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 #include <net/vrf.h> 19 20 static struct xfrm_policy_afinfo xfrm4_policy_afinfo; 21 22 static struct dst_entry *__xfrm4_dst_lookup(struct net *net, struct flowi4 *fl4, 23 int tos, int oif, 24 const xfrm_address_t *saddr, 25 const xfrm_address_t *daddr) 26 { 27 struct rtable *rt; 28 29 memset(fl4, 0, sizeof(*fl4)); 30 fl4->daddr = daddr->a4; 31 fl4->flowi4_tos = tos; 32 fl4->flowi4_oif = oif; 33 if (saddr) 34 fl4->saddr = saddr->a4; 35 36 rt = __ip_route_output_key(net, fl4); 37 if (!IS_ERR(rt)) 38 return &rt->dst; 39 40 return ERR_CAST(rt); 41 } 42 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) 46 { 47 struct flowi4 fl4; 48 49 return __xfrm4_dst_lookup(net, &fl4, tos, oif, saddr, daddr); 50 } 51 52 static int xfrm4_get_saddr(struct net *net, int oif, 53 xfrm_address_t *saddr, xfrm_address_t *daddr) 54 { 55 struct dst_entry *dst; 56 struct flowi4 fl4; 57 58 dst = __xfrm4_dst_lookup(net, &fl4, 0, oif, NULL, daddr); 59 if (IS_ERR(dst)) 60 return -EHOSTUNREACH; 61 62 saddr->a4 = fl4.saddr; 63 dst_release(dst); 64 return 0; 65 } 66 67 static int xfrm4_get_tos(const struct flowi *fl) 68 { 69 return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */ 70 } 71 72 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst, 73 int nfheader_len) 74 { 75 return 0; 76 } 77 78 static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev, 79 const struct flowi *fl) 80 { 81 struct rtable *rt = (struct rtable *)xdst->route; 82 const struct flowi4 *fl4 = &fl->u.ip4; 83 84 xdst->u.rt.rt_iif = fl4->flowi4_iif; 85 86 xdst->u.dst.dev = dev; 87 dev_hold(dev); 88 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 | 93 RTCF_LOCAL); 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); 100 101 return 0; 102 } 103 104 static void 105 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse) 106 { 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; 110 int oif = 0; 111 112 if (skb_dst(skb)) { 113 oif = vrf_master_ifindex(skb_dst(skb)->dev) ? 114 : skb_dst(skb)->dev->ifindex; 115 } 116 117 memset(fl4, 0, sizeof(struct flowi4)); 118 fl4->flowi4_mark = skb->mark; 119 fl4->flowi4_oif = reverse ? skb->skb_iif : oif; 120 121 if (!ip_is_fragment(iph)) { 122 switch (iph->protocol) { 123 case IPPROTO_UDP: 124 case IPPROTO_UDPLITE: 125 case IPPROTO_TCP: 126 case IPPROTO_SCTP: 127 case IPPROTO_DCCP: 128 if (xprth + 4 < skb->data || 129 pskb_may_pull(skb, xprth + 4 - skb->data)) { 130 __be16 *ports = (__be16 *)xprth; 131 132 fl4->fl4_sport = ports[!!reverse]; 133 fl4->fl4_dport = ports[!reverse]; 134 } 135 break; 136 137 case IPPROTO_ICMP: 138 if (pskb_may_pull(skb, xprth + 2 - skb->data)) { 139 u8 *icmp = xprth; 140 141 fl4->fl4_icmp_type = icmp[0]; 142 fl4->fl4_icmp_code = icmp[1]; 143 } 144 break; 145 146 case IPPROTO_ESP: 147 if (pskb_may_pull(skb, xprth + 4 - skb->data)) { 148 __be32 *ehdr = (__be32 *)xprth; 149 150 fl4->fl4_ipsec_spi = ehdr[0]; 151 } 152 break; 153 154 case IPPROTO_AH: 155 if (pskb_may_pull(skb, xprth + 8 - skb->data)) { 156 __be32 *ah_hdr = (__be32 *)xprth; 157 158 fl4->fl4_ipsec_spi = ah_hdr[1]; 159 } 160 break; 161 162 case IPPROTO_COMP: 163 if (pskb_may_pull(skb, xprth + 4 - skb->data)) { 164 __be16 *ipcomp_hdr = (__be16 *)xprth; 165 166 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1])); 167 } 168 break; 169 170 case IPPROTO_GRE: 171 if (pskb_may_pull(skb, xprth + 12 - skb->data)) { 172 __be16 *greflags = (__be16 *)xprth; 173 __be32 *gre_hdr = (__be32 *)xprth; 174 175 if (greflags[0] & GRE_KEY) { 176 if (greflags[0] & GRE_CSUM) 177 gre_hdr++; 178 fl4->fl4_gre_key = gre_hdr[1]; 179 } 180 } 181 break; 182 183 default: 184 fl4->fl4_ipsec_spi = 0; 185 break; 186 } 187 } 188 fl4->flowi4_proto = iph->protocol; 189 fl4->daddr = reverse ? iph->saddr : iph->daddr; 190 fl4->saddr = reverse ? iph->daddr : iph->saddr; 191 fl4->flowi4_tos = iph->tos; 192 } 193 194 static inline int xfrm4_garbage_collect(struct dst_ops *ops) 195 { 196 struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops); 197 198 xfrm4_policy_afinfo.garbage_collect(net); 199 return (dst_entries_get_slow(ops) > ops->gc_thresh * 2); 200 } 201 202 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk, 203 struct sk_buff *skb, u32 mtu) 204 { 205 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 206 struct dst_entry *path = xdst->route; 207 208 path->ops->update_pmtu(path, sk, skb, mtu); 209 } 210 211 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk, 212 struct sk_buff *skb) 213 { 214 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 215 struct dst_entry *path = xdst->route; 216 217 path->ops->redirect(path, sk, skb); 218 } 219 220 static void xfrm4_dst_destroy(struct dst_entry *dst) 221 { 222 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 223 224 dst_destroy_metrics_generic(dst); 225 226 xfrm_dst_destroy(xdst); 227 } 228 229 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, 230 int unregister) 231 { 232 if (!unregister) 233 return; 234 235 xfrm_dst_ifdown(dst, dev); 236 } 237 238 static struct dst_ops xfrm4_dst_ops = { 239 .family = AF_INET, 240 .gc = xfrm4_garbage_collect, 241 .update_pmtu = xfrm4_update_pmtu, 242 .redirect = xfrm4_redirect, 243 .cow_metrics = dst_cow_metrics_generic, 244 .destroy = xfrm4_dst_destroy, 245 .ifdown = xfrm4_dst_ifdown, 246 .local_out = __ip_local_out, 247 .gc_thresh = 32768, 248 }; 249 250 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = { 251 .family = AF_INET, 252 .dst_ops = &xfrm4_dst_ops, 253 .dst_lookup = xfrm4_dst_lookup, 254 .get_saddr = xfrm4_get_saddr, 255 .decode_session = _decode_session4, 256 .get_tos = xfrm4_get_tos, 257 .init_path = xfrm4_init_path, 258 .fill_dst = xfrm4_fill_dst, 259 .blackhole_route = ipv4_blackhole_route, 260 }; 261 262 #ifdef CONFIG_SYSCTL 263 static struct ctl_table xfrm4_policy_table[] = { 264 { 265 .procname = "xfrm4_gc_thresh", 266 .data = &init_net.xfrm.xfrm4_dst_ops.gc_thresh, 267 .maxlen = sizeof(int), 268 .mode = 0644, 269 .proc_handler = proc_dointvec, 270 }, 271 { } 272 }; 273 274 static int __net_init xfrm4_net_init(struct net *net) 275 { 276 struct ctl_table *table; 277 struct ctl_table_header *hdr; 278 279 table = xfrm4_policy_table; 280 if (!net_eq(net, &init_net)) { 281 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL); 282 if (!table) 283 goto err_alloc; 284 285 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh; 286 } 287 288 hdr = register_net_sysctl(net, "net/ipv4", table); 289 if (!hdr) 290 goto err_reg; 291 292 net->ipv4.xfrm4_hdr = hdr; 293 return 0; 294 295 err_reg: 296 if (!net_eq(net, &init_net)) 297 kfree(table); 298 err_alloc: 299 return -ENOMEM; 300 } 301 302 static void __net_exit xfrm4_net_exit(struct net *net) 303 { 304 struct ctl_table *table; 305 306 if (!net->ipv4.xfrm4_hdr) 307 return; 308 309 table = net->ipv4.xfrm4_hdr->ctl_table_arg; 310 unregister_net_sysctl_table(net->ipv4.xfrm4_hdr); 311 if (!net_eq(net, &init_net)) 312 kfree(table); 313 } 314 315 static struct pernet_operations __net_initdata xfrm4_net_ops = { 316 .init = xfrm4_net_init, 317 .exit = xfrm4_net_exit, 318 }; 319 #endif 320 321 static void __init xfrm4_policy_init(void) 322 { 323 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo); 324 } 325 326 void __init xfrm4_init(void) 327 { 328 dst_entries_init(&xfrm4_dst_ops); 329 330 xfrm4_state_init(); 331 xfrm4_policy_init(); 332 xfrm4_protocol_init(); 333 #ifdef CONFIG_SYSCTL 334 register_pernet_subsys(&xfrm4_net_ops); 335 #endif 336 } 337 338