1
2
3
4
5
6
7
8
9
10
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#if IS_ENABLED(CONFIG_IPV6_MIP6)
24#include <net/mip6.h>
25#endif
26
27static struct xfrm_policy_afinfo xfrm6_policy_afinfo;
28
29static struct dst_entry *xfrm6_dst_lookup(struct net *net, int tos,
30 const xfrm_address_t *saddr,
31 const xfrm_address_t *daddr)
32{
33 struct flowi6 fl6;
34 struct dst_entry *dst;
35 int err;
36
37 memset(&fl6, 0, sizeof(fl6));
38 memcpy(&fl6.daddr, daddr, sizeof(fl6.daddr));
39 if (saddr)
40 memcpy(&fl6.saddr, saddr, sizeof(fl6.saddr));
41
42 dst = ip6_route_output(net, NULL, &fl6);
43
44 err = dst->error;
45 if (dst->error) {
46 dst_release(dst);
47 dst = ERR_PTR(err);
48 }
49
50 return dst;
51}
52
53static int xfrm6_get_saddr(struct net *net,
54 xfrm_address_t *saddr, xfrm_address_t *daddr)
55{
56 struct dst_entry *dst;
57 struct net_device *dev;
58
59 dst = xfrm6_dst_lookup(net, 0, NULL, daddr);
60 if (IS_ERR(dst))
61 return -EHOSTUNREACH;
62
63 dev = ip6_dst_idev(dst)->dev;
64 ipv6_dev_get_saddr(dev_net(dev), dev,
65 (struct in6_addr *)&daddr->a6, 0,
66 (struct in6_addr *)&saddr->a6);
67 dst_release(dst);
68 return 0;
69}
70
71static int xfrm6_get_tos(const struct flowi *fl)
72{
73 return 0;
74}
75
76static void xfrm6_init_dst(struct net *net, struct xfrm_dst *xdst)
77{
78 struct rt6_info *rt = (struct rt6_info *)xdst;
79
80 rt6_init_peer(rt, net->ipv6.peers);
81}
82
83static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
84 int nfheader_len)
85{
86 if (dst->ops->family == AF_INET6) {
87 struct rt6_info *rt = (struct rt6_info *)dst;
88 if (rt->rt6i_node)
89 path->path_cookie = rt->rt6i_node->fn_sernum;
90 }
91
92 path->u.rt6.rt6i_nfheader_len = nfheader_len;
93
94 return 0;
95}
96
97static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
98 const struct flowi *fl)
99{
100 struct rt6_info *rt = (struct rt6_info *)xdst->route;
101
102 xdst->u.dst.dev = dev;
103 dev_hold(dev);
104
105 xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
106 if (!xdst->u.rt6.rt6i_idev) {
107 dev_put(dev);
108 return -ENODEV;
109 }
110
111 rt6_transfer_peer(&xdst->u.rt6, rt);
112
113
114
115 xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
116 RTF_LOCAL);
117 xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
118 xdst->u.rt6.rt6i_node = rt->rt6i_node;
119 if (rt->rt6i_node)
120 xdst->route_cookie = rt->rt6i_node->fn_sernum;
121 xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
122 xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
123 xdst->u.rt6.rt6i_src = rt->rt6i_src;
124
125 return 0;
126}
127
128static inline void
129_decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
130{
131 struct flowi6 *fl6 = &fl->u.ip6;
132 int onlyproto = 0;
133 const struct ipv6hdr *hdr = ipv6_hdr(skb);
134 u16 offset = sizeof(*hdr);
135 struct ipv6_opt_hdr *exthdr;
136 const unsigned char *nh = skb_network_header(skb);
137 u16 nhoff = IP6CB(skb)->nhoff;
138 int oif = 0;
139 u8 nexthdr;
140
141 if (!nhoff)
142 nhoff = offsetof(struct ipv6hdr, nexthdr);
143
144 nexthdr = nh[nhoff];
145
146 if (skb_dst(skb))
147 oif = skb_dst(skb)->dev->ifindex;
148
149 memset(fl6, 0, sizeof(struct flowi6));
150 fl6->flowi6_mark = skb->mark;
151 fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
152
153 fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
154 fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
155
156 while (nh + offset + 1 < skb->data ||
157 pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
158 nh = skb_network_header(skb);
159 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
160
161 switch (nexthdr) {
162 case NEXTHDR_FRAGMENT:
163 onlyproto = 1;
164 case NEXTHDR_ROUTING:
165 case NEXTHDR_HOP:
166 case NEXTHDR_DEST:
167 offset += ipv6_optlen(exthdr);
168 nexthdr = exthdr->nexthdr;
169 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
170 break;
171
172 case IPPROTO_UDP:
173 case IPPROTO_UDPLITE:
174 case IPPROTO_TCP:
175 case IPPROTO_SCTP:
176 case IPPROTO_DCCP:
177 if (!onlyproto && (nh + offset + 4 < skb->data ||
178 pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
179 __be16 *ports;
180
181 nh = skb_network_header(skb);
182 ports = (__be16 *)(nh + offset);
183 fl6->fl6_sport = ports[!!reverse];
184 fl6->fl6_dport = ports[!reverse];
185 }
186 fl6->flowi6_proto = nexthdr;
187 return;
188
189 case IPPROTO_ICMPV6:
190 if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
191 u8 *icmp;
192
193 nh = skb_network_header(skb);
194 icmp = (u8 *)(nh + offset);
195 fl6->fl6_icmp_type = icmp[0];
196 fl6->fl6_icmp_code = icmp[1];
197 }
198 fl6->flowi6_proto = nexthdr;
199 return;
200
201#if IS_ENABLED(CONFIG_IPV6_MIP6)
202 case IPPROTO_MH:
203 offset += ipv6_optlen(exthdr);
204 if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
205 struct ip6_mh *mh;
206
207 nh = skb_network_header(skb);
208 mh = (struct ip6_mh *)(nh + offset);
209 fl6->fl6_mh_type = mh->ip6mh_type;
210 }
211 fl6->flowi6_proto = nexthdr;
212 return;
213#endif
214
215
216 case IPPROTO_AH:
217 case IPPROTO_ESP:
218 case IPPROTO_COMP:
219 default:
220 fl6->fl6_ipsec_spi = 0;
221 fl6->flowi6_proto = nexthdr;
222 return;
223 }
224 }
225}
226
227static inline int xfrm6_garbage_collect(struct dst_ops *ops)
228{
229 struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
230
231 xfrm6_policy_afinfo.garbage_collect(net);
232 return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
233}
234
235static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
236 struct sk_buff *skb, u32 mtu)
237{
238 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
239 struct dst_entry *path = xdst->route;
240
241 path->ops->update_pmtu(path, sk, skb, mtu);
242}
243
244static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
245 struct sk_buff *skb)
246{
247 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
248 struct dst_entry *path = xdst->route;
249
250 path->ops->redirect(path, sk, skb);
251}
252
253static void xfrm6_dst_destroy(struct dst_entry *dst)
254{
255 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
256
257 if (likely(xdst->u.rt6.rt6i_idev))
258 in6_dev_put(xdst->u.rt6.rt6i_idev);
259 dst_destroy_metrics_generic(dst);
260 if (rt6_has_peer(&xdst->u.rt6)) {
261 struct inet_peer *peer = rt6_peer_ptr(&xdst->u.rt6);
262 inet_putpeer(peer);
263 }
264 xfrm_dst_destroy(xdst);
265}
266
267static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
268 int unregister)
269{
270 struct xfrm_dst *xdst;
271
272 if (!unregister)
273 return;
274
275 xdst = (struct xfrm_dst *)dst;
276 if (xdst->u.rt6.rt6i_idev->dev == dev) {
277 struct inet6_dev *loopback_idev =
278 in6_dev_get(dev_net(dev)->loopback_dev);
279 BUG_ON(!loopback_idev);
280
281 do {
282 in6_dev_put(xdst->u.rt6.rt6i_idev);
283 xdst->u.rt6.rt6i_idev = loopback_idev;
284 in6_dev_hold(loopback_idev);
285 xdst = (struct xfrm_dst *)xdst->u.dst.child;
286 } while (xdst->u.dst.xfrm);
287
288 __in6_dev_put(loopback_idev);
289 }
290
291 xfrm_dst_ifdown(dst, dev);
292}
293
294static struct dst_ops xfrm6_dst_ops = {
295 .family = AF_INET6,
296 .protocol = cpu_to_be16(ETH_P_IPV6),
297 .gc = xfrm6_garbage_collect,
298 .update_pmtu = xfrm6_update_pmtu,
299 .redirect = xfrm6_redirect,
300 .cow_metrics = dst_cow_metrics_generic,
301 .destroy = xfrm6_dst_destroy,
302 .ifdown = xfrm6_dst_ifdown,
303 .local_out = __ip6_local_out,
304 .gc_thresh = 32768,
305};
306
307static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
308 .family = AF_INET6,
309 .dst_ops = &xfrm6_dst_ops,
310 .dst_lookup = xfrm6_dst_lookup,
311 .get_saddr = xfrm6_get_saddr,
312 .decode_session = _decode_session6,
313 .get_tos = xfrm6_get_tos,
314 .init_dst = xfrm6_init_dst,
315 .init_path = xfrm6_init_path,
316 .fill_dst = xfrm6_fill_dst,
317 .blackhole_route = ip6_blackhole_route,
318};
319
320static int __init xfrm6_policy_init(void)
321{
322 return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
323}
324
325static void xfrm6_policy_fini(void)
326{
327 xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
328}
329
330#ifdef CONFIG_SYSCTL
331static struct ctl_table xfrm6_policy_table[] = {
332 {
333 .procname = "xfrm6_gc_thresh",
334 .data = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
335 .maxlen = sizeof(int),
336 .mode = 0644,
337 .proc_handler = proc_dointvec,
338 },
339 { }
340};
341
342static int __net_init xfrm6_net_init(struct net *net)
343{
344 struct ctl_table *table;
345 struct ctl_table_header *hdr;
346
347 table = xfrm6_policy_table;
348 if (!net_eq(net, &init_net)) {
349 table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
350 if (!table)
351 goto err_alloc;
352
353 table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
354 }
355
356 hdr = register_net_sysctl(net, "net/ipv6", table);
357 if (!hdr)
358 goto err_reg;
359
360 net->ipv6.sysctl.xfrm6_hdr = hdr;
361 return 0;
362
363err_reg:
364 if (!net_eq(net, &init_net))
365 kfree(table);
366err_alloc:
367 return -ENOMEM;
368}
369
370static void __net_exit xfrm6_net_exit(struct net *net)
371{
372 struct ctl_table *table;
373
374 if (net->ipv6.sysctl.xfrm6_hdr == NULL)
375 return;
376
377 table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
378 unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
379 if (!net_eq(net, &init_net))
380 kfree(table);
381}
382
383static struct pernet_operations xfrm6_net_ops = {
384 .init = xfrm6_net_init,
385 .exit = xfrm6_net_exit,
386};
387#endif
388
389int __init xfrm6_init(void)
390{
391 int ret;
392
393 dst_entries_init(&xfrm6_dst_ops);
394
395 ret = xfrm6_policy_init();
396 if (ret) {
397 dst_entries_destroy(&xfrm6_dst_ops);
398 goto out;
399 }
400 ret = xfrm6_state_init();
401 if (ret)
402 goto out_policy;
403
404 ret = xfrm6_protocol_init();
405 if (ret)
406 goto out_state;
407
408#ifdef CONFIG_SYSCTL
409 register_pernet_subsys(&xfrm6_net_ops);
410#endif
411out:
412 return ret;
413out_state:
414 xfrm6_state_fini();
415out_policy:
416 xfrm6_policy_fini();
417 goto out;
418}
419
420void xfrm6_fini(void)
421{
422#ifdef CONFIG_SYSCTL
423 unregister_pernet_subsys(&xfrm6_net_ops);
424#endif
425 xfrm6_protocol_fini();
426 xfrm6_policy_fini();
427 xfrm6_state_fini();
428 dst_entries_destroy(&xfrm6_dst_ops);
429}
430
