6_hash(struct sock *sk)
63 {
64 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num &
65 (RAWV6_HTABLE_SIZE - 1)];
66
67 write_lock_bh(&raw_v6_lock);
68 sk_add_node(sk, list);
69 sock_prot_inc_use(sk->sk_prot);
70 write_unlock_bh(&raw_v6_lock);
71 }
72
73 static void raw_v6_unhash(struct sock *sk)
74 {
75 write_lock_bh(&raw_v6_lock);
76 if (sk_del_node_init(sk))
77 sock_prot_dec_use(sk->sk_prot);
78 write_unlock_bh(&raw_v6_lock);
79 }
80
81
82 /* Grumble... icmp and ip_input want to get at this... */
83 struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
84 struct in6_addr *loc_addr, struct in6_addr *rmt_addr)
85 {
86 struct hlist_node *node;
87 int is_multicast = ipv6_addr_is_multicast(loc_addr);
88
89 sk_for_each_from(sk, node)
90 if (inet_sk(sk)->num == num) {
91 struct ipv6_pinfo *np = inet6_sk(sk);
92
93 if (!ipv6_addr_any(&np->daddr) &&
94 !ipv6_addr_equal(&np->daddr, rmt_addr))
95 continue;
96
97 if (!ipv6_addr_any(&np->rcv_saddr)) {
98 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
99 goto found;
100 if (is_multicast &&
101 inet6_mc_check(sk, loc_addr, rmt_addr))
102 goto found;
103 continue;
104 }
105 goto found;
106 }
107 sk = NULL;
108 found:
109 return sk;
110 }
111
112 /*
113 * 0 - deliver
114 * 1 - block
115 */
116 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
117 {
118 struct icmp6hdr *icmph;
119 struct raw6_sock *rp = raw6_sk(sk);
120
121 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
122 __u32 *data = &rp->filter.data[0];
123 int bit_nr;
124
125 icmph = (struct icmp6hdr *) skb->data;
126 bit_nr = icmph->icmp6_type;
127
128 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
129 }
130 return 0;
131 }
132
133 /*
134 * demultiplex raw sockets.
135 * (should consider queueing the skb in the sock receive_queue
136 * without calling rawv6.c)
137 *
138 * Caller owns SKB so we must make clones.
139 */
140 void ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
141 {
142 struct in6_addr *saddr;
143 struct in6_addr *daddr;
144 struct sock *sk;
145 __u8 hash;
146
147 saddr = &skb->nh.ipv6h->saddr;
148 daddr = saddr + 1;
149
150 hash = nexthdr & (MAX_INET_PROTOS - 1);
151
152 read_lock(&raw_v6_lock);
153 sk = sk_head(&raw_v6_htable[hash]);
154
155 /*
156 * The first socket found will be delivered after
157 * delivery to transport protocols.
158 */
159
160 if (sk == NULL)
161 goto out;
162
163 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr);
164
165 while (sk) {
166 if (nexthdr != IPPROTO_ICMPV6 || !icmpv6_filter(sk, skb)) {
167 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
168
169 /* Not releasing hash table! */
170 if (clone)
171 rawv6_rcv(sk, clone);
172 }
173 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr);
174 }
175 out:
176 read_unlock(&raw_v6_lock);
177 }
178
179 /* This cleans up af_inet6 a bit. -DaveM */
180 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
181 {
182 struct inet_sock *inet = inet_sk(sk);
183 struct ipv6_pinfo *np = inet6_sk(sk);
184 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
185 __u32 v4addr = 0;
186 int addr_type;
187 int err;
188
189 if (addr_len < SIN6_LEN_RFC2133)
190 return -EINVAL;
191 addr_type = ipv6_addr_type(&addr->sin6_addr);
192
193 /* Raw sockets are IPv6 only */
194 if (addr_type == IPV6_ADDR_MAPPED)
195 return(-EADDRNOTAVAIL);
196
197 lock_sock(sk);
198
199 err = -EINVAL;
200 if (sk->sk_state != TCP_CLOSE)
201 goto out;
202
203 /* Check if the address belongs to the host. */
204 if (addr_type != IPV6_ADDR_ANY) {
205 struct net_device *dev = NULL;
206
207 if (addr_type & IPV6_ADDR_LINKLOCAL) {
208 if (addr_len >= sizeof(struct sockaddr_in6) &&
209 addr->sin6_scope_id) {
210 /* Override any existing binding, if another
211 * one is supplied by user.
212 */
213 sk->sk_bound_dev_if = addr->sin6_scope_id;
214 }
215
216 /* Binding to link-local address requires an interface */
217 if (!sk->sk_bound_dev_if)
218 goto out;
219
220 dev = dev_get_by_index(sk->sk_bound_dev_if);
221 if (!dev) {
222 err = -ENODEV;
223 goto out;
224 }
225 }
226
227 /* ipv4 addr of the socket is invalid. Only the
228 * unspecified and mapped address have a v4 equivalent.
229 */
230 v4addr = LOOPBACK4_IPV6;
231 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
232 err = -EADDRNOTAVAIL;
233 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
234 if (dev)
235 dev_put(dev);
236 goto out;
237 }
238 }
239 if (dev)
240 dev_put(dev);
241 }
242
243 inet->rcv_saddr = inet->saddr = v4addr;
244 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
245 if (!(addr_type & IPV6_ADDR_MULTICAST))
246 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
247 err = 0;
248 out:
249 release_sock(sk);
250 return err;
251 }
252
253 void rawv6_err(struct sock *sk, struct sk_buff *skb,
254 struct inet6_skb_parm *opt,
255 int type, int code, int offset, u32 info)
256 {
257 struct inet_sock *inet = inet_sk(sk);
258 struct ipv6_pinfo *np = inet6_sk(sk);
259 int err;
260 int harderr;
261
262 /* Report error on raw socket, if:
263 1. User requested recverr.
264 2. Socket is connected (otherwise the error indication
265 is useless without recverr and error is hard.
266 */
267 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
268 return;
269
270 harderr = icmpv6_err_convert(type, code, &err);
271 if (type == ICMPV6_PKT_TOOBIG)
272 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
273
274 if (np->recverr) {
275 u8 *payload = skb->data;
276 if (!inet->hdrincl)
277 payload += offset;
278 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
279 }
280
281 if (np->recverr || harderr) {
282 sk->sk_err = err;
283 sk->sk_error_report(sk);
284 }
285 }
286
287 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
288 {
289 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
290 skb->ip_summed != CHECKSUM_UNNECESSARY) {
291 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
292 /* FIXME: increment a raw6 drops counter here */
293 kfree_skb(skb);
294 return 0;
295 }
296 skb->ip_summed = CHECKSUM_UNNECESSARY;
297 }
298
299 /* Charge it to the socket. */
300 if (sock_queue_rcv_skb(sk,skb)<0) {
301 /* FIXME: increment a raw6 drops counter here */
302 kfree_skb(skb);
303 return 0;
304 }
305
306 return 0;
307 }
308
309 /*
310 * This is next to useless...
311 * if we demultiplex in network layer we don't need the extra call
312 * just to queue the skb...
313 * maybe we could have the network decide upon a hint if it
314 * should call raw_rcv for demultiplexing
315 */
316 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
317 {
318 struct inet_sock *inet = inet_sk(sk);
319 struct raw6_sock *rp = raw6_sk(sk);
320
321 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
322 kfree_skb(skb);
323 return NET_RX_DROP;
324 }
325
326 if (!rp->checksum)
327 skb->ip_summed = CHECKSUM_UNNECESSARY;
328
329 if (skb->ip_summed != CHECKSUM_UNNECESSARY) {
330 if (skb->ip_summed == CHECKSUM_HW) {
331 skb_postpull_rcsum(skb, skb->nh.raw,
332 skb->h.raw - skb->nh.raw);
333 skb->ip_summed = CHECKSUM_UNNECESSARY;
334 if (csum_ipv6_magic(&skb->nh.ipv6h->saddr,
335 &skb->nh.ipv6h->daddr,
336 skb->len, inet->num, skb->csum)) {
337 LIMIT_NETDEBUG(
338 printk(KERN_DEBUG "raw v6 hw csum failure.\n"));
339 skb->ip_summed = CHECKSUM_NONE;
340 }
341 }
342 if (skb->ip_summed == CHECKSUM_NONE)
343 skb->csum = ~csum_ipv6_magic(&skb->nh.ipv6h->saddr,
344 &skb->nh.ipv6h->daddr,
345 skb->len, inet->num, 0);
346 }
347
348 if (inet->hdrincl) {
349 if (skb->ip_summed != CHECKSUM_UNNECESSARY &&
350 (unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
351 /* FIXME: increment a raw6 drops counter here */
352 kfree_skb(skb);
353 return 0;
354 }
355 skb->ip_summed = CHECKSUM_UNNECESSARY;
356 }
357
358 rawv6_rcv_skb(sk, skb);
359 return 0;
360 }
361
362
363 /*
364 * This should be easy, if there is something there
365 * we return it, otherwise we block.
366 */
367
368 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
369 struct msghdr *msg, size_t len,
370 int noblock, int flags, int *addr_len)
371 {
372 struct ipv6_pinfo *np = inet6_sk(sk);
373 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
374 struct sk_buff *skb;
375 size_t copied;
376 int err;
377
378 if (flags & MSG_OOB)
379 return -EOPNOTSUPP;
380
381 if (addr_len)
382 *addr_len=sizeof(*sin6);
383
384 if (flags & MSG_ERRQUEUE)
385 return ipv6_recv_error(sk, msg, len);
386
387 skb = skb_recv_datagram(sk, flags, noblock, &err);
388 if (!skb)
389 goto out;
390
391 copied = skb->len;
392 if (copied > len) {
393 copied = len;
394 msg->msg_flags |= MSG_TRUNC;
395 }
396
397 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
398 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
399 } else if (msg->msg_flags&MSG_TRUNC) {
400 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum)))
401 goto csum_copy_err;
402 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
403 } else {
404 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
405 if (err == -EINVAL)
406 goto csum_copy_err;
407 }
408 if (err)
409 goto out_free;
410
411 /* Copy the address. */
412 if (sin6) {
413 sin6->sin6_family = AF_INET6;
414 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
415 sin6->sin6_flowinfo = 0;
416 sin6->sin6_scope_id = 0;
417 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
418 sin6->sin6_scope_id = IP6CB(skb)->iif;
419 }
420
421 sock_recv_timestamp(msg, sk, skb);
422
423 if (np->rxopt.all)
424 datagram_recv_ctl(sk, msg, skb);
425
426 err = copied;
427 if (flags & MSG_TRUNC)
428 err = skb->len;
429
430 out_free:
431 skb_free_datagram(sk, skb);
432 out:
433 return err;
434
435 csum_copy_err:
436 /* Clear queue. */
437 if (flags&MSG_PEEK) {
438 int clear = 0;
439 spin_lock_bh(&sk->sk_receive_queue.lock);
440 if (skb == skb_peek(&sk->sk_receive_queue)) {
441 __skb_unlink(skb, &sk->sk_receive_queue);
442 clear = 1;
443 }
444 spin_unlock_bh(&sk->sk_receive_queue.lock);
445 if (clear)
446 kfree_skb(skb);
447 }
448
449 /* Error for blocking case is chosen to masquerade
450 as some normal condition.
451 */
452 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
453 /* FIXME: increment a raw6 drops counter here */
454 goto out_free;
455 }
456
457 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
458 struct raw6_sock *rp)
459 {
460 struct sk_buff *skb;
461 int err = 0;
462 int offset;
463 int len;
464 int total_len;
465 u32 tmp_csum;
466 u16 csum;
467
468 if (!rp->checksum)
469 goto send;
470
471 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
472 goto out;
473
474 offset = rp->offset;
475 total_len = inet_sk(sk)->cork.length - (skb->nh.raw - skb->data);
476 if (offset >= total_len - 1) {
477 err = -EINVAL;
478 ip6_flush_pending_frames(sk);
479 goto out;
480 }
481
482 /* should be check HW csum miyazawa */
483 if (skb_queue_len(&sk->sk_write_queue) == 1) {
484 /*
485 * Only one fragment on the socket.
486 */
487 tmp_csum = skb->csum;
488 } else {
489 struct sk_buff *csum_skb = NULL;
490 tmp_csum = 0;
491
492 skb_queue_walk(&sk->sk_write_queue, skb) {
493 tmp_csum = csum_add(tmp_csum, skb->csum);
494
495 if (csum_skb)
496 continue;
497
498 len = skb->len - (skb->h.raw - skb->data);
499 if (offset >= len) {
500 offset -= len;
501 continue;
502 }
503
504 csum_skb = skb;
505 }
506
507 skb = csum_skb;
508 }
509
510 offset += skb->h.raw - skb->data;
511 if (skb_copy_bits(skb, offset, &csum, 2))
512 BUG();
513
514 /* in case cksum was not initialized */
515 if (unlikely(csum))
516 tmp_csum = csum_sub(tmp_csum, csum);
517
518 tmp_csum = csum_ipv6_magic(&fl->fl6_src,
519 &fl->fl6_dst,
520 total_len, fl->proto, tmp_csum);
521
522 if (tmp_csum == 0)
523 tmp_csum = -1;
524
525 csum = tmp_csum;
526 if (skb_store_bits(skb, offset, &csum, 2))
527 BUG();
528
529 send:
530 err = ip6_push_pending_frames(sk);
531 out:
532 return err;
533 }
534
535 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
536 struct flowi *fl, struct rt6_info *rt,
537 unsigned int flags)
538 {
539 struct ipv6_pinfo *np = inet6_sk(sk);
540 struct ipv6hdr *iph;
541 struct sk_buff *skb;
542 unsigned int hh_len;
543 int err;
544
545 if (length > rt->u.dst.dev->mtu) {
546 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
547 return -EMSGSIZE;
548 }
549 if (flags&MSG_PROBE)
550 goto out;
551
552 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
553
554 skb = sock_alloc_send_skb(sk, length+hh_len+15,
555 flags&MSG_DONTWAIT, &err);
556 if (skb == NULL)
557 goto error;
558 skb_reserve(skb, hh_len);
559
560 skb->priority = sk->sk_priority;
561 skb->dst = dst_clone(&rt->u.dst);
562
563 skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length);
564
565 skb->ip_summed = CHECKSUM_NONE;
566
567 skb->h.raw = skb->nh.raw;
568 err = memcpy_fromiovecend((void *)iph, from, 0, length);
569 if (err)
570 goto error_fault;
571
572 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
573 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
574 dst_output);
575 if (err > 0)
576 err = np->recverr ? net_xmit_errno(err) : 0;
577 if (err)
578 goto error;
579 out:
580 return 0;
581
582 error_fault:
583 err = -EFAULT;
584 kfree_skb(skb);
585 error:
586 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
587 return err;
588 }
589
590 static void rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
591 {
592 struct iovec *iov;
593 u8 __user *type = NULL;
594 u8 __user *code = NULL;
595 int probed = 0;
596 int i;
597
598 if (!msg->msg_iov)
599 return;
600
601 for (i = 0; i < msg->msg_iovlen; i++) {
602 iov = &msg->msg_iov[i];
603 if (!iov)
604 continue;
605
606 switch (fl->proto) {
607 case IPPROTO_ICMPV6:
608 /* check if one-byte field is readable or not. */
609 if (iov->iov_base && iov->iov_len < 1)
610 break;
611
612 if (!type) {
613 type = iov->iov_base;
614 /* check if code field is readable or not. */
615 if (iov->iov_len > 1)
616 code = type + 1;
617 } else if (!code)
618 code = iov->iov_base;
619
620 if (type && code) {
621 get_user(fl->fl_icmp_type, type);
622 __get_user(fl->fl_icmp_code, code);
623 probed = 1;
624 }
625 break;
626 default:
627 probed = 1;
628 break;
629 }
630 if (probed)
631 break;
632 }
633 }
634
635 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
636 struct msghdr *msg, size_t len)
637 {
638 struct ipv6_txoptions opt_space;
639 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
640 struct in6_addr *daddr, *final_p = NULL, final;
641 struct inet_sock *inet = inet_sk(sk);
642 struct ipv6_pinfo *np = inet6_sk(sk);
643 struct raw6_sock *rp = raw6_sk(sk);
644 struct ipv6_txoptions *opt = NULL;
645 struct ip6_flowlabel *flowlabel = NULL;
646 struct dst_entry *dst = NULL;
647 struct flowi fl;
648 int addr_len = msg->msg_namelen;
649 int hlimit = -1;
650 u16 proto;
651 int err;
652
653 /* Rough check on arithmetic overflow,
654 better check is made in ip6_build_xmit
655 */
656 if (len < 0)
657 return -EMSGSIZE;
658
659 /* Mirror BSD error message compatibility */
660 if (msg->msg_flags & MSG_OOB)
661 return -EOPNOTSUPP;
662
663 /*
664 * Get and verify the address.
665 */
666 memset(&fl, 0, sizeof(fl));
667
668 if (sin6) {
669 if (addr_len < SIN6_LEN_RFC2133)
670 return -EINVAL;
671
672 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
673 return(-EAFNOSUPPORT);
674
675 /* port is the proto value [0..255] carried in nexthdr */
676 proto = ntohs(sin6->sin6_port);
677
678 if (!proto)
679 proto = inet->num;
680 else if (proto != inet->num)
681 return(-EINVAL);
682
683 if (proto > 255)
684 return(-EINVAL);
685
686 daddr = &sin6->sin6_addr;
687 if (np->sndflow) {
688 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
689 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
690 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
691 if (flowlabel == NULL)
692 return -EINVAL;
693 daddr = &flowlabel->dst;
694 }
695 }
696
697 /*
698 * Otherwise it will be difficult to maintain
699 * sk->sk_dst_cache.
700 */
701 if (sk->sk_state == TCP_ESTABLISHED &&
702 ipv6_addr_equal(daddr, &np->daddr))
703 daddr = &np->daddr;
704
705 if (addr_len >= sizeof(struct sockaddr_in6) &&
706 sin6->sin6_scope_id &&
707 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
708 fl.oif = sin6->sin6_scope_id;
709 } else {
710 if (sk->sk_state != TCP_ESTABLISHED)
711 return -EDESTADDRREQ;
712
713 proto = inet->num;
714 daddr = &np->daddr;
715 fl.fl6_flowlabel = np->flow_label;
716 }
717
718 if (ipv6_addr_any(daddr)) {
719 /*
720 * unspecified destination address
721 * treated as error... is this correct ?
722 */
723 fl6_sock_release(flowlabel);
724 return(-EINVAL);
725 }
726
727 if (fl.oif == 0)
728 fl.oif = sk->sk_bound_dev_if;
729
730 if (msg->msg_controllen) {
731 opt = &opt_space;
732 memset(opt, 0, sizeof(struct ipv6_txoptions));
733 opt->tot_len = sizeof(struct ipv6_txoptions);
734
735 err = datagram_send_ctl(msg, &fl, opt, &hlimit);
736 if (err < 0) {
737 fl6_sock_release(flowlabel);
738 return err;
739 }
740 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
741 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
742 if (flowlabel == NULL)
743 return -EINVAL;
744 }
745 if (!(opt->opt_nflen|opt->opt_flen))
746 opt = NULL;
747 }
748 if (opt == NULL)
749 opt = np->opt;
750 if (flowlabel)
751 opt = fl6_merge_options(&opt_space, flowlabel, opt);
752
753 fl.proto = proto;
754 rawv6_probe_proto_opt(&fl, msg);
755
756 ipv6_addr_copy(&fl.fl6_dst, daddr);
757 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
758 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
759
760 /* merge ip6_build_xmit from ip6_output */
761 if (opt && opt->srcrt) {
762 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
763 ipv6_addr_copy(&final, &fl.fl6_dst);
764 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
765 final_p = &final;
766 }
767
768 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
769 fl.oif = np->mcast_oif;
770
771 err = ip6_dst_lookup(sk, &dst, &fl);
772 if (err)
773 goto out;
774 if (final_p)
775 ipv6_addr_copy(&fl.fl6_dst, final_p);
776
777 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
778 dst_release(dst);
779 goto out;
780 }
781
782 if (hlimit < 0) {
783 if (ipv6_addr_is_multicast(&fl.fl6_dst))
784 hlimit = np->mcast_hops;
785 else
786 hlimit = np->hop_limit;
787 if (hlimit < 0)
788 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
789 if (hlimit < 0)
790 hlimit = ipv6_get_hoplimit(dst->dev);
791 }
792
793 if (msg->msg_flags&MSG_CONFIRM)
794 goto do_confirm;
795
796 back_from_confirm:
797 if (inet->hdrincl) {
798 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
799 } else {
800 lock_sock(sk);
801 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
802 hlimit, opt, &fl, (struct rt6_info*)dst, msg->msg_flags);
803
804 if (err)
805 ip6_flush_pending_frames(sk);
806 else if (!(msg->msg_flags & MSG_MORE))
807 err = rawv6_push_pending_frames(sk, &fl, rp);
808 }
809 done:
810 ip6_dst_store(sk, dst,
811 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
812 &np->daddr : NULL);
813
814 release_sock(sk);
815 out:
816 fl6_sock_release(flowlabel);
817 return err<0?err:len;
818 do_confirm:
819 dst_confirm(dst);
820 if (!(msg->msg_flags & MSG_PROBE) || len)
821 goto back_from_confirm;
822 err = 0;
823 goto done;
824 }
825
826 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
827 char __user *optval, int optlen)
828 {
829 switch (optname) {
830 case ICMPV6_FILTER:
831 if (optlen > sizeof(struct icmp6_filter))
832 optlen = sizeof(struct icmp6_filter);
833 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
834 return -EFAULT;
835 return 0;
836 default:
837 return -ENOPROTOOPT;
838 };
839
840 return 0;
841 }
842
843 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
844 char __user *optval, int __user *optlen)
845 {
846 int len;
847
848 switch (optname) {
849 case ICMPV6_FILTER:
850 if (get_user(len, optlen))
851 return -EFAULT;
852 if (len < 0)
853 return -EINVAL;
854 if (len > sizeof(struct icmp6_filter))
855 len = sizeof(struct icmp6_filter);
856 if (put_user(len, optlen))
857 return -EFAULT;
858 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
859 return -EFAULT;
860 return 0;
861 default:
862 return -ENOPROTOOPT;
863 };
864
865 return 0;
866 }
867
868
869 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
870 char __user *optval, int optlen)
871 {
872 struct raw6_sock *rp = raw6_sk(sk);
873 int val;
874
875 switch(level) {
876 case SOL_RAW:
877 break;
878
879 case SOL_ICMPV6:
880 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
881 return -EOPNOTSUPP;
882 return rawv6_seticmpfilter(sk, level, optname, optval,
883 optlen);
884 case SOL_IPV6:
885 if (optname == IPV6_CHECKSUM)
886 break;
887 default:
888 return ipv6_setsockopt(sk, level, optname, optval,
889 optlen);
890 };
891
892 if (get_user(val, (int __user *)optval))
893 return -EFAULT;
894
895 switch (optname) {
896 case IPV6_CHECKSUM:
897 /* You may get strange result with a positive odd offset;
898 RFC2292bis agrees with me. */
899 if (val > 0 && (val&1))
900 return(-EINVAL);
901 if (val < 0) {
902 rp->checksum = 0;
903 } else {
904 rp->checksum = 1;
905 rp->offset = val;
906 }
907
908 return 0;
909 break;
910
911 default:
912 return(-ENOPROTOOPT);
913 }
914 }
915
916 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
917 char __user *optval, int __user *optlen)
918 {
919 struct raw6_sock *rp = raw6_sk(sk);
920 int val, len;
921
922 switch(level) {
923 case SOL_RAW:
924 break;
925
926 case SOL_ICMPV6:
927 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
928 return -EOPNOTSUPP;
929 return rawv6_geticmpfilter(sk, level, optname, optval,
930 optlen);
931 case SOL_IPV6:
932 if (optname == IPV6_CHECKSUM)
933 break;
934 default:
935 return ipv6_getsockopt(sk, level, optname, optval,
936 optlen);
937 };
938
939 if (get_user(len,optlen))
940 return -EFAULT;
941
942 switch (optname) {
943 case IPV6_CHECKSUM:
944 if (rp->checksum == 0)
945 val = -1;
946 else
947 val = rp->offset;
948 break;
949
950 default:
951 return -ENOPROTOOPT;
952 }
953
954 len = min_t(unsigned int, sizeof(int), len);
955
956 if (put_user(len, optlen))
957 return -EFAULT;
958 if (copy_to_user(optval,&val,len))
959 return -EFAULT;
960 return 0;
961 }
962
963 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
964 {
965 switch(cmd) {
966 case SIOCOUTQ:
967 {
968 int amount = atomic_read(&sk->sk_wmem_alloc);
969 return put_user(amount, (int __user *)arg);
970 }
971 case SIOCINQ:
972 {
973 struct sk_buff *skb;
974 int amount = 0;
975
976 spin_lock_bh(&sk->sk_receive_queue.lock);
977 skb = skb_peek(&sk->sk_receive_queue);
978 if (skb != NULL)
979 amount = skb->tail - skb->h.raw;
980 spin_unlock_bh(&sk->sk_receive_queue.lock);
981 return put_user(amount, (int __user *)arg);
982 }
983
984 default:
985 return -ENOIOCTLCMD;
986 }
987 }
988
989 static void rawv6_close(struct sock *sk, long timeout)
990 {
991 if (inet_sk(sk)->num == IPPROTO_RAW)
992 ip6_ra_control(sk, -1, NULL);
993
994 sk_common_release(sk);
995 }
996
997 static int rawv6_init_sk(struct sock *sk)
998 {
999 if (inet_sk(sk)->num == IPPROTO_ICMPV6) {
1000 struct raw6_sock *rp = raw6_sk(sk);
1001 rp->checksum = 1;
1002 rp->offset = 2;
1003 }
1004 return(0);
1005 }
1006
1007 struct proto rawv6_prot = {
1008 .name = "RAWv6",
1009 .owner = THIS_MODULE,
1010 .close = rawv6_close,
1011 .connect = ip6_datagram_connect,
1012 .disconnect = udp_disconnect,
1013 .ioctl = rawv6_ioctl,
1014 .init = rawv6_init_sk,
1015 .destroy = inet6_destroy_sock,
1016 .setsockopt = rawv6_setsockopt,
1017 .getsockopt = rawv6_getsockopt,
1018 .sendmsg = rawv6_sendmsg,
1019 .recvmsg = rawv6_recvmsg,
1020 .bind = rawv6_bind,
1021 .backlog_rcv = rawv6_rcv_skb,
1022 .hash = raw_v6_hash,
1023 .unhash = raw_v6_unhash,
1024 .obj_size = sizeof(struct raw6_sock),
1025 };
1026
1027 #ifdef CONFIG_PROC_FS
1028 struct raw6_iter_state {
1029 int bucket;
1030 };
1031
1032 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1033
1034 static struct sock *raw6_get_first(struct seq_file *seq)
1035 {
1036 struct sock *sk;
1037 struct hlist_node *node;
1038 struct raw6_iter_state* state = raw6_seq_private(seq);
1039
1040 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1041 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1042 if (sk->sk_family == PF_INET6)
1043 goto out;
1044 sk = NULL;
1045 out:
1046 return sk;
1047 }
1048
1049 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1050 {
1051 struct raw6_iter_state* state = raw6_seq_private(seq);
1052
1053 do {
1054 sk = sk_next(sk);
1055 try_again:
1056 ;
1057 } while (sk && sk->sk_family != PF_INET6);
1058
1059 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1060 sk = sk_head(&raw_v6_htable[state->bucket]);
1061 goto try_again;
1062 }
1063 return sk;
1064 }
1065
1066 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1067 {
1068 struct sock *sk = raw6_get_first(seq);
1069 if (sk)
1070 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1071 --pos;
1072 return pos ? NULL : sk;
1073 }
1074
1075 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1076 {
1077 read_lock(&raw_v6_lock);
1078 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1079 }
1080
1081 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1082 {
1083 struct sock *sk;
1084
1085 if (v == SEQ_START_TOKEN)
1086 sk = raw6_get_first(seq);
1087 else
1088 sk = raw6_get_next(seq, v);
1089 ++*pos;
1090 return sk;
1091 }
1092
1093 static void raw6_seq_stop(struct seq_file *seq, void *v)
1094 {
1095 read_unlock(&raw_v6_lock);
1096 }
1097
1098 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1099 {
1100 struct ipv6_pinfo *np = inet6_sk(sp);
1101 struct in6_addr *dest, *src;
1102 __u16 destp, srcp;
1103
1104 dest = &np->daddr;
1105 src = &np->rcv_saddr;
1106 destp = 0;
1107 srcp = inet_sk(sp)->num;
1108 seq_printf(seq,
1109 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1110 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1111 i,
1112 src->s6_addr32[0], src->s6_addr32[1],
1113 src->s6_addr32[2], src->s6_addr32[3], srcp,
1114 dest->s6_addr32[0], dest->s6_addr32[1],
1115 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1116 sp->sk_state,
1117 atomic_read(&sp->sk_wmem_alloc),
1118 atomic_read(&sp->sk_rmem_alloc),
1119 0, 0L, 0,
1120 sock_i_uid(sp), 0,
1121 sock_i_ino(sp),
1122 atomic_read(&sp->sk_refcnt), sp);
1123 }
1124
1125 static int raw6_seq_show(struct seq_file *seq, void *v)
1126 {
1127 if (v == SEQ_START_TOKEN)
1128 seq_printf(seq,
1129 " sl "
1130 "local_address "
1131 "remote_address "
1132 "st tx_queue rx_queue tr tm->when retrnsmt"
1133 " uid timeout inode\n");
1134 else
1135 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1136 return 0;
1137 }
1138
1139 static struct seq_operations raw6_seq_ops = {
1140 .start = raw6_seq_start,
1141 .next = raw6_seq_next,
1142 .stop = raw6_seq_stop,
1143 .show = raw6_seq_show,
1144 };
1145
1146 static int raw6_seq_open(struct inode *inode, struct file *file)
1147 {
1148 struct seq_file *seq;
1149 int rc = -ENOMEM;
1150 struct raw6_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
1151 if (!s)
1152 goto out;
1153 rc = seq_open(file, &raw6_seq_ops);
1154 if (rc)
1155 goto out_kfree;
1156 seq = file->private_data;
1157 seq->private = s;
1158 memset(s, 0, sizeof(*s));
1159 out:
1160 return rc;
1161 out_kfree:
1162 kfree(s);
1163 goto out;
1164 }
1165
1166 static struct file_operations raw6_seq_fops = {
1167 .owner = THIS_MODULE,
1168 .open = raw6_seq_open,
1169 .read = seq_read,
1170 .llseek = seq_lseek,
1171 .release = seq_release_private,
1172 };
1173
1174 int __init raw6_proc_init(void)
1175 {
1176 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1177 return -ENOMEM;
1178 return 0;
1179 }
1180
1181 void raw6_proc_exit(void)
1182 {
1183 proc_net_remove("raw6");
1184 }
1185 #endif /* CONFIG_PROC_FS */
Linux kernel - Deliverables
This page took 0.05604 seconds and 6 git commands to generate.