| 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 | xfrm_address_t *saddr, |
| 31 | xfrm_address_t *daddr) |
| 32 | { |
| 33 | struct flowi fl = {}; |
| 34 | struct dst_entry *dst; |
| 35 | int err; |
| 36 | |
| 37 | memcpy(&fl.fl6_dst, daddr, sizeof(fl.fl6_dst)); |
| 38 | if (saddr) |
| 39 | memcpy(&fl.fl6_src, saddr, sizeof(fl.fl6_src)); |
| 40 | |
| 41 | dst = ip6_route_output(net, NULL, &fl); |
| 42 | |
| 43 | err = dst->error; |
| 44 | if (dst->error) { |
| 45 | dst_release(dst); |
| 46 | dst = ERR_PTR(err); |
| 47 | } |
| 48 | |
| 49 | return dst; |
| 50 | } |
| 51 | |
| 52 | static int xfrm6_get_saddr(struct net *net, |
| 53 | xfrm_address_t *saddr, xfrm_address_t *daddr) |
| 54 | { |
| 55 | struct dst_entry *dst; |
| 56 | struct net_device *dev; |
| 57 | |
| 58 | dst = xfrm6_dst_lookup(net, 0, NULL, daddr); |
| 59 | if (IS_ERR(dst)) |
| 60 | return -EHOSTUNREACH; |
| 61 | |
| 62 | dev = ip6_dst_idev(dst)->dev; |
| 63 | ipv6_dev_get_saddr(dev_net(dev), dev, |
| 64 | (struct in6_addr *)&daddr->a6, 0, |
| 65 | (struct in6_addr *)&saddr->a6); |
| 66 | dst_release(dst); |
| 67 | return 0; |
| 68 | } |
| 69 | |
| 70 | static int xfrm6_get_tos(const struct flowi *fl) |
| 71 | { |
| 72 | return 0; |
| 73 | } |
| 74 | |
| 75 | static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst, |
| 76 | int nfheader_len) |
| 77 | { |
| 78 | if (dst->ops->family == AF_INET6) { |
| 79 | struct rt6_info *rt = (struct rt6_info*)dst; |
| 80 | if (rt->rt6i_node) |
| 81 | path->path_cookie = rt->rt6i_node->fn_sernum; |
| 82 | } |
| 83 | |
| 84 | path->u.rt6.rt6i_nfheader_len = nfheader_len; |
| 85 | |
| 86 | return 0; |
| 87 | } |
| 88 | |
| 89 | static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev, |
| 90 | struct flowi *fl) |
| 91 | { |
| 92 | struct rt6_info *rt = (struct rt6_info*)xdst->route; |
| 93 | |
| 94 | xdst->u.dst.dev = dev; |
| 95 | dev_hold(dev); |
| 96 | |
| 97 | xdst->u.rt6.rt6i_idev = in6_dev_get(dev); |
| 98 | if (!xdst->u.rt6.rt6i_idev) |
| 99 | return -ENODEV; |
| 100 | |
| 101 | xdst->u.rt6.rt6i_peer = rt->rt6i_peer; |
| 102 | if (rt->rt6i_peer) |
| 103 | atomic_inc(&rt->rt6i_peer->refcnt); |
| 104 | |
| 105 | /* Sheit... I remember I did this right. Apparently, |
| 106 | * it was magically lost, so this code needs audit */ |
| 107 | xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST | |
| 108 | RTF_LOCAL); |
| 109 | xdst->u.rt6.rt6i_metric = rt->rt6i_metric; |
| 110 | xdst->u.rt6.rt6i_node = rt->rt6i_node; |
| 111 | if (rt->rt6i_node) |
| 112 | xdst->route_cookie = rt->rt6i_node->fn_sernum; |
| 113 | xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway; |
| 114 | xdst->u.rt6.rt6i_dst = rt->rt6i_dst; |
| 115 | xdst->u.rt6.rt6i_src = rt->rt6i_src; |
| 116 | |
| 117 | return 0; |
| 118 | } |
| 119 | |
| 120 | static inline void |
| 121 | _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse) |
| 122 | { |
| 123 | int onlyproto = 0; |
| 124 | u16 offset = skb_network_header_len(skb); |
| 125 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
| 126 | struct ipv6_opt_hdr *exthdr; |
| 127 | const unsigned char *nh = skb_network_header(skb); |
| 128 | u8 nexthdr = nh[IP6CB(skb)->nhoff]; |
| 129 | |
| 130 | memset(fl, 0, sizeof(struct flowi)); |
| 131 | fl->mark = skb->mark; |
| 132 | |
| 133 | ipv6_addr_copy(&fl->fl6_dst, reverse ? &hdr->saddr : &hdr->daddr); |
| 134 | ipv6_addr_copy(&fl->fl6_src, reverse ? &hdr->daddr : &hdr->saddr); |
| 135 | |
| 136 | while (nh + offset + 1 < skb->data || |
| 137 | pskb_may_pull(skb, nh + offset + 1 - skb->data)) { |
| 138 | nh = skb_network_header(skb); |
| 139 | exthdr = (struct ipv6_opt_hdr *)(nh + offset); |
| 140 | |
| 141 | switch (nexthdr) { |
| 142 | case NEXTHDR_FRAGMENT: |
| 143 | onlyproto = 1; |
| 144 | case NEXTHDR_ROUTING: |
| 145 | case NEXTHDR_HOP: |
| 146 | case NEXTHDR_DEST: |
| 147 | offset += ipv6_optlen(exthdr); |
| 148 | nexthdr = exthdr->nexthdr; |
| 149 | exthdr = (struct ipv6_opt_hdr *)(nh + offset); |
| 150 | break; |
| 151 | |
| 152 | case IPPROTO_UDP: |
| 153 | case IPPROTO_UDPLITE: |
| 154 | case IPPROTO_TCP: |
| 155 | case IPPROTO_SCTP: |
| 156 | case IPPROTO_DCCP: |
| 157 | if (!onlyproto && (nh + offset + 4 < skb->data || |
| 158 | pskb_may_pull(skb, nh + offset + 4 - skb->data))) { |
| 159 | __be16 *ports = (__be16 *)exthdr; |
| 160 | |
| 161 | fl->fl_ip_sport = ports[!!reverse]; |
| 162 | fl->fl_ip_dport = ports[!reverse]; |
| 163 | } |
| 164 | fl->proto = nexthdr; |
| 165 | return; |
| 166 | |
| 167 | case IPPROTO_ICMPV6: |
| 168 | if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) { |
| 169 | u8 *icmp = (u8 *)exthdr; |
| 170 | |
| 171 | fl->fl_icmp_type = icmp[0]; |
| 172 | fl->fl_icmp_code = icmp[1]; |
| 173 | } |
| 174 | fl->proto = nexthdr; |
| 175 | return; |
| 176 | |
| 177 | #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) |
| 178 | case IPPROTO_MH: |
| 179 | if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) { |
| 180 | struct ip6_mh *mh; |
| 181 | mh = (struct ip6_mh *)exthdr; |
| 182 | |
| 183 | fl->fl_mh_type = mh->ip6mh_type; |
| 184 | } |
| 185 | fl->proto = nexthdr; |
| 186 | return; |
| 187 | #endif |
| 188 | |
| 189 | /* XXX Why are there these headers? */ |
| 190 | case IPPROTO_AH: |
| 191 | case IPPROTO_ESP: |
| 192 | case IPPROTO_COMP: |
| 193 | default: |
| 194 | fl->fl_ipsec_spi = 0; |
| 195 | fl->proto = nexthdr; |
| 196 | return; |
| 197 | } |
| 198 | } |
| 199 | } |
| 200 | |
| 201 | static inline int xfrm6_garbage_collect(struct dst_ops *ops) |
| 202 | { |
| 203 | struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops); |
| 204 | |
| 205 | xfrm6_policy_afinfo.garbage_collect(net); |
| 206 | return dst_entries_get_fast(ops) > ops->gc_thresh * 2; |
| 207 | } |
| 208 | |
| 209 | static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu) |
| 210 | { |
| 211 | struct xfrm_dst *xdst = (struct xfrm_dst *)dst; |
| 212 | struct dst_entry *path = xdst->route; |
| 213 | |
| 214 | path->ops->update_pmtu(path, mtu); |
| 215 | } |
| 216 | |
| 217 | static void xfrm6_dst_destroy(struct dst_entry *dst) |
| 218 | { |
| 219 | struct xfrm_dst *xdst = (struct xfrm_dst *)dst; |
| 220 | |
| 221 | if (likely(xdst->u.rt6.rt6i_idev)) |
| 222 | in6_dev_put(xdst->u.rt6.rt6i_idev); |
| 223 | dst_destroy_metrics_generic(dst); |
| 224 | if (likely(xdst->u.rt6.rt6i_peer)) |
| 225 | inet_putpeer(xdst->u.rt6.rt6i_peer); |
| 226 | xfrm_dst_destroy(xdst); |
| 227 | } |
| 228 | |
| 229 | static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev, |
| 230 | int unregister) |
| 231 | { |
| 232 | struct xfrm_dst *xdst; |
| 233 | |
| 234 | if (!unregister) |
| 235 | return; |
| 236 | |
| 237 | xdst = (struct xfrm_dst *)dst; |
| 238 | if (xdst->u.rt6.rt6i_idev->dev == dev) { |
| 239 | struct inet6_dev *loopback_idev = |
| 240 | in6_dev_get(dev_net(dev)->loopback_dev); |
| 241 | BUG_ON(!loopback_idev); |
| 242 | |
| 243 | do { |
| 244 | in6_dev_put(xdst->u.rt6.rt6i_idev); |
| 245 | xdst->u.rt6.rt6i_idev = loopback_idev; |
| 246 | in6_dev_hold(loopback_idev); |
| 247 | xdst = (struct xfrm_dst *)xdst->u.dst.child; |
| 248 | } while (xdst->u.dst.xfrm); |
| 249 | |
| 250 | __in6_dev_put(loopback_idev); |
| 251 | } |
| 252 | |
| 253 | xfrm_dst_ifdown(dst, dev); |
| 254 | } |
| 255 | |
| 256 | static struct dst_ops xfrm6_dst_ops = { |
| 257 | .family = AF_INET6, |
| 258 | .protocol = cpu_to_be16(ETH_P_IPV6), |
| 259 | .gc = xfrm6_garbage_collect, |
| 260 | .update_pmtu = xfrm6_update_pmtu, |
| 261 | .cow_metrics = dst_cow_metrics_generic, |
| 262 | .destroy = xfrm6_dst_destroy, |
| 263 | .ifdown = xfrm6_dst_ifdown, |
| 264 | .local_out = __ip6_local_out, |
| 265 | .gc_thresh = 1024, |
| 266 | }; |
| 267 | |
| 268 | static struct xfrm_policy_afinfo xfrm6_policy_afinfo = { |
| 269 | .family = AF_INET6, |
| 270 | .dst_ops = &xfrm6_dst_ops, |
| 271 | .dst_lookup = xfrm6_dst_lookup, |
| 272 | .get_saddr = xfrm6_get_saddr, |
| 273 | .decode_session = _decode_session6, |
| 274 | .get_tos = xfrm6_get_tos, |
| 275 | .init_path = xfrm6_init_path, |
| 276 | .fill_dst = xfrm6_fill_dst, |
| 277 | }; |
| 278 | |
| 279 | static int __init xfrm6_policy_init(void) |
| 280 | { |
| 281 | return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo); |
| 282 | } |
| 283 | |
| 284 | static void xfrm6_policy_fini(void) |
| 285 | { |
| 286 | xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo); |
| 287 | } |
| 288 | |
| 289 | #ifdef CONFIG_SYSCTL |
| 290 | static struct ctl_table xfrm6_policy_table[] = { |
| 291 | { |
| 292 | .procname = "xfrm6_gc_thresh", |
| 293 | .data = &init_net.xfrm.xfrm6_dst_ops.gc_thresh, |
| 294 | .maxlen = sizeof(int), |
| 295 | .mode = 0644, |
| 296 | .proc_handler = proc_dointvec, |
| 297 | }, |
| 298 | { } |
| 299 | }; |
| 300 | |
| 301 | static struct ctl_table_header *sysctl_hdr; |
| 302 | #endif |
| 303 | |
| 304 | int __init xfrm6_init(void) |
| 305 | { |
| 306 | int ret; |
| 307 | unsigned int gc_thresh; |
| 308 | |
| 309 | /* |
| 310 | * We need a good default value for the xfrm6 gc threshold. |
| 311 | * In ipv4 we set it to the route hash table size * 8, which |
| 312 | * is half the size of the maximaum route cache for ipv4. It |
| 313 | * would be good to do the same thing for v6, except the table is |
| 314 | * constructed differently here. Here each table for a net namespace |
| 315 | * can have FIB_TABLE_HASHSZ entries, so lets go with the same |
| 316 | * computation that we used for ipv4 here. Also, lets keep the initial |
| 317 | * gc_thresh to a minimum of 1024, since, the ipv6 route cache defaults |
| 318 | * to that as a minimum as well |
| 319 | */ |
| 320 | gc_thresh = FIB6_TABLE_HASHSZ * 8; |
| 321 | xfrm6_dst_ops.gc_thresh = (gc_thresh < 1024) ? 1024 : gc_thresh; |
| 322 | dst_entries_init(&xfrm6_dst_ops); |
| 323 | |
| 324 | ret = xfrm6_policy_init(); |
| 325 | if (ret) { |
| 326 | dst_entries_destroy(&xfrm6_dst_ops); |
| 327 | goto out; |
| 328 | } |
| 329 | ret = xfrm6_state_init(); |
| 330 | if (ret) |
| 331 | goto out_policy; |
| 332 | |
| 333 | #ifdef CONFIG_SYSCTL |
| 334 | sysctl_hdr = register_net_sysctl_table(&init_net, net_ipv6_ctl_path, |
| 335 | xfrm6_policy_table); |
| 336 | #endif |
| 337 | out: |
| 338 | return ret; |
| 339 | out_policy: |
| 340 | xfrm6_policy_fini(); |
| 341 | goto out; |
| 342 | } |
| 343 | |
| 344 | void xfrm6_fini(void) |
| 345 | { |
| 346 | #ifdef CONFIG_SYSCTL |
| 347 | if (sysctl_hdr) |
| 348 | unregister_net_sysctl_table(sysctl_hdr); |
| 349 | #endif |
| 350 | //xfrm6_input_fini(); |
| 351 | xfrm6_policy_fini(); |
| 352 | xfrm6_state_fini(); |
| 353 | dst_entries_destroy(&xfrm6_dst_ops); |
| 354 | } |