ia64/linux-2.6.18-xen.hg

view net/ipv6/raw.c @ 673:3161879fdf22

[IA64] xencomm: support XENMEM_add_to_physmap and XENMEM_remove_from_phsymap

support XENMEM_add_to_physmap and XENMEM_remove_from_phsymap.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
author Isaku Yamahata <yamahata@valinux.co.jp>
date Tue Sep 16 21:26:15 2008 +0900 (2008-09-16)
parents 831230e53067
children
line source
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 */
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>
39 #include <net/ip.h>
40 #include <net/sock.h>
41 #include <net/snmp.h>
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>
54 #include <net/rawv6.h>
55 #include <net/xfrm.h>
57 #include <linux/proc_fs.h>
58 #include <linux/seq_file.h>
60 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
61 DEFINE_RWLOCK(raw_v6_lock);
63 static void raw_v6_hash(struct sock *sk)
64 {
65 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num &
66 (RAWV6_HTABLE_SIZE - 1)];
68 write_lock_bh(&raw_v6_lock);
69 sk_add_node(sk, list);
70 sock_prot_inc_use(sk->sk_prot);
71 write_unlock_bh(&raw_v6_lock);
72 }
74 static void raw_v6_unhash(struct sock *sk)
75 {
76 write_lock_bh(&raw_v6_lock);
77 if (sk_del_node_init(sk))
78 sock_prot_dec_use(sk->sk_prot);
79 write_unlock_bh(&raw_v6_lock);
80 }
83 /* Grumble... icmp and ip_input want to get at this... */
84 struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
85 struct in6_addr *loc_addr, struct in6_addr *rmt_addr,
86 int dif)
87 {
88 struct hlist_node *node;
89 int is_multicast = ipv6_addr_is_multicast(loc_addr);
91 sk_for_each_from(sk, node)
92 if (inet_sk(sk)->num == num) {
93 struct ipv6_pinfo *np = inet6_sk(sk);
95 if (!ipv6_addr_any(&np->daddr) &&
96 !ipv6_addr_equal(&np->daddr, rmt_addr))
97 continue;
99 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
100 continue;
102 if (!ipv6_addr_any(&np->rcv_saddr)) {
103 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
104 goto found;
105 if (is_multicast &&
106 inet6_mc_check(sk, loc_addr, rmt_addr))
107 goto found;
108 continue;
109 }
110 goto found;
111 }
112 sk = NULL;
113 found:
114 return sk;
115 }
117 /*
118 * 0 - deliver
119 * 1 - block
120 */
121 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
122 {
123 struct icmp6hdr *icmph;
124 struct raw6_sock *rp = raw6_sk(sk);
126 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
127 __u32 *data = &rp->filter.data[0];
128 int bit_nr;
130 icmph = (struct icmp6hdr *) skb->data;
131 bit_nr = icmph->icmp6_type;
133 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
134 }
135 return 0;
136 }
138 /*
139 * demultiplex raw sockets.
140 * (should consider queueing the skb in the sock receive_queue
141 * without calling rawv6.c)
142 *
143 * Caller owns SKB so we must make clones.
144 */
145 int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
146 {
147 struct in6_addr *saddr;
148 struct in6_addr *daddr;
149 struct sock *sk;
150 int delivered = 0;
151 __u8 hash;
153 saddr = &skb->nh.ipv6h->saddr;
154 daddr = saddr + 1;
156 hash = nexthdr & (MAX_INET_PROTOS - 1);
158 read_lock(&raw_v6_lock);
159 sk = sk_head(&raw_v6_htable[hash]);
161 /*
162 * The first socket found will be delivered after
163 * delivery to transport protocols.
164 */
166 if (sk == NULL)
167 goto out;
169 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
171 while (sk) {
172 delivered = 1;
173 if (nexthdr != IPPROTO_ICMPV6 || !icmpv6_filter(sk, skb)) {
174 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
176 /* Not releasing hash table! */
177 if (clone) {
178 nf_reset(clone);
179 rawv6_rcv(sk, clone);
180 }
181 }
182 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr,
183 IP6CB(skb)->iif);
184 }
185 out:
186 read_unlock(&raw_v6_lock);
187 return delivered;
188 }
190 /* This cleans up af_inet6 a bit. -DaveM */
191 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
192 {
193 struct inet_sock *inet = inet_sk(sk);
194 struct ipv6_pinfo *np = inet6_sk(sk);
195 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
196 __u32 v4addr = 0;
197 int addr_type;
198 int err;
200 if (addr_len < SIN6_LEN_RFC2133)
201 return -EINVAL;
202 addr_type = ipv6_addr_type(&addr->sin6_addr);
204 /* Raw sockets are IPv6 only */
205 if (addr_type == IPV6_ADDR_MAPPED)
206 return(-EADDRNOTAVAIL);
208 lock_sock(sk);
210 err = -EINVAL;
211 if (sk->sk_state != TCP_CLOSE)
212 goto out;
214 /* Check if the address belongs to the host. */
215 if (addr_type != IPV6_ADDR_ANY) {
216 struct net_device *dev = NULL;
218 if (addr_type & IPV6_ADDR_LINKLOCAL) {
219 if (addr_len >= sizeof(struct sockaddr_in6) &&
220 addr->sin6_scope_id) {
221 /* Override any existing binding, if another
222 * one is supplied by user.
223 */
224 sk->sk_bound_dev_if = addr->sin6_scope_id;
225 }
227 /* Binding to link-local address requires an interface */
228 if (!sk->sk_bound_dev_if)
229 goto out;
231 dev = dev_get_by_index(sk->sk_bound_dev_if);
232 if (!dev) {
233 err = -ENODEV;
234 goto out;
235 }
236 }
238 /* ipv4 addr of the socket is invalid. Only the
239 * unspecified and mapped address have a v4 equivalent.
240 */
241 v4addr = LOOPBACK4_IPV6;
242 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
243 err = -EADDRNOTAVAIL;
244 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
245 if (dev)
246 dev_put(dev);
247 goto out;
248 }
249 }
250 if (dev)
251 dev_put(dev);
252 }
254 inet->rcv_saddr = inet->saddr = v4addr;
255 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
256 if (!(addr_type & IPV6_ADDR_MULTICAST))
257 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
258 err = 0;
259 out:
260 release_sock(sk);
261 return err;
262 }
264 void rawv6_err(struct sock *sk, struct sk_buff *skb,
265 struct inet6_skb_parm *opt,
266 int type, int code, int offset, u32 info)
267 {
268 struct inet_sock *inet = inet_sk(sk);
269 struct ipv6_pinfo *np = inet6_sk(sk);
270 int err;
271 int harderr;
273 /* Report error on raw socket, if:
274 1. User requested recverr.
275 2. Socket is connected (otherwise the error indication
276 is useless without recverr and error is hard.
277 */
278 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
279 return;
281 harderr = icmpv6_err_convert(type, code, &err);
282 if (type == ICMPV6_PKT_TOOBIG)
283 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
285 if (np->recverr) {
286 u8 *payload = skb->data;
287 if (!inet->hdrincl)
288 payload += offset;
289 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
290 }
292 if (np->recverr || harderr) {
293 sk->sk_err = err;
294 sk->sk_error_report(sk);
295 }
296 }
298 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
299 {
300 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
301 skb_checksum_complete(skb)) {
302 /* FIXME: increment a raw6 drops counter here */
303 kfree_skb(skb);
304 return 0;
305 }
307 /* Charge it to the socket. */
308 if (sock_queue_rcv_skb(sk,skb)<0) {
309 /* FIXME: increment a raw6 drops counter here */
310 kfree_skb(skb);
311 return 0;
312 }
314 return 0;
315 }
317 /*
318 * This is next to useless...
319 * if we demultiplex in network layer we don't need the extra call
320 * just to queue the skb...
321 * maybe we could have the network decide upon a hint if it
322 * should call raw_rcv for demultiplexing
323 */
324 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
325 {
326 struct inet_sock *inet = inet_sk(sk);
327 struct raw6_sock *rp = raw6_sk(sk);
329 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
330 kfree_skb(skb);
331 return NET_RX_DROP;
332 }
334 if (!rp->checksum)
335 skb->ip_summed = CHECKSUM_UNNECESSARY;
337 if (skb->ip_summed == CHECKSUM_HW) {
338 skb_postpull_rcsum(skb, skb->nh.raw,
339 skb->h.raw - skb->nh.raw);
340 if (!csum_ipv6_magic(&skb->nh.ipv6h->saddr,
341 &skb->nh.ipv6h->daddr,
342 skb->len, inet->num, skb->csum))
343 skb->ip_summed = CHECKSUM_UNNECESSARY;
344 }
345 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
346 skb->csum = ~csum_ipv6_magic(&skb->nh.ipv6h->saddr,
347 &skb->nh.ipv6h->daddr,
348 skb->len, inet->num, 0);
350 if (inet->hdrincl) {
351 if (skb_checksum_complete(skb)) {
352 /* FIXME: increment a raw6 drops counter here */
353 kfree_skb(skb);
354 return 0;
355 }
356 }
358 rawv6_rcv_skb(sk, skb);
359 return 0;
360 }
363 /*
364 * This should be easy, if there is something there
365 * we return it, otherwise we block.
366 */
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;
378 if (flags & MSG_OOB)
379 return -EOPNOTSUPP;
381 if (addr_len)
382 *addr_len=sizeof(*sin6);
384 if (flags & MSG_ERRQUEUE)
385 return ipv6_recv_error(sk, msg, len);
387 skb = skb_recv_datagram(sk, flags, noblock, &err);
388 if (!skb)
389 goto out;
391 copied = skb->len;
392 if (copied > len) {
393 copied = len;
394 msg->msg_flags |= MSG_TRUNC;
395 }
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 (__skb_checksum_complete(skb))
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;
411 /* Copy the address. */
412 if (sin6) {
413 sin6->sin6_family = AF_INET6;
414 sin6->sin6_port = 0;
415 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
416 sin6->sin6_flowinfo = 0;
417 sin6->sin6_scope_id = 0;
418 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
419 sin6->sin6_scope_id = IP6CB(skb)->iif;
420 }
422 sock_recv_timestamp(msg, sk, skb);
424 if (np->rxopt.all)
425 datagram_recv_ctl(sk, msg, skb);
427 err = copied;
428 if (flags & MSG_TRUNC)
429 err = skb->len;
431 out_free:
432 skb_free_datagram(sk, skb);
433 out:
434 return err;
436 csum_copy_err:
437 skb_kill_datagram(sk, skb, flags);
439 /* Error for blocking case is chosen to masquerade
440 as some normal condition.
441 */
442 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
443 /* FIXME: increment a raw6 drops counter here */
444 goto out;
445 }
447 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
448 struct raw6_sock *rp)
449 {
450 struct sk_buff *skb;
451 int err = 0;
452 int offset;
453 int len;
454 int total_len;
455 u32 tmp_csum;
456 u16 csum;
458 if (!rp->checksum)
459 goto send;
461 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
462 goto out;
464 offset = rp->offset;
465 total_len = inet_sk(sk)->cork.length - (skb->nh.raw - skb->data);
466 if (offset >= total_len - 1) {
467 err = -EINVAL;
468 ip6_flush_pending_frames(sk);
469 goto out;
470 }
472 /* should be check HW csum miyazawa */
473 if (skb_queue_len(&sk->sk_write_queue) == 1) {
474 /*
475 * Only one fragment on the socket.
476 */
477 tmp_csum = skb->csum;
478 } else {
479 struct sk_buff *csum_skb = NULL;
480 tmp_csum = 0;
482 skb_queue_walk(&sk->sk_write_queue, skb) {
483 tmp_csum = csum_add(tmp_csum, skb->csum);
485 if (csum_skb)
486 continue;
488 len = skb->len - (skb->h.raw - skb->data);
489 if (offset >= len) {
490 offset -= len;
491 continue;
492 }
494 csum_skb = skb;
495 }
497 skb = csum_skb;
498 }
500 offset += skb->h.raw - skb->data;
501 if (skb_copy_bits(skb, offset, &csum, 2))
502 BUG();
504 /* in case cksum was not initialized */
505 if (unlikely(csum))
506 tmp_csum = csum_sub(tmp_csum, csum);
508 tmp_csum = csum_ipv6_magic(&fl->fl6_src,
509 &fl->fl6_dst,
510 total_len, fl->proto, tmp_csum);
512 if (tmp_csum == 0)
513 tmp_csum = -1;
515 csum = tmp_csum;
516 if (skb_store_bits(skb, offset, &csum, 2))
517 BUG();
519 send:
520 err = ip6_push_pending_frames(sk);
521 out:
522 return err;
523 }
525 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
526 struct flowi *fl, struct rt6_info *rt,
527 unsigned int flags)
528 {
529 struct ipv6_pinfo *np = inet6_sk(sk);
530 struct ipv6hdr *iph;
531 struct sk_buff *skb;
532 unsigned int hh_len;
533 int err;
535 if (length > rt->u.dst.dev->mtu) {
536 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
537 return -EMSGSIZE;
538 }
539 if (flags&MSG_PROBE)
540 goto out;
542 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
544 skb = sock_alloc_send_skb(sk, length+hh_len+15,
545 flags&MSG_DONTWAIT, &err);
546 if (skb == NULL)
547 goto error;
548 skb_reserve(skb, hh_len);
550 skb->priority = sk->sk_priority;
551 skb->dst = dst_clone(&rt->u.dst);
553 skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length);
555 skb->ip_summed = CHECKSUM_NONE;
557 skb->h.raw = skb->nh.raw;
558 err = memcpy_fromiovecend((void *)iph, from, 0, length);
559 if (err)
560 goto error_fault;
562 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
563 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
564 dst_output);
565 if (err > 0)
566 err = np->recverr ? net_xmit_errno(err) : 0;
567 if (err)
568 goto error;
569 out:
570 return 0;
572 error_fault:
573 err = -EFAULT;
574 kfree_skb(skb);
575 error:
576 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
577 return err;
578 }
580 static void rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
581 {
582 struct iovec *iov;
583 u8 __user *type = NULL;
584 u8 __user *code = NULL;
585 int probed = 0;
586 int i;
588 if (!msg->msg_iov)
589 return;
591 for (i = 0; i < msg->msg_iovlen; i++) {
592 iov = &msg->msg_iov[i];
593 if (!iov)
594 continue;
596 switch (fl->proto) {
597 case IPPROTO_ICMPV6:
598 /* check if one-byte field is readable or not. */
599 if (iov->iov_base && iov->iov_len < 1)
600 break;
602 if (!type) {
603 type = iov->iov_base;
604 /* check if code field is readable or not. */
605 if (iov->iov_len > 1)
606 code = type + 1;
607 } else if (!code)
608 code = iov->iov_base;
610 if (type && code) {
611 get_user(fl->fl_icmp_type, type);
612 get_user(fl->fl_icmp_code, code);
613 probed = 1;
614 }
615 break;
616 default:
617 probed = 1;
618 break;
619 }
620 if (probed)
621 break;
622 }
623 }
625 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
626 struct msghdr *msg, size_t len)
627 {
628 struct ipv6_txoptions opt_space;
629 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
630 struct in6_addr *daddr, *final_p = NULL, final;
631 struct inet_sock *inet = inet_sk(sk);
632 struct ipv6_pinfo *np = inet6_sk(sk);
633 struct raw6_sock *rp = raw6_sk(sk);
634 struct ipv6_txoptions *opt = NULL;
635 struct ip6_flowlabel *flowlabel = NULL;
636 struct dst_entry *dst = NULL;
637 struct flowi fl;
638 int addr_len = msg->msg_namelen;
639 int hlimit = -1;
640 int tclass = -1;
641 u16 proto;
642 int err;
644 /* Rough check on arithmetic overflow,
645 better check is made in ip6_build_xmit
646 */
647 if (len < 0)
648 return -EMSGSIZE;
650 /* Mirror BSD error message compatibility */
651 if (msg->msg_flags & MSG_OOB)
652 return -EOPNOTSUPP;
654 /*
655 * Get and verify the address.
656 */
657 memset(&fl, 0, sizeof(fl));
659 if (sin6) {
660 if (addr_len < SIN6_LEN_RFC2133)
661 return -EINVAL;
663 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
664 return(-EAFNOSUPPORT);
666 /* port is the proto value [0..255] carried in nexthdr */
667 proto = ntohs(sin6->sin6_port);
669 if (!proto)
670 proto = inet->num;
671 else if (proto != inet->num)
672 return(-EINVAL);
674 if (proto > 255)
675 return(-EINVAL);
677 daddr = &sin6->sin6_addr;
678 if (np->sndflow) {
679 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
680 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
681 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
682 if (flowlabel == NULL)
683 return -EINVAL;
684 daddr = &flowlabel->dst;
685 }
686 }
688 /*
689 * Otherwise it will be difficult to maintain
690 * sk->sk_dst_cache.
691 */
692 if (sk->sk_state == TCP_ESTABLISHED &&
693 ipv6_addr_equal(daddr, &np->daddr))
694 daddr = &np->daddr;
696 if (addr_len >= sizeof(struct sockaddr_in6) &&
697 sin6->sin6_scope_id &&
698 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
699 fl.oif = sin6->sin6_scope_id;
700 } else {
701 if (sk->sk_state != TCP_ESTABLISHED)
702 return -EDESTADDRREQ;
704 proto = inet->num;
705 daddr = &np->daddr;
706 fl.fl6_flowlabel = np->flow_label;
707 }
709 if (ipv6_addr_any(daddr)) {
710 /*
711 * unspecified destination address
712 * treated as error... is this correct ?
713 */
714 fl6_sock_release(flowlabel);
715 return(-EINVAL);
716 }
718 if (fl.oif == 0)
719 fl.oif = sk->sk_bound_dev_if;
721 if (msg->msg_controllen) {
722 opt = &opt_space;
723 memset(opt, 0, sizeof(struct ipv6_txoptions));
724 opt->tot_len = sizeof(struct ipv6_txoptions);
726 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
727 if (err < 0) {
728 fl6_sock_release(flowlabel);
729 return err;
730 }
731 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
732 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
733 if (flowlabel == NULL)
734 return -EINVAL;
735 }
736 if (!(opt->opt_nflen|opt->opt_flen))
737 opt = NULL;
738 }
739 if (opt == NULL)
740 opt = np->opt;
741 if (flowlabel)
742 opt = fl6_merge_options(&opt_space, flowlabel, opt);
743 opt = ipv6_fixup_options(&opt_space, opt);
745 fl.proto = proto;
746 rawv6_probe_proto_opt(&fl, msg);
748 ipv6_addr_copy(&fl.fl6_dst, daddr);
749 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
750 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
752 /* merge ip6_build_xmit from ip6_output */
753 if (opt && opt->srcrt) {
754 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
755 ipv6_addr_copy(&final, &fl.fl6_dst);
756 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
757 final_p = &final;
758 }
760 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
761 fl.oif = np->mcast_oif;
763 err = ip6_dst_lookup(sk, &dst, &fl);
764 if (err)
765 goto out;
766 if (final_p)
767 ipv6_addr_copy(&fl.fl6_dst, final_p);
769 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
770 goto out;
772 if (hlimit < 0) {
773 if (ipv6_addr_is_multicast(&fl.fl6_dst))
774 hlimit = np->mcast_hops;
775 else
776 hlimit = np->hop_limit;
777 if (hlimit < 0)
778 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
779 if (hlimit < 0)
780 hlimit = ipv6_get_hoplimit(dst->dev);
781 }
783 if (tclass < 0) {
784 tclass = np->tclass;
785 if (tclass < 0)
786 tclass = 0;
787 }
789 if (msg->msg_flags&MSG_CONFIRM)
790 goto do_confirm;
792 back_from_confirm:
793 if (inet->hdrincl) {
794 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
795 } else {
796 lock_sock(sk);
797 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
798 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
799 msg->msg_flags);
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 dst_release(dst);
808 release_sock(sk);
809 out:
810 fl6_sock_release(flowlabel);
811 return err<0?err:len;
812 do_confirm:
813 dst_confirm(dst);
814 if (!(msg->msg_flags & MSG_PROBE) || len)
815 goto back_from_confirm;
816 err = 0;
817 goto done;
818 }
820 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
821 char __user *optval, int optlen)
822 {
823 switch (optname) {
824 case ICMPV6_FILTER:
825 if (optlen > sizeof(struct icmp6_filter))
826 optlen = sizeof(struct icmp6_filter);
827 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
828 return -EFAULT;
829 return 0;
830 default:
831 return -ENOPROTOOPT;
832 };
834 return 0;
835 }
837 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
838 char __user *optval, int __user *optlen)
839 {
840 int len;
842 switch (optname) {
843 case ICMPV6_FILTER:
844 if (get_user(len, optlen))
845 return -EFAULT;
846 if (len < 0)
847 return -EINVAL;
848 if (len > sizeof(struct icmp6_filter))
849 len = sizeof(struct icmp6_filter);
850 if (put_user(len, optlen))
851 return -EFAULT;
852 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
853 return -EFAULT;
854 return 0;
855 default:
856 return -ENOPROTOOPT;
857 };
859 return 0;
860 }
863 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
864 char __user *optval, int optlen)
865 {
866 struct raw6_sock *rp = raw6_sk(sk);
867 int val;
869 if (get_user(val, (int __user *)optval))
870 return -EFAULT;
872 switch (optname) {
873 case IPV6_CHECKSUM:
874 /* You may get strange result with a positive odd offset;
875 RFC2292bis agrees with me. */
876 if (val > 0 && (val&1))
877 return(-EINVAL);
878 if (val < 0) {
879 rp->checksum = 0;
880 } else {
881 rp->checksum = 1;
882 rp->offset = val;
883 }
885 return 0;
886 break;
888 default:
889 return(-ENOPROTOOPT);
890 }
891 }
893 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
894 char __user *optval, int optlen)
895 {
896 switch(level) {
897 case SOL_RAW:
898 break;
900 case SOL_ICMPV6:
901 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
902 return -EOPNOTSUPP;
903 return rawv6_seticmpfilter(sk, level, optname, optval,
904 optlen);
905 case SOL_IPV6:
906 if (optname == IPV6_CHECKSUM)
907 break;
908 default:
909 return ipv6_setsockopt(sk, level, optname, optval,
910 optlen);
911 };
912 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
913 }
915 #ifdef CONFIG_COMPAT
916 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
917 char __user *optval, int optlen)
918 {
919 switch (level) {
920 case SOL_RAW:
921 break;
922 case SOL_ICMPV6:
923 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
924 return -EOPNOTSUPP;
925 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
926 case SOL_IPV6:
927 if (optname == IPV6_CHECKSUM)
928 break;
929 default:
930 return compat_ipv6_setsockopt(sk, level, optname,
931 optval, optlen);
932 };
933 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
934 }
935 #endif
937 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
938 char __user *optval, int __user *optlen)
939 {
940 struct raw6_sock *rp = raw6_sk(sk);
941 int val, len;
943 if (get_user(len,optlen))
944 return -EFAULT;
946 switch (optname) {
947 case IPV6_CHECKSUM:
948 if (rp->checksum == 0)
949 val = -1;
950 else
951 val = rp->offset;
952 break;
954 default:
955 return -ENOPROTOOPT;
956 }
958 len = min_t(unsigned int, sizeof(int), len);
960 if (put_user(len, optlen))
961 return -EFAULT;
962 if (copy_to_user(optval,&val,len))
963 return -EFAULT;
964 return 0;
965 }
967 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
968 char __user *optval, int __user *optlen)
969 {
970 switch(level) {
971 case SOL_RAW:
972 break;
974 case SOL_ICMPV6:
975 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
976 return -EOPNOTSUPP;
977 return rawv6_geticmpfilter(sk, level, optname, optval,
978 optlen);
979 case SOL_IPV6:
980 if (optname == IPV6_CHECKSUM)
981 break;
982 default:
983 return ipv6_getsockopt(sk, level, optname, optval,
984 optlen);
985 };
986 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
987 }
989 #ifdef CONFIG_COMPAT
990 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
991 char __user *optval, int __user *optlen)
992 {
993 switch (level) {
994 case SOL_RAW:
995 break;
996 case SOL_ICMPV6:
997 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
998 return -EOPNOTSUPP;
999 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1000 case SOL_IPV6:
1001 if (optname == IPV6_CHECKSUM)
1002 break;
1003 default:
1004 return compat_ipv6_getsockopt(sk, level, optname,
1005 optval, optlen);
1006 };
1007 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1009 #endif
1011 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1013 switch(cmd) {
1014 case SIOCOUTQ:
1016 int amount = atomic_read(&sk->sk_wmem_alloc);
1017 return put_user(amount, (int __user *)arg);
1019 case SIOCINQ:
1021 struct sk_buff *skb;
1022 int amount = 0;
1024 spin_lock_bh(&sk->sk_receive_queue.lock);
1025 skb = skb_peek(&sk->sk_receive_queue);
1026 if (skb != NULL)
1027 amount = skb->tail - skb->h.raw;
1028 spin_unlock_bh(&sk->sk_receive_queue.lock);
1029 return put_user(amount, (int __user *)arg);
1032 default:
1033 return -ENOIOCTLCMD;
1037 static void rawv6_close(struct sock *sk, long timeout)
1039 if (inet_sk(sk)->num == IPPROTO_RAW)
1040 ip6_ra_control(sk, -1, NULL);
1042 sk_common_release(sk);
1045 static int rawv6_init_sk(struct sock *sk)
1047 if (inet_sk(sk)->num == IPPROTO_ICMPV6) {
1048 struct raw6_sock *rp = raw6_sk(sk);
1049 rp->checksum = 1;
1050 rp->offset = 2;
1052 return(0);
1055 struct proto rawv6_prot = {
1056 .name = "RAWv6",
1057 .owner = THIS_MODULE,
1058 .close = rawv6_close,
1059 .connect = ip6_datagram_connect,
1060 .disconnect = udp_disconnect,
1061 .ioctl = rawv6_ioctl,
1062 .init = rawv6_init_sk,
1063 .destroy = inet6_destroy_sock,
1064 .setsockopt = rawv6_setsockopt,
1065 .getsockopt = rawv6_getsockopt,
1066 .sendmsg = rawv6_sendmsg,
1067 .recvmsg = rawv6_recvmsg,
1068 .bind = rawv6_bind,
1069 .backlog_rcv = rawv6_rcv_skb,
1070 .hash = raw_v6_hash,
1071 .unhash = raw_v6_unhash,
1072 .obj_size = sizeof(struct raw6_sock),
1073 #ifdef CONFIG_COMPAT
1074 .compat_setsockopt = compat_rawv6_setsockopt,
1075 .compat_getsockopt = compat_rawv6_getsockopt,
1076 #endif
1077 };
1079 #ifdef CONFIG_PROC_FS
1080 struct raw6_iter_state {
1081 int bucket;
1082 };
1084 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1086 static struct sock *raw6_get_first(struct seq_file *seq)
1088 struct sock *sk;
1089 struct hlist_node *node;
1090 struct raw6_iter_state* state = raw6_seq_private(seq);
1092 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1093 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1094 if (sk->sk_family == PF_INET6)
1095 goto out;
1096 sk = NULL;
1097 out:
1098 return sk;
1101 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1103 struct raw6_iter_state* state = raw6_seq_private(seq);
1105 do {
1106 sk = sk_next(sk);
1107 try_again:
1109 } while (sk && sk->sk_family != PF_INET6);
1111 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1112 sk = sk_head(&raw_v6_htable[state->bucket]);
1113 goto try_again;
1115 return sk;
1118 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1120 struct sock *sk = raw6_get_first(seq);
1121 if (sk)
1122 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1123 --pos;
1124 return pos ? NULL : sk;
1127 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1129 read_lock(&raw_v6_lock);
1130 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1133 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1135 struct sock *sk;
1137 if (v == SEQ_START_TOKEN)
1138 sk = raw6_get_first(seq);
1139 else
1140 sk = raw6_get_next(seq, v);
1141 ++*pos;
1142 return sk;
1145 static void raw6_seq_stop(struct seq_file *seq, void *v)
1147 read_unlock(&raw_v6_lock);
1150 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1152 struct ipv6_pinfo *np = inet6_sk(sp);
1153 struct in6_addr *dest, *src;
1154 __u16 destp, srcp;
1156 dest = &np->daddr;
1157 src = &np->rcv_saddr;
1158 destp = 0;
1159 srcp = inet_sk(sp)->num;
1160 seq_printf(seq,
1161 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1162 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1163 i,
1164 src->s6_addr32[0], src->s6_addr32[1],
1165 src->s6_addr32[2], src->s6_addr32[3], srcp,
1166 dest->s6_addr32[0], dest->s6_addr32[1],
1167 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1168 sp->sk_state,
1169 atomic_read(&sp->sk_wmem_alloc),
1170 atomic_read(&sp->sk_rmem_alloc),
1171 0, 0L, 0,
1172 sock_i_uid(sp), 0,
1173 sock_i_ino(sp),
1174 atomic_read(&sp->sk_refcnt), sp);
1177 static int raw6_seq_show(struct seq_file *seq, void *v)
1179 if (v == SEQ_START_TOKEN)
1180 seq_printf(seq,
1181 " sl "
1182 "local_address "
1183 "remote_address "
1184 "st tx_queue rx_queue tr tm->when retrnsmt"
1185 " uid timeout inode\n");
1186 else
1187 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1188 return 0;
1191 static struct seq_operations raw6_seq_ops = {
1192 .start = raw6_seq_start,
1193 .next = raw6_seq_next,
1194 .stop = raw6_seq_stop,
1195 .show = raw6_seq_show,
1196 };
1198 static int raw6_seq_open(struct inode *inode, struct file *file)
1200 struct seq_file *seq;
1201 int rc = -ENOMEM;
1202 struct raw6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
1203 if (!s)
1204 goto out;
1205 rc = seq_open(file, &raw6_seq_ops);
1206 if (rc)
1207 goto out_kfree;
1208 seq = file->private_data;
1209 seq->private = s;
1210 out:
1211 return rc;
1212 out_kfree:
1213 kfree(s);
1214 goto out;
1217 static struct file_operations raw6_seq_fops = {
1218 .owner = THIS_MODULE,
1219 .open = raw6_seq_open,
1220 .read = seq_read,
1221 .llseek = seq_lseek,
1222 .release = seq_release_private,
1223 };
1225 int __init raw6_proc_init(void)
1227 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1228 return -ENOMEM;
1229 return 0;
1232 void raw6_proc_exit(void)
1234 proc_net_remove("raw6");
1236 #endif /* CONFIG_PROC_FS */