1// SPDX-License-Identifier: GPL-2.0
2/*
3 * xfrm4_policy.c
4 *
5 * Changes:
6 * Kazunori MIYAZAWA @USAGI
7 * YOSHIFUJI Hideaki @USAGI
8 * Split up af-specific portion
9 *
10 */
11
12#include <linux/err.h>
13#include <linux/kernel.h>
14#include <linux/inetdevice.h>
15#include <linux/if_tunnel.h>
16#include <net/dst.h>
17#include <net/xfrm.h>
18#include <net/ip.h>
19#include <net/l3mdev.h>
20
21static struct dst_entry *__xfrm4_dst_lookup(struct net *net, struct flowi4 *fl4,
22 int tos, int oif,
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_oif = l3mdev_master_ifindex_by_index(net, oif);
33 fl4->flowi4_mark = mark;
34 if (saddr)
35 fl4->saddr = saddr->a4;
36
37 fl4->flowi4_flags = FLOWI_FLAG_SKIP_NH_OIF;
38
39 rt = __ip_route_output_key(net, fl4);
40 if (!IS_ERR(rt))
41 return &rt->dst;
42
43 return ERR_CAST(rt);
44}
45
46static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos, int oif,
47 const xfrm_address_t *saddr,
48 const xfrm_address_t *daddr,
49 u32 mark)
50{
51 struct flowi4 fl4;
52
53 return __xfrm4_dst_lookup(net, &fl4, tos, oif, saddr, daddr, mark);
54}
55
56static int xfrm4_get_saddr(struct net *net, int oif,
57 xfrm_address_t *saddr, xfrm_address_t *daddr,
58 u32 mark)
59{
60 struct dst_entry *dst;
61 struct flowi4 fl4;
62
63 dst = __xfrm4_dst_lookup(net, &fl4, 0, oif, NULL, daddr, mark);
64 if (IS_ERR(dst))
65 return -EHOSTUNREACH;
66
67 saddr->a4 = fl4.saddr;
68 dst_release(dst);
69 return 0;
70}
71
72static int xfrm4_get_tos(const struct flowi *fl)
73{
74 return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */
75}
76
77static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
78 int nfheader_len)
79{
80 return 0;
81}
82
83static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
84 const struct flowi *fl)
85{
86 struct rtable *rt = (struct rtable *)xdst->route;
87 const struct flowi4 *fl4 = &fl->u.ip4;
88
89 xdst->u.rt.rt_iif = fl4->flowi4_iif;
90
91 xdst->u.dst.dev = dev;
92 dev_hold(dev);
93
94 /* Sheit... I remember I did this right. Apparently,
95 * it was magically lost, so this code needs audit */
96 xdst->u.rt.rt_is_input = rt->rt_is_input;
97 xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
98 RTCF_LOCAL);
99 xdst->u.rt.rt_type = rt->rt_type;
100 xdst->u.rt.rt_gateway = rt->rt_gateway;
101 xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway;
102 xdst->u.rt.rt_pmtu = rt->rt_pmtu;
103 xdst->u.rt.rt_mtu_locked = rt->rt_mtu_locked;
104 INIT_LIST_HEAD(&xdst->u.rt.rt_uncached);
105 rt_add_uncached_list(&xdst->u.rt);
106
107 return 0;
108}
109
110static void
111_decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
112{
113 const struct iphdr *iph = ip_hdr(skb);
114 u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
115 struct flowi4 *fl4 = &fl->u.ip4;
116 int oif = 0;
117
118 if (skb_dst(skb))
119 oif = skb_dst(skb)->dev->ifindex;
120
121 memset(fl4, 0, sizeof(struct flowi4));
122 fl4->flowi4_mark = skb->mark;
123 fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
124
125 if (!ip_is_fragment(iph)) {
126 switch (iph->protocol) {
127 case IPPROTO_UDP:
128 case IPPROTO_UDPLITE:
129 case IPPROTO_TCP:
130 case IPPROTO_SCTP:
131 case IPPROTO_DCCP:
132 if (xprth + 4 < skb->data ||
133 pskb_may_pull(skb, xprth + 4 - skb->data)) {
134 __be16 *ports;
135
136 xprth = skb_network_header(skb) + iph->ihl * 4;
137 ports = (__be16 *)xprth;
138
139 fl4->fl4_sport = ports[!!reverse];
140 fl4->fl4_dport = ports[!reverse];
141 }
142 break;
143
144 case IPPROTO_ICMP:
145 if (xprth + 2 < skb->data ||
146 pskb_may_pull(skb, xprth + 2 - skb->data)) {
147 u8 *icmp;
148
149 xprth = skb_network_header(skb) + iph->ihl * 4;
150 icmp = xprth;
151
152 fl4->fl4_icmp_type = icmp[0];
153 fl4->fl4_icmp_code = icmp[1];
154 }
155 break;
156
157 case IPPROTO_ESP:
158 if (xprth + 4 < skb->data ||
159 pskb_may_pull(skb, xprth + 4 - skb->data)) {
160 __be32 *ehdr;
161
162 xprth = skb_network_header(skb) + iph->ihl * 4;
163 ehdr = (__be32 *)xprth;
164
165 fl4->fl4_ipsec_spi = ehdr[0];
166 }
167 break;
168
169 case IPPROTO_AH:
170 if (xprth + 8 < skb->data ||
171 pskb_may_pull(skb, xprth + 8 - skb->data)) {
172 __be32 *ah_hdr;
173
174 xprth = skb_network_header(skb) + iph->ihl * 4;
175 ah_hdr = (__be32 *)xprth;
176
177 fl4->fl4_ipsec_spi = ah_hdr[1];
178 }
179 break;
180
181 case IPPROTO_COMP:
182 if (xprth + 4 < skb->data ||
183 pskb_may_pull(skb, xprth + 4 - skb->data)) {
184 __be16 *ipcomp_hdr;
185
186 xprth = skb_network_header(skb) + iph->ihl * 4;
187 ipcomp_hdr = (__be16 *)xprth;
188
189 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
190 }
191 break;
192
193 case IPPROTO_GRE:
194 if (xprth + 12 < skb->data ||
195 pskb_may_pull(skb, xprth + 12 - skb->data)) {
196 __be16 *greflags;
197 __be32 *gre_hdr;
198
199 xprth = skb_network_header(skb) + iph->ihl * 4;
200 greflags = (__be16 *)xprth;
201 gre_hdr = (__be32 *)xprth;
202
203 if (greflags[0] & GRE_KEY) {
204 if (greflags[0] & GRE_CSUM)
205 gre_hdr++;
206 fl4->fl4_gre_key = gre_hdr[1];
207 }
208 }
209 break;
210
211 default:
212 fl4->fl4_ipsec_spi = 0;
213 break;
214 }
215 }
216 fl4->flowi4_proto = iph->protocol;
217 fl4->daddr = reverse ? iph->saddr : iph->daddr;
218 fl4->saddr = reverse ? iph->daddr : iph->saddr;
219 fl4->flowi4_tos = iph->tos;
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 if (xdst->u.rt.rt_uncached_list)
246 rt_del_uncached_list(&xdst->u.rt);
247 xfrm_dst_destroy(xdst);
248}
249
250static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
251 int unregister)
252{
253 if (!unregister)
254 return;
255
256 xfrm_dst_ifdown(dst, dev);
257}
258
259static struct dst_ops xfrm4_dst_ops_template = {
260 .family = AF_INET,
261 .update_pmtu = xfrm4_update_pmtu,
262 .redirect = xfrm4_redirect,
263 .cow_metrics = dst_cow_metrics_generic,
264 .destroy = xfrm4_dst_destroy,
265 .ifdown = xfrm4_dst_ifdown,
266 .local_out = __ip_local_out,
267 .gc_thresh = 32768,
268};
269
270static const struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
271 .dst_ops = &xfrm4_dst_ops_template,
272 .dst_lookup = xfrm4_dst_lookup,
273 .get_saddr = xfrm4_get_saddr,
274 .decode_session = _decode_session4,
275 .get_tos = xfrm4_get_tos,
276 .init_path = xfrm4_init_path,
277 .fill_dst = xfrm4_fill_dst,
278 .blackhole_route = ipv4_blackhole_route,
279};
280
281#ifdef CONFIG_SYSCTL
282static struct ctl_table xfrm4_policy_table[] = {
283 {
284 .procname = "xfrm4_gc_thresh",
285 .data = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
286 .maxlen = sizeof(int),
287 .mode = 0644,
288 .proc_handler = proc_dointvec,
289 },
290 { }
291};
292
293static __net_init int xfrm4_net_sysctl_init(struct net *net)
294{
295 struct ctl_table *table;
296 struct ctl_table_header *hdr;
297
298 table = xfrm4_policy_table;
299 if (!net_eq(net, &init_net)) {
300 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
301 if (!table)
302 goto err_alloc;
303
304 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
305 }
306
307 hdr = register_net_sysctl(net, "net/ipv4", table);
308 if (!hdr)
309 goto err_reg;
310
311 net->ipv4.xfrm4_hdr = hdr;
312 return 0;
313
314err_reg:
315 if (!net_eq(net, &init_net))
316 kfree(table);
317err_alloc:
318 return -ENOMEM;
319}
320
321static __net_exit void xfrm4_net_sysctl_exit(struct net *net)
322{
323 struct ctl_table *table;
324
325 if (!net->ipv4.xfrm4_hdr)
326 return;
327
328 table = net->ipv4.xfrm4_hdr->ctl_table_arg;
329 unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
330 if (!net_eq(net, &init_net))
331 kfree(table);
332}
333#else /* CONFIG_SYSCTL */
334static inline int xfrm4_net_sysctl_init(struct net *net)
335{
336 return 0;
337}
338
339static inline void xfrm4_net_sysctl_exit(struct net *net)
340{
341}
342#endif
343
344static int __net_init xfrm4_net_init(struct net *net)
345{
346 int ret;
347
348 memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template,
349 sizeof(xfrm4_dst_ops_template));
350 ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops);
351 if (ret)
352 return ret;
353
354 ret = xfrm4_net_sysctl_init(net);
355 if (ret)
356 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
357
358 return ret;
359}
360
361static void __net_exit xfrm4_net_exit(struct net *net)
362{
363 xfrm4_net_sysctl_exit(net);
364 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
365}
366
367static struct pernet_operations __net_initdata xfrm4_net_ops = {
368 .init = xfrm4_net_init,
369 .exit = xfrm4_net_exit,
370};
371
372static void __init xfrm4_policy_init(void)
373{
374 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo, AF_INET);
375}
376
377void __init xfrm4_init(void)
378{
379 xfrm4_state_init();
380 xfrm4_policy_init();
381 xfrm4_protocol_init();
382 register_pernet_subsys(&xfrm4_net_ops);
383}
384