xfrm6_policy.c revision c82f963efe823d3cacaf1f1b7f1a35cc9628b188 
1/*
2 * xfrm6_policy.c: based on xfrm4_policy.c
3 *
4 * Authors:
5 *	Mitsuru KANDA @USAGI
6 * 	Kazunori MIYAZAWA @USAGI
7 * 	Kunihiro Ishiguro <kunihiro@ipinfusion.com>
8 * 		IPv6 support
9 * 	YOSHIFUJI Hideaki
10 * 		Split up af-specific portion
11 *
12 */
13
14#include <linux/compiler.h>
15#include <linux/netdevice.h>
16#include <net/addrconf.h>
17#include <net/xfrm.h>
18#include <net/ip.h>
19#include <net/ipv6.h>
20#include <net/ip6_route.h>
21#ifdef CONFIG_IPV6_MIP6
22#include <net/mip6.h>
23#endif
24
25static struct dst_ops xfrm6_dst_ops;
26static struct xfrm_policy_afinfo xfrm6_policy_afinfo;
27
28static int xfrm6_dst_lookup(struct xfrm_dst **xdst, struct flowi *fl)
29{
30	struct dst_entry *dst = ip6_route_output(NULL, fl);
31	int err = dst->error;
32	if (!err)
33		*xdst = (struct xfrm_dst *) dst;
34	else
35		dst_release(dst);
36	return err;
37}
38
39static int xfrm6_get_saddr(xfrm_address_t *saddr, xfrm_address_t *daddr)
40{
41	struct rt6_info *rt;
42	struct flowi fl_tunnel = {
43		.nl_u = {
44			.ip6_u = {
45				.daddr = *(struct in6_addr *)&daddr->a6,
46			},
47		},
48	};
49
50	if (!xfrm6_dst_lookup((struct xfrm_dst **)&rt, &fl_tunnel)) {
51		ipv6_get_saddr(&rt->u.dst, (struct in6_addr *)&daddr->a6,
52			       (struct in6_addr *)&saddr->a6);
53		dst_release(&rt->u.dst);
54		return 0;
55	}
56	return -EHOSTUNREACH;
57}
58
59static struct dst_entry *
60__xfrm6_find_bundle(struct flowi *fl, struct xfrm_policy *policy)
61{
62	struct dst_entry *dst;
63
64	/* Still not clear if we should set fl->fl6_{src,dst}... */
65	read_lock_bh(&policy->lock);
66	for (dst = policy->bundles; dst; dst = dst->next) {
67		struct xfrm_dst *xdst = (struct xfrm_dst*)dst;
68		struct in6_addr fl_dst_prefix, fl_src_prefix;
69
70		ipv6_addr_prefix(&fl_dst_prefix,
71				 &fl->fl6_dst,
72				 xdst->u.rt6.rt6i_dst.plen);
73		ipv6_addr_prefix(&fl_src_prefix,
74				 &fl->fl6_src,
75				 xdst->u.rt6.rt6i_src.plen);
76		if (ipv6_addr_equal(&xdst->u.rt6.rt6i_dst.addr, &fl_dst_prefix) &&
77		    ipv6_addr_equal(&xdst->u.rt6.rt6i_src.addr, &fl_src_prefix) &&
78		    xfrm_bundle_ok(policy, xdst, fl, AF_INET6,
79				   (xdst->u.rt6.rt6i_dst.plen != 128 ||
80				    xdst->u.rt6.rt6i_src.plen != 128))) {
81			dst_clone(dst);
82			break;
83		}
84	}
85	read_unlock_bh(&policy->lock);
86	return dst;
87}
88
89static inline struct in6_addr*
90__xfrm6_bundle_addr_remote(struct xfrm_state *x, struct in6_addr *addr)
91{
92	return (x->type->remote_addr) ?
93		(struct in6_addr*)x->type->remote_addr(x, (xfrm_address_t *)addr) :
94		(struct in6_addr*)&x->id.daddr;
95}
96
97static inline struct in6_addr*
98__xfrm6_bundle_addr_local(struct xfrm_state *x, struct in6_addr *addr)
99{
100	return (x->type->local_addr) ?
101		(struct in6_addr*)x->type->local_addr(x, (xfrm_address_t *)addr) :
102		(struct in6_addr*)&x->props.saddr;
103}
104
105static inline void
106__xfrm6_bundle_len_inc(int *len, int *nflen, struct xfrm_state *x)
107{
108	if (x->type->flags & XFRM_TYPE_NON_FRAGMENT)
109		*nflen += x->props.header_len;
110	else
111		*len += x->props.header_len;
112}
113
114static inline void
115__xfrm6_bundle_len_dec(int *len, int *nflen, struct xfrm_state *x)
116{
117	if (x->type->flags & XFRM_TYPE_NON_FRAGMENT)
118		*nflen -= x->props.header_len;
119	else
120		*len -= x->props.header_len;
121}
122
123/* Allocate chain of dst_entry's, attach known xfrm's, calculate
124 * all the metrics... Shortly, bundle a bundle.
125 */
126
127static int
128__xfrm6_bundle_create(struct xfrm_policy *policy, struct xfrm_state **xfrm, int nx,
129		      struct flowi *fl, struct dst_entry **dst_p)
130{
131	struct dst_entry *dst, *dst_prev;
132	struct rt6_info *rt0 = (struct rt6_info*)(*dst_p);
133	struct rt6_info *rt  = rt0;
134	struct flowi fl_tunnel = {
135		.nl_u = {
136			.ip6_u = {
137				.saddr = fl->fl6_src,
138				.daddr = fl->fl6_dst,
139			}
140		}
141	};
142	int i;
143	int err = 0;
144	int header_len = 0;
145	int nfheader_len = 0;
146	int trailer_len = 0;
147
148	dst = dst_prev = NULL;
149	dst_hold(&rt->u.dst);
150
151	for (i = 0; i < nx; i++) {
152		struct dst_entry *dst1 = dst_alloc(&xfrm6_dst_ops);
153		struct xfrm_dst *xdst;
154
155		if (unlikely(dst1 == NULL)) {
156			err = -ENOBUFS;
157			dst_release(&rt->u.dst);
158			goto error;
159		}
160
161		if (!dst)
162			dst = dst1;
163		else {
164			dst_prev->child = dst1;
165			dst1->flags |= DST_NOHASH;
166			dst_clone(dst1);
167		}
168
169		xdst = (struct xfrm_dst *)dst1;
170		xdst->route = &rt->u.dst;
171		xdst->genid = xfrm[i]->genid;
172		if (rt->rt6i_node)
173			xdst->route_cookie = rt->rt6i_node->fn_sernum;
174
175		dst1->next = dst_prev;
176		dst_prev = dst1;
177
178		__xfrm6_bundle_len_inc(&header_len, &nfheader_len, xfrm[i]);
179		trailer_len += xfrm[i]->props.trailer_len;
180
181		if (xfrm[i]->props.mode == XFRM_MODE_TUNNEL) {
182			unsigned short encap_family = xfrm[i]->props.family;
183			switch(encap_family) {
184			case AF_INET:
185				fl_tunnel.fl4_dst = xfrm[i]->id.daddr.a4;
186				fl_tunnel.fl4_src = xfrm[i]->props.saddr.a4;
187				break;
188			case AF_INET6:
189				ipv6_addr_copy(&fl_tunnel.fl6_dst, (struct in6_addr*)&xfrm[i]->id.daddr.a6);
190				ipv6_addr_copy(&fl_tunnel.fl6_src, (struct in6_addr*)&xfrm[i]->props.saddr.a6);
191				break;
192			default:
193				BUG_ON(1);
194			}
195
196			err = xfrm_dst_lookup((struct xfrm_dst **) &rt,
197					      &fl_tunnel, encap_family);
198			if (err)
199				goto error;
200		} else
201			dst_hold(&rt->u.dst);
202	}
203
204	dst_prev->child = &rt->u.dst;
205	dst->path = &rt->u.dst;
206	if (rt->rt6i_node)
207		((struct xfrm_dst *)dst)->path_cookie = rt->rt6i_node->fn_sernum;
208
209	*dst_p = dst;
210	dst = dst_prev;
211
212	dst_prev = *dst_p;
213	i = 0;
214	for (; dst_prev != &rt->u.dst; dst_prev = dst_prev->child) {
215		struct xfrm_dst *x = (struct xfrm_dst*)dst_prev;
216		struct xfrm_state_afinfo *afinfo;
217
218		dst_prev->xfrm = xfrm[i++];
219		dst_prev->dev = rt->u.dst.dev;
220		if (rt->u.dst.dev)
221			dev_hold(rt->u.dst.dev);
222		dst_prev->obsolete	= -1;
223		dst_prev->flags	       |= DST_HOST;
224		dst_prev->lastuse	= jiffies;
225		dst_prev->header_len	= header_len;
226		dst_prev->nfheader_len	= nfheader_len;
227		dst_prev->trailer_len	= trailer_len;
228		memcpy(&dst_prev->metrics, &x->route->metrics, sizeof(dst_prev->metrics));
229
230		/* Copy neighbour for reachability confirmation */
231		dst_prev->neighbour	= neigh_clone(rt->u.dst.neighbour);
232		dst_prev->input		= rt->u.dst.input;
233		/* XXX: When IPv4 is implemented as module and can be unloaded,
234		 * we should manage reference to xfrm4_output in afinfo->output.
235		 * Miyazawa
236		 */
237		afinfo = xfrm_state_get_afinfo(dst_prev->xfrm->props.family);
238		if (!afinfo) {
239			dst = *dst_p;
240			goto error;
241		};
242		dst_prev->output = afinfo->output;
243		xfrm_state_put_afinfo(afinfo);
244		/* Sheit... I remember I did this right. Apparently,
245		 * it was magically lost, so this code needs audit */
246		x->u.rt6.rt6i_flags    = rt0->rt6i_flags&(RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL);
247		x->u.rt6.rt6i_metric   = rt0->rt6i_metric;
248		x->u.rt6.rt6i_node     = rt0->rt6i_node;
249		x->u.rt6.rt6i_gateway  = rt0->rt6i_gateway;
250		memcpy(&x->u.rt6.rt6i_gateway, &rt0->rt6i_gateway, sizeof(x->u.rt6.rt6i_gateway));
251		x->u.rt6.rt6i_dst      = rt0->rt6i_dst;
252		x->u.rt6.rt6i_src      = rt0->rt6i_src;
253		x->u.rt6.rt6i_idev     = rt0->rt6i_idev;
254		in6_dev_hold(rt0->rt6i_idev);
255		__xfrm6_bundle_len_dec(&header_len, &nfheader_len, x->u.dst.xfrm);
256		trailer_len -= x->u.dst.xfrm->props.trailer_len;
257	}
258
259	xfrm_init_pmtu(dst);
260	return 0;
261
262error:
263	if (dst)
264		dst_free(dst);
265	return err;
266}
267
268static inline void
269_decode_session6(struct sk_buff *skb, struct flowi *fl)
270{
271	u16 offset = skb->h.raw - skb->nh.raw;
272	struct ipv6hdr *hdr = skb->nh.ipv6h;
273	struct ipv6_opt_hdr *exthdr;
274	u8 nexthdr = skb->nh.raw[IP6CB(skb)->nhoff];
275
276	memset(fl, 0, sizeof(struct flowi));
277	ipv6_addr_copy(&fl->fl6_dst, &hdr->daddr);
278	ipv6_addr_copy(&fl->fl6_src, &hdr->saddr);
279
280	while (pskb_may_pull(skb, skb->nh.raw + offset + 1 - skb->data)) {
281		exthdr = (struct ipv6_opt_hdr*)(skb->nh.raw + offset);
282
283		switch (nexthdr) {
284		case NEXTHDR_ROUTING:
285		case NEXTHDR_HOP:
286		case NEXTHDR_DEST:
287			offset += ipv6_optlen(exthdr);
288			nexthdr = exthdr->nexthdr;
289			exthdr = (struct ipv6_opt_hdr*)(skb->nh.raw + offset);
290			break;
291
292		case IPPROTO_UDP:
293		case IPPROTO_UDPLITE:
294		case IPPROTO_TCP:
295		case IPPROTO_SCTP:
296		case IPPROTO_DCCP:
297			if (pskb_may_pull(skb, skb->nh.raw + offset + 4 - skb->data)) {
298				__be16 *ports = (__be16 *)exthdr;
299
300				fl->fl_ip_sport = ports[0];
301				fl->fl_ip_dport = ports[1];
302			}
303			fl->proto = nexthdr;
304			return;
305
306		case IPPROTO_ICMPV6:
307			if (pskb_may_pull(skb, skb->nh.raw + offset + 2 - skb->data)) {
308				u8 *icmp = (u8 *)exthdr;
309
310				fl->fl_icmp_type = icmp[0];
311				fl->fl_icmp_code = icmp[1];
312			}
313			fl->proto = nexthdr;
314			return;
315
316#ifdef CONFIG_IPV6_MIP6
317		case IPPROTO_MH:
318			if (pskb_may_pull(skb, skb->nh.raw + offset + 3 - skb->data)) {
319				struct ip6_mh *mh;
320				mh = (struct ip6_mh *)exthdr;
321
322				fl->fl_mh_type = mh->ip6mh_type;
323			}
324			fl->proto = nexthdr;
325			return;
326#endif
327
328		/* XXX Why are there these headers? */
329		case IPPROTO_AH:
330		case IPPROTO_ESP:
331		case IPPROTO_COMP:
332		default:
333			fl->fl_ipsec_spi = 0;
334			fl->proto = nexthdr;
335			return;
336		};
337	}
338}
339
340static inline int xfrm6_garbage_collect(void)
341{
342	xfrm6_policy_afinfo.garbage_collect();
343	return (atomic_read(&xfrm6_dst_ops.entries) > xfrm6_dst_ops.gc_thresh*2);
344}
345
346static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu)
347{
348	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
349	struct dst_entry *path = xdst->route;
350
351	path->ops->update_pmtu(path, mtu);
352}
353
354static void xfrm6_dst_destroy(struct dst_entry *dst)
355{
356	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
357
358	if (likely(xdst->u.rt6.rt6i_idev))
359		in6_dev_put(xdst->u.rt6.rt6i_idev);
360	xfrm_dst_destroy(xdst);
361}
362
363static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
364			     int unregister)
365{
366	struct xfrm_dst *xdst;
367
368	if (!unregister)
369		return;
370
371	xdst = (struct xfrm_dst *)dst;
372	if (xdst->u.rt6.rt6i_idev->dev == dev) {
373		struct inet6_dev *loopback_idev = in6_dev_get(&loopback_dev);
374		BUG_ON(!loopback_idev);
375
376		do {
377			in6_dev_put(xdst->u.rt6.rt6i_idev);
378			xdst->u.rt6.rt6i_idev = loopback_idev;
379			in6_dev_hold(loopback_idev);
380			xdst = (struct xfrm_dst *)xdst->u.dst.child;
381		} while (xdst->u.dst.xfrm);
382
383		__in6_dev_put(loopback_idev);
384	}
385
386	xfrm_dst_ifdown(dst, dev);
387}
388
389static struct dst_ops xfrm6_dst_ops = {
390	.family =		AF_INET6,
391	.protocol =		__constant_htons(ETH_P_IPV6),
392	.gc =			xfrm6_garbage_collect,
393	.update_pmtu =		xfrm6_update_pmtu,
394	.destroy =		xfrm6_dst_destroy,
395	.ifdown =		xfrm6_dst_ifdown,
396	.gc_thresh =		1024,
397	.entry_size =		sizeof(struct xfrm_dst),
398};
399
400static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
401	.family =		AF_INET6,
402	.dst_ops =		&xfrm6_dst_ops,
403	.dst_lookup =		xfrm6_dst_lookup,
404	.get_saddr = 		xfrm6_get_saddr,
405	.find_bundle =		__xfrm6_find_bundle,
406	.bundle_create =	__xfrm6_bundle_create,
407	.decode_session =	_decode_session6,
408};
409
410static void __init xfrm6_policy_init(void)
411{
412	xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
413}
414
415static void xfrm6_policy_fini(void)
416{
417	xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
418}
419
420void __init xfrm6_init(void)
421{
422	xfrm6_policy_init();
423	xfrm6_state_init();
424}
425
426void xfrm6_fini(void)
427{
428	//xfrm6_input_fini();
429	xfrm6_policy_fini();
430	xfrm6_state_fini();
431}
432