xref: /linux/net/ipv6/xfrm6_policy.c (revision f3a8b6645dc2e60d11f20c1c23afd964ff4e55ae)
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 #include <net/l3mdev.h>
24 #if IS_ENABLED(CONFIG_IPV6_MIP6)
25 #include <net/mip6.h>
26 #endif
27 
28 static struct xfrm_policy_afinfo xfrm6_policy_afinfo;
29 
30 static struct dst_entry *xfrm6_dst_lookup(struct net *net, int tos, int oif,
31 					  const xfrm_address_t *saddr,
32 					  const xfrm_address_t *daddr)
33 {
34 	struct flowi6 fl6;
35 	struct dst_entry *dst;
36 	int err;
37 
38 	memset(&fl6, 0, sizeof(fl6));
39 	fl6.flowi6_oif = l3mdev_master_ifindex_by_index(net, oif);
40 	fl6.flowi6_flags = FLOWI_FLAG_SKIP_NH_OIF;
41 	memcpy(&fl6.daddr, daddr, sizeof(fl6.daddr));
42 	if (saddr)
43 		memcpy(&fl6.saddr, saddr, sizeof(fl6.saddr));
44 
45 	dst = ip6_route_output(net, NULL, &fl6);
46 
47 	err = dst->error;
48 	if (dst->error) {
49 		dst_release(dst);
50 		dst = ERR_PTR(err);
51 	}
52 
53 	return dst;
54 }
55 
56 static int xfrm6_get_saddr(struct net *net, int oif,
57 			   xfrm_address_t *saddr, xfrm_address_t *daddr)
58 {
59 	struct dst_entry *dst;
60 	struct net_device *dev;
61 
62 	dst = xfrm6_dst_lookup(net, 0, oif, NULL, daddr);
63 	if (IS_ERR(dst))
64 		return -EHOSTUNREACH;
65 
66 	dev = ip6_dst_idev(dst)->dev;
67 	ipv6_dev_get_saddr(dev_net(dev), dev, &daddr->in6, 0, &saddr->in6);
68 	dst_release(dst);
69 	return 0;
70 }
71 
72 static int xfrm6_get_tos(const struct flowi *fl)
73 {
74 	return 0;
75 }
76 
77 static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
78 			   int nfheader_len)
79 {
80 	if (dst->ops->family == AF_INET6) {
81 		struct rt6_info *rt = (struct rt6_info *)dst;
82 		path->path_cookie = rt6_get_cookie(rt);
83 	}
84 
85 	path->u.rt6.rt6i_nfheader_len = nfheader_len;
86 
87 	return 0;
88 }
89 
90 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
91 			  const struct flowi *fl)
92 {
93 	struct rt6_info *rt = (struct rt6_info *)xdst->route;
94 
95 	xdst->u.dst.dev = dev;
96 	dev_hold(dev);
97 
98 	xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
99 	if (!xdst->u.rt6.rt6i_idev) {
100 		dev_put(dev);
101 		return -ENODEV;
102 	}
103 
104 	/* Sheit... I remember I did this right. Apparently,
105 	 * it was magically lost, so this code needs audit */
106 	xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
107 						   RTF_LOCAL);
108 	xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
109 	xdst->u.rt6.rt6i_node = rt->rt6i_node;
110 	xdst->route_cookie = rt6_get_cookie(rt);
111 	xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
112 	xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
113 	xdst->u.rt6.rt6i_src = rt->rt6i_src;
114 
115 	return 0;
116 }
117 
118 static inline void
119 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
120 {
121 	struct flowi6 *fl6 = &fl->u.ip6;
122 	int onlyproto = 0;
123 	const struct ipv6hdr *hdr = ipv6_hdr(skb);
124 	u16 offset = sizeof(*hdr);
125 	struct ipv6_opt_hdr *exthdr;
126 	const unsigned char *nh = skb_network_header(skb);
127 	u16 nhoff = IP6CB(skb)->nhoff;
128 	int oif = 0;
129 	u8 nexthdr;
130 
131 	if (!nhoff)
132 		nhoff = offsetof(struct ipv6hdr, nexthdr);
133 
134 	nexthdr = nh[nhoff];
135 
136 	if (skb_dst(skb))
137 		oif = skb_dst(skb)->dev->ifindex;
138 
139 	memset(fl6, 0, sizeof(struct flowi6));
140 	fl6->flowi6_mark = skb->mark;
141 	fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
142 
143 	fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
144 	fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
145 
146 	while (nh + offset + 1 < skb->data ||
147 	       pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
148 		nh = skb_network_header(skb);
149 		exthdr = (struct ipv6_opt_hdr *)(nh + offset);
150 
151 		switch (nexthdr) {
152 		case NEXTHDR_FRAGMENT:
153 			onlyproto = 1;
154 		case NEXTHDR_ROUTING:
155 		case NEXTHDR_HOP:
156 		case NEXTHDR_DEST:
157 			offset += ipv6_optlen(exthdr);
158 			nexthdr = exthdr->nexthdr;
159 			exthdr = (struct ipv6_opt_hdr *)(nh + offset);
160 			break;
161 
162 		case IPPROTO_UDP:
163 		case IPPROTO_UDPLITE:
164 		case IPPROTO_TCP:
165 		case IPPROTO_SCTP:
166 		case IPPROTO_DCCP:
167 			if (!onlyproto && (nh + offset + 4 < skb->data ||
168 			     pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
169 				__be16 *ports;
170 
171 				nh = skb_network_header(skb);
172 				ports = (__be16 *)(nh + offset);
173 				fl6->fl6_sport = ports[!!reverse];
174 				fl6->fl6_dport = ports[!reverse];
175 			}
176 			fl6->flowi6_proto = nexthdr;
177 			return;
178 
179 		case IPPROTO_ICMPV6:
180 			if (!onlyproto && (nh + offset + 2 < skb->data ||
181 			    pskb_may_pull(skb, nh + offset + 2 - skb->data))) {
182 				u8 *icmp;
183 
184 				nh = skb_network_header(skb);
185 				icmp = (u8 *)(nh + offset);
186 				fl6->fl6_icmp_type = icmp[0];
187 				fl6->fl6_icmp_code = icmp[1];
188 			}
189 			fl6->flowi6_proto = nexthdr;
190 			return;
191 
192 #if IS_ENABLED(CONFIG_IPV6_MIP6)
193 		case IPPROTO_MH:
194 			offset += ipv6_optlen(exthdr);
195 			if (!onlyproto && (nh + offset + 3 < skb->data ||
196 			    pskb_may_pull(skb, nh + offset + 3 - skb->data))) {
197 				struct ip6_mh *mh;
198 
199 				nh = skb_network_header(skb);
200 				mh = (struct ip6_mh *)(nh + offset);
201 				fl6->fl6_mh_type = mh->ip6mh_type;
202 			}
203 			fl6->flowi6_proto = nexthdr;
204 			return;
205 #endif
206 
207 		/* XXX Why are there these headers? */
208 		case IPPROTO_AH:
209 		case IPPROTO_ESP:
210 		case IPPROTO_COMP:
211 		default:
212 			fl6->fl6_ipsec_spi = 0;
213 			fl6->flowi6_proto = nexthdr;
214 			return;
215 		}
216 	}
217 }
218 
219 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
220 {
221 	struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
222 
223 	xfrm6_policy_afinfo.garbage_collect(net);
224 	return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
225 }
226 
227 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
228 			      struct sk_buff *skb, u32 mtu)
229 {
230 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
231 	struct dst_entry *path = xdst->route;
232 
233 	path->ops->update_pmtu(path, sk, skb, mtu);
234 }
235 
236 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
237 			   struct sk_buff *skb)
238 {
239 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
240 	struct dst_entry *path = xdst->route;
241 
242 	path->ops->redirect(path, sk, skb);
243 }
244 
245 static void xfrm6_dst_destroy(struct dst_entry *dst)
246 {
247 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
248 
249 	if (likely(xdst->u.rt6.rt6i_idev))
250 		in6_dev_put(xdst->u.rt6.rt6i_idev);
251 	dst_destroy_metrics_generic(dst);
252 	xfrm_dst_destroy(xdst);
253 }
254 
255 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
256 			     int unregister)
257 {
258 	struct xfrm_dst *xdst;
259 
260 	if (!unregister)
261 		return;
262 
263 	xdst = (struct xfrm_dst *)dst;
264 	if (xdst->u.rt6.rt6i_idev->dev == dev) {
265 		struct inet6_dev *loopback_idev =
266 			in6_dev_get(dev_net(dev)->loopback_dev);
267 		BUG_ON(!loopback_idev);
268 
269 		do {
270 			in6_dev_put(xdst->u.rt6.rt6i_idev);
271 			xdst->u.rt6.rt6i_idev = loopback_idev;
272 			in6_dev_hold(loopback_idev);
273 			xdst = (struct xfrm_dst *)xdst->u.dst.child;
274 		} while (xdst->u.dst.xfrm);
275 
276 		__in6_dev_put(loopback_idev);
277 	}
278 
279 	xfrm_dst_ifdown(dst, dev);
280 }
281 
282 static struct dst_ops xfrm6_dst_ops_template = {
283 	.family =		AF_INET6,
284 	.gc =			xfrm6_garbage_collect,
285 	.update_pmtu =		xfrm6_update_pmtu,
286 	.redirect =		xfrm6_redirect,
287 	.cow_metrics =		dst_cow_metrics_generic,
288 	.destroy =		xfrm6_dst_destroy,
289 	.ifdown =		xfrm6_dst_ifdown,
290 	.local_out =		__ip6_local_out,
291 	.gc_thresh =		INT_MAX,
292 };
293 
294 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
295 	.family =		AF_INET6,
296 	.dst_ops =		&xfrm6_dst_ops_template,
297 	.dst_lookup =		xfrm6_dst_lookup,
298 	.get_saddr =		xfrm6_get_saddr,
299 	.decode_session =	_decode_session6,
300 	.get_tos =		xfrm6_get_tos,
301 	.init_path =		xfrm6_init_path,
302 	.fill_dst =		xfrm6_fill_dst,
303 	.blackhole_route =	ip6_blackhole_route,
304 };
305 
306 static int __init xfrm6_policy_init(void)
307 {
308 	return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
309 }
310 
311 static void xfrm6_policy_fini(void)
312 {
313 	xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
314 }
315 
316 #ifdef CONFIG_SYSCTL
317 static struct ctl_table xfrm6_policy_table[] = {
318 	{
319 		.procname       = "xfrm6_gc_thresh",
320 		.data		= &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
321 		.maxlen		= sizeof(int),
322 		.mode		= 0644,
323 		.proc_handler   = proc_dointvec,
324 	},
325 	{ }
326 };
327 
328 static int __net_init xfrm6_net_sysctl_init(struct net *net)
329 {
330 	struct ctl_table *table;
331 	struct ctl_table_header *hdr;
332 
333 	table = xfrm6_policy_table;
334 	if (!net_eq(net, &init_net)) {
335 		table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
336 		if (!table)
337 			goto err_alloc;
338 
339 		table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
340 	}
341 
342 	hdr = register_net_sysctl(net, "net/ipv6", table);
343 	if (!hdr)
344 		goto err_reg;
345 
346 	net->ipv6.sysctl.xfrm6_hdr = hdr;
347 	return 0;
348 
349 err_reg:
350 	if (!net_eq(net, &init_net))
351 		kfree(table);
352 err_alloc:
353 	return -ENOMEM;
354 }
355 
356 static void __net_exit xfrm6_net_sysctl_exit(struct net *net)
357 {
358 	struct ctl_table *table;
359 
360 	if (!net->ipv6.sysctl.xfrm6_hdr)
361 		return;
362 
363 	table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
364 	unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
365 	if (!net_eq(net, &init_net))
366 		kfree(table);
367 }
368 #else /* CONFIG_SYSCTL */
369 static inline int xfrm6_net_sysctl_init(struct net *net)
370 {
371 	return 0;
372 }
373 
374 static inline void xfrm6_net_sysctl_exit(struct net *net)
375 {
376 }
377 #endif
378 
379 static int __net_init xfrm6_net_init(struct net *net)
380 {
381 	int ret;
382 
383 	memcpy(&net->xfrm.xfrm6_dst_ops, &xfrm6_dst_ops_template,
384 	       sizeof(xfrm6_dst_ops_template));
385 	ret = dst_entries_init(&net->xfrm.xfrm6_dst_ops);
386 	if (ret)
387 		return ret;
388 
389 	ret = xfrm6_net_sysctl_init(net);
390 	if (ret)
391 		dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
392 
393 	return ret;
394 }
395 
396 static void __net_exit xfrm6_net_exit(struct net *net)
397 {
398 	xfrm6_net_sysctl_exit(net);
399 	dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
400 }
401 
402 static struct pernet_operations xfrm6_net_ops = {
403 	.init	= xfrm6_net_init,
404 	.exit	= xfrm6_net_exit,
405 };
406 
407 int __init xfrm6_init(void)
408 {
409 	int ret;
410 
411 	ret = xfrm6_policy_init();
412 	if (ret)
413 		goto out;
414 	ret = xfrm6_state_init();
415 	if (ret)
416 		goto out_policy;
417 
418 	ret = xfrm6_protocol_init();
419 	if (ret)
420 		goto out_state;
421 
422 	register_pernet_subsys(&xfrm6_net_ops);
423 out:
424 	return ret;
425 out_state:
426 	xfrm6_state_fini();
427 out_policy:
428 	xfrm6_policy_fini();
429 	goto out;
430 }
431 
432 void xfrm6_fini(void)
433 {
434 	unregister_pernet_subsys(&xfrm6_net_ops);
435 	xfrm6_protocol_fini();
436 	xfrm6_policy_fini();
437 	xfrm6_state_fini();
438 }
439