net/ipv6/xfrm6_policy.c

   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 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
  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 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
 106  * all the metrics... Shortly, bundle a bundle.
 107  */
 108
 109 static int
 110 __xfrm6_bundle_create(struct xfrm_policy *policy, struct xfrm_state **xfrm, int nx,
 111                       struct flowi *fl, struct dst_entry **dst_p)
 112 {
 113         struct dst_entry *dst, *dst_prev;
 114         struct rt6_info *rt0 = (struct rt6_info*)(*dst_p);
 115         struct rt6_info *rt  = rt0;
 116         struct flowi fl_tunnel = {
 117                 .nl_u = {
 118                         .ip6_u = {
 119                                 .saddr = fl->fl6_src,
 120                                 .daddr = fl->fl6_dst,
 121                         }
 122                 }
 123         };
 124         int i;
 125         int err = 0;
 126         int header_len = 0;
 127         int trailer_len = 0;
 128
 129         dst = dst_prev = NULL;
 130         dst_hold(&rt->u.dst);
 131
 132         for (i = 0; i < nx; i++) {
 133                 struct dst_entry *dst1 = dst_alloc(&xfrm6_dst_ops);
 134                 struct xfrm_dst *xdst;
 135
 136                 if (unlikely(dst1 == NULL)) {
 137                         err = -ENOBUFS;
 138                         dst_release(&rt->u.dst);
 139                         goto error;
 140                 }
 141
 142                 if (!dst)
 143                         dst = dst1;
 144                 else {
 145                         dst_prev->child = dst1;
 146                         dst1->flags |= DST_NOHASH;
 147                         dst_clone(dst1);
 148                 }
 149
 150                 xdst = (struct xfrm_dst *)dst1;
 151                 xdst->route = &rt->u.dst;
 152                 xdst->genid = xfrm[i]->genid;
 153                 if (rt->rt6i_node)
 154                         xdst->route_cookie = rt->rt6i_node->fn_sernum;
 155
 156                 dst1->next = dst_prev;
 157                 dst_prev = dst1;
 158
 159                 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
 160                         ((struct rt6_info *)dst)->nfheader_len +=
 161                                 xfrm[i]->props.header_len;
 162                 header_len += xfrm[i]->props.header_len;
 163                 trailer_len += xfrm[i]->props.trailer_len;
 164
 165                 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
 166                         unsigned short encap_family = xfrm[i]->props.family;
 167                         switch(encap_family) {
 168                         case AF_INET:
 169                                 fl_tunnel.fl4_dst = xfrm[i]->id.daddr.a4;
 170                                 fl_tunnel.fl4_src = xfrm[i]->props.saddr.a4;
 171                                 break;
 172                         case AF_INET6:
 173                                 ipv6_addr_copy(&fl_tunnel.fl6_dst, __xfrm6_bundle_addr_remote(xfrm[i], &fl->fl6_dst));
 174
 175                                 ipv6_addr_copy(&fl_tunnel.fl6_src, __xfrm6_bundle_addr_local(xfrm[i], &fl->fl6_src));
 176                                 break;
 177                         default:
 178                                 BUG_ON(1);
 179                         }
 180
 181                         err = xfrm_dst_lookup((struct xfrm_dst **) &rt,
 182                                               &fl_tunnel, encap_family);
 183                         if (err)
 184                                 goto error;
 185                 } else
 186                         dst_hold(&rt->u.dst);
 187         }
 188
 189         dst_prev->child = &rt->u.dst;
 190         dst->path = &rt->u.dst;
 191
 192         /* Copy neighbour for reachability confirmation */
 193         dst->neighbour = neigh_clone(rt->u.dst.neighbour);
 194
 195         if (rt->rt6i_node)
 196                 ((struct xfrm_dst *)dst)->path_cookie = rt->rt6i_node->fn_sernum;
 197
 198         *dst_p = dst;
 199         dst = dst_prev;
 200
 201         dst_prev = *dst_p;
 202         i = 0;
 203         for (; dst_prev != &rt->u.dst; dst_prev = dst_prev->child) {
 204                 struct xfrm_dst *x = (struct xfrm_dst*)dst_prev;
 205
 206                 dst_prev->xfrm = xfrm[i++];
 207                 dst_prev->dev = rt->u.dst.dev;
 208                 if (rt->u.dst.dev)
 209                         dev_hold(rt->u.dst.dev);
 210                 dst_prev->obsolete      = -1;
 211                 dst_prev->flags        |= DST_HOST;
 212                 dst_prev->lastuse       = jiffies;
 213                 dst_prev->header_len    = header_len;
 214                 dst_prev->trailer_len   = trailer_len;
 215                 memcpy(&dst_prev->metrics, &x->route->metrics, sizeof(dst_prev->metrics));
 216
 217                 dst_prev->input         = rt->u.dst.input;
 218                 dst_prev->output = dst_prev->xfrm->outer_mode->afinfo->output;
 219                 /* Sheit... I remember I did this right. Apparently,
 220                  * it was magically lost, so this code needs audit */
 221                 x->u.rt6.rt6i_flags    = rt0->rt6i_flags&(RTF_ANYCAST|RTF_LOCAL);
 222                 x->u.rt6.rt6i_metric   = rt0->rt6i_metric;
 223                 x->u.rt6.rt6i_node     = rt0->rt6i_node;
 224                 x->u.rt6.rt6i_gateway  = rt0->rt6i_gateway;
 225                 memcpy(&x->u.rt6.rt6i_gateway, &rt0->rt6i_gateway, sizeof(x->u.rt6.rt6i_gateway));
 226                 x->u.rt6.rt6i_dst      = rt0->rt6i_dst;
 227                 x->u.rt6.rt6i_src      = rt0->rt6i_src;
 228                 x->u.rt6.rt6i_idev     = rt0->rt6i_idev;
 229                 in6_dev_hold(rt0->rt6i_idev);
 230                 header_len -= x->u.dst.xfrm->props.header_len;
 231                 trailer_len -= x->u.dst.xfrm->props.trailer_len;
 232         }
 233
 234         xfrm_init_pmtu(dst);
 235         return 0;
 236
 237 error:
 238         if (dst)
 239                 dst_free(dst);
 240         return err;
 241 }
 242
 243 static inline void
 244 _decode_session6(struct sk_buff *skb, struct flowi *fl)
 245 {
 246         u16 offset = skb_network_header_len(skb);
 247         struct ipv6hdr *hdr = ipv6_hdr(skb);
 248         struct ipv6_opt_hdr *exthdr;
 249         const unsigned char *nh = skb_network_header(skb);
 250         u8 nexthdr = nh[IP6CB(skb)->nhoff];
 251
 252         memset(fl, 0, sizeof(struct flowi));
 253         ipv6_addr_copy(&fl->fl6_dst, &hdr->daddr);
 254         ipv6_addr_copy(&fl->fl6_src, &hdr->saddr);
 255
 256         while (pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
 257                 nh = skb_network_header(skb);
 258                 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 259
 260                 switch (nexthdr) {
 261                 case NEXTHDR_ROUTING:
 262                 case NEXTHDR_HOP:
 263                 case NEXTHDR_DEST:
 264                         offset += ipv6_optlen(exthdr);
 265                         nexthdr = exthdr->nexthdr;
 266                         exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 267                         break;
 268
 269                 case IPPROTO_UDP:
 270                 case IPPROTO_UDPLITE:
 271                 case IPPROTO_TCP:
 272                 case IPPROTO_SCTP:
 273                 case IPPROTO_DCCP:
 274                         if (pskb_may_pull(skb, nh + offset + 4 - skb->data)) {
 275                                 __be16 *ports = (__be16 *)exthdr;
 276
 277                                 fl->fl_ip_sport = ports[0];
 278                                 fl->fl_ip_dport = ports[1];
 279                         }
 280                         fl->proto = nexthdr;
 281                         return;
 282
 283                 case IPPROTO_ICMPV6:
 284                         if (pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
 285                                 u8 *icmp = (u8 *)exthdr;
 286
 287                                 fl->fl_icmp_type = icmp[0];
 288                                 fl->fl_icmp_code = icmp[1];
 289                         }
 290                         fl->proto = nexthdr;
 291                         return;
 292
 293 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
 294                 case IPPROTO_MH:
 295                         if (pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
 296                                 struct ip6_mh *mh;
 297                                 mh = (struct ip6_mh *)exthdr;
 298
 299                                 fl->fl_mh_type = mh->ip6mh_type;
 300                         }
 301                         fl->proto = nexthdr;
 302                         return;
 303 #endif
 304
 305                 /* XXX Why are there these headers? */
 306                 case IPPROTO_AH:
 307                 case IPPROTO_ESP:
 308                 case IPPROTO_COMP:
 309                 default:
 310                         fl->fl_ipsec_spi = 0;
 311                         fl->proto = nexthdr;
 312                         return;
 313                 }
 314         }
 315 }
 316
 317 static inline int xfrm6_garbage_collect(void)
 318 {
 319         xfrm6_policy_afinfo.garbage_collect();
 320         return (atomic_read(&xfrm6_dst_ops.entries) > xfrm6_dst_ops.gc_thresh*2);
 321 }
 322
 323 static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu)
 324 {
 325         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 326         struct dst_entry *path = xdst->route;
 327
 328         path->ops->update_pmtu(path, mtu);
 329 }
 330
 331 static void xfrm6_dst_destroy(struct dst_entry *dst)
 332 {
 333         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 334
 335         if (likely(xdst->u.rt6.rt6i_idev))
 336                 in6_dev_put(xdst->u.rt6.rt6i_idev);
 337         xfrm_dst_destroy(xdst);
 338 }
 339
 340 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 341                              int unregister)
 342 {
 343         struct xfrm_dst *xdst;
 344
 345         if (!unregister)
 346                 return;
 347
 348         xdst = (struct xfrm_dst *)dst;
 349         if (xdst->u.rt6.rt6i_idev->dev == dev) {
 350                 struct inet6_dev *loopback_idev = in6_dev_get(init_net.loopback_dev);
 351                 BUG_ON(!loopback_idev);
 352
 353                 do {
 354                         in6_dev_put(xdst->u.rt6.rt6i_idev);
 355                         xdst->u.rt6.rt6i_idev = loopback_idev;
 356                         in6_dev_hold(loopback_idev);
 357                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
 358                 } while (xdst->u.dst.xfrm);
 359
 360                 __in6_dev_put(loopback_idev);
 361         }
 362
 363         xfrm_dst_ifdown(dst, dev);
 364 }
 365
 366 static struct dst_ops xfrm6_dst_ops = {
 367         .family =               AF_INET6,
 368         .protocol =             __constant_htons(ETH_P_IPV6),
 369         .gc =                   xfrm6_garbage_collect,
 370         .update_pmtu =          xfrm6_update_pmtu,
 371         .destroy =              xfrm6_dst_destroy,
 372         .ifdown =               xfrm6_dst_ifdown,
 373         .gc_thresh =            1024,
 374         .entry_size =           sizeof(struct xfrm_dst),
 375 };
 376
 377 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
 378         .family =               AF_INET6,
 379         .dst_ops =              &xfrm6_dst_ops,
 380         .dst_lookup =           xfrm6_dst_lookup,
 381         .get_saddr =            xfrm6_get_saddr,
 382         .find_bundle =          __xfrm6_find_bundle,
 383         .bundle_create =        __xfrm6_bundle_create,
 384         .decode_session =       _decode_session6,
 385 };
 386
 387 static void __init xfrm6_policy_init(void)
 388 {
 389         xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
 390 }
 391
 392 static void xfrm6_policy_fini(void)
 393 {
 394         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 395 }
 396
 397 void __init xfrm6_init(void)
 398 {
 399         xfrm6_policy_init();
 400         xfrm6_state_init();
 401 }
 402
 403 void xfrm6_fini(void)
 404 {
 405         //xfrm6_input_fini();
 406         xfrm6_policy_fini();
 407         xfrm6_state_fini();
 408 }