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/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 void xfrm6_init_dst(struct net *net, struct xfrm_dst *xdst)
  77 {
  78         struct rt6_info *rt = (struct rt6_info *)xdst;
  79
  80         rt6_init_peer(rt, net->ipv6.peers);
  81 }
  82
  83 static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
  84                            int nfheader_len)
  85 {
  86         if (dst->ops->family == AF_INET6) {
  87                 struct rt6_info *rt = (struct rt6_info*)dst;
  88                 if (rt->rt6i_node)
  89                         path->path_cookie = rt->rt6i_node->fn_sernum;
  90         }
  91
  92         path->u.rt6.rt6i_nfheader_len = nfheader_len;
  93
  94         return 0;
  95 }
  96
  97 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
  98                           const struct flowi *fl)
  99 {
 100         struct rt6_info *rt = (struct rt6_info*)xdst->route;
 101
 102         xdst->u.dst.dev = dev;
 103         dev_hold(dev);
 104
 105         xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
 106         if (!xdst->u.rt6.rt6i_idev)
 107                 return -ENODEV;
 108
 109         rt6_transfer_peer(&xdst->u.rt6, rt);
 110
 111         /* Sheit... I remember I did this right. Apparently,
 112          * it was magically lost, so this code needs audit */
 113         xdst->u.rt6.n = neigh_clone(rt->n);
 114         xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
 115                                                    RTF_LOCAL);
 116         xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
 117         xdst->u.rt6.rt6i_node = rt->rt6i_node;
 118         if (rt->rt6i_node)
 119                 xdst->route_cookie = rt->rt6i_node->fn_sernum;
 120         xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
 121         xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
 122         xdst->u.rt6.rt6i_src = rt->rt6i_src;
 123
 124         return 0;
 125 }
 126
 127 static inline void
 128 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
 129 {
 130         struct flowi6 *fl6 = &fl->u.ip6;
 131         int onlyproto = 0;
 132         u16 offset = skb_network_header_len(skb);
 133         const struct ipv6hdr *hdr = ipv6_hdr(skb);
 134         struct ipv6_opt_hdr *exthdr;
 135         const unsigned char *nh = skb_network_header(skb);
 136         u8 nexthdr = nh[IP6CB(skb)->nhoff];
 137
 138         memset(fl6, 0, sizeof(struct flowi6));
 139         fl6->flowi6_mark = skb->mark;
 140
 141         fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
 142         fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
 143
 144         while (nh + offset + 1 < skb->data ||
 145                pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
 146                 nh = skb_network_header(skb);
 147                 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 148
 149                 switch (nexthdr) {
 150                 case NEXTHDR_FRAGMENT:
 151                         onlyproto = 1;
 152                 case NEXTHDR_ROUTING:
 153                 case NEXTHDR_HOP:
 154                 case NEXTHDR_DEST:
 155                         offset += ipv6_optlen(exthdr);
 156                         nexthdr = exthdr->nexthdr;
 157                         exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 158                         break;
 159
 160                 case IPPROTO_UDP:
 161                 case IPPROTO_UDPLITE:
 162                 case IPPROTO_TCP:
 163                 case IPPROTO_SCTP:
 164                 case IPPROTO_DCCP:
 165                         if (!onlyproto && (nh + offset + 4 < skb->data ||
 166                              pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
 167                                 __be16 *ports = (__be16 *)exthdr;
 168
 169                                 fl6->fl6_sport = ports[!!reverse];
 170                                 fl6->fl6_dport = ports[!reverse];
 171                         }
 172                         fl6->flowi6_proto = nexthdr;
 173                         return;
 174
 175                 case IPPROTO_ICMPV6:
 176                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
 177                                 u8 *icmp = (u8 *)exthdr;
 178
 179                                 fl6->fl6_icmp_type = icmp[0];
 180                                 fl6->fl6_icmp_code = icmp[1];
 181                         }
 182                         fl6->flowi6_proto = nexthdr;
 183                         return;
 184
 185 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
 186                 case IPPROTO_MH:
 187                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
 188                                 struct ip6_mh *mh;
 189                                 mh = (struct ip6_mh *)exthdr;
 190
 191                                 fl6->fl6_mh_type = mh->ip6mh_type;
 192                         }
 193                         fl6->flowi6_proto = nexthdr;
 194                         return;
 195 #endif
 196
 197                 /* XXX Why are there these headers? */
 198                 case IPPROTO_AH:
 199                 case IPPROTO_ESP:
 200                 case IPPROTO_COMP:
 201                 default:
 202                         fl6->fl6_ipsec_spi = 0;
 203                         fl6->flowi6_proto = nexthdr;
 204                         return;
 205                 }
 206         }
 207 }
 208
 209 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
 210 {
 211         struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
 212
 213         xfrm6_policy_afinfo.garbage_collect(net);
 214         return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
 215 }
 216
 217 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
 218                               struct sk_buff *skb, u32 mtu)
 219 {
 220         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 221         struct dst_entry *path = xdst->route;
 222
 223         path->ops->update_pmtu(path, sk, skb, mtu);
 224 }
 225
 226 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
 227                            struct sk_buff *skb)
 228 {
 229         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 230         struct dst_entry *path = xdst->route;
 231
 232         path->ops->redirect(path, sk, skb);
 233 }
 234
 235 static void xfrm6_dst_destroy(struct dst_entry *dst)
 236 {
 237         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 238
 239         if (likely(xdst->u.rt6.rt6i_idev))
 240                 in6_dev_put(xdst->u.rt6.rt6i_idev);
 241         dst_destroy_metrics_generic(dst);
 242         if (rt6_has_peer(&xdst->u.rt6)) {
 243                 struct inet_peer *peer = rt6_peer_ptr(&xdst->u.rt6);
 244                 inet_putpeer(peer);
 245         }
 246         xfrm_dst_destroy(xdst);
 247 }
 248
 249 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 250                              int unregister)
 251 {
 252         struct xfrm_dst *xdst;
 253
 254         if (!unregister)
 255                 return;
 256
 257         xdst = (struct xfrm_dst *)dst;
 258         if (xdst->u.rt6.rt6i_idev->dev == dev) {
 259                 struct inet6_dev *loopback_idev =
 260                         in6_dev_get(dev_net(dev)->loopback_dev);
 261                 BUG_ON(!loopback_idev);
 262
 263                 do {
 264                         in6_dev_put(xdst->u.rt6.rt6i_idev);
 265                         xdst->u.rt6.rt6i_idev = loopback_idev;
 266                         in6_dev_hold(loopback_idev);
 267                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
 268                 } while (xdst->u.dst.xfrm);
 269
 270                 __in6_dev_put(loopback_idev);
 271         }
 272
 273         xfrm_dst_ifdown(dst, dev);
 274 }
 275
 276 static struct dst_ops xfrm6_dst_ops = {
 277         .family =               AF_INET6,
 278         .protocol =             cpu_to_be16(ETH_P_IPV6),
 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 =            1024,
 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_dst =             xfrm6_init_dst,
 297         .init_path =            xfrm6_init_path,
 298         .fill_dst =             xfrm6_fill_dst,
 299         .blackhole_route =      ip6_blackhole_route,
 300 };
 301
 302 static int __init xfrm6_policy_init(void)
 303 {
 304         return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
 305 }
 306
 307 static void xfrm6_policy_fini(void)
 308 {
 309         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 310 }
 311
 312 #ifdef CONFIG_SYSCTL
 313 static struct ctl_table xfrm6_policy_table[] = {
 314         {
 315                 .procname       = "xfrm6_gc_thresh",
 316                 .data           = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
 317                 .maxlen         = sizeof(int),
 318                 .mode           = 0644,
 319                 .proc_handler   = proc_dointvec,
 320         },
 321         { }
 322 };
 323
 324 static struct ctl_table_header *sysctl_hdr;
 325 #endif
 326
 327 int __init xfrm6_init(void)
 328 {
 329         int ret;
 330         unsigned int gc_thresh;
 331
 332         /*
 333          * We need a good default value for the xfrm6 gc threshold.
 334          * In ipv4 we set it to the route hash table size * 8, which
 335          * is half the size of the maximaum route cache for ipv4.  It
 336          * would be good to do the same thing for v6, except the table is
 337          * constructed differently here.  Here each table for a net namespace
 338          * can have FIB_TABLE_HASHSZ entries, so lets go with the same
 339          * computation that we used for ipv4 here.  Also, lets keep the initial
 340          * gc_thresh to a minimum of 1024, since, the ipv6 route cache defaults
 341          * to that as a minimum as well
 342          */
 343         gc_thresh = FIB6_TABLE_HASHSZ * 8;
 344         xfrm6_dst_ops.gc_thresh = (gc_thresh < 1024) ? 1024 : gc_thresh;
 345         dst_entries_init(&xfrm6_dst_ops);
 346
 347         ret = xfrm6_policy_init();
 348         if (ret) {
 349                 dst_entries_destroy(&xfrm6_dst_ops);
 350                 goto out;
 351         }
 352         ret = xfrm6_state_init();
 353         if (ret)
 354                 goto out_policy;
 355
 356 #ifdef CONFIG_SYSCTL
 357         sysctl_hdr = register_net_sysctl(&init_net, "net/ipv6",
 358                                          xfrm6_policy_table);
 359 #endif
 360 out:
 361         return ret;
 362 out_policy:
 363         xfrm6_policy_fini();
 364         goto out;
 365 }
 366
 367 void xfrm6_fini(void)
 368 {
 369 #ifdef CONFIG_SYSCTL
 370         if (sysctl_hdr)
 371                 unregister_net_sysctl_table(sysctl_hdr);
 372 #endif
 373         //xfrm6_input_fini();
 374         xfrm6_policy_fini();
 375         xfrm6_state_fini();
 376         dst_entries_destroy(&xfrm6_dst_ops);
 377 }