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,
  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                 dev_put(dev);
 108                 return -ENODEV;
 109         }
 110
 111         rt6_transfer_peer(&xdst->u.rt6, rt);
 112
 113         /* Sheit... I remember I did this right. Apparently,
 114          * it was magically lost, so this code needs audit */
 115         xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
 116                                                    RTF_LOCAL);
 117         xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
 118         xdst->u.rt6.rt6i_node = rt->rt6i_node;
 119         if (rt->rt6i_node)
 120                 xdst->route_cookie = rt->rt6i_node->fn_sernum;
 121         xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
 122         xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
 123         xdst->u.rt6.rt6i_src = rt->rt6i_src;
 124
 125         return 0;
 126 }
 127
 128 static inline void
 129 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
 130 {
 131         struct flowi6 *fl6 = &fl->u.ip6;
 132         int onlyproto = 0;
 133         u16 offset = skb_network_header_len(skb);
 134         const struct ipv6hdr *hdr = ipv6_hdr(skb);
 135         struct ipv6_opt_hdr *exthdr;
 136         const unsigned char *nh = skb_network_header(skb);
 137         u8 nexthdr = nh[IP6CB(skb)->nhoff];
 138
 139         memset(fl6, 0, sizeof(struct flowi6));
 140         fl6->flowi6_mark = skb->mark;
 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 = (__be16 *)exthdr;
 169
 170                                 fl6->fl6_sport = ports[!!reverse];
 171                                 fl6->fl6_dport = ports[!reverse];
 172                         }
 173                         fl6->flowi6_proto = nexthdr;
 174                         return;
 175
 176                 case IPPROTO_ICMPV6:
 177                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
 178                                 u8 *icmp = (u8 *)exthdr;
 179
 180                                 fl6->fl6_icmp_type = icmp[0];
 181                                 fl6->fl6_icmp_code = icmp[1];
 182                         }
 183                         fl6->flowi6_proto = nexthdr;
 184                         return;
 185
 186 #if IS_ENABLED(CONFIG_IPV6_MIP6)
 187                 case IPPROTO_MH:
 188                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
 189                                 struct ip6_mh *mh;
 190                                 mh = (struct ip6_mh *)exthdr;
 191
 192                                 fl6->fl6_mh_type = mh->ip6mh_type;
 193                         }
 194                         fl6->flowi6_proto = nexthdr;
 195                         return;
 196 #endif
 197
 198                 /* XXX Why are there these headers? */
 199                 case IPPROTO_AH:
 200                 case IPPROTO_ESP:
 201                 case IPPROTO_COMP:
 202                 default:
 203                         fl6->fl6_ipsec_spi = 0;
 204                         fl6->flowi6_proto = nexthdr;
 205                         return;
 206                 }
 207         }
 208 }
 209
 210 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
 211 {
 212         struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
 213
 214         xfrm6_policy_afinfo.garbage_collect(net);
 215         return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
 216 }
 217
 218 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
 219                               struct sk_buff *skb, u32 mtu)
 220 {
 221         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 222         struct dst_entry *path = xdst->route;
 223
 224         path->ops->update_pmtu(path, sk, skb, mtu);
 225 }
 226
 227 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
 228                            struct sk_buff *skb)
 229 {
 230         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 231         struct dst_entry *path = xdst->route;
 232
 233         path->ops->redirect(path, sk, skb);
 234 }
 235
 236 static void xfrm6_dst_destroy(struct dst_entry *dst)
 237 {
 238         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 239
 240         if (likely(xdst->u.rt6.rt6i_idev))
 241                 in6_dev_put(xdst->u.rt6.rt6i_idev);
 242         dst_destroy_metrics_generic(dst);
 243         if (rt6_has_peer(&xdst->u.rt6)) {
 244                 struct inet_peer *peer = rt6_peer_ptr(&xdst->u.rt6);
 245                 inet_putpeer(peer);
 246         }
 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         .redirect =             xfrm6_redirect,
 283         .cow_metrics =          dst_cow_metrics_generic,
 284         .destroy =              xfrm6_dst_destroy,
 285         .ifdown =               xfrm6_dst_ifdown,
 286         .local_out =            __ip6_local_out,
 287         .gc_thresh =            32768,
 288 };
 289
 290 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
 291         .family =               AF_INET6,
 292         .dst_ops =              &xfrm6_dst_ops,
 293         .dst_lookup =           xfrm6_dst_lookup,
 294         .get_saddr =            xfrm6_get_saddr,
 295         .decode_session =       _decode_session6,
 296         .get_tos =              xfrm6_get_tos,
 297         .init_dst =             xfrm6_init_dst,
 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 == NULL)
 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 #ifdef CONFIG_SYSCTL
 388         register_pernet_subsys(&xfrm6_net_ops);
 389 #endif
 390 out:
 391         return ret;
 392 out_policy:
 393         xfrm6_policy_fini();
 394         goto out;
 395 }
 396
 397 void xfrm6_fini(void)
 398 {
 399 #ifdef CONFIG_SYSCTL
 400         unregister_pernet_subsys(&xfrm6_net_ops);
 401 #endif
 402         xfrm6_policy_fini();
 403         xfrm6_state_fini();
 404         dst_entries_destroy(&xfrm6_dst_ops);
 405 }