1 /*
2 * xfrm6_policy.c: based on xfrm4_policy.c
3 *
4 * Authors:
5 * Mitsuru KANDA @USAGI
6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
8 * IPv6 support
9 * YOSHIFUJI Hideaki
10 * Split up af-specific portion
11 *
12 */
13
14 #include <linux/err.h>
15 #include <linux/kernel.h>
16 #include <linux/netdevice.h>
17 #include <net/addrconf.h>
18 #include <net/dst.h>
19 #include <net/xfrm.h>
20 #include <net/ip.h>
21 #include <net/ipv6.h>
22 #include <net/ip6_route.h>
23 #include <net/l3mdev.h>
24 #if IS_ENABLED(CONFIG_IPV6_MIP6)
25 #include <net/mip6.h>
26 #endif
27
28 static struct xfrm_policy_afinfo xfrm6_policy_afinfo;
29
xfrm6_dst_lookup(struct net * net,int tos,int oif,const xfrm_address_t * saddr,const xfrm_address_t * daddr,u32 mark)30 static struct dst_entry *xfrm6_dst_lookup(struct net *net, int tos, int oif,
31 const xfrm_address_t *saddr,
32 const xfrm_address_t *daddr,
33 u32 mark)
34 {
35 struct flowi6 fl6;
36 struct dst_entry *dst;
37 int err;
38
39 memset(&fl6, 0, sizeof(fl6));
40 fl6.flowi6_oif = l3mdev_master_ifindex_by_index(net, oif);
41 fl6.flowi6_flags = FLOWI_FLAG_SKIP_NH_OIF;
42 fl6.flowi6_mark = mark;
43 memcpy(&fl6.daddr, daddr, sizeof(fl6.daddr));
44 if (saddr)
45 memcpy(&fl6.saddr, saddr, sizeof(fl6.saddr));
46
47 dst = ip6_route_output(net, NULL, &fl6);
48
49 err = dst->error;
50 if (dst->error) {
51 dst_release(dst);
52 dst = ERR_PTR(err);
53 }
54
55 return dst;
56 }
57
xfrm6_get_saddr(struct net * net,int oif,xfrm_address_t * saddr,xfrm_address_t * daddr,u32 mark)58 static int xfrm6_get_saddr(struct net *net, int oif,
59 xfrm_address_t *saddr, xfrm_address_t *daddr,
60 u32 mark)
61 {
62 struct dst_entry *dst;
63 struct net_device *dev;
64
65 dst = xfrm6_dst_lookup(net, 0, oif, NULL, daddr, mark);
66 if (IS_ERR(dst))
67 return -EHOSTUNREACH;
68
69 dev = ip6_dst_idev(dst)->dev;
70 ipv6_dev_get_saddr(dev_net(dev), dev, &daddr->in6, 0, &saddr->in6);
71 dst_release(dst);
72 return 0;
73 }
74
xfrm6_get_tos(const struct flowi * fl)75 static int xfrm6_get_tos(const struct flowi *fl)
76 {
77 return 0;
78 }
79
xfrm6_init_path(struct xfrm_dst * path,struct dst_entry * dst,int nfheader_len)80 static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
81 int nfheader_len)
82 {
83 if (dst->ops->family == AF_INET6) {
84 struct rt6_info *rt = (struct rt6_info *)dst;
85 path->path_cookie = rt6_get_cookie(rt);
86 }
87
88 path->u.rt6.rt6i_nfheader_len = nfheader_len;
89
90 return 0;
91 }
92
xfrm6_fill_dst(struct xfrm_dst * xdst,struct net_device * dev,const struct flowi * fl)93 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
94 const struct flowi *fl)
95 {
96 struct rt6_info *rt = (struct rt6_info *)xdst->route;
97
98 xdst->u.dst.dev = dev;
99 dev_hold(dev);
100
101 xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
102 if (!xdst->u.rt6.rt6i_idev) {
103 dev_put(dev);
104 return -ENODEV;
105 }
106
107 /* Sheit... I remember I did this right. Apparently,
108 * it was magically lost, so this code needs audit */
109 xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
110 RTF_LOCAL);
111 xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
112 xdst->u.rt6.rt6i_node = rt->rt6i_node;
113 xdst->route_cookie = rt6_get_cookie(rt);
114 xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
115 xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
116 xdst->u.rt6.rt6i_src = rt->rt6i_src;
117
118 return 0;
119 }
120
121 static inline void
_decode_session6(struct sk_buff * skb,struct flowi * fl,int reverse)122 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
123 {
124 struct flowi6 *fl6 = &fl->u.ip6;
125 int onlyproto = 0;
126 const struct ipv6hdr *hdr = ipv6_hdr(skb);
127 u16 offset = sizeof(*hdr);
128 struct ipv6_opt_hdr *exthdr;
129 const unsigned char *nh = skb_network_header(skb);
130 u16 nhoff = IP6CB(skb)->nhoff;
131 int oif = 0;
132 u8 nexthdr;
133
134 if (!nhoff)
135 nhoff = offsetof(struct ipv6hdr, nexthdr);
136
137 nexthdr = nh[nhoff];
138
139 if (skb_dst(skb))
140 oif = skb_dst(skb)->dev->ifindex;
141
142 memset(fl6, 0, sizeof(struct flowi6));
143 fl6->flowi6_mark = skb->mark;
144 fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
145
146 fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
147 fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
148
149 while (nh + offset + 1 < skb->data ||
150 pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
151 nh = skb_network_header(skb);
152 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
153
154 switch (nexthdr) {
155 case NEXTHDR_FRAGMENT:
156 onlyproto = 1;
157 case NEXTHDR_ROUTING:
158 case NEXTHDR_HOP:
159 case NEXTHDR_DEST:
160 offset += ipv6_optlen(exthdr);
161 nexthdr = exthdr->nexthdr;
162 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
163 break;
164
165 case IPPROTO_UDP:
166 case IPPROTO_UDPLITE:
167 case IPPROTO_TCP:
168 case IPPROTO_SCTP:
169 case IPPROTO_DCCP:
170 if (!onlyproto && (nh + offset + 4 < skb->data ||
171 pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
172 __be16 *ports;
173
174 nh = skb_network_header(skb);
175 ports = (__be16 *)(nh + offset);
176 fl6->fl6_sport = ports[!!reverse];
177 fl6->fl6_dport = ports[!reverse];
178 }
179 fl6->flowi6_proto = nexthdr;
180 return;
181
182 case IPPROTO_ICMPV6:
183 if (!onlyproto && (nh + offset + 2 < skb->data ||
184 pskb_may_pull(skb, nh + offset + 2 - skb->data))) {
185 u8 *icmp;
186
187 nh = skb_network_header(skb);
188 icmp = (u8 *)(nh + offset);
189 fl6->fl6_icmp_type = icmp[0];
190 fl6->fl6_icmp_code = icmp[1];
191 }
192 fl6->flowi6_proto = nexthdr;
193 return;
194
195 #if IS_ENABLED(CONFIG_IPV6_MIP6)
196 case IPPROTO_MH:
197 offset += ipv6_optlen(exthdr);
198 if (!onlyproto && (nh + offset + 3 < skb->data ||
199 pskb_may_pull(skb, nh + offset + 3 - skb->data))) {
200 struct ip6_mh *mh;
201
202 nh = skb_network_header(skb);
203 mh = (struct ip6_mh *)(nh + offset);
204 fl6->fl6_mh_type = mh->ip6mh_type;
205 }
206 fl6->flowi6_proto = nexthdr;
207 return;
208 #endif
209
210 /* XXX Why are there these headers? */
211 case IPPROTO_AH:
212 case IPPROTO_ESP:
213 case IPPROTO_COMP:
214 default:
215 fl6->fl6_ipsec_spi = 0;
216 fl6->flowi6_proto = nexthdr;
217 return;
218 }
219 }
220 }
221
xfrm6_garbage_collect(struct dst_ops * ops)222 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
223 {
224 struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
225
226 xfrm6_policy_afinfo.garbage_collect(net);
227 return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
228 }
229
xfrm6_update_pmtu(struct dst_entry * dst,struct sock * sk,struct sk_buff * skb,u32 mtu)230 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
231 struct sk_buff *skb, u32 mtu)
232 {
233 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
234 struct dst_entry *path = xdst->route;
235
236 path->ops->update_pmtu(path, sk, skb, mtu);
237 }
238
xfrm6_redirect(struct dst_entry * dst,struct sock * sk,struct sk_buff * skb)239 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
240 struct sk_buff *skb)
241 {
242 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
243 struct dst_entry *path = xdst->route;
244
245 path->ops->redirect(path, sk, skb);
246 }
247
xfrm6_dst_destroy(struct dst_entry * dst)248 static void xfrm6_dst_destroy(struct dst_entry *dst)
249 {
250 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
251
252 if (likely(xdst->u.rt6.rt6i_idev))
253 in6_dev_put(xdst->u.rt6.rt6i_idev);
254 dst_destroy_metrics_generic(dst);
255 xfrm_dst_destroy(xdst);
256 }
257
xfrm6_dst_ifdown(struct dst_entry * dst,struct net_device * dev,int unregister)258 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
259 int unregister)
260 {
261 struct xfrm_dst *xdst;
262
263 if (!unregister)
264 return;
265
266 xdst = (struct xfrm_dst *)dst;
267 if (xdst->u.rt6.rt6i_idev->dev == dev) {
268 struct inet6_dev *loopback_idev =
269 in6_dev_get(dev_net(dev)->loopback_dev);
270 BUG_ON(!loopback_idev);
271
272 do {
273 in6_dev_put(xdst->u.rt6.rt6i_idev);
274 xdst->u.rt6.rt6i_idev = loopback_idev;
275 in6_dev_hold(loopback_idev);
276 xdst = (struct xfrm_dst *)xdst->u.dst.child;
277 } while (xdst->u.dst.xfrm);
278
279 __in6_dev_put(loopback_idev);
280 }
281
282 xfrm_dst_ifdown(dst, dev);
283 }
284
285 static struct dst_ops xfrm6_dst_ops_template = {
286 .family = AF_INET6,
287 .gc = xfrm6_garbage_collect,
288 .update_pmtu = xfrm6_update_pmtu,
289 .redirect = xfrm6_redirect,
290 .cow_metrics = dst_cow_metrics_generic,
291 .destroy = xfrm6_dst_destroy,
292 .ifdown = xfrm6_dst_ifdown,
293 .local_out = __ip6_local_out,
294 .gc_thresh = INT_MAX,
295 };
296
297 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
298 .family = AF_INET6,
299 .dst_ops = &xfrm6_dst_ops_template,
300 .dst_lookup = xfrm6_dst_lookup,
301 .get_saddr = xfrm6_get_saddr,
302 .decode_session = _decode_session6,
303 .get_tos = xfrm6_get_tos,
304 .init_path = xfrm6_init_path,
305 .fill_dst = xfrm6_fill_dst,
306 .blackhole_route = ip6_blackhole_route,
307 };
308
xfrm6_policy_init(void)309 static int __init xfrm6_policy_init(void)
310 {
311 return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
312 }
313
xfrm6_policy_fini(void)314 static void xfrm6_policy_fini(void)
315 {
316 xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
317 }
318
319 #ifdef CONFIG_SYSCTL
320 static struct ctl_table xfrm6_policy_table[] = {
321 {
322 .procname = "xfrm6_gc_thresh",
323 .data = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
324 .maxlen = sizeof(int),
325 .mode = 0644,
326 .proc_handler = proc_dointvec,
327 },
328 { }
329 };
330
xfrm6_net_sysctl_init(struct net * net)331 static int __net_init xfrm6_net_sysctl_init(struct net *net)
332 {
333 struct ctl_table *table;
334 struct ctl_table_header *hdr;
335
336 table = xfrm6_policy_table;
337 if (!net_eq(net, &init_net)) {
338 table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
339 if (!table)
340 goto err_alloc;
341
342 table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
343 }
344
345 hdr = register_net_sysctl(net, "net/ipv6", table);
346 if (!hdr)
347 goto err_reg;
348
349 net->ipv6.sysctl.xfrm6_hdr = hdr;
350 return 0;
351
352 err_reg:
353 if (!net_eq(net, &init_net))
354 kfree(table);
355 err_alloc:
356 return -ENOMEM;
357 }
358
xfrm6_net_sysctl_exit(struct net * net)359 static void __net_exit xfrm6_net_sysctl_exit(struct net *net)
360 {
361 struct ctl_table *table;
362
363 if (!net->ipv6.sysctl.xfrm6_hdr)
364 return;
365
366 table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
367 unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
368 if (!net_eq(net, &init_net))
369 kfree(table);
370 }
371 #else /* CONFIG_SYSCTL */
xfrm6_net_sysctl_init(struct net * net)372 static inline int xfrm6_net_sysctl_init(struct net *net)
373 {
374 return 0;
375 }
376
xfrm6_net_sysctl_exit(struct net * net)377 static inline void xfrm6_net_sysctl_exit(struct net *net)
378 {
379 }
380 #endif
381
xfrm6_net_init(struct net * net)382 static int __net_init xfrm6_net_init(struct net *net)
383 {
384 int ret;
385
386 memcpy(&net->xfrm.xfrm6_dst_ops, &xfrm6_dst_ops_template,
387 sizeof(xfrm6_dst_ops_template));
388 ret = dst_entries_init(&net->xfrm.xfrm6_dst_ops);
389 if (ret)
390 return ret;
391
392 ret = xfrm6_net_sysctl_init(net);
393 if (ret)
394 dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
395
396 return ret;
397 }
398
xfrm6_net_exit(struct net * net)399 static void __net_exit xfrm6_net_exit(struct net *net)
400 {
401 xfrm6_net_sysctl_exit(net);
402 dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
403 }
404
405 static struct pernet_operations xfrm6_net_ops = {
406 .init = xfrm6_net_init,
407 .exit = xfrm6_net_exit,
408 };
409
xfrm6_init(void)410 int __init xfrm6_init(void)
411 {
412 int ret;
413
414 ret = xfrm6_policy_init();
415 if (ret)
416 goto out;
417 ret = xfrm6_state_init();
418 if (ret)
419 goto out_policy;
420
421 ret = xfrm6_protocol_init();
422 if (ret)
423 goto out_state;
424
425 register_pernet_subsys(&xfrm6_net_ops);
426 out:
427 return ret;
428 out_state:
429 xfrm6_state_fini();
430 out_policy:
431 xfrm6_policy_fini();
432 goto out;
433 }
434
xfrm6_fini(void)435 void xfrm6_fini(void)
436 {
437 unregister_pernet_subsys(&xfrm6_net_ops);
438 xfrm6_protocol_fini();
439 xfrm6_policy_fini();
440 xfrm6_state_fini();
441 }
442