1
2
3
4
5
6
7
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
19static struct xfrm_policy_afinfo xfrm4_policy_afinfo;
20
21static 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{
26 struct rtable *rt;
27
28 memset(fl4, 0, sizeof(*fl4));
29 fl4->daddr = daddr->a4;
30 fl4->flowi4_tos = tos;
31 if (saddr)
32 fl4->saddr = saddr->a4;
33
34 rt = __ip_route_output_key(net, fl4);
35 if (!IS_ERR(rt))
36 return &rt->dst;
37
38 return ERR_CAST(rt);
39}
40
41static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos,
42 const xfrm_address_t *saddr,
43 const xfrm_address_t *daddr)
44{
45 struct flowi4 fl4;
46
47 return __xfrm4_dst_lookup(net, &fl4, tos, saddr, daddr);
48}
49
50static int xfrm4_get_saddr(struct net *net,
51 xfrm_address_t *saddr, xfrm_address_t *daddr)
52{
53 struct dst_entry *dst;
54 struct flowi4 fl4;
55
56 dst = __xfrm4_dst_lookup(net, &fl4, 0, NULL, daddr);
57 if (IS_ERR(dst))
58 return -EHOSTUNREACH;
59
60 saddr->a4 = fl4.saddr;
61 dst_release(dst);
62 return 0;
63}
64
65static int xfrm4_get_tos(const struct flowi *fl)
66{
67 return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos;
68}
69
70static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
71 int nfheader_len)
72{
73 return 0;
74}
75
76static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
77 const struct flowi *fl)
78{
79 struct rtable *rt = (struct rtable *)xdst->route;
80 const struct flowi4 *fl4 = &fl->u.ip4;
81
82 xdst->u.rt.rt_iif = fl4->flowi4_iif;
83
84 xdst->u.dst.dev = dev;
85 dev_hold(dev);
86
87
88
89 xdst->u.rt.rt_is_input = rt->rt_is_input;
90 xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
91 RTCF_LOCAL);
92 xdst->u.rt.rt_type = rt->rt_type;
93 xdst->u.rt.rt_gateway = rt->rt_gateway;
94 xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway;
95 xdst->u.rt.rt_pmtu = rt->rt_pmtu;
96 xdst->u.rt.rt_mtu_locked = rt->rt_mtu_locked;
97 INIT_LIST_HEAD(&xdst->u.rt.rt_uncached);
98
99 return 0;
100}
101
102static void
103_decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
104{
105 const struct iphdr *iph = ip_hdr(skb);
106 u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
107 struct flowi4 *fl4 = &fl->u.ip4;
108 int oif = 0;
109
110 if (skb_dst(skb))
111 oif = skb_dst(skb)->dev->ifindex;
112
113 memset(fl4, 0, sizeof(struct flowi4));
114 fl4->flowi4_mark = skb->mark;
115 fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
116
117 if (!ip_is_fragment(iph)) {
118 switch (iph->protocol) {
119 case IPPROTO_UDP:
120 case IPPROTO_UDPLITE:
121 case IPPROTO_TCP:
122 case IPPROTO_SCTP:
123 case IPPROTO_DCCP:
124 if (xprth + 4 < skb->data ||
125 pskb_may_pull(skb, xprth + 4 - skb->data)) {
126 __be16 *ports;
127
128 xprth = skb_network_header(skb) + iph->ihl * 4;
129 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 (xprth + 2 < skb->data ||
138 pskb_may_pull(skb, xprth + 2 - skb->data)) {
139 u8 *icmp;
140
141 xprth = skb_network_header(skb) + iph->ihl * 4;
142 icmp = xprth;
143
144 fl4->fl4_icmp_type = icmp[0];
145 fl4->fl4_icmp_code = icmp[1];
146 }
147 break;
148
149 case IPPROTO_ESP:
150 if (xprth + 4 < skb->data ||
151 pskb_may_pull(skb, xprth + 4 - skb->data)) {
152 __be32 *ehdr;
153
154 xprth = skb_network_header(skb) + iph->ihl * 4;
155 ehdr = (__be32 *)xprth;
156
157 fl4->fl4_ipsec_spi = ehdr[0];
158 }
159 break;
160
161 case IPPROTO_AH:
162 if (xprth + 8 < skb->data ||
163 pskb_may_pull(skb, xprth + 8 - skb->data)) {
164 __be32 *ah_hdr;
165
166 xprth = skb_network_header(skb) + iph->ihl * 4;
167 ah_hdr = (__be32 *)xprth;
168
169 fl4->fl4_ipsec_spi = ah_hdr[1];
170 }
171 break;
172
173 case IPPROTO_COMP:
174 if (xprth + 4 < skb->data ||
175 pskb_may_pull(skb, xprth + 4 - skb->data)) {
176 __be16 *ipcomp_hdr;
177
178 xprth = skb_network_header(skb) + iph->ihl * 4;
179 ipcomp_hdr = (__be16 *)xprth;
180
181 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
182 }
183 break;
184
185 case IPPROTO_GRE:
186 if (xprth + 12 < skb->data ||
187 pskb_may_pull(skb, xprth + 12 - skb->data)) {
188 __be16 *greflags;
189 __be32 *gre_hdr;
190
191 xprth = skb_network_header(skb) + iph->ihl * 4;
192 greflags = (__be16 *)xprth;
193 gre_hdr = (__be32 *)xprth;
194
195 if (greflags[0] & GRE_KEY) {
196 if (greflags[0] & GRE_CSUM)
197 gre_hdr++;
198 fl4->fl4_gre_key = gre_hdr[1];
199 }
200 }
201 break;
202
203 default:
204 fl4->fl4_ipsec_spi = 0;
205 break;
206 }
207 }
208 fl4->flowi4_proto = iph->protocol;
209 fl4->daddr = reverse ? iph->saddr : iph->daddr;
210 fl4->saddr = reverse ? iph->daddr : iph->saddr;
211 fl4->flowi4_tos = iph->tos;
212}
213
214static inline int xfrm4_garbage_collect(struct dst_ops *ops)
215{
216 struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
217
218 xfrm4_policy_afinfo.garbage_collect(net);
219 return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
220}
221
222static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
223 struct sk_buff *skb, u32 mtu)
224{
225 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
226 struct dst_entry *path = xdst->route;
227
228 path->ops->update_pmtu(path, sk, skb, mtu);
229}
230
231static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
232 struct sk_buff *skb)
233{
234 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
235 struct dst_entry *path = xdst->route;
236
237 path->ops->redirect(path, sk, skb);
238}
239
240static void xfrm4_dst_destroy(struct dst_entry *dst)
241{
242 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
243
244 dst_destroy_metrics_generic(dst);
245
246 xfrm_dst_destroy(xdst);
247}
248
249static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
250 int unregister)
251{
252 if (!unregister)
253 return;
254
255 xfrm_dst_ifdown(dst, dev);
256}
257
258static struct dst_ops xfrm4_dst_ops_template = {
259 .family = AF_INET,
260 .protocol = cpu_to_be16(ETH_P_IP),
261 .gc = xfrm4_garbage_collect,
262 .update_pmtu = xfrm4_update_pmtu,
263 .redirect = xfrm4_redirect,
264 .cow_metrics = dst_cow_metrics_generic,
265 .destroy = xfrm4_dst_destroy,
266 .ifdown = xfrm4_dst_ifdown,
267 .local_out = __ip_local_out,
268 .gc_thresh = 32768,
269};
270
271static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
272 .family = AF_INET,
273 .dst_ops = &xfrm4_dst_ops_template,
274 .dst_lookup = xfrm4_dst_lookup,
275 .get_saddr = xfrm4_get_saddr,
276 .decode_session = _decode_session4,
277 .get_tos = xfrm4_get_tos,
278 .init_path = xfrm4_init_path,
279 .fill_dst = xfrm4_fill_dst,
280 .blackhole_route = ipv4_blackhole_route,
281};
282
283#ifdef CONFIG_SYSCTL
284static struct ctl_table xfrm4_policy_table[] = {
285 {
286 .procname = "xfrm4_gc_thresh",
287 .data = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
288 .maxlen = sizeof(int),
289 .mode = 0644,
290 .proc_handler = proc_dointvec,
291 },
292 { }
293};
294
295static int __net_init xfrm4_net_sysctl_init(struct net *net)
296{
297 struct ctl_table *table;
298 struct ctl_table_header *hdr;
299
300 table = xfrm4_policy_table;
301 if (!net_eq(net, &init_net)) {
302 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
303 if (!table)
304 goto err_alloc;
305
306 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
307 }
308
309 hdr = register_net_sysctl(net, "net/ipv4", table);
310 if (!hdr)
311 goto err_reg;
312
313 net->ipv4.xfrm4_hdr = hdr;
314 return 0;
315
316err_reg:
317 if (!net_eq(net, &init_net))
318 kfree(table);
319err_alloc:
320 return -ENOMEM;
321}
322
323static void __net_exit xfrm4_net_sysctl_exit(struct net *net)
324{
325 struct ctl_table *table;
326
327 if (net->ipv4.xfrm4_hdr == NULL)
328 return;
329
330 table = net->ipv4.xfrm4_hdr->ctl_table_arg;
331 unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
332 if (!net_eq(net, &init_net))
333 kfree(table);
334}
335#else
336static int inline xfrm4_net_sysctl_init(struct net *net)
337{
338 return 0;
339}
340
341static void inline xfrm4_net_sysctl_exit(struct net *net)
342{
343}
344#endif
345
346static int __net_init xfrm4_net_init(struct net *net)
347{
348 int ret;
349
350 memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template,
351 sizeof(xfrm4_dst_ops_template));
352 ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops);
353 if (ret)
354 return ret;
355
356 ret = xfrm4_net_sysctl_init(net);
357 if (ret)
358 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
359
360 return ret;
361}
362
363static void __net_exit xfrm4_net_exit(struct net *net)
364{
365 xfrm4_net_sysctl_exit(net);
366 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
367}
368
369static struct pernet_operations __net_initdata xfrm4_net_ops = {
370 .init = xfrm4_net_init,
371 .exit = xfrm4_net_exit,
372};
373
374static void __init xfrm4_policy_init(void)
375{
376 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
377}
378
379void __init xfrm4_init(void)
380{
381 xfrm4_state_init();
382 xfrm4_policy_init();
383 xfrm4_protocol_init();
384 register_pernet_subsys(&xfrm4_net_ops);
385}
386
387
