xref: /openbmc/linux/net/ipv6/xfrm6_policy.c (revision b24413180f5600bcb3bb70fbed5cf186b60864bd)
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 
117 	return 0;
118 }
119 
120 static inline void
121 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
122 {
123 	struct flowi6 *fl6 = &fl->u.ip6;
124 	int onlyproto = 0;
125 	const struct ipv6hdr *hdr = ipv6_hdr(skb);
126 	u16 offset = sizeof(*hdr);
127 	struct ipv6_opt_hdr *exthdr;
128 	const unsigned char *nh = skb_network_header(skb);
129 	u16 nhoff = IP6CB(skb)->nhoff;
130 	int oif = 0;
131 	u8 nexthdr;
132 
133 	if (!nhoff)
134 		nhoff = offsetof(struct ipv6hdr, nexthdr);
135 
136 	nexthdr = nh[nhoff];
137 
138 	if (skb_dst(skb))
139 		oif = skb_dst(skb)->dev->ifindex;
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 void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
222 			      struct sk_buff *skb, u32 mtu)
223 {
224 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
225 	struct dst_entry *path = xdst->route;
226 
227 	path->ops->update_pmtu(path, sk, skb, mtu);
228 }
229 
230 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
231 			   struct sk_buff *skb)
232 {
233 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
234 	struct dst_entry *path = xdst->route;
235 
236 	path->ops->redirect(path, sk, skb);
237 }
238 
239 static void xfrm6_dst_destroy(struct dst_entry *dst)
240 {
241 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
242 
243 	if (likely(xdst->u.rt6.rt6i_idev))
244 		in6_dev_put(xdst->u.rt6.rt6i_idev);
245 	dst_destroy_metrics_generic(dst);
246 	xfrm_dst_destroy(xdst);
247 }
248 
249 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
250 			     int unregister)
251 {
252 	struct xfrm_dst *xdst;
253 
254 	if (!unregister)
255 		return;
256 
257 	xdst = (struct xfrm_dst *)dst;
258 	if (xdst->u.rt6.rt6i_idev->dev == dev) {
259 		struct inet6_dev *loopback_idev =
260 			in6_dev_get(dev_net(dev)->loopback_dev);
261 		BUG_ON(!loopback_idev);
262 
263 		do {
264 			in6_dev_put(xdst->u.rt6.rt6i_idev);
265 			xdst->u.rt6.rt6i_idev = loopback_idev;
266 			in6_dev_hold(loopback_idev);
267 			xdst = (struct xfrm_dst *)xdst->u.dst.child;
268 		} while (xdst->u.dst.xfrm);
269 
270 		__in6_dev_put(loopback_idev);
271 	}
272 
273 	xfrm_dst_ifdown(dst, dev);
274 }
275 
276 static struct dst_ops xfrm6_dst_ops_template = {
277 	.family =		AF_INET6,
278 	.update_pmtu =		xfrm6_update_pmtu,
279 	.redirect =		xfrm6_redirect,
280 	.cow_metrics =		dst_cow_metrics_generic,
281 	.destroy =		xfrm6_dst_destroy,
282 	.ifdown =		xfrm6_dst_ifdown,
283 	.local_out =		__ip6_local_out,
284 	.gc_thresh =		32768,
285 };
286 
287 static const struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
288 	.dst_ops =		&xfrm6_dst_ops_template,
289 	.dst_lookup =		xfrm6_dst_lookup,
290 	.get_saddr =		xfrm6_get_saddr,
291 	.decode_session =	_decode_session6,
292 	.get_tos =		xfrm6_get_tos,
293 	.init_path =		xfrm6_init_path,
294 	.fill_dst =		xfrm6_fill_dst,
295 	.blackhole_route =	ip6_blackhole_route,
296 };
297 
298 static int __init xfrm6_policy_init(void)
299 {
300 	return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo, AF_INET6);
301 }
302 
303 static void xfrm6_policy_fini(void)
304 {
305 	xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
306 }
307 
308 #ifdef CONFIG_SYSCTL
309 static struct ctl_table xfrm6_policy_table[] = {
310 	{
311 		.procname       = "xfrm6_gc_thresh",
312 		.data		= &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
313 		.maxlen		= sizeof(int),
314 		.mode		= 0644,
315 		.proc_handler   = proc_dointvec,
316 	},
317 	{ }
318 };
319 
320 static int __net_init xfrm6_net_sysctl_init(struct net *net)
321 {
322 	struct ctl_table *table;
323 	struct ctl_table_header *hdr;
324 
325 	table = xfrm6_policy_table;
326 	if (!net_eq(net, &init_net)) {
327 		table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
328 		if (!table)
329 			goto err_alloc;
330 
331 		table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
332 	}
333 
334 	hdr = register_net_sysctl(net, "net/ipv6", table);
335 	if (!hdr)
336 		goto err_reg;
337 
338 	net->ipv6.sysctl.xfrm6_hdr = hdr;
339 	return 0;
340 
341 err_reg:
342 	if (!net_eq(net, &init_net))
343 		kfree(table);
344 err_alloc:
345 	return -ENOMEM;
346 }
347 
348 static void __net_exit xfrm6_net_sysctl_exit(struct net *net)
349 {
350 	struct ctl_table *table;
351 
352 	if (!net->ipv6.sysctl.xfrm6_hdr)
353 		return;
354 
355 	table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
356 	unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
357 	if (!net_eq(net, &init_net))
358 		kfree(table);
359 }
360 #else /* CONFIG_SYSCTL */
361 static inline int xfrm6_net_sysctl_init(struct net *net)
362 {
363 	return 0;
364 }
365 
366 static inline void xfrm6_net_sysctl_exit(struct net *net)
367 {
368 }
369 #endif
370 
371 static int __net_init xfrm6_net_init(struct net *net)
372 {
373 	int ret;
374 
375 	memcpy(&net->xfrm.xfrm6_dst_ops, &xfrm6_dst_ops_template,
376 	       sizeof(xfrm6_dst_ops_template));
377 	ret = dst_entries_init(&net->xfrm.xfrm6_dst_ops);
378 	if (ret)
379 		return ret;
380 
381 	ret = xfrm6_net_sysctl_init(net);
382 	if (ret)
383 		dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
384 
385 	return ret;
386 }
387 
388 static void __net_exit xfrm6_net_exit(struct net *net)
389 {
390 	xfrm6_net_sysctl_exit(net);
391 	dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
392 }
393 
394 static struct pernet_operations xfrm6_net_ops = {
395 	.init	= xfrm6_net_init,
396 	.exit	= xfrm6_net_exit,
397 };
398 
399 int __init xfrm6_init(void)
400 {
401 	int ret;
402 
403 	ret = xfrm6_policy_init();
404 	if (ret)
405 		goto out;
406 	ret = xfrm6_state_init();
407 	if (ret)
408 		goto out_policy;
409 
410 	ret = xfrm6_protocol_init();
411 	if (ret)
412 		goto out_state;
413 
414 	register_pernet_subsys(&xfrm6_net_ops);
415 out:
416 	return ret;
417 out_state:
418 	xfrm6_state_fini();
419 out_policy:
420 	xfrm6_policy_fini();
421 	goto out;
422 }
423 
424 void xfrm6_fini(void)
425 {
426 	unregister_pernet_subsys(&xfrm6_net_ops);
427 	xfrm6_protocol_fini();
428 	xfrm6_policy_fini();
429 	xfrm6_state_fini();
430 }
431