1 /* 2 * xfrm6_policy.c: based on xfrm4_policy.c 3 * 4 * Authors: 5 * Mitsuru KANDA @USAGI 6 * Kazunori MIYAZAWA @USAGI 7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com> 8 * IPv6 support 9 * YOSHIFUJI Hideaki 10 * Split up af-specific portion 11 * 12 */ 13 14 #include <linux/err.h> 15 #include <linux/kernel.h> 16 #include <linux/netdevice.h> 17 #include <net/addrconf.h> 18 #include <net/dst.h> 19 #include <net/xfrm.h> 20 #include <net/ip.h> 21 #include <net/ipv6.h> 22 #include <net/ip6_route.h> 23 #if IS_ENABLED(CONFIG_IPV6_MIP6) 24 #include <net/mip6.h> 25 #endif 26 27 static struct xfrm_policy_afinfo xfrm6_policy_afinfo; 28 29 static struct dst_entry *xfrm6_dst_lookup(struct net *net, int tos, 30 const xfrm_address_t *saddr, 31 const xfrm_address_t *daddr) 32 { 33 struct flowi6 fl6; 34 struct dst_entry *dst; 35 int err; 36 37 memset(&fl6, 0, sizeof(fl6)); 38 memcpy(&fl6.daddr, daddr, sizeof(fl6.daddr)); 39 if (saddr) 40 memcpy(&fl6.saddr, saddr, sizeof(fl6.saddr)); 41 42 dst = ip6_route_output(net, NULL, &fl6); 43 44 err = dst->error; 45 if (dst->error) { 46 dst_release(dst); 47 dst = ERR_PTR(err); 48 } 49 50 return dst; 51 } 52 53 static int xfrm6_get_saddr(struct net *net, 54 xfrm_address_t *saddr, xfrm_address_t *daddr) 55 { 56 struct dst_entry *dst; 57 struct net_device *dev; 58 59 dst = xfrm6_dst_lookup(net, 0, NULL, daddr); 60 if (IS_ERR(dst)) 61 return -EHOSTUNREACH; 62 63 dev = ip6_dst_idev(dst)->dev; 64 ipv6_dev_get_saddr(dev_net(dev), dev, &daddr->in6, 0, &saddr->in6); 65 dst_release(dst); 66 return 0; 67 } 68 69 static int xfrm6_get_tos(const struct flowi *fl) 70 { 71 return 0; 72 } 73 74 static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst, 75 int nfheader_len) 76 { 77 if (dst->ops->family == AF_INET6) { 78 struct rt6_info *rt = (struct rt6_info *)dst; 79 path->path_cookie = rt6_get_cookie(rt); 80 } 81 82 path->u.rt6.rt6i_nfheader_len = nfheader_len; 83 84 return 0; 85 } 86 87 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev, 88 const struct flowi *fl) 89 { 90 struct rt6_info *rt = (struct rt6_info *)xdst->route; 91 92 xdst->u.dst.dev = dev; 93 dev_hold(dev); 94 95 xdst->u.rt6.rt6i_idev = in6_dev_get(dev); 96 if (!xdst->u.rt6.rt6i_idev) { 97 dev_put(dev); 98 return -ENODEV; 99 } 100 101 /* Sheit... I remember I did this right. Apparently, 102 * it was magically lost, so this code needs audit */ 103 xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST | 104 RTF_LOCAL); 105 xdst->u.rt6.rt6i_metric = rt->rt6i_metric; 106 xdst->u.rt6.rt6i_node = rt->rt6i_node; 107 xdst->route_cookie = rt6_get_cookie(rt); 108 xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway; 109 xdst->u.rt6.rt6i_dst = rt->rt6i_dst; 110 xdst->u.rt6.rt6i_src = rt->rt6i_src; 111 112 return 0; 113 } 114 115 static inline void 116 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse) 117 { 118 struct flowi6 *fl6 = &fl->u.ip6; 119 int onlyproto = 0; 120 const struct ipv6hdr *hdr = ipv6_hdr(skb); 121 u16 offset = sizeof(*hdr); 122 struct ipv6_opt_hdr *exthdr; 123 const unsigned char *nh = skb_network_header(skb); 124 u16 nhoff = IP6CB(skb)->nhoff; 125 int oif = 0; 126 u8 nexthdr; 127 128 if (!nhoff) 129 nhoff = offsetof(struct ipv6hdr, nexthdr); 130 131 nexthdr = nh[nhoff]; 132 133 if (skb_dst(skb)) 134 oif = skb_dst(skb)->dev->ifindex; 135 136 memset(fl6, 0, sizeof(struct flowi6)); 137 fl6->flowi6_mark = skb->mark; 138 fl6->flowi6_oif = reverse ? skb->skb_iif : oif; 139 140 fl6->daddr = reverse ? hdr->saddr : hdr->daddr; 141 fl6->saddr = reverse ? hdr->daddr : hdr->saddr; 142 143 while (nh + offset + 1 < skb->data || 144 pskb_may_pull(skb, nh + offset + 1 - skb->data)) { 145 nh = skb_network_header(skb); 146 exthdr = (struct ipv6_opt_hdr *)(nh + offset); 147 148 switch (nexthdr) { 149 case NEXTHDR_FRAGMENT: 150 onlyproto = 1; 151 case NEXTHDR_ROUTING: 152 case NEXTHDR_HOP: 153 case NEXTHDR_DEST: 154 offset += ipv6_optlen(exthdr); 155 nexthdr = exthdr->nexthdr; 156 exthdr = (struct ipv6_opt_hdr *)(nh + offset); 157 break; 158 159 case IPPROTO_UDP: 160 case IPPROTO_UDPLITE: 161 case IPPROTO_TCP: 162 case IPPROTO_SCTP: 163 case IPPROTO_DCCP: 164 if (!onlyproto && (nh + offset + 4 < skb->data || 165 pskb_may_pull(skb, nh + offset + 4 - skb->data))) { 166 __be16 *ports; 167 168 nh = skb_network_header(skb); 169 ports = (__be16 *)(nh + offset); 170 fl6->fl6_sport = ports[!!reverse]; 171 fl6->fl6_dport = ports[!reverse]; 172 } 173 fl6->flowi6_proto = nexthdr; 174 return; 175 176 case IPPROTO_ICMPV6: 177 if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) { 178 u8 *icmp; 179 180 nh = skb_network_header(skb); 181 icmp = (u8 *)(nh + offset); 182 fl6->fl6_icmp_type = icmp[0]; 183 fl6->fl6_icmp_code = icmp[1]; 184 } 185 fl6->flowi6_proto = nexthdr; 186 return; 187 188 #if IS_ENABLED(CONFIG_IPV6_MIP6) 189 case IPPROTO_MH: 190 offset += ipv6_optlen(exthdr); 191 if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) { 192 struct ip6_mh *mh; 193 194 nh = skb_network_header(skb); 195 mh = (struct ip6_mh *)(nh + offset); 196 fl6->fl6_mh_type = mh->ip6mh_type; 197 } 198 fl6->flowi6_proto = nexthdr; 199 return; 200 #endif 201 202 /* XXX Why are there these headers? */ 203 case IPPROTO_AH: 204 case IPPROTO_ESP: 205 case IPPROTO_COMP: 206 default: 207 fl6->fl6_ipsec_spi = 0; 208 fl6->flowi6_proto = nexthdr; 209 return; 210 } 211 } 212 } 213 214 static inline int xfrm6_garbage_collect(struct dst_ops *ops) 215 { 216 struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops); 217 218 xfrm6_policy_afinfo.garbage_collect(net); 219 return dst_entries_get_fast(ops) > ops->gc_thresh * 2; 220 } 221 222 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk, 223 struct sk_buff *skb, u32 mtu) 224 { 225 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 226 struct dst_entry *path = xdst->route; 227 228 path->ops->update_pmtu(path, sk, skb, mtu); 229 } 230 231 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk, 232 struct sk_buff *skb) 233 { 234 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 235 struct dst_entry *path = xdst->route; 236 237 path->ops->redirect(path, sk, skb); 238 } 239 240 static void xfrm6_dst_destroy(struct dst_entry *dst) 241 { 242 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 243 244 if (likely(xdst->u.rt6.rt6i_idev)) 245 in6_dev_put(xdst->u.rt6.rt6i_idev); 246 dst_destroy_metrics_generic(dst); 247 xfrm_dst_destroy(xdst); 248 } 249 250 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev, 251 int unregister) 252 { 253 struct xfrm_dst *xdst; 254 255 if (!unregister) 256 return; 257 258 xdst = (struct xfrm_dst *)dst; 259 if (xdst->u.rt6.rt6i_idev->dev == dev) { 260 struct inet6_dev *loopback_idev = 261 in6_dev_get(dev_net(dev)->loopback_dev); 262 BUG_ON(!loopback_idev); 263 264 do { 265 in6_dev_put(xdst->u.rt6.rt6i_idev); 266 xdst->u.rt6.rt6i_idev = loopback_idev; 267 in6_dev_hold(loopback_idev); 268 xdst = (struct xfrm_dst *)xdst->u.dst.child; 269 } while (xdst->u.dst.xfrm); 270 271 __in6_dev_put(loopback_idev); 272 } 273 274 xfrm_dst_ifdown(dst, dev); 275 } 276 277 static struct dst_ops xfrm6_dst_ops = { 278 .family = AF_INET6, 279 .gc = xfrm6_garbage_collect, 280 .update_pmtu = xfrm6_update_pmtu, 281 .redirect = xfrm6_redirect, 282 .cow_metrics = dst_cow_metrics_generic, 283 .destroy = xfrm6_dst_destroy, 284 .ifdown = xfrm6_dst_ifdown, 285 .local_out = __ip6_local_out, 286 .gc_thresh = 32768, 287 }; 288 289 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = { 290 .family = AF_INET6, 291 .dst_ops = &xfrm6_dst_ops, 292 .dst_lookup = xfrm6_dst_lookup, 293 .get_saddr = xfrm6_get_saddr, 294 .decode_session = _decode_session6, 295 .get_tos = xfrm6_get_tos, 296 .init_path = xfrm6_init_path, 297 .fill_dst = xfrm6_fill_dst, 298 .blackhole_route = ip6_blackhole_route, 299 }; 300 301 static int __init xfrm6_policy_init(void) 302 { 303 return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo); 304 } 305 306 static void xfrm6_policy_fini(void) 307 { 308 xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo); 309 } 310 311 #ifdef CONFIG_SYSCTL 312 static struct ctl_table xfrm6_policy_table[] = { 313 { 314 .procname = "xfrm6_gc_thresh", 315 .data = &init_net.xfrm.xfrm6_dst_ops.gc_thresh, 316 .maxlen = sizeof(int), 317 .mode = 0644, 318 .proc_handler = proc_dointvec, 319 }, 320 { } 321 }; 322 323 static int __net_init xfrm6_net_init(struct net *net) 324 { 325 struct ctl_table *table; 326 struct ctl_table_header *hdr; 327 328 table = xfrm6_policy_table; 329 if (!net_eq(net, &init_net)) { 330 table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL); 331 if (!table) 332 goto err_alloc; 333 334 table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh; 335 } 336 337 hdr = register_net_sysctl(net, "net/ipv6", table); 338 if (!hdr) 339 goto err_reg; 340 341 net->ipv6.sysctl.xfrm6_hdr = hdr; 342 return 0; 343 344 err_reg: 345 if (!net_eq(net, &init_net)) 346 kfree(table); 347 err_alloc: 348 return -ENOMEM; 349 } 350 351 static void __net_exit xfrm6_net_exit(struct net *net) 352 { 353 struct ctl_table *table; 354 355 if (!net->ipv6.sysctl.xfrm6_hdr) 356 return; 357 358 table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg; 359 unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr); 360 if (!net_eq(net, &init_net)) 361 kfree(table); 362 } 363 364 static struct pernet_operations xfrm6_net_ops = { 365 .init = xfrm6_net_init, 366 .exit = xfrm6_net_exit, 367 }; 368 #endif 369 370 int __init xfrm6_init(void) 371 { 372 int ret; 373 374 dst_entries_init(&xfrm6_dst_ops); 375 376 ret = xfrm6_policy_init(); 377 if (ret) { 378 dst_entries_destroy(&xfrm6_dst_ops); 379 goto out; 380 } 381 ret = xfrm6_state_init(); 382 if (ret) 383 goto out_policy; 384 385 ret = xfrm6_protocol_init(); 386 if (ret) 387 goto out_state; 388 389 #ifdef CONFIG_SYSCTL 390 register_pernet_subsys(&xfrm6_net_ops); 391 #endif 392 out: 393 return ret; 394 out_state: 395 xfrm6_state_fini(); 396 out_policy: 397 xfrm6_policy_fini(); 398 goto out; 399 } 400 401 void xfrm6_fini(void) 402 { 403 #ifdef CONFIG_SYSCTL 404 unregister_pernet_subsys(&xfrm6_net_ops); 405 #endif 406 xfrm6_protocol_fini(); 407 xfrm6_policy_fini(); 408 xfrm6_state_fini(); 409 dst_entries_destroy(&xfrm6_dst_ops); 410 } 411