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 IS_ENABLED(CONFIG_IPV6_MIP6)
  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, &daddr->in6, 0, &saddr->in6);
  65         dst_release(dst);
  66         return 0;
  67 }
  68
  69 static int xfrm6_get_tos(const struct flowi *fl)
  70 {
  71         return 0;
  72 }
  73
  74 static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
  75                            int nfheader_len)
  76 {
  77         if (dst->ops->family == AF_INET6) {
  78                 struct rt6_info *rt = (struct rt6_info *)dst;
  79                 if (rt->rt6i_node)
  80                         path->path_cookie = rt->rt6i_node->fn_sernum;
  81         }
  82
  83         path->u.rt6.rt6i_nfheader_len = nfheader_len;
  84
  85         return 0;
  86 }
  87
  88 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
  89                           const struct flowi *fl)
  90 {
  91         struct rt6_info *rt = (struct rt6_info *)xdst->route;
  92
  93         xdst->u.dst.dev = dev;
  94         dev_hold(dev);
  95
  96         xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
  97         if (!xdst->u.rt6.rt6i_idev) {
  98                 dev_put(dev);
  99                 return -ENODEV;
 100         }
 101
 102         /* Sheit... I remember I did this right. Apparently,
 103          * it was magically lost, so this code needs audit */
 104         xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
 105                                                    RTF_LOCAL);
 106         xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
 107         xdst->u.rt6.rt6i_node = rt->rt6i_node;
 108         if (rt->rt6i_node)
 109                 xdst->route_cookie = rt->rt6i_node->fn_sernum;
 110         xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
 111         xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
 112         xdst->u.rt6.rt6i_src = rt->rt6i_src;
 113
 114         return 0;
 115 }
 116
 117 static inline void
 118 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
 119 {
 120         struct flowi6 *fl6 = &fl->u.ip6;
 121         int onlyproto = 0;
 122         const struct ipv6hdr *hdr = ipv6_hdr(skb);
 123         u16 offset = sizeof(*hdr);
 124         struct ipv6_opt_hdr *exthdr;
 125         const unsigned char *nh = skb_network_header(skb);
 126         u16 nhoff = IP6CB(skb)->nhoff;
 127         int oif = 0;
 128         u8 nexthdr;
 129
 130         if (!nhoff)
 131                 nhoff = offsetof(struct ipv6hdr, nexthdr);
 132
 133         nexthdr = nh[nhoff];
 134
 135         if (skb_dst(skb))
 136                 oif = skb_dst(skb)->dev->ifindex;
 137
 138         memset(fl6, 0, sizeof(struct flowi6));
 139         fl6->flowi6_mark = skb->mark;
 140         fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
 141
 142         fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
 143         fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
 144
 145         while (nh + offset + 1 < skb->data ||
 146                pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
 147                 nh = skb_network_header(skb);
 148                 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 149
 150                 switch (nexthdr) {
 151                 case NEXTHDR_FRAGMENT:
 152                         onlyproto = 1;
 153                 case NEXTHDR_ROUTING:
 154                 case NEXTHDR_HOP:
 155                 case NEXTHDR_DEST:
 156                         offset += ipv6_optlen(exthdr);
 157                         nexthdr = exthdr->nexthdr;
 158                         exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 159                         break;
 160
 161                 case IPPROTO_UDP:
 162                 case IPPROTO_UDPLITE:
 163                 case IPPROTO_TCP:
 164                 case IPPROTO_SCTP:
 165                 case IPPROTO_DCCP:
 166                         if (!onlyproto && (nh + offset + 4 < skb->data ||
 167                              pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
 168                                 __be16 *ports;
 169
 170                                 nh = skb_network_header(skb);
 171                                 ports = (__be16 *)(nh + offset);
 172                                 fl6->fl6_sport = ports[!!reverse];
 173                                 fl6->fl6_dport = ports[!reverse];
 174                         }
 175                         fl6->flowi6_proto = nexthdr;
 176                         return;
 177
 178                 case IPPROTO_ICMPV6:
 179                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
 180                                 u8 *icmp;
 181
 182                                 nh = skb_network_header(skb);
 183                                 icmp = (u8 *)(nh + offset);
 184                                 fl6->fl6_icmp_type = icmp[0];
 185                                 fl6->fl6_icmp_code = icmp[1];
 186                         }
 187                         fl6->flowi6_proto = nexthdr;
 188                         return;
 189
 190 #if IS_ENABLED(CONFIG_IPV6_MIP6)
 191                 case IPPROTO_MH:
 192                         offset += ipv6_optlen(exthdr);
 193                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
 194                                 struct ip6_mh *mh;
 195
 196                                 nh = skb_network_header(skb);
 197                                 mh = (struct ip6_mh *)(nh + offset);
 198                                 fl6->fl6_mh_type = mh->ip6mh_type;
 199                         }
 200                         fl6->flowi6_proto = nexthdr;
 201                         return;
 202 #endif
 203
 204                 /* XXX Why are there these headers? */
 205                 case IPPROTO_AH:
 206                 case IPPROTO_ESP:
 207                 case IPPROTO_COMP:
 208                 default:
 209                         fl6->fl6_ipsec_spi = 0;
 210                         fl6->flowi6_proto = nexthdr;
 211                         return;
 212                 }
 213         }
 214 }
 215
 216 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
 217 {
 218         struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
 219
 220         xfrm6_policy_afinfo.garbage_collect(net);
 221         return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
 222 }
 223
 224 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
 225                               struct sk_buff *skb, u32 mtu)
 226 {
 227         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 228         struct dst_entry *path = xdst->route;
 229
 230         path->ops->update_pmtu(path, sk, skb, mtu);
 231 }
 232
 233 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
 234                            struct sk_buff *skb)
 235 {
 236         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 237         struct dst_entry *path = xdst->route;
 238
 239         path->ops->redirect(path, sk, skb);
 240 }
 241
 242 static void xfrm6_dst_destroy(struct dst_entry *dst)
 243 {
 244         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 245
 246         if (likely(xdst->u.rt6.rt6i_idev))
 247                 in6_dev_put(xdst->u.rt6.rt6i_idev);
 248         dst_destroy_metrics_generic(dst);
 249         xfrm_dst_destroy(xdst);
 250 }
 251
 252 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 253                              int unregister)
 254 {
 255         struct xfrm_dst *xdst;
 256
 257         if (!unregister)
 258                 return;
 259
 260         xdst = (struct xfrm_dst *)dst;
 261         if (xdst->u.rt6.rt6i_idev->dev == dev) {
 262                 struct inet6_dev *loopback_idev =
 263                         in6_dev_get(dev_net(dev)->loopback_dev);
 264                 BUG_ON(!loopback_idev);
 265
 266                 do {
 267                         in6_dev_put(xdst->u.rt6.rt6i_idev);
 268                         xdst->u.rt6.rt6i_idev = loopback_idev;
 269                         in6_dev_hold(loopback_idev);
 270                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
 271                 } while (xdst->u.dst.xfrm);
 272
 273                 __in6_dev_put(loopback_idev);
 274         }
 275
 276         xfrm_dst_ifdown(dst, dev);
 277 }
 278
 279 static struct dst_ops xfrm6_dst_ops = {
 280         .family =               AF_INET6,
 281         .gc =                   xfrm6_garbage_collect,
 282         .update_pmtu =          xfrm6_update_pmtu,
 283         .redirect =             xfrm6_redirect,
 284         .cow_metrics =          dst_cow_metrics_generic,
 285         .destroy =              xfrm6_dst_destroy,
 286         .ifdown =               xfrm6_dst_ifdown,
 287         .local_out =            __ip6_local_out,
 288         .gc_thresh =            32768,
 289 };
 290
 291 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
 292         .family =               AF_INET6,
 293         .dst_ops =              &xfrm6_dst_ops,
 294         .dst_lookup =           xfrm6_dst_lookup,
 295         .get_saddr =            xfrm6_get_saddr,
 296         .decode_session =       _decode_session6,
 297         .get_tos =              xfrm6_get_tos,
 298         .init_path =            xfrm6_init_path,
 299         .fill_dst =             xfrm6_fill_dst,
 300         .blackhole_route =      ip6_blackhole_route,
 301 };
 302
 303 static int __init xfrm6_policy_init(void)
 304 {
 305         return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
 306 }
 307
 308 static void xfrm6_policy_fini(void)
 309 {
 310         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 311 }
 312
 313 #ifdef CONFIG_SYSCTL
 314 static struct ctl_table xfrm6_policy_table[] = {
 315         {
 316                 .procname       = "xfrm6_gc_thresh",
 317                 .data           = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
 318                 .maxlen         = sizeof(int),
 319                 .mode           = 0644,
 320                 .proc_handler   = proc_dointvec,
 321         },
 322         { }
 323 };
 324
 325 static int __net_init xfrm6_net_init(struct net *net)
 326 {
 327         struct ctl_table *table;
 328         struct ctl_table_header *hdr;
 329
 330         table = xfrm6_policy_table;
 331         if (!net_eq(net, &init_net)) {
 332                 table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
 333                 if (!table)
 334                         goto err_alloc;
 335
 336                 table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
 337         }
 338
 339         hdr = register_net_sysctl(net, "net/ipv6", table);
 340         if (!hdr)
 341                 goto err_reg;
 342
 343         net->ipv6.sysctl.xfrm6_hdr = hdr;
 344         return 0;
 345
 346 err_reg:
 347         if (!net_eq(net, &init_net))
 348                 kfree(table);
 349 err_alloc:
 350         return -ENOMEM;
 351 }
 352
 353 static void __net_exit xfrm6_net_exit(struct net *net)
 354 {
 355         struct ctl_table *table;
 356
 357         if (!net->ipv6.sysctl.xfrm6_hdr)
 358                 return;
 359
 360         table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
 361         unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
 362         if (!net_eq(net, &init_net))
 363                 kfree(table);
 364 }
 365
 366 static struct pernet_operations xfrm6_net_ops = {
 367         .init   = xfrm6_net_init,
 368         .exit   = xfrm6_net_exit,
 369 };
 370 #endif
 371
 372 int __init xfrm6_init(void)
 373 {
 374         int ret;
 375
 376         dst_entries_init(&xfrm6_dst_ops);
 377
 378         ret = xfrm6_policy_init();
 379         if (ret) {
 380                 dst_entries_destroy(&xfrm6_dst_ops);
 381                 goto out;
 382         }
 383         ret = xfrm6_state_init();
 384         if (ret)
 385                 goto out_policy;
 386
 387         ret = xfrm6_protocol_init();
 388         if (ret)
 389                 goto out_state;
 390
 391 #ifdef CONFIG_SYSCTL
 392         register_pernet_subsys(&xfrm6_net_ops);
 393 #endif
 394 out:
 395         return ret;
 396 out_state:
 397         xfrm6_state_fini();
 398 out_policy:
 399         xfrm6_policy_fini();
 400         goto out;
 401 }
 402
 403 void xfrm6_fini(void)
 404 {
 405 #ifdef CONFIG_SYSCTL
 406         unregister_pernet_subsys(&xfrm6_net_ops);
 407 #endif
 408         xfrm6_protocol_fini();
 409         xfrm6_policy_fini();
 410         xfrm6_state_fini();
 411         dst_entries_destroy(&xfrm6_dst_ops);
 412 }