5 * Kazunori MIYAZAWA @USAGI
6 * YOSHIFUJI Hideaki @USAGI
7 * Split up af-specific portion
11 #include <linux/err.h>
12 #include <linux/kernel.h>
13 #include <linux/inetdevice.h>
14 #include <linux/if_tunnel.h>
18 #include <net/l3mdev.h>
20 static struct dst_entry *__xfrm4_dst_lookup(struct net *net, struct flowi4 *fl4,
22 const xfrm_address_t *saddr,
23 const xfrm_address_t *daddr)
27 memset(fl4, 0, sizeof(*fl4));
28 fl4->daddr = daddr->a4;
29 fl4->flowi4_tos = tos;
30 fl4->flowi4_oif = l3mdev_master_ifindex_by_index(net, oif);
32 fl4->saddr = saddr->a4;
34 fl4->flowi4_flags = FLOWI_FLAG_SKIP_NH_OIF;
36 rt = __ip_route_output_key(net, fl4);
43 static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos, int oif,
44 const xfrm_address_t *saddr,
45 const xfrm_address_t *daddr)
49 return __xfrm4_dst_lookup(net, &fl4, tos, oif, saddr, daddr);
52 static int xfrm4_get_saddr(struct net *net, int oif,
53 xfrm_address_t *saddr, xfrm_address_t *daddr)
55 struct dst_entry *dst;
58 dst = __xfrm4_dst_lookup(net, &fl4, 0, oif, NULL, daddr);
62 saddr->a4 = fl4.saddr;
67 static int xfrm4_get_tos(const struct flowi *fl)
69 return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */
72 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
78 static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
79 const struct flowi *fl)
81 struct rtable *rt = (struct rtable *)xdst->route;
82 const struct flowi4 *fl4 = &fl->u.ip4;
84 xdst->u.rt.rt_iif = fl4->flowi4_iif;
86 xdst->u.dst.dev = dev;
89 /* Sheit... I remember I did this right. Apparently,
90 * it was magically lost, so this code needs audit */
91 xdst->u.rt.rt_is_input = rt->rt_is_input;
92 xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
94 xdst->u.rt.rt_type = rt->rt_type;
95 xdst->u.rt.rt_gateway = rt->rt_gateway;
96 xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway;
97 xdst->u.rt.rt_pmtu = rt->rt_pmtu;
98 xdst->u.rt.rt_table_id = rt->rt_table_id;
99 INIT_LIST_HEAD(&xdst->u.rt.rt_uncached);
105 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
107 const struct iphdr *iph = ip_hdr(skb);
108 u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
109 struct flowi4 *fl4 = &fl->u.ip4;
113 oif = skb_dst(skb)->dev->ifindex;
115 memset(fl4, 0, sizeof(struct flowi4));
116 fl4->flowi4_mark = skb->mark;
117 fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
119 if (!ip_is_fragment(iph)) {
120 switch (iph->protocol) {
122 case IPPROTO_UDPLITE:
126 if (xprth + 4 < skb->data ||
127 pskb_may_pull(skb, xprth + 4 - skb->data)) {
130 xprth = skb_network_header(skb) + iph->ihl * 4;
131 ports = (__be16 *)xprth;
133 fl4->fl4_sport = ports[!!reverse];
134 fl4->fl4_dport = ports[!reverse];
139 if (xprth + 2 < skb->data ||
140 pskb_may_pull(skb, xprth + 2 - skb->data)) {
143 xprth = skb_network_header(skb) + iph->ihl * 4;
146 fl4->fl4_icmp_type = icmp[0];
147 fl4->fl4_icmp_code = icmp[1];
152 if (xprth + 4 < skb->data ||
153 pskb_may_pull(skb, xprth + 4 - skb->data)) {
156 xprth = skb_network_header(skb) + iph->ihl * 4;
157 ehdr = (__be32 *)xprth;
159 fl4->fl4_ipsec_spi = ehdr[0];
164 if (xprth + 8 < skb->data ||
165 pskb_may_pull(skb, xprth + 8 - skb->data)) {
168 xprth = skb_network_header(skb) + iph->ihl * 4;
169 ah_hdr = (__be32 *)xprth;
171 fl4->fl4_ipsec_spi = ah_hdr[1];
176 if (xprth + 4 < skb->data ||
177 pskb_may_pull(skb, xprth + 4 - skb->data)) {
180 xprth = skb_network_header(skb) + iph->ihl * 4;
181 ipcomp_hdr = (__be16 *)xprth;
183 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
188 if (xprth + 12 < skb->data ||
189 pskb_may_pull(skb, xprth + 12 - skb->data)) {
193 xprth = skb_network_header(skb) + iph->ihl * 4;
194 greflags = (__be16 *)xprth;
195 gre_hdr = (__be32 *)xprth;
197 if (greflags[0] & GRE_KEY) {
198 if (greflags[0] & GRE_CSUM)
200 fl4->fl4_gre_key = gre_hdr[1];
206 fl4->fl4_ipsec_spi = 0;
210 fl4->flowi4_proto = iph->protocol;
211 fl4->daddr = reverse ? iph->saddr : iph->daddr;
212 fl4->saddr = reverse ? iph->daddr : iph->saddr;
213 fl4->flowi4_tos = iph->tos;
216 static inline int xfrm4_garbage_collect(struct dst_ops *ops)
218 struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
220 xfrm_garbage_collect_deferred(net);
221 return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
224 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
225 struct sk_buff *skb, u32 mtu)
227 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
228 struct dst_entry *path = xdst->route;
230 path->ops->update_pmtu(path, sk, skb, mtu);
233 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
236 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
237 struct dst_entry *path = xdst->route;
239 path->ops->redirect(path, sk, skb);
242 static void xfrm4_dst_destroy(struct dst_entry *dst)
244 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
246 dst_destroy_metrics_generic(dst);
248 xfrm_dst_destroy(xdst);
251 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
257 xfrm_dst_ifdown(dst, dev);
260 static struct dst_ops xfrm4_dst_ops_template = {
262 .gc = xfrm4_garbage_collect,
263 .update_pmtu = xfrm4_update_pmtu,
264 .redirect = xfrm4_redirect,
265 .cow_metrics = dst_cow_metrics_generic,
266 .destroy = xfrm4_dst_destroy,
267 .ifdown = xfrm4_dst_ifdown,
268 .local_out = __ip_local_out,
269 .gc_thresh = INT_MAX,
272 static const struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
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,
284 static struct ctl_table xfrm4_policy_table[] = {
286 .procname = "xfrm4_gc_thresh",
287 .data = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
288 .maxlen = sizeof(int),
290 .proc_handler = proc_dointvec,
295 static __net_init int xfrm4_net_sysctl_init(struct net *net)
297 struct ctl_table *table;
298 struct ctl_table_header *hdr;
300 table = xfrm4_policy_table;
301 if (!net_eq(net, &init_net)) {
302 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
306 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
309 hdr = register_net_sysctl(net, "net/ipv4", table);
313 net->ipv4.xfrm4_hdr = hdr;
317 if (!net_eq(net, &init_net))
323 static __net_exit void xfrm4_net_sysctl_exit(struct net *net)
325 struct ctl_table *table;
327 if (!net->ipv4.xfrm4_hdr)
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))
335 #else /* CONFIG_SYSCTL */
336 static inline int xfrm4_net_sysctl_init(struct net *net)
341 static inline void xfrm4_net_sysctl_exit(struct net *net)
346 static int __net_init xfrm4_net_init(struct net *net)
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);
356 ret = xfrm4_net_sysctl_init(net);
358 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
363 static void __net_exit xfrm4_net_exit(struct net *net)
365 xfrm4_net_sysctl_exit(net);
366 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
369 static struct pernet_operations __net_initdata xfrm4_net_ops = {
370 .init = xfrm4_net_init,
371 .exit = xfrm4_net_exit,
374 static void __init xfrm4_policy_init(void)
376 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo, AF_INET);
379 void __init xfrm4_init(void)
383 xfrm4_protocol_init();
384 register_pernet_subsys(&xfrm4_net_ops);