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         const struct ipv6hdr *hdr = ipv6_hdr(skb);
 134         u16 offset = sizeof(*hdr);
 135         struct ipv6_opt_hdr *exthdr;
 136         const unsigned char *nh = skb_network_header(skb);
 137         u16 nhoff = IP6CB(skb)->nhoff;
 138         int oif = 0;
 139         u8 nexthdr;
 140
 141         if (!nhoff)
 142                 nhoff = offsetof(struct ipv6hdr, nexthdr);
 143
 144         nexthdr = nh[nhoff];
 145
 146         if (skb_dst(skb))
 147                 oif = skb_dst(skb)->dev->ifindex;
 148
 149         memset(fl6, 0, sizeof(struct flowi6));
 150         fl6->flowi6_mark = skb->mark;
 151         fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
 152
 153         fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
 154         fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
 155
 156         while (nh + offset + 1 < skb->data ||
 157                pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
 158                 nh = skb_network_header(skb);
 159                 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 160
 161                 switch (nexthdr) {
 162                 case NEXTHDR_FRAGMENT:
 163                         onlyproto = 1;
 164                 case NEXTHDR_ROUTING:
 165                 case NEXTHDR_HOP:
 166                 case NEXTHDR_DEST:
 167                         offset += ipv6_optlen(exthdr);
 168                         nexthdr = exthdr->nexthdr;
 169                         exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 170                         break;
 171
 172                 case IPPROTO_UDP:
 173                 case IPPROTO_UDPLITE:
 174                 case IPPROTO_TCP:
 175                 case IPPROTO_SCTP:
 176                 case IPPROTO_DCCP:
 177                         if (!onlyproto && (nh + offset + 4 < skb->data ||
 178                              pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
 179                                 __be16 *ports;
 180
 181                                 nh = skb_network_header(skb);
 182                                 ports = (__be16 *)(nh + offset);
 183                                 fl6->fl6_sport = ports[!!reverse];
 184                                 fl6->fl6_dport = ports[!reverse];
 185                         }
 186                         fl6->flowi6_proto = nexthdr;
 187                         return;
 188
 189                 case IPPROTO_ICMPV6:
 190                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
 191                                 u8 *icmp;
 192
 193                                 nh = skb_network_header(skb);
 194                                 icmp = (u8 *)(nh + offset);
 195                                 fl6->fl6_icmp_type = icmp[0];
 196                                 fl6->fl6_icmp_code = icmp[1];
 197                         }
 198                         fl6->flowi6_proto = nexthdr;
 199                         return;
 200
 201 #if IS_ENABLED(CONFIG_IPV6_MIP6)
 202                 case IPPROTO_MH:
 203                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
 204                                 struct ip6_mh *mh;
 205
 206                                 nh = skb_network_header(skb);
 207                                 mh = (struct ip6_mh *)(nh + offset);
 208                                 fl6->fl6_mh_type = mh->ip6mh_type;
 209                         }
 210                         fl6->flowi6_proto = nexthdr;
 211                         return;
 212 #endif
 213
 214                 /* XXX Why are there these headers? */
 215                 case IPPROTO_AH:
 216                 case IPPROTO_ESP:
 217                 case IPPROTO_COMP:
 218                 default:
 219                         fl6->fl6_ipsec_spi = 0;
 220                         fl6->flowi6_proto = nexthdr;
 221                         return;
 222                 }
 223         }
 224 }
 225
 226 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
 227 {
 228         struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
 229
 230         xfrm6_policy_afinfo.garbage_collect(net);
 231         return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
 232 }
 233
 234 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
 235                               struct sk_buff *skb, u32 mtu)
 236 {
 237         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 238         struct dst_entry *path = xdst->route;
 239
 240         path->ops->update_pmtu(path, sk, skb, mtu);
 241 }
 242
 243 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
 244                            struct sk_buff *skb)
 245 {
 246         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 247         struct dst_entry *path = xdst->route;
 248
 249         path->ops->redirect(path, sk, skb);
 250 }
 251
 252 static void xfrm6_dst_destroy(struct dst_entry *dst)
 253 {
 254         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 255
 256         if (likely(xdst->u.rt6.rt6i_idev))
 257                 in6_dev_put(xdst->u.rt6.rt6i_idev);
 258         dst_destroy_metrics_generic(dst);
 259         if (rt6_has_peer(&xdst->u.rt6)) {
 260                 struct inet_peer *peer = rt6_peer_ptr(&xdst->u.rt6);
 261                 inet_putpeer(peer);
 262         }
 263         xfrm_dst_destroy(xdst);
 264 }
 265
 266 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 267                              int unregister)
 268 {
 269         struct xfrm_dst *xdst;
 270
 271         if (!unregister)
 272                 return;
 273
 274         xdst = (struct xfrm_dst *)dst;
 275         if (xdst->u.rt6.rt6i_idev->dev == dev) {
 276                 struct inet6_dev *loopback_idev =
 277                         in6_dev_get(dev_net(dev)->loopback_dev);
 278                 BUG_ON(!loopback_idev);
 279
 280                 do {
 281                         in6_dev_put(xdst->u.rt6.rt6i_idev);
 282                         xdst->u.rt6.rt6i_idev = loopback_idev;
 283                         in6_dev_hold(loopback_idev);
 284                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
 285                 } while (xdst->u.dst.xfrm);
 286
 287                 __in6_dev_put(loopback_idev);
 288         }
 289
 290         xfrm_dst_ifdown(dst, dev);
 291 }
 292
 293 static struct dst_ops xfrm6_dst_ops = {
 294         .family =               AF_INET6,
 295         .protocol =             cpu_to_be16(ETH_P_IPV6),
 296         .gc =                   xfrm6_garbage_collect,
 297         .update_pmtu =          xfrm6_update_pmtu,
 298         .redirect =             xfrm6_redirect,
 299         .cow_metrics =          dst_cow_metrics_generic,
 300         .destroy =              xfrm6_dst_destroy,
 301         .ifdown =               xfrm6_dst_ifdown,
 302         .local_out =            __ip6_local_out,
 303         .gc_thresh =            32768,
 304 };
 305
 306 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
 307         .family =               AF_INET6,
 308         .dst_ops =              &xfrm6_dst_ops,
 309         .dst_lookup =           xfrm6_dst_lookup,
 310         .get_saddr =            xfrm6_get_saddr,
 311         .decode_session =       _decode_session6,
 312         .get_tos =              xfrm6_get_tos,
 313         .init_dst =             xfrm6_init_dst,
 314         .init_path =            xfrm6_init_path,
 315         .fill_dst =             xfrm6_fill_dst,
 316         .blackhole_route =      ip6_blackhole_route,
 317 };
 318
 319 static int __init xfrm6_policy_init(void)
 320 {
 321         return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
 322 }
 323
 324 static void xfrm6_policy_fini(void)
 325 {
 326         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 327 }
 328
 329 #ifdef CONFIG_SYSCTL
 330 static struct ctl_table xfrm6_policy_table[] = {
 331         {
 332                 .procname       = "xfrm6_gc_thresh",
 333                 .data           = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
 334                 .maxlen         = sizeof(int),
 335                 .mode           = 0644,
 336                 .proc_handler   = proc_dointvec,
 337         },
 338         { }
 339 };
 340
 341 static int __net_init xfrm6_net_init(struct net *net)
 342 {
 343         struct ctl_table *table;
 344         struct ctl_table_header *hdr;
 345
 346         table = xfrm6_policy_table;
 347         if (!net_eq(net, &init_net)) {
 348                 table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
 349                 if (!table)
 350                         goto err_alloc;
 351
 352                 table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
 353         }
 354
 355         hdr = register_net_sysctl(net, "net/ipv6", table);
 356         if (!hdr)
 357                 goto err_reg;
 358
 359         net->ipv6.sysctl.xfrm6_hdr = hdr;
 360         return 0;
 361
 362 err_reg:
 363         if (!net_eq(net, &init_net))
 364                 kfree(table);
 365 err_alloc:
 366         return -ENOMEM;
 367 }
 368
 369 static void __net_exit xfrm6_net_exit(struct net *net)
 370 {
 371         struct ctl_table *table;
 372
 373         if (net->ipv6.sysctl.xfrm6_hdr == NULL)
 374                 return;
 375
 376         table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
 377         unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
 378         if (!net_eq(net, &init_net))
 379                 kfree(table);
 380 }
 381
 382 static struct pernet_operations xfrm6_net_ops = {
 383         .init   = xfrm6_net_init,
 384         .exit   = xfrm6_net_exit,
 385 };
 386 #endif
 387
 388 int __init xfrm6_init(void)
 389 {
 390         int ret;
 391
 392         dst_entries_init(&xfrm6_dst_ops);
 393
 394         ret = xfrm6_policy_init();
 395         if (ret) {
 396                 dst_entries_destroy(&xfrm6_dst_ops);
 397                 goto out;
 398         }
 399         ret = xfrm6_state_init();
 400         if (ret)
 401                 goto out_policy;
 402
 403         ret = xfrm6_protocol_init();
 404         if (ret)
 405                 goto out_state;
 406
 407 #ifdef CONFIG_SYSCTL
 408         register_pernet_subsys(&xfrm6_net_ops);
 409 #endif
 410 out:
 411         return ret;
 412 out_state:
 413         xfrm6_state_fini();
 414 out_policy:
 415         xfrm6_policy_fini();
 416         goto out;
 417 }
 418
 419 void xfrm6_fini(void)
 420 {
 421 #ifdef CONFIG_SYSCTL
 422         unregister_pernet_subsys(&xfrm6_net_ops);
 423 #endif
 424         xfrm6_protocol_fini();
 425         xfrm6_policy_fini();
 426         xfrm6_state_fini();
 427         dst_entries_destroy(&xfrm6_dst_ops);
 428 }