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/l3mdev.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 fl4->flowi4_flags = FLOWI_FLAG_SKIP_NH_OIF; 37 38 rt = __ip_route_output_key(net, fl4); 39 if (!IS_ERR(rt)) 40 return &rt->dst; 41 42 return ERR_CAST(rt); 43 } 44 45 static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos, int oif, 46 const xfrm_address_t *saddr, 47 const xfrm_address_t *daddr) 48 { 49 struct flowi4 fl4; 50 51 return __xfrm4_dst_lookup(net, &fl4, tos, oif, saddr, daddr); 52 } 53 54 static int xfrm4_get_saddr(struct net *net, int oif, 55 xfrm_address_t *saddr, xfrm_address_t *daddr) 56 { 57 struct dst_entry *dst; 58 struct flowi4 fl4; 59 60 dst = __xfrm4_dst_lookup(net, &fl4, 0, oif, NULL, daddr); 61 if (IS_ERR(dst)) 62 return -EHOSTUNREACH; 63 64 saddr->a4 = fl4.saddr; 65 dst_release(dst); 66 return 0; 67 } 68 69 static int xfrm4_get_tos(const struct flowi *fl) 70 { 71 return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */ 72 } 73 74 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst, 75 int nfheader_len) 76 { 77 return 0; 78 } 79 80 static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev, 81 const struct flowi *fl) 82 { 83 struct rtable *rt = (struct rtable *)xdst->route; 84 const struct flowi4 *fl4 = &fl->u.ip4; 85 86 xdst->u.rt.rt_iif = fl4->flowi4_iif; 87 88 xdst->u.dst.dev = dev; 89 dev_hold(dev); 90 91 /* Sheit... I remember I did this right. Apparently, 92 * it was magically lost, so this code needs audit */ 93 xdst->u.rt.rt_is_input = rt->rt_is_input; 94 xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | 95 RTCF_LOCAL); 96 xdst->u.rt.rt_type = rt->rt_type; 97 xdst->u.rt.rt_gateway = rt->rt_gateway; 98 xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway; 99 xdst->u.rt.rt_pmtu = rt->rt_pmtu; 100 xdst->u.rt.rt_table_id = rt->rt_table_id; 101 INIT_LIST_HEAD(&xdst->u.rt.rt_uncached); 102 103 return 0; 104 } 105 106 static void 107 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse) 108 { 109 const struct iphdr *iph = ip_hdr(skb); 110 u8 *xprth = skb_network_header(skb) + iph->ihl * 4; 111 struct flowi4 *fl4 = &fl->u.ip4; 112 int oif = 0; 113 114 if (skb_dst(skb)) 115 oif = l3mdev_fib_oif(skb_dst(skb)->dev); 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; 131 132 xprth = skb_network_header(skb) + iph->ihl * 4; 133 ports = (__be16 *)xprth; 134 135 fl4->fl4_sport = ports[!!reverse]; 136 fl4->fl4_dport = ports[!reverse]; 137 } 138 break; 139 140 case IPPROTO_ICMP: 141 if (xprth + 2 < skb->data || 142 pskb_may_pull(skb, xprth + 2 - skb->data)) { 143 u8 *icmp; 144 145 xprth = skb_network_header(skb) + iph->ihl * 4; 146 icmp = xprth; 147 148 fl4->fl4_icmp_type = icmp[0]; 149 fl4->fl4_icmp_code = icmp[1]; 150 } 151 break; 152 153 case IPPROTO_ESP: 154 if (xprth + 4 < skb->data || 155 pskb_may_pull(skb, xprth + 4 - skb->data)) { 156 __be32 *ehdr; 157 158 xprth = skb_network_header(skb) + iph->ihl * 4; 159 ehdr = (__be32 *)xprth; 160 161 fl4->fl4_ipsec_spi = ehdr[0]; 162 } 163 break; 164 165 case IPPROTO_AH: 166 if (xprth + 8 < skb->data || 167 pskb_may_pull(skb, xprth + 8 - skb->data)) { 168 __be32 *ah_hdr; 169 170 xprth = skb_network_header(skb) + iph->ihl * 4; 171 ah_hdr = (__be32 *)xprth; 172 173 fl4->fl4_ipsec_spi = ah_hdr[1]; 174 } 175 break; 176 177 case IPPROTO_COMP: 178 if (xprth + 4 < skb->data || 179 pskb_may_pull(skb, xprth + 4 - skb->data)) { 180 __be16 *ipcomp_hdr; 181 182 xprth = skb_network_header(skb) + iph->ihl * 4; 183 ipcomp_hdr = (__be16 *)xprth; 184 185 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1])); 186 } 187 break; 188 189 case IPPROTO_GRE: 190 if (xprth + 12 < skb->data || 191 pskb_may_pull(skb, xprth + 12 - skb->data)) { 192 __be16 *greflags; 193 __be32 *gre_hdr; 194 195 xprth = skb_network_header(skb) + iph->ihl * 4; 196 greflags = (__be16 *)xprth; 197 gre_hdr = (__be32 *)xprth; 198 199 if (greflags[0] & GRE_KEY) { 200 if (greflags[0] & GRE_CSUM) 201 gre_hdr++; 202 fl4->fl4_gre_key = gre_hdr[1]; 203 } 204 } 205 break; 206 207 default: 208 fl4->fl4_ipsec_spi = 0; 209 break; 210 } 211 } 212 fl4->flowi4_proto = iph->protocol; 213 fl4->daddr = reverse ? iph->saddr : iph->daddr; 214 fl4->saddr = reverse ? iph->daddr : iph->saddr; 215 fl4->flowi4_tos = iph->tos; 216 } 217 218 static inline int xfrm4_garbage_collect(struct dst_ops *ops) 219 { 220 struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops); 221 222 xfrm4_policy_afinfo.garbage_collect(net); 223 return (dst_entries_get_slow(ops) > ops->gc_thresh * 2); 224 } 225 226 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk, 227 struct sk_buff *skb, u32 mtu) 228 { 229 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 230 struct dst_entry *path = xdst->route; 231 232 path->ops->update_pmtu(path, sk, skb, mtu); 233 } 234 235 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk, 236 struct sk_buff *skb) 237 { 238 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 239 struct dst_entry *path = xdst->route; 240 241 path->ops->redirect(path, sk, skb); 242 } 243 244 static void xfrm4_dst_destroy(struct dst_entry *dst) 245 { 246 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 247 248 dst_destroy_metrics_generic(dst); 249 250 xfrm_dst_destroy(xdst); 251 } 252 253 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, 254 int unregister) 255 { 256 if (!unregister) 257 return; 258 259 xfrm_dst_ifdown(dst, dev); 260 } 261 262 static struct dst_ops xfrm4_dst_ops_template = { 263 .family = AF_INET, 264 .gc = xfrm4_garbage_collect, 265 .update_pmtu = xfrm4_update_pmtu, 266 .redirect = xfrm4_redirect, 267 .cow_metrics = dst_cow_metrics_generic, 268 .destroy = xfrm4_dst_destroy, 269 .ifdown = xfrm4_dst_ifdown, 270 .local_out = __ip_local_out, 271 .gc_thresh = INT_MAX, 272 }; 273 274 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = { 275 .family = AF_INET, 276 .dst_ops = &xfrm4_dst_ops_template, 277 .dst_lookup = xfrm4_dst_lookup, 278 .get_saddr = xfrm4_get_saddr, 279 .decode_session = _decode_session4, 280 .get_tos = xfrm4_get_tos, 281 .init_path = xfrm4_init_path, 282 .fill_dst = xfrm4_fill_dst, 283 .blackhole_route = ipv4_blackhole_route, 284 }; 285 286 #ifdef CONFIG_SYSCTL 287 static struct ctl_table xfrm4_policy_table[] = { 288 { 289 .procname = "xfrm4_gc_thresh", 290 .data = &init_net.xfrm.xfrm4_dst_ops.gc_thresh, 291 .maxlen = sizeof(int), 292 .mode = 0644, 293 .proc_handler = proc_dointvec, 294 }, 295 { } 296 }; 297 298 static int __net_init xfrm4_net_sysctl_init(struct net *net) 299 { 300 struct ctl_table *table; 301 struct ctl_table_header *hdr; 302 303 table = xfrm4_policy_table; 304 if (!net_eq(net, &init_net)) { 305 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL); 306 if (!table) 307 goto err_alloc; 308 309 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh; 310 } 311 312 hdr = register_net_sysctl(net, "net/ipv4", table); 313 if (!hdr) 314 goto err_reg; 315 316 net->ipv4.xfrm4_hdr = hdr; 317 return 0; 318 319 err_reg: 320 if (!net_eq(net, &init_net)) 321 kfree(table); 322 err_alloc: 323 return -ENOMEM; 324 } 325 326 static void __net_exit xfrm4_net_sysctl_exit(struct net *net) 327 { 328 struct ctl_table *table; 329 330 if (!net->ipv4.xfrm4_hdr) 331 return; 332 333 table = net->ipv4.xfrm4_hdr->ctl_table_arg; 334 unregister_net_sysctl_table(net->ipv4.xfrm4_hdr); 335 if (!net_eq(net, &init_net)) 336 kfree(table); 337 } 338 #else /* CONFIG_SYSCTL */ 339 static int inline xfrm4_net_sysctl_init(struct net *net) 340 { 341 return 0; 342 } 343 344 static void inline xfrm4_net_sysctl_exit(struct net *net) 345 { 346 } 347 #endif 348 349 static int __net_init xfrm4_net_init(struct net *net) 350 { 351 int ret; 352 353 memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template, 354 sizeof(xfrm4_dst_ops_template)); 355 ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops); 356 if (ret) 357 return ret; 358 359 ret = xfrm4_net_sysctl_init(net); 360 if (ret) 361 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops); 362 363 return ret; 364 } 365 366 static void __net_exit xfrm4_net_exit(struct net *net) 367 { 368 xfrm4_net_sysctl_exit(net); 369 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops); 370 } 371 372 static struct pernet_operations __net_initdata xfrm4_net_ops = { 373 .init = xfrm4_net_init, 374 .exit = xfrm4_net_exit, 375 }; 376 377 static void __init xfrm4_policy_init(void) 378 { 379 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo); 380 } 381 382 void __init xfrm4_init(void) 383 { 384 xfrm4_state_init(); 385 xfrm4_policy_init(); 386 xfrm4_protocol_init(); 387 register_pernet_subsys(&xfrm4_net_ops); 388 } 389 390