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
19 static struct xfrm_policy_afinfo xfrm4_policy_afinfo;
20
__xfrm4_dst_lookup(struct net * net,struct flowi4 * fl4,int tos,const xfrm_address_t * saddr,const xfrm_address_t * daddr,u32 mark)21 static struct dst_entry *__xfrm4_dst_lookup(struct net *net, struct flowi4 *fl4,
22 int tos,
23 const xfrm_address_t *saddr,
24 const xfrm_address_t *daddr,
25 u32 mark)
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_mark = mark;
33 if (saddr)
34 fl4->saddr = saddr->a4;
35
36 rt = __ip_route_output_key(net, fl4);
37 if (!IS_ERR(rt))
38 return &rt->dst;
39
40 return ERR_CAST(rt);
41 }
42
xfrm4_dst_lookup(struct net * net,int tos,const xfrm_address_t * saddr,const xfrm_address_t * daddr,u32 mark)43 static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos,
44 const xfrm_address_t *saddr,
45 const xfrm_address_t *daddr,
46 u32 mark)
47 {
48 struct flowi4 fl4;
49
50 return __xfrm4_dst_lookup(net, &fl4, tos, saddr, daddr, mark);
51 }
52
xfrm4_get_saddr(struct net * net,xfrm_address_t * saddr,xfrm_address_t * daddr,u32 mark)53 static int xfrm4_get_saddr(struct net *net,
54 xfrm_address_t *saddr, xfrm_address_t *daddr,
55 u32 mark)
56 {
57 struct dst_entry *dst;
58 struct flowi4 fl4;
59
60 dst = __xfrm4_dst_lookup(net, &fl4, 0, NULL, daddr, mark);
61 if (IS_ERR(dst))
62 return -EHOSTUNREACH;
63
64 saddr->a4 = fl4.saddr;
65 dst_release(dst);
66 return 0;
67 }
68
xfrm4_get_tos(const struct flowi * fl)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
xfrm4_init_path(struct xfrm_dst * path,struct dst_entry * dst,int nfheader_len)74 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
75 int nfheader_len)
76 {
77 return 0;
78 }
79
xfrm4_fill_dst(struct xfrm_dst * xdst,struct net_device * dev,const struct flowi * fl)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
_decode_session4(struct sk_buff * skb,struct flowi * fl,int reverse)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 = skb_dst(skb)->dev->ifindex;
115
116 memset(fl4, 0, sizeof(struct flowi4));
117 fl4->flowi4_mark = skb->mark;
118 fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
119
120 if (!ip_is_fragment(iph)) {
121 switch (iph->protocol) {
122 case IPPROTO_UDP:
123 case IPPROTO_UDPLITE:
124 case IPPROTO_TCP:
125 case IPPROTO_SCTP:
126 case IPPROTO_DCCP:
127 if (xprth + 4 < skb->data ||
128 pskb_may_pull(skb, xprth + 4 - skb->data)) {
129 __be16 *ports = (__be16 *)xprth;
130
131 fl4->fl4_sport = ports[!!reverse];
132 fl4->fl4_dport = ports[!reverse];
133 }
134 break;
135
136 case IPPROTO_ICMP:
137 if (pskb_may_pull(skb, xprth + 2 - skb->data)) {
138 u8 *icmp = xprth;
139
140 fl4->fl4_icmp_type = icmp[0];
141 fl4->fl4_icmp_code = icmp[1];
142 }
143 break;
144
145 case IPPROTO_ESP:
146 if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
147 __be32 *ehdr = (__be32 *)xprth;
148
149 fl4->fl4_ipsec_spi = ehdr[0];
150 }
151 break;
152
153 case IPPROTO_AH:
154 if (pskb_may_pull(skb, xprth + 8 - skb->data)) {
155 __be32 *ah_hdr = (__be32 *)xprth;
156
157 fl4->fl4_ipsec_spi = ah_hdr[1];
158 }
159 break;
160
161 case IPPROTO_COMP:
162 if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
163 __be16 *ipcomp_hdr = (__be16 *)xprth;
164
165 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
166 }
167 break;
168
169 case IPPROTO_GRE:
170 if (pskb_may_pull(skb, xprth + 12 - skb->data)) {
171 __be16 *greflags = (__be16 *)xprth;
172 __be32 *gre_hdr = (__be32 *)xprth;
173
174 if (greflags[0] & GRE_KEY) {
175 if (greflags[0] & GRE_CSUM)
176 gre_hdr++;
177 fl4->fl4_gre_key = gre_hdr[1];
178 }
179 }
180 break;
181
182 default:
183 fl4->fl4_ipsec_spi = 0;
184 break;
185 }
186 }
187 fl4->flowi4_proto = iph->protocol;
188 fl4->daddr = reverse ? iph->saddr : iph->daddr;
189 fl4->saddr = reverse ? iph->daddr : iph->saddr;
190 fl4->flowi4_tos = iph->tos;
191 }
192
xfrm4_garbage_collect(struct dst_ops * ops)193 static inline int xfrm4_garbage_collect(struct dst_ops *ops)
194 {
195 struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
196
197 xfrm4_policy_afinfo.garbage_collect(net);
198 return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
199 }
200
xfrm4_update_pmtu(struct dst_entry * dst,struct sock * sk,struct sk_buff * skb,u32 mtu)201 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
202 struct sk_buff *skb, u32 mtu)
203 {
204 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
205 struct dst_entry *path = xdst->route;
206
207 path->ops->update_pmtu(path, sk, skb, mtu);
208 }
209
xfrm4_redirect(struct dst_entry * dst,struct sock * sk,struct sk_buff * skb)210 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
211 struct sk_buff *skb)
212 {
213 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
214 struct dst_entry *path = xdst->route;
215
216 path->ops->redirect(path, sk, skb);
217 }
218
xfrm4_dst_destroy(struct dst_entry * dst)219 static void xfrm4_dst_destroy(struct dst_entry *dst)
220 {
221 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
222
223 dst_destroy_metrics_generic(dst);
224
225 xfrm_dst_destroy(xdst);
226 }
227
xfrm4_dst_ifdown(struct dst_entry * dst,struct net_device * dev,int unregister)228 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
229 int unregister)
230 {
231 if (!unregister)
232 return;
233
234 xfrm_dst_ifdown(dst, dev);
235 }
236
237 static struct dst_ops xfrm4_dst_ops_template = {
238 .family = AF_INET,
239 .protocol = cpu_to_be16(ETH_P_IP),
240 .gc = xfrm4_garbage_collect,
241 .update_pmtu = xfrm4_update_pmtu,
242 .redirect = xfrm4_redirect,
243 .cow_metrics = dst_cow_metrics_generic,
244 .destroy = xfrm4_dst_destroy,
245 .ifdown = xfrm4_dst_ifdown,
246 .local_out = __ip_local_out,
247 .gc_thresh = 32768,
248 };
249
250 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
251 .family = AF_INET,
252 .dst_ops = &xfrm4_dst_ops_template,
253 .dst_lookup = xfrm4_dst_lookup,
254 .get_saddr = xfrm4_get_saddr,
255 .decode_session = _decode_session4,
256 .get_tos = xfrm4_get_tos,
257 .init_path = xfrm4_init_path,
258 .fill_dst = xfrm4_fill_dst,
259 .blackhole_route = ipv4_blackhole_route,
260 };
261
262 #ifdef CONFIG_SYSCTL
263 static struct ctl_table xfrm4_policy_table[] = {
264 {
265 .procname = "xfrm4_gc_thresh",
266 .data = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
267 .maxlen = sizeof(int),
268 .mode = 0644,
269 .proc_handler = proc_dointvec,
270 },
271 { }
272 };
273
xfrm4_net_sysctl_init(struct net * net)274 static int __net_init xfrm4_net_sysctl_init(struct net *net)
275 {
276 struct ctl_table *table;
277 struct ctl_table_header *hdr;
278
279 table = xfrm4_policy_table;
280 if (!net_eq(net, &init_net)) {
281 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
282 if (!table)
283 goto err_alloc;
284
285 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
286 }
287
288 hdr = register_net_sysctl(net, "net/ipv4", table);
289 if (!hdr)
290 goto err_reg;
291
292 net->ipv4.xfrm4_hdr = hdr;
293 return 0;
294
295 err_reg:
296 if (!net_eq(net, &init_net))
297 kfree(table);
298 err_alloc:
299 return -ENOMEM;
300 }
301
xfrm4_net_sysctl_exit(struct net * net)302 static void __net_exit xfrm4_net_sysctl_exit(struct net *net)
303 {
304 struct ctl_table *table;
305
306 if (net->ipv4.xfrm4_hdr == NULL)
307 return;
308
309 table = net->ipv4.xfrm4_hdr->ctl_table_arg;
310 unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
311 if (!net_eq(net, &init_net))
312 kfree(table);
313 }
314 #else /* CONFIG_SYSCTL */
xfrm4_net_sysctl_init(struct net * net)315 static int inline xfrm4_net_sysctl_init(struct net *net)
316 {
317 return 0;
318 }
319
xfrm4_net_sysctl_exit(struct net * net)320 static void inline xfrm4_net_sysctl_exit(struct net *net)
321 {
322 }
323 #endif
324
xfrm4_net_init(struct net * net)325 static int __net_init xfrm4_net_init(struct net *net)
326 {
327 int ret;
328
329 memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template,
330 sizeof(xfrm4_dst_ops_template));
331 ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops);
332 if (ret)
333 return ret;
334
335 ret = xfrm4_net_sysctl_init(net);
336 if (ret)
337 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
338
339 return ret;
340 }
341
xfrm4_net_exit(struct net * net)342 static void __net_exit xfrm4_net_exit(struct net *net)
343 {
344 xfrm4_net_sysctl_exit(net);
345 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
346 }
347
348 static struct pernet_operations __net_initdata xfrm4_net_ops = {
349 .init = xfrm4_net_init,
350 .exit = xfrm4_net_exit,
351 };
352
xfrm4_policy_init(void)353 static void __init xfrm4_policy_init(void)
354 {
355 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
356 }
357
xfrm4_init(void)358 void __init xfrm4_init(void)
359 {
360 xfrm4_state_init();
361 xfrm4_policy_init();
362 xfrm4_protocol_init();
363 register_pernet_subsys(&xfrm4_net_ops);
364 }
365
366