1/*
2 * xfrm4_policy.c
3 *
4 * Changes:
5 *	Kazunori MIYAZAWA @USAGI
6 * 	YOSHIFUJI Hideaki @USAGI
7 *		Split up af-specific portion
8 *
9 */
10
11#include <linux/err.h>
12#include <linux/kernel.h>
13#include <linux/inetdevice.h>
14#include <linux/if_tunnel.h>
15#include <net/dst.h>
16#include <net/xfrm.h>
17#include <net/ip.h>
18
19static struct xfrm_policy_afinfo xfrm4_policy_afinfo;
20
21static struct dst_entry *__xfrm4_dst_lookup(struct net *net, struct flowi4 *fl4,
22					    int tos,
23					    const xfrm_address_t *saddr,
24					    const xfrm_address_t *daddr)
25{
26	struct rtable *rt;
27
28	memset(fl4, 0, sizeof(*fl4));
29	fl4->daddr = daddr->a4;
30	fl4->flowi4_tos = tos;
31	if (saddr)
32		fl4->saddr = saddr->a4;
33
34	rt = __ip_route_output_key(net, fl4);
35	if (!IS_ERR(rt))
36		return &rt->dst;
37
38	return ERR_CAST(rt);
39}
40
41static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos,
42					  const xfrm_address_t *saddr,
43					  const xfrm_address_t *daddr)
44{
45	struct flowi4 fl4;
46
47	return __xfrm4_dst_lookup(net, &fl4, tos, saddr, daddr);
48}
49
50static int xfrm4_get_saddr(struct net *net,
51			   xfrm_address_t *saddr, xfrm_address_t *daddr)
52{
53	struct dst_entry *dst;
54	struct flowi4 fl4;
55
56	dst = __xfrm4_dst_lookup(net, &fl4, 0, NULL, daddr);
57	if (IS_ERR(dst))
58		return -EHOSTUNREACH;
59
60	saddr->a4 = fl4.saddr;
61	dst_release(dst);
62	return 0;
63}
64
65static int xfrm4_get_tos(const struct flowi *fl)
66{
67	return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */
68}
69
70static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
71			   int nfheader_len)
72{
73	return 0;
74}
75
76static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
77			  const struct flowi *fl)
78{
79	struct rtable *rt = (struct rtable *)xdst->route;
80	const struct flowi4 *fl4 = &fl->u.ip4;
81
82	xdst->u.rt.rt_iif = fl4->flowi4_iif;
83
84	xdst->u.dst.dev = dev;
85	dev_hold(dev);
86
87	/* Sheit... I remember I did this right. Apparently,
88	 * it was magically lost, so this code needs audit */
89	xdst->u.rt.rt_is_input = rt->rt_is_input;
90	xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
91					      RTCF_LOCAL);
92	xdst->u.rt.rt_type = rt->rt_type;
93	xdst->u.rt.rt_gateway = rt->rt_gateway;
94	xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway;
95	xdst->u.rt.rt_pmtu = rt->rt_pmtu;
96	INIT_LIST_HEAD(&xdst->u.rt.rt_uncached);
97
98	return 0;
99}
100
101static void
102_decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
103{
104	const struct iphdr *iph = ip_hdr(skb);
105	u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
106	struct flowi4 *fl4 = &fl->u.ip4;
107	int oif = 0;
108
109	if (skb_dst(skb))
110		oif = skb_dst(skb)->dev->ifindex;
111
112	memset(fl4, 0, sizeof(struct flowi4));
113	fl4->flowi4_mark = skb->mark;
114	fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
115
116	if (!ip_is_fragment(iph)) {
117		switch (iph->protocol) {
118		case IPPROTO_UDP:
119		case IPPROTO_UDPLITE:
120		case IPPROTO_TCP:
121		case IPPROTO_SCTP:
122		case IPPROTO_DCCP:
123			if (xprth + 4 < skb->data ||
124			    pskb_may_pull(skb, xprth + 4 - skb->data)) {
125				__be16 *ports = (__be16 *)xprth;
126
127				fl4->fl4_sport = ports[!!reverse];
128				fl4->fl4_dport = ports[!reverse];
129			}
130			break;
131
132		case IPPROTO_ICMP:
133			if (pskb_may_pull(skb, xprth + 2 - skb->data)) {
134				u8 *icmp = xprth;
135
136				fl4->fl4_icmp_type = icmp[0];
137				fl4->fl4_icmp_code = icmp[1];
138			}
139			break;
140
141		case IPPROTO_ESP:
142			if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
143				__be32 *ehdr = (__be32 *)xprth;
144
145				fl4->fl4_ipsec_spi = ehdr[0];
146			}
147			break;
148
149		case IPPROTO_AH:
150			if (pskb_may_pull(skb, xprth + 8 - skb->data)) {
151				__be32 *ah_hdr = (__be32 *)xprth;
152
153				fl4->fl4_ipsec_spi = ah_hdr[1];
154			}
155			break;
156
157		case IPPROTO_COMP:
158			if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
159				__be16 *ipcomp_hdr = (__be16 *)xprth;
160
161				fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
162			}
163			break;
164
165		case IPPROTO_GRE:
166			if (pskb_may_pull(skb, xprth + 12 - skb->data)) {
167				__be16 *greflags = (__be16 *)xprth;
168				__be32 *gre_hdr = (__be32 *)xprth;
169
170				if (greflags[0] & GRE_KEY) {
171					if (greflags[0] & GRE_CSUM)
172						gre_hdr++;
173					fl4->fl4_gre_key = gre_hdr[1];
174				}
175			}
176			break;
177
178		default:
179			fl4->fl4_ipsec_spi = 0;
180			break;
181		}
182	}
183	fl4->flowi4_proto = iph->protocol;
184	fl4->daddr = reverse ? iph->saddr : iph->daddr;
185	fl4->saddr = reverse ? iph->daddr : iph->saddr;
186	fl4->flowi4_tos = iph->tos;
187}
188
189static inline int xfrm4_garbage_collect(struct dst_ops *ops)
190{
191	struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
192
193	xfrm4_policy_afinfo.garbage_collect(net);
194	return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
195}
196
197static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
198			      struct sk_buff *skb, u32 mtu)
199{
200	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
201	struct dst_entry *path = xdst->route;
202
203	path->ops->update_pmtu(path, sk, skb, mtu);
204}
205
206static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
207			   struct sk_buff *skb)
208{
209	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
210	struct dst_entry *path = xdst->route;
211
212	path->ops->redirect(path, sk, skb);
213}
214
215static void xfrm4_dst_destroy(struct dst_entry *dst)
216{
217	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
218
219	dst_destroy_metrics_generic(dst);
220
221	xfrm_dst_destroy(xdst);
222}
223
224static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
225			     int unregister)
226{
227	if (!unregister)
228		return;
229
230	xfrm_dst_ifdown(dst, dev);
231}
232
233static struct dst_ops xfrm4_dst_ops_template = {
234	.family =		AF_INET,
235	.gc =			xfrm4_garbage_collect,
236	.update_pmtu =		xfrm4_update_pmtu,
237	.redirect =		xfrm4_redirect,
238	.cow_metrics =		dst_cow_metrics_generic,
239	.destroy =		xfrm4_dst_destroy,
240	.ifdown =		xfrm4_dst_ifdown,
241	.local_out =		__ip_local_out,
242	.gc_thresh =		32768,
243};
244
245static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
246	.family = 		AF_INET,
247	.dst_ops =		&xfrm4_dst_ops_template,
248	.dst_lookup =		xfrm4_dst_lookup,
249	.get_saddr =		xfrm4_get_saddr,
250	.decode_session =	_decode_session4,
251	.get_tos =		xfrm4_get_tos,
252	.init_path =		xfrm4_init_path,
253	.fill_dst =		xfrm4_fill_dst,
254	.blackhole_route =	ipv4_blackhole_route,
255};
256
257#ifdef CONFIG_SYSCTL
258static struct ctl_table xfrm4_policy_table[] = {
259	{
260		.procname       = "xfrm4_gc_thresh",
261		.data           = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
262		.maxlen         = sizeof(int),
263		.mode           = 0644,
264		.proc_handler   = proc_dointvec,
265	},
266	{ }
267};
268
269static int __net_init xfrm4_net_sysctl_init(struct net *net)
270{
271	struct ctl_table *table;
272	struct ctl_table_header *hdr;
273
274	table = xfrm4_policy_table;
275	if (!net_eq(net, &init_net)) {
276		table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
277		if (!table)
278			goto err_alloc;
279
280		table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
281	}
282
283	hdr = register_net_sysctl(net, "net/ipv4", table);
284	if (!hdr)
285		goto err_reg;
286
287	net->ipv4.xfrm4_hdr = hdr;
288	return 0;
289
290err_reg:
291	if (!net_eq(net, &init_net))
292		kfree(table);
293err_alloc:
294	return -ENOMEM;
295}
296
297static void __net_exit xfrm4_net_sysctl_exit(struct net *net)
298{
299	struct ctl_table *table;
300
301	if (!net->ipv4.xfrm4_hdr)
302		return;
303
304	table = net->ipv4.xfrm4_hdr->ctl_table_arg;
305	unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
306	if (!net_eq(net, &init_net))
307		kfree(table);
308}
309#else /* CONFIG_SYSCTL */
310static int inline xfrm4_net_sysctl_init(struct net *net)
311{
312	return 0;
313}
314
315static void inline xfrm4_net_sysctl_exit(struct net *net)
316{
317}
318#endif
319
320static int __net_init xfrm4_net_init(struct net *net)
321{
322	int ret;
323
324	memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template,
325	       sizeof(xfrm4_dst_ops_template));
326	ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops);
327	if (ret)
328		return ret;
329
330	ret = xfrm4_net_sysctl_init(net);
331	if (ret)
332		dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
333
334	return ret;
335}
336
337static void __net_exit xfrm4_net_exit(struct net *net)
338{
339	xfrm4_net_sysctl_exit(net);
340	dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
341}
342
343static struct pernet_operations __net_initdata xfrm4_net_ops = {
344	.init	= xfrm4_net_init,
345	.exit	= xfrm4_net_exit,
346};
347
348static void __init xfrm4_policy_init(void)
349{
350	xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
351}
352
353void __init xfrm4_init(void)
354{
355	xfrm4_state_init();
356	xfrm4_policy_init();
357	xfrm4_protocol_init();
358	register_pernet_subsys(&xfrm4_net_ops);
359}
360
361