2 * xfrm6_policy.c: based on xfrm4_policy.c
6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
10 * Split up af-specific portion
14 #include <linux/err.h>
15 #include <linux/kernel.h>
16 #include <linux/netdevice.h>
17 #include <net/addrconf.h>
22 #include <net/ip6_route.h>
23 #include <net/l3mdev.h>
24 #if IS_ENABLED(CONFIG_IPV6_MIP6)
28 static struct xfrm_policy_afinfo xfrm6_policy_afinfo;
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)
35 struct dst_entry *dst;
38 memset(&fl6, 0, sizeof(fl6));
40 fl6.flowi6_flags = FLOWI_FLAG_SKIP_NH_OIF;
41 memcpy(&fl6.daddr, daddr, sizeof(fl6.daddr));
43 memcpy(&fl6.saddr, saddr, sizeof(fl6.saddr));
45 dst = ip6_route_output(net, NULL, &fl6);
56 static int xfrm6_get_saddr(struct net *net, int oif,
57 xfrm_address_t *saddr, xfrm_address_t *daddr)
59 struct dst_entry *dst;
60 struct net_device *dev;
62 dst = xfrm6_dst_lookup(net, 0, oif, NULL, daddr);
66 dev = ip6_dst_idev(dst)->dev;
67 ipv6_dev_get_saddr(dev_net(dev), dev, &daddr->in6, 0, &saddr->in6);
72 static int xfrm6_get_tos(const struct flowi *fl)
77 static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
80 if (dst->ops->family == AF_INET6) {
81 struct rt6_info *rt = (struct rt6_info *)dst;
82 path->path_cookie = rt6_get_cookie(rt);
85 path->u.rt6.rt6i_nfheader_len = nfheader_len;
90 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
91 const struct flowi *fl)
93 struct rt6_info *rt = (struct rt6_info *)xdst->route;
95 xdst->u.dst.dev = dev;
98 xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
99 if (!xdst->u.rt6.rt6i_idev) {
104 /* Sheit... I remember I did this right. Apparently,
105 * it was magically lost, so this code needs audit */
106 xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
108 xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
109 xdst->u.rt6.rt6i_node = rt->rt6i_node;
110 xdst->route_cookie = rt6_get_cookie(rt);
111 xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
112 xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
113 xdst->u.rt6.rt6i_src = rt->rt6i_src;
119 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
121 struct flowi6 *fl6 = &fl->u.ip6;
123 const struct ipv6hdr *hdr = ipv6_hdr(skb);
124 u16 offset = sizeof(*hdr);
125 struct ipv6_opt_hdr *exthdr;
126 const unsigned char *nh = skb_network_header(skb);
127 u16 nhoff = IP6CB(skb)->nhoff;
132 nhoff = offsetof(struct ipv6hdr, nexthdr);
137 oif = l3mdev_fib_oif(skb_dst(skb)->dev);
139 memset(fl6, 0, sizeof(struct flowi6));
140 fl6->flowi6_mark = skb->mark;
141 fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
143 fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
144 fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
146 while (nh + offset + 1 < skb->data ||
147 pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
148 nh = skb_network_header(skb);
149 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
152 case NEXTHDR_FRAGMENT:
154 case NEXTHDR_ROUTING:
157 offset += ipv6_optlen(exthdr);
158 nexthdr = exthdr->nexthdr;
159 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
163 case IPPROTO_UDPLITE:
167 if (!onlyproto && (nh + offset + 4 < skb->data ||
168 pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
171 nh = skb_network_header(skb);
172 ports = (__be16 *)(nh + offset);
173 fl6->fl6_sport = ports[!!reverse];
174 fl6->fl6_dport = ports[!reverse];
176 fl6->flowi6_proto = nexthdr;
180 if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
183 nh = skb_network_header(skb);
184 icmp = (u8 *)(nh + offset);
185 fl6->fl6_icmp_type = icmp[0];
186 fl6->fl6_icmp_code = icmp[1];
188 fl6->flowi6_proto = nexthdr;
191 #if IS_ENABLED(CONFIG_IPV6_MIP6)
193 offset += ipv6_optlen(exthdr);
194 if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
197 nh = skb_network_header(skb);
198 mh = (struct ip6_mh *)(nh + offset);
199 fl6->fl6_mh_type = mh->ip6mh_type;
201 fl6->flowi6_proto = nexthdr;
205 /* XXX Why are there these headers? */
210 fl6->fl6_ipsec_spi = 0;
211 fl6->flowi6_proto = nexthdr;
217 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
219 struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
221 xfrm6_policy_afinfo.garbage_collect(net);
222 return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
225 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
226 struct sk_buff *skb, u32 mtu)
228 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
229 struct dst_entry *path = xdst->route;
231 path->ops->update_pmtu(path, sk, skb, mtu);
234 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
237 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
238 struct dst_entry *path = xdst->route;
240 path->ops->redirect(path, sk, skb);
243 static void xfrm6_dst_destroy(struct dst_entry *dst)
245 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
247 if (likely(xdst->u.rt6.rt6i_idev))
248 in6_dev_put(xdst->u.rt6.rt6i_idev);
249 dst_destroy_metrics_generic(dst);
250 xfrm_dst_destroy(xdst);
253 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
256 struct xfrm_dst *xdst;
261 xdst = (struct xfrm_dst *)dst;
262 if (xdst->u.rt6.rt6i_idev->dev == dev) {
263 struct inet6_dev *loopback_idev =
264 in6_dev_get(dev_net(dev)->loopback_dev);
265 BUG_ON(!loopback_idev);
268 in6_dev_put(xdst->u.rt6.rt6i_idev);
269 xdst->u.rt6.rt6i_idev = loopback_idev;
270 in6_dev_hold(loopback_idev);
271 xdst = (struct xfrm_dst *)xdst->u.dst.child;
272 } while (xdst->u.dst.xfrm);
274 __in6_dev_put(loopback_idev);
277 xfrm_dst_ifdown(dst, dev);
280 static struct dst_ops xfrm6_dst_ops = {
282 .gc = xfrm6_garbage_collect,
283 .update_pmtu = xfrm6_update_pmtu,
284 .redirect = xfrm6_redirect,
285 .cow_metrics = dst_cow_metrics_generic,
286 .destroy = xfrm6_dst_destroy,
287 .ifdown = xfrm6_dst_ifdown,
288 .local_out = __ip6_local_out,
292 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
294 .dst_ops = &xfrm6_dst_ops,
295 .dst_lookup = xfrm6_dst_lookup,
296 .get_saddr = xfrm6_get_saddr,
297 .decode_session = _decode_session6,
298 .get_tos = xfrm6_get_tos,
299 .init_path = xfrm6_init_path,
300 .fill_dst = xfrm6_fill_dst,
301 .blackhole_route = ip6_blackhole_route,
304 static int __init xfrm6_policy_init(void)
306 return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
309 static void xfrm6_policy_fini(void)
311 xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
315 static struct ctl_table xfrm6_policy_table[] = {
317 .procname = "xfrm6_gc_thresh",
318 .data = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
319 .maxlen = sizeof(int),
321 .proc_handler = proc_dointvec,
326 static int __net_init xfrm6_net_init(struct net *net)
328 struct ctl_table *table;
329 struct ctl_table_header *hdr;
331 table = xfrm6_policy_table;
332 if (!net_eq(net, &init_net)) {
333 table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
337 table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
340 hdr = register_net_sysctl(net, "net/ipv6", table);
344 net->ipv6.sysctl.xfrm6_hdr = hdr;
348 if (!net_eq(net, &init_net))
354 static void __net_exit xfrm6_net_exit(struct net *net)
356 struct ctl_table *table;
358 if (!net->ipv6.sysctl.xfrm6_hdr)
361 table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
362 unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
363 if (!net_eq(net, &init_net))
367 static struct pernet_operations xfrm6_net_ops = {
368 .init = xfrm6_net_init,
369 .exit = xfrm6_net_exit,
373 int __init xfrm6_init(void)
377 dst_entries_init(&xfrm6_dst_ops);
379 ret = xfrm6_policy_init();
381 dst_entries_destroy(&xfrm6_dst_ops);
384 ret = xfrm6_state_init();
388 ret = xfrm6_protocol_init();
393 register_pernet_subsys(&xfrm6_net_ops);
404 void xfrm6_fini(void)
407 unregister_pernet_subsys(&xfrm6_net_ops);
409 xfrm6_protocol_fini();
412 dst_entries_destroy(&xfrm6_dst_ops);