xref: /openbmc/linux/net/ipv4/xfrm4_policy.c (revision b96fc2f3)
1 /*
2  * xfrm4_policy.c
3  *
4  * Changes:
5  *	Kazunori MIYAZAWA @USAGI
6  * 	YOSHIFUJI Hideaki @USAGI
7  *		Split up af-specific portion
8  *
9  */
10 
11 #include <linux/err.h>
12 #include <linux/kernel.h>
13 #include <linux/inetdevice.h>
14 #include <linux/if_tunnel.h>
15 #include <net/dst.h>
16 #include <net/xfrm.h>
17 #include <net/ip.h>
18 #include <net/vrf.h>
19 
20 static struct xfrm_policy_afinfo xfrm4_policy_afinfo;
21 
22 static struct dst_entry *__xfrm4_dst_lookup(struct net *net, struct flowi4 *fl4,
23 					    int tos, int oif,
24 					    const xfrm_address_t *saddr,
25 					    const xfrm_address_t *daddr)
26 {
27 	struct rtable *rt;
28 
29 	memset(fl4, 0, sizeof(*fl4));
30 	fl4->daddr = daddr->a4;
31 	fl4->flowi4_tos = tos;
32 	fl4->flowi4_oif = oif;
33 	if (saddr)
34 		fl4->saddr = saddr->a4;
35 
36 	fl4->flowi4_flags = FLOWI_FLAG_SKIP_NH_OIF;
37 
38 	rt = __ip_route_output_key(net, fl4);
39 	if (!IS_ERR(rt))
40 		return &rt->dst;
41 
42 	return ERR_CAST(rt);
43 }
44 
45 static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos, int oif,
46 					  const xfrm_address_t *saddr,
47 					  const xfrm_address_t *daddr)
48 {
49 	struct flowi4 fl4;
50 
51 	return __xfrm4_dst_lookup(net, &fl4, tos, oif, saddr, daddr);
52 }
53 
54 static int xfrm4_get_saddr(struct net *net, int oif,
55 			   xfrm_address_t *saddr, xfrm_address_t *daddr)
56 {
57 	struct dst_entry *dst;
58 	struct flowi4 fl4;
59 
60 	dst = __xfrm4_dst_lookup(net, &fl4, 0, oif, NULL, daddr);
61 	if (IS_ERR(dst))
62 		return -EHOSTUNREACH;
63 
64 	saddr->a4 = fl4.saddr;
65 	dst_release(dst);
66 	return 0;
67 }
68 
69 static int xfrm4_get_tos(const struct flowi *fl)
70 {
71 	return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */
72 }
73 
74 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
75 			   int nfheader_len)
76 {
77 	return 0;
78 }
79 
80 static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
81 			  const struct flowi *fl)
82 {
83 	struct rtable *rt = (struct rtable *)xdst->route;
84 	const struct flowi4 *fl4 = &fl->u.ip4;
85 
86 	xdst->u.rt.rt_iif = fl4->flowi4_iif;
87 
88 	xdst->u.dst.dev = dev;
89 	dev_hold(dev);
90 
91 	/* Sheit... I remember I did this right. Apparently,
92 	 * it was magically lost, so this code needs audit */
93 	xdst->u.rt.rt_is_input = rt->rt_is_input;
94 	xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
95 					      RTCF_LOCAL);
96 	xdst->u.rt.rt_type = rt->rt_type;
97 	xdst->u.rt.rt_gateway = rt->rt_gateway;
98 	xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway;
99 	xdst->u.rt.rt_pmtu = rt->rt_pmtu;
100 	INIT_LIST_HEAD(&xdst->u.rt.rt_uncached);
101 
102 	return 0;
103 }
104 
105 static void
106 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
107 {
108 	const struct iphdr *iph = ip_hdr(skb);
109 	u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
110 	struct flowi4 *fl4 = &fl->u.ip4;
111 	int oif = 0;
112 
113 	if (skb_dst(skb)) {
114 		oif = vrf_master_ifindex(skb_dst(skb)->dev) ?
115 			: skb_dst(skb)->dev->ifindex;
116 	}
117 
118 	memset(fl4, 0, sizeof(struct flowi4));
119 	fl4->flowi4_mark = skb->mark;
120 	fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
121 
122 	if (!ip_is_fragment(iph)) {
123 		switch (iph->protocol) {
124 		case IPPROTO_UDP:
125 		case IPPROTO_UDPLITE:
126 		case IPPROTO_TCP:
127 		case IPPROTO_SCTP:
128 		case IPPROTO_DCCP:
129 			if (xprth + 4 < skb->data ||
130 			    pskb_may_pull(skb, xprth + 4 - skb->data)) {
131 				__be16 *ports = (__be16 *)xprth;
132 
133 				fl4->fl4_sport = ports[!!reverse];
134 				fl4->fl4_dport = ports[!reverse];
135 			}
136 			break;
137 
138 		case IPPROTO_ICMP:
139 			if (pskb_may_pull(skb, xprth + 2 - skb->data)) {
140 				u8 *icmp = xprth;
141 
142 				fl4->fl4_icmp_type = icmp[0];
143 				fl4->fl4_icmp_code = icmp[1];
144 			}
145 			break;
146 
147 		case IPPROTO_ESP:
148 			if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
149 				__be32 *ehdr = (__be32 *)xprth;
150 
151 				fl4->fl4_ipsec_spi = ehdr[0];
152 			}
153 			break;
154 
155 		case IPPROTO_AH:
156 			if (pskb_may_pull(skb, xprth + 8 - skb->data)) {
157 				__be32 *ah_hdr = (__be32 *)xprth;
158 
159 				fl4->fl4_ipsec_spi = ah_hdr[1];
160 			}
161 			break;
162 
163 		case IPPROTO_COMP:
164 			if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
165 				__be16 *ipcomp_hdr = (__be16 *)xprth;
166 
167 				fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
168 			}
169 			break;
170 
171 		case IPPROTO_GRE:
172 			if (pskb_may_pull(skb, xprth + 12 - skb->data)) {
173 				__be16 *greflags = (__be16 *)xprth;
174 				__be32 *gre_hdr = (__be32 *)xprth;
175 
176 				if (greflags[0] & GRE_KEY) {
177 					if (greflags[0] & GRE_CSUM)
178 						gre_hdr++;
179 					fl4->fl4_gre_key = gre_hdr[1];
180 				}
181 			}
182 			break;
183 
184 		default:
185 			fl4->fl4_ipsec_spi = 0;
186 			break;
187 		}
188 	}
189 	fl4->flowi4_proto = iph->protocol;
190 	fl4->daddr = reverse ? iph->saddr : iph->daddr;
191 	fl4->saddr = reverse ? iph->daddr : iph->saddr;
192 	fl4->flowi4_tos = iph->tos;
193 }
194 
195 static inline int xfrm4_garbage_collect(struct dst_ops *ops)
196 {
197 	struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
198 
199 	xfrm4_policy_afinfo.garbage_collect(net);
200 	return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
201 }
202 
203 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
204 			      struct sk_buff *skb, u32 mtu)
205 {
206 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
207 	struct dst_entry *path = xdst->route;
208 
209 	path->ops->update_pmtu(path, sk, skb, mtu);
210 }
211 
212 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
213 			   struct sk_buff *skb)
214 {
215 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
216 	struct dst_entry *path = xdst->route;
217 
218 	path->ops->redirect(path, sk, skb);
219 }
220 
221 static void xfrm4_dst_destroy(struct dst_entry *dst)
222 {
223 	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
224 
225 	dst_destroy_metrics_generic(dst);
226 
227 	xfrm_dst_destroy(xdst);
228 }
229 
230 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
231 			     int unregister)
232 {
233 	if (!unregister)
234 		return;
235 
236 	xfrm_dst_ifdown(dst, dev);
237 }
238 
239 static struct dst_ops xfrm4_dst_ops = {
240 	.family =		AF_INET,
241 	.gc =			xfrm4_garbage_collect,
242 	.update_pmtu =		xfrm4_update_pmtu,
243 	.redirect =		xfrm4_redirect,
244 	.cow_metrics =		dst_cow_metrics_generic,
245 	.destroy =		xfrm4_dst_destroy,
246 	.ifdown =		xfrm4_dst_ifdown,
247 	.local_out =		__ip_local_out,
248 	.gc_thresh =		32768,
249 };
250 
251 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
252 	.family = 		AF_INET,
253 	.dst_ops =		&xfrm4_dst_ops,
254 	.dst_lookup =		xfrm4_dst_lookup,
255 	.get_saddr =		xfrm4_get_saddr,
256 	.decode_session =	_decode_session4,
257 	.get_tos =		xfrm4_get_tos,
258 	.init_path =		xfrm4_init_path,
259 	.fill_dst =		xfrm4_fill_dst,
260 	.blackhole_route =	ipv4_blackhole_route,
261 };
262 
263 #ifdef CONFIG_SYSCTL
264 static struct ctl_table xfrm4_policy_table[] = {
265 	{
266 		.procname       = "xfrm4_gc_thresh",
267 		.data           = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
268 		.maxlen         = sizeof(int),
269 		.mode           = 0644,
270 		.proc_handler   = proc_dointvec,
271 	},
272 	{ }
273 };
274 
275 static int __net_init xfrm4_net_init(struct net *net)
276 {
277 	struct ctl_table *table;
278 	struct ctl_table_header *hdr;
279 
280 	table = xfrm4_policy_table;
281 	if (!net_eq(net, &init_net)) {
282 		table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
283 		if (!table)
284 			goto err_alloc;
285 
286 		table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
287 	}
288 
289 	hdr = register_net_sysctl(net, "net/ipv4", table);
290 	if (!hdr)
291 		goto err_reg;
292 
293 	net->ipv4.xfrm4_hdr = hdr;
294 	return 0;
295 
296 err_reg:
297 	if (!net_eq(net, &init_net))
298 		kfree(table);
299 err_alloc:
300 	return -ENOMEM;
301 }
302 
303 static void __net_exit xfrm4_net_exit(struct net *net)
304 {
305 	struct ctl_table *table;
306 
307 	if (!net->ipv4.xfrm4_hdr)
308 		return;
309 
310 	table = net->ipv4.xfrm4_hdr->ctl_table_arg;
311 	unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
312 	if (!net_eq(net, &init_net))
313 		kfree(table);
314 }
315 
316 static struct pernet_operations __net_initdata xfrm4_net_ops = {
317 	.init	= xfrm4_net_init,
318 	.exit	= xfrm4_net_exit,
319 };
320 #endif
321 
322 static void __init xfrm4_policy_init(void)
323 {
324 	xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
325 }
326 
327 void __init xfrm4_init(void)
328 {
329 	dst_entries_init(&xfrm4_dst_ops);
330 
331 	xfrm4_state_init();
332 	xfrm4_policy_init();
333 	xfrm4_protocol_init();
334 #ifdef CONFIG_SYSCTL
335 	register_pernet_subsys(&xfrm4_net_ops);
336 #endif
337 }
338 
339