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 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                 dev_put(dev);
 101                 return -ENODEV;
 102         }
 103
 104         xdst->u.rt6.rt6i_peer = rt->rt6i_peer;
 105         if (rt->rt6i_peer)
 106                 atomic_inc(&rt->rt6i_peer->refcnt);
 107
 108         /* Sheit... I remember I did this right. Apparently,
 109          * it was magically lost, so this code needs audit */
 110         xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
 111                                                    RTF_LOCAL);
 112         xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
 113         xdst->u.rt6.rt6i_node = rt->rt6i_node;
 114         if (rt->rt6i_node)
 115                 xdst->route_cookie = rt->rt6i_node->fn_sernum;
 116         xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
 117         xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
 118         xdst->u.rt6.rt6i_src = rt->rt6i_src;
 119
 120         return 0;
 121 }
 122
 123 static inline void
 124 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
 125 {
 126         struct flowi6 *fl6 = &fl->u.ip6;
 127         int onlyproto = 0;
 128         u16 offset = skb_network_header_len(skb);
 129         const struct ipv6hdr *hdr = ipv6_hdr(skb);
 130         struct ipv6_opt_hdr *exthdr;
 131         const unsigned char *nh = skb_network_header(skb);
 132         u8 nexthdr = nh[IP6CB(skb)->nhoff];
 133
 134         memset(fl6, 0, sizeof(struct flowi6));
 135         fl6->flowi6_mark = skb->mark;
 136
 137         ipv6_addr_copy(&fl6->daddr, reverse ? &hdr->saddr : &hdr->daddr);
 138         ipv6_addr_copy(&fl6->saddr, reverse ? &hdr->daddr : &hdr->saddr);
 139
 140         while (nh + offset + 1 < skb->data ||
 141                pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
 142                 nh = skb_network_header(skb);
 143                 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 144
 145                 switch (nexthdr) {
 146                 case NEXTHDR_FRAGMENT:
 147                         onlyproto = 1;
 148                 case NEXTHDR_ROUTING:
 149                 case NEXTHDR_HOP:
 150                 case NEXTHDR_DEST:
 151                         offset += ipv6_optlen(exthdr);
 152                         nexthdr = exthdr->nexthdr;
 153                         exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 154                         break;
 155
 156                 case IPPROTO_UDP:
 157                 case IPPROTO_UDPLITE:
 158                 case IPPROTO_TCP:
 159                 case IPPROTO_SCTP:
 160                 case IPPROTO_DCCP:
 161                         if (!onlyproto && (nh + offset + 4 < skb->data ||
 162                              pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
 163                                 __be16 *ports = (__be16 *)exthdr;
 164
 165                                 fl6->fl6_sport = ports[!!reverse];
 166                                 fl6->fl6_dport = ports[!reverse];
 167                         }
 168                         fl6->flowi6_proto = nexthdr;
 169                         return;
 170
 171                 case IPPROTO_ICMPV6:
 172                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
 173                                 u8 *icmp = (u8 *)exthdr;
 174
 175                                 fl6->fl6_icmp_type = icmp[0];
 176                                 fl6->fl6_icmp_code = icmp[1];
 177                         }
 178                         fl6->flowi6_proto = nexthdr;
 179                         return;
 180
 181 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
 182                 case IPPROTO_MH:
 183                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
 184                                 struct ip6_mh *mh;
 185                                 mh = (struct ip6_mh *)exthdr;
 186
 187                                 fl6->fl6_mh_type = mh->ip6mh_type;
 188                         }
 189                         fl6->flowi6_proto = nexthdr;
 190                         return;
 191 #endif
 192
 193                 /* XXX Why are there these headers? */
 194                 case IPPROTO_AH:
 195                 case IPPROTO_ESP:
 196                 case IPPROTO_COMP:
 197                 default:
 198                         fl6->fl6_ipsec_spi = 0;
 199                         fl6->flowi6_proto = nexthdr;
 200                         return;
 201                 }
 202         }
 203 }
 204
 205 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
 206 {
 207         struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
 208
 209         xfrm6_policy_afinfo.garbage_collect(net);
 210         return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
 211 }
 212
 213 static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu)
 214 {
 215         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 216         struct dst_entry *path = xdst->route;
 217
 218         path->ops->update_pmtu(path, mtu);
 219 }
 220
 221 static void xfrm6_dst_destroy(struct dst_entry *dst)
 222 {
 223         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 224
 225         if (likely(xdst->u.rt6.rt6i_idev))
 226                 in6_dev_put(xdst->u.rt6.rt6i_idev);
 227         dst_destroy_metrics_generic(dst);
 228         if (likely(xdst->u.rt6.rt6i_peer))
 229                 inet_putpeer(xdst->u.rt6.rt6i_peer);
 230         xfrm_dst_destroy(xdst);
 231 }
 232
 233 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 234                              int unregister)
 235 {
 236         struct xfrm_dst *xdst;
 237
 238         if (!unregister)
 239                 return;
 240
 241         xdst = (struct xfrm_dst *)dst;
 242         if (xdst->u.rt6.rt6i_idev->dev == dev) {
 243                 struct inet6_dev *loopback_idev =
 244                         in6_dev_get(dev_net(dev)->loopback_dev);
 245                 BUG_ON(!loopback_idev);
 246
 247                 do {
 248                         in6_dev_put(xdst->u.rt6.rt6i_idev);
 249                         xdst->u.rt6.rt6i_idev = loopback_idev;
 250                         in6_dev_hold(loopback_idev);
 251                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
 252                 } while (xdst->u.dst.xfrm);
 253
 254                 __in6_dev_put(loopback_idev);
 255         }
 256
 257         xfrm_dst_ifdown(dst, dev);
 258 }
 259
 260 static struct dst_ops xfrm6_dst_ops = {
 261         .family =               AF_INET6,
 262         .protocol =             cpu_to_be16(ETH_P_IPV6),
 263         .gc =                   xfrm6_garbage_collect,
 264         .update_pmtu =          xfrm6_update_pmtu,
 265         .cow_metrics =          dst_cow_metrics_generic,
 266         .destroy =              xfrm6_dst_destroy,
 267         .ifdown =               xfrm6_dst_ifdown,
 268         .local_out =            __ip6_local_out,
 269         .gc_thresh =            1024,
 270 };
 271
 272 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
 273         .family =               AF_INET6,
 274         .dst_ops =              &xfrm6_dst_ops,
 275         .dst_lookup =           xfrm6_dst_lookup,
 276         .get_saddr =            xfrm6_get_saddr,
 277         .decode_session =       _decode_session6,
 278         .get_tos =              xfrm6_get_tos,
 279         .init_path =            xfrm6_init_path,
 280         .fill_dst =             xfrm6_fill_dst,
 281         .blackhole_route =      ip6_blackhole_route,
 282 };
 283
 284 static int __init xfrm6_policy_init(void)
 285 {
 286         return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
 287 }
 288
 289 static void xfrm6_policy_fini(void)
 290 {
 291         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 292 }
 293
 294 #ifdef CONFIG_SYSCTL
 295 static struct ctl_table xfrm6_policy_table[] = {
 296         {
 297                 .procname       = "xfrm6_gc_thresh",
 298                 .data           = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
 299                 .maxlen         = sizeof(int),
 300                 .mode           = 0644,
 301                 .proc_handler   = proc_dointvec,
 302         },
 303         { }
 304 };
 305
 306 static struct ctl_table_header *sysctl_hdr;
 307 #endif
 308
 309 int __init xfrm6_init(void)
 310 {
 311         int ret;
 312         unsigned int gc_thresh;
 313
 314         /*
 315          * We need a good default value for the xfrm6 gc threshold.
 316          * In ipv4 we set it to the route hash table size * 8, which
 317          * is half the size of the maximaum route cache for ipv4.  It
 318          * would be good to do the same thing for v6, except the table is
 319          * constructed differently here.  Here each table for a net namespace
 320          * can have FIB_TABLE_HASHSZ entries, so lets go with the same
 321          * computation that we used for ipv4 here.  Also, lets keep the initial
 322          * gc_thresh to a minimum of 1024, since, the ipv6 route cache defaults
 323          * to that as a minimum as well
 324          */
 325         gc_thresh = FIB6_TABLE_HASHSZ * 8;
 326         xfrm6_dst_ops.gc_thresh = (gc_thresh < 1024) ? 1024 : gc_thresh;
 327         dst_entries_init(&xfrm6_dst_ops);
 328
 329         ret = xfrm6_policy_init();
 330         if (ret) {
 331                 dst_entries_destroy(&xfrm6_dst_ops);
 332                 goto out;
 333         }
 334         ret = xfrm6_state_init();
 335         if (ret)
 336                 goto out_policy;
 337
 338 #ifdef CONFIG_SYSCTL
 339         sysctl_hdr = register_net_sysctl_table(&init_net, net_ipv6_ctl_path,
 340                                                 xfrm6_policy_table);
 341 #endif
 342 out:
 343         return ret;
 344 out_policy:
 345         xfrm6_policy_fini();
 346         goto out;
 347 }
 348
 349 void xfrm6_fini(void)
 350 {
 351 #ifdef CONFIG_SYSCTL
 352         if (sysctl_hdr)
 353                 unregister_net_sysctl_table(sysctl_hdr);
 354 #endif
 355         //xfrm6_input_fini();
 356         xfrm6_policy_fini();
 357         xfrm6_state_fini();
 358         dst_entries_destroy(&xfrm6_dst_ops);
 359 }