blob: e2b848ec98513ac9ababcc9f3d588c0fd93efc84 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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>
Herbert Xu357b40a2005-04-19 22:30:14 -070037#include <asm/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
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
59struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
60DEFINE_RWLOCK(raw_v6_lock);
61
62static 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
73static 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... */
83struct 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;
108found:
109 return sk;
110}
111
112/*
113 * 0 - deliver
114 * 1 - block
115 */
116static __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 */
140void 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 }
175out:
176 read_unlock(&raw_v6_lock);
177}
178
179/* This cleans up af_inet6 a bit. -DaveM */
180static 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;
248out:
249 release_sock(sk);
250 return err;
251}
252
253void 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
287static 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 */
316int 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->ip_summed = CHECKSUM_UNNECESSARY;
332 if (csum_ipv6_magic(&skb->nh.ipv6h->saddr,
333 &skb->nh.ipv6h->daddr,
334 skb->len, inet->num, skb->csum)) {
335 LIMIT_NETDEBUG(
336 printk(KERN_DEBUG "raw v6 hw csum failure.\n"));
337 skb->ip_summed = CHECKSUM_NONE;
338 }
339 }
340 if (skb->ip_summed == CHECKSUM_NONE)
341 skb->csum = ~csum_ipv6_magic(&skb->nh.ipv6h->saddr,
342 &skb->nh.ipv6h->daddr,
343 skb->len, inet->num, 0);
344 }
345
346 if (inet->hdrincl) {
347 if (skb->ip_summed != CHECKSUM_UNNECESSARY &&
348 (unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
349 /* FIXME: increment a raw6 drops counter here */
350 kfree_skb(skb);
351 return 0;
352 }
353 skb->ip_summed = CHECKSUM_UNNECESSARY;
354 }
355
356 rawv6_rcv_skb(sk, skb);
357 return 0;
358}
359
360
361/*
362 * This should be easy, if there is something there
363 * we return it, otherwise we block.
364 */
365
366static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
367 struct msghdr *msg, size_t len,
368 int noblock, int flags, int *addr_len)
369{
370 struct ipv6_pinfo *np = inet6_sk(sk);
371 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
372 struct sk_buff *skb;
373 size_t copied;
374 int err;
375
376 if (flags & MSG_OOB)
377 return -EOPNOTSUPP;
378
379 if (addr_len)
380 *addr_len=sizeof(*sin6);
381
382 if (flags & MSG_ERRQUEUE)
383 return ipv6_recv_error(sk, msg, len);
384
385 skb = skb_recv_datagram(sk, flags, noblock, &err);
386 if (!skb)
387 goto out;
388
389 copied = skb->len;
390 if (copied > len) {
391 copied = len;
392 msg->msg_flags |= MSG_TRUNC;
393 }
394
395 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
396 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
397 } else if (msg->msg_flags&MSG_TRUNC) {
398 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum)))
399 goto csum_copy_err;
400 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
401 } else {
402 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
403 if (err == -EINVAL)
404 goto csum_copy_err;
405 }
406 if (err)
407 goto out_free;
408
409 /* Copy the address. */
410 if (sin6) {
411 sin6->sin6_family = AF_INET6;
412 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
413 sin6->sin6_flowinfo = 0;
414 sin6->sin6_scope_id = 0;
415 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
416 sin6->sin6_scope_id = IP6CB(skb)->iif;
417 }
418
419 sock_recv_timestamp(msg, sk, skb);
420
421 if (np->rxopt.all)
422 datagram_recv_ctl(sk, msg, skb);
423
424 err = copied;
425 if (flags & MSG_TRUNC)
426 err = skb->len;
427
428out_free:
429 skb_free_datagram(sk, skb);
430out:
431 return err;
432
433csum_copy_err:
434 /* Clear queue. */
435 if (flags&MSG_PEEK) {
436 int clear = 0;
Herbert Xue0f9f852005-06-18 22:56:18 -0700437 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 if (skb == skb_peek(&sk->sk_receive_queue)) {
439 __skb_unlink(skb, &sk->sk_receive_queue);
440 clear = 1;
441 }
Herbert Xue0f9f852005-06-18 22:56:18 -0700442 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 if (clear)
444 kfree_skb(skb);
445 }
446
447 /* Error for blocking case is chosen to masquerade
448 as some normal condition.
449 */
450 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
451 /* FIXME: increment a raw6 drops counter here */
452 goto out_free;
453}
454
455static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
Herbert Xu357b40a2005-04-19 22:30:14 -0700456 struct raw6_sock *rp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457{
458 struct sk_buff *skb;
459 int err = 0;
Herbert Xu357b40a2005-04-19 22:30:14 -0700460 int offset;
461 int len;
Herbert Xu679a8732005-05-03 14:24:36 -0700462 int total_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 u32 tmp_csum;
Herbert Xu357b40a2005-04-19 22:30:14 -0700464 u16 csum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
466 if (!rp->checksum)
467 goto send;
468
469 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
470 goto out;
471
Herbert Xu357b40a2005-04-19 22:30:14 -0700472 offset = rp->offset;
Herbert Xu679a8732005-05-03 14:24:36 -0700473 total_len = inet_sk(sk)->cork.length - (skb->nh.raw - skb->data);
474 if (offset >= total_len - 1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475 err = -EINVAL;
Herbert Xu357b40a2005-04-19 22:30:14 -0700476 ip6_flush_pending_frames(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 goto out;
478 }
479
480 /* should be check HW csum miyazawa */
481 if (skb_queue_len(&sk->sk_write_queue) == 1) {
482 /*
483 * Only one fragment on the socket.
484 */
485 tmp_csum = skb->csum;
486 } else {
Herbert Xu357b40a2005-04-19 22:30:14 -0700487 struct sk_buff *csum_skb = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 tmp_csum = 0;
489
490 skb_queue_walk(&sk->sk_write_queue, skb) {
491 tmp_csum = csum_add(tmp_csum, skb->csum);
Herbert Xu357b40a2005-04-19 22:30:14 -0700492
493 if (csum_skb)
494 continue;
495
496 len = skb->len - (skb->h.raw - skb->data);
497 if (offset >= len) {
498 offset -= len;
499 continue;
500 }
501
502 csum_skb = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 }
Herbert Xu357b40a2005-04-19 22:30:14 -0700504
505 skb = csum_skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 }
507
Herbert Xu357b40a2005-04-19 22:30:14 -0700508 offset += skb->h.raw - skb->data;
509 if (skb_copy_bits(skb, offset, &csum, 2))
510 BUG();
511
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 /* in case cksum was not initialized */
Herbert Xu357b40a2005-04-19 22:30:14 -0700513 if (unlikely(csum))
514 tmp_csum = csum_sub(tmp_csum, csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
Herbert Xu357b40a2005-04-19 22:30:14 -0700516 tmp_csum = csum_ipv6_magic(&fl->fl6_src,
517 &fl->fl6_dst,
Herbert Xu679a8732005-05-03 14:24:36 -0700518 total_len, fl->proto, tmp_csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519
Herbert Xu357b40a2005-04-19 22:30:14 -0700520 if (tmp_csum == 0)
521 tmp_csum = -1;
522
523 csum = tmp_csum;
524 if (skb_store_bits(skb, offset, &csum, 2))
525 BUG();
526
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527send:
528 err = ip6_push_pending_frames(sk);
529out:
530 return err;
531}
532
533static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
534 struct flowi *fl, struct rt6_info *rt,
535 unsigned int flags)
536{
Herbert Xu3320da82005-04-19 22:32:22 -0700537 struct ipv6_pinfo *np = inet6_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 struct ipv6hdr *iph;
539 struct sk_buff *skb;
540 unsigned int hh_len;
541 int err;
542
543 if (length > rt->u.dst.dev->mtu) {
544 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
545 return -EMSGSIZE;
546 }
547 if (flags&MSG_PROBE)
548 goto out;
549
550 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
551
552 skb = sock_alloc_send_skb(sk, length+hh_len+15,
553 flags&MSG_DONTWAIT, &err);
554 if (skb == NULL)
555 goto error;
556 skb_reserve(skb, hh_len);
557
558 skb->priority = sk->sk_priority;
559 skb->dst = dst_clone(&rt->u.dst);
560
561 skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length);
562
563 skb->ip_summed = CHECKSUM_NONE;
564
565 skb->h.raw = skb->nh.raw;
566 err = memcpy_fromiovecend((void *)iph, from, 0, length);
567 if (err)
568 goto error_fault;
569
570 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
571 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
572 dst_output);
573 if (err > 0)
Herbert Xu3320da82005-04-19 22:32:22 -0700574 err = np->recverr ? net_xmit_errno(err) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 if (err)
576 goto error;
577out:
578 return 0;
579
580error_fault:
581 err = -EFAULT;
582 kfree_skb(skb);
583error:
584 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
585 return err;
586}
587
588static void rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
589{
590 struct iovec *iov;
591 u8 __user *type = NULL;
592 u8 __user *code = NULL;
593 int probed = 0;
594 int i;
595
596 if (!msg->msg_iov)
597 return;
598
599 for (i = 0; i < msg->msg_iovlen; i++) {
600 iov = &msg->msg_iov[i];
601 if (!iov)
602 continue;
603
604 switch (fl->proto) {
605 case IPPROTO_ICMPV6:
606 /* check if one-byte field is readable or not. */
607 if (iov->iov_base && iov->iov_len < 1)
608 break;
609
610 if (!type) {
611 type = iov->iov_base;
612 /* check if code field is readable or not. */
613 if (iov->iov_len > 1)
614 code = type + 1;
615 } else if (!code)
616 code = iov->iov_base;
617
618 if (type && code) {
619 get_user(fl->fl_icmp_type, type);
620 __get_user(fl->fl_icmp_code, code);
621 probed = 1;
622 }
623 break;
624 default:
625 probed = 1;
626 break;
627 }
628 if (probed)
629 break;
630 }
631}
632
633static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
634 struct msghdr *msg, size_t len)
635{
636 struct ipv6_txoptions opt_space;
637 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
638 struct in6_addr *daddr, *final_p = NULL, final;
639 struct inet_sock *inet = inet_sk(sk);
640 struct ipv6_pinfo *np = inet6_sk(sk);
641 struct raw6_sock *rp = raw6_sk(sk);
642 struct ipv6_txoptions *opt = NULL;
643 struct ip6_flowlabel *flowlabel = NULL;
644 struct dst_entry *dst = NULL;
645 struct flowi fl;
646 int addr_len = msg->msg_namelen;
647 int hlimit = -1;
648 u16 proto;
649 int err;
650
651 /* Rough check on arithmetic overflow,
652 better check is made in ip6_build_xmit
653 */
654 if (len < 0)
655 return -EMSGSIZE;
656
657 /* Mirror BSD error message compatibility */
658 if (msg->msg_flags & MSG_OOB)
659 return -EOPNOTSUPP;
660
661 /*
662 * Get and verify the address.
663 */
664 memset(&fl, 0, sizeof(fl));
665
666 if (sin6) {
667 if (addr_len < SIN6_LEN_RFC2133)
668 return -EINVAL;
669
670 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
671 return(-EAFNOSUPPORT);
672
673 /* port is the proto value [0..255] carried in nexthdr */
674 proto = ntohs(sin6->sin6_port);
675
676 if (!proto)
677 proto = inet->num;
678 else if (proto != inet->num)
679 return(-EINVAL);
680
681 if (proto > 255)
682 return(-EINVAL);
683
684 daddr = &sin6->sin6_addr;
685 if (np->sndflow) {
686 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
687 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
688 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
689 if (flowlabel == NULL)
690 return -EINVAL;
691 daddr = &flowlabel->dst;
692 }
693 }
694
695 /*
696 * Otherwise it will be difficult to maintain
697 * sk->sk_dst_cache.
698 */
699 if (sk->sk_state == TCP_ESTABLISHED &&
700 ipv6_addr_equal(daddr, &np->daddr))
701 daddr = &np->daddr;
702
703 if (addr_len >= sizeof(struct sockaddr_in6) &&
704 sin6->sin6_scope_id &&
705 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
706 fl.oif = sin6->sin6_scope_id;
707 } else {
708 if (sk->sk_state != TCP_ESTABLISHED)
709 return -EDESTADDRREQ;
710
711 proto = inet->num;
712 daddr = &np->daddr;
713 fl.fl6_flowlabel = np->flow_label;
714 }
715
716 if (ipv6_addr_any(daddr)) {
717 /*
718 * unspecified destination address
719 * treated as error... is this correct ?
720 */
721 fl6_sock_release(flowlabel);
722 return(-EINVAL);
723 }
724
725 if (fl.oif == 0)
726 fl.oif = sk->sk_bound_dev_if;
727
728 if (msg->msg_controllen) {
729 opt = &opt_space;
730 memset(opt, 0, sizeof(struct ipv6_txoptions));
731 opt->tot_len = sizeof(struct ipv6_txoptions);
732
733 err = datagram_send_ctl(msg, &fl, opt, &hlimit);
734 if (err < 0) {
735 fl6_sock_release(flowlabel);
736 return err;
737 }
738 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
739 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
740 if (flowlabel == NULL)
741 return -EINVAL;
742 }
743 if (!(opt->opt_nflen|opt->opt_flen))
744 opt = NULL;
745 }
746 if (opt == NULL)
747 opt = np->opt;
748 if (flowlabel)
749 opt = fl6_merge_options(&opt_space, flowlabel, opt);
750
751 fl.proto = proto;
752 rawv6_probe_proto_opt(&fl, msg);
753
754 ipv6_addr_copy(&fl.fl6_dst, daddr);
755 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
756 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
757
758 /* merge ip6_build_xmit from ip6_output */
759 if (opt && opt->srcrt) {
760 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
761 ipv6_addr_copy(&final, &fl.fl6_dst);
762 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
763 final_p = &final;
764 }
765
766 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
767 fl.oif = np->mcast_oif;
768
769 err = ip6_dst_lookup(sk, &dst, &fl);
770 if (err)
771 goto out;
772 if (final_p)
773 ipv6_addr_copy(&fl.fl6_dst, final_p);
774
775 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
776 dst_release(dst);
777 goto out;
778 }
779
780 if (hlimit < 0) {
781 if (ipv6_addr_is_multicast(&fl.fl6_dst))
782 hlimit = np->mcast_hops;
783 else
784 hlimit = np->hop_limit;
785 if (hlimit < 0)
786 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
787 if (hlimit < 0)
788 hlimit = ipv6_get_hoplimit(dst->dev);
789 }
790
791 if (msg->msg_flags&MSG_CONFIRM)
792 goto do_confirm;
793
794back_from_confirm:
795 if (inet->hdrincl) {
796 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
797 } else {
798 lock_sock(sk);
799 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
800 hlimit, opt, &fl, (struct rt6_info*)dst, msg->msg_flags);
801
802 if (err)
803 ip6_flush_pending_frames(sk);
804 else if (!(msg->msg_flags & MSG_MORE))
Herbert Xu357b40a2005-04-19 22:30:14 -0700805 err = rawv6_push_pending_frames(sk, &fl, rp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 }
807done:
808 ip6_dst_store(sk, dst,
809 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
810 &np->daddr : NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811
812 release_sock(sk);
813out:
814 fl6_sock_release(flowlabel);
815 return err<0?err:len;
816do_confirm:
817 dst_confirm(dst);
818 if (!(msg->msg_flags & MSG_PROBE) || len)
819 goto back_from_confirm;
820 err = 0;
821 goto done;
822}
823
824static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
825 char __user *optval, int optlen)
826{
827 switch (optname) {
828 case ICMPV6_FILTER:
829 if (optlen > sizeof(struct icmp6_filter))
830 optlen = sizeof(struct icmp6_filter);
831 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
832 return -EFAULT;
833 return 0;
834 default:
835 return -ENOPROTOOPT;
836 };
837
838 return 0;
839}
840
841static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
842 char __user *optval, int __user *optlen)
843{
844 int len;
845
846 switch (optname) {
847 case ICMPV6_FILTER:
848 if (get_user(len, optlen))
849 return -EFAULT;
850 if (len < 0)
851 return -EINVAL;
852 if (len > sizeof(struct icmp6_filter))
853 len = sizeof(struct icmp6_filter);
854 if (put_user(len, optlen))
855 return -EFAULT;
856 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
857 return -EFAULT;
858 return 0;
859 default:
860 return -ENOPROTOOPT;
861 };
862
863 return 0;
864}
865
866
867static int rawv6_setsockopt(struct sock *sk, int level, int optname,
868 char __user *optval, int optlen)
869{
870 struct raw6_sock *rp = raw6_sk(sk);
871 int val;
872
873 switch(level) {
874 case SOL_RAW:
875 break;
876
877 case SOL_ICMPV6:
878 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
879 return -EOPNOTSUPP;
880 return rawv6_seticmpfilter(sk, level, optname, optval,
881 optlen);
882 case SOL_IPV6:
883 if (optname == IPV6_CHECKSUM)
884 break;
885 default:
886 return ipv6_setsockopt(sk, level, optname, optval,
887 optlen);
888 };
889
890 if (get_user(val, (int __user *)optval))
891 return -EFAULT;
892
893 switch (optname) {
894 case IPV6_CHECKSUM:
895 /* You may get strange result with a positive odd offset;
896 RFC2292bis agrees with me. */
897 if (val > 0 && (val&1))
898 return(-EINVAL);
899 if (val < 0) {
900 rp->checksum = 0;
901 } else {
902 rp->checksum = 1;
903 rp->offset = val;
904 }
905
906 return 0;
907 break;
908
909 default:
910 return(-ENOPROTOOPT);
911 }
912}
913
914static int rawv6_getsockopt(struct sock *sk, int level, int optname,
915 char __user *optval, int __user *optlen)
916{
917 struct raw6_sock *rp = raw6_sk(sk);
918 int val, len;
919
920 switch(level) {
921 case SOL_RAW:
922 break;
923
924 case SOL_ICMPV6:
925 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
926 return -EOPNOTSUPP;
927 return rawv6_geticmpfilter(sk, level, optname, optval,
928 optlen);
929 case SOL_IPV6:
930 if (optname == IPV6_CHECKSUM)
931 break;
932 default:
933 return ipv6_getsockopt(sk, level, optname, optval,
934 optlen);
935 };
936
937 if (get_user(len,optlen))
938 return -EFAULT;
939
940 switch (optname) {
941 case IPV6_CHECKSUM:
942 if (rp->checksum == 0)
943 val = -1;
944 else
945 val = rp->offset;
946 break;
947
948 default:
949 return -ENOPROTOOPT;
950 }
951
952 len = min_t(unsigned int, sizeof(int), len);
953
954 if (put_user(len, optlen))
955 return -EFAULT;
956 if (copy_to_user(optval,&val,len))
957 return -EFAULT;
958 return 0;
959}
960
961static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
962{
963 switch(cmd) {
964 case SIOCOUTQ:
965 {
966 int amount = atomic_read(&sk->sk_wmem_alloc);
967 return put_user(amount, (int __user *)arg);
968 }
969 case SIOCINQ:
970 {
971 struct sk_buff *skb;
972 int amount = 0;
973
Herbert Xue0f9f852005-06-18 22:56:18 -0700974 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 skb = skb_peek(&sk->sk_receive_queue);
976 if (skb != NULL)
977 amount = skb->tail - skb->h.raw;
Herbert Xue0f9f852005-06-18 22:56:18 -0700978 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 return put_user(amount, (int __user *)arg);
980 }
981
982 default:
983 return -ENOIOCTLCMD;
984 }
985}
986
987static void rawv6_close(struct sock *sk, long timeout)
988{
989 if (inet_sk(sk)->num == IPPROTO_RAW)
990 ip6_ra_control(sk, -1, NULL);
991
992 sk_common_release(sk);
993}
994
995static int rawv6_init_sk(struct sock *sk)
996{
997 if (inet_sk(sk)->num == IPPROTO_ICMPV6) {
998 struct raw6_sock *rp = raw6_sk(sk);
999 rp->checksum = 1;
1000 rp->offset = 2;
1001 }
1002 return(0);
1003}
1004
1005struct proto rawv6_prot = {
1006 .name = "RAWv6",
1007 .owner = THIS_MODULE,
1008 .close = rawv6_close,
1009 .connect = ip6_datagram_connect,
1010 .disconnect = udp_disconnect,
1011 .ioctl = rawv6_ioctl,
1012 .init = rawv6_init_sk,
1013 .destroy = inet6_destroy_sock,
1014 .setsockopt = rawv6_setsockopt,
1015 .getsockopt = rawv6_getsockopt,
1016 .sendmsg = rawv6_sendmsg,
1017 .recvmsg = rawv6_recvmsg,
1018 .bind = rawv6_bind,
1019 .backlog_rcv = rawv6_rcv_skb,
1020 .hash = raw_v6_hash,
1021 .unhash = raw_v6_unhash,
1022 .obj_size = sizeof(struct raw6_sock),
1023};
1024
1025#ifdef CONFIG_PROC_FS
1026struct raw6_iter_state {
1027 int bucket;
1028};
1029
1030#define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1031
1032static struct sock *raw6_get_first(struct seq_file *seq)
1033{
1034 struct sock *sk;
1035 struct hlist_node *node;
1036 struct raw6_iter_state* state = raw6_seq_private(seq);
1037
1038 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1039 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1040 if (sk->sk_family == PF_INET6)
1041 goto out;
1042 sk = NULL;
1043out:
1044 return sk;
1045}
1046
1047static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1048{
1049 struct raw6_iter_state* state = raw6_seq_private(seq);
1050
1051 do {
1052 sk = sk_next(sk);
1053try_again:
1054 ;
1055 } while (sk && sk->sk_family != PF_INET6);
1056
1057 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1058 sk = sk_head(&raw_v6_htable[state->bucket]);
1059 goto try_again;
1060 }
1061 return sk;
1062}
1063
1064static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1065{
1066 struct sock *sk = raw6_get_first(seq);
1067 if (sk)
1068 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1069 --pos;
1070 return pos ? NULL : sk;
1071}
1072
1073static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1074{
1075 read_lock(&raw_v6_lock);
1076 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1077}
1078
1079static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1080{
1081 struct sock *sk;
1082
1083 if (v == SEQ_START_TOKEN)
1084 sk = raw6_get_first(seq);
1085 else
1086 sk = raw6_get_next(seq, v);
1087 ++*pos;
1088 return sk;
1089}
1090
1091static void raw6_seq_stop(struct seq_file *seq, void *v)
1092{
1093 read_unlock(&raw_v6_lock);
1094}
1095
1096static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1097{
1098 struct ipv6_pinfo *np = inet6_sk(sp);
1099 struct in6_addr *dest, *src;
1100 __u16 destp, srcp;
1101
1102 dest = &np->daddr;
1103 src = &np->rcv_saddr;
1104 destp = 0;
1105 srcp = inet_sk(sp)->num;
1106 seq_printf(seq,
1107 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1108 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1109 i,
1110 src->s6_addr32[0], src->s6_addr32[1],
1111 src->s6_addr32[2], src->s6_addr32[3], srcp,
1112 dest->s6_addr32[0], dest->s6_addr32[1],
1113 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1114 sp->sk_state,
1115 atomic_read(&sp->sk_wmem_alloc),
1116 atomic_read(&sp->sk_rmem_alloc),
1117 0, 0L, 0,
1118 sock_i_uid(sp), 0,
1119 sock_i_ino(sp),
1120 atomic_read(&sp->sk_refcnt), sp);
1121}
1122
1123static int raw6_seq_show(struct seq_file *seq, void *v)
1124{
1125 if (v == SEQ_START_TOKEN)
1126 seq_printf(seq,
1127 " sl "
1128 "local_address "
1129 "remote_address "
1130 "st tx_queue rx_queue tr tm->when retrnsmt"
1131 " uid timeout inode\n");
1132 else
1133 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1134 return 0;
1135}
1136
1137static struct seq_operations raw6_seq_ops = {
1138 .start = raw6_seq_start,
1139 .next = raw6_seq_next,
1140 .stop = raw6_seq_stop,
1141 .show = raw6_seq_show,
1142};
1143
1144static int raw6_seq_open(struct inode *inode, struct file *file)
1145{
1146 struct seq_file *seq;
1147 int rc = -ENOMEM;
1148 struct raw6_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
1149 if (!s)
1150 goto out;
1151 rc = seq_open(file, &raw6_seq_ops);
1152 if (rc)
1153 goto out_kfree;
1154 seq = file->private_data;
1155 seq->private = s;
1156 memset(s, 0, sizeof(*s));
1157out:
1158 return rc;
1159out_kfree:
1160 kfree(s);
1161 goto out;
1162}
1163
1164static struct file_operations raw6_seq_fops = {
1165 .owner = THIS_MODULE,
1166 .open = raw6_seq_open,
1167 .read = seq_read,
1168 .llseek = seq_lseek,
1169 .release = seq_release_private,
1170};
1171
1172int __init raw6_proc_init(void)
1173{
1174 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1175 return -ENOMEM;
1176 return 0;
1177}
1178
1179void raw6_proc_exit(void)
1180{
1181 proc_net_remove("raw6");
1182}
1183#endif /* CONFIG_PROC_FS */