xfrm6_policy.c revision 44b451f1633896de15d2d52e1a2bd462e80b7814 
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/err.h>
15#include <linux/kernel.h>
16#include <linux/netdevice.h>
17#include <net/addrconf.h>
18#include <net/dst.h>
19#include <net/xfrm.h>
20#include <net/ip.h>
21#include <net/ipv6.h>
22#include <net/ip6_route.h>
23#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
24#include <net/mip6.h>
25#endif
26
27static struct xfrm_policy_afinfo xfrm6_policy_afinfo;
28
29static struct dst_entry *xfrm6_dst_lookup(struct net *net, int tos,
30					  xfrm_address_t *saddr,
31					  xfrm_address_t *daddr)
32{
33	struct flowi fl = {};
34	struct dst_entry *dst;
35	int err;
36
37	memcpy(&fl.fl6_dst, daddr, sizeof(fl.fl6_dst));
38	if (saddr)
39		memcpy(&fl.fl6_src, saddr, sizeof(fl.fl6_src));
40
41	dst = ip6_route_output(net, NULL, &fl);
42
43	err = dst->error;
44	if (dst->error) {
45		dst_release(dst);
46		dst = ERR_PTR(err);
47	}
48
49	return dst;
50}
51
52static int xfrm6_get_saddr(struct net *net,
53			   xfrm_address_t *saddr, xfrm_address_t *daddr)
54{
55	struct dst_entry *dst;
56	struct net_device *dev;
57
58	dst = xfrm6_dst_lookup(net, 0, NULL, daddr);
59	if (IS_ERR(dst))
60		return -EHOSTUNREACH;
61
62	dev = ip6_dst_idev(dst)->dev;
63	ipv6_dev_get_saddr(dev_net(dev), dev,
64			   (struct in6_addr *)&daddr->a6, 0,
65			   (struct in6_addr *)&saddr->a6);
66	dst_release(dst);
67	return 0;
68}
69
70static int xfrm6_get_tos(struct flowi *fl)
71{
72	return 0;
73}
74
75static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
76			   int nfheader_len)
77{
78	if (dst->ops->family == AF_INET6) {
79		struct rt6_info *rt = (struct rt6_info*)dst;
80		if (rt->rt6i_node)
81			path->path_cookie = rt->rt6i_node->fn_sernum;
82	}
83
84	path->u.rt6.rt6i_nfheader_len = nfheader_len;
85
86	return 0;
87}
88
89static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
90			  struct flowi *fl)
91{
92	struct rt6_info *rt = (struct rt6_info*)xdst->route;
93
94	xdst->u.dst.dev = dev;
95	dev_hold(dev);
96
97	xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
98	if (!xdst->u.rt6.rt6i_idev)
99		return -ENODEV;
100
101	/* Sheit... I remember I did this right. Apparently,
102	 * it was magically lost, so this code needs audit */
103	xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
104						   RTF_LOCAL);
105	xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
106	xdst->u.rt6.rt6i_node = rt->rt6i_node;
107	if (rt->rt6i_node)
108		xdst->route_cookie = rt->rt6i_node->fn_sernum;
109	xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
110	xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
111	xdst->u.rt6.rt6i_src = rt->rt6i_src;
112
113	return 0;
114}
115
116static inline void
117_decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
118{
119	int onlyproto = 0;
120	u16 offset = skb_network_header_len(skb);
121	struct ipv6hdr *hdr = ipv6_hdr(skb);
122	struct ipv6_opt_hdr *exthdr;
123	const unsigned char *nh = skb_network_header(skb);
124	u8 nexthdr = nh[IP6CB(skb)->nhoff];
125
126	memset(fl, 0, sizeof(struct flowi));
127	fl->mark = skb->mark;
128
129	ipv6_addr_copy(&fl->fl6_dst, reverse ? &hdr->saddr : &hdr->daddr);
130	ipv6_addr_copy(&fl->fl6_src, reverse ? &hdr->daddr : &hdr->saddr);
131
132	while (nh + offset + 1 < skb->data ||
133	       pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
134		nh = skb_network_header(skb);
135		exthdr = (struct ipv6_opt_hdr *)(nh + offset);
136
137		switch (nexthdr) {
138		case NEXTHDR_FRAGMENT:
139			onlyproto = 1;
140		case NEXTHDR_ROUTING:
141		case NEXTHDR_HOP:
142		case NEXTHDR_DEST:
143			offset += ipv6_optlen(exthdr);
144			nexthdr = exthdr->nexthdr;
145			exthdr = (struct ipv6_opt_hdr *)(nh + offset);
146			break;
147
148		case IPPROTO_UDP:
149		case IPPROTO_UDPLITE:
150		case IPPROTO_TCP:
151		case IPPROTO_SCTP:
152		case IPPROTO_DCCP:
153			if (!onlyproto && (nh + offset + 4 < skb->data ||
154			     pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
155				__be16 *ports = (__be16 *)exthdr;
156
157				fl->fl_ip_sport = ports[!!reverse];
158				fl->fl_ip_dport = ports[!reverse];
159			}
160			fl->proto = nexthdr;
161			return;
162
163		case IPPROTO_ICMPV6:
164			if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
165				u8 *icmp = (u8 *)exthdr;
166
167				fl->fl_icmp_type = icmp[0];
168				fl->fl_icmp_code = icmp[1];
169			}
170			fl->proto = nexthdr;
171			return;
172
173#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
174		case IPPROTO_MH:
175			if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
176				struct ip6_mh *mh;
177				mh = (struct ip6_mh *)exthdr;
178
179				fl->fl_mh_type = mh->ip6mh_type;
180			}
181			fl->proto = nexthdr;
182			return;
183#endif
184
185		/* XXX Why are there these headers? */
186		case IPPROTO_AH:
187		case IPPROTO_ESP:
188		case IPPROTO_COMP:
189		default:
190			fl->fl_ipsec_spi = 0;
191			fl->proto = nexthdr;
192			return;
193		}
194	}
195}
196
197static inline int xfrm6_garbage_collect(struct dst_ops *ops)
198{
199	struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
200
201	xfrm6_policy_afinfo.garbage_collect(net);
202	return (atomic_read(&ops->entries) > ops->gc_thresh * 2);
203}
204
205static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu)
206{
207	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
208	struct dst_entry *path = xdst->route;
209
210	path->ops->update_pmtu(path, mtu);
211}
212
213static void xfrm6_dst_destroy(struct dst_entry *dst)
214{
215	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
216
217	if (likely(xdst->u.rt6.rt6i_idev))
218		in6_dev_put(xdst->u.rt6.rt6i_idev);
219	xfrm_dst_destroy(xdst);
220}
221
222static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
223			     int unregister)
224{
225	struct xfrm_dst *xdst;
226
227	if (!unregister)
228		return;
229
230	xdst = (struct xfrm_dst *)dst;
231	if (xdst->u.rt6.rt6i_idev->dev == dev) {
232		struct inet6_dev *loopback_idev =
233			in6_dev_get(dev_net(dev)->loopback_dev);
234		BUG_ON(!loopback_idev);
235
236		do {
237			in6_dev_put(xdst->u.rt6.rt6i_idev);
238			xdst->u.rt6.rt6i_idev = loopback_idev;
239			in6_dev_hold(loopback_idev);
240			xdst = (struct xfrm_dst *)xdst->u.dst.child;
241		} while (xdst->u.dst.xfrm);
242
243		__in6_dev_put(loopback_idev);
244	}
245
246	xfrm_dst_ifdown(dst, dev);
247}
248
249static struct dst_ops xfrm6_dst_ops = {
250	.family =		AF_INET6,
251	.protocol =		cpu_to_be16(ETH_P_IPV6),
252	.gc =			xfrm6_garbage_collect,
253	.update_pmtu =		xfrm6_update_pmtu,
254	.destroy =		xfrm6_dst_destroy,
255	.ifdown =		xfrm6_dst_ifdown,
256	.local_out =		__ip6_local_out,
257	.gc_thresh =		1024,
258	.entries =		ATOMIC_INIT(0),
259};
260
261static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
262	.family =		AF_INET6,
263	.dst_ops =		&xfrm6_dst_ops,
264	.dst_lookup =		xfrm6_dst_lookup,
265	.get_saddr = 		xfrm6_get_saddr,
266	.decode_session =	_decode_session6,
267	.get_tos =		xfrm6_get_tos,
268	.init_path =		xfrm6_init_path,
269	.fill_dst =		xfrm6_fill_dst,
270};
271
272static int __init xfrm6_policy_init(void)
273{
274	return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
275}
276
277static void xfrm6_policy_fini(void)
278{
279	xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
280}
281
282#ifdef CONFIG_SYSCTL
283static struct ctl_table xfrm6_policy_table[] = {
284	{
285		.procname       = "xfrm6_gc_thresh",
286		.data	   	= &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
287		.maxlen	 	= sizeof(int),
288		.mode	   	= 0644,
289		.proc_handler   = proc_dointvec,
290	},
291	{ }
292};
293
294static struct ctl_table_header *sysctl_hdr;
295#endif
296
297int __init xfrm6_init(void)
298{
299	int ret;
300	unsigned int gc_thresh;
301
302	/*
303	 * We need a good default value for the xfrm6 gc threshold.
304	 * In ipv4 we set it to the route hash table size * 8, which
305	 * is half the size of the maximaum route cache for ipv4.  It
306	 * would be good to do the same thing for v6, except the table is
307	 * constructed differently here.  Here each table for a net namespace
308	 * can have FIB_TABLE_HASHSZ entries, so lets go with the same
309	 * computation that we used for ipv4 here.  Also, lets keep the initial
310	 * gc_thresh to a minimum of 1024, since, the ipv6 route cache defaults
311	 * to that as a minimum as well
312	 */
313	gc_thresh = FIB6_TABLE_HASHSZ * 8;
314	xfrm6_dst_ops.gc_thresh = (gc_thresh < 1024) ? 1024 : gc_thresh;
315
316	ret = xfrm6_policy_init();
317	if (ret)
318		goto out;
319
320	ret = xfrm6_state_init();
321	if (ret)
322		goto out_policy;
323
324#ifdef CONFIG_SYSCTL
325	sysctl_hdr = register_net_sysctl_table(&init_net, net_ipv6_ctl_path,
326						xfrm6_policy_table);
327#endif
328out:
329	return ret;
330out_policy:
331	xfrm6_policy_fini();
332	goto out;
333}
334
335void xfrm6_fini(void)
336{
337#ifdef CONFIG_SYSCTL
338	if (sysctl_hdr)
339		unregister_net_sysctl_table(sysctl_hdr);
340#endif
341	//xfrm6_input_fini();
342	xfrm6_policy_fini();
343	xfrm6_state_fini();
344}
345