xref: /linux/net/ipv6/xfrm6_policy.c (revision 791d3ef2e11100449837dc0b6fe884e60ca3a484)
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->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 	INIT_LIST_HEAD(&xdst->u.rt6.rt6i_uncached);
115 	rt6_uncached_list_add(&xdst->u.rt6);
116 	atomic_inc(&dev_net(dev)->ipv6.rt6_stats->fib_rt_uncache);
117 
118 	return 0;
119 }
120 
121 static inline void
122 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
123 {
124 	struct flowi6 *fl6 = &fl->u.ip6;
125 	int onlyproto = 0;
126 	const struct ipv6hdr *hdr = ipv6_hdr(skb);
127 	u32 offset = sizeof(*hdr);
128 	struct ipv6_opt_hdr *exthdr;
129 	const unsigned char *nh = skb_network_header(skb);
130 	u16 nhoff = IP6CB(skb)->nhoff;
131 	int oif = 0;
132 	u8 nexthdr;
133 
134 	if (!nhoff)
135 		nhoff = offsetof(struct ipv6hdr, nexthdr);
136 
137 	nexthdr = nh[nhoff];
138 
139 	if (skb_dst(skb))
140 		oif = skb_dst(skb)->dev->ifindex;
141 
142 	memset(fl6, 0, sizeof(struct flowi6));
143 	fl6->flowi6_mark = skb->mark;
144 	fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
145 
146 	fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
147 	fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
148 
149 	while (nh + offset + 1 < skb->data ||
150 	       pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
151 		nh = skb_network_header(skb);
152 		exthdr = (struct ipv6_opt_hdr *)(nh + offset);
153 
154 		switch (nexthdr) {
155 		case NEXTHDR_FRAGMENT:
156 			onlyproto = 1;
157 			/* fall through */
158 		case NEXTHDR_ROUTING:
159 		case NEXTHDR_HOP:
160 		case NEXTHDR_DEST:
161 			offset += ipv6_optlen(exthdr);
162 			nexthdr = exthdr->nexthdr;
163 			exthdr = (struct ipv6_opt_hdr *)(nh + offset);
164 			break;
165 
166 		case IPPROTO_UDP:
167 		case IPPROTO_UDPLITE:
168 		case IPPROTO_TCP:
169 		case IPPROTO_SCTP:
170 		case IPPROTO_DCCP:
171 			if (!onlyproto && (nh + offset + 4 < skb->data ||
172 			     pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
173 				__be16 *ports;
174 
175 				nh = skb_network_header(skb);
176 				ports = (__be16 *)(nh + offset);
177 				fl6->fl6_sport = ports[!!reverse];
178 				fl6->fl6_dport = ports[!reverse];
179 			}
180 			fl6->flowi6_proto = nexthdr;
181 			return;
182 
183 		case IPPROTO_ICMPV6:
184 			if (!onlyproto && (nh + offset + 2 < skb->data ||
185 			    pskb_may_pull(skb, nh + offset + 2 - skb->data))) {
186 				u8 *icmp;
187 
188 				nh = skb_network_header(skb);
189 				icmp = (u8 *)(nh + offset);
190 				fl6->fl6_icmp_type = icmp[0];
191 				fl6->fl6_icmp_code = icmp[1];
192 			}
193 			fl6->flowi6_proto = nexthdr;
194 			return;
195 
196 #if IS_ENABLED(CONFIG_IPV6_MIP6)
197 		case IPPROTO_MH:
198 			offset += ipv6_optlen(exthdr);
199 			if (!onlyproto && (nh + offset + 3 < skb->data ||
200 			    pskb_may_pull(skb, nh + offset + 3 - skb->data))) {
201 				struct ip6_mh *mh;
202 
203 				nh = skb_network_header(skb);
204 				mh = (struct ip6_mh *)(nh + offset);
205 				fl6->fl6_mh_type = mh->ip6mh_type;
206 			}
207 			fl6->flowi6_proto = nexthdr;
208 			return;
209 #endif
210 
211 		/* XXX Why are there these headers? */
212 		case IPPROTO_AH:
213 		case IPPROTO_ESP:
214 		case IPPROTO_COMP:
215 		default:
216 			fl6->fl6_ipsec_spi = 0;
217 			fl6->flowi6_proto = nexthdr;
218 			return;
219 		}
220 	}
221 }
222 
223 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
224 			      struct sk_buff *skb, u32 mtu)
225 {
226 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
227 	struct dst_entry *path = xdst->route;
228 
229 	path->ops->update_pmtu(path, sk, skb, mtu);
230 }
231 
232 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
233 			   struct sk_buff *skb)
234 {
235 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
236 	struct dst_entry *path = xdst->route;
237 
238 	path->ops->redirect(path, sk, skb);
239 }
240 
241 static void xfrm6_dst_destroy(struct dst_entry *dst)
242 {
243 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
244 
245 	if (likely(xdst->u.rt6.rt6i_idev))
246 		in6_dev_put(xdst->u.rt6.rt6i_idev);
247 	dst_destroy_metrics_generic(dst);
248 	if (xdst->u.rt6.rt6i_uncached_list)
249 		rt6_uncached_list_del(&xdst->u.rt6);
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 *)xfrm_dst_child(&xdst->u.dst);
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 	.update_pmtu =		xfrm6_update_pmtu,
283 	.redirect =		xfrm6_redirect,
284 	.cow_metrics =		dst_cow_metrics_generic,
285 	.destroy =		xfrm6_dst_destroy,
286 	.ifdown =		xfrm6_dst_ifdown,
287 	.local_out =		__ip6_local_out,
288 	.gc_thresh =		32768,
289 };
290 
291 static const struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
292 	.dst_ops =		&xfrm6_dst_ops_template,
293 	.dst_lookup =		xfrm6_dst_lookup,
294 	.get_saddr =		xfrm6_get_saddr,
295 	.decode_session =	_decode_session6,
296 	.get_tos =		xfrm6_get_tos,
297 	.init_path =		xfrm6_init_path,
298 	.fill_dst =		xfrm6_fill_dst,
299 	.blackhole_route =	ip6_blackhole_route,
300 };
301 
302 static int __init xfrm6_policy_init(void)
303 {
304 	return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo, AF_INET6);
305 }
306 
307 static void xfrm6_policy_fini(void)
308 {
309 	xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
310 }
311 
312 #ifdef CONFIG_SYSCTL
313 static struct ctl_table xfrm6_policy_table[] = {
314 	{
315 		.procname       = "xfrm6_gc_thresh",
316 		.data		= &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
317 		.maxlen		= sizeof(int),
318 		.mode		= 0644,
319 		.proc_handler   = proc_dointvec,
320 	},
321 	{ }
322 };
323 
324 static int __net_init xfrm6_net_sysctl_init(struct net *net)
325 {
326 	struct ctl_table *table;
327 	struct ctl_table_header *hdr;
328 
329 	table = xfrm6_policy_table;
330 	if (!net_eq(net, &init_net)) {
331 		table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
332 		if (!table)
333 			goto err_alloc;
334 
335 		table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
336 	}
337 
338 	hdr = register_net_sysctl(net, "net/ipv6", table);
339 	if (!hdr)
340 		goto err_reg;
341 
342 	net->ipv6.sysctl.xfrm6_hdr = hdr;
343 	return 0;
344 
345 err_reg:
346 	if (!net_eq(net, &init_net))
347 		kfree(table);
348 err_alloc:
349 	return -ENOMEM;
350 }
351 
352 static void __net_exit xfrm6_net_sysctl_exit(struct net *net)
353 {
354 	struct ctl_table *table;
355 
356 	if (!net->ipv6.sysctl.xfrm6_hdr)
357 		return;
358 
359 	table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
360 	unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
361 	if (!net_eq(net, &init_net))
362 		kfree(table);
363 }
364 #else /* CONFIG_SYSCTL */
365 static inline int xfrm6_net_sysctl_init(struct net *net)
366 {
367 	return 0;
368 }
369 
370 static inline void xfrm6_net_sysctl_exit(struct net *net)
371 {
372 }
373 #endif
374 
375 static int __net_init xfrm6_net_init(struct net *net)
376 {
377 	int ret;
378 
379 	memcpy(&net->xfrm.xfrm6_dst_ops, &xfrm6_dst_ops_template,
380 	       sizeof(xfrm6_dst_ops_template));
381 	ret = dst_entries_init(&net->xfrm.xfrm6_dst_ops);
382 	if (ret)
383 		return ret;
384 
385 	ret = xfrm6_net_sysctl_init(net);
386 	if (ret)
387 		dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
388 
389 	return ret;
390 }
391 
392 static void __net_exit xfrm6_net_exit(struct net *net)
393 {
394 	xfrm6_net_sysctl_exit(net);
395 	dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
396 }
397 
398 static struct pernet_operations xfrm6_net_ops = {
399 	.init	= xfrm6_net_init,
400 	.exit	= xfrm6_net_exit,
401 };
402 
403 int __init xfrm6_init(void)
404 {
405 	int ret;
406 
407 	ret = xfrm6_policy_init();
408 	if (ret)
409 		goto out;
410 	ret = xfrm6_state_init();
411 	if (ret)
412 		goto out_policy;
413 
414 	ret = xfrm6_protocol_init();
415 	if (ret)
416 		goto out_state;
417 
418 	register_pernet_subsys(&xfrm6_net_ops);
419 out:
420 	return ret;
421 out_state:
422 	xfrm6_state_fini();
423 out_policy:
424 	xfrm6_policy_fini();
425 	goto out;
426 }
427 
428 void xfrm6_fini(void)
429 {
430 	unregister_pernet_subsys(&xfrm6_net_ops);
431 	xfrm6_protocol_fini();
432 	xfrm6_policy_fini();
433 	xfrm6_state_fini();
434 }
435