File: /usr/src/linux/net/ipv6/sit.c

1     /*
2      *	IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
3      *	Linux INET6 implementation
4      *
5      *	Authors:
6      *	Pedro Roque		<roque@di.fc.ul.pt>	
7      *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
8      *
9      *	$Id: sit.c,v 1.52 2001/09/01 00:31:50 davem Exp $
10      *
11      *	This program is free software; you can redistribute it and/or
12      *      modify it under the terms of the GNU General Public License
13      *      as published by the Free Software Foundation; either version
14      *      2 of the License, or (at your option) any later version.
15      *
16      *	Changes:
17      * Roger Venning <r.venning@telstra.com>:	6to4 support
18      * Nate Thompson <nate@thebog.net>:		6to4 support
19      */
20     
21     #define __NO_VERSION__
22     #include <linux/config.h>
23     #include <linux/module.h>
24     #include <linux/errno.h>
25     #include <linux/types.h>
26     #include <linux/socket.h>
27     #include <linux/sockios.h>
28     #include <linux/sched.h>
29     #include <linux/net.h>
30     #include <linux/in6.h>
31     #include <linux/netdevice.h>
32     #include <linux/if_arp.h>
33     #include <linux/icmp.h>
34     #include <asm/uaccess.h>
35     #include <linux/init.h>
36     #include <linux/netfilter_ipv4.h>
37     
38     #include <net/sock.h>
39     #include <net/snmp.h>
40     
41     #include <net/ipv6.h>
42     #include <net/protocol.h>
43     #include <net/transp_v6.h>
44     #include <net/ip6_fib.h>
45     #include <net/ip6_route.h>
46     #include <net/ndisc.h>
47     #include <net/addrconf.h>
48     #include <net/ip.h>
49     #include <net/udp.h>
50     #include <net/icmp.h>
51     #include <net/ipip.h>
52     #include <net/inet_ecn.h>
53     
54     /*
55        This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
56     
57        For comments look at net/ipv4/ip_gre.c --ANK
58      */
59     
60     #define HASH_SIZE  16
61     #define HASH(addr) ((addr^(addr>>4))&0xF)
62     
63     static int ipip6_fb_tunnel_init(struct net_device *dev);
64     static int ipip6_tunnel_init(struct net_device *dev);
65     
66     static struct net_device ipip6_fb_tunnel_dev = {
67     	"sit0", 0x0, 0x0, 0x0, 0x0, 0, 0, 0, 0, 0, NULL, ipip6_fb_tunnel_init,
68     };
69     
70     static struct ip_tunnel ipip6_fb_tunnel = {
71     	NULL, &ipip6_fb_tunnel_dev, {0, }, 0, 0, 0, 0, 0, 0, 0, {"sit0", }
72     };
73     
74     static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
75     static struct ip_tunnel *tunnels_r[HASH_SIZE];
76     static struct ip_tunnel *tunnels_l[HASH_SIZE];
77     static struct ip_tunnel *tunnels_wc[1];
78     static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
79     
80     static rwlock_t ipip6_lock = RW_LOCK_UNLOCKED;
81     
82     static struct ip_tunnel * ipip6_tunnel_lookup(u32 remote, u32 local)
83     {
84     	unsigned h0 = HASH(remote);
85     	unsigned h1 = HASH(local);
86     	struct ip_tunnel *t;
87     
88     	for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
89     		if (local == t->parms.iph.saddr &&
90     		    remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
91     			return t;
92     	}
93     	for (t = tunnels_r[h0]; t; t = t->next) {
94     		if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
95     			return t;
96     	}
97     	for (t = tunnels_l[h1]; t; t = t->next) {
98     		if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
99     			return t;
100     	}
101     	if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
102     		return t;
103     	return NULL;
104     }
105     
106     static struct ip_tunnel ** ipip6_bucket(struct ip_tunnel *t)
107     {
108     	u32 remote = t->parms.iph.daddr;
109     	u32 local = t->parms.iph.saddr;
110     	unsigned h = 0;
111     	int prio = 0;
112     
113     	if (remote) {
114     		prio |= 2;
115     		h ^= HASH(remote);
116     	}
117     	if (local) {
118     		prio |= 1;
119     		h ^= HASH(local);
120     	}
121     	return &tunnels[prio][h];
122     }
123     
124     static void ipip6_tunnel_unlink(struct ip_tunnel *t)
125     {
126     	struct ip_tunnel **tp;
127     
128     	for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
129     		if (t == *tp) {
130     			write_lock_bh(&ipip6_lock);
131     			*tp = t->next;
132     			write_unlock_bh(&ipip6_lock);
133     			break;
134     		}
135     	}
136     }
137     
138     static void ipip6_tunnel_link(struct ip_tunnel *t)
139     {
140     	struct ip_tunnel **tp = ipip6_bucket(t);
141     
142     	write_lock_bh(&ipip6_lock);
143     	t->next = *tp;
144     	write_unlock_bh(&ipip6_lock);
145     	*tp = t;
146     }
147     
148     struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
149     {
150     	u32 remote = parms->iph.daddr;
151     	u32 local = parms->iph.saddr;
152     	struct ip_tunnel *t, **tp, *nt;
153     	struct net_device *dev;
154     	unsigned h = 0;
155     	int prio = 0;
156     
157     	if (remote) {
158     		prio |= 2;
159     		h ^= HASH(remote);
160     	}
161     	if (local) {
162     		prio |= 1;
163     		h ^= HASH(local);
164     	}
165     	for (tp = &tunnels[prio][h]; (t = *tp) != NULL; tp = &t->next) {
166     		if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
167     			return t;
168     	}
169     	if (!create)
170     		return NULL;
171     
172     	MOD_INC_USE_COUNT;
173     	dev = kmalloc(sizeof(*dev) + sizeof(*t), GFP_KERNEL);
174     	if (dev == NULL) {
175     		MOD_DEC_USE_COUNT;
176     		return NULL;
177     	}
178     	memset(dev, 0, sizeof(*dev) + sizeof(*t));
179     	dev->priv = (void*)(dev+1);
180     	nt = (struct ip_tunnel*)dev->priv;
181     	nt->dev = dev;
182     	dev->init = ipip6_tunnel_init;
183     	dev->features |= NETIF_F_DYNALLOC;
184     	memcpy(&nt->parms, parms, sizeof(*parms));
185     	nt->parms.name[IFNAMSIZ-1] = '\0';
186     	strcpy(dev->name, nt->parms.name);
187     	if (dev->name[0] == 0) {
188     		int i;
189     		for (i=1; i<100; i++) {
190     			sprintf(dev->name, "sit%d", i);
191     			if (__dev_get_by_name(dev->name) == NULL)
192     				break;
193     		}
194     		if (i==100)
195     			goto failed;
196     		memcpy(nt->parms.name, dev->name, IFNAMSIZ);
197     	}
198     	if (register_netdevice(dev) < 0)
199     		goto failed;
200     
201     	dev_hold(dev);
202     	ipip6_tunnel_link(nt);
203     	/* Do not decrement MOD_USE_COUNT here. */
204     	return nt;
205     
206     failed:
207     	kfree(dev);
208     	MOD_DEC_USE_COUNT;
209     	return NULL;
210     }
211     
212     static void ipip6_tunnel_destructor(struct net_device *dev)
213     {
214     	if (dev != &ipip6_fb_tunnel_dev) {
215     		MOD_DEC_USE_COUNT;
216     	}
217     }
218     
219     static void ipip6_tunnel_uninit(struct net_device *dev)
220     {
221     	if (dev == &ipip6_fb_tunnel_dev) {
222     		write_lock_bh(&ipip6_lock);
223     		tunnels_wc[0] = NULL;
224     		write_unlock_bh(&ipip6_lock);
225     		dev_put(dev);
226     	} else {
227     		ipip6_tunnel_unlink((struct ip_tunnel*)dev->priv);
228     		dev_put(dev);
229     	}
230     }
231     
232     
233     void ipip6_err(struct sk_buff *skb, u32 info)
234     {
235     #ifndef I_WISH_WORLD_WERE_PERFECT
236     
237     /* It is not :-( All the routers (except for Linux) return only
238        8 bytes of packet payload. It means, that precise relaying of
239        ICMP in the real Internet is absolutely infeasible.
240      */
241     	struct iphdr *iph = (struct iphdr*)skb->data;
242     	int type = skb->h.icmph->type;
243     	int code = skb->h.icmph->code;
244     	struct ip_tunnel *t;
245     
246     	switch (type) {
247     	default:
248     	case ICMP_PARAMETERPROB:
249     		return;
250     
251     	case ICMP_DEST_UNREACH:
252     		switch (code) {
253     		case ICMP_SR_FAILED:
254     		case ICMP_PORT_UNREACH:
255     			/* Impossible event. */
256     			return;
257     		case ICMP_FRAG_NEEDED:
258     			/* Soft state for pmtu is maintained by IP core. */
259     			return;
260     		default:
261     			/* All others are translated to HOST_UNREACH.
262     			   rfc2003 contains "deep thoughts" about NET_UNREACH,
263     			   I believe they are just ether pollution. --ANK
264     			 */
265     			break;
266     		}
267     		break;
268     	case ICMP_TIME_EXCEEDED:
269     		if (code != ICMP_EXC_TTL)
270     			return;
271     		break;
272     	}
273     
274     	read_lock(&ipip6_lock);
275     	t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
276     	if (t == NULL || t->parms.iph.daddr == 0)
277     		goto out;
278     	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
279     		goto out;
280     
281     	if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
282     		t->err_count++;
283     	else
284     		t->err_count = 1;
285     	t->err_time = jiffies;
286     out:
287     	read_unlock(&ipip6_lock);
288     	return;
289     #else
290     	struct iphdr *iph = (struct iphdr*)dp;
291     	int hlen = iph->ihl<<2;
292     	struct ipv6hdr *iph6;
293     	int type = skb->h.icmph->type;
294     	int code = skb->h.icmph->code;
295     	int rel_type = 0;
296     	int rel_code = 0;
297     	int rel_info = 0;
298     	struct sk_buff *skb2;
299     	struct rt6_info *rt6i;
300     
301     	if (len < hlen + sizeof(struct ipv6hdr))
302     		return;
303     	iph6 = (struct ipv6hdr*)(dp + hlen);
304     
305     	switch (type) {
306     	default:
307     		return;
308     	case ICMP_PARAMETERPROB:
309     		if (skb->h.icmph->un.gateway < hlen)
310     			return;
311     
312     		/* So... This guy found something strange INSIDE encapsulated
313     		   packet. Well, he is fool, but what can we do ?
314     		 */
315     		rel_type = ICMPV6_PARAMPROB;
316     		rel_info = skb->h.icmph->un.gateway - hlen;
317     		break;
318     
319     	case ICMP_DEST_UNREACH:
320     		switch (code) {
321     		case ICMP_SR_FAILED:
322     		case ICMP_PORT_UNREACH:
323     			/* Impossible event. */
324     			return;
325     		case ICMP_FRAG_NEEDED:
326     			/* Too complicated case ... */
327     			return;
328     		default:
329     			/* All others are translated to HOST_UNREACH.
330     			   rfc2003 contains "deep thoughts" about NET_UNREACH,
331     			   I believe, it is just ether pollution. --ANK
332     			 */
333     			rel_type = ICMPV6_DEST_UNREACH;
334     			rel_code = ICMPV6_ADDR_UNREACH;
335     			break;
336     		}
337     		break;
338     	case ICMP_TIME_EXCEEDED:
339     		if (code != ICMP_EXC_TTL)
340     			return;
341     		rel_type = ICMPV6_TIME_EXCEED;
342     		rel_code = ICMPV6_EXC_HOPLIMIT;
343     		break;
344     	}
345     
346     	/* Prepare fake skb to feed it to icmpv6_send */
347     	skb2 = skb_clone(skb, GFP_ATOMIC);
348     	if (skb2 == NULL)
349     		return;
350     	dst_release(skb2->dst);
351     	skb2->dst = NULL;
352     	skb_pull(skb2, skb->data - (u8*)iph6);
353     	skb2->nh.raw = skb2->data;
354     
355     	/* Try to guess incoming interface */
356     	rt6i = rt6_lookup(&iph6->saddr, NULL, NULL, 0);
357     	if (rt6i && rt6i->rt6i_dev) {
358     		skb2->dev = rt6i->rt6i_dev;
359     
360     		rt6i = rt6_lookup(&iph6->daddr, &iph6->saddr, NULL, 0);
361     
362     		if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
363     			struct ip_tunnel * t = (struct ip_tunnel*)rt6i->rt6i_dev->priv;
364     			if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
365     				rel_type = ICMPV6_DEST_UNREACH;
366     				rel_code = ICMPV6_ADDR_UNREACH;
367     			}
368     			icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
369     		}
370     	}
371     	kfree_skb(skb2);
372     	return;
373     #endif
374     }
375     
376     static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
377     {
378     	if (INET_ECN_is_ce(iph->tos) &&
379     	    INET_ECN_is_not_ce(ip6_get_dsfield(skb->nh.ipv6h)))
380     		IP6_ECN_set_ce(skb->nh.ipv6h);
381     }
382     
383     int ipip6_rcv(struct sk_buff *skb)
384     {
385     	struct iphdr *iph;
386     	struct ip_tunnel *tunnel;
387     
388     	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
389     		goto out;
390     
391     	iph = skb->nh.iph;
392     
393     	read_lock(&ipip6_lock);
394     	if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
395     		skb->mac.raw = skb->nh.raw;
396     		skb->nh.raw = skb->data;
397     		memset(&(IPCB(skb)->opt), 0, sizeof(struct ip_options));
398     		skb->protocol = __constant_htons(ETH_P_IPV6);
399     		skb->pkt_type = PACKET_HOST;
400     		tunnel->stat.rx_packets++;
401     		tunnel->stat.rx_bytes += skb->len;
402     		skb->dev = tunnel->dev;
403     		dst_release(skb->dst);
404     		skb->dst = NULL;
405     #ifdef CONFIG_NETFILTER
406     		nf_conntrack_put(skb->nfct);
407     		skb->nfct = NULL;
408     #ifdef CONFIG_NETFILTER_DEBUG
409     		skb->nf_debug = 0;
410     #endif
411     #endif
412     		ipip6_ecn_decapsulate(iph, skb);
413     		netif_rx(skb);
414     		read_unlock(&ipip6_lock);
415     		return 0;
416     	}
417     
418     	icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PROT_UNREACH, 0);
419     	kfree_skb(skb);
420     	read_unlock(&ipip6_lock);
421     out:
422     	return 0;
423     }
424     
425     /* Need this wrapper because NF_HOOK takes the function address */
426     static inline int do_ip_send(struct sk_buff *skb)
427     {
428     	return ip_send(skb);
429     }
430     
431     
432     /* Returns the embedded IPv4 address if the IPv6 address
433        comes from 6to4 (draft-ietf-ngtrans-6to4-04) addr space */
434     
435     static inline u32 try_6to4(struct in6_addr *v6dst)
436     {
437     	u32 dst = 0;
438     
439     	if (v6dst->s6_addr16[0] == htons(0x2002)) {
440     	        /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
441     		memcpy(&dst, &v6dst->s6_addr16[1], 4);
442     	}
443     	return dst;
444     }
445     
446     /*
447      *	This function assumes it is being called from dev_queue_xmit()
448      *	and that skb is filled properly by that function.
449      */
450     
451     static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
452     {
453     	struct ip_tunnel *tunnel = (struct ip_tunnel*)dev->priv;
454     	struct net_device_stats *stats = &tunnel->stat;
455     	struct iphdr  *tiph = &tunnel->parms.iph;
456     	struct ipv6hdr *iph6 = skb->nh.ipv6h;
457     	u8     tos = tunnel->parms.iph.tos;
458     	struct rtable *rt;     			/* Route to the other host */
459     	struct net_device *tdev;			/* Device to other host */
460     	struct iphdr  *iph;			/* Our new IP header */
461     	int    max_headroom;			/* The extra header space needed */
462     	u32    dst = tiph->daddr;
463     	int    mtu;
464     	struct in6_addr *addr6;	
465     	int addr_type;
466     
467     	if (tunnel->recursion++) {
468     		tunnel->stat.collisions++;
469     		goto tx_error;
470     	}
471     
472     	if (skb->protocol != __constant_htons(ETH_P_IPV6))
473     		goto tx_error;
474     
475     	if (!dst)
476     		dst = try_6to4(&iph6->daddr);
477     
478     	if (!dst) {
479     		struct neighbour *neigh = NULL;
480     
481     		if (skb->dst)
482     			neigh = skb->dst->neighbour;
483     
484     		if (neigh == NULL) {
485     			if (net_ratelimit())
486     				printk(KERN_DEBUG "sit: nexthop == NULL\n");
487     			goto tx_error;
488     		}
489     
490     		addr6 = (struct in6_addr*)&neigh->primary_key;
491     		addr_type = ipv6_addr_type(addr6);
492     
493     		if (addr_type == IPV6_ADDR_ANY) {
494     			addr6 = &skb->nh.ipv6h->daddr;
495     			addr_type = ipv6_addr_type(addr6);
496     		}
497     
498     		if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
499     			goto tx_error_icmp;
500     
501     		dst = addr6->s6_addr32[3];
502     	}
503     
504     	if (ip_route_output(&rt, dst, tiph->saddr, RT_TOS(tos), tunnel->parms.link)) {
505     		tunnel->stat.tx_carrier_errors++;
506     		goto tx_error_icmp;
507     	}
508     	if (rt->rt_type != RTN_UNICAST) {
509     		tunnel->stat.tx_carrier_errors++;
510     		goto tx_error_icmp;
511     	}
512     	tdev = rt->u.dst.dev;
513     
514     	if (tdev == dev) {
515     		ip_rt_put(rt);
516     		tunnel->stat.collisions++;
517     		goto tx_error;
518     	}
519     
520     	mtu = rt->u.dst.pmtu - sizeof(struct iphdr);
521     	if (mtu < 68) {
522     		tunnel->stat.collisions++;
523     		ip_rt_put(rt);
524     		goto tx_error;
525     	}
526     	if (mtu < IPV6_MIN_MTU)
527     		mtu = IPV6_MIN_MTU;
528     	if (skb->dst && mtu < skb->dst->pmtu) {
529     		struct rt6_info *rt6 = (struct rt6_info*)skb->dst;
530     		if (mtu < rt6->u.dst.pmtu) {
531     			if (tunnel->parms.iph.daddr || rt6->rt6i_dst.plen == 128) {
532     				rt6->rt6i_flags |= RTF_MODIFIED;
533     				rt6->u.dst.pmtu = mtu;
534     			}
535     		}
536     	}
537     	if (skb->len > mtu) {
538     		icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
539     		ip_rt_put(rt);
540     		goto tx_error;
541     	}
542     
543     	if (tunnel->err_count > 0) {
544     		if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
545     			tunnel->err_count--;
546     			dst_link_failure(skb);
547     		} else
548     			tunnel->err_count = 0;
549     	}
550     
551     	skb->h.raw = skb->nh.raw;
552     
553     	/*
554     	 * Okay, now see if we can stuff it in the buffer as-is.
555     	 */
556     	max_headroom = (((tdev->hard_header_len+15)&~15)+sizeof(struct iphdr));
557     
558     	if (skb_headroom(skb) < max_headroom || skb_cloned(skb) || skb_shared(skb)) {
559     		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
560     		if (!new_skb) {
561     			ip_rt_put(rt);
562       			stats->tx_dropped++;
563     			dev_kfree_skb(skb);
564     			tunnel->recursion--;
565     			return 0;
566     		}
567     		if (skb->sk)
568     			skb_set_owner_w(new_skb, skb->sk);
569     		dev_kfree_skb(skb);
570     		skb = new_skb;
571     	}
572     
573     	skb->nh.raw = skb_push(skb, sizeof(struct iphdr));
574     	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
575     	dst_release(skb->dst);
576     	skb->dst = &rt->u.dst;
577     
578     	/*
579     	 *	Push down and install the IPIP header.
580     	 */
581     
582     	iph 			=	skb->nh.iph;
583     	iph->version		=	4;
584     	iph->ihl		=	sizeof(struct iphdr)>>2;
585     	if (mtu > IPV6_MIN_MTU)
586     		iph->frag_off	=	__constant_htons(IP_DF);
587     	else
588     		iph->frag_off	=	0;
589     
590     	iph->protocol		=	IPPROTO_IPV6;
591     	iph->tos		=	INET_ECN_encapsulate(tos, ip6_get_dsfield(iph6));
592     	iph->daddr		=	rt->rt_dst;
593     	iph->saddr		=	rt->rt_src;
594     
595     	if ((iph->ttl = tiph->ttl) == 0)
596     		iph->ttl	=	iph6->hop_limit;
597     
598     #ifdef CONFIG_NETFILTER
599     	nf_conntrack_put(skb->nfct);
600     	skb->nfct = NULL;
601     #ifdef CONFIG_NETFILTER_DEBUG
602     	skb->nf_debug = 0;
603     #endif
604     #endif
605     
606     	IPTUNNEL_XMIT();
607     	tunnel->recursion--;
608     	return 0;
609     
610     tx_error_icmp:
611     	dst_link_failure(skb);
612     tx_error:
613     	stats->tx_errors++;
614     	dev_kfree_skb(skb);
615     	tunnel->recursion--;
616     	return 0;
617     }
618     
619     static int
620     ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
621     {
622     	int err = 0;
623     	struct ip_tunnel_parm p;
624     	struct ip_tunnel *t;
625     
626     	MOD_INC_USE_COUNT;
627     
628     	switch (cmd) {
629     	case SIOCGETTUNNEL:
630     		t = NULL;
631     		if (dev == &ipip6_fb_tunnel_dev) {
632     			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
633     				err = -EFAULT;
634     				break;
635     			}
636     			t = ipip6_tunnel_locate(&p, 0);
637     		}
638     		if (t == NULL)
639     			t = (struct ip_tunnel*)dev->priv;
640     		memcpy(&p, &t->parms, sizeof(p));
641     		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
642     			err = -EFAULT;
643     		break;
644     
645     	case SIOCADDTUNNEL:
646     	case SIOCCHGTUNNEL:
647     		err = -EPERM;
648     		if (!capable(CAP_NET_ADMIN))
649     			goto done;
650     
651     		err = -EFAULT;
652     		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
653     			goto done;
654     
655     		err = -EINVAL;
656     		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
657     		    p.iph.ihl != 5 || (p.iph.frag_off&__constant_htons(~IP_DF)))
658     			goto done;
659     		if (p.iph.ttl)
660     			p.iph.frag_off |= __constant_htons(IP_DF);
661     
662     		t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
663     
664     		if (dev != &ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL &&
665     		    t != &ipip6_fb_tunnel) {
666     			if (t != NULL) {
667     				if (t->dev != dev) {
668     					err = -EEXIST;
669     					break;
670     				}
671     			} else {
672     				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
673     				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
674     					err = -EINVAL;
675     					break;
676     				}
677     				t = (struct ip_tunnel*)dev->priv;
678     				ipip6_tunnel_unlink(t);
679     				t->parms.iph.saddr = p.iph.saddr;
680     				t->parms.iph.daddr = p.iph.daddr;
681     				memcpy(dev->dev_addr, &p.iph.saddr, 4);
682     				memcpy(dev->broadcast, &p.iph.daddr, 4);
683     				ipip6_tunnel_link(t);
684     				netdev_state_change(dev);
685     			}
686     		}
687     
688     		if (t) {
689     			err = 0;
690     			if (cmd == SIOCCHGTUNNEL) {
691     				t->parms.iph.ttl = p.iph.ttl;
692     				t->parms.iph.tos = p.iph.tos;
693     			}
694     			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
695     				err = -EFAULT;
696     		} else
697     			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
698     		break;
699     
700     	case SIOCDELTUNNEL:
701     		err = -EPERM;
702     		if (!capable(CAP_NET_ADMIN))
703     			goto done;
704     
705     		if (dev == &ipip6_fb_tunnel_dev) {
706     			err = -EFAULT;
707     			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
708     				goto done;
709     			err = -ENOENT;
710     			if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
711     				goto done;
712     			err = -EPERM;
713     			if (t == &ipip6_fb_tunnel)
714     				goto done;
715     		}
716     		err = unregister_netdevice(dev);
717     		break;
718     
719     	default:
720     		err = -EINVAL;
721     	}
722     
723     done:
724     	MOD_DEC_USE_COUNT;
725     	return err;
726     }
727     
728     static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
729     {
730     	return &(((struct ip_tunnel*)dev->priv)->stat);
731     }
732     
733     static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
734     {
735     	if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
736     		return -EINVAL;
737     	dev->mtu = new_mtu;
738     	return 0;
739     }
740     
741     static void ipip6_tunnel_init_gen(struct net_device *dev)
742     {
743     	struct ip_tunnel *t = (struct ip_tunnel*)dev->priv;
744     
745     	dev->destructor		= ipip6_tunnel_destructor;
746     	dev->uninit		= ipip6_tunnel_uninit;
747     	dev->hard_start_xmit	= ipip6_tunnel_xmit;
748     	dev->get_stats		= ipip6_tunnel_get_stats;
749     	dev->do_ioctl		= ipip6_tunnel_ioctl;
750     	dev->change_mtu		= ipip6_tunnel_change_mtu;
751     
752     	dev->type		= ARPHRD_SIT;
753     	dev->hard_header_len 	= LL_MAX_HEADER + sizeof(struct iphdr);
754     	dev->mtu		= 1500 - sizeof(struct iphdr);
755     	dev->flags		= IFF_NOARP;
756     	dev->iflink		= 0;
757     	dev->addr_len		= 4;
758     	memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
759     	memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
760     }
761     
762     static int ipip6_tunnel_init(struct net_device *dev)
763     {
764     	struct net_device *tdev = NULL;
765     	struct ip_tunnel *tunnel;
766     	struct iphdr *iph;
767     
768     	tunnel = (struct ip_tunnel*)dev->priv;
769     	iph = &tunnel->parms.iph;
770     
771     	ipip6_tunnel_init_gen(dev);
772     
773     	if (iph->daddr) {
774     		struct rtable *rt;
775     		if (!ip_route_output(&rt, iph->daddr, iph->saddr, RT_TOS(iph->tos), tunnel->parms.link)) {
776     			tdev = rt->u.dst.dev;
777     			ip_rt_put(rt);
778     		}
779     		dev->flags |= IFF_POINTOPOINT;
780     	}
781     
782     	if (!tdev && tunnel->parms.link)
783     		tdev = __dev_get_by_index(tunnel->parms.link);
784     
785     	if (tdev) {
786     		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
787     		dev->mtu = tdev->mtu - sizeof(struct iphdr);
788     		if (dev->mtu < IPV6_MIN_MTU)
789     			dev->mtu = IPV6_MIN_MTU;
790     	}
791     	dev->iflink = tunnel->parms.link;
792     
793     	return 0;
794     }
795     
796     #ifdef MODULE
797     static int ipip6_fb_tunnel_open(struct net_device *dev)
798     {
799     	MOD_INC_USE_COUNT;
800     	return 0;
801     }
802     
803     static int ipip6_fb_tunnel_close(struct net_device *dev)
804     {
805     	MOD_DEC_USE_COUNT;
806     	return 0;
807     }
808     #endif
809     
810     int __init ipip6_fb_tunnel_init(struct net_device *dev)
811     {
812     	struct iphdr *iph;
813     
814     	ipip6_tunnel_init_gen(dev);
815     #ifdef MODULE
816     	dev->open		= ipip6_fb_tunnel_open;
817     	dev->stop		= ipip6_fb_tunnel_close;
818     #endif
819     
820     	iph = &ipip6_fb_tunnel.parms.iph;
821     	iph->version		= 4;
822     	iph->protocol		= IPPROTO_IPV6;
823     	iph->ihl		= 5;
824     	iph->ttl		= 64;
825     
826     	dev_hold(dev);
827     	tunnels_wc[0]		= &ipip6_fb_tunnel;
828     	return 0;
829     }
830     
831     static struct inet_protocol sit_protocol = {
832     	ipip6_rcv,
833     	ipip6_err,
834     	0,
835     	IPPROTO_IPV6,
836     	0,
837     	NULL,
838     	"IPv6"
839     };
840     
841     #ifdef MODULE
842     void sit_cleanup(void)
843     {
844     	inet_del_protocol(&sit_protocol);
845     	unregister_netdev(&ipip6_fb_tunnel_dev);
846     }
847     #endif
848     
849     int __init sit_init(void)
850     {
851     	printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
852     
853     	ipip6_fb_tunnel_dev.priv = (void*)&ipip6_fb_tunnel;
854     	strcpy(ipip6_fb_tunnel_dev.name, ipip6_fb_tunnel.parms.name);
855     	register_netdev(&ipip6_fb_tunnel_dev);
856     	inet_add_protocol(&sit_protocol);
857     	return 0;
858     }
859