?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[PATCH] USB: drivers/usb/net/zd1201.c: make some code static
[linux-2.6/kvm.git] / net / ipv6 / raw.c
blob1352c1d9bf4d35f6619d3dce4996b69578bdb2d2
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->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
366 static 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
428 out_free:
429 skb_free_datagram(sk, skb);
430 out:
431 return err;
432
433 csum_copy_err:
434 /* Clear queue. */
435 if (flags&MSG_PEEK) {
436 int clear = 0;
437 spin_lock_irq(&sk->sk_receive_queue.lock);
438 if (skb == skb_peek(&sk->sk_receive_queue)) {
439 __skb_unlink(skb, &sk->sk_receive_queue);
440 clear = 1;
441 }
442 spin_unlock_irq(&sk->sk_receive_queue.lock);
443 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
455 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
456 struct raw6_sock *rp)
457 {
458 struct inet_sock *inet = inet_sk(sk);
459 struct sk_buff *skb;
460 int err = 0;
461 int offset;
462 int len;
463 u32 tmp_csum;
464 u16 csum;
465
466 if (!rp->checksum)
467 goto send;
468
469 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
470 goto out;
471
472 offset = rp->offset;
473 if (offset >= inet->cork.length - 1) {
474 err = -EINVAL;
475 ip6_flush_pending_frames(sk);
476 goto out;
477 }
478
479 /* should be check HW csum miyazawa */
480 if (skb_queue_len(&sk->sk_write_queue) == 1) {
481 /*
482 * Only one fragment on the socket.
483 */
484 tmp_csum = skb->csum;
485 } else {
486 struct sk_buff *csum_skb = NULL;
487 tmp_csum = 0;
488
489 skb_queue_walk(&sk->sk_write_queue, skb) {
490 tmp_csum = csum_add(tmp_csum, skb->csum);
491
492 if (csum_skb)
493 continue;
494
495 len = skb->len - (skb->h.raw - skb->data);
496 if (offset >= len) {
497 offset -= len;
498 continue;
499 }
500
501 csum_skb = skb;
502 }
503
504 skb = csum_skb;
505 }
506
507 offset += skb->h.raw - skb->data;
508 if (skb_copy_bits(skb, offset, &csum, 2))
509 BUG();
510
511 /* in case cksum was not initialized */
512 if (unlikely(csum))
513 tmp_csum = csum_sub(tmp_csum, csum);
514
515 tmp_csum = csum_ipv6_magic(&fl->fl6_src,
516 &fl->fl6_dst,
517 inet->cork.length, fl->proto, tmp_csum);
518
519 if (tmp_csum == 0)
520 tmp_csum = -1;
521
522 csum = tmp_csum;
523 if (skb_store_bits(skb, offset, &csum, 2))
524 BUG();
525
526 send:
527 err = ip6_push_pending_frames(sk);
528 out:
529 return err;
530 }
531
532 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
533 struct flowi *fl, struct rt6_info *rt,
534 unsigned int flags)
535 {
536 struct ipv6_pinfo *np = inet6_sk(sk);
537 struct ipv6hdr *iph;
538 struct sk_buff *skb;
539 unsigned int hh_len;
540 int err;
541
542 if (length > rt->u.dst.dev->mtu) {
543 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
544 return -EMSGSIZE;
545 }
546 if (flags&MSG_PROBE)
547 goto out;
548
549 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
550
551 skb = sock_alloc_send_skb(sk, length+hh_len+15,
552 flags&MSG_DONTWAIT, &err);
553 if (skb == NULL)
554 goto error;
555 skb_reserve(skb, hh_len);
556
557 skb->priority = sk->sk_priority;
558 skb->dst = dst_clone(&rt->u.dst);
559
560 skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length);
561
562 skb->ip_summed = CHECKSUM_NONE;
563
564 skb->h.raw = skb->nh.raw;
565 err = memcpy_fromiovecend((void *)iph, from, 0, length);
566 if (err)
567 goto error_fault;
568
569 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
570 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
571 dst_output);
572 if (err > 0)
573 err = np->recverr ? net_xmit_errno(err) : 0;
574 if (err)
575 goto error;
576 out:
577 return 0;
578
579 error_fault:
580 err = -EFAULT;
581 kfree_skb(skb);
582 error:
583 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
584 return err;
585 }
586
587 static void rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
588 {
589 struct iovec *iov;
590 u8 __user *type = NULL;
591 u8 __user *code = NULL;
592 int probed = 0;
593 int i;
594
595 if (!msg->msg_iov)
596 return;
597
598 for (i = 0; i < msg->msg_iovlen; i++) {
599 iov = &msg->msg_iov[i];
600 if (!iov)
601 continue;
602
603 switch (fl->proto) {
604 case IPPROTO_ICMPV6:
605 /* check if one-byte field is readable or not. */
606 if (iov->iov_base && iov->iov_len < 1)
607 break;
608
609 if (!type) {
610 type = iov->iov_base;
611 /* check if code field is readable or not. */
612 if (iov->iov_len > 1)
613 code = type + 1;
614 } else if (!code)
615 code = iov->iov_base;
616
617 if (type && code) {
618 get_user(fl->fl_icmp_type, type);
619 __get_user(fl->fl_icmp_code, code);
620 probed = 1;
621 }
622 break;
623 default:
624 probed = 1;
625 break;
626 }
627 if (probed)
628 break;
629 }
630 }
631
632 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
633 struct msghdr *msg, size_t len)
634 {
635 struct ipv6_txoptions opt_space;
636 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
637 struct in6_addr *daddr, *final_p = NULL, final;
638 struct inet_sock *inet = inet_sk(sk);
639 struct ipv6_pinfo *np = inet6_sk(sk);
640 struct raw6_sock *rp = raw6_sk(sk);
641 struct ipv6_txoptions *opt = NULL;
642 struct ip6_flowlabel *flowlabel = NULL;
643 struct dst_entry *dst = NULL;
644 struct flowi fl;
645 int addr_len = msg->msg_namelen;
646 int hlimit = -1;
647 u16 proto;
648 int err;
649
650 /* Rough check on arithmetic overflow,
651 better check is made in ip6_build_xmit
652 */
653 if (len < 0)
654 return -EMSGSIZE;
655
656 /* Mirror BSD error message compatibility */
657 if (msg->msg_flags & MSG_OOB)
658 return -EOPNOTSUPP;
659
660 /*
661 * Get and verify the address.
662 */
663 memset(&fl, 0, sizeof(fl));
664
665 if (sin6) {
666 if (addr_len < SIN6_LEN_RFC2133)
667 return -EINVAL;
668
669 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
670 return(-EAFNOSUPPORT);
671
672 /* port is the proto value [0..255] carried in nexthdr */
673 proto = ntohs(sin6->sin6_port);
674
675 if (!proto)
676 proto = inet->num;
677 else if (proto != inet->num)
678 return(-EINVAL);
679
680 if (proto > 255)
681 return(-EINVAL);
682
683 daddr = &sin6->sin6_addr;
684 if (np->sndflow) {
685 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
686 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
687 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
688 if (flowlabel == NULL)
689 return -EINVAL;
690 daddr = &flowlabel->dst;
691 }
692 }
693
694 /*
695 * Otherwise it will be difficult to maintain
696 * sk->sk_dst_cache.
697 */
698 if (sk->sk_state == TCP_ESTABLISHED &&
699 ipv6_addr_equal(daddr, &np->daddr))
700 daddr = &np->daddr;
701
702 if (addr_len >= sizeof(struct sockaddr_in6) &&
703 sin6->sin6_scope_id &&
704 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
705 fl.oif = sin6->sin6_scope_id;
706 } else {
707 if (sk->sk_state != TCP_ESTABLISHED)
708 return -EDESTADDRREQ;
709
710 proto = inet->num;
711 daddr = &np->daddr;
712 fl.fl6_flowlabel = np->flow_label;
713 }
714
715 if (ipv6_addr_any(daddr)) {
716 /*
717 * unspecified destination address
718 * treated as error... is this correct ?
719 */
720 fl6_sock_release(flowlabel);
721 return(-EINVAL);
722 }
723
724 if (fl.oif == 0)
725 fl.oif = sk->sk_bound_dev_if;
726
727 if (msg->msg_controllen) {
728 opt = &opt_space;
729 memset(opt, 0, sizeof(struct ipv6_txoptions));
730 opt->tot_len = sizeof(struct ipv6_txoptions);
731
732 err = datagram_send_ctl(msg, &fl, opt, &hlimit);
733 if (err < 0) {
734 fl6_sock_release(flowlabel);
735 return err;
736 }
737 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
738 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
739 if (flowlabel == NULL)
740 return -EINVAL;
741 }
742 if (!(opt->opt_nflen|opt->opt_flen))
743 opt = NULL;
744 }
745 if (opt == NULL)
746 opt = np->opt;
747 if (flowlabel)
748 opt = fl6_merge_options(&opt_space, flowlabel, opt);
749
750 fl.proto = proto;
751 rawv6_probe_proto_opt(&fl, msg);
752
753 ipv6_addr_copy(&fl.fl6_dst, daddr);
754 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
755 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
756
757 /* merge ip6_build_xmit from ip6_output */
758 if (opt && opt->srcrt) {
759 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
760 ipv6_addr_copy(&final, &fl.fl6_dst);
761 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
762 final_p = &final;
763 }
764
765 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
766 fl.oif = np->mcast_oif;
767
768 err = ip6_dst_lookup(sk, &dst, &fl);
769 if (err)
770 goto out;
771 if (final_p)
772 ipv6_addr_copy(&fl.fl6_dst, final_p);
773
774 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
775 dst_release(dst);
776 goto out;
777 }
778
779 if (hlimit < 0) {
780 if (ipv6_addr_is_multicast(&fl.fl6_dst))
781 hlimit = np->mcast_hops;
782 else
783 hlimit = np->hop_limit;
784 if (hlimit < 0)
785 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
786 if (hlimit < 0)
787 hlimit = ipv6_get_hoplimit(dst->dev);
788 }
789
790 if (msg->msg_flags&MSG_CONFIRM)
791 goto do_confirm;
792
793 back_from_confirm:
794 if (inet->hdrincl) {
795 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
796 } else {
797 lock_sock(sk);
798 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
799 hlimit, opt, &fl, (struct rt6_info*)dst, msg->msg_flags);
800
801 if (err)
802 ip6_flush_pending_frames(sk);
803 else if (!(msg->msg_flags & MSG_MORE))
804 err = rawv6_push_pending_frames(sk, &fl, rp);
805 }
806 done:
807 ip6_dst_store(sk, dst,
808 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
809 &np->daddr : NULL);
810
811 release_sock(sk);
812 out:
813 fl6_sock_release(flowlabel);
814 return err<0?err:len;
815 do_confirm:
816 dst_confirm(dst);
817 if (!(msg->msg_flags & MSG_PROBE) || len)
818 goto back_from_confirm;
819 err = 0;
820 goto done;
821 }
822
823 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
824 char __user *optval, int optlen)
825 {
826 switch (optname) {
827 case ICMPV6_FILTER:
828 if (optlen > sizeof(struct icmp6_filter))
829 optlen = sizeof(struct icmp6_filter);
830 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
831 return -EFAULT;
832 return 0;
833 default:
834 return -ENOPROTOOPT;
835 };
836
837 return 0;
838 }
839
840 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
841 char __user *optval, int __user *optlen)
842 {
843 int len;
844
845 switch (optname) {
846 case ICMPV6_FILTER:
847 if (get_user(len, optlen))
848 return -EFAULT;
849 if (len < 0)
850 return -EINVAL;
851 if (len > sizeof(struct icmp6_filter))
852 len = sizeof(struct icmp6_filter);
853 if (put_user(len, optlen))
854 return -EFAULT;
855 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
856 return -EFAULT;
857 return 0;
858 default:
859 return -ENOPROTOOPT;
860 };
861
862 return 0;
863 }
864
865
866 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
867 char __user *optval, int optlen)
868 {
869 struct raw6_sock *rp = raw6_sk(sk);
870 int val;
871
872 switch(level) {
873 case SOL_RAW:
874 break;
875
876 case SOL_ICMPV6:
877 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
878 return -EOPNOTSUPP;
879 return rawv6_seticmpfilter(sk, level, optname, optval,
880 optlen);
881 case SOL_IPV6:
882 if (optname == IPV6_CHECKSUM)
883 break;
884 default:
885 return ipv6_setsockopt(sk, level, optname, optval,
886 optlen);
887 };
888
889 if (get_user(val, (int __user *)optval))
890 return -EFAULT;
891
892 switch (optname) {
893 case IPV6_CHECKSUM:
894 /* You may get strange result with a positive odd offset;
895 RFC2292bis agrees with me. */
896 if (val > 0 && (val&1))
897 return(-EINVAL);
898 if (val < 0) {
899 rp->checksum = 0;
900 } else {
901 rp->checksum = 1;
902 rp->offset = val;
903 }
904
905 return 0;
906 break;
907
908 default:
909 return(-ENOPROTOOPT);
910 }
911 }
912
913 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
914 char __user *optval, int __user *optlen)
915 {
916 struct raw6_sock *rp = raw6_sk(sk);
917 int val, len;
918
919 switch(level) {
920 case SOL_RAW:
921 break;
922
923 case SOL_ICMPV6:
924 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
925 return -EOPNOTSUPP;
926 return rawv6_geticmpfilter(sk, level, optname, optval,
927 optlen);
928 case SOL_IPV6:
929 if (optname == IPV6_CHECKSUM)
930 break;
931 default:
932 return ipv6_getsockopt(sk, level, optname, optval,
933 optlen);
934 };
935
936 if (get_user(len,optlen))
937 return -EFAULT;
938
939 switch (optname) {
940 case IPV6_CHECKSUM:
941 if (rp->checksum == 0)
942 val = -1;
943 else
944 val = rp->offset;
945 break;
946
947 default:
948 return -ENOPROTOOPT;
949 }
950
951 len = min_t(unsigned int, sizeof(int), len);
952
953 if (put_user(len, optlen))
954 return -EFAULT;
955 if (copy_to_user(optval,&val,len))
956 return -EFAULT;
957 return 0;
958 }
959
960 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
961 {
962 switch(cmd) {
963 case SIOCOUTQ:
964 {
965 int amount = atomic_read(&sk->sk_wmem_alloc);
966 return put_user(amount, (int __user *)arg);
967 }
968 case SIOCINQ:
969 {
970 struct sk_buff *skb;
971 int amount = 0;
972
973 spin_lock_irq(&sk->sk_receive_queue.lock);
974 skb = skb_peek(&sk->sk_receive_queue);
975 if (skb != NULL)
976 amount = skb->tail - skb->h.raw;
977 spin_unlock_irq(&sk->sk_receive_queue.lock);
978 return put_user(amount, (int __user *)arg);
979 }
980
981 default:
982 return -ENOIOCTLCMD;
983 }
984 }
985
986 static void rawv6_close(struct sock *sk, long timeout)
987 {
988 if (inet_sk(sk)->num == IPPROTO_RAW)
989 ip6_ra_control(sk, -1, NULL);
990
991 sk_common_release(sk);
992 }
993
994 static int rawv6_init_sk(struct sock *sk)
995 {
996 if (inet_sk(sk)->num == IPPROTO_ICMPV6) {
997 struct raw6_sock *rp = raw6_sk(sk);
998 rp->checksum = 1;
999 rp->offset = 2;
1000 }
1001 return(0);
1002 }
1003
1004 struct proto rawv6_prot = {
1005 .name = "RAWv6",
1006 .owner = THIS_MODULE,
1007 .close = rawv6_close,
1008 .connect = ip6_datagram_connect,
1009 .disconnect = udp_disconnect,
1010 .ioctl = rawv6_ioctl,
1011 .init = rawv6_init_sk,
1012 .destroy = inet6_destroy_sock,
1013 .setsockopt = rawv6_setsockopt,
1014 .getsockopt = rawv6_getsockopt,
1015 .sendmsg = rawv6_sendmsg,
1016 .recvmsg = rawv6_recvmsg,
1017 .bind = rawv6_bind,
1018 .backlog_rcv = rawv6_rcv_skb,
1019 .hash = raw_v6_hash,
1020 .unhash = raw_v6_unhash,
1021 .obj_size = sizeof(struct raw6_sock),
1022 };
1023
1024 #ifdef CONFIG_PROC_FS
1025 struct raw6_iter_state {
1026 int bucket;
1027 };
1028
1029 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1030
1031 static struct sock *raw6_get_first(struct seq_file *seq)
1032 {
1033 struct sock *sk;
1034 struct hlist_node *node;
1035 struct raw6_iter_state* state = raw6_seq_private(seq);
1036
1037 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1038 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1039 if (sk->sk_family == PF_INET6)
1040 goto out;
1041 sk = NULL;
1042 out:
1043 return sk;
1044 }
1045
1046 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1047 {
1048 struct raw6_iter_state* state = raw6_seq_private(seq);
1049
1050 do {
1051 sk = sk_next(sk);
1052 try_again:
1053 ;
1054 } while (sk && sk->sk_family != PF_INET6);
1055
1056 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1057 sk = sk_head(&raw_v6_htable[state->bucket]);
1058 goto try_again;
1059 }
1060 return sk;
1061 }
1062
1063 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1064 {
1065 struct sock *sk = raw6_get_first(seq);
1066 if (sk)
1067 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1068 --pos;
1069 return pos ? NULL : sk;
1070 }
1071
1072 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1073 {
1074 read_lock(&raw_v6_lock);
1075 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1076 }
1077
1078 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1079 {
1080 struct sock *sk;
1081
1082 if (v == SEQ_START_TOKEN)
1083 sk = raw6_get_first(seq);
1084 else
1085 sk = raw6_get_next(seq, v);
1086 ++*pos;
1087 return sk;
1088 }
1089
1090 static void raw6_seq_stop(struct seq_file *seq, void *v)
1091 {
1092 read_unlock(&raw_v6_lock);
1093 }
1094
1095 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1096 {
1097 struct ipv6_pinfo *np = inet6_sk(sp);
1098 struct in6_addr *dest, *src;
1099 __u16 destp, srcp;
1100
1101 dest = &np->daddr;
1102 src = &np->rcv_saddr;
1103 destp = 0;
1104 srcp = inet_sk(sp)->num;
1105 seq_printf(seq,
1106 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1107 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1108 i,
1109 src->s6_addr32[0], src->s6_addr32[1],
1110 src->s6_addr32[2], src->s6_addr32[3], srcp,
1111 dest->s6_addr32[0], dest->s6_addr32[1],
1112 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1113 sp->sk_state,
1114 atomic_read(&sp->sk_wmem_alloc),
1115 atomic_read(&sp->sk_rmem_alloc),
1116 0, 0L, 0,
1117 sock_i_uid(sp), 0,
1118 sock_i_ino(sp),
1119 atomic_read(&sp->sk_refcnt), sp);
1120 }
1121
1122 static int raw6_seq_show(struct seq_file *seq, void *v)
1123 {
1124 if (v == SEQ_START_TOKEN)
1125 seq_printf(seq,
1126 " sl "
1127 "local_address "
1128 "remote_address "
1129 "st tx_queue rx_queue tr tm->when retrnsmt"
1130 " uid timeout inode\n");
1131 else
1132 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1133 return 0;
1134 }
1135
1136 static struct seq_operations raw6_seq_ops = {
1137 .start = raw6_seq_start,
1138 .next = raw6_seq_next,
1139 .stop = raw6_seq_stop,
1140 .show = raw6_seq_show,
1141 };
1142
1143 static int raw6_seq_open(struct inode *inode, struct file *file)
1144 {
1145 struct seq_file *seq;
1146 int rc = -ENOMEM;
1147 struct raw6_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
1148 if (!s)
1149 goto out;
1150 rc = seq_open(file, &raw6_seq_ops);
1151 if (rc)
1152 goto out_kfree;
1153 seq = file->private_data;
1154 seq->private = s;
1155 memset(s, 0, sizeof(*s));
1156 out:
1157 return rc;
1158 out_kfree:
1159 kfree(s);
1160 goto out;
1161 }
1162
1163 static struct file_operations raw6_seq_fops = {
1164 .owner = THIS_MODULE,
1165 .open = raw6_seq_open,
1166 .read = seq_read,
1167 .llseek = seq_lseek,
1168 .release = seq_release_private,
1169 };
1170
1171 int __init raw6_proc_init(void)
1172 {
1173 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1174 return -ENOMEM;
1175 return 0;
1176 }
1177
1178 void raw6_proc_exit(void)
1179 {
1180 proc_net_remove("raw6");
1181 }
1182 #endif /* CONFIG_PROC_FS */
kernel-based virtual machine