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/compiler.h> 15 #include <linux/netdevice.h> 16 #include <net/addrconf.h> 17 #include <net/xfrm.h> 18 #include <net/ip.h> 19 #include <net/ipv6.h> 20 #include <net/ip6_route.h> 21 #ifdef CONFIG_IPV6_MIP6 22 #include <net/mip6.h> 23 #endif 24 25 static struct dst_ops xfrm6_dst_ops; 26 static struct xfrm_policy_afinfo xfrm6_policy_afinfo; 27 28 static int xfrm6_dst_lookup(struct xfrm_dst **xdst, struct flowi *fl) 29 { 30 struct dst_entry *dst = ip6_route_output(NULL, fl); 31 int err = dst->error; 32 if (!err) 33 *xdst = (struct xfrm_dst *) dst; 34 else 35 dst_release(dst); 36 return err; 37 } 38 39 static int xfrm6_get_saddr(xfrm_address_t *saddr, xfrm_address_t *daddr) 40 { 41 struct rt6_info *rt; 42 struct flowi fl_tunnel = { 43 .nl_u = { 44 .ip6_u = { 45 .daddr = *(struct in6_addr *)&daddr->a6, 46 }, 47 }, 48 }; 49 50 if (!xfrm6_dst_lookup((struct xfrm_dst **)&rt, &fl_tunnel)) { 51 ipv6_get_saddr(&rt->u.dst, (struct in6_addr *)&daddr->a6, 52 (struct in6_addr *)&saddr->a6); 53 dst_release(&rt->u.dst); 54 return 0; 55 } 56 return -EHOSTUNREACH; 57 } 58 59 static struct dst_entry * 60 __xfrm6_find_bundle(struct flowi *fl, struct xfrm_policy *policy) 61 { 62 struct dst_entry *dst; 63 64 /* Still not clear if we should set fl->fl6_{src,dst}... */ 65 read_lock_bh(&policy->lock); 66 for (dst = policy->bundles; dst; dst = dst->next) { 67 struct xfrm_dst *xdst = (struct xfrm_dst*)dst; 68 struct in6_addr fl_dst_prefix, fl_src_prefix; 69 70 ipv6_addr_prefix(&fl_dst_prefix, 71 &fl->fl6_dst, 72 xdst->u.rt6.rt6i_dst.plen); 73 ipv6_addr_prefix(&fl_src_prefix, 74 &fl->fl6_src, 75 xdst->u.rt6.rt6i_src.plen); 76 if (ipv6_addr_equal(&xdst->u.rt6.rt6i_dst.addr, &fl_dst_prefix) && 77 ipv6_addr_equal(&xdst->u.rt6.rt6i_src.addr, &fl_src_prefix) && 78 xfrm_bundle_ok(policy, xdst, fl, AF_INET6, 79 (xdst->u.rt6.rt6i_dst.plen != 128 || 80 xdst->u.rt6.rt6i_src.plen != 128))) { 81 dst_clone(dst); 82 break; 83 } 84 } 85 read_unlock_bh(&policy->lock); 86 return dst; 87 } 88 89 static inline struct in6_addr* 90 __xfrm6_bundle_addr_remote(struct xfrm_state *x, struct in6_addr *addr) 91 { 92 return (x->type->remote_addr) ? 93 (struct in6_addr*)x->type->remote_addr(x, (xfrm_address_t *)addr) : 94 (struct in6_addr*)&x->id.daddr; 95 } 96 97 static inline struct in6_addr* 98 __xfrm6_bundle_addr_local(struct xfrm_state *x, struct in6_addr *addr) 99 { 100 return (x->type->local_addr) ? 101 (struct in6_addr*)x->type->local_addr(x, (xfrm_address_t *)addr) : 102 (struct in6_addr*)&x->props.saddr; 103 } 104 105 static inline void 106 __xfrm6_bundle_len_inc(int *len, int *nflen, struct xfrm_state *x) 107 { 108 if (x->type->flags & XFRM_TYPE_NON_FRAGMENT) 109 *nflen += x->props.header_len; 110 else 111 *len += x->props.header_len; 112 } 113 114 static inline void 115 __xfrm6_bundle_len_dec(int *len, int *nflen, struct xfrm_state *x) 116 { 117 if (x->type->flags & XFRM_TYPE_NON_FRAGMENT) 118 *nflen -= x->props.header_len; 119 else 120 *len -= x->props.header_len; 121 } 122 123 /* Allocate chain of dst_entry's, attach known xfrm's, calculate 124 * all the metrics... Shortly, bundle a bundle. 125 */ 126 127 static int 128 __xfrm6_bundle_create(struct xfrm_policy *policy, struct xfrm_state **xfrm, int nx, 129 struct flowi *fl, struct dst_entry **dst_p) 130 { 131 struct dst_entry *dst, *dst_prev; 132 struct rt6_info *rt0 = (struct rt6_info*)(*dst_p); 133 struct rt6_info *rt = rt0; 134 struct flowi fl_tunnel = { 135 .nl_u = { 136 .ip6_u = { 137 .saddr = fl->fl6_src, 138 .daddr = fl->fl6_dst, 139 } 140 } 141 }; 142 int i; 143 int err = 0; 144 int header_len = 0; 145 int nfheader_len = 0; 146 int trailer_len = 0; 147 148 dst = dst_prev = NULL; 149 dst_hold(&rt->u.dst); 150 151 for (i = 0; i < nx; i++) { 152 struct dst_entry *dst1 = dst_alloc(&xfrm6_dst_ops); 153 struct xfrm_dst *xdst; 154 155 if (unlikely(dst1 == NULL)) { 156 err = -ENOBUFS; 157 dst_release(&rt->u.dst); 158 goto error; 159 } 160 161 if (!dst) 162 dst = dst1; 163 else { 164 dst_prev->child = dst1; 165 dst1->flags |= DST_NOHASH; 166 dst_clone(dst1); 167 } 168 169 xdst = (struct xfrm_dst *)dst1; 170 xdst->route = &rt->u.dst; 171 xdst->genid = xfrm[i]->genid; 172 if (rt->rt6i_node) 173 xdst->route_cookie = rt->rt6i_node->fn_sernum; 174 175 dst1->next = dst_prev; 176 dst_prev = dst1; 177 178 __xfrm6_bundle_len_inc(&header_len, &nfheader_len, xfrm[i]); 179 trailer_len += xfrm[i]->props.trailer_len; 180 181 if (xfrm[i]->props.mode == XFRM_MODE_TUNNEL || 182 xfrm[i]->props.mode == XFRM_MODE_ROUTEOPTIMIZATION) { 183 unsigned short encap_family = xfrm[i]->props.family; 184 switch(encap_family) { 185 case AF_INET: 186 fl_tunnel.fl4_dst = xfrm[i]->id.daddr.a4; 187 fl_tunnel.fl4_src = xfrm[i]->props.saddr.a4; 188 break; 189 case AF_INET6: 190 ipv6_addr_copy(&fl_tunnel.fl6_dst, __xfrm6_bundle_addr_remote(xfrm[i], &fl->fl6_dst)); 191 192 ipv6_addr_copy(&fl_tunnel.fl6_src, __xfrm6_bundle_addr_local(xfrm[i], &fl->fl6_src)); 193 break; 194 default: 195 BUG_ON(1); 196 } 197 198 err = xfrm_dst_lookup((struct xfrm_dst **) &rt, 199 &fl_tunnel, encap_family); 200 if (err) 201 goto error; 202 } else 203 dst_hold(&rt->u.dst); 204 } 205 206 dst_prev->child = &rt->u.dst; 207 dst->path = &rt->u.dst; 208 if (rt->rt6i_node) 209 ((struct xfrm_dst *)dst)->path_cookie = rt->rt6i_node->fn_sernum; 210 211 *dst_p = dst; 212 dst = dst_prev; 213 214 dst_prev = *dst_p; 215 i = 0; 216 for (; dst_prev != &rt->u.dst; dst_prev = dst_prev->child) { 217 struct xfrm_dst *x = (struct xfrm_dst*)dst_prev; 218 struct xfrm_state_afinfo *afinfo; 219 220 dst_prev->xfrm = xfrm[i++]; 221 dst_prev->dev = rt->u.dst.dev; 222 if (rt->u.dst.dev) 223 dev_hold(rt->u.dst.dev); 224 dst_prev->obsolete = -1; 225 dst_prev->flags |= DST_HOST; 226 dst_prev->lastuse = jiffies; 227 dst_prev->header_len = header_len; 228 dst_prev->nfheader_len = nfheader_len; 229 dst_prev->trailer_len = trailer_len; 230 memcpy(&dst_prev->metrics, &x->route->metrics, sizeof(dst_prev->metrics)); 231 232 /* Copy neighbour for reachability confirmation */ 233 dst_prev->neighbour = neigh_clone(rt->u.dst.neighbour); 234 dst_prev->input = rt->u.dst.input; 235 /* XXX: When IPv4 is implemented as module and can be unloaded, 236 * we should manage reference to xfrm4_output in afinfo->output. 237 * Miyazawa 238 */ 239 afinfo = xfrm_state_get_afinfo(dst_prev->xfrm->props.family); 240 if (!afinfo) { 241 dst = *dst_p; 242 goto error; 243 }; 244 dst_prev->output = afinfo->output; 245 xfrm_state_put_afinfo(afinfo); 246 /* Sheit... I remember I did this right. Apparently, 247 * it was magically lost, so this code needs audit */ 248 x->u.rt6.rt6i_flags = rt0->rt6i_flags&(RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL); 249 x->u.rt6.rt6i_metric = rt0->rt6i_metric; 250 x->u.rt6.rt6i_node = rt0->rt6i_node; 251 x->u.rt6.rt6i_gateway = rt0->rt6i_gateway; 252 memcpy(&x->u.rt6.rt6i_gateway, &rt0->rt6i_gateway, sizeof(x->u.rt6.rt6i_gateway)); 253 x->u.rt6.rt6i_dst = rt0->rt6i_dst; 254 x->u.rt6.rt6i_src = rt0->rt6i_src; 255 x->u.rt6.rt6i_idev = rt0->rt6i_idev; 256 in6_dev_hold(rt0->rt6i_idev); 257 __xfrm6_bundle_len_dec(&header_len, &nfheader_len, x->u.dst.xfrm); 258 trailer_len -= x->u.dst.xfrm->props.trailer_len; 259 } 260 261 xfrm_init_pmtu(dst); 262 return 0; 263 264 error: 265 if (dst) 266 dst_free(dst); 267 return err; 268 } 269 270 static inline void 271 _decode_session6(struct sk_buff *skb, struct flowi *fl) 272 { 273 u16 offset = skb->h.raw - skb->nh.raw; 274 struct ipv6hdr *hdr = skb->nh.ipv6h; 275 struct ipv6_opt_hdr *exthdr; 276 u8 nexthdr = skb->nh.raw[IP6CB(skb)->nhoff]; 277 278 memset(fl, 0, sizeof(struct flowi)); 279 ipv6_addr_copy(&fl->fl6_dst, &hdr->daddr); 280 ipv6_addr_copy(&fl->fl6_src, &hdr->saddr); 281 282 while (pskb_may_pull(skb, skb->nh.raw + offset + 1 - skb->data)) { 283 exthdr = (struct ipv6_opt_hdr*)(skb->nh.raw + offset); 284 285 switch (nexthdr) { 286 case NEXTHDR_ROUTING: 287 case NEXTHDR_HOP: 288 case NEXTHDR_DEST: 289 offset += ipv6_optlen(exthdr); 290 nexthdr = exthdr->nexthdr; 291 exthdr = (struct ipv6_opt_hdr*)(skb->nh.raw + offset); 292 break; 293 294 case IPPROTO_UDP: 295 case IPPROTO_UDPLITE: 296 case IPPROTO_TCP: 297 case IPPROTO_SCTP: 298 case IPPROTO_DCCP: 299 if (pskb_may_pull(skb, skb->nh.raw + offset + 4 - skb->data)) { 300 __be16 *ports = (__be16 *)exthdr; 301 302 fl->fl_ip_sport = ports[0]; 303 fl->fl_ip_dport = ports[1]; 304 } 305 fl->proto = nexthdr; 306 return; 307 308 case IPPROTO_ICMPV6: 309 if (pskb_may_pull(skb, skb->nh.raw + offset + 2 - skb->data)) { 310 u8 *icmp = (u8 *)exthdr; 311 312 fl->fl_icmp_type = icmp[0]; 313 fl->fl_icmp_code = icmp[1]; 314 } 315 fl->proto = nexthdr; 316 return; 317 318 #ifdef CONFIG_IPV6_MIP6 319 case IPPROTO_MH: 320 if (pskb_may_pull(skb, skb->nh.raw + offset + 3 - skb->data)) { 321 struct ip6_mh *mh; 322 mh = (struct ip6_mh *)exthdr; 323 324 fl->fl_mh_type = mh->ip6mh_type; 325 } 326 fl->proto = nexthdr; 327 return; 328 #endif 329 330 /* XXX Why are there these headers? */ 331 case IPPROTO_AH: 332 case IPPROTO_ESP: 333 case IPPROTO_COMP: 334 default: 335 fl->fl_ipsec_spi = 0; 336 fl->proto = nexthdr; 337 return; 338 }; 339 } 340 } 341 342 static inline int xfrm6_garbage_collect(void) 343 { 344 xfrm6_policy_afinfo.garbage_collect(); 345 return (atomic_read(&xfrm6_dst_ops.entries) > xfrm6_dst_ops.gc_thresh*2); 346 } 347 348 static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu) 349 { 350 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 351 struct dst_entry *path = xdst->route; 352 353 path->ops->update_pmtu(path, mtu); 354 } 355 356 static void xfrm6_dst_destroy(struct dst_entry *dst) 357 { 358 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 359 360 if (likely(xdst->u.rt6.rt6i_idev)) 361 in6_dev_put(xdst->u.rt6.rt6i_idev); 362 xfrm_dst_destroy(xdst); 363 } 364 365 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev, 366 int unregister) 367 { 368 struct xfrm_dst *xdst; 369 370 if (!unregister) 371 return; 372 373 xdst = (struct xfrm_dst *)dst; 374 if (xdst->u.rt6.rt6i_idev->dev == dev) { 375 struct inet6_dev *loopback_idev = in6_dev_get(&loopback_dev); 376 BUG_ON(!loopback_idev); 377 378 do { 379 in6_dev_put(xdst->u.rt6.rt6i_idev); 380 xdst->u.rt6.rt6i_idev = loopback_idev; 381 in6_dev_hold(loopback_idev); 382 xdst = (struct xfrm_dst *)xdst->u.dst.child; 383 } while (xdst->u.dst.xfrm); 384 385 __in6_dev_put(loopback_idev); 386 } 387 388 xfrm_dst_ifdown(dst, dev); 389 } 390 391 static struct dst_ops xfrm6_dst_ops = { 392 .family = AF_INET6, 393 .protocol = __constant_htons(ETH_P_IPV6), 394 .gc = xfrm6_garbage_collect, 395 .update_pmtu = xfrm6_update_pmtu, 396 .destroy = xfrm6_dst_destroy, 397 .ifdown = xfrm6_dst_ifdown, 398 .gc_thresh = 1024, 399 .entry_size = sizeof(struct xfrm_dst), 400 }; 401 402 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = { 403 .family = AF_INET6, 404 .dst_ops = &xfrm6_dst_ops, 405 .dst_lookup = xfrm6_dst_lookup, 406 .get_saddr = xfrm6_get_saddr, 407 .find_bundle = __xfrm6_find_bundle, 408 .bundle_create = __xfrm6_bundle_create, 409 .decode_session = _decode_session6, 410 }; 411 412 static void __init xfrm6_policy_init(void) 413 { 414 xfrm_policy_register_afinfo(&xfrm6_policy_afinfo); 415 } 416 417 static void xfrm6_policy_fini(void) 418 { 419 xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo); 420 } 421 422 void __init xfrm6_init(void) 423 { 424 xfrm6_policy_init(); 425 xfrm6_state_init(); 426 } 427 428 void xfrm6_fini(void) 429 { 430 //xfrm6_input_fini(); 431 xfrm6_policy_fini(); 432 xfrm6_state_fini(); 433 } 434