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 defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) 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, 65 (struct in6_addr *)&daddr->a6, 0, 66 (struct in6_addr *)&saddr->a6); 67 dst_release(dst); 68 return 0; 69 } 70 71 static int xfrm6_get_tos(const struct flowi *fl) 72 { 73 return 0; 74 } 75 76 static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst, 77 int nfheader_len) 78 { 79 if (dst->ops->family == AF_INET6) { 80 struct rt6_info *rt = (struct rt6_info*)dst; 81 if (rt->rt6i_node) 82 path->path_cookie = rt->rt6i_node->fn_sernum; 83 } 84 85 path->u.rt6.rt6i_nfheader_len = nfheader_len; 86 87 return 0; 88 } 89 90 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev, 91 const struct flowi *fl) 92 { 93 struct rt6_info *rt = (struct rt6_info*)xdst->route; 94 95 xdst->u.dst.dev = dev; 96 dev_hold(dev); 97 98 xdst->u.rt6.rt6i_idev = in6_dev_get(dev); 99 if (!xdst->u.rt6.rt6i_idev) 100 return -ENODEV; 101 102 xdst->u.rt6.rt6i_peer = rt->rt6i_peer; 103 if (rt->rt6i_peer) 104 atomic_inc(&rt->rt6i_peer->refcnt); 105 106 /* Sheit... I remember I did this right. Apparently, 107 * it was magically lost, so this code needs audit */ 108 xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST | 109 RTF_LOCAL); 110 xdst->u.rt6.rt6i_metric = rt->rt6i_metric; 111 xdst->u.rt6.rt6i_node = rt->rt6i_node; 112 if (rt->rt6i_node) 113 xdst->route_cookie = rt->rt6i_node->fn_sernum; 114 xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway; 115 xdst->u.rt6.rt6i_dst = rt->rt6i_dst; 116 xdst->u.rt6.rt6i_src = rt->rt6i_src; 117 118 return 0; 119 } 120 121 static inline void 122 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse) 123 { 124 struct flowi6 *fl6 = &fl->u.ip6; 125 int onlyproto = 0; 126 u16 offset = skb_network_header_len(skb); 127 const struct ipv6hdr *hdr = ipv6_hdr(skb); 128 struct ipv6_opt_hdr *exthdr; 129 const unsigned char *nh = skb_network_header(skb); 130 u8 nexthdr = nh[IP6CB(skb)->nhoff]; 131 132 memset(fl6, 0, sizeof(struct flowi6)); 133 fl6->flowi6_mark = skb->mark; 134 135 fl6->daddr = reverse ? hdr->saddr : hdr->daddr; 136 fl6->saddr = reverse ? hdr->daddr : hdr->saddr; 137 138 while (nh + offset + 1 < skb->data || 139 pskb_may_pull(skb, nh + offset + 1 - skb->data)) { 140 nh = skb_network_header(skb); 141 exthdr = (struct ipv6_opt_hdr *)(nh + offset); 142 143 switch (nexthdr) { 144 case NEXTHDR_FRAGMENT: 145 onlyproto = 1; 146 case NEXTHDR_ROUTING: 147 case NEXTHDR_HOP: 148 case NEXTHDR_DEST: 149 offset += ipv6_optlen(exthdr); 150 nexthdr = exthdr->nexthdr; 151 exthdr = (struct ipv6_opt_hdr *)(nh + offset); 152 break; 153 154 case IPPROTO_UDP: 155 case IPPROTO_UDPLITE: 156 case IPPROTO_TCP: 157 case IPPROTO_SCTP: 158 case IPPROTO_DCCP: 159 if (!onlyproto && (nh + offset + 4 < skb->data || 160 pskb_may_pull(skb, nh + offset + 4 - skb->data))) { 161 __be16 *ports = (__be16 *)exthdr; 162 163 fl6->fl6_sport = ports[!!reverse]; 164 fl6->fl6_dport = ports[!reverse]; 165 } 166 fl6->flowi6_proto = nexthdr; 167 return; 168 169 case IPPROTO_ICMPV6: 170 if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) { 171 u8 *icmp = (u8 *)exthdr; 172 173 fl6->fl6_icmp_type = icmp[0]; 174 fl6->fl6_icmp_code = icmp[1]; 175 } 176 fl6->flowi6_proto = nexthdr; 177 return; 178 179 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) 180 case IPPROTO_MH: 181 if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) { 182 struct ip6_mh *mh; 183 mh = (struct ip6_mh *)exthdr; 184 185 fl6->fl6_mh_type = mh->ip6mh_type; 186 } 187 fl6->flowi6_proto = nexthdr; 188 return; 189 #endif 190 191 /* XXX Why are there these headers? */ 192 case IPPROTO_AH: 193 case IPPROTO_ESP: 194 case IPPROTO_COMP: 195 default: 196 fl6->fl6_ipsec_spi = 0; 197 fl6->flowi6_proto = nexthdr; 198 return; 199 } 200 } 201 } 202 203 static inline int xfrm6_garbage_collect(struct dst_ops *ops) 204 { 205 struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops); 206 207 xfrm6_policy_afinfo.garbage_collect(net); 208 return dst_entries_get_fast(ops) > ops->gc_thresh * 2; 209 } 210 211 static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu) 212 { 213 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 214 struct dst_entry *path = xdst->route; 215 216 path->ops->update_pmtu(path, mtu); 217 } 218 219 static void xfrm6_dst_destroy(struct dst_entry *dst) 220 { 221 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 222 223 if (likely(xdst->u.rt6.rt6i_idev)) 224 in6_dev_put(xdst->u.rt6.rt6i_idev); 225 dst_destroy_metrics_generic(dst); 226 if (likely(xdst->u.rt6.rt6i_peer)) 227 inet_putpeer(xdst->u.rt6.rt6i_peer); 228 xfrm_dst_destroy(xdst); 229 } 230 231 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev, 232 int unregister) 233 { 234 struct xfrm_dst *xdst; 235 236 if (!unregister) 237 return; 238 239 xdst = (struct xfrm_dst *)dst; 240 if (xdst->u.rt6.rt6i_idev->dev == dev) { 241 struct inet6_dev *loopback_idev = 242 in6_dev_get(dev_net(dev)->loopback_dev); 243 BUG_ON(!loopback_idev); 244 245 do { 246 in6_dev_put(xdst->u.rt6.rt6i_idev); 247 xdst->u.rt6.rt6i_idev = loopback_idev; 248 in6_dev_hold(loopback_idev); 249 xdst = (struct xfrm_dst *)xdst->u.dst.child; 250 } while (xdst->u.dst.xfrm); 251 252 __in6_dev_put(loopback_idev); 253 } 254 255 xfrm_dst_ifdown(dst, dev); 256 } 257 258 static struct dst_ops xfrm6_dst_ops = { 259 .family = AF_INET6, 260 .protocol = cpu_to_be16(ETH_P_IPV6), 261 .gc = xfrm6_garbage_collect, 262 .update_pmtu = xfrm6_update_pmtu, 263 .cow_metrics = dst_cow_metrics_generic, 264 .destroy = xfrm6_dst_destroy, 265 .ifdown = xfrm6_dst_ifdown, 266 .local_out = __ip6_local_out, 267 .gc_thresh = 1024, 268 }; 269 270 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = { 271 .family = AF_INET6, 272 .dst_ops = &xfrm6_dst_ops, 273 .dst_lookup = xfrm6_dst_lookup, 274 .get_saddr = xfrm6_get_saddr, 275 .decode_session = _decode_session6, 276 .get_tos = xfrm6_get_tos, 277 .init_path = xfrm6_init_path, 278 .fill_dst = xfrm6_fill_dst, 279 .blackhole_route = ip6_blackhole_route, 280 }; 281 282 static int __init xfrm6_policy_init(void) 283 { 284 return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo); 285 } 286 287 static void xfrm6_policy_fini(void) 288 { 289 xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo); 290 } 291 292 #ifdef CONFIG_SYSCTL 293 static struct ctl_table xfrm6_policy_table[] = { 294 { 295 .procname = "xfrm6_gc_thresh", 296 .data = &init_net.xfrm.xfrm6_dst_ops.gc_thresh, 297 .maxlen = sizeof(int), 298 .mode = 0644, 299 .proc_handler = proc_dointvec, 300 }, 301 { } 302 }; 303 304 static struct ctl_table_header *sysctl_hdr; 305 #endif 306 307 int __init xfrm6_init(void) 308 { 309 int ret; 310 unsigned int gc_thresh; 311 312 /* 313 * We need a good default value for the xfrm6 gc threshold. 314 * In ipv4 we set it to the route hash table size * 8, which 315 * is half the size of the maximaum route cache for ipv4. It 316 * would be good to do the same thing for v6, except the table is 317 * constructed differently here. Here each table for a net namespace 318 * can have FIB_TABLE_HASHSZ entries, so lets go with the same 319 * computation that we used for ipv4 here. Also, lets keep the initial 320 * gc_thresh to a minimum of 1024, since, the ipv6 route cache defaults 321 * to that as a minimum as well 322 */ 323 gc_thresh = FIB6_TABLE_HASHSZ * 8; 324 xfrm6_dst_ops.gc_thresh = (gc_thresh < 1024) ? 1024 : gc_thresh; 325 dst_entries_init(&xfrm6_dst_ops); 326 327 ret = xfrm6_policy_init(); 328 if (ret) { 329 dst_entries_destroy(&xfrm6_dst_ops); 330 goto out; 331 } 332 ret = xfrm6_state_init(); 333 if (ret) 334 goto out_policy; 335 336 #ifdef CONFIG_SYSCTL 337 sysctl_hdr = register_net_sysctl_table(&init_net, net_ipv6_ctl_path, 338 xfrm6_policy_table); 339 #endif 340 out: 341 return ret; 342 out_policy: 343 xfrm6_policy_fini(); 344 goto out; 345 } 346 347 void xfrm6_fini(void) 348 { 349 #ifdef CONFIG_SYSCTL 350 if (sysctl_hdr) 351 unregister_net_sysctl_table(sysctl_hdr); 352 #endif 353 //xfrm6_input_fini(); 354 xfrm6_policy_fini(); 355 xfrm6_state_fini(); 356 dst_entries_destroy(&xfrm6_dst_ops); 357 } 358