xref: /linux/net/ipv6/xfrm6_policy.c (revision ca55b2fef3a9373fcfc30f82fd26bc7fccbda732)
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/vrf.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 = 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 = vrf_master_ifindex(skb_dst(skb)->dev) ?
138 			: skb_dst(skb)->dev->ifindex;
139 	}
140 
141 	memset(fl6, 0, sizeof(struct flowi6));
142 	fl6->flowi6_mark = skb->mark;
143 	fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
144 
145 	fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
146 	fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
147 
148 	while (nh + offset + 1 < skb->data ||
149 	       pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
150 		nh = skb_network_header(skb);
151 		exthdr = (struct ipv6_opt_hdr *)(nh + offset);
152 
153 		switch (nexthdr) {
154 		case NEXTHDR_FRAGMENT:
155 			onlyproto = 1;
156 		case NEXTHDR_ROUTING:
157 		case NEXTHDR_HOP:
158 		case NEXTHDR_DEST:
159 			offset += ipv6_optlen(exthdr);
160 			nexthdr = exthdr->nexthdr;
161 			exthdr = (struct ipv6_opt_hdr *)(nh + offset);
162 			break;
163 
164 		case IPPROTO_UDP:
165 		case IPPROTO_UDPLITE:
166 		case IPPROTO_TCP:
167 		case IPPROTO_SCTP:
168 		case IPPROTO_DCCP:
169 			if (!onlyproto && (nh + offset + 4 < skb->data ||
170 			     pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
171 				__be16 *ports;
172 
173 				nh = skb_network_header(skb);
174 				ports = (__be16 *)(nh + offset);
175 				fl6->fl6_sport = ports[!!reverse];
176 				fl6->fl6_dport = ports[!reverse];
177 			}
178 			fl6->flowi6_proto = nexthdr;
179 			return;
180 
181 		case IPPROTO_ICMPV6:
182 			if (!onlyproto && (nh + offset + 2 < skb->data ||
183 			    pskb_may_pull(skb, nh + offset + 2 - skb->data))) {
184 				u8 *icmp;
185 
186 				nh = skb_network_header(skb);
187 				icmp = (u8 *)(nh + offset);
188 				fl6->fl6_icmp_type = icmp[0];
189 				fl6->fl6_icmp_code = icmp[1];
190 			}
191 			fl6->flowi6_proto = nexthdr;
192 			return;
193 
194 #if IS_ENABLED(CONFIG_IPV6_MIP6)
195 		case IPPROTO_MH:
196 			offset += ipv6_optlen(exthdr);
197 			if (!onlyproto && (nh + offset + 3 < skb->data ||
198 			    pskb_may_pull(skb, nh + offset + 3 - skb->data))) {
199 				struct ip6_mh *mh;
200 
201 				nh = skb_network_header(skb);
202 				mh = (struct ip6_mh *)(nh + offset);
203 				fl6->fl6_mh_type = mh->ip6mh_type;
204 			}
205 			fl6->flowi6_proto = nexthdr;
206 			return;
207 #endif
208 
209 		/* XXX Why are there these headers? */
210 		case IPPROTO_AH:
211 		case IPPROTO_ESP:
212 		case IPPROTO_COMP:
213 		default:
214 			fl6->fl6_ipsec_spi = 0;
215 			fl6->flowi6_proto = nexthdr;
216 			return;
217 		}
218 	}
219 }
220 
221 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
222 {
223 	struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
224 
225 	xfrm6_policy_afinfo.garbage_collect(net);
226 	return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
227 }
228 
229 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
230 			      struct sk_buff *skb, u32 mtu)
231 {
232 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
233 	struct dst_entry *path = xdst->route;
234 
235 	path->ops->update_pmtu(path, sk, skb, mtu);
236 }
237 
238 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
239 			   struct sk_buff *skb)
240 {
241 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
242 	struct dst_entry *path = xdst->route;
243 
244 	path->ops->redirect(path, sk, skb);
245 }
246 
247 static void xfrm6_dst_destroy(struct dst_entry *dst)
248 {
249 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
250 
251 	if (likely(xdst->u.rt6.rt6i_idev))
252 		in6_dev_put(xdst->u.rt6.rt6i_idev);
253 	dst_destroy_metrics_generic(dst);
254 	xfrm_dst_destroy(xdst);
255 }
256 
257 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
258 			     int unregister)
259 {
260 	struct xfrm_dst *xdst;
261 
262 	if (!unregister)
263 		return;
264 
265 	xdst = (struct xfrm_dst *)dst;
266 	if (xdst->u.rt6.rt6i_idev->dev == dev) {
267 		struct inet6_dev *loopback_idev =
268 			in6_dev_get(dev_net(dev)->loopback_dev);
269 		BUG_ON(!loopback_idev);
270 
271 		do {
272 			in6_dev_put(xdst->u.rt6.rt6i_idev);
273 			xdst->u.rt6.rt6i_idev = loopback_idev;
274 			in6_dev_hold(loopback_idev);
275 			xdst = (struct xfrm_dst *)xdst->u.dst.child;
276 		} while (xdst->u.dst.xfrm);
277 
278 		__in6_dev_put(loopback_idev);
279 	}
280 
281 	xfrm_dst_ifdown(dst, dev);
282 }
283 
284 static struct dst_ops xfrm6_dst_ops = {
285 	.family =		AF_INET6,
286 	.gc =			xfrm6_garbage_collect,
287 	.update_pmtu =		xfrm6_update_pmtu,
288 	.redirect =		xfrm6_redirect,
289 	.cow_metrics =		dst_cow_metrics_generic,
290 	.destroy =		xfrm6_dst_destroy,
291 	.ifdown =		xfrm6_dst_ifdown,
292 	.local_out =		__ip6_local_out,
293 	.gc_thresh =		32768,
294 };
295 
296 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
297 	.family =		AF_INET6,
298 	.dst_ops =		&xfrm6_dst_ops,
299 	.dst_lookup =		xfrm6_dst_lookup,
300 	.get_saddr =		xfrm6_get_saddr,
301 	.decode_session =	_decode_session6,
302 	.get_tos =		xfrm6_get_tos,
303 	.init_path =		xfrm6_init_path,
304 	.fill_dst =		xfrm6_fill_dst,
305 	.blackhole_route =	ip6_blackhole_route,
306 };
307 
308 static int __init xfrm6_policy_init(void)
309 {
310 	return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
311 }
312 
313 static void xfrm6_policy_fini(void)
314 {
315 	xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
316 }
317 
318 #ifdef CONFIG_SYSCTL
319 static struct ctl_table xfrm6_policy_table[] = {
320 	{
321 		.procname       = "xfrm6_gc_thresh",
322 		.data		= &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
323 		.maxlen		= sizeof(int),
324 		.mode		= 0644,
325 		.proc_handler   = proc_dointvec,
326 	},
327 	{ }
328 };
329 
330 static int __net_init xfrm6_net_init(struct net *net)
331 {
332 	struct ctl_table *table;
333 	struct ctl_table_header *hdr;
334 
335 	table = xfrm6_policy_table;
336 	if (!net_eq(net, &init_net)) {
337 		table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
338 		if (!table)
339 			goto err_alloc;
340 
341 		table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
342 	}
343 
344 	hdr = register_net_sysctl(net, "net/ipv6", table);
345 	if (!hdr)
346 		goto err_reg;
347 
348 	net->ipv6.sysctl.xfrm6_hdr = hdr;
349 	return 0;
350 
351 err_reg:
352 	if (!net_eq(net, &init_net))
353 		kfree(table);
354 err_alloc:
355 	return -ENOMEM;
356 }
357 
358 static void __net_exit xfrm6_net_exit(struct net *net)
359 {
360 	struct ctl_table *table;
361 
362 	if (!net->ipv6.sysctl.xfrm6_hdr)
363 		return;
364 
365 	table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
366 	unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
367 	if (!net_eq(net, &init_net))
368 		kfree(table);
369 }
370 
371 static struct pernet_operations xfrm6_net_ops = {
372 	.init	= xfrm6_net_init,
373 	.exit	= xfrm6_net_exit,
374 };
375 #endif
376 
377 int __init xfrm6_init(void)
378 {
379 	int ret;
380 
381 	dst_entries_init(&xfrm6_dst_ops);
382 
383 	ret = xfrm6_policy_init();
384 	if (ret) {
385 		dst_entries_destroy(&xfrm6_dst_ops);
386 		goto out;
387 	}
388 	ret = xfrm6_state_init();
389 	if (ret)
390 		goto out_policy;
391 
392 	ret = xfrm6_protocol_init();
393 	if (ret)
394 		goto out_state;
395 
396 #ifdef CONFIG_SYSCTL
397 	register_pernet_subsys(&xfrm6_net_ops);
398 #endif
399 out:
400 	return ret;
401 out_state:
402 	xfrm6_state_fini();
403 out_policy:
404 	xfrm6_policy_fini();
405 	goto out;
406 }
407 
408 void xfrm6_fini(void)
409 {
410 #ifdef CONFIG_SYSCTL
411 	unregister_pernet_subsys(&xfrm6_net_ops);
412 #endif
413 	xfrm6_protocol_fini();
414 	xfrm6_policy_fini();
415 	xfrm6_state_fini();
416 	dst_entries_destroy(&xfrm6_dst_ops);
417 }
418