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                                           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 struct dst_entry *
  71 __xfrm6_find_bundle(struct flowi *fl, struct xfrm_policy *policy)
  72 {
  73         struct dst_entry *dst;
  74
  75         /* Still not clear if we should set fl->fl6_{src,dst}... */
  76         read_lock_bh(&policy->lock);
  77         for (dst = policy->bundles; dst; dst = dst->next) {
  78                 struct xfrm_dst *xdst = (struct xfrm_dst*)dst;
  79                 struct in6_addr fl_dst_prefix, fl_src_prefix;
  80
  81                 ipv6_addr_prefix(&fl_dst_prefix,
  82                                  &fl->fl6_dst,
  83                                  xdst->u.rt6.rt6i_dst.plen);
  84                 ipv6_addr_prefix(&fl_src_prefix,
  85                                  &fl->fl6_src,
  86                                  xdst->u.rt6.rt6i_src.plen);
  87                 if (ipv6_addr_equal(&xdst->u.rt6.rt6i_dst.addr, &fl_dst_prefix) &&
  88                     ipv6_addr_equal(&xdst->u.rt6.rt6i_src.addr, &fl_src_prefix) &&
  89                     xfrm_bundle_ok(policy, xdst, fl, AF_INET6,
  90                                    (xdst->u.rt6.rt6i_dst.plen != 128 ||
  91                                     xdst->u.rt6.rt6i_src.plen != 128))) {
  92                         dst_clone(dst);
  93                         break;
  94                 }
  95         }
  96         read_unlock_bh(&policy->lock);
  97         return dst;
  98 }
  99
 100 static int xfrm6_get_tos(struct flowi *fl)
 101 {
 102         return 0;
 103 }
 104
 105 static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
 106                            int nfheader_len)
 107 {
 108         if (dst->ops->family == AF_INET6) {
 109                 struct rt6_info *rt = (struct rt6_info*)dst;
 110                 if (rt->rt6i_node)
 111                         path->path_cookie = rt->rt6i_node->fn_sernum;
 112         }
 113
 114         path->u.rt6.rt6i_nfheader_len = nfheader_len;
 115
 116         return 0;
 117 }
 118
 119 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
 120                           struct flowi *fl)
 121 {
 122         struct rt6_info *rt = (struct rt6_info*)xdst->route;
 123
 124         xdst->u.dst.dev = dev;
 125         dev_hold(dev);
 126
 127         xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
 128         if (!xdst->u.rt6.rt6i_idev)
 129                 return -ENODEV;
 130
 131         /* Sheit... I remember I did this right. Apparently,
 132          * it was magically lost, so this code needs audit */
 133         xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
 134                                                    RTF_LOCAL);
 135         xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
 136         xdst->u.rt6.rt6i_node = rt->rt6i_node;
 137         if (rt->rt6i_node)
 138                 xdst->route_cookie = rt->rt6i_node->fn_sernum;
 139         xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
 140         xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
 141         xdst->u.rt6.rt6i_src = rt->rt6i_src;
 142
 143         return 0;
 144 }
 145
 146 static inline void
 147 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
 148 {
 149         int onlyproto = 0;
 150         u16 offset = skb_network_header_len(skb);
 151         struct ipv6hdr *hdr = ipv6_hdr(skb);
 152         struct ipv6_opt_hdr *exthdr;
 153         const unsigned char *nh = skb_network_header(skb);
 154         u8 nexthdr = nh[IP6CB(skb)->nhoff];
 155
 156         memset(fl, 0, sizeof(struct flowi));
 157         ipv6_addr_copy(&fl->fl6_dst, reverse ? &hdr->saddr : &hdr->daddr);
 158         ipv6_addr_copy(&fl->fl6_src, reverse ? &hdr->daddr : &hdr->saddr);
 159
 160         while (nh + offset + 1 < skb->data ||
 161                pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
 162                 nh = skb_network_header(skb);
 163                 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 164
 165                 switch (nexthdr) {
 166                 case NEXTHDR_FRAGMENT:
 167                         onlyproto = 1;
 168                 case NEXTHDR_ROUTING:
 169                 case NEXTHDR_HOP:
 170                 case NEXTHDR_DEST:
 171                         offset += ipv6_optlen(exthdr);
 172                         nexthdr = exthdr->nexthdr;
 173                         exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 174                         break;
 175
 176                 case IPPROTO_UDP:
 177                 case IPPROTO_UDPLITE:
 178                 case IPPROTO_TCP:
 179                 case IPPROTO_SCTP:
 180                 case IPPROTO_DCCP:
 181                         if (!onlyproto && (nh + offset + 4 < skb->data ||
 182                              pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
 183                                 __be16 *ports = (__be16 *)exthdr;
 184
 185                                 fl->fl_ip_sport = ports[!!reverse];
 186                                 fl->fl_ip_dport = ports[!reverse];
 187                         }
 188                         fl->proto = nexthdr;
 189                         return;
 190
 191                 case IPPROTO_ICMPV6:
 192                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
 193                                 u8 *icmp = (u8 *)exthdr;
 194
 195                                 fl->fl_icmp_type = icmp[0];
 196                                 fl->fl_icmp_code = icmp[1];
 197                         }
 198                         fl->proto = nexthdr;
 199                         return;
 200
 201 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
 202                 case IPPROTO_MH:
 203                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
 204                                 struct ip6_mh *mh;
 205                                 mh = (struct ip6_mh *)exthdr;
 206
 207                                 fl->fl_mh_type = mh->ip6mh_type;
 208                         }
 209                         fl->proto = nexthdr;
 210                         return;
 211 #endif
 212
 213                 /* XXX Why are there these headers? */
 214                 case IPPROTO_AH:
 215                 case IPPROTO_ESP:
 216                 case IPPROTO_COMP:
 217                 default:
 218                         fl->fl_ipsec_spi = 0;
 219                         fl->proto = nexthdr;
 220                         return;
 221                 }
 222         }
 223 }
 224
 225 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
 226 {
 227         struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
 228
 229         xfrm6_policy_afinfo.garbage_collect(net);
 230         return (atomic_read(&ops->entries) > ops->gc_thresh * 2);
 231 }
 232
 233 static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu)
 234 {
 235         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 236         struct dst_entry *path = xdst->route;
 237
 238         path->ops->update_pmtu(path, mtu);
 239 }
 240
 241 static void xfrm6_dst_destroy(struct dst_entry *dst)
 242 {
 243         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 244
 245         if (likely(xdst->u.rt6.rt6i_idev))
 246                 in6_dev_put(xdst->u.rt6.rt6i_idev);
 247         xfrm_dst_destroy(xdst);
 248 }
 249
 250 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 251                              int unregister)
 252 {
 253         struct xfrm_dst *xdst;
 254
 255         if (!unregister)
 256                 return;
 257
 258         xdst = (struct xfrm_dst *)dst;
 259         if (xdst->u.rt6.rt6i_idev->dev == dev) {
 260                 struct inet6_dev *loopback_idev =
 261                         in6_dev_get(dev_net(dev)->loopback_dev);
 262                 BUG_ON(!loopback_idev);
 263
 264                 do {
 265                         in6_dev_put(xdst->u.rt6.rt6i_idev);
 266                         xdst->u.rt6.rt6i_idev = loopback_idev;
 267                         in6_dev_hold(loopback_idev);
 268                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
 269                 } while (xdst->u.dst.xfrm);
 270
 271                 __in6_dev_put(loopback_idev);
 272         }
 273
 274         xfrm_dst_ifdown(dst, dev);
 275 }
 276
 277 static struct dst_ops xfrm6_dst_ops = {
 278         .family =               AF_INET6,
 279         .protocol =             cpu_to_be16(ETH_P_IPV6),
 280         .gc =                   xfrm6_garbage_collect,
 281         .update_pmtu =          xfrm6_update_pmtu,
 282         .destroy =              xfrm6_dst_destroy,
 283         .ifdown =               xfrm6_dst_ifdown,
 284         .local_out =            __ip6_local_out,
 285         .gc_thresh =            1024,
 286         .entries =              ATOMIC_INIT(0),
 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         .find_bundle =          __xfrm6_find_bundle,
 295         .decode_session =       _decode_session6,
 296         .get_tos =              xfrm6_get_tos,
 297         .init_path =            xfrm6_init_path,
 298         .fill_dst =             xfrm6_fill_dst,
 299 };
 300
 301 static int __init xfrm6_policy_init(void)
 302 {
 303         return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
 304 }
 305
 306 static void xfrm6_policy_fini(void)
 307 {
 308         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 309 }
 310
 311 #ifdef CONFIG_SYSCTL
 312 static struct ctl_table xfrm6_policy_table[] = {
 313         {
 314                 .procname       = "xfrm6_gc_thresh",
 315                 .data           = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
 316                 .maxlen         = sizeof(int),
 317                 .mode           = 0644,
 318                 .proc_handler   = proc_dointvec,
 319         },
 320         { }
 321 };
 322
 323 static struct ctl_table_header *sysctl_hdr;
 324 #endif
 325
 326 int __init xfrm6_init(void)
 327 {
 328         int ret;
 329         unsigned int gc_thresh;
 330
 331         /*
 332          * We need a good default value for the xfrm6 gc threshold.
 333          * In ipv4 we set it to the route hash table size * 8, which
 334          * is half the size of the maximaum route cache for ipv4.  It
 335          * would be good to do the same thing for v6, except the table is
 336          * constructed differently here.  Here each table for a net namespace
 337          * can have FIB_TABLE_HASHSZ entries, so lets go with the same
 338          * computation that we used for ipv4 here.  Also, lets keep the initial
 339          * gc_thresh to a minimum of 1024, since, the ipv6 route cache defaults
 340          * to that as a minimum as well
 341          */
 342         gc_thresh = FIB6_TABLE_HASHSZ * 8;
 343         xfrm6_dst_ops.gc_thresh = (gc_thresh < 1024) ? 1024 : gc_thresh;
 344
 345         ret = xfrm6_policy_init();
 346         if (ret)
 347                 goto out;
 348
 349         ret = xfrm6_state_init();
 350         if (ret)
 351                 goto out_policy;
 352
 353 #ifdef CONFIG_SYSCTL
 354         sysctl_hdr = register_net_sysctl_table(&init_net, net_ipv6_ctl_path,
 355                                                 xfrm6_policy_table);
 356 #endif
 357 out:
 358         return ret;
 359 out_policy:
 360         xfrm6_policy_fini();
 361         goto out;
 362 }
 363
 364 void xfrm6_fini(void)
 365 {
 366 #ifdef CONFIG_SYSCTL
 367         if (sysctl_hdr)
 368                 unregister_net_sysctl_table(sysctl_hdr);
 369 #endif
 370         //xfrm6_input_fini();
 371         xfrm6_policy_fini();
 372         xfrm6_state_fini();
 373 }