1 /* 2 * xfrm4_policy.c 3 * 4 * Changes: 5 * Kazunori MIYAZAWA @USAGI 6 * YOSHIFUJI Hideaki @USAGI 7 * Split up af-specific portion 8 * 9 */ 10 11 #include <linux/err.h> 12 #include <linux/kernel.h> 13 #include <linux/inetdevice.h> 14 #include <linux/if_tunnel.h> 15 #include <net/dst.h> 16 #include <net/xfrm.h> 17 #include <net/ip.h> 18 19 static struct xfrm_policy_afinfo xfrm4_policy_afinfo; 20 21 static struct dst_entry *__xfrm4_dst_lookup(struct net *net, struct flowi4 *fl4, 22 int tos, 23 const xfrm_address_t *saddr, 24 const xfrm_address_t *daddr) 25 { 26 struct rtable *rt; 27 28 memset(fl4, 0, sizeof(*fl4)); 29 fl4->daddr = daddr->a4; 30 fl4->flowi4_tos = tos; 31 if (saddr) 32 fl4->saddr = saddr->a4; 33 34 rt = __ip_route_output_key(net, fl4); 35 if (!IS_ERR(rt)) 36 return &rt->dst; 37 38 return ERR_CAST(rt); 39 } 40 41 static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos, 42 const xfrm_address_t *saddr, 43 const xfrm_address_t *daddr) 44 { 45 struct flowi4 fl4; 46 47 return __xfrm4_dst_lookup(net, &fl4, tos, saddr, daddr); 48 } 49 50 static int xfrm4_get_saddr(struct net *net, 51 xfrm_address_t *saddr, xfrm_address_t *daddr) 52 { 53 struct dst_entry *dst; 54 struct flowi4 fl4; 55 56 dst = __xfrm4_dst_lookup(net, &fl4, 0, NULL, daddr); 57 if (IS_ERR(dst)) 58 return -EHOSTUNREACH; 59 60 saddr->a4 = fl4.saddr; 61 dst_release(dst); 62 return 0; 63 } 64 65 static int xfrm4_get_tos(const struct flowi *fl) 66 { 67 return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */ 68 } 69 70 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst, 71 int nfheader_len) 72 { 73 return 0; 74 } 75 76 static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev, 77 const struct flowi *fl) 78 { 79 struct rtable *rt = (struct rtable *)xdst->route; 80 const struct flowi4 *fl4 = &fl->u.ip4; 81 82 xdst->u.rt.rt_iif = fl4->flowi4_iif; 83 84 xdst->u.dst.dev = dev; 85 dev_hold(dev); 86 87 /* Sheit... I remember I did this right. Apparently, 88 * it was magically lost, so this code needs audit */ 89 xdst->u.rt.rt_is_input = rt->rt_is_input; 90 xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | 91 RTCF_LOCAL); 92 xdst->u.rt.rt_type = rt->rt_type; 93 xdst->u.rt.rt_gateway = rt->rt_gateway; 94 xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway; 95 xdst->u.rt.rt_pmtu = rt->rt_pmtu; 96 INIT_LIST_HEAD(&xdst->u.rt.rt_uncached); 97 98 return 0; 99 } 100 101 static void 102 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse) 103 { 104 const struct iphdr *iph = ip_hdr(skb); 105 u8 *xprth = skb_network_header(skb) + iph->ihl * 4; 106 struct flowi4 *fl4 = &fl->u.ip4; 107 108 memset(fl4, 0, sizeof(struct flowi4)); 109 fl4->flowi4_mark = skb->mark; 110 fl4->flowi4_oif = skb_dst(skb)->dev->ifindex; 111 112 if (!ip_is_fragment(iph)) { 113 switch (iph->protocol) { 114 case IPPROTO_UDP: 115 case IPPROTO_UDPLITE: 116 case IPPROTO_TCP: 117 case IPPROTO_SCTP: 118 case IPPROTO_DCCP: 119 if (xprth + 4 < skb->data || 120 pskb_may_pull(skb, xprth + 4 - skb->data)) { 121 __be16 *ports = (__be16 *)xprth; 122 123 fl4->fl4_sport = ports[!!reverse]; 124 fl4->fl4_dport = ports[!reverse]; 125 } 126 break; 127 128 case IPPROTO_ICMP: 129 if (pskb_may_pull(skb, xprth + 2 - skb->data)) { 130 u8 *icmp = xprth; 131 132 fl4->fl4_icmp_type = icmp[0]; 133 fl4->fl4_icmp_code = icmp[1]; 134 } 135 break; 136 137 case IPPROTO_ESP: 138 if (pskb_may_pull(skb, xprth + 4 - skb->data)) { 139 __be32 *ehdr = (__be32 *)xprth; 140 141 fl4->fl4_ipsec_spi = ehdr[0]; 142 } 143 break; 144 145 case IPPROTO_AH: 146 if (pskb_may_pull(skb, xprth + 8 - skb->data)) { 147 __be32 *ah_hdr = (__be32 *)xprth; 148 149 fl4->fl4_ipsec_spi = ah_hdr[1]; 150 } 151 break; 152 153 case IPPROTO_COMP: 154 if (pskb_may_pull(skb, xprth + 4 - skb->data)) { 155 __be16 *ipcomp_hdr = (__be16 *)xprth; 156 157 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1])); 158 } 159 break; 160 161 case IPPROTO_GRE: 162 if (pskb_may_pull(skb, xprth + 12 - skb->data)) { 163 __be16 *greflags = (__be16 *)xprth; 164 __be32 *gre_hdr = (__be32 *)xprth; 165 166 if (greflags[0] & GRE_KEY) { 167 if (greflags[0] & GRE_CSUM) 168 gre_hdr++; 169 fl4->fl4_gre_key = gre_hdr[1]; 170 } 171 } 172 break; 173 174 default: 175 fl4->fl4_ipsec_spi = 0; 176 break; 177 } 178 } 179 fl4->flowi4_proto = iph->protocol; 180 fl4->daddr = reverse ? iph->saddr : iph->daddr; 181 fl4->saddr = reverse ? iph->daddr : iph->saddr; 182 fl4->flowi4_tos = iph->tos; 183 } 184 185 static inline int xfrm4_garbage_collect(struct dst_ops *ops) 186 { 187 struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops); 188 189 xfrm4_policy_afinfo.garbage_collect(net); 190 return (dst_entries_get_slow(ops) > ops->gc_thresh * 2); 191 } 192 193 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk, 194 struct sk_buff *skb, u32 mtu) 195 { 196 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 197 struct dst_entry *path = xdst->route; 198 199 path->ops->update_pmtu(path, sk, skb, mtu); 200 } 201 202 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk, 203 struct sk_buff *skb) 204 { 205 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 206 struct dst_entry *path = xdst->route; 207 208 path->ops->redirect(path, sk, skb); 209 } 210 211 static void xfrm4_dst_destroy(struct dst_entry *dst) 212 { 213 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 214 215 dst_destroy_metrics_generic(dst); 216 217 xfrm_dst_destroy(xdst); 218 } 219 220 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, 221 int unregister) 222 { 223 if (!unregister) 224 return; 225 226 xfrm_dst_ifdown(dst, dev); 227 } 228 229 static struct dst_ops xfrm4_dst_ops = { 230 .family = AF_INET, 231 .protocol = cpu_to_be16(ETH_P_IP), 232 .gc = xfrm4_garbage_collect, 233 .update_pmtu = xfrm4_update_pmtu, 234 .redirect = xfrm4_redirect, 235 .cow_metrics = dst_cow_metrics_generic, 236 .destroy = xfrm4_dst_destroy, 237 .ifdown = xfrm4_dst_ifdown, 238 .local_out = __ip_local_out, 239 .gc_thresh = 1024, 240 }; 241 242 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = { 243 .family = AF_INET, 244 .dst_ops = &xfrm4_dst_ops, 245 .dst_lookup = xfrm4_dst_lookup, 246 .get_saddr = xfrm4_get_saddr, 247 .decode_session = _decode_session4, 248 .get_tos = xfrm4_get_tos, 249 .init_path = xfrm4_init_path, 250 .fill_dst = xfrm4_fill_dst, 251 .blackhole_route = ipv4_blackhole_route, 252 }; 253 254 #ifdef CONFIG_SYSCTL 255 static struct ctl_table xfrm4_policy_table[] = { 256 { 257 .procname = "xfrm4_gc_thresh", 258 .data = &init_net.xfrm.xfrm4_dst_ops.gc_thresh, 259 .maxlen = sizeof(int), 260 .mode = 0644, 261 .proc_handler = proc_dointvec, 262 }, 263 { } 264 }; 265 266 static int __net_init xfrm4_net_init(struct net *net) 267 { 268 struct ctl_table *table; 269 struct ctl_table_header *hdr; 270 271 table = xfrm4_policy_table; 272 if (!net_eq(net, &init_net)) { 273 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL); 274 if (!table) 275 goto err_alloc; 276 277 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh; 278 } 279 280 hdr = register_net_sysctl(net, "net/ipv4", table); 281 if (!hdr) 282 goto err_reg; 283 284 net->ipv4.xfrm4_hdr = hdr; 285 return 0; 286 287 err_reg: 288 if (!net_eq(net, &init_net)) 289 kfree(table); 290 err_alloc: 291 return -ENOMEM; 292 } 293 294 static void __net_exit xfrm4_net_exit(struct net *net) 295 { 296 struct ctl_table *table; 297 298 if (net->ipv4.xfrm4_hdr == NULL) 299 return; 300 301 table = net->ipv4.xfrm4_hdr->ctl_table_arg; 302 unregister_net_sysctl_table(net->ipv4.xfrm4_hdr); 303 if (!net_eq(net, &init_net)) 304 kfree(table); 305 } 306 307 static struct pernet_operations __net_initdata xfrm4_net_ops = { 308 .init = xfrm4_net_init, 309 .exit = xfrm4_net_exit, 310 }; 311 #endif 312 313 static void __init xfrm4_policy_init(void) 314 { 315 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo); 316 } 317 318 void __init xfrm4_init(void) 319 { 320 dst_entries_init(&xfrm4_dst_ops); 321 322 xfrm4_state_init(); 323 xfrm4_policy_init(); 324 #ifdef CONFIG_SYSCTL 325 register_pernet_subsys(&xfrm4_net_ops); 326 #endif 327 } 328 329