xref: /linux/net/ipv6/xfrm6_policy.c (revision a8b70ccf10e38775785d9cb12ead916474549f99)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * xfrm6_policy.c: based on xfrm4_policy.c
4  *
5  * Authors:
6  *	Mitsuru KANDA @USAGI
7  *	Kazunori MIYAZAWA @USAGI
8  *	Kunihiro Ishiguro <kunihiro@ipinfusion.com>
9  *		IPv6 support
10  *	YOSHIFUJI Hideaki
11  *		Split up af-specific portion
12  *
13  */
14 
15 #include <linux/err.h>
16 #include <linux/kernel.h>
17 #include <linux/netdevice.h>
18 #include <net/addrconf.h>
19 #include <net/dst.h>
20 #include <net/xfrm.h>
21 #include <net/ip.h>
22 #include <net/ipv6.h>
23 #include <net/ip6_route.h>
24 #include <net/l3mdev.h>
25 #if IS_ENABLED(CONFIG_IPV6_MIP6)
26 #include <net/mip6.h>
27 #endif
28 
29 static struct dst_entry *xfrm6_dst_lookup(struct net *net, int tos, int oif,
30 					  const xfrm_address_t *saddr,
31 					  const xfrm_address_t *daddr,
32 					  u32 mark)
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 	fl6.flowi6_mark = mark;
42 	memcpy(&fl6.daddr, daddr, sizeof(fl6.daddr));
43 	if (saddr)
44 		memcpy(&fl6.saddr, saddr, sizeof(fl6.saddr));
45 
46 	dst = ip6_route_output(net, NULL, &fl6);
47 
48 	err = dst->error;
49 	if (dst->error) {
50 		dst_release(dst);
51 		dst = ERR_PTR(err);
52 	}
53 
54 	return dst;
55 }
56 
57 static int xfrm6_get_saddr(struct net *net, int oif,
58 			   xfrm_address_t *saddr, xfrm_address_t *daddr,
59 			   u32 mark)
60 {
61 	struct dst_entry *dst;
62 	struct net_device *dev;
63 
64 	dst = xfrm6_dst_lookup(net, 0, oif, NULL, daddr, mark);
65 	if (IS_ERR(dst))
66 		return -EHOSTUNREACH;
67 
68 	dev = ip6_dst_idev(dst)->dev;
69 	ipv6_dev_get_saddr(dev_net(dev), dev, &daddr->in6, 0, &saddr->in6);
70 	dst_release(dst);
71 	return 0;
72 }
73 
74 static int xfrm6_get_tos(const struct flowi *fl)
75 {
76 	return 0;
77 }
78 
79 static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
80 			   int nfheader_len)
81 {
82 	if (dst->ops->family == AF_INET6) {
83 		struct rt6_info *rt = (struct rt6_info *)dst;
84 		path->path_cookie = rt6_get_cookie(rt);
85 	}
86 
87 	path->u.rt6.rt6i_nfheader_len = nfheader_len;
88 
89 	return 0;
90 }
91 
92 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
93 			  const struct flowi *fl)
94 {
95 	struct rt6_info *rt = (struct rt6_info *)xdst->route;
96 
97 	xdst->u.dst.dev = dev;
98 	dev_hold(dev);
99 
100 	xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
101 	if (!xdst->u.rt6.rt6i_idev) {
102 		dev_put(dev);
103 		return -ENODEV;
104 	}
105 
106 	/* Sheit... I remember I did this right. Apparently,
107 	 * it was magically lost, so this code needs audit */
108 	xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
109 						   RTF_LOCAL);
110 	xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
111 	xdst->u.rt6.rt6i_node = rt->rt6i_node;
112 	xdst->route_cookie = rt6_get_cookie(rt);
113 	xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
114 	xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
115 	xdst->u.rt6.rt6i_src = rt->rt6i_src;
116 	INIT_LIST_HEAD(&xdst->u.rt6.rt6i_uncached);
117 	rt6_uncached_list_add(&xdst->u.rt6);
118 	atomic_inc(&dev_net(dev)->ipv6.rt6_stats->fib_rt_uncache);
119 
120 	return 0;
121 }
122 
123 static inline void
124 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
125 {
126 	struct flowi6 *fl6 = &fl->u.ip6;
127 	int onlyproto = 0;
128 	const struct ipv6hdr *hdr = ipv6_hdr(skb);
129 	u16 offset = sizeof(*hdr);
130 	struct ipv6_opt_hdr *exthdr;
131 	const unsigned char *nh = skb_network_header(skb);
132 	u16 nhoff = IP6CB(skb)->nhoff;
133 	int oif = 0;
134 	u8 nexthdr;
135 
136 	if (!nhoff)
137 		nhoff = offsetof(struct ipv6hdr, nexthdr);
138 
139 	nexthdr = nh[nhoff];
140 
141 	if (skb_dst(skb))
142 		oif = skb_dst(skb)->dev->ifindex;
143 
144 	memset(fl6, 0, sizeof(struct flowi6));
145 	fl6->flowi6_mark = skb->mark;
146 	fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
147 
148 	fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
149 	fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
150 
151 	while (nh + offset + 1 < skb->data ||
152 	       pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
153 		nh = skb_network_header(skb);
154 		exthdr = (struct ipv6_opt_hdr *)(nh + offset);
155 
156 		switch (nexthdr) {
157 		case NEXTHDR_FRAGMENT:
158 			onlyproto = 1;
159 			/* fall through */
160 		case NEXTHDR_ROUTING:
161 		case NEXTHDR_HOP:
162 		case NEXTHDR_DEST:
163 			offset += ipv6_optlen(exthdr);
164 			nexthdr = exthdr->nexthdr;
165 			exthdr = (struct ipv6_opt_hdr *)(nh + offset);
166 			break;
167 
168 		case IPPROTO_UDP:
169 		case IPPROTO_UDPLITE:
170 		case IPPROTO_TCP:
171 		case IPPROTO_SCTP:
172 		case IPPROTO_DCCP:
173 			if (!onlyproto && (nh + offset + 4 < skb->data ||
174 			     pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
175 				__be16 *ports;
176 
177 				nh = skb_network_header(skb);
178 				ports = (__be16 *)(nh + offset);
179 				fl6->fl6_sport = ports[!!reverse];
180 				fl6->fl6_dport = ports[!reverse];
181 			}
182 			fl6->flowi6_proto = nexthdr;
183 			return;
184 
185 		case IPPROTO_ICMPV6:
186 			if (!onlyproto && (nh + offset + 2 < skb->data ||
187 			    pskb_may_pull(skb, nh + offset + 2 - skb->data))) {
188 				u8 *icmp;
189 
190 				nh = skb_network_header(skb);
191 				icmp = (u8 *)(nh + offset);
192 				fl6->fl6_icmp_type = icmp[0];
193 				fl6->fl6_icmp_code = icmp[1];
194 			}
195 			fl6->flowi6_proto = nexthdr;
196 			return;
197 
198 #if IS_ENABLED(CONFIG_IPV6_MIP6)
199 		case IPPROTO_MH:
200 			offset += ipv6_optlen(exthdr);
201 			if (!onlyproto && (nh + offset + 3 < skb->data ||
202 			    pskb_may_pull(skb, nh + offset + 3 - skb->data))) {
203 				struct ip6_mh *mh;
204 
205 				nh = skb_network_header(skb);
206 				mh = (struct ip6_mh *)(nh + offset);
207 				fl6->fl6_mh_type = mh->ip6mh_type;
208 			}
209 			fl6->flowi6_proto = nexthdr;
210 			return;
211 #endif
212 
213 		/* XXX Why are there these headers? */
214 		case IPPROTO_AH:
215 		case IPPROTO_ESP:
216 		case IPPROTO_COMP:
217 		default:
218 			fl6->fl6_ipsec_spi = 0;
219 			fl6->flowi6_proto = nexthdr;
220 			return;
221 		}
222 	}
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 	if (xdst->u.rt6.rt6i_uncached_list)
251 		rt6_uncached_list_del(&xdst->u.rt6);
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 *)xfrm_dst_child(&xdst->u.dst);
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 	.update_pmtu =		xfrm6_update_pmtu,
285 	.redirect =		xfrm6_redirect,
286 	.cow_metrics =		dst_cow_metrics_generic,
287 	.destroy =		xfrm6_dst_destroy,
288 	.ifdown =		xfrm6_dst_ifdown,
289 	.local_out =		__ip6_local_out,
290 	.gc_thresh =		32768,
291 };
292 
293 static const struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
294 	.dst_ops =		&xfrm6_dst_ops_template,
295 	.dst_lookup =		xfrm6_dst_lookup,
296 	.get_saddr =		xfrm6_get_saddr,
297 	.decode_session =	_decode_session6,
298 	.get_tos =		xfrm6_get_tos,
299 	.init_path =		xfrm6_init_path,
300 	.fill_dst =		xfrm6_fill_dst,
301 	.blackhole_route =	ip6_blackhole_route,
302 };
303 
304 static int __init xfrm6_policy_init(void)
305 {
306 	return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo, AF_INET6);
307 }
308 
309 static void xfrm6_policy_fini(void)
310 {
311 	xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
312 }
313 
314 #ifdef CONFIG_SYSCTL
315 static struct ctl_table xfrm6_policy_table[] = {
316 	{
317 		.procname       = "xfrm6_gc_thresh",
318 		.data		= &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
319 		.maxlen		= sizeof(int),
320 		.mode		= 0644,
321 		.proc_handler   = proc_dointvec,
322 	},
323 	{ }
324 };
325 
326 static int __net_init xfrm6_net_sysctl_init(struct net *net)
327 {
328 	struct ctl_table *table;
329 	struct ctl_table_header *hdr;
330 
331 	table = xfrm6_policy_table;
332 	if (!net_eq(net, &init_net)) {
333 		table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
334 		if (!table)
335 			goto err_alloc;
336 
337 		table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
338 	}
339 
340 	hdr = register_net_sysctl(net, "net/ipv6", table);
341 	if (!hdr)
342 		goto err_reg;
343 
344 	net->ipv6.sysctl.xfrm6_hdr = hdr;
345 	return 0;
346 
347 err_reg:
348 	if (!net_eq(net, &init_net))
349 		kfree(table);
350 err_alloc:
351 	return -ENOMEM;
352 }
353 
354 static void __net_exit xfrm6_net_sysctl_exit(struct net *net)
355 {
356 	struct ctl_table *table;
357 
358 	if (!net->ipv6.sysctl.xfrm6_hdr)
359 		return;
360 
361 	table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
362 	unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
363 	if (!net_eq(net, &init_net))
364 		kfree(table);
365 }
366 #else /* CONFIG_SYSCTL */
367 static inline int xfrm6_net_sysctl_init(struct net *net)
368 {
369 	return 0;
370 }
371 
372 static inline void xfrm6_net_sysctl_exit(struct net *net)
373 {
374 }
375 #endif
376 
377 static int __net_init xfrm6_net_init(struct net *net)
378 {
379 	int ret;
380 
381 	memcpy(&net->xfrm.xfrm6_dst_ops, &xfrm6_dst_ops_template,
382 	       sizeof(xfrm6_dst_ops_template));
383 	ret = dst_entries_init(&net->xfrm.xfrm6_dst_ops);
384 	if (ret)
385 		return ret;
386 
387 	ret = xfrm6_net_sysctl_init(net);
388 	if (ret)
389 		dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
390 
391 	return ret;
392 }
393 
394 static void __net_exit xfrm6_net_exit(struct net *net)
395 {
396 	xfrm6_net_sysctl_exit(net);
397 	dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
398 }
399 
400 static struct pernet_operations xfrm6_net_ops = {
401 	.init	= xfrm6_net_init,
402 	.exit	= xfrm6_net_exit,
403 };
404 
405 int __init xfrm6_init(void)
406 {
407 	int ret;
408 
409 	ret = xfrm6_policy_init();
410 	if (ret)
411 		goto out;
412 	ret = xfrm6_state_init();
413 	if (ret)
414 		goto out_policy;
415 
416 	ret = xfrm6_protocol_init();
417 	if (ret)
418 		goto out_state;
419 
420 	register_pernet_subsys(&xfrm6_net_ops);
421 out:
422 	return ret;
423 out_state:
424 	xfrm6_state_fini();
425 out_policy:
426 	xfrm6_policy_fini();
427 	goto out;
428 }
429 
430 void xfrm6_fini(void)
431 {
432 	unregister_pernet_subsys(&xfrm6_net_ops);
433 	xfrm6_protocol_fini();
434 	xfrm6_policy_fini();
435 	xfrm6_state_fini();
436 }
437