xref: /openbmc/linux/net/ipv6/xfrm6_policy.c (revision 4f6cce39)
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 dst_entry *xfrm6_dst_lookup(struct net *net, int tos, int oif,
29 					  const xfrm_address_t *saddr,
30 					  const xfrm_address_t *daddr)
31 {
32 	struct flowi6 fl6;
33 	struct dst_entry *dst;
34 	int err;
35 
36 	memset(&fl6, 0, sizeof(fl6));
37 	fl6.flowi6_oif = l3mdev_master_ifindex_by_index(net, oif);
38 	fl6.flowi6_flags = FLOWI_FLAG_SKIP_NH_OIF;
39 	memcpy(&fl6.daddr, daddr, sizeof(fl6.daddr));
40 	if (saddr)
41 		memcpy(&fl6.saddr, saddr, sizeof(fl6.saddr));
42 
43 	dst = ip6_route_output(net, NULL, &fl6);
44 
45 	err = dst->error;
46 	if (dst->error) {
47 		dst_release(dst);
48 		dst = ERR_PTR(err);
49 	}
50 
51 	return dst;
52 }
53 
54 static int xfrm6_get_saddr(struct net *net, int oif,
55 			   xfrm_address_t *saddr, xfrm_address_t *daddr)
56 {
57 	struct dst_entry *dst;
58 	struct net_device *dev;
59 
60 	dst = xfrm6_dst_lookup(net, 0, oif, NULL, daddr);
61 	if (IS_ERR(dst))
62 		return -EHOSTUNREACH;
63 
64 	dev = ip6_dst_idev(dst)->dev;
65 	ipv6_dev_get_saddr(dev_net(dev), dev, &daddr->in6, 0, &saddr->in6);
66 	dst_release(dst);
67 	return 0;
68 }
69 
70 static int xfrm6_get_tos(const struct flowi *fl)
71 {
72 	return 0;
73 }
74 
75 static 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 		path->path_cookie = rt6_get_cookie(rt);
81 	}
82 
83 	path->u.rt6.rt6i_nfheader_len = nfheader_len;
84 
85 	return 0;
86 }
87 
88 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
89 			  const struct flowi *fl)
90 {
91 	struct rt6_info *rt = (struct rt6_info *)xdst->route;
92 
93 	xdst->u.dst.dev = dev;
94 	dev_hold(dev);
95 
96 	xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
97 	if (!xdst->u.rt6.rt6i_idev) {
98 		dev_put(dev);
99 		return -ENODEV;
100 	}
101 
102 	/* Sheit... I remember I did this right. Apparently,
103 	 * it was magically lost, so this code needs audit */
104 	xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
105 						   RTF_LOCAL);
106 	xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
107 	xdst->u.rt6.rt6i_node = rt->rt6i_node;
108 	xdst->route_cookie = rt6_get_cookie(rt);
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 
116 static inline void
117 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
118 {
119 	struct flowi6 *fl6 = &fl->u.ip6;
120 	int onlyproto = 0;
121 	const struct ipv6hdr *hdr = ipv6_hdr(skb);
122 	u16 offset = sizeof(*hdr);
123 	struct ipv6_opt_hdr *exthdr;
124 	const unsigned char *nh = skb_network_header(skb);
125 	u16 nhoff = IP6CB(skb)->nhoff;
126 	int oif = 0;
127 	u8 nexthdr;
128 
129 	if (!nhoff)
130 		nhoff = offsetof(struct ipv6hdr, nexthdr);
131 
132 	nexthdr = nh[nhoff];
133 
134 	if (skb_dst(skb))
135 		oif = skb_dst(skb)->dev->ifindex;
136 
137 	memset(fl6, 0, sizeof(struct flowi6));
138 	fl6->flowi6_mark = skb->mark;
139 	fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
140 
141 	fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
142 	fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
143 
144 	while (nh + offset + 1 < skb->data ||
145 	       pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
146 		nh = skb_network_header(skb);
147 		exthdr = (struct ipv6_opt_hdr *)(nh + offset);
148 
149 		switch (nexthdr) {
150 		case NEXTHDR_FRAGMENT:
151 			onlyproto = 1;
152 		case NEXTHDR_ROUTING:
153 		case NEXTHDR_HOP:
154 		case NEXTHDR_DEST:
155 			offset += ipv6_optlen(exthdr);
156 			nexthdr = exthdr->nexthdr;
157 			exthdr = (struct ipv6_opt_hdr *)(nh + offset);
158 			break;
159 
160 		case IPPROTO_UDP:
161 		case IPPROTO_UDPLITE:
162 		case IPPROTO_TCP:
163 		case IPPROTO_SCTP:
164 		case IPPROTO_DCCP:
165 			if (!onlyproto && (nh + offset + 4 < skb->data ||
166 			     pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
167 				__be16 *ports;
168 
169 				nh = skb_network_header(skb);
170 				ports = (__be16 *)(nh + offset);
171 				fl6->fl6_sport = ports[!!reverse];
172 				fl6->fl6_dport = ports[!reverse];
173 			}
174 			fl6->flowi6_proto = nexthdr;
175 			return;
176 
177 		case IPPROTO_ICMPV6:
178 			if (!onlyproto && (nh + offset + 2 < skb->data ||
179 			    pskb_may_pull(skb, nh + offset + 2 - skb->data))) {
180 				u8 *icmp;
181 
182 				nh = skb_network_header(skb);
183 				icmp = (u8 *)(nh + offset);
184 				fl6->fl6_icmp_type = icmp[0];
185 				fl6->fl6_icmp_code = icmp[1];
186 			}
187 			fl6->flowi6_proto = nexthdr;
188 			return;
189 
190 #if IS_ENABLED(CONFIG_IPV6_MIP6)
191 		case IPPROTO_MH:
192 			offset += ipv6_optlen(exthdr);
193 			if (!onlyproto && (nh + offset + 3 < skb->data ||
194 			    pskb_may_pull(skb, nh + offset + 3 - skb->data))) {
195 				struct ip6_mh *mh;
196 
197 				nh = skb_network_header(skb);
198 				mh = (struct ip6_mh *)(nh + offset);
199 				fl6->fl6_mh_type = mh->ip6mh_type;
200 			}
201 			fl6->flowi6_proto = nexthdr;
202 			return;
203 #endif
204 
205 		/* XXX Why are there these headers? */
206 		case IPPROTO_AH:
207 		case IPPROTO_ESP:
208 		case IPPROTO_COMP:
209 		default:
210 			fl6->fl6_ipsec_spi = 0;
211 			fl6->flowi6_proto = nexthdr;
212 			return;
213 		}
214 	}
215 }
216 
217 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
218 {
219 	struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
220 
221 	xfrm_garbage_collect_deferred(net);
222 	return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
223 }
224 
225 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
226 			      struct sk_buff *skb, u32 mtu)
227 {
228 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
229 	struct dst_entry *path = xdst->route;
230 
231 	path->ops->update_pmtu(path, sk, skb, mtu);
232 }
233 
234 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
235 			   struct sk_buff *skb)
236 {
237 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
238 	struct dst_entry *path = xdst->route;
239 
240 	path->ops->redirect(path, sk, skb);
241 }
242 
243 static void xfrm6_dst_destroy(struct dst_entry *dst)
244 {
245 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
246 
247 	if (likely(xdst->u.rt6.rt6i_idev))
248 		in6_dev_put(xdst->u.rt6.rt6i_idev);
249 	dst_destroy_metrics_generic(dst);
250 	xfrm_dst_destroy(xdst);
251 }
252 
253 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
254 			     int unregister)
255 {
256 	struct xfrm_dst *xdst;
257 
258 	if (!unregister)
259 		return;
260 
261 	xdst = (struct xfrm_dst *)dst;
262 	if (xdst->u.rt6.rt6i_idev->dev == dev) {
263 		struct inet6_dev *loopback_idev =
264 			in6_dev_get(dev_net(dev)->loopback_dev);
265 		BUG_ON(!loopback_idev);
266 
267 		do {
268 			in6_dev_put(xdst->u.rt6.rt6i_idev);
269 			xdst->u.rt6.rt6i_idev = loopback_idev;
270 			in6_dev_hold(loopback_idev);
271 			xdst = (struct xfrm_dst *)xdst->u.dst.child;
272 		} while (xdst->u.dst.xfrm);
273 
274 		__in6_dev_put(loopback_idev);
275 	}
276 
277 	xfrm_dst_ifdown(dst, dev);
278 }
279 
280 static struct dst_ops xfrm6_dst_ops_template = {
281 	.family =		AF_INET6,
282 	.gc =			xfrm6_garbage_collect,
283 	.update_pmtu =		xfrm6_update_pmtu,
284 	.redirect =		xfrm6_redirect,
285 	.cow_metrics =		dst_cow_metrics_generic,
286 	.destroy =		xfrm6_dst_destroy,
287 	.ifdown =		xfrm6_dst_ifdown,
288 	.local_out =		__ip6_local_out,
289 	.gc_thresh =		INT_MAX,
290 };
291 
292 static const struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
293 	.dst_ops =		&xfrm6_dst_ops_template,
294 	.dst_lookup =		xfrm6_dst_lookup,
295 	.get_saddr =		xfrm6_get_saddr,
296 	.decode_session =	_decode_session6,
297 	.get_tos =		xfrm6_get_tos,
298 	.init_path =		xfrm6_init_path,
299 	.fill_dst =		xfrm6_fill_dst,
300 	.blackhole_route =	ip6_blackhole_route,
301 };
302 
303 static int __init xfrm6_policy_init(void)
304 {
305 	return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo, AF_INET6);
306 }
307 
308 static void xfrm6_policy_fini(void)
309 {
310 	xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
311 }
312 
313 #ifdef CONFIG_SYSCTL
314 static struct ctl_table xfrm6_policy_table[] = {
315 	{
316 		.procname       = "xfrm6_gc_thresh",
317 		.data		= &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
318 		.maxlen		= sizeof(int),
319 		.mode		= 0644,
320 		.proc_handler   = proc_dointvec,
321 	},
322 	{ }
323 };
324 
325 static int __net_init xfrm6_net_sysctl_init(struct net *net)
326 {
327 	struct ctl_table *table;
328 	struct ctl_table_header *hdr;
329 
330 	table = xfrm6_policy_table;
331 	if (!net_eq(net, &init_net)) {
332 		table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
333 		if (!table)
334 			goto err_alloc;
335 
336 		table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
337 	}
338 
339 	hdr = register_net_sysctl(net, "net/ipv6", table);
340 	if (!hdr)
341 		goto err_reg;
342 
343 	net->ipv6.sysctl.xfrm6_hdr = hdr;
344 	return 0;
345 
346 err_reg:
347 	if (!net_eq(net, &init_net))
348 		kfree(table);
349 err_alloc:
350 	return -ENOMEM;
351 }
352 
353 static void __net_exit xfrm6_net_sysctl_exit(struct net *net)
354 {
355 	struct ctl_table *table;
356 
357 	if (!net->ipv6.sysctl.xfrm6_hdr)
358 		return;
359 
360 	table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
361 	unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
362 	if (!net_eq(net, &init_net))
363 		kfree(table);
364 }
365 #else /* CONFIG_SYSCTL */
366 static inline int xfrm6_net_sysctl_init(struct net *net)
367 {
368 	return 0;
369 }
370 
371 static inline void xfrm6_net_sysctl_exit(struct net *net)
372 {
373 }
374 #endif
375 
376 static int __net_init xfrm6_net_init(struct net *net)
377 {
378 	int ret;
379 
380 	memcpy(&net->xfrm.xfrm6_dst_ops, &xfrm6_dst_ops_template,
381 	       sizeof(xfrm6_dst_ops_template));
382 	ret = dst_entries_init(&net->xfrm.xfrm6_dst_ops);
383 	if (ret)
384 		return ret;
385 
386 	ret = xfrm6_net_sysctl_init(net);
387 	if (ret)
388 		dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
389 
390 	return ret;
391 }
392 
393 static void __net_exit xfrm6_net_exit(struct net *net)
394 {
395 	xfrm6_net_sysctl_exit(net);
396 	dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
397 }
398 
399 static struct pernet_operations xfrm6_net_ops = {
400 	.init	= xfrm6_net_init,
401 	.exit	= xfrm6_net_exit,
402 };
403 
404 int __init xfrm6_init(void)
405 {
406 	int ret;
407 
408 	ret = xfrm6_policy_init();
409 	if (ret)
410 		goto out;
411 	ret = xfrm6_state_init();
412 	if (ret)
413 		goto out_policy;
414 
415 	ret = xfrm6_protocol_init();
416 	if (ret)
417 		goto out_state;
418 
419 	register_pernet_subsys(&xfrm6_net_ops);
420 out:
421 	return ret;
422 out_state:
423 	xfrm6_state_fini();
424 out_policy:
425 	xfrm6_policy_fini();
426 	goto out;
427 }
428 
429 void xfrm6_fini(void)
430 {
431 	unregister_pernet_subsys(&xfrm6_net_ops);
432 	xfrm6_protocol_fini();
433 	xfrm6_policy_fini();
434 	xfrm6_state_fini();
435 }
436